British Nuclear Group Sellafield Ltd v Gemeinschaftskernkraftwerk Grohnde GmbH & Anor

Service, which included power station operators in Japan and Switzerland as well as Germany and the UK. BNFL was itself a customer, as the owner of certain Advanced Gas Cooled Reactor power stations in the UK. Collectively, the members of the group contracted for the reprocessing of 6,000tU of spent fuel by Service Agreements in substantially identical terms (save as to individual quantities) entered into between each group member and BNFL between 1980 and 1983. The expression “tU” means tonnes of uranium, and is a measure for the quantification of spent fuel of the type in issue, generally known as Oxide fuel. After full economic exploitation in a power station Oxide fuel still contains 96% unreacted Uranium, as well as 1% Plutonium and 3% waste products. The Service Agreements provided for payment to BNFL on a cost plus basis for the design, construction, commissioning and operation of THORP and the necessary ancillary plant at Sellafield, for the whole of the period necessary for the processing of that 6,000tU, or for an operational period of ten years, whichever should prove to be the longer.

BNFL began reprocessing spent Oxide fuel at THORP in 1994. By the end of its first ten years of operation only 5,074tU had been reprocessed. In the year beginning April 2004 a further 604tU was reprocessed. In the year beginning April 2005 a serious malfunction caused the closure of THORP after the reprocessing of only 51tU. The combination of the steps necessary to investigate and remedy that malfunction and other matters, of the detail of which I was not informed, have prevented any further reprocessing at THORP until now. The result is that the original 6,000tU baseload has not been fully reprocessed, and reprocessing of the additional 1,000tU has yet to start.

The Issues

This is a liability only trial, and the issues which I have to decide may be grouped under three headings. They have come to be labelled as follows:

The RTR Issue

This issue relates to the construction, validity and enforceability of the Side Letters. The defendants claim first, that it was a term of the Side Letters that the additional 1,000tU would be reprocessed within the first ten years of THORP’s operation. Alternatively, they say that they are entitled to, and have, rescinded the Side Letters by reason of misrepresentation made by BNFL which induced each of the defendants to contract for the reprocessing of their share of the additional 1,000tU on the terms of the Side Letters. The misrepresentation case is put in various ways, in pleadings, skeleton arguments and oral submissions, but may for present purposes conveniently be summarised as follows. As reflected in paragraph 2 of the Side Letters which I have recited above, BNFL first stated its expectation that THORP was capable of reprocessing 7,000tU during its first ten years of operation at a meeting with BLCs held in December 1986, stating that this expectation was the result of “a rigorous technical review”. It is alleged (and not seriously disputed) that this statement included first, an express representation that there had been a “rigorous technical review” and secondly that, by implication, BNFL’s confidence in the ability of THORP to reprocess 7,000tU within ten years was based on reasonable grounds.

Both of those representations are alleged to have been false. The defendants’ case is that there was never any “rigorous technical review” as suggested, and that in any event there did not exist, either in December 1986 or at any time prior to the making of the Side Letters, reasonable grounds for BNFL’s stated confidence in THORP’s 7,000tU reprocessing capacity within ten years.

BNFL’s response is, first, that by December 1986, and a fortiori by the making of Side Letters, there had been a rigorous technical review within the meaning of the statement to that effect, properly understood, and secondly that although THORP did not in the event justify the expectations which had been placed upon it, there were reasonable grounds for believing that it would, both in December 1986 and when the Side Letters were made.

The defendants advanced two alternatives to their rescission claim. The first is that should rescission be impossible, damages in lieu should be awarded. The second consists of an allegation that the Side Letters are void for common mistake, the relevant mistake being the parties’ alleged shared assumption that the design for THORP as at the date of the Side Letters was such that it was capable (barring unforeseeable circumstances) of reprocessing 7,000tU within ten years, and that, as a matter of fact, there had been a rigorous technical review which demonstrated that capability.

The resolution of the RTR issue has involved the parties in presenting very detailed factual and expert evidence about the process of the design of THORP and its ancillary plant, of the review of that design, of discussion about it between the parties and of the reasonableness or otherwise of the grounds for BNFL’s confidence in THORP’s 7,000tU ten year reprocessing capacity, all focussed upon a period which ended in late 1989, more than 15 years ago, and which preceded the completion of the construction of the plant, let alone its commissioning and operation. The bulk of the documents (which exceeded 160 lever arch files in aggregate) and of the factual and expert evidence which occupied many hundreds of pages and 10 days of oral evidence) was directed to this issue. Furthermore, the underlying question whether BNFL’s capacity expectations of THORP were based on reasonable grounds required the court to receive a crash course in the history, development and technical aspects of the reprocessing of spent nuclear fuel at Sellafield and (to a limited extent) elsewhere, only a fraction of which will be found set out in the remainder of this judgment. The RTR issue has also involved the court in hearing factual and expert evidence about the use of stochastic simulation modelling as a design and/or design review technique, and evidence as to the limitations upon the availability of that technique caused by the then relatively rudimentary state of computer technology in the 1980’s.

The MAGNOX Prioritisation Issue

Stripped to its essentials, this issue may be summarised as follows. Although the main parts of the machinery for reprocessing Oxide fuel at Sellafield were designed and constructed within a dedicated facility (namely THORP), parts of the process took place in ancillary plant which was also used for the reprocessing of spent MAGNOX fuel (that is fuel from the MAGNOX power stations in the UK), which had been reprocessed at Sellafield for many years prior the commissioning of THORP, and which continues to this day.

I have already described Oxide fuel as consisting in essence of Uranium, Plutonium and waste. The essence of reprocessing consists of the extraction, for re-use, of Uranium and Plutonium and the treatment of the residual waste in a manner calculated to reduce, to the extent possible with present technology, the risks associated with its collection and long-term storage.

Various streams of solid, gaseous and liquid waste are produced by the reprocessing of spent nuclear fuel (whether of the Oxide or MAGNOX type), of varying degrees of radioactivity. The most dangerous type of waste emanating from the process is known as Highly Active Raffinate, or “HAR”. It contains, as its name implies, highly radioactive material with half-lives varying from months to millions of years. In all years prior to 1990, its treatment involved evaporation (reducing its volume but increasing its concentration) into Highly Active Liquor or “HAL” and long-term storage in biologically shielded tanks. Both evaporation and storage took place in a building at Sellafield known as B215, the functional name of which is “HALES”, an abbreviation for Highly Active Liquor Evaporation and Storage.

By the 1980s, it was recognised (for reasons which will become apparent later in this judgment) that long-term storage of HAL in tanks gave rise to risks which were best mitigated by a process called vitrification. This was achieved first by transforming HAL into powder (called calcine) by heating it, and then by melting the calcine and mixing it with molten glass, then pouring the mixture into stainless steel drums and allowing it to cool and solidify. This process had been pioneered in France and was planned to be (and was) commissioned at Sellafield, prior to the coming on stream of THORP, in a plant down-stream from HALES called the Waste Vitrification Plant or “WVP”. That part of the reprocessing of Oxide fuel which requires the evaporation, storage and vitrification of HAR/HAL was to be, and was, carried out in facilities shared with the reprocessing of HAR/HAL derived from MAGNOX fuel. Indeed, it was recognised that advantages in terms of reducing the volume of vitrified waste could be obtained by mixing Oxide and MAGNOX HAL prior to vitrification.

In February 2000 the Nuclear Installations Inspectorate or “NII” published a report on the storage of HAL at Sellafield which, while commending the process of vitrification, called upon BNFL to propose a programme for the reduction of stored HAL to a buffer stock level by 2015 with a view to agreement of the programme with, and its subsequent monitoring and regulation by, NII. After detailed discussions between BNFL and NII during the remainder of 2000, NII issued a Specification under statutory powers in January 2001, imposing a programme for the reduction of stored HAL to a buffer stock of 200 cubic metres by 2015, in accordance with a detailed set of reducing annual limits, both for the aggregate of Oxide and MAGNOX HAL, and specifically for Oxide HAL. Thereafter, BNFL were under a statutory obligation to keep stocks of HAL within those limits.

It is common ground that, at least from the financial year beginning April 2002, BNFL deliberately reduced the rate of Oxide reprocessing at THORP below that which the plant would otherwise have been able to achieve, as a means of avoiding a breach of the HAL storage limits imposed by the Specification, and that it imposed no similar reduction in MAGNOX reprocessing during the same period. The defendants claimed that this amounted to a “prioritisation” of MAGNOX reprocessing over Oxide reprocessing which involved BNFL in committing breaches of the Service Agreements, and that the resulting delay in the completion of the 6,000tU baseload caused loss to the defendants, inherent in the propensity for any extension of the time necessary to complete the baseload to increase the unit cost per tU of reprocessing, under the payment structure provided for by the Service Agreements.

BNFL denies any such breach. The substance of its case is that it had no alternative than to make the reductions in reprocessing at THORP which it did, so that those reductions did not involve the breach of any contractual obligation. Further, BNFL relied on force majeure provisions in the Service Agreements. It accepts that, if there was a breach, damage was caused, depending upon the length of reprocessing delay thereby occasioned. There is a case management issue whether I should assess the period of delay at this trial.

The defendants accept (at least by implication) that if BNFL had no alternative than to reduce the reprocessing rate at THORP in order to comply with the Specification, then it committed no breach of the Service Agreements. But the defendants claim that BNFL did have alternatives, both in terms of negotiating a less restrictive specification than that which NII imposed, and in terms of taking other steps to keep HAL storage levels at Sellafield within the specified limits.

The MAGNOX prioritisation issue therefore gives rise first to questions as to the true construction of the Service Agreements and thereafter to issues as to the alternative options (if any) available to BNFL after the publication of the NII 2000 Report, for the control of the levels of HAL storage. Leaving aside the questions of construction, the remainder of those issues have necessitated a detailed technical and historical examination of the processes of highly active waste evaporation, storage and vitrification at Sellafield, an examination with the assistance of competing expert evidence of the negotiating options available to BNFL vis a vis NII in 2000, and of the alternatives steps which the defendants allege that BNFL should have taken thereafter. The examination of those issues requires a measure of technical understanding both of the differing characteristics of Oxide and MAGNOX spent fuel, and in particular its HAL derivative, and of the evaporation, storage and vitrification processes. Again, this necessitated a crash course in the relevant technicalities for the court, by reference to a much smaller body of documentation, and to a slightly smaller body of written and oral evidence.

Although there is some overlap in terms of technical appraisal with the RTR issue, the focus of the factual examination necessary for deciding the MAGNOX prioritisation issue relates to the period 2000 and thereafter, during the operation rather than design and construction of THORP. Furthermore, it concentrates upon the ancillary plant (that is HALES and the WVP) rather than upon the THORP plant itself.

The accounting issue

The Service Agreement contains detailed provision both by way of formula and machinery for the costing of the reprocessing service and for invoicing for it. In clause 12.2.1, it is provided that:

“Charges to be levied by the Reprocessor and paid by the Company in respect of FUEL SERVICES and RESIDUE SERVICES will comprise the recorded costs incurred by the Reprocessor in the construction and operation of facilities provided under this Agreement therefor (below in this Clause 12 and in Clause 13 called “COSTS”) together with the LAND UTILISATION CHARGE and twenty per cent fee described in Clauses 12.2.2.1 and 12.2.2.2 below. COSTS shall be determined by the Reprocessor in accordance with its normal costing procedures as described in Appendix 16 Part B.”

Part B of Appendix 16 headed “Costing Procedures” begins as follows;

“The Reprocessor operates a process costing system which embodies absorption costing principles. ”

The Service Agreements also provide for BNFL’s Costs calculations to be audited by accountants acting on behalf of the BLCs. The auditors chosen were Coopers & Lybrand, later PWC. The defendants’ expectation of the audit process was expressed to Coopers & Lybrand on behalf of the BLCs as: “you provide the bullets so we can shoot”.

Bullets were duly provided by the auditors from time to time, either in the form of matters by reference to which an audit certificate was qualified for a particular year or as matters which, although not leading to a qualification of the audit certificate, were nonetheless considered worth drawing to the BLCs’ attention.

In detailed pleadings, the defendants challenged BNFL’s invoices on a large number of grounds. Further, the defendants claimed that BNFL should be required solemnly to prove each and every item of cost alleged to form the basis of the invoices in dispute. Alternatively, they claimed that if the examination of the specific objections discloses a sufficiently serious failure by BNFL correctly to apply the charging provisions of the Service Agreements that would justify a general enquiry as to the adequacy of the invoices. As a third more cautious alternative, the defendants submit that in any enquiry following this trial arising from successful challenges to particular items of charge, the defendants should be at liberty to raise any further questionable items which may appear during that enquiry.

There was a very substantial narrowing of differences during the preparation for this trial. In some instances items were conceded by BNFL. In others, objections were withdrawn. In a third group, where the issue related to the extent to which particular expenditure related to baseload or post-baseload reprocessing, the parties agreed that the question should be adjourned.

There remained three groups of objections calling for resolution in principle (but not as to quantum) at this trial. Two of them raised a common question of construction, namely whether the precise meaning of “COSTS” in clause 12.2.1 is to be derived entirely by reference to BNFL’s normal costing procedures as described in Appendix 16 Part B, or whether the word “COSTS” carries its own inherent restrictions, derived from the rest of clause 12.2.1, such that even if an item falls within BNFL’s normal costing procedures, it is nonetheless capable of being rejected, on the grounds that it has nothing to do with the provision of the FUEL SERVICES or RESIDUE SERVICES as defined in the Service Agreement.

The three groups of outstanding objections are as follows:

None of those specific objections gave rise to any serious disputes of primary fact, nor did they involve the consideration of lengthy technical material, substantial documents or significantly challenged written or oral evidence. Primarily they involve the construction and then the application of the Service Agreements to the particular items of cost which I have identified. The accounting issue is therefore self-contained, and wholly distinct from either of the other two much larger issues.

It will be apparent from this thumbnail sketch of the issues falling for decision that, divided into those three groups, they are each self contained, both by reference to subject matter and (for the most part) date. It is therefore both possible, and in my judgment convenient, to deal with them separately, both in terms of appraisal of the evidence, fact finding and analysis. To an extent, the determination of the second and third groups (MAGNOX prioritisation and accounting) is facilitated by an understanding of the history and background both of spent fuel reprocessing, and of the relationship between the parties, all of which is essential to the determination of the first and largest issue, to which I now turn.

RTR - RIGOROUS TECHNICAL REVIEW

The Evidence

Documents

Consistent with its stance that the “Rigorous Technical Review” referred to by its representative at the December 1986 Joint Committee Meeting identified in paragraph 2 of the Side Letter was intended to be a reference to the whole of the process of the design and appraisal of the THORP project which had occurred prior to that date, BNFL disclosed a vast bulk of documentation generated during that process, only a small fragment of which was referred to or analysed in any detail during the trial. Much of it is highly technical, and virtually all of it called for expert interpretation, both to enable the court to understand its apparent meaning, but also, (and importantly), to provide a proper understanding of its meaning to the intended recipients of such documentation, almost all of whom were themselves technically qualified officers and employees trained and managed to appraise the documentation for particular purposes.

It is important to note however that since the focus of the court’s factual enquiry in relation to the RTR issue is upon a period which ended in 1989, the surviving documentation could not be expected to be, and is not in fact, a complete record. As will appear, certain documents have gone missing, including one which, had it survived, might have been of real importance.

It cannot therefore be concluded that, because there is no document in the trial bundles recording a review or analysis of a particular item of equipment in the enormously complicated systems being designed for Oxide reprocessing, the design of that item was not in fact reviewed or appraised. As will appear, a significant proportion of the design appraisal was carried out by means of stochastic simulation modelling on computers. While a substantial amount of written reporting of the outcome of that process is available, it is of course impossible to reconstruct those programmes in their entirety, let alone to run them.

It also emerged from the cross examination of BNFL’s main factual and expert witness on this issue, Mr Nigel Donaldson, that there probably is surviving documentation recording some relevant parts of the design studies which has not found its way into the trial bundles. A relevant example is the documentation generated during the study of the design of evaporator C in the HALES plant in Building B215, which was the only one of the three evaporators in that building designed to be capable of evaporating Oxide rather than MAGNOX HAR. The build up of crystalline deposits in that evaporator was a significant contributor to un-planned shut downs or “outages” of the THORP plant, and therefore to the under-performance in terms of throughput of the Oxide reprocessing system as a whole.

I make no criticism of the absence of these documents from the trial bundles, and nor did the defendants complain of their non–disclosure. It is simply an inevitable feature of the enormity of the task of reconstructing a design process which for present purposes was largely concluded by the end of the 1980s (although aspects of it continued thereafter). The point is that the surviving and available documents do not provide an exhaustive record of the design process, nor a complete source from which to judge the quality of the process, or the reasonableness of every underlying judgment or assumption made within it.

By contrast, the surviving documents do provide a reasonably complete record of those communications between the parties which are relevant for the purpose of assessing the precise meaning to be attributed to the statements made on behalf of BNFL on which the defendants’ misrepresentation case is founded, and for the purpose of reaching a safe conclusion as to questions of reliance by the defendants on those representations.

Clause 2.9 of the Service Agreements provided for the establishment of a Joint Committee consisting of representatives of BNFL, COGEMA, and each of the non-UK BLCs for the purposes of consultation on matters pertaining to the operation of the Agreement. Its terms of reference were set out in Appendix 18. In addition, BNFL and the non-UK BLC’s established a Techno- Economic Sub-Committee (“TEC”) for more detailed consultation on technical aspects of the design and the detailed costs structure of the project. The regular meetings of both those committees were minuted and relevant papers presented to them have survived.

This is not a case in which any of the parties suggest that anything turns on particular communications between BNFL and these three defendants. To the extent that the communications between the parties are relevant to the issues which I have to decide, it is common ground that it is the communications passing between BNFL on one side and the non-UK BLC’s as a body on the other side which matter.

Witnesses

BNFL called five witnesses of fact in relation to the RTR issue. One of those, Mr Donaldson, was also their main expert witness on this issue, but they also called an independent expert in relation to the discipline of stochastic modelling.

Dealing firstly with their purely factual witnesses, BNFL first called Mr John Tindle. He graduated with a physics degree in 1958 and joined UKAEA as a reactor physicist. He worked on the commissioning of the Calder Hall power station, and thereafter had a variety of assignments within UKAEA and BNFL, from its foundation in 1971. From late 1976 until the end of 1986 he was BNFL’s principal negotiator with BLCs in relation to the reprocessing of their spent fuel at THORP, reporting to Mr Terry Panton, BNFL’s Head of Overseas Commercial work. Mr Panton was from June 1979 until June 1986 also Chairman of the Joint Committee established pursuant to the Service Agreements, and Mr Tindle was his successor. It was Mr Tindle who first made the statement to BLCs recorded in the Side Letter, in December 1986.

I found Mr Tindle to be a forthright, open and generally relaxed witness, although he appeared a little defensive about his preparation of an important document for BNFL’s main board in August 1986, to which I shall return. He had clearly prepared carefully for giving oral evidence, and I found his account generally to be honest, logical and reliable.

Dr Michael Jeal was the next witness. Having obtained a first class degree and PhD in Chemical Engineering and Fuel Technology, he worked for seven years for ICI on conventional chemical plant research on explosion hazards and plant risk assessment, on the detailed design of new chemical plant and the “de-bottlenecking” of existing plant. After five years working for the UKAEA at its Dounreay plant in Scotland, he joined BNFL in 1981. Having gained experience of the use of stochastic modelling in plant design and appraisal at ICI, he was placed in charge of the development and use of the stochastic models in the design and appraisal of the THORP project by the Project Leader Mr Bill Bennett. In that capacity he was one of the key BNFL employees called upon to express a view about the capacity of THORP to handle an additional 1,000 tU as part of its baseload in the second half of 1986, and a memorandum prepared by him for Mr Bennett on 5^th September 1986 in which he set out his views is one of the most important documents in the case.

I found Dr Jeal to be an entirely honest and generally reliable witness. He was articulate and lost no time in providing precise responses to cross examination from a base of obvious expertise and high intelligence. He gave his evidence with a combination of relaxation and occasional quiet humour which suggested no anxiety or concern to mould his evidence in a manner designed to limit damage. Apart from an understandable desire to defend his own work from what he regarded as unjustified criticism, he showed no partiality towards BNFL, by whom he had long ceased to be employed. He appeared to me to have an above average recollection of events which occurred long ago, and to have taken real trouble to prepare. I found his response to cross examination both reasonable and fair. He seemed to me to be concerned to assist the court, and in this objective he succeeded.

Next to be called was Dr William Wilkinson. He graduated in Chemical Engineering from Cambridge, with a PhD and ScD. After an early career lecturing in chemical engineering, he joined the Technical Department of the UKAEA, working in its nuclear fuel division. After eight years he became Professor of Chemical Engineering at Bradford University, working at the same time as a consultant for UKAEA and for BNFL. He is a Fellow and past President of the Institution of Chemical Engineers, a Fellow of the Royal Academy of Engineering and a Fellow of the Royal Society.

He joined BNFL full time in 1979 as a Director of Research and Development for the Reprocessing Division. In 1982 he became a Director of Engineering for the Reprocessing Division, and in 1984 was appointed to the main board of BNFL as its Company Technical Director, responsible for the technical aspects of all BNFL’s operations. He became Deputy Chief Executive in 1986, and the Director of Spent Fuel Management Services with responsibility for all engineering and operations on the Sellafield site. He retired as a full time main board director in 1992, but remained as a non-executive director until 1994.

Dr Wilkinson played a central role in BNFL’s decision in 1986 that the baseload capacity of THORP could be increased from 6000tU to 7,000tU. It was upon his advice and recommendation that BNFL’s main board approved the increase in baseload in principle on 28^th August 1986, and he provided the technical justification for that increase to the Company Executive of BNFL at meetings in the period September to December 1986. He was, as the responsible main board director, the senior officer of BNFL in connection with the THORP project, and he described himself in cross examination as having been in more or less daily contact with the senior members of BNFL’s staff engaged in that project, including Mr Bennett and Mr Grahame Smith (to whom I shall shortly refer).

I found Dr Wilkinson to be a plainly honest, frank and forthcoming witness. He showed no sign of defensiveness or anxiety in being the most senior officer of BNFL responsible for its commitment to an Oxide reprocessing baseload which (after his retirement) THORP failed to achieve.

Although he described his recollection of events in the 1980s as being good, he did not in fact exhibit powers of recollection of detail anything like those manifested by Dr Jeal or Mr Donaldson. I suspect that this was, in part, the result of having been asked to carry out a much less in-depth process of memory refreshment from documents than that which was undertaken by either of his former subordinates. It was perhaps unfortunate for him that a key document to which the minutes of the BNFL Executive meetings show that he spoke on more than one occasion, and which in an amended form was intended to have constituted the company’s record of the technical justification for its decision to increase THORP’s baseload, has not survived. He was understandably unable to recall its contents, or even to be sure whether a revised version of it, or an addendum to it which he agreed to provide for the Executive, was in fact prepared.

Nonetheless, Dr Wilkinson gave a reasonably clear and consistent (albeit less than detailed) account of his role in that important change of company policy, and of the manner in which he went about consulting with his subordinate colleagues in the process of forming the professional technical opinion as to THORP’s potential baseload capacity, upon which both the company’s main board and Executive plainly relied. I have no hesitation in accepting his factual account as honest, unvarnished and reliable. He expressed the opinion that his view as to THORP’s baseload capacity was based on reasonable grounds, but since he was not called as an expert this was not something which it was necessary for Mr Paul Darling QC who appeared for the defendants to challenge in cross examination. That challenge he quite properly reserved for BNFL’s expert witnesses.

BNFL’s final purely factual witness was Mr Grahame Smith. He graduated in 1959 from Imperial College, London University with a BSc in Chemistry and joined the UKAEA as a technical assistant, before becoming a plant production manager after two years. In 1984, (by which time UKAEA had been replaced by BNFL as operator of the Sellafield site) he became the General Manager of Windscale Works, Sellafield, with responsibility for the whole of the Sellafield site including the (by then) existing MAGNOX reprocessing operation and THORP (by then under construction). He became an executive director and the Head of Sellafield in 1989 and was appointed to the BNFL main board in the mid to late 1990’s, retiring in 1999 after one year as a non-executive main board director.

Mr Smith was an excellent witness. He had no agenda, either to support his former employer or to protect his own reputation. He maintained throughout his evidence an admirable clarity in distinguishing between, on the one hand, facts which he could recall or derive from documents, and assumptions about the forgotten past, on the other. Even his assumptions were generally balanced and reliable. He was of particular assistance in interpreting an important memorandum he wrote to Dr Wilkinson in December 1986 to which I refer later as “the Smith Memorandum”, and which provides the best written insight into the thinking within BNFL about THORP’s capacity immediately prior to its offer of the additional 1,000tU to the BLCs.

BNFL’s main witness was Mr Nigel Donaldson, called both as a witness of fact and as an expert. He was originally intended to constitute BNFL’s only expert witness in relation to the RTR issue, although because of the detailed and technical focus placed by the defendants’ expert evidence on stochastic modelling, BNFL chose to call a separate expert in relation to that discipline as well. I have kept in mind the need to distinguish between Mr Donaldson’s qualities as a witness of fact, and his qualities as an expert.

Upon analysis, his role may fairly be divided into three. First, he gave direct factual evidence about aspects of the design and appraisal of the THORP project in which he had been personally involved at the relevant time. Secondly, he provided largely hearsay factual evidence about a whole range of matters relevant to the RTR issue which he had collected from a detailed appraisal of the relevant documents, and from discussion with current and former colleagues at BNFL, several of whom did not give evidence themselves. Thirdly, he gave expert evidence in the form of his opinion as to the reasonableness of the conclusions reached by BNFL in reaching its view that THORP could reprocess a 7,000tU baseload in ten years.

Mr Donaldson graduated with first class honours in Chemical Engineering from the University of Bradford in 1976. Initially he worked in the MAGNOX reprocessing plant as Sellafield undertaking diagnosis and rectification of plant problems, improvement and changes to plant operation and general support to routine operation. After about eighteen months he took over the day to day management of the Oxide fuel storage facility at Sellafield in building B27, and responsibility for the care and maintenance of the Oxide reprocessing plant at building B204 which had by then already ceased operation but, being in a highly radioactive state, continued to require on-going management of various fully operational systems and infrastructure for the control of the storage environment and the monitoring and control of potential radioactive discharges.

This experience led him to be charged with a liaison responsibility in connection with the design of the first fuel storage pond at Sellafield dedicated to the storage of fuel from Advanced Gas Cooled Reactors (“AGRs”) in building B310. By 1980 Mr Donaldson had therefore gained experience relevant to the storage and reprocessing of spent fuel of both the Oxide and MAGNOX types. As the design of the THORP project developed, he became involved in the provision of liaison from the perspective of operations management to the design team, initially with particular reference to the early stages of the process consisting of the receipt and storage of spent fuel rods in what became known as the THORP building.

His involvement in the design process broadened over time and in 1986 he became a member of the Oxide Fuels Chemical Plant Design Committee (“OFCPDC”), which had overall responsibility for the design of THORP and for the approval of key policy issues. He was also part of the THORP Operations Group which in the late 1980’s was responsible for the preparation of the commissioning and operation of THORP. He was given specific responsibility for the commissioning strategy for the Head End Plant (“HEP”) and Effluent Treatment Process Areas.

From 1992 to 1997 Mr Donaldson worked as an Area Commissioning Manager responsible for the actual commissioning of parts of the THORP plant. Thereafter he took over responsibility for the development of technologies for future deployment in the nuclear fuel cycle. In 2004, when due to structural changes the development portfolio was to be separated from the management of the Sellafield site, Mr Donaldson was assigned to the THORP Operating Unit Technical Department, of which he is now the head. That department provides technical support to the ongoing operational management of THORP, including process trouble shooting, process development, safety case management, nuclear materials safeguards and process evaluation.

It will be apparent from the foregoing that Mr Donaldson has the benefit of long and broad experience in the technical management of both Oxide and MAGNOX fuel reprocessing, relevant detailed personal knowledge of the design, commissioning and operation of the THORP plant, in particular the HEP, and as a continuing senior employee of BNFL, the ability by site visit, discussions with colleagues and document appraisal to refresh his memory on the details of the design and operation of THORP, not shared either by retired employees or by independent experts. By contrast however with Dr Wilkinson, Dr Jeal and Mr Smith, Mr Donaldson was not personally involved in the decision making process which led to BNFL’s commitment to the 7,000tU baseload for THORP in 1986.

Mr Donaldson was a factual witness of quite exceptional quality. In addition to providing a highly detailed account of the design of THORP and of the ancillary plant necessary for a complete Oxide fuel reprocessing system in his first witness statement, he evidently carried out the most through preparation for the provision of further factual information while being cross examined, and in numerous respects amplified his written account with substantial and precise further detail, wherever requested either by counsel or by the court. While he made no secret of the fact that his personal involvement (and therefore experience) related mainly to the Receipt and Storage Plant and the HEP rather than the other main part of the THORP complex, the Chemical Separation Plant or “SEP”, he appeared to have taken the trouble to inform himself in great detail both of the workings and design of the SEP and also of the design and operation of the various ancillary plants such as HALES, the WVP and the Salt Evaporation Plant. In short, there appeared to be no aspect of the Oxide or MAGNOX reprocessing plants at Sellafield about which Mr Donaldson had not carried out the necessary preparatory work to be able to provide well researched and reliable factual answers to the technical issues thrown up by the RTR issue.

Mr Donaldson gave his evidence with care, patience, courtesy and a becoming lack of arrogance. His account of matters of factual detail was not successfully challenged at any stage in cross examination. He was also a highly articulate and patient teacher of technical detail to the uninitiated, including of course myself, and very occasionally counsel. All in all, he was a witness of primary fact in whom I found it possible to place complete reliance.

As an expert, Mr Donaldson started with an inevitable disadvantage in being a full-time employee of BNFL. Prior to the decision of the Court of Appeal in R (Factortame Ltd) v Secretary of State for Transport, Local Government and the Regions (No 8) [2003] QB 381, there were some who thought that full-time employment by the party calling him was a crippling disability for an expert, sufficient to render his opinions inadmissible as expert evidence: see for example Liverpool Roman Catholic Archdiocesan Trustees v Goldberg (No 3) [2001] 1WLR2337.

While it is now clear that full-time employment by a party is no longer a disqualifying characteristic for the giving of expert evidence, the lack of an independent relationship between the expert and the party calling him nonetheless remains a major factor going to the weight of the expert evidence, and one which it has been necessary for me to bear constantly in mind. Nuclear fuel reprocessing is a subject in relation to which it is no doubt difficult to find experts with relevant experience and qualifications, and although there are no doubt other reprocessing plants abroad, each will operate under its own commercial, technological and in particular regulatory constraints which render experience abroad intrinsically less valuable to the expert than experience at Sellafield itself. Furthermore, it may persuasively be argued that there is little substitute, in terms of a reliable perception of the realities, for a long day-to-day involvement in the process of the design and commissioning of the very plant under review.

Balancing those considerations, Mr Donaldson’s thorough preparation, wide ranging experience and personal involvement in the design, operation and maintenance of the THORP plant and its predecessor at Sellafield largely made up for his lack of de facto independence from BNFL. Furthermore, it appeared to me that he had both taken on board and resolved to take seriously his responsibility to the court as an expert witness. He was much criticised in cross examination for the perhaps argumentative tone of parts of his second witness statement, in which he took the defendants’ experts to task in robust terms for, as he regarded it, ignoring or failing to take account of material matters, including matters as to which he had given factual evidence available to them before they wrote their reports. That cross examination of Mr Donaldson was directed to persuading me that he was arguing a case rather than providing an objective opinion. I found his oral evidence to be an encouraging contrast to the signs of argumentativeness in his second witness statement. On the whole, he seemed to me to maintain both balance and objectivity under cross examination to an extent which led me to conclude that the apparent argumentativeness of parts of his second witness statement was more a matter of style than substance.

In conclusion, Mr Donaldson’s lack of de facto independence means that I have been obliged to treat with caution his expressions of opinion. Since he was in fact not personally engaged in the decision making process in 1986 which led to BNFL’s decision to increase THORP’s baseload, he was able to an extent to stand apart from that process and comment upon it. Nonetheless, as will appear, I did not accept in full his expert analysis of the way in which he regarded the decision making process as objectively justified.

BNFL called a second expert, Professor Stewart Robinson, to deal with stochastic modelling issues. He is Professor of Operational Research at the Warwick Business School, which is part the University of Warwick. He has a BSc in Management Science (Operational Research) and a PhD in Management Science, both from Lancaster University, and his PhD focussed specifically on the practice of discrete-event stochastic simulation modelling. He has lectured in stochastic simulation modelling for 14 years, and is the author or co-author of 3 books and numerous articles on the subject. He is an active member of the Operational Research Society, being co-chair both of its Simulation Special Interest Group and of its Simulation Workshop (a biennial conference on stochastic simulation modelling). He is the co-founder of the Society’s Journal of Simulation.

His practical experience of stochastic simulation modelling has included working for ISTEL, Ford, Jaguar, British Airports Authority, BT and North West Water. He has been retained as a consultant in simulation modelling and operational research by a number of well-known businesses. His experience in the nuclear industry consists of carrying out a number of discrete modelling projects and reviews for BNFL, but it was not suggested that this detracted from his independence.

I found Professor Robinson’s evidence to be straightforward, precise, concise and free from guesswork. He appeared to have no agenda to pursue, hidden or otherwise. His opinions were of course challenged in cross examination, but without conspicuous success. It is a fair comment that he was not as deeply grounded in the academic mathematical underpinning and theory of stochastic modelling as was the defendants’ expert Professor Von Collani, but what little he lacked in this respect he more than made up for by his particular expertise in the simulation aspect of stochastic modelling, and by his realistic perception of the practicalities of the use of stochastic simulation modelling as a business tool. All in all, I found him to be an expert upon whose opinions I could generally rely with confidence.

The defendants called four witnesses of fact, mainly to describe from their perspective as employees of BLCs the making by BNFL of its Rigorous Technical Review statements about THORP’s increased baseload, the negotiation of the Side Letters in reliance on them, and their understanding of the meaning of the Side Letters. For none of them was English their first language, and they all therefore gave their evidence through, or with the assistance of, interpreters.

In fact, the representations made by BNFL about the increased baseload are proved by documentary evidence, such as the approved minutes of the meetings at which they were made. The subjective views of the defendants’ witnesses as to what they thought they meant are of limited relevance, and their evidence as to what they thought the Side Letters meant is inadmissible as an aid to construction. The result is that the evidence of the defendants’ witnesses is of significant relevance only to the issue of reliance, and it is not seriously in dispute that BNFL’s representations about the increased baseload were intended to be, and were in fact, relied on. It is not therefore necessary for me to review the weight of the evidence of the defendants’ factual witnesses in detail. All of them were plainly honest, and they intended to assist the court.

The first to be called was Mr Hans-Albert Oppenborn. He had in 1986 been head of the Commerce department of PreussenElektra AG, responsible for the oversight of the company’s economic situation, in particular in relation to its nuclear generation business. PreussenElektra merged with Bayernwerk AG in 2000 to become E.ON Energie AG (“E.ON”). Mr Oppenborn retired in 2003.

Although Mr Oppenborn attended one relevant meeting with BNFL in March 1987, most of his evidence was directed to his own appreciation, at one remove, of what BNFL were offering in terms of the additional 1,000 tU, from what was reported to him by others more closely involved. The quality of his recollection was mixed, and his main perception was that the real dispute lay in the true construction of the Side Letters. Nonetheless his evidence that he placed reliance in his decision making on BNFL’s statement that its confidence in THORP’s increased capacity was based on a rigorous technical review was not shaken in cross examination.

Mr Oppenborn was followed by Mr Jurgen Schkodda. He is now Managing Director of Gesellschaft fur Nuklear-Service mbH, a company owned by a number of German nuclear power station operators, in which E.ON has a controlling stake. In the mid 1980s he was a member of the Accounts and Nuclear Cycle Reprocessing Contracts and Waste Management Department of RWE Energie AG. He acted as secretary for the BLCs on two of the sub-committees established by the Joint Committee under the Services Agreements, namely the Techno-Economic Sub Committee (“TEC”) and the Financial Working Group (“FWG”). He gave evidence in relation to the RTR issue and the Accounting issue.

Mr Schkodda’s recollection of events was better than that of Mr Oppenborn. Cross examination showed that some of the assertions which he made in his witness statement needed to be read with caution, as having been too broadly stated. After initial over-confidence, Mr Schkodda proved to be prepared to make sensible concessions in cross examination, so that those parts of his evidence which survived cross examination were generally reliable.

The defendants’ third factual witness was Mr Dieter Schroder, now the Head of Personnel at the third defendant. At the material time he was a member of the Accounting Department at PreussenElektra, and a member of the FWG. He was a careful, honest and precise witness but, having had no relevant involvement with the relationship between his employer and BNFL since 1992, had little relevant recollection beyond what appears from the documents. Occasionally his recollection conflicted with the documents and he very frankly accepted that I should resolve any such conflict by relying on the documents. He was realistic about the shortcomings of memory over long periods, and of his in particular. As a result he had despite his best intentions little of factual significance to contribute.

Finally, and at a later stage in the trial, the defendants called Mr Jurgen Huber, a member of the Nuclear Department of Bayernwerk AG at the material time. He retired in 2000. He attended the JC as a representative of his employer, and was an active member of certain other sub committees. Like Mr Oppenborn he had firm views about the meaning of the Side Letters and their negotiation which are of course inadmissible as aids to construction.

Mr Huber was an open and honest witness, not given to guessing about what he could not remember. He gave sensible and persuasive evidence about the reliance which he placed upon BNFL’s confidence as experts in THORP’s increased capacity, while recognising that in reality no prediction about future performance can ever be treated as guaranteed in the nuclear industry. He had little to offer about the exchanges between the parties beyond what appears from the documents.

I turn to the Defendants’ nuclear expert, Mr James McGuigan. After graduating with a BSc in Physics, he joined the nuclear industry in 1980, working for BNFL as a radiation safety specialist at Sellafield. In 1983 he joined the UKAEA, undertaking various nuclear safety related functions at its establishment at Harwell. He was involved in the obtaining of the first nuclear site licence for Harwell from the Nuclear Installations Inspectorate (“NII”) between 1988 and 1990. From then until 1995 he advised the directors of Harwell on licence compliance and other regulatory issues, and liaised with NII among others in relation to such matters.

In 1995 Mr McGuigan joined RM Consultants Ltd, a company providing nuclear safety related services to the nuclear industry in the UK, and has been a director of that company since 1999. He is the independent member of the Nuclear Safety Committees of UKAEA (Culham), GE Healthcare, NIREX and BAe Systems (Barrow) which builds nuclear submarines. He has been a member of nuclear industry committees relating to Dounreay, Devonport Naval Base and Harwell.

The above summary demonstrates that Mr McGuigan has very considerable expertise and experience in the safety aspects of nuclear installations, and has as a result come into regular contact with NII. This qualified him well for the giving of his opinions in relation to the MAGNOX Prioritisation issue, for reasons which will appear, and provided him with a general understanding of the design and operation of nuclear plant for the purpose of giving opinions about the RTR issue. In this latter respect he was less experienced in plant design than Mr Donaldson, and of course had no direct experience of the design or operation of THORP. He had to assemble his detailed understanding of THORP and its ancillary plant from the considerable published information about it, and from the voluminous documentation disclosed in these proceedings by BNFL, as well as from the factual evidence of BNFL’s witnesses. This was an inevitable consequence of giving expert evidence for the defendants, let alone independent expert evidence.

Mr McGuigan was a less impressive witness than Mr Donaldson on matters of detail. To an extent this was the natural consequence of his being much less well acquainted with THORP and its ancillary plant, but a remorseless cross examination coupled with Mr Donaldson’s evidence in reply showed that a significant number of errors of detail were the consequence of Mr McGuigan misreading or misinterpreting the material which he did have. He tended during cross examination not to argue points of detail, but rather to retire to a level of generality at which such differences, as he put it, did not affect his opinion on the main points in issue. There were also occasions where he was unable to identify his sources for certain assertions made in his report and under cross examination, which suggested a less than perfect application of the disciplines of an expert.

As to independence, I found Mr McGuigan to be a generally independent expert rather than an advocate for the defendants’ case, although there were moments when his response to questions which challenged his own opinions was rather defensive. His opinions were plainly his own however, and the matters of detail on which he was successfully cross examined did not often undermine those opinions directly. All in all, I found him a less impressive expert than either Mr Donaldson (on the RTR issue) or Dr Robson (BNFL’s expert on the MAGNOX Prioritisation issue), mainly because his lesser command of, or attention to, detail inevitably undermined the general quality of his work. Nonetheless he was of a stature which required me to address his opinions on their merits, and as will appear I have not resolved every difference between him and BNFL’s experts in the favour of the latter.

The defendants’ second expert, on stochastic modelling, was Professor Elart Von Collani, from the University of Wurzburg. He has both a PhD equivalent and a more advanced professorial qualification from that university, respectively in mathematics and applied mathematics, and holds the professorial chair of stochastics and statistics there.

He is a prolific author and co-author of books and articles in international scientific journals, and a managing editor of the journal Economic Quality Control. Since the early 1990s he has been a permanent consultant of the German Post and Telecommunications regulator, and also has regulatory experience in relation to energy. He has provided expert advice on stochastic modelling to two international cancer projects, and in 1997 became a member and project team leader of the European Committee for Standardisation. In 2002 he founded Stochastikon GmbH, a company providing stochastic modelling services, and on its behalf has provided technical reviews, expert reports and seminars in connection with the nuclear industry among others.

Professor Von Collani demonstrated an academic and historical understanding of the discipline and development of stochastics which it would be hard for anyone to emulate. He was an ardent advocate of stochastics as the best (if not the only reliable) technique for predicting future events, including but by no means limited to the throughput of a complex industrial plant at the design stage. He was an articulate exponent of the discipline from a mathematical and academic standpoint, and a very good educator to the uninitiated like myself. In many respects his detailed and clear account of the detail of the discipline, given in what was not his first language but without an interpreter was almost impossible to fault. He gave his evidence firmly, openly, emphatically and without the least sign of partiality.

His opinions on the main points in issue were however seriously undermined by two characteristics, both of which stemmed from his academic background. The first was a lack of practical perception of the point where the value of stochastic modelling becomes so undermined by the paucity of the input data that it ceases to be of practical use, by comparison with the application of experience and judgment. The second was a tendency towards dogmatic perfectionism in his opinions of the quality and reasonableness of the work of others. An example of the latter point lay in his virtual incomprehension and disgust at what he regarded as the sloppy attitude of BNFL in its recording in documents of the process of design review, and his unrealistic view that the use of the phrase ‘rigorous technical review’ incorporated a recognised set of international standards in terms of process, independence and documentary reporting.

These characteristics were in marked contrast to the more practical and realistic attitude to the discipline of stochastics manifested by Professor Robinson, and notwithstanding Professor Von Collani’s eminence, and the frank and engaging way in which he gave his evidence, I found him to be a less valuable expert on the central issues about the use and quality of BNFL’s stochastic modelling than his colleague. Again however, his opinions merited serious consideration on their merits. The high standards which he advocated command real respect, but it by no means follows that, in falling short of them, BNFL acted unreasonably by the standards of the time.

The Facts

Technical Background

A proper appreciation of the challenge facing BNFL in arriving at a reliable prediction of the throughput of the THORP plant requires an understanding, at least at a high level of generality, of the way in which spent nuclear fuel is reprocessed, and in particular of the major differences between spent Oxide fuel and spent MAGNOX fuel, and the consequential differences in the reprocessing techniques required for those different types of fuel. By the time it came to design THORP, BNFL had considerable experience of the reprocessing of MAGNOX fuel, and some much less satisfactory experience of reprocessing Oxide fuel.

As its name implies, spent MAGNOX fuel is derived from MAGNOX nuclear power stations. The first of these to be built was the Calder Hall power station at Sellafield itself. It was the first commercial nuclear power station in the world. Since then, ten MAGNOX nuclear power stations of a similar design have been built and operated in the UK. Some have now closed, and all will have closed by 2010. MAGNOX power stations were also built outside the UK, in particular in Italy and Japan. All those have already closed.

In 1964 BNFL commissioned a MAGNOX fuel reprocessing plant at Sellafield. It is generally named after its building “B205”. This reprocessing plant has operated, generally satisfactorily, throughout the period of the design, commissioning and operation of THORP. It has reprocessed all spent fuel generated by UK MAGNOX power stations as well as spent fuel from Italian and Japanese MAGNOX reactors. It is expected to continue in operation for the reprocessing of all fuel extracted from the UK MAGNOX power stations, including fuel which, upon the de-commissioning of a power station, is removed prior to being fully “spent”.

I shall have to describe the system for reprocessing Oxide fuel centred on THORP in considerable detail. In many respects the system used for reprocessing MAGNOX fuel is very similar. In particular, both use substantially the same “Butex” based solvent extraction process. For present purposes, it is necessary only to describe in any detail those aspects of the make up and properties of MAGNOX fuel and its component parts (including derived waste) which differ significantly from Oxide fuel.

It is a feature of the MAGNOX reactor design, by comparison with more recent designs, that its fuel becomes “spent” at a lower level of commercial use. Fuel is spent when the build-up of the products of nuclear fission interfere with the fission process to an extent which renders it inefficient as a generator of heat. The result is that MAGNOX fuel has less radioactive waste per tU than Oxide fuel, so that, per tU of fuel, the HAR derived from its reprocessing presents an inherently lower long-term risk to the environment than HAR derived from spent Oxide fuel.

A second important difference between MAGNOX and Oxide fuel is the different form of construction of the fuel assemblies. MAGNOX fuel consists, in essence, of a uranium metal rod clad in a jacket consisting of a magnesium-based alloy. The first stage of its reprocessing therefore consists of the mechanical stripping of its magnesium jacket. This takes place under water.

By contrast, Oxide fuel assemblies have a rather more complex structure, in which the uranium and plutonium fuel is encased in pellet form in a stainless steel container, the structure of which is such that it cannot simply be stripped from the active part of the fuel, but rather has to be chopped up as the first stage in its re-processing.

The storage of spent fuel of both types takes place under water in what are called “ponds”, to protect the environment from radiation and to provide cooling. An important distinction between the two types of spent fuel is that the magnesium alloy jacket enclosing MAGNOX fuel is liable to corrosion in contact with water, whereas the stainless steel canister enclosing Oxide fuel is not. The result is that spent MAGNOX fuel can only safely be stored under water for limited periods, whereas spent Oxide fuel is subject to no such disadvantage, and is commonly stored under water for many years. No satisfactory dry storage system for spent fuel has been found. It is normal practice for spent Oxide fuel to be stored for 5 years (either at the reactor site or in storage ponds at Sellafield) to allow for a “cooling” process. Some of the radioactive isotopes which constitute the waste by-product of fission in the reactor have a relatively short half-life and the radioactivity of those parts of the waste is greatly reduced during the cooling period.

Oxide fuel takes its name from the fact that its active element consists of pellets of uranium oxide and plutonium oxide, contained in a sealed tube. Fuel assemblies of this type are used on Advanced Gas Cooled Reactors in the UK (“AGRs”), Pressurised Water Reactors in the UK and abroad (“PWRs”), and in Boiling Water Reactors outside the UK. Operators of those three types of reactor are therefore all customers for an Oxide fuel reprocessing service. In the UK the requirement for such a service initially arose from the commissioning of AGRs, and BNFL first sought to meet the demand for such a service by modifying an earlier reprocessing facility which had originally been built to deal with spent fuel from an experimental reactor (known as the Windscale Piles) and housed in building B204. It operated on a relatively small scale between 1969 and 1973, reprocessing about 90 tU of Oxide fuel. Its use was discontinued in 1973 as the result of an incident which caused radioactivity to escape from its biological shielding into (but not beyond) the B204 building. The relevance of this reprocessing operation is that it provided BNFL with some, albeit less than satisfactory, experience of the design and operation of an Oxide reprocessing plant, in advance of the design and commissioning of THORP.

In order to enable the reader to understand the account of the design and operation of THORP which follows, it is necessary first to give a brief description of the process of spent Oxide fuel reprocessing, for which the plant within the THORP building provided the main, but by no means the only apparatus at the Sellafield site. I shall describe the process as it was designed to be, and was, achieved at Sellafield, pointing out where relevant, differences in the on-going parallel process for the reprocessing of spent MAGNOX fuel.

Spent fuel assemblies arrive at Sellafield stacked within what are called multi-element bottles (“MEBs”), each containing a number of spent fuel canisters. The MEBs themselves are transported by sea, rail and road in larger assemblies called flasks, from storage at power stations to Sellafield. The flasks provide a radiation shield during transport, but at Sellafield the MEBs are stored under water in ponds which provide both a radiation shield and cooling.

The THORP building contains its own dedicated receipt and storage pond, but there had already been storage of spent Oxide fuel at Sellafield (for example from UK sited AGRs), and the site contains an internal railway system for the transport of spent Oxide fuel assemblies from existing storage ponds to the pond at THORP, known as the Receipt and Storage facility.

At B205, spent MAGNOX fuel rods have their casings stripped away under water. At THORP, spent Oxide fuel assemblies are removed from their MEBs in a deep part of the storage pond and then transported up a ramp into a biologically shielded shear cave, where thick concrete replaces water as the radiation shield. There, a machine performing the same function as a large guillotine cuts the fuel assemblies into short lengths, from where they are dropped into one of three dissolvers in another concrete shielded area, in which the uranium, plutonium and waste are dissolved in concentrated, heated nitric acid and piped away, leaving behind the insoluble elements of the stainless steel casings and other metallic parts of the fuel assembly.

Each dissolver consists of a basket-like structure immersed in concentrated nitric acid. Some small insoluble elements known as fines escape from the basket together with the soluble elements and are later removed by centrifuging. Each dissolver is situated in its own shielded enclosure known as a cell and the THORP design provided for a spare cell in which a further dissolver could be installed ahead of any anticipated failure of one of the existing dissolvers. Each dissolver is fed separately from the shear cave, but can be operated in parallel with the others, since the dissolution process takes longer than the cutting process, and since there is no “buffer” for storage of cut-up fuel canisters, pending dissolution. The processes of cutting, dissolution and separation into insoluble and soluble streams which I have just described take place in the THORP Head End Plant or “HEP”.

The insoluble elements of the fuel assemblies, collected either in the baskets or by centrifuging, are removed from the THORP building for treatment as intermediate level waste in other plant at Sellafield, including in particular the Waste Encapsulation Plant. The liquid product of the dissolution process, consisting of solvents, dissolved uranium and plutonium and dissolved waste elements, passes to the next and final part of the THORP building itself, known as the Chemical Separation Plant or SEP. The dissolution process also produces a gaseous stream which is separately collected and treated, in part in the HEP and in part elsewhere.

Between the HEP and SEP there is a system of buffer tanks designed to convert the intermittent production of liquid products from the HEP into a smooth continuous input into the SEP. The first and most important process in the SEP consists of the extraction of uranium and plutonium from the incoming liquid, into two separate streams, leaving HAR as a separate and distinct waste product. I shall not attempt to describe this process, known as the Highly Active Cycle or “HAC”, but it needs to be noted that although a similar result is achieved in relation to spent MAGNOX fuel in building B205, the HAC in THORP made use of a new process. In B205 the process was achieved in settlement tanks, whereas in THORP it was achieved by the use of pulsed columns, of the operation of which BNFL had no previous experience.

In bare outline the HAC produced 3 liquid output streams, namely dissolved uranium, dissolved plutonium and HAR. Each of the uranium and plutonium streams were subjected to further processes of purification and finishing, during which solvents and other effluents were removed. This was designed to yield uranium and plutonium which could be re- used in new fuel assemblies. By contrast, the stream of HAR was piped to a further series of buffer tanks in building B212, for evaporation and storage in B215, the HALES plant.

The various processes undertaken in the SEP also produced various gaseous outputs calling for treatment outside the THORP building, mainly in the THORP Aerial Effluent plant. There were other liquid outputs in the form, for example, of solvents piped away from the SEP for treatment elsewhere within the site, for example at the Salt Evaporator in building B303.

The three principal novelties in the THORP design may be summarised as follows. First, the processes of the Receipt and Storage plant, the HEP and the SEP were all integrated within the same building. Secondly, the process involved the cutting up of fuel assemblies out of water prior to dissolution. Thirdly, at the heart of the SEP lay the Highly Active Cycle based on pulsed columns.

I have already briefly described the processes of the evaporation of HAR so as to produce HAL, its storage at HALES and vitrification at the WVP. In addition to the uranium and plutonium (in oxide form) which the BLCs receive as the re-usable end-product of the reprocessing service, BNFL is also entitled under the Service Agreements to return the vitrified waste canisters to BLCs for storage in their respective countries of origin. The advantage of vitrification is that the highly radioactive waste product of the reprocessing service is returned in an inert form, which receives sufficient cooling from natural ventilation, rather than needing active cooling in storage tanks, and which is much less vulnerable than liquid waste to risks of contamination of the environment.

THORP – Design History

The first stage in the development of THORP was the obtaining of appropriate planning permission from the Copeland Borough Council and the Cumbria County Council. That process required the preparation of conceptual design proposals of a detail sufficient to enable the planning authorities to understand both the physical nature of the project, and its implications in terms of financial and other benefit to the locality and to the national economy. The planning process culminated in a lengthy public enquiry in 1977 known as the Windscale Inquiry, for which a design known as the Reference Design of the plant was provided, as a result of work done during the two preceding years. It was not a design of a physical plant so much as a design concept based on process flowsheets. Detail does not matter, but two points about the Reference Design need to be noted.

The first is that the Reference Design was aimed at achieving a nominal throughput of 1,200tU per year. “Nominal throughput” means the amount of spent fuel theoretically capable of being reprocessed during a year in which the plant operates continuously save for planned maintenance periods, and at full capacity. It represents the ideal rather than the expectation. The reality is that complex plant, and in particular plant which contains novel elements and is subject to extensive regulatory control rarely, if ever, achieves its nominal throughput. In practice, designers expect it to achieve a throughput derived from the application of a percentage often described as “availability” to its nominal throughput.

The second point is that the Reference Design assumed that BNFL would apply its then established philosophy of including spare or duplicate plant in critical process areas. That philosophy is by no means confined to plant designed for handling radioactive substances, but it is of particular relevance to such plant because of the acute difficulties caused by radioactive contamination of those parts of the plant which come into contact with spent fuel. While in operation those parts are shielded either by water or by thick concrete. When they break down they cannot simply be repaired by maintenance staff. Within the shielded areas, maintenance, repair and replacement have to be carried out by machinery operated under remote control. If human access is required, it must be preceded by long periods of decontamination. Furthermore, the remotely controlled handling machinery also tends to be of complex and novel design, with its own inherent susceptibility to breakdown due to wear and tear and other causes.

The philosophy of including spare plant in critical process areas had been applied by BNFL in its design of the MAGNOX reprocessing plant, where there was duplication of the dissolver plant, the first part of the HAC and of the salt free evaporators (in a separate building) as well as the mechanisms for the transfer and handling of the fuel and its constituent parts throughout the process. Nonetheless the philosophy was not, even within the MAGNOX reprocessing plant, taken to the logical extreme of the provision of a complete and wholly separate spare plant. It was not a two line plant, but a one line plant with very substantial duplication, and sometimes even triplication, of critical parts.

At the time of the Reference Design for THORP, no firm decisions had been taken as to the precise extent to which that established philosophy would be applied. The requirement to keep a reprocessing plant within the bounds of commercially acceptable cost gives rise to an obvious trade-off between cost and availability, reflected in the extent to which the philosophy of duplication is applied.

It is convenient next to take a snap-shot of the progress of the design of THORP as at March 1982, by reference to a paper prepared by Mr Bennett for the Reprocessing Division Technical Committee entitled “Capital Cost Status Alternative Plant Provisions for the THORP Head End Chemical Plant” and numbered RDTC (82) P19. By this time the negotiation of what became the Service Agreements had refined the original throughput requirement of 1,200tU per year to a required baseload of 6,000tU in ten years. The paper appears to have been prompted by a perception that the then current global capital cost estimate for the THORP reference design, as by then revised, at £1,180 Million was 2.5 times in excess of current budget provisions, and by a consequential need to assess how best to achieve a design which would achieve the required baseload within ten years by the most efficient use of capital.

Mr Bennett’s paper contained, under the heading “Discussion”, the following important passage:

“Normal BNFL practice in design of reprocessing facilities would be the provision of spare plant for any section considered to be subject to severe environmental or operating conditions. Such a practice is essential when dealing with irradiated fuels, such as Magnox, which display unstable characteristics during long term storage. However such provision is expensive in both capital cost and engineering effort; the practice can, with the more stable stainless steel or zircaloy clad Oxide fuels, be reconsidered.

The capital provision of THORP is based upon the reprocessing of a set quantity of fuel, about 6,000 tonnes, to be undertaken over a suggested period of about 10 years. Following normal practice this would require a reprocessing rate of about 600tU/y guaranteed by the provision of spare plant. However an alternative would be the provision of a single line plant designed for a greater instantaneous throughout (sic), obtaining the guarantee of reprocessing by building in a time contingency to permit major repair should the need arise.

From Table 4 it would be seen that this alternative is more cost effective than the previous normal practice because of the cost capacity index derived from the THORP global estimate.”

It is apparent from tables 4 and 5 annexed to paper (82)P19 that the “time contingency to permit major repair” referred to by Mr Bennett was at that stage was being thought of as about two and a half years within the ten year period. It is also apparent from the same document that the design processes had by March 1982 settled upon using HALES (with suitable modification) for THORP’s HAR evaporation and storage, but that it was recognised that there was limited scope for the use of other existing plant within Sellafield for parts of the Oxide reprocessing service, other than the Plutonium Finishing Line in building B299.

Mr Bennett invited the RDTC to endorse (among others) following recommendations:

I infer that Mr Bennett’s recommendations were approved by the RDTC because, in a paper dated July 1982 and entitled “Design Concept for THORP” Mr Bennett stated as part of the introduction:

“The contracts for THORP call for the reprocessing of 6,000 tonnes of CAGR and LWR fuel in the first ten years of operation. Design work to meet this commitment is well advanced on the basis of a single stream plant of 1,200 tonnes per year instantaneous capacity. This has been shown to be the most cost effective option to meet the contractual commitments and to provide a base for future profitable reprocessing business beyond the first ten years of operation.”

The paper noted that the cost of a THORP plant built according to BNFL’s previous philosophy of full duplication of all highly active sections would far exceed the estimate made at the time of the Windscale Inquiry, and it introduced the revised design philosophy as one of a number of elements designed to reduce costs. The paper continued:

“Two design concepts were considered to meet the contractual commitment of 6,000 tonnes in ten years, viz:-

1.

A plant reduced in guaranteed throughput from a nominal 1200 tonnes to a nominal 600 tonnes per year but retaining the philosophy of duplicated plant in essential areas.

2.

A plant in which the 1200 tonnes per year capacity of the main unit operations was retained but with the elimination of duplicate plant.

The retention of duplicate plant in the first option would avoid major shutdowns for repair, and give a high chance of completing the first 6000 tonnes in the contractual ten-year period by continuous operation. The elimination of duplicate plant in the second option would limit the capability of operating for ten years without major shutdown for repair, but the greater throughput would nevertheless ensure the achievement of 6000 tonnes production over the ten-year period”

Having noted that the first of those options offered only limited scope for cost reduction, and that a 600 tonnes per year plant without duplication would be highly unlikely to be capable of meeting the contractual commitments, Mr Bennett continued as follows:

“An assessment of the forecast production from THORP over the first 10 year period must take account of:-

1.

potential commissioning and ‘run-up’ to achieve steady operation

2.

predicted overall availability thereafter

3.

major unplanned shutdowns.”

Pausing there, reference to half a year for run-up reflects the fact that a crude average of the throughput during a run-up period starting at zero and ending with full normal working capacity will be half that capacity. The above analysis constitutes the earliest surviving written example of a BNFL throughput estimate based upon the combination of the three elements of (1) commissioning and run-up (or ramp-up as it has been described at trial), (2) overall availability thereafter and (3) major unplanned shut downs (or “major outage”). This three point analysis continued to dominate BNFL’s thinking about throughput predictions thereafter, although the precise identification and estimation of the consequences of each element frequently changed.

Before concluding that the preferred philosophy of high nominal capacity but minimal duplication was the most cost effective means of achieving the 6,000 tU baseload in ten years, Mr Bennett added this, under the heading Commercial Considerations:

“The assessment of the capacity of THORP based on the preferred design concept, shows that the plant will be able to reprocess the contracted 6000 tonnes of fuel in ten years with an adequate contingency. If the plant does not suffer a major breakdown, or if the availability of the Head–End plant can be improved beyond 50%, then the capacity over ten years could be substantially greater than 6000 tonnes, perhaps approaching 10,000 tonnes. This potential surplus capacity which may become apparent after a few years from start-up could possibly be used to the commercial advantage of BNFL in a number of ways, although it is too early to speculate on the preferred strategy to follow at this stage.”

It is evident from the fact that as late as mid 1982 fundamental questions of the type reflected in the alternative philosophies debated in these two documents were still being resolved, that the design for THORP was still at a very early stage. Thereafter, although the decision to go for a high nominal capacity but without duplication was followed in principle, such that for example THORP continued to have a nominal design capacity of 1200tU per year, the decision in principle to depart from the duplication philosophy was not applied with anything like full rigour. Furthermore, in numerous respects the plant was designed so as to minimise the risks of major unplanned shut down, with the benefit of BNFL’s previous experience. It is necessary therefore briefly to describe those aspects where duplication was retained, and some of the features designed to minimise the risk of unplanned shut down.

First, provision was made in the design for monitoring and assessment while in service of features of plant which it was considered were vulnerable to failure within 10 years (however well designed). At the same time, provision was made so as to enable replacements to be constructed and then assembled within shielded areas during the time when assessment and monitoring showed that failure was approaching, the objective being to avoid an unplanned shut down by having the replacement plant available on–line in time.

This approach was applied for example to the dissolver baskets in the HEP, to the Medium Active Salt Free Evaporator, and to the systems for removing solids from the decanters, in the latter case by the provision from the outset of a third unused transfer route together with erosion monitoring in the two primary routes.

As at the MAGNOX plant, the liquor transfer routes and key pumping systems were also duplicated at THORP. Further, a modular approach was adopted in the design of the mechanical process plant in radioactive areas, so that spare modules could be assembled and maintained in advance of possible failures in the duty modules, so as to minimise the outages which would otherwise be caused by the need to carry out detailed or intricate repairs within radioactive areas.

The fine detail of these various design features is neither contentious nor relevant. Their relevance is that, taken as a whole, their development and refinement during the period between mid 1982 and the end of 1986 led to a substantial increase in BNFL’s level of confidence in the ability of the design as a whole to minimise the risk of unplanned major shut down, and therefore to the reduction in the time contingency for major shut down which had to be built into any throughput estimate, by comparison with the three year period referred to by Mr Bennett in July 1982.

Besides “pure” design, two important additional processes were being carried on within the more general concept of the design of THORP, namely testing and stochastic modelling. Testing has not been a matter of contention in these proceedings, but stochastic modelling has. I shall briefly describe the testing processes used in the design of THORP, but their importance as part of the basis upon which BNFL developed its confidence in THORP’s potential throughput is not to be underestimated merely because it can be briefly described. Testing processes included the use of an existing test rig at Dounreay, the construction at Sellafield of numerous test rigs for various parts of the THORP plant, the construction of a miniature pilot plant for advance study of the chemical processes to be operated in the pulsed columns in the HAC, and the assembly or part- assembly in advance of mechanical plant at its place of manufacture, before dismantling and reassembly on site.

These testing processes were not merely applied to plant in the form as finally designed, but as part of the design process itself, so that necessary alterations or improvements could be added before the manufacture and the assembly of the plant at THORP.

Stochastic Modelling

A full description of the concept and practical application of stochastic modelling could be as long as my description of the design and operation of the THORP plant, but I shall only summarise it. Professor Von Collani provided some useful definitions, which were not in dispute. A model is simply a replica of something else. It may be a physical replica, like a model railway or model boat, or a mathematical replica. A simulation is not itself a model, but an imitation of a process or operation of a particular entity. Thus, a railway journey may be simulated by running a model train around or from end to end of a model railway, and the behaviour of a ship (including its resistance to movement at different speeds) may be simulated by the use of a physical model in a test tank.

Industrial processes may be simulated on a physical model (such as a test rig) or on a mathematical model, either by manual calculation or (as is now normal) by the use of computers. The capacity of a computer to accommodate a mathematical model is a function of the size of its memory. Its ability to simulate a process by the use of a mathematical model is a function of its power.

Mathematical models may be either deterministic or stochastic. In the more traditional deterministic model, the events to be simulated (in terms both of their frequency and duration) are simply inserted in the model by reference to known or pre-determined parameters. By contrast, as is implied by its name, stochastic modelling incorporates the essential randomness of future events, thereby taking into account the fact that the frequency and timing of many types of future event cannot be precisely pre-determined. As Professor Von Collani put it: “A stochastic model includes random elements describing the omnipresent variability in the course of future developments.”

This case is concerned with the stochastic simulation modelling by computer of a partly designed and hugely complex industrial process for the stated purpose of reviewing its operability. The primary purpose of that review was to assist in the refinement and development of aspects of the design so as to increase the operational performance of the plant under review. A by-product of such a modelling process, and indeed one of the principal measures of the operability of the plant under review, is the stream of throughput statistics emerging from the simulations, and from the statistically derived averages, standard deviations and confidence levels with which throughput is capable of being analysed. That by-product was available to BNFL when seeking to review the ability of the developing design, at various points in time, to meet the objective of reprocessing a baseload of 6000tU in ten years and, as will appear, an increased baseload of 7,000tU in ten years.

Although an important part of the material relied upon by BNFL in developing its changing views as to THORP’s throughput capacity over ten years, stochastic modelling was not by any means the sufficient or sole basis for BNFL’s conclusions from time to time. In particular stochastic modelling was not at any stage applied to the prediction of the necessary commissioning and ramp–up period, nor to the contingency required to address the risk of major outages, two of the three main factors identified by Mr Bennett in July 1982 as central to a ten year throughput prediction. It was however the main source of enlightenment as to the third item, namely the tonnage which the THORP process could be expected to reprocess in a “normal” year of operation, that is a year unaffected by commissioning or ramp- up, in which no major outage occurs. It is to be noted however that BNFL did not attempt a stochastic model of the entirety of the Oxide reprocessing plant centred on THORP until 1988, after it first announced its confidence in THORP’s ability to process a baseload of 7,000tU in ten years, but before the making of the Side Letters.

The essential structure of the stochastic simulation models used by BNFL in connection with its design of THORP may be summarised as follows. First, the nominal throughput rates of the processes within the part of the plant under review were input in deterministic form. Secondly, a list was made of all those foreseeable equipment failures within that part of the plant about which BNFL had some reliable data, in particular in relation to the likelihood or frequency of failure and the time likely to be required to rectify it. Thirdly, those failure events were inserted into the model using a random number generator. Finally, the model was “run” for a number of simulated reprocessing “campaigns” so as to produce a spread of outcomes in terms of tonnage reprocessed per campaign which could then be subjected to statistical analysis.

That analysis might show, for example, that although an average throughput of a particular plant was satisfactory, an adverse deviation from the average consisting of an increased number of failures would lead to the frequency of breakdowns rendering it inoperable, and to consequential design changes directed either to reducing the risk of that level of breakdown to an acceptable extent, or to the increase in the size of buffer storage between items of plant designed to prevent stoppages in one plant compromising the throughput of plants upstream or downstream, and thereby detracting from the throughput of the reprocessing system as a whole. The process of review by stochastic modelling was of course iterative, in a sense that design changes introduced as the result of one series of simulations would require a restructuring of the model, and the production of a revised set of throughput data as a means of testing the efficacy of those changes.

Before describing the history of BNFL’s use of stochastic modelling in the development of its design of THORP, it is necessary to focus on some of the limitations of the modelling process. First, (and inevitably), the quality of the output data from any stochastic simulation model is critically dependent upon the quality of the inputs. In relation to failure rates and repair times, the quality of the inputs depended upon BNFL having access to reliable data or experience about the failure of identical or similar pieces of equipment in existing plants or, in relation to new plants, the availability of failure data to be derived from rigorous testing, or even data derived from experience of the use of individual component parts in different existing plant.

The defendants mounted no sustained detailed attack on the quality of BNFL’s input data, but concentrated their focus on two more general points. The first is that, precisely because THORP was a brand new integrated design for which there was no comparable precedent other than (perhaps) in France for which BNFL had detailed data or experience, and because THORP was designed to eliminate many of the defects which had been experienced in earlier plant, considerable caution was required before treating the outputs of the model as a reliable guide to future performance.

The second and more fundamental point is that the requirement for reliable input data about defects and repair times necessarily limited the class of defects capable of being put into those stochastic models to those which were both foreseeable and which gave rise to relatively short repair times, measured on average in hours (but ranging from minutes at one end up to about seven days at the other) rather than, for example, either months or years.

The consequence was therefore to exclude altogether from stochastic modelling those breakdowns which were either wholly unforeseeable or which, although in theory foreseeable, could not be quantified in terms of outage because they had never previously occurred. In practice, breakdowns liable to cause outage of more than a week were therefore simply not modelled, and the stochastic model therefore provided throughput information relevant only to plant operation unaffected by such outages. The non-modelled outages included not merely the disaster scenarios which might require the closure of the plant for anything up to three years, but much more modest outages interfering with the smooth operation of THORP for one or more months. All these had therefore to be accommodated within the major outage contingency, in any ten year throughput prediction.

The next preliminary point is that the proper interpretation of the output of a stochastic model needs to be based upon an appreciation of the purpose for which the model was designed. In the present case, the primary purpose of the models prepared and run by BNFL was the refinement of THORP’s design, rather than the review of its potential ten year baseload. In short, they were aimed at identifying and curing potential bottlenecks by design improvement, rather than at a review of the question whether the final design could be expected to meet any specified throughput capacity over ten years. Thus for example, the longest specific model run was of a half yearly campaign lasting approximately 150 days, rather than of a continuous period of one or more, still less ten, years. Furthermore, the distribution technique designed to model the time between the failures used (so far as is ascertainable) an exponential rather than Weibull method, thereby taking no specific account of the effect of ageing of equipment over a substantial period upon the frequency or pattern of breakdowns. The extent to which these consequences of the aims for which the models were constructed significantly affects the throughput data derived from their operation is a matter of controversy, but it is common ground that the use of the modelling process as a guide to the prediction of achievable baseload over ten years called for caution, having regard for the different primary purpose for which the modelling had been instituted.

BNFL had not used stochastic modelling for any design purpose prior to its design of THORP. The evidence does not show precisely when such work started in relation to THORP, but it was probably in 1977. Some of it therefore preceded Dr Jeal’s arrival at BNFL in 1981, and it is evident that the work had started by the time it was mentioned by Mr Bennett in his July 1982 paper, to which I have referred.

The earliest written description of the stochastic modelling work which I was shown appears in a report on the SEP stochastic model dated March 1982 and numbered OFCPDC (82) P116. It was issued with Dr Jeal’s approval. In its introduction, the requirement of the modelling task is described as being to assist in deciding the throughputs required in each section of the plant, and how the operation of each section should be isolated from adjacent sections, for example by buffering. The overall design requirement is identified as providing reprocessing of 6,000tU over 10 years while providing a time contingency for major outage of a maximum of 4 years, and therefore a capacity of 1,000tU per year. The text does not distinguish expressly between nominal and what I have labelled as ‘normal’ capacity, but I infer that readers of the report would have assumed it meant the latter.

A fuller report is to be found in the following year in a document entitled “Stochastic Modelling of THORP – Progress to Date” presented in November 1983 to the OFCPDC and generally known in these proceedings as “P275”. It was written by a Mr Thomas of BNFL’S Operational Research department (OR) and approved by Dr Jeal. Its purpose was to summarise for the OFCPDC the progress of modelling to date, but it noted that at an informal level the results of the work so far done had already been communicated to the design team, and that it had led to the incorporation of appropriate changes to the design.

A reading of P275 reveals the following relevant facts, as at November 1983, among a much larger body of conclusions and recommendations. Firstly, most parts of the THORP plant had by then been subjected to some stochastic modelling, but at varying levels of detail. By “the THORP plant” I mean the plant within the THORP building itself, not the ancillary plant such as HALES or the WVP. The paper carried a health warning at paragraph 13 in the following terms:

“There is still a need to investigate crucial areas at a high level of detail to verify assumptions made in previous work. These detailed models would be used to check areas of uncertainty and their sensitivity to changes in assumptions where the data is not accurately known.

Although the various studies have used common data and assumptions wherever possible….they have necessarily progressed independently. It would now be appropriate to ensure that the separate results and recommendations do not conflict. To do this, the development of an “overall” model of the plant is recommended to supersede the work reported in reference 1. Such a model could also be used to display the operational characteristics of the plant and assist in the evolution of operating strategies.”

Secondly, P275 shows that the modelling had proceeded on the basis of assumptions firstly that reprocessing would take place in two campaigns per year each of 150 days’ operation of the HEP and SEP, staggered to allow the HEP to generate sufficient feed for the SEP, and secondly that the required throughput was 1,000tU per year or 500tU per campaign.

Thirdly, the modelling of the HEP had used typical plant throughput (assuming 100% availability) based upon a weighted average of the different fuels which would under the then contractual arrangements be presented for reprocessing, leading to an average nominal throughput of 6.4tU per day.

Fourthly, the modelling suggested an average availability of the HEP of 59% which, when applied to the average daily throughput (just described) provided an annual throughput of 1095tU. Nonetheless it was concluded that because weekly availability was expected to range between 10% and 80%, large HEP/ SEP buffers were needed to avoid disrupting the continuous operation of the SEP.

Fifthly, the modelling of the SEP (and in particular of the HAC) suggested that an average throughput of 554tU per campaign was achievable, but was vulnerable to increased breakdown frequency which needed further study, and that a margin above the minimum required daily throughput of 4.5tU would require design change.

Among the recommendations, by way of further modelling beyond that currently in hand, was the preparation of an overall model of the SEP and downstream plant to examine interactions. Dr Jeal explained in evidence that “downstream plant” included HALES, WVP, the Off-Gas treatment plant and the plant for the treatment medium and low level waste.

Broadly speaking, P275 and the detailed but at that stage by no means final stochastic modelling upon which it was based, gave grounds for cautious optimism that the plant within the THORP building itself would be capable of achieving a “normal” throughput of the required 1,000tU per year. By “normal” I mean (as I have already noted) a year unaffected by commissioning or ramp-up, or by individual outages of more than a week. It contained clear warning of the need for further design work, further stochastic analysis, and in particular sensitivity analysis arising from the non-availability of complete input data about failure rates and repair times.

By September 1986 it is clear that a substantial amount of further stochastic modelling had been carried out, in particular in relation to the SEP, but it had not been summarised or presented in any comprehensive report by way of an update from P275, nor had any similar progress been made towards modelling the ancillary plant outside the proposed THORP building itself. That much is apparent from a memorandum to Mr Bennett from Dr Jeal dated 5^th September 1986, known in the proceedings as “the Jeal Memorandum”, to which I shall have to return in detail. The result is that, although the primary intended function of stochastic modelling in terms of assisting in the refinement of the design had continued, there had by mid 1986 been no further overall review of its implications in terms of THORP’s throughput capacity over 10 years.

Ramp-up and Major Outage

In the mean time however, developments had been taking place giving rise to an adjustment in the allowances which Mr Bennett had thought necessary in July 1982 in respect of commissioning / ramp-up and major outage, of one year (with a half-year effect) and three years respectively. I have thus far said little about commissioning or ramp-up. In order to operate a nuclear plant such as THORP, BNFL needed an operating licence from NII, and such a license would only be obtainable once the commissioning of each element of the new plant by the processing of test samples of spent fuel had demonstrated its compliance with relevant safety criteria. The time likely to be taken by that process could in 1986 only be predicted by an exercise of judgment, based upon previous experience of dealing with NII, discounted by reference to the novelty and much greater sophistication of the integrated THORP plant. By mid 1986 it had been adjudged at Chief Executive level within BNFL that the likely commissioning period should be 21 months, a period more than double that estimated earlier at 9 months. The reasonableness of that judgment has not been challenged, and accordingly it has been unnecessary for BNFL to deploy evidence as to the process by which it was reached.

Ramp-up is a process which (according to varying usages of the phrase) may include the whole of the commissioning period, but extends substantially beyond it in relation to complex plant such as THORP, because it includes the period of fully licensed operation when an increase in throughput from the test levels necessary for licensing up to the plant’s full “normal” capacity is undertaken, by staff and managers who are, of necessity, still learning the ropes and refining their operational techniques in the light of experience.

Nothing in the evidence which I have seen and heard suggests that any specific or detailed consideration was given within BNFL to the question how long beyond the end of commissioning ramp-up would take, at least by the middle of 1986. For convenience I shall refer to both commissioning and ramp-up as ramp-up, save where it is necessary to distinguish between the two.

As to the development of BNFL’s perception after July 1982 of the appropriate time contingency to allow for major outage, there is some limited evidence, but it is by no means clear or consistent. I shall have to describe that evidence in my chronological account of BNFL’s internal appraisal of a possible 1,000 tonne increase in the 10 year throughput requirement. The most that can be said as part of this description of the background is that advances in the design process had led to an increase in BNFL’s confidence that it had considerably reduced the risk of major outage, sufficient to lead to a conclusion that a dedicated 3 year contingency for major outage was excessive.

The Service Agreements

Before describing in detail the circumstances in which BNFL came to offer an additional 1,000tU processing service to BLCs, it is necessary to look slightly more closely at the terms of the Service Agreements, by which the parties had become bound by the end of 1983. Although the RTR issue does not give rise to any question as to the true construction of the Service Agreements, they form an important part of the background, both in relation to the construction of the Side Letters and to the questions of misrepresentation, reliance and mistake.

Clause 2.1 required the relevant BLC described as “the Company” to deliver its proportion of the 6,000tU baseload within a precisely stated time frame. By clause 2.3, BNFL granted the BLCs “an option of first refusal” in relation to surplus capacity at THORP beyond the initial 6,000tU baseload. The right of first refusal entitled the BLCs to be first in the queue for any additional reprocessing capacity (in proportion to their respective contributions to the 6,000tU baseload) but not to any preferential terms, better than the market terms upon which BNFL might at that time wish to offer a reprocessing service.

By clause 2.5 BNFL was obliged to store the spent fuel delivered by the BLCs at Sellafield until it should decide to reprocess it. Clause 2.6 provided that property in spent fuel, and any uranium, plutonium waste or residue derived from it should be and remain the property of the Company unless otherwise agreed. Clause 2.7 required BNFL to use its reasonable endeavours to undertake storage, reprocessing and conversion of waste into residue in an economical manner having regard to the state of the art and plant requirements.

Clause 5.1 largely replicated clause 2.5 in terms of BNFL’s storage obligation. Clause 5.2 provided as follows:

“Without affecting the Company’s allocation of priorities pursuant to Appendix 5 hereof the Company’s FUEL shall be REPROCESSED by the Reprocessor at the REPROCESSING PLANT in such quantities at such time as may be determined by the Reprocessor. ”

Appendix 5 laid down certain overall priority rules for the transport and storage of fuel and for the making available of uranium and plutonium relating to fuel delivered to Sellafield by BLCs. The precise rules do not matter for present purposes, but they were based on certain express assumptions including the following:

Returning to the main part of the Agreements, clause 6.1 provided that fuel should be reprocessed in such quantities and at such times as might be determined by BNFL. Clause 6.2 is of central importance, and is as follows:

“Without affecting the Company’s rights under Appendix 6 Part B, Appendix 6 Part A hererof sets out a schedule for REPROCESSING of all fuel assemblies on behalf of BASELOAD CUSTOMERS (which schedule is below called “the TARGET SCHEDULE”). In the event that the Reprocessor fails to adhere to the TARGET SCHEDULE and to the extent described in the said Appendix 6 Part A the Company shall be entitled to a reduction in the fee referred to in Clause 12 hereof but the Company’s rights in respect of such delay shall be limited solely to such remedy and such remedy shall be accepted by the Company in full satisfaction for any such delay as aforesaid.”

I have already made brief reference to clause 12 at the beginning of this Judgment. For present purposes it is sufficient to note that the charges to be levied on BLCs contained 3 elements. The first is payment of BNFL’s Costs (as defined, and the subject of the accounting issue). The second is a “Land Utilisation Charge” or “LUC” consisting of an amount equivalent to 4% of the Costs. The third element is a fee comprising 20% of the aggregate of the Costs and LUC.

Part A of Appendix 6 contains what is at first sight a curious definition of the Target Schedule. Its unusual form was necessitated by the fact that different BLCs signed the Service Agreements with BNFL at different times during a lengthy period, and the schedule had to be identical in each Service Agreement. Paragraph 2 laid down a “target programme” for reprocessing 6,000tU at THORP assuming the completion of construction in 1987. As in Appendix 5, it assumed a throughput of 300tU in the first year, 500tU in the second year and 650tU in each subsequent year, starting in 1987/88 and ending in 1996/97.

Paragraph 3 of Appendix 6 sets out an amended programme assuming a delayed start in 1990/91 and, by paragraph 4, was based on an assumption that Service Agreements for the reprocessing of a 6,000tU baseload had been signed by 31 December 1978. It contemplated the reprocessing of 300tU in the first year and 650tU in each subsequent year, with a balancing amount of 500tU in the final year.

That table constituted the Target Schedule, within the meaning of clause 6.2, subject to provision for further delay in commencement equivalent to the time which might elapse between 31 December 1978 and the actual date on which the last relevant Service Agreement came into effect. Since the Service Agreements were made in late 1983, the Target Schedule contractually defined by the Service Agreement had a start date in 1995/6.

Paragraph 5 of Appendix 6 set out a formula whereby any shortfall in actual reprocessing in any year below the amount stated in the Target Schedule would give rise to a reduction in BNFL’s Fee subject to a floor of 15%. Paragraph 6 provided for a partial abatement of the Fee (but not LUC or Costs) in the event that the reprocessing plant should fail to operate for reasons attributable to BNFL for a cumulative excess of three years. It also provided for a deferment of the date of completion of the Target Schedule “until such time as the total quantity of fuel specified therein has been reprocessed”.

An expressly stated “principle” of the Service Agreement was that BNFL should not bear financial risk thereunder: see clauses 10.7.3 and 18.2.1.4. The practical operation of that principle is reflected in the provisions to which I have referred above in that :

Clause 18 contained a detailed force majeure clause to which I shall have to return in detail in relation to the MAGNOX prioritisation issue. For present purposes it is sufficient to note, first, that in clause 18.1.2 it was stated to be a principle of negotiation between the parties that “the Reprocessor would use every endeavour to provide facilities and services as specified herein”. Secondly, clause 18.2.2 gave limited rights to the BLCs to terminate the Service Agreement in the event either that suspension, delays or shut-downs caused a cost escalation double that of the real cost of construction or that after 13 years from commencement, the plant should be shut down without completing the reprocessing of 6,000tU in circumstances where the BLCs considered that operations would not be able to continue. So far as I am aware, those limited provisions have not been invoked.

The essential contractual background to the discussions which led to the making of the Side Letters is therefore that by the end of 1983 the BLCs (including the defendants) had committed themselves to funding the continued design, construction and operation of THORP and the ancillary plant at Sellafield necessary for the reprocessing of a 6,000tU baseload of Oxide fuel on cost-plus terms which gave them rights of consultation, but not control over the process, and which exposed them to the expressly contemplated risk that the reprocessing of the 6,000tU baseload would take substantially more than 10 years to complete and thereby expose them to funding the ongoing operation of the plant for an indefinite period, with very limited provisions for abatement of fees or termination. Although such a contractual structure might appear at first sight to be rather one-sided, I must assume (and there is no evidence to the contrary), that it was a proper reflection of the perceived spent fuel reprocessing market at the time.

Steps Leading to the December 1986 RTR Statement

The consideration by BNFL of a possible increase above the 6,000tU baseload contracted for in the Service Agreements appears to have been prompted by a suggestion from the representative of a Swiss BLC namely KKG at the 12^th meeting of the Joint Committee held on 13^th June 1984. The immediate background was that at the previous meeting of the Joint Committee in December 1983, BNFL has presented its October 1983 Business Plan for THORP which included a 30% increase in the unit cost of reprocessing fuel, above that implicit in the previous year’s Business Plan. The unit price per tU in the October 1983 Business Plan had risen to £468,000. BNFL had attempted to sweeten an otherwise bitter pill by suggesting certain concessions to BLCs worth in aggregate approximately £5,000 per tU.

KKG’s representative Mr Vogt reminded the meeting that BNFL’s French competitor COGEMA (which had a substantial reprocessing facility at Cap la Hague near Cherbourg in Normandy) had, when faced with having to announce a similar cost increase, alleviated it by offering to increase the baseload offered to its customers in respect of its planned new plant called UP3 from 6,000tU to 7,000tU. COGEMA was technologically ahead of BNFL in 1984 because it already had an operating Oxide reprocessing plant called “UP2” and was further advanced in its design of UP3 than was BNFL in relation to THORP.

BNFL’s response at the 12^th Joint Committee meeting, expressed by its Chairman Mr Panton, was that although BNFL expected to be able to reprocess more than 6,000tU in the first 10 years of the operation of THORP, it was not in a position to commit to any specific increase in baseload, and preferred to negotiate favourable terms for a second and subsequent baseload of 6,000tU to follow on from completion of the first. No significant change to this commercial policy occurred between 1984 and the holding of the 16^th Joint Committee Meeting with BLCs on 24^th June 1986, at which Mr Panton’s replacement by Mr Tindle as chairman of the Joint Committee was announced to BLCs.

Shortly thereafter it came to Mr Tindle’s attention for the first time that it would be necessary to announce to BLCs a two year delay in the planned start-up date for THORP, coupled with a further increase in the estimated unit cost of reprocessing. This led to a perception by Mr Tindle and other members of BNFL’s commercial and financial staff responsible for the negotiation and performance of the Service Agreements that these announcements would threaten the continuing commitment of the BLCs to the THORP project (notwithstanding their contractual obligation) which would be unlikely to be assuaged by modest concessions of the type suggested in 1984. Accordingly, Mr Tindle prepared a strategy paper for consideration by BNFL’s main board, proposing an increase in THORP’s planned throughput in its first 10 years of operation from 6,000tU to 7,000tU, so as to reduce unit prices. The paper entitled “Commercial Strategy for THORP Baseload” and numbered BNFL/B/86/74 referred to delayed commencement of operation from 1990 to 1992, and to COGEMA’s solution of offering an increased baseload at its UP3 plant. It continued, under the heading ‘Strategy ‘as follows;

“The increases in unit prices which customers would see in the October 1986 issue of the Business Plans will be unacceptable to them and there is a real risk that overseas customers would become so dissatisfied that their commitment to THORP is put at risk. It is therefore proposed that BNFL should increase its planned throughput of 6,000tU in the first 10 years of THORP operation to 7,000tU in order to reduce unit prices. Such a throughput is judged to be realistic by BNFL’s designers and plant operators. “

Mr Tindle wrote this paper so that it could be presented to the BNFL main board by Dr Wilkinson. Forensic attention was naturally focussed at trial upon the source of Mr Tindle’s statement that a 7,000tU throughput in 10 years was judged to be realistic by BNFL’s designers and plant operators. The answer emerged as the result of an internal inquiry made by Mr Smith, the senior “plant operator” as General Manager of the Sellafield site, when he saw the paper for the first time in September 1986, having not previously been consulted about the making of such a statement to the main board. Cross examination by reference to the memoranda generated by Mr Smith’s inquiry, and to Mr Tindle’s surviving diary notes, show that the source of the statement was Mr Tindle’s discussion with his superior Dr Wilkinson, and with Mr Smith’s immediate superior Mr Gordon Steele while preparing the paper “in a hurry”.

Mr Tindle reviewed a draft of the paper with Dr Wilkinson on about 3rd August and obtained his approval to it. Dr Wilkinson said in cross examination that his almost day to day contact with the designers of THORP and the operators at Sellafield enabled him to put his name (as he did) to the strategy paper, and in particular to the judgment that a 7,000tU baseload in 10 years was realistic, without having to consult with his subordinates. It was he said a judgment which he had already formed as a result of his constant on-going contact with them. In its final form, the strategy paper identifies Dr Wilkinson as the first of its two co-authors. The other was Mr Alan Johnson, the Commercial Director. Mr Tindle is not mentioned.

Dr Wilkinson and Mr Johnson tabled and spoke to the strategy paper at the meeting of the main board of BNFL on 28^th August 1986. The minutes of that meeting show that Mr Johnson sought an immediate decision on the principle of extending the THORP baseload to 7,000tU, in order to meet the timescale for submitting a revised Business Plan to BLCs, and explained that its motivation was to reduce the unit cost of reprocessing to a level below that notified to BLCs in the previous year, and below that then assumed for COGEMA’s UP3 plant. Dr Wilkinson stated that:

“the plant should be technically capable of processing the additional 1,000te within the first 10 years of operation since the 6,000te baseload assumed only 50% utilisation of the plant.”

This was of course a reference to THORP’s nominal capacity of 1,200tU per year, and constituted an explanation for his confidence that the increase could be achieved at what I can only describe as the highest conceivable level of generality. In his defence it may fairly be said that the minute is no doubt an abbreviated summary of what he actually said, and that a high level of generality was perfectly appropriate in a non-technical environment, at which it was only a decision in principle that was being requested. Dr Wilkinson said in evidence that as BNFL’s Technical Director the board looked to him for the relevant technical judgment.

The main board’s decision was to approve the principle of extending the THORP baseload by 1,000tU to 7,000tU, and to note that “detailed proposals would be discussed Executively”. It therefore fell to Dr Wilkinson and Mr Johnson to prepare detailed proposals for the consideration of BNFL’s Executive, at its next meeting on 16th September. They did so in the form of a paper for which the reference is CX(86)98, of which there is unfortunately no surviving copy, and, unsurprisingly, no reliable recollection as to its contents. To the extent relevant, these must be a matter of inference. The critical question for present purposes is whether the paper contained any technical summary or justification of the basis for Dr Wilkinson’s confidence, together with the reported confidence of THORP’s designers and the plant operators, in the ability of the THORP design to reprocess 7,000tU in 10 years.

In my judgment, the paper CX(86)98 concentrated upon the commercial justification for the extension of the baseload to 7,000tU, and contained no or no significant technical justification. That seems to me to be the clear inference to be drawn from the minutes of the 16^th September meeting of the company Executive at which one of its members, a Mr Chamberlain, asked “whether BNFL now had additional information or greater confidence in the potential of THORP to reprocess at a higher rate”, since he evidently considered that if such an increase was offered to BLCs after the recent announcement of a two year delay in commencement, they would be likely to treat it with considerable suspicion. The minutes do not suggest that Mr Chamberlain obtained any satisfactory answer to his question because they record that he asked that the paper CX(86)98 should be revised so as to include a technical justification for extension of the baseload.

The Executive’s decision is recorded as follows:

“The Executive

(1)

endorsed the commercial strategy proposed in this paper, subject to :

(2)

Dr Wilkinson providing a technical justification for extension of the baseload by 1,000tU.

(3)

asked that CX(86)98 be withdrawn and reissued as CX(86)98 Revised to take account of the Chief Executive’s comments and to include the technical justification.”

Again, I do not doubt the Executive minutes record only a small part of what was said, and it may well be that Dr Wilkinson provided some further explanation for his judgment, and for the confidence of THORP’s designers and the plant operators. But it seems to me clear that no such technical justification was given in the paper itself.

In reaching that conclusion I have not ignored the fact that by 28^th September various of Dr Wilkinson’s subordinates had begun a process designed to enable a decision to be made whether, and if so with what degree of confidence, BNFL should formally express to BLCs an opinion that THORP was capable of handling an additional 1,000tU baseload within the 10 year period. It is reasonable to assume that this process began shortly after the main board reached its decision in principle on 28^th August, but it may have started a little earlier. The first documentary evidence of the process consists of Dr Jeal’s Memorandum dated 5th September. That document shows that he had been requested by Mr Bennett at an earlier meeting to review “the stochastic model work with the intention of establishing a single “guaranteed throughput”, and that before preparing his Memorandum he had carried out certain enquiries and performed or caused to be performed by his team some modelling work specifically directed to a response to Mr Bennett’s inquiry.

Dr Jeal told me, and I accept, that Mr Bennett asked him two questions, to be addressed by reference to the conclusions to be derived from stochastic modelling, of which Dr Jeal was in charge. The first was as to the “guaranteed” throughput, in terms of tU per year. By “guaranteed” Dr Jeal understood Mr Bennett to mean a level of which BNFL could be “reasonably sure” and which would be satisfied for example by its achievement in nine out of ten campaigns. Both he and Mr Bennett were well aware that there could be no absolute guarantee of any level of throughput by THORP. The second question was whether, in the light of the answer to the first question, THORP could handle 7,000tU in 10 years.

Dr Jeal and his small team addressed those questions first by going to see each section of the very large THORP design team (amounting to more than 1000 staff), to ascertain whether any relevant stochastic model had been altered since their previous involvement with it, and, separately, whether any design changes had occurred which called for a change to a relevant model, which had yet to be implemented. They took as their starting point the document known as P275, since there had been no comprehensive written summary of the process of stochastic modelling since that date, and they reviewed all the changes and further modelling work which had in fact been done since then, but which had not been fully reported. Since the provisional throughput data in P275 suggested that THORP could achieve a throughput of 1,000tU per year, they concentrated on design and modelling changes which might lead to a reduction in capacity, rather than those which might increase it. They carried out a limited number of model runs to ascertain the consequences of those changes in terms of throughput.

The results of that exercise are set out in a compressed summary on the first page of the Jeal Memorandum. Three potentially adverse changes of design to the HEP were identified. One of them was an increase in the likely duration of control system breakdowns. Unfortunately, when inserting that change into the then current HEP model, Dr Jeal’s team inadvertently increased the number as well as the duration of such breakdowns, and the model unsurprisingly then delivered what Dr Jeal regarded as a “clearly very pessimistic” throughput, averaging 445tU per campaign or 890tU per year. Rather than re-run the HEP model with corrected input data, Dr Jeal simply expressed a judgment that, having regard to the error, the outcome of those runs did not undermine the conclusion to be drawn from P275, which he described as “still directly applicable” and indicating a range of 966-1300tU per year, and justifying a conclusion that the HEP was capable of 1,000tU per year, based on a range of availabilities from 52% to 65%. His expression of the throughput per campaign (upon which the annual throughput was based) contained a typographical error, identifying 403tU per campaign rather than 483tU per campaign as the bottom of the range, but this was not repeated in the figure stated as the bottom of the annual range, and cannot have misled anyone.

Dr Jeal concluded his summary of his analysis of the HEP model with a sideways look at papers available to him suggesting that the availability in a particular year of the shear and dissolution parts of the French UP2 plant was 62%, which he described as “lending some support to our calculated figures”.

Dr Jeal has been heavily criticised, in particular by Professor Von Collani, for the lack of rigour or discipline inherent in that summary of the review of the HEP model. I shall address those criticisms in due course. For the purposes of these factual findings, I merely note that the computers then in use by Dr Jeal’s stochastic modelling team had already displayed signs of being at the limits of their capacity and, as is familiar to any computer user, that leads to a reduction in speed of operation. The evidence suggests that runs of the stochastic models could take many hours, whereas today they might easily be assumed to take minutes.

In relation to the SEP, Dr Jeal’s team input certain additional breakdown data beyond that in the then existing HAC model, and derived a range of campaign throughputs between 510 and 580 tU with an average of 546tU and an availability of 85.7% based on 10 campaigns. Dr Jeal described this as a worst case, and based on more pessimistic breakdown data than were being input into the new HAC model then in development. While acknowledging that the 10 run sample was rather small for statistical reliability, he nonetheless recommended the adoption of a “guaranteed” throughput of 1,000tU per year for the SEP.

Again, Dr Jeal has been roundly criticised by the defendants for seeking to derive any reliable conclusions from only 10 runs of the old HAC model, but the same computer speed background must be borne in mind when addressing those criticisms.

Dr Jeal’s overall conclusion in relation to the first question was that, for the HEP and SEP plants, a guaranteed throughput of 1,000tU per year should be adopted. As to the second question, Dr Jeal expressed his view as follows:

“I believe that this throughput guarantee will enable an additional 1,000tU of base load fuel to be processed through HE & SEP. However, I have not checked any of the following:-

i.

HA evaporator/ vitrification

ii.

Salt Evaporation

iii.

Central Effluent Plants

iv.

Solid Waste Handling

v.

Ponds

vi.

Environmental impact considerations

Should you wish me to do any work in these areas please let me know. I note that Ponds are clearly a major problem.”

The point underlining Dr Jeal’s reference to the 6 stated items of plant not having been checked was that no stochastic modelling of any of those items had been undertaken by September 1986. Items (i) to (iv) were either existing plant or plant designed to be outside the THORP building itself. Although there were one or more Ponds within the THORP Receipt and Storage Facility, the Oxide reprocessing process also interacted with and to a limited extent relied upon existing ponds lying outside the THORP building. The final item was not of course a reference to any specific plant, but to a separate consideration capable of impacting upon THORP’s potential throughput which had not been modelled.

The reader of the Jeal Memorandum might wonder how Dr Jeal thought it possible to extrapolate from a guaranteed throughput of 1,000tU per year to a throughput of 7,000tU in ten years. He explained in his evidence that he had factored in a commissioning period of 21 months as (by then) laid down by the Chief Executive of BNFL (although he thought that 12 months would be more realistic), leaving a contingency in excess of one year to recover from any unforeseen malfunctions. But as he was at pains to point out, and I have no doubt that Mr Bennett well understood, issues of commissioning, ramp-up and major outage did not fall within Dr Jeal’s area of responsibility, since none of them had been input into the stochastic models, and Mr Bennett already knew enough about them, from his continual involvement in the design project, to be under no illusions in that regard. It follows that Dr Jeal did not expect Mr Bennett to place much reliance upon his answer to the second question, other than in the negative sense that his summary did not identify any specific warning signs to be derived from the stochastic modelling itself, in connection with the extrapolation of a 10 year throughput estimate from models designed to replicate the possible outcomes of a single typical campaign. Again, with the emphatic support of Professor Von Collani, the defendants strongly criticised the use of single campaign stochastic modelling (regardless of the number of replications) as the basis for any reliable extrapolation of a 10 year throughput.

It is clear that Dr Wilkinson, Mr Bennett and Mr Smith (the three members of BNFL’s senior staff involved in the appraisal of the 1,000 tU baseload increase in late 1986), did not regard stochastic modelling as the sole basis for forming a reliable throughput estimate. They were alert to the warning implicit in the second page of the Jeal Memorandum (if, which I doubt, they needed it) about a throughput estimate needed to address the whole of the Oxide reprocessing plant, not merely that part of it lying within the THORP building itself, which was the only part that had already been modelled. The output data from stochastic modelling of the plant within the THORP building was only one part of the necessary analysis of a multi-faceted problem. The other parts of it were regarded within BNFL as either being inherently, or at least at that stage, incapable of being addressed by a stochastic approach. I have already explained why (pace Professor Von Collani, to whose contrary opinion I shall return) the effect on a 10 year throughput of major outage was inherently unsuitable for stochastic modelling, due to the absence of reliable input data. For similar reasons, commissioning and ramp-up were regarded within BNFL as similarly unsuitable. By contrast, the performance of specific plant outside the THORP building itself was no less suitable for stochastic modelling than plant within that building. The difficulty in late 1986 was that the ancillary plant had simply not by then been modelled, and parts of it were at a stage of design which would render the throughput data from any models unreliable. For those facets of the problem, BNFL decided to rely upon judgment, based upon its experience, and upon the specialised skill of its staff.

For that purpose, but on a date which is not now possible to identify, a meeting was convened at Mr Smith’s office at Sellafield to address the 10 year throughput question with the added benefit of the perspective and experience of plant operators. No minutes were taken, or if they were, they have not survived. Since it took place more than 20 years ago, recollections are at best sketchy. The most that can be said about it within any confidence is that it was well attended (in fact better than Mr Smith had expected), that representatives from a wide variety of disciplines, including both plant operation and design, attended to offer their input, and that it was conducted in an atmosphere in which the objective of forming a reliable 10 year throughput estimate was uninfluenced by commercial pressure. The meeting was conducted by way of an informal discussion between experts in their respective disciplines.

The nearest that it is possible to get after 21 years to the outcome of that meeting (to which I will refer as “the Smith Meeting”) is to be found in a memorandum by Mr Smith to Dr Wilkinson dated 5^th December 1986 (“the Smith Memorandum”). In evidence, Mr Smith thought that some time had elapsed between the date of the meeting and the Smith Memorandum, so that the Smith Memorandum represents not merely the wisdom gained at the meeting, but also the accumulation of further work by and input to him after it.

Before describing the Smith Memorandum in detail, I must mention certain other events which had occurred in the meantime. It will be recalled that Dr Wilkinson had been charged by the Executive meeting on 16^th September with the provision of a technical justification for the extension of the baseload and with the consequential reissue of CX(86)98. The next meeting of the Executive occurred on 7^th October, at which Dr Wilkinson reported (according to the minutes) that:

“the first stage of this exercise had been completed and that an OR [Operational Research] exercise had demonstrated that THORP could cope with a 7,000t baseload over the first 10 years of operation. Work is now being carried out to assess whether all ancillary plants will be able to cope with the increased workload. The results from this exercise will be distributed to members of Executive as soon as they are available and by no later than 17^th October.”

On 7^th November there was held the 17^th meeting of the Techno-Economic Sub-Committee between non-UK BLCs and BNFL, preceded by a pre-meeting on 6^th November between BLCs only. By then, BNFL had issued its October 1986 THORP Business Plan which, for the first time, depicted a series of annual throughputs the effect of which in aggregate was to achieve a 6,000tU baseload in 9 rather than 10 years. Whereas all previous business plans had contemplated a normal (i.e. post ramp-up) annual throughput of 650tU, the October 1986 Business Plan increased that to 900tU in years 6, 7 and 8, preceded by a much slower ramp-up of 5 years, and, as anticipated, a two year delay in commencement, from 1990/ 1991 to 1992/93. The Business Plan also predicated an increase in reprocessing costs and prices. As the minutes of the 17^th TEC show, BNFL’s expectation that the publication of this Business Plan would cause concern to BLCs was well founded.

Paragraphs 4 to 17 of the minutes of that meeting summarised opening remarks made by Mr Tindle on behalf of BNFL. Rather than paraphrase, I quote the relevant passages below in full:

“11.

Mr Tindle said the Business Plan assumed a slower build up in throughput once the reprocessing plant came into operation and levelled out at a higher throughput. BNFL was confident that THORP could reprocess more than 6,000t in the first 10 years of operation. The assumption in the Business Plan that 6,000t would be reprocessed in approximately nine years permitted an easy extension to produce a Business Plan for the reprocessing of 7,000t in ten years.

12.

Mr Tindle added that careful studies within BNFL had indicated that THORP was capable of meeting this target. Checks were currently being undertaken on peripheral items of plant and BNFL would hope to make a more positive statement of this subject at the Joint Committee.

……

16.

In response to a question from Mr Rieger, Mr Tindle said BNFL would hope to issue a 7,000t Business Plan in January 1987. He suggested that it might be possible to have an extra TEC meeting in the early part of 1987 at which the Business Plan could be discussed and a decision taken on how best to proceed.”

It is a fair but not inevitable inference that by this time the Smith Meeting had occurred, or at least that Mr Smith had started to bring his opinions about throughput to bear on the developing analysis. He was the principal advocate of a slow ramp-up, as the means for achieving the best long-term yearly throughput in a highly complex plant such as THORP. Mr Tindle had some recollection of having attended the Smith Meeting, and he had certainly spoken to Mr Smith before the 17^th TEC meeting. He said that his reference to “careful studies” was intended to refer both to the stochastic modelling and to the analysis conducted at the Smith Meeting.

On the same day as the TEC pre–meeting (6^th November) there was a briefing of the BNFL Executive at which, consistent with Mr Tindle’s statement to the BLCs, the Executive was informed by Dr Wilkinson that “Site staff are assessing whether the peripheral plant can cope; they are unwilling to give an ill-considered response but anticipate resolving the issue well before the next JRC meeting.”

On 21^st November the Executive was informed by Dr Wilkinson that:

“a study of THORP and all peripheral plants had now been completed and that feasibility of increasing throughput in the first 10 years to 7,000tU had been approved. He further reported that he would now make arrangements to reissue CX(86)98 as CX(86)98 Revised.”

Like its predecessor, CX(86)98 Revised (if ever prepared) has not survived. At a further meeting of the Executive on 3^rd December, the minutes show that Dr Wilkinson:

“reported that Reprocessing Operations Division had confirmed that it would be possible to reprocess 7,000tU during the first 10 years of operation of THORP.”

The important 17^th meeting of the Joint Committee was scheduled to take place in Japan on Tuesday 9^th December. Mr Smith sent his Memorandum to Dr Wilkinson on the previous Friday, 5^th December, after Mr Tindle had left for Japan, and two days after the meeting of the Executive at which Dr Wilkinson had reported that the review into THORP’s increased reprocessing ability had been completed. I infer therefore that Mr Smith had already reported orally to Dr Wilkinson broadly in the terms reflected in his Memorandum of 5^th December, and that the Memorandum was designed as an aide memoire for Dr Wilkinson, both in briefing Mr Tindle in Japan and (perhaps but probably not) for preparation of the promised Addendum. It therefore constitutes the best surviving written record of the view of a senior member of BNFL’s staff as to THORP’s ability to meet a 7,000tU in ten years baseload, immediately prior to Mr Tindle’s Rigorous Technical Review statement to BLCs at the Joint Committee. It therefore merits close scrutiny, albeit conditioned by the need to bear in mind the likelihood that, when briefing Mr Tindle as to what to say at the meeting in Japan, Dr Wilkinson had access to other contemporaneous written material. I bear in mind also that Dr Wilkinson is likely to have been in regular contact with the THORP design team throughout, and briefed on further developments, including stochastic modelling, which had been going on in the meantime. I set out the Smith Memorandum below in full:

The Smith Memorandum needs careful interpretation in the light both of Mr Smith’s evidence as its author, and Dr Wilkinson’s evidence at its intended recipient. Two general points need to be noted before considering its detail. The first is that it is likely that neither Mr Smith’s opinion, nor the basis for it set out in the Memorandum, was unknown to Dr Wilkinson prior to his receipt of it. He said (and I accept) that he must have received Mr Smith’s views orally before the Executive meeting on 3^rd December, to have been able to inform his colleagues that “Reprocessing Operations Division” had confirmed the possibility of reprocessing the increased baseload in 10 years, since Mr Smith was the head of that Division, and since he was not in the habit of lying to his Executive colleagues.

Secondly, both Mr Smith and Dr Wilkinson said that there was an element of hidden agenda in the Memorandum. Dr Wilkinson described it as having built into it an element of self-protection. Mr Smith was, as a typical cautious plant operator, “covering himself”. Mr Smith described the agenda rather differently. He said that, while he had full confidence in THORP’s ability to reprocess 7,000tU in 10 years, he recognised that the resolution of the “already identified problems” referred to in paragraphs 5 and 7 of the Memorandum would require “energy and enthusiasm”, and one purpose of his Memorandum was to emphasise that point, so as to ensure that his superiors would provide the necessary financial and other support in those areas.

Either of those hidden agendas is likely to mean that an objective reading of the Memorandum would reflect a lower degree of confidence in THORP’s ability to meet a 7,000tU in 10 year baseload than its author actually believed. I bear in mind that, as a potentially damaging document, when compared with what Mr Tindle told BLCs on the following Tuesday, it would be tempting for loyal former employees and officers of BNFL to wish to apply a little corrective spin. But in my judgment both Dr Wilkinson’s and Mr Smith’s references to their interpretation of that hidden agenda are both credible and reliable, and I accept both of them. They are different but not incompatible.

I turn to the interpretation of the detail, beginning with paragraph 2. It will be recalled that Mr Smith only became the General Manager of Sellafield in 1984. He told me that he had not been involved in the early planning of THORP, and that his references to business plans were to details reported to him, rather than to documents which he had studied himself. There had not in fact been a Business Plan prior to December 1986 which, in terms, proposed a throughput of 1,000tU per annum with 2 years outage. I consider that this was a reference by Mr Smith to throughput assumptions based, as I have shown, on a “normal” throughput of 1,000tU per annum, with allowances for ramp-up and major outage. The surviving documents show major outage allowances of four and three years. It may be that, between 1983 and 1986, a three year allowance was refined down to two years, but I have not found any specific documentary evidence of it. Nor is there any evidence of the major outage concept being “dropped” other than the Smith Memorandum itself. The evidence of all BNFL’s witnesses was that an initial perhaps resigned acceptance that there would be a major outage had been replaced by a growing confidence that, as it developed, the design was such as substantially to reduce that risk, leaving it to be included, with other risks not susceptible to stochastic modelling, within a general contingency, rather than a dedicated major outage allowance.

Mr Smith’s reference to the “advent of MOX” is to a form of nuclear fuel for which the Service Agreements did not provide a reprocessing service. Accordingly, its advent did not need to be considered.

Mr Smith did not claim to be either an avid student or expert in the art of stochastic modelling. He told me (and I accept) that its outputs were reported to him from time to time and in particular at the Smith Meeting. He could not recall whether he had received or studied the Jeal Memorandum, but considered that the substance of it had been explained to him.

The final sentence of paragraph 2 only makes sense if Mr Smith’s reference to the Business Plan was to a plan earlier than BNFL’s October 1986 Plan. That plan did, where as the earlier plans did not, include a ramp–up provision sufficient to allow for a 21 months commissioning period.

As for paragraph 3, it is at first sight curious to find a statement that, at the same time, Civil Work had only 2 years to go to completion, in relation to a highly complex plant which was “not yet designed”. In my judgment this terse comment meant and was understood to mean no more than that the design process was not complete. The evidence was that the design of different parts of THORP proceeded at different speeds, such that, at any given time, parts were designed and built, other parts were fully designed, and other parts were still the subject of flow-sheets, rather than precise physical design. The extent to which BLCs were aware of this reality is matter to which I shall have to return.

Mr Smith refers in paragraph 4 to a production program, which he sets out on a year by year basis. He made it clear in evidence that it was merely a rough and ready illustration of how a 7,000tU in 10 years baseload might be achieved, consistent with the Chief Executive’s decision about commissioning, and his own views about the wisdom of a slow ramp-up. It was not, he said, a programme likely to find its way into a Business Plan, not least because it provided for the completion of a 7,000tU baseload in only 9 years. To understand the table, it is necessary to bear in mind that it was based upon an assumption that reprocessing would start (for commissioning purposes only) in the fourth quarter of the 1993/4 business year, which itself ran from April to March, i.e. in or about January 1994. The 10 year period would therefore conclude at the end of December 2003. Mr Smith’s figure of 720tU for the business year 2002/3 assumed, not a deterioration in annual throughput, but that the 7,000 tU baseload would be completed just before the end of the third quarter of that business year, i.e. by the end of December 2002, leaving the last quarter of the 2002/3 year and three quarters of the 2003/4 year as a contingency within the 10 year period.

Properly understood therefore, Mr Smith’s table projects a theoretical 7,000tU baseload in 10 years, is based upon a “normal” annual throughput of 1,000tU, as advised by Dr Jeal, a commissioning period of 21 months, within a total ramp-up period of 5 years, and a general contingency of 1 year or, or at “normal” throughput, 1,000tU.

Mr Smith’s reference in paragraph 5 to areas of uncertainty is similar but not identical to Dr Jeal’s list of “not checked” items, but is based upon Mr Smith’s much greater experience of the operational implications of the increased baseload upon plant outside the THORP building itself. No detailed written explanation of the process of examination of those areas of uncertainty survives, nor was any witness’s recollection of them explored in minute detail. As I have already stated, it is reasonable to suppose that Dr Wilkinson was aware of the nature of the areas of uncertainty and the “already identified problems” referred to in paragraph 7, and did not need a detailed written explanation of why the solution of the problems was likely to be as efficacious for a 7,000tU baseload as it would be for a 6,000tU baseload.

The explanation which I was offered in part by Mr Smith and in part by Mr Donaldson (and which I accept) was as follows. First, all the items of plant identified in paragraph 5 were to be shared by the MAGNOX and Oxide reprocessing streams. For example, the phrase “HA Evaporator/ Vitrification” is a reference to HALES and the WVP which, as I have already described, were designed to take and to blend HAR from both reprocessing streams. Accordingly, whereas the addition of 1,000tU to an original baseload of 6,000tU represents a substantial increase in the burden, the addition of 1,000tU to the combined baseload of THORP and the MAGNOX plant is of much less significance.

Secondly, most of the ancillary plant identified in paragraph 5, with the exception of the WVP, already existed, but needed enlargement or re-design and alteration to cope with the addition of an Oxide stream to an existing MAGNOX stream, regardless of the specific projected throughput of the Oxide stream. Thus for example, one of the three evaporators in the HALES plant was being re-designed to enable it to accept Oxide HAR. This involved detailed planning and testing, because of the different chemical make-up of Oxide HAR, before a single tonne of that material could be evaporated into HAL. As another example, the Sellafield site had an extensive internal railway system. Part of the Oxide fuel to be reprocessed at THORP has to be transported from ponds outside the THORP building into the THORP Receipt and Storage Facility, which itself included a large pond. That transport was to take place in flasks carried by rail, and detailed design and testing had to be carried out to ensure that this could be done safely, and without causing bottlenecks.

Mr Smith’s and Mr Donaldson’s evidence was that the design processes necessary to accommodate the addition of the THORP Oxide reprocessing stream had not by December 1986 reached the stage at which the planned solutions to those problems could be stochastically modelled, but they had reached a stage at which it could be said with confidence that solutions were capable of being designed, that the proposed increase in THORP’s baseload did not present a challenge to the efficacy of those solutions, and that, as Mr Smith put it, with appropriate energy and enthusiasm, the problems would be overcome.

Paragraph 6 is impenetrable unless unravelled. The acronym “ALARP” is a reference to the regulatory principle constituted by the phrase “As Low As Reasonably Practicable”. It summarises a principle that all available safety measures in the nuclear context must be taken unless the economic cost of implementing them would be grossly disproportionate to the benefit. Since the objective of the increased baseload was, as Mr Smith summarised in paragraph 1, to reduce the unit cost of reprocessing spent fuel, he was recognising in paragraph 6 the possibility that reduced unit costs might unlock and therefore require the implementation of safety measures which at the higher unit cost would have been grossly disproportionate in cost benefit terms. It was a prudent warning, but one which has not featured in any part of the defendants’ criticism of BNFL’s throughput estimates.

Finally, paragraph 7 needs the obvious correction of “date” to “data” and its cautious phraseology to be interpreted in the light of the hidden agendas to which I have referred. Mr Smith did not intend it to be interpreted as casting doubt on an affirmative outcome to the process of appraisal of the increased baseload in which he had been engaged, and Dr Wilkinson did not so interpret it. Nonetheless, the Smith Memorandum contained pointers which any person seeking to carry out a retrospective examination of the level of confidence with which BNFL could reasonably have predicted the proposed increased 10 year baseload would ignore at his peril. Needless to say, the defendants rely heavily on it, and I shall have to return to its cautionary pointers in due course.

I mentioned that Dr Wilkinson probably had other available sources of information with which to form a judgment as to an appropriate level of confidence in the increased baseload, when briefing Mr Tindle ahead of the (by then imminent) Joint Committee meeting. I have in mind in particular the fact that the stochastic modelling process had not been standing still between the beginning of September and the beginning of December 1986. The main emphasis during that period had been upon perfecting the revised HAC model which, it will be recalled, was not in a sufficiently final state to be able to provide much assistance to Dr Jeal when preparing the Jeal Memorandum in September.

A new HAC model had been substantially completed, sufficient for it to be able to begin to deliver preliminary results, by the beginning of December 1986. A report on those preliminary results, under the reference OFCPDC P413, was circulated on 12^th December, ahead of a meeting of that Committee on 18^th December. It was a much more sophisticated model than that reported on in P275, including in particular a greatly increased number of breakdowns, and its object as defined on the title page of the report was aimed at confirming the operability of the HA cycle. The conclusion reads as follows:

“The preliminary results of the HA Stochastic Model Study have been presented. These data should be treated with care since they represent only a stage in the development of the study. However, they show throughput requirements are achievable with existing buffer sizes.”

Dr Jeal said that the new model in its form as reported in P413 was constructed on a worst case or pessimistic basis, and accordingly in his view produced a pessimistic response. That view is reflected in P413 itself. Nonetheless, the report stated that the model nonetheless produced an average throughput per HAC campaign of 525tU, once certain typographical errors in its text are corrected, as they were at the meeting of the OFCPDC on 18^th December. By reference to the histogram enclosed at Figure 6 to the report, 40 simulated campaigns produced throughputs varying between 420tU and 600tU, with the bulk occurring between 500 and 580tU. Of course, this model replicated only one part of the THORP process, but the HAC contained the most important parts of the untried new technology, and the parts which BNFL’s designers were most concerned to study and refine.

The evidence does not permit any safe conclusion to be drawn as to the extent to which the preliminary output of the new HAC model, and its provision of additional comfort beyond that reported by the Jeal Memorandum, was reported to Dr Wilkinson prior to the Joint Committee meeting on 9^th December. Dr Jeal could not recall being asked for any further input to the increased throughput issue after the preparation of his Memorandum, but I consider it probable that some intimation of the preliminary results of the running of the new HAC model was reported by him, at least to Mr Bennett, as a matter of ordinary routine, as and when they emerged. The dating of the preparation and circulation of P413 was tasked by reference to the approaching date of the next OFCPDC meeting, rather than to the Joint Committee meeting in Japan. It is reasonable to assume at least that Mr Bennett and Dr Wilkinson had some awareness that the ongoing process of stochastic modelling had not in any way undermined the judgment expressed by Dr Jeal as to the normal annual throughput to be expected of THORP in his September Memorandum.

The December 1986 RTR Statement

Against that background, I turn to the Joint Committee meeting itself. It was the first to be chaired by Mr Tindle. The minutes record him as saying this, as part of his introduction:

“It had been mentioned at the TEC that BNFL may be commercially willing to extend the throughput of THORP for the first 10 years to 7,000t. Mr Tindle reported that a rigorous technical review of the peripheral facilities as well as the THORP Head End and Chemical Plant had now been completed, and he was able to confirm officially that it had been accepted at the highest level in BNFL, by both the designers and the Sellafield operators, that a throughput of 7,000t could be achieved in the first 10 years. A 7,000t Business Plan was being prepared to show the effect on prices. It was believed that the result would be a reduction in unit prices to those in last year’s Business Plan.

Mr Tindle said he was aware that utilities [by which he meant the BLCs] could not confirm immediately their acceptance of a share of the increased throughput. He foresaw discussions over a period of some two to three years to determine details. BNFL would show flexibility in order that all customers could benefit.”

Mr Tindle’s statements were made to a meeting attended by representatives of many of the BLCs but, perhaps ironically, not including the defendants. The minutes were circulated both for approval, and for the information of all BLCs, including the defendants, and may be taken as a concise but reliable record of what he said. There is certainly no different or better record.

Mr Tindle was cross examined as to the basis for the statements which he made, both as to the nature of the review and as to the expression of BNFL’s confidence in the 7,000tU in 10 years throughput which it contained. He said that, as far as he could recall, the phrase “rigorous technical review” was his own, rather than one which had been given to him. He said that the basis for his belief for making the statement reported in the minutes consisted of information given to him by, or obtained by him speaking to, Dr Wilkinson and Mr Smith, and that he had Dr Wilkinson’s authority to express the gist of what he told the BLCs, although not necessarily the precise words which he used. He was sure that the body of information upon which he based his statement was greater than that which he could recall, over 20 years later. I have no doubt that he was correct in that respect.

Dr Wilkinson confirmed that he thought it unlikely that he would have invited Mr Tindle to use the phrase ‘rigorous technical review’, but that he would specifically have authorised Mr Tindle to make, in the precise terms in which he made it, the rest of his statement. In cross examination he said he thought that the phrase ‘rigorous technical review’ was a fair description of the work which had been going on in relation to the THORP project for several years, and that although he regarded Mr Tindle as probably reasonably conversant with it, he would have summarised its essential features to Mr Tindle when briefing him before the Joint Committee meeting. There was no hint in Dr Wilkinson’s evidence that in using the phrase ‘rigorous technical review’, Mr Tindle had exceeded his authority. He had merely used his own phrase.

Like Dr Wilkinson, Mr Tindle said that he used the expression ‘rigorous technical review’ as a reference not merely to the work carried out in late 1986 in addressing the specific question of the increase in THORP’s throughput, but as a reference to all the work in the nature of review which had been going on for several years. He said that he was intending thereby to include the stochastic modelling of the plant within the THORP building, and that he was aware that this modelling had not extended to the peripheral plant. He regarded the peripheral plant as having nonetheless being carefully studied, with a view in particular to ascertaining whether its design or operation might give rise to bottlenecks which would put the achievement of a 7,000 tU baseload at risk.

Mr Tindle said that he was unaware of any special meaning of the phrase ‘rigorous technical review’, and meant by his use of it only that there had been a review, that it had been rigorous and that it had been technical. In cross examination he said that he was amazed at the superstructure which, in their case, the defendants had placed upon the phrase. When asked to expand upon his use of ‘rigorous’ he said that he meant careful and thorough. I shall in due course have to decide what that phrase, and the statement of which it formed part, should be taken to have meant to his audience, for the purposes of founding a claim in misrepresentation or mistake. With his greater direct knowledge, Mr Smith said in cross examination that, as at December 1986, and regardless of how far back in time he looked, there had not been a rigorous technical review of the peripheral plant. Nor did he regard the analysis which he had personally carried out in late 1986, the results of which was reflected in Smith Memorandum, as amounting to a rigorous technical review of the HEP or SEP.

1987 to 1989 – the lead up to the Side letters

Since it was almost another 3 years before the Side Letters were made, it is necessary to examine the continuing process of the design of THORP and its ancillary plant during the intervening period, and the continued communications between BNFL and the BLCs in relation to it.

It will be recalled that at the 17^th TEC meeting on 7^th November 1986, Mr Tindle had in response to requests from BLCs suggested that a further TEC meeting could be held in early 1987 for the purposes of discussing a 7,000tU Business Plan. The task of preparing and then speaking to a paper to that effect fell to Mr Trevor Moulding, the Head of Future Plant Department at Sellafield. His Report was entitled “Operational Implications of Increasing the THORP Throughput from 6,000 to 7,000tU in the First Ten Years of Operation”, with the reference number TEC(87)1. He presented it to the 18^th meeting of the TEC on 13^th March 1987, held in Essen, Germany. The report had attached to it as Table 1, a revised reprocessing program intended to be included in the 1987 Business Plan. An earlier draft of the report had a Table 1 which also included a “Contingency Programme” in a separate column. The Contingency Programme was a precise replica of the programme set out in paragraph 4 of the Smith Memorandum, but excluding the blank amount for 2003/4. It had been intended to include the Contingency Programme to show what could be achieved by the end of 2002/3 if THORP’s start-up was delayed until 1993/4.

The 1987 Business Plan programme included as Table 1 in the final version of Mr Moulding’s Report assumed a commencement of reprocessing in December 1992. Like Mr Smith’s programme the 1987 Business Plan programme also provided for a 5 year ramp-up, to achieve a normal throughput of 900tU, for 4 years during the business years 1988/9 to 2001/2. The three quarters of the final year was programmed at the same rate, so as to achieve 720tU by the 10^th anniversary of commencement. Just as the normal throughput was 100tU less per year than that of Mr Smith’s programme, so the ramp-up rate was also slower. After a 30tU throughput in the final quarter of the 1992/3 year, the throughputs in each of the following 5 years were, respectively, 100tU, 150tU, 150tU, 100tU and 100tU less than those in Mr Smith’s programme.

For convenience, I have set out both tables, as they appear in the first draft of Table 1 of Mr Moulding’s Report, below.

TABLE 1

THORP REPROCESSING FEED PROGRAMMES

The effect of the changes made in 1987 Business Plan programme by comparison with Mr Smith’s programme was to take out entirely his final year’s free contingency, but to spread it throughout the 10 year programme, by using throughput figures lower in each year by 100 or 150tU than those used by Mr Smith.

If the 1987 Business Plan programme is compared with the programmes which were set out in all business plans prior to 1986, it will be seen that they all involve the same approach to the inclusion of a general contingency, or contingency for major outage. Rather than programme a throughput designed to leave a free contingency or major outage period at the end, they cut it up and allocated it year by year. Thus for example, all Business Plans prior to 1986 showed a maximum normal throughput of 650tU, well below the throughput being considered, both as a design requirement, and as indicated by stochastic modelling. The only exception to that pattern is to be found in the October 1986 Business Plan, which was specifically designed to demonstrate that THORP could reprocess 6,000tU in 9 years.

The natural inference from that analysis is that the formulation of the 1987 Business Plan programme involved no change in assumptions about the ramp-up rate or the general contingency allowance from those which appear more clearly from Mr Smith’s programme. Both the ramp-up rate and the “normal” operating rate thereafter are included at a level usually 100tU below what Mr Smith predicted and 150tU below his prediction in 2 ramp-up years. That inference finds some, but by no means unambiguous, support in the evidence.

Mr Moulding’s report, and his commentary on it at the 18^th TEC meeting bear close scrutiny, because they constituted the main occasion upon which BNFL gave the BLCs any detail as to the basis on which it had come to upgrade its estimate of THORP’s 10 year baseload throughput. I shall quote the key passages in full, and summarise the remainder.

Under “introduction” Mr Moulding wrote as follows:

Under the heading “Forward Program” the Report explained how the MAGNOX programme was intended to be run side by side with the THORP reprocessing plants, at a level of 1,000tU per annum, but with a 50% fall off in throughput in the final year 2002/3. It then introduced the 1987 Business Plan programme, the assumption of the commencement of active operation in December 1992, and introduced in Table 1. The report continued:

“Thus the program considered requires annual throughputs of 900t per annum over a period of several years towards the end of the 10 year period. This compares with annual throughputs of 650t per annum in the earlier programmes. The assessment of THORP reprocessing capabilities was considered against 1,000t per annum to give some contingency. ”

The report continued under the heading “Thorp Plant Capacity” to describe the nominal quoted capacity of THORP as 1,000tU per annum being “the capacity that would be achieved in a good year when the overall plant reliability is high”. There followed a brief description of the stochastic modelling process, which stated that the HEP and SEP had been “extensively studied”, that initial block models have been replaced by more detailed assessment of individual plant sections, and that to obtain statistically valid results, the models had be run for “extended operating periods – usually a minimum of 20 campaigns.” After a reference to the source of availability and reliability data and to the use of sensitivity analyses “to ensure that conclusions drawn from the studies are robust” this section continued as follows:

“8.

Stochastic modelling for THORP concludes that throughputs in the range 1020-1160tU per annum are achievable within the framework of two 150 day campaigns. 1000t per annum is achievable whilst still retaining a margin.

9.

Although the daily design capacity of the plant is set at 5tU per day, a throughput of 1,000tU per annum is achievable with a highly active cycle throughput of 4.25tU per day. That implies running for 235 days per year plus allowances for start –up and run-down. Since the peak throughput comes after a number of years of establishing confidence in plant operations, this is seen as reasonable, and the required throughput of 1,000t per annum can therefore be considered to be attainable.”

Pausing there, Mr Moulding did not give evidence, and his working papers did not survive. The origin of the range of throughputs from 1020-1160tU per annum described in paragraph 8 of the Report as the conclusion from the stochastic modelling for the plant as a whole appears to have been the throughput figures emerging from Dr Jeal’s run of the (by then superseded) HAC model which he used to derive his range of 510/580tU per campaign in his Memorandum of 5^th September 1986, a result which the Jeal Memorandum itself describes as being of questionable statistical significance due to the smallness of the sample.

Mr Moulding’s report then provided a brief summary of BNFL’s review of four items of ancillary plant, namely (i) Ponds, (ii) Salt Evaporators, (iii) Low Level Liquid Effluent Treatment and (iv) Highly Active Evaporation and Vitrification, before concluding, for reasons briefly stated, that none of those items were considered likely to cause problems. Having regard to what later occurred, it is worth quoting what the report stated about Highly Active Evaporation and Vitrification:

“15.

No problem is foreseen with highly active evaporation, and an Operational Research model has been run to test impact on concentrate storage and Vitrification Plant capacity. Based on a Vitrification Plant start-up of July 1990 no problems are foreseen with the storage of highly active concentrate.”

Under the heading “Summary” the report concluded:

“16.

In summary an increase in the baseload from 6,000 to 7,000tU results in an extension of the number of reprocessing days per year rather than an increase in the daily throughput. Even at an average throughput of 4.25tU per day an annual throughput of 1,000t per year in the latter years of the baseload program is achievable. ”

When Mr Moulding’s report was tabled at the TEC meeting on 13^th March, the minuted reaction of the BLCs was that it was “self explanatory and reassuring”. Nonetheless the minutes show (no doubt in summary form) that Mr Moulding continued with the following oral explanation:

“Mr Moulding said that he would like to give some explanation. Table 1 indicate a 7,000t program which the paper showed was achievable over the first ten years based on a build up of confidence in the early years. It should not be read from paragraphs 8 and 9 that stochastic modelling gave confidence of 1,000t per annum for 9 consecutive years and therefore that a 9,000t throughput was possible over the period. It was also worth recording that excess capacity over nominal program was needed to allow from recovery from a postulated shut down of, say, up to two years. 7,000t was not a new absolute limit but was a throughput which BNFL had confidence that it could achieve in a ten year period ”

It is evident from an earlier part of the minutes that the 7,000tU Business Plan had been received by BLCs in sufficient time prior to the meeting for them to be able to study it. Another representative of BNFL, Mr Grimodby, said this:

“…BNFL had recognised that the increase in costs indicated in the 6,000tU Business Plan issued in November 1986 was unwelcome to some BLCs and had considered ways of alleviating the increase. Following a rigorous technical review, BNFL had been able to confirm at the last Joint Committee meeting that a throughput of 7,000tU could be achieved in the first ten years of plant operation. The review had covered the peripheral facilities as well as the Head End and Chemical Plant. The sole purpose of the Provisional 7,000t Business Plan which had been distributed in February 1987 was to provide indicative cash flow and pricing information.”

I infer from that exchange that the BLCs’ had been given the programme summarised in Table 1 to Mr Moulding’s Report in February, but not the report itself.

In response to a question as to the contractual character to be attributed to the proposed additional business, in the light of uncertainty among BLCs whether it was to be regarded as an extension of the existing Service Agreement or the subject of a separate contract, Mr Tindle said that it was offered as part of the existing Service Agreement, but not as “new business” within the terms of the contract, rather as a concession to bring unit prices back to what they were before the recent increase. He said that “the objective was to work towards terms which were equivalent financially to a 7,000t Baseload.” In response to further questions he said that “BNFL were endeavouring to find a way of working towards a situation which was financially equivalent to an increase in Baseload from 6,000t to 7,000t in a manner which was equitable for all customers.”

Slightly earlier, Mr Schleef, a representative from Preussen Elektra had asked what would happen if a 7,000 tU throughput was not achieved within the 10 years. Mr Tindle replied that:

“the position would be the same as if we currently did not reach 6,000t. Operating costs would be payable until the 7,000t was complete. It should not affect the marginal costs of the extra 1,000t and the risk already existed. Dr Hauser postulated a theoretical situation where the plant operated smoothly to reprocess the 6,000t but shortly after beginning the reprocessing of the additional 1,000t things went wrong. In these circumstances the cost of the extra 1,000t could increase rapidly. Mr Tindle said that this could occur only in that particular theoretical situation.”

I have quoted the above exchange to illustrate the fact that BNFL were not at this stage offering a contractual commitment or guarantee to reprocess 7,000tU in 10 years, any more than there was a pre-existing contractual commitment or guarantee in the Service Agreements in relation to the 6,000tU baseload. Whether BNFL did undertake a contractual commitment to that effect by the Side Letters is a question of construction to which I shall return.

I said earlier that the inference that the 1987 Business Plan programme was based upon an assumed general contingency requirement equivalent to that reflected in Mr Smith’s programme was supported, but not unambiguously, by the evidence. The description in Mr Moulding’s report, and his oral explanation of it recorded in the minutes which I have quoted in full, both show that BNFL regarded the 1987 Business Plan programme as having built-in general contingency, in addition to the 5 year ramp-up indicated by Table 1. His reference, recorded in the minutes, to a postulated shut down of, “say, up to two years” and the statement in paragraph 8 of the report that “1,000t per annum is achievable whilst still retaining a margin” suggests that Mr Moulding’s approach to contingency was that it could be found between a normal operating throughput somewhere between the range 1020-1160tU per annum and the annual throughput stated in the 1987 Business Plan programme. Since the crude average between 1020 and 1160 is 1090, it is possible to see where he may have derived his “up to 2 years” contingency, but there is no evidence that this was an amount of contingency specifically or carefully identified as necessary by any process of calculation or considered judgment. Mr Moulding’s objective in his oral supplement to his report, recorded in the minutes, was to fend off any attempt by BLCs to seek more than an additional 1,000tU by way of compensation for projected cost increases, and I find that his reference to “recovery from a postulated shut down of, say, up to 2 years” was no more than an off-the-cuff explanation of the reason why a 1,000tU annual throughput derived from stochastic modelling could not justify a 9,000tU throughput in 10 years estimate.

Mr Barnes submitted that some support for treating Mr Moulding’s approach as more thoroughly considered was to be derived from the fact that an Operational Research Group study of HAR treatment dated 3^rd March 1987 showed that the storage tank capacity at HALES was being considered in relation not only to an assumed throughput at THORP peaking at 900tU per annum, but also in relation to an accelerated programme peaking at 1,100tU per annum after only one and a quarter years’ ramp-up. The report’s conclusion, together with the accompanying table, show that it was envisaged that an accelerated program at THORP might under certain operating conditions give rise to a need for up to 4 extra storage tanks (HASTs) at HALES, and that further detailed analysis was called for in that respect. In the event, only 2 further HASTs were constructed. In my judgment the fact that an accelerated programme was considered comes nowhere near justifying the assumption that BNFL thought 1100tU was a sufficiently probable throughput that a contingency margin could properly be built on it.

Mr Donaldson was pressed hard in cross examination as to the basis for Mr Moulding’s approach for “up to 2 years” contingency margin, and was unable to explain how it could be justified. His evidence was that Mr Moulding must have made his contingency margin on a normal (post ramp-up) capacity of 1,000tU per annum, as he put it : “because the 1,000 tonnes was the figure which the designers had said would be achievable with confidence. I.e. it would be on average at least 1,000 tonnes that the plant would be expected to achieve within a year.” (Day 4 page 88 of the transcript.) He was then asked how from that starting point a 2 year contingency margin could be identified in the figures, and his reply was as follows:

“That is why I believe there was some explanation at the meeting, because there was not sufficient data here to actually come to the conclusion that it is 2 years. He must have illustrated to the meeting that he was expecting successful ramp-up to some degree and 1,000 tonnes a year as opposed to 900 tonnes a year in the subsequent period.”

In more general terms, all that Mr Donaldson was able to say about this issue, and in this he was supported by Mr Smith, was that the progressive improvement and refinement of the design of THORP over time had led to a steady increase in BNFL’s confidence in the ability of the plant to avoid or withstand the consequences of major breakdowns, such that it was appropriate to reduce, but not altogether to remove, the contingency margin from the calculations.

Dr Jeal’s evidence was that in extrapolating from a “guaranteed” annual throughput of 1,000tU to a 7,000tU throughput in 10 years, he was working on the basis of a contingency margin “in excess of a year to recover from any unforeseen malfunctions”. But this was not a judgment of his as to the reasonable requirement in relation to an issue which lay outside his competence. It was rather an assumption which he made as to the then thinking of his colleagues.

One final piece of Mr Donaldson’s evidence is revealing on this issue. He was asked whether Mr Moulding’s reference to “up to 2 years” suggested that a view had been taken within BNFL that it was necessary to allow for a shut down of up to 2 years. His answer was as follows:

“No, I read it as the converse. What he is suggesting is this nominal programme, if adjusted, would, in his view, give a protection against shut down of up to two years i.e the two years is a consequence, rather than part of the definition.

Q.A consequence rather than a part of the definition; what do you mean by that?

A It is a consequence of how much time is left over, having ramped up and achieved the 1,000 tonnes a year, rather than being a prerequisite built into the programme. I said earlier, I do not believe that it has ever been possible to define a specific period” (day 4 page 87 of the transcript)

This passage in my judgment gives the key to the attitude to a contingency margin within BNFL at the time. It was not treated as something which had to be identified in terms of amount and then included in the figures used to calculate the 10 year throughput estimate. Rather, the 7,000tU in 10 years objective was compared with the throughput derived from normal operation at 1,000tU per annum, discounted by ramp-up, and the difference between the two, once identified, simply subjected to the question: “is this a sufficient contingency margin?”. To that question, the responsible senior staff, principally Dr Wilkinson, Mr Bennett and Mr Smith, appear to have concluded, without any separate structured, detailed or thorough analysis, but purely as a matter of judgment and experience, that it was.

While the BNFL witnesses were unanimous in their recollection that the earlier perceived requirement of a specific contingency for a single major outage had been replaced by a combination of greater design assurance against major outage coupled with the need to retain a more modest but general contingency, the documents did not disclose, and the witnesses could not recall, any more specific identification of the required general contingency period.

In my judgment, the best evidence as to the contingency period regarded as built into BNFL’s ten year throughput estimates is not to be found in Mr Moulding’s “say, up to two years comment” but in the logic behind the final year’s gap in Mr Smith’s programme. He was not proceeding on the basis that a throughput of 1,000tU left a safe annual contingency margin above that figure, or that his ramp-up rate left any safe annual contingency margin above that either. He probably did think it possible that the THORP plant would out-perform its modelled normal annual throughput of 1,000tU, and that ramp–up might proceed more quickly than his programme suggested. But contingency margins are not built upon possibilities. As I have said, the contingency margin apparent from Mr Smith’s programme was one year’s throughput at the normal rate, namely 1,000tU. That contingency is the nearest which, after the passage of more than 20 years, it is now possible to come as exemplifying the attitude towards contingency margin which BNFL had reached by the end of 1986, and in my judgment, nothing changed by March 1987. Whether that was a reasonable general contingency margin is a critical question to which, after reviewing the expert evidence, I shall return.

Stochastic modelling continued, as anticipated, following Report P413. The next significant report was in a document entitled “Highly Active Cycle Stochastic Model Interim Results”, with the reference number OR/87/19/11, prepared by a Mr Murphy of the Operational Research Department and dated 9^th June 1987. It reported further refinement and use of the new HAC model, together with the results of a sensitivity analysis designed to identify the extent of the effect upon the plant’s performance of departures from the assumed reliability of plant equipment. By this time, the HAC model simulated the activity of over 140 separate equipment systems. The conclusion of the exercise is summarised in the following passages:

“With all pieces of equipment failing twice as often the plant would be totally inoperable.”

“The plant’s performance is fairly sensitive to the given reliability of the equipment as a whole. If all the equipment is less reliable by a factor of 1.5 the plant is inoperable.”

“Superficially, 1.5 does not seem a large margin for error. But considering that there are more than 140 equipment systems involved and that the reliability assessments err on the pessimistic side, it is extremely unlikely that each piece of equipment would fail 1.5 times more frequently than stated.”

The report went on to identify those specific items of equipment where a shortfall against assumed reliability would impact adversely on the performance of the HAC, and also to identify design changes which would produce good improvements in operability.

For present purposes, the analysis of that part of BNFL’s review of its design of the THORP plant, and its ancillary plant, prior to the making of the Side Letters, may be taken as having concluded, and as being expressed in, a further report entitled “ A Summary of THORP Availability and Throughput Studies by Stochastic Model” with the reference number OFCPDC/P466, produced in August 1988. It superseded P275 and P413, and set out to provide a summary of the stage which the whole of the stochastic modelling process had reached by that date.

P466 shows that by August 1988, three main stochastic models had been developed and used for repeated simulation of 150 day campaigns. The first two, to which I have already referred, are the HEP model and the HAC model. The third is the THORP purification, Finishing and Effluent Treatment Model. This modelled all the main plant in the main THORP building downstream of the HAC, including in particular the uranium and plutonium extraction and finishing lines, but not any ancillary plant (ie outside the THORP building).

The report also shows that by August 1988 Stochastic modelling of ancillary plant had been undertaken so as to assess the interactions between THORP and all those pieces of ancillary plant which were considered to have the propensity, should they under-perform, adversely to affect THORP’s throughput. In particular, the report shows that both HALES and the WVP had been modelled. Mr Donaldson said that it had been impossible to conduct this stochastic modelling of the ancillary plant in 1986 or early 1987, because its design (or modification) had not by then reached a sufficient stage to enable meaningful models to be prepared.

P466 shows that the 3 main models were run both separately, to refine BNFL’s understanding of their individual throughput characteristics, and together, so as to facilitate a stochastic prediction of the throughput of the main THORP plant as a whole.

The conclusions set out in P466 may be summarised as follows. First, each of the three individual parts of the main THORP plant separately modelled were predicted to achieve reliable throughputs in the region of 1,200tU per year, in line with their original design targets. Secondly, the main THORP plant was in the aggregate predicted to achieve 1,100tU per year, that reduction from the 1,200tU design being due to interactions between the plant areas. Thirdly, the overall throughput of the main THORP plant did not appear to be sensitive to modest overall reductions in availability, or to reductions in individual plant equipment reliabilities, save in two stated respects (the details of which do not matter). Fourthly, two items of ancillary plant (the details of which, again, do not matter) were identified as having the potential significantly to affect THORP throughput. Further study of those interactions was to be undertaken and recommendations were made for the provision of additional bufferage so as to detach the performance of THORP from any underperformance by one of them.

The final section of P466, headed “Further stochastic modelling of THORP” stated as follows:

“The design of THORP is now sufficiently advanced that no further changes are planned that would significantly change the design capacity or availability. Obviously changes may occur in external plants interacting with THORP, but here again it must be expected that these would not significantly compromise THORP throughput. Therefore, this paper is intended to be the final statement on the overall availability of THORP HECSP.”

The passage quoted above shows that by August 1988 BNFL had, to its satisfaction, reached a stage in its stochastic modelling sufficient to enable it to form firm views about THORP’s predicted throughput under normal conditions. That state of finality may be contrasted with the clearly provisional stage which the stochastic modelling had reached by the end of 1986, or March 1987, when referred to in Mr Moulding’s report to the TEC.

For completeness, I should add that the ongoing process of testing which I have already described continued during the same period. Since no issue as to the adequacy or otherwise of BNFL’s testing processes arises for decision, I need say no more about it.

Concern had been expressed at the TEC meeting in March 1987 at the ongoing cost of the design of THORP, for which of course the BLCs were paying on a cost-plus basis. It was resolved to set up a new committee named the Research and Development Working Group (“RDWG”) but in fact it met only twice during 1987. Its only relevance is that it operated as a forum in which BNFL could, and did, report to BLCs in considerable detail the on-going process of the design of the plant. Again, the detail does not matter, but BNFL’s reports to that committee, together with other less detailed reports to the JC and the TEC, show that BLCs were kept informed of the progress of the design process during the period of the negotiation of the Side Letters, and can have been under no illusion that (as has been alleged) they thought there was a completed design for THORP by December 1986.

The Side Letters

I have already quoted the recital in the Side Letters at the beginning of this Judgment. Since the letters are short, and so much turns on them, it is worth quoting them in full:

“Dear Sirs

1.

We refer to the agreement signed between British Nuclear Fuels Limited, now British Nuclear Fuels PLC (“BNFL”) and Kernkraftwerk Brokdorf GmbH (“the Company”) on 30 March 1983 and to the supplemental agreements thereto (hereinafter collectively called the “the Service Agreement”)

2.

When BNFL signed the Service Agreement with the Company relating to the reprocessing of 64 tonnes uranium, we expected that the reprocessing capacity of the Thermal Oxide Reprocessing Plant (THORP) would be 6000 tonnes uranium over the first ten years of operation. As the design and construction of THORP progressed, BNFL undertook a rigorous technical review which re-assessed the THORP reprocessing capacity as 7000 tonnes uranium over the first ten years of operation. This was reported to Baseload Customers at the Seventeenth Joint Committee meeting held in December 1986.

3.

The purpose of the present letter is therefore to agree on the provisions relating to the additional capacity to be offered to the Company in respect to the Service Agreement.

3.1

As a consequence of the additional capacity, each Baseload Customer can deliver to BNFL an additional quantity of fuel corresponding to its share in the additional 1000 tonnes uranium such share being determined by the same percentage as the share in the 6000 tonnes uranium. This situation of additional capacity is covered by the Service Agreement between BNFL and the Company and does not require amendments to the Service Agreement.

3.2

The additional quantity to be delivered from the Company will be 10.7 tonnes uranium before irradiation, this being one-sixth of the quantity of 64 tonnes uranium specified in the Company’s Service Agreement

3.3

Regarding financial aspects, we confirmed that no addition payment of construction costs for the THORP Head End and Chemical Reprocessing Plant will be required in connection with the additional 1000 tonnes uranium, since the initial investment has been paid in connection with the first 6000 tonnes uranium and any re-investment costs will be treated as operating costs. An agreed allocation of costs and corresponding Land Utilisation Charge (LUC) and any fee will be payable in respect of marginal additional capital arising for example, from any additional pond capacity and multi-element bottles which are required. Waste management capital costs and corresponding LUC and fee will be allocated in accordance with the presently established practices.

It is presently estimated that the unit price of the Company’s addition quantity, in January 1989 money values, is approximately £300,000 per tonne uranium. As part of its obligations under the Service Agreement BNFL will inform customers annually of its most recent estimate of the price of this additional quantity.

The payment to be made in connection with the additional operating cost and corresponding LUC and fee will be made in accordance with the provisions of the exiting Service Agreement. However, BNFL will accept Letters of Guarantee from the Company’s bank with respect to all of the prepayments due on delivery of the additional fuel.

3.4

The Company shall deliver the fuel after the last delivery of its share of 6000 tonnes uranium according to a schedule to be agreed with BNFL which facilitates continuous operation of the THORP Head End and Chemical Reprocessing Plant.

3.5

Other terms and conditions of the existing Service Agreement remain applicable to the additional quantity.

You are invited to signify your acceptance of the above by signing and dating both copies of this letter and thereafter returning one copy to BNFL.

Yours faithfully”

THORP’s Actual Performance

The THORP plant began ‘active’ commissioning (i.e. using spent fuel rather than inactive test materials) during the last quarter of the 1994/5 business year. The first HAR from THORP reached the HALES plant in March 1995. By that time, the throughput during the commissioning and ramp-up period contemplated in the January 1987 Business Plan had been slightly remodelled, but without any alteration in the aggregate throughput planned. It had originally been 30:150:400:600:700:800. In the October 1992 Business Plan it was revised to 30:210:340:600:700:800, by the addition of 60 tonnes in the first full year, and the subtraction of the same amount in the second full year. Consistent with both Mr Smith’s program and Mr Moulding’s statement to the TEC in (March) 1987, the expectation was, or should have been, that THORP would out-perform this rate, and thereby build up a general contingency margin, providing some cushion, for example against major outage. I say “or should have been” because no evidence was adduced as to whether this expectation persisted after 1987. The further stochastic modelling which I have described suggests that it did, because, as is common ground, nothing in the design, testing and modelling process after March 1987 detracted from the confidence in THORP’s performance then expressed to BLCs.

In fact, THORP did slightly out-perform both the 1987 commissioning and ramp-up plan, and the amended version introduced in October 1992, but only for the first three full operating business years. The actual figures were as follows;

The result of that performance was to exceed the aggregate planned by the end of the third full year of operation by 282tU. But it fell short of Mr Smith’s prediction (against which the business plans were intended to leave a margin) by 118tU. In other words, THORP achieved a little over two-thirds of what may be supposed to have been the anticipated out-performance of its planned commissioning and ramp-up throughput by the end of its third full year of operation. Bearing in mind its expected “normal” annual throughput after ramp-up of 1,000tU per annum, this provided a cushion against, for example, unplanned outage of slightly more than 3 months. This margin had been built up notwithstanding unplanned outages in the 1997/8 year which reduced its period of operation to only 280 days, which Mr Donaldson described as caused by 60 days’ forced outages. The details of what he described as unforeseeable equipment failures do not matter. They all occurred in the HEP.

In the meantime, THORP completed its commissioning period on the obtaining of formal Consent to Operate from the NII in August 1997, but of course ramp-up continued.

The business year 1998/9 proved to be a disappointment. Mr Smith had predicted a throughput of 800tU, and all the business plans had, in order to allow for a margin, allocated 700tU to the fourth complete operating year. In fact THORP achieved only 461tU, as the combined result of two events which caused the plant to be off-line for over 7 calendar months. During its 140 operating days, the plant achieved a daily output of approximately 3.3tU which, had it been achieved consistently during a 300 day planned operating period would have produced an annual throughput of about 990tU, only just below the predicted normal level.

Again, both the problems occurred in the HEP. The first, and most serious, was a leak in piping which carried coarse fines (ie solid residues) produced in the shearing (ie cutting up) process away to transport containers. This leak occurred due to an unexpected level of erosion in the thickened steel pipes which was itself caused by the transport of fines in superheated steam at very high speeds, rather than in liquid suspension, at much slower speeds, as had been intended. This was itself caused by damage to the steam ejector which was designed to start the fines on the process down the piping. This damage had itself been caused by erosion as the result of the degrading of ceramic plugs used in certain types of spent fuel into a form of highly abrasive powder. That degradation, and the consequential erosion of the steam ejector, had not been anticipated, and although the piping systems had been designed with ultrasonic probes installed at their first bend where there was expected to be the highest risk of erosion, the leaks actually occurred at the second and third bends which had not been monitored.

Recovery from what might have caused a much longer outage due to the need to decontaminate and then enter the shielded area in which the leak occurred, was achieved by the bringing into operation of a third unused transfer route. This was an area where there had been designed duplication.

The second serious breakdown concerned a blockage of another piping system, known as a decanter, and itself led to a 3 month stoppage in that year, and a 1 month stoppage in the following year. The blockage was caused by slender needle-like fragments of fuel casings which had escaped from the dissolver baskets and became entwined like a sort of bird’s nest in the piping. The blockage and the underlying problem was resolved by the obtaining of remotely controlled access to the relevant shielded area, the clearance of the pipe and the installation of a filter between the dissolvers and the decanters, designed to catch such fragments.

Both these breakdowns were described to me as typical examples of the general proposition that all that can be said about unforeseeable risks is that one or more of them may occur. They were precisely the sort of risks for which a general contingency had been built into the throughput predictions and business plans, since they had not been and, I accept, could not have been the subject of stochastic modelling so as to include them as a factor affecting “normal” throughput.

Nonetheless, THORP was still 43tU ahead of the January 1987 (and subsequent) Business Plans by the end of March 1999. It was however 457tU behind Mr Smith’s aggregate prediction by that stage in the 10 year process. It may therefore fairly be said to have used almost all of the contingency cushion which it had built up during its previous years’ good behaviour.

The business year 1999/2000 saw a return to satisfactory performance after the first month’s outage caused by the decanter blockage which I have described. THORP achieved 272 days of normal operation (that is including predictable and therefore modelled unavailability), and achieved an average daily throughput of about 3.2tU, equating to a throughput of about 960tU, a little below its expected normal rate. The annual tonnage was however 879tU, because of the stoppage at the beginning of the year.

By April 2000 therefore, THORP was 122tU ahead of the 1987 Business Plan, but had slightly increased its lag behind Mr Smith’s prediction to 478tU, because its actual throughput fell slightly short of Mr Smith’s prediction of 900tU. That was, for THORP, the end of the planned ramp-up period, and, more importantly, the end of the good times. From then on, Mr Smith’s prediction was that THORP would normally achieve 1,000tU per year, and the business plans had, to allow for a general contingency, proposed 900tU per year. THORP never achieved either of those throughputs in any subsequent year, or anything like them.

The 2000/1 business year started with a 41 day planned engineering outage, and then ran normally for 135 days, subject to a loss of 16 days due to the repetition of a problem in the Waste Encapsulation Plant, part of the ancillary plant, which treated the cut up stainless steel residues of the spent fuel canisters known as sheared hulls.

Much more serious for THORP throughput however were two successive difficulties which occurred within or in connection with the HALES plant, downstream of the main THORP plant, which received its output of HAR.

The design changes to the HALES (which, it will be recalled, pre-existed the THORP plant by many years) involved the adaptation of one of its three evaporators so as to be able to receive Oxide rather than MAGNOX HAR. This was known as evaporator C. The other two evaporators A and B remained un-modified and were therefore capable only of treating MAGNOX HAR. Detailed study and non-radioactive testing had been applied to the modification of evaporator C so as to ensure as far as possible that no difficulties would be encountered once, for the first time, Oxide HAR became available for evaporation into HAL upon the beginning of the commissioning of THORP. Oxide HAR became available in March 1995, for the first time. In particular, the study and testing had addressed the question whether the different chemical constitution of Oxide HAR would cause crystallisation problems within evaporator C, but the assumption that it would not had to be based upon a combination of scientific analysis and testing with the most nearly equivalent non-radioactive substitute, since Oxide HAR was of course not available.

The evaporation of Oxide HAR for the first 5 years after March 1995 suggested that the scientific analysis and testing had been correct, but in September 2000 the disentrainment packing of evaporator C became blocked by crystalline deposits from the Oxide HAR, and evaporator C became unavailable between late September 2000 and the end of January 2001, while remedial measures, consisting of a system of dissolving and flushing out the obstructive crystals were designed and tested. The only protection against an outage at HALES disrupting THORP’s throughput consisted of HAR buffer tanks between THORP and HALES, in Building B212. They provided very little cushion, and THORP was forced off-line in early October 2000. In the meantime, work to connect the HALES plant to a new ventilation treatment plant hit unexpected snags, such that although evaporator C became available again at the end of January 2001, it took a further 3 months before the HALES plant was ready to receive Oxide HAR from THORP.

The result was that, having closed in early October, THORP remained closed for the rest of that business year. It operated only for 123 days, and processed 362tU, at an average rate of just over 2.9tU per day. That was itself a slow rate which would only have produced 882tU in a 300 day normal working period of two 150 day campaigns. Mr Donaldson said that the reason for this was that a significant part of a particular and very small type of used fuel was re-processed at a rate of only slightly over 1tU per day during a 36 day period. But for that, he said (and I accept) a rate of 940-950tU might have been achieved had THORP been in operation for the full planned 300 days that year.

Again, all these cumulative difficulties had to be accommodated within the general contingency, because none of them had been modelled as part of “normal” operation, and THORP had passed the end of its ramp-up period. By this time however there was an insufficient contingency cushion to avoid THORP falling, for the first time, behind the programme reflected in the Business Plans. The aggregate throughput by the end of the 2000/1 business year was 3164tU. The aggregate identified by the Business Plans was 3580tU, and Mr Smith’s prediction based on continuous normal operation without unplanned outage (leaving contingency to the end) was 4280tU. By this time, THORP had been in operation for 6 ¼ years. Even if it had performed at its full normal rate of 1,000tU per annum for the remaining 3 ¾ years, and therefore processed 3750tU, it would still have fallen short of the 7,000tU baseload by 86 tU. In other words, THORP had by the end of its first year of post ramp-up operation not merely exhausted the cushion against contingency which it had previously built up, but exhausted more than the whole of the contingency margin contemplated by an assumed normal operating rate of 1,000tU per year after ramp-up.

The business year 2001/02 was much better than its predecessor, but still only saw a throughput of 736tU, i.e. less than three quarters of the predicted normal rate. This was caused in part by a one month delay in commencement of operations and as the result of a need to comply with directions issued by NII arising out of the decanter blockage which had adversely affected the years 1998/9 and 1999/2000. Thereafter, an average daily throughput of 2.9tU was achieved, which in a 300 day pair of campaigns would have produced 866tU for the year, below both the Business Plan rate of 900tU and the normal rate of 1,000tU. Mr Donaldson attributed this to there having been a high number of campaign changes (that is changes to accommodate the successive reprocessing of different types of fuel). This is something which, it seems to me, might have been modelled had the modelling been directed to the prediction of a 10 year period rather than campaign by campaign throughput, but it was not. That reason for under-performance therefore needed to be accommodated within a general contingency margin.

By April 2002 therefore, THORP’s aggregate throughput to date was 3900tU, falling short of the Business Plan aggregate by 580tU, and short of Mr Smith’s normal aggregate prediction by 1380tU. There was, of course, no reasonable prospect that this shortfall could be made good in the 2¾ years remaining of the 10 year baseload period.

It is necessary to break off this account of the actual performance of THORP because, in subsequent years, THORP’s throughput was adversely, and deliberately, reduced by the need to avoid breach of the Specification imposed by NII in January 2001 for the phased reduction in HAL stocks at Sellafield. It is, necessarily, a matter of speculation as to what THORP might have achieved had that limitation not been imposed. For present purposes I merely note that THORP’s throughputs in the remaining business years before it suffered the outage from which it has yet fully to recover, in 2005, were as follows:

ANALYSIS AND DETERMINATAION OF THE RTR ISSUE

It will be recalled that the defendants put their case under three headings:

Contractual Term

The defendants’ claim was that it was a term of each of the Side Letters that BNFL would reprocess the additional 1,000tU within the first 10 years of THORP’s operation. It is said to be identifiable either as an express or implied term of the Side Letters.

While it is true that paragraph 2 of the Side Letters contained the phrase “…THORP Reprocessing capacity as 7000 tonnes uranium over the first 10 years of operation…”, this appears only by way of recital forming part of the description of the background to the making of an agreement for the reprocessing of the additional 1,000tU. It is a statement of THORP’s capacity rather than a contractual obligation.

Paragraph 3.3 of the Side Letters contains a statement of the present estimate of the unit price to be paid by the BLCs for the additional capacity, calculated on the assumption that the additional 1,000tU would be processed within 10 years. Again however, this is stated as a present estimate and as an approximate figure, and is followed by an express obligation on the part of BNFL to inform customers annually “of its most recent estimate of the price of this additional quantity”.

There is no express statement in the Side Letters of the period within which BNFL is to reprocess the additional 1,000tU. In paragraph 3.1, the entitlement of each BLC to deliver its share of the additional quantity is followed by the phrase “this situation of additional capacity is covered by the Service Agreement between BNFL and the Company and does not require amendments to the Service Agreement”. Paragraph 3.4 requires the BLC to deliver its share of the additional fuel after the last delivery of its share of the initial 6,000 tU, “according to a schedule to be agreed with BNFL….”. Paragraph 3.5 provides simply that “other terms and conditions of the existing Service Agreement remain applicable to the additional quantity.”

There is of course no contractual obligation in the Service Agreements as to the time by which BNFL is to have reprocessed the initial 6,000tU, although the Target Schedule in Appendix 6 contemplated that this would be completed within 10 years. Clause 6.2 of the Service Agreements contained a provision for fee abatement if this was not achieved. The Service Agreements are of course silent as to the time within which any tonnage in excess of 6,000 tU was to be reprocessed, and it does not appear that the parties agreed any revision to the Target Schedule in Appendix 6 so as to accommodate the additional 1,000tU within a ten year programme. As contemplated by paragraph 3.1 of the Side Letters, the Service Agreements did contain provision in relation to additional capacity, at clause 2.3, but this contained nothing by of timetable.

Mr Darling QC for the defendants reminded me, correctly, that every contract must be construed with the assistance of appropriate reference to its factual matrix. In the present case, the relevant factual matrix is in my judgment that:

Mr Darling submitted that since the achievement of a cost advantage for the BLCs was entirely dependent upon the increased baseload being reprocessed within the 10 year target period, the court should construe the Side Letters as including whether expressly by necessary implication, an obligation on BNFL to deliver that benefit by adhering to that target. I disagree. In my judgment there is nothing in the factual matrix to compel the court to identify that contractual obligation on BNFL if no such obligation is either expressed or necessarily to be implied in the Side Letters. A contrary conclusion, namely that there is no such obligation, and therefore a commercial risk that the BLCs might not obtain the unit cost reduction which would arise if the additional baseload were not reprocessed within the 10 year target period, comes nowhere near to amounting to an absurd or irrational result which the parties (including BNFL) could not have intended, such that the court is compelled to strain the words of the contract, or identify a necessarily implied term.

In my judgment it is plain that there is no express term in the Side Letters that BNFL would process the additional 1,000tU in the 10 year period. It is well settled that no term will be implied which would contradict an express term. The terms for the reprocessing of the additional 1,000tU were, to the extent not specifically set out in the Side Letters, to be derived by reference to the Service Agreements: see paragraph 3.5 of the Side Letters. The Service Agreements, which were not amended either for the purposes of dealing with the additional 1,000tU or with the original 6,000tU are in my judgment wholly inconsistent with the existence of any contractual obligation to reprocess the increased 7,000tU baseload within 10 years. For example, such an obligation would make nonsense of the fee abatement provisions of clauses 6.2 and 12 of the Service Agreements. In my judgment no properly informed officious bystander would have been told, in answer to his question “is the additional 1,000tU to be reprocessed within 10 years as a matter of contractual obligation”, “ Oh, of course it is”. Accordingly the contractual term claim of the defendants fails.

Common Mistake

Both Mr Barnes QC and Mr Darling QC invited me to treat as a sufficient summary of the law the following passage in the judgment of Steyn J in Associated Japanese Bank (International) Ltd v Credit du Nord SA [1989] 1WLR 255 at 268, approved (with one immaterial qualification) in Great Peace Shipping Ltd v Tsavlaris Salvage (International) Ltd [2002] EWCA Civ 140 [2003] QB679:

“The first imperative must be that the law ought to uphold rather than destroy apparent contracts. Secondly, the common law rules as to a mistake regarding the quality of the subject matter, like the common law rules regarding commercial frustration, are designed to cope with the impact of unexpected and wholly exceptional circumstances on apparent contracts. Thirdly, such a mistake in order to attract legal consequences must substantially be shared by both parties, and must relate to facts as they existed at the time the contract was made. Fourthly, and this is the point established by Bell v Lever Brothers Ltd [1932] AC161, the mistake must render the subject matter of the contract essentially and radically different from the subject matter which the parties believed to exist…”

The defendants’ case was that at the time of the making of Side Letters, there was a common shared assumption as to an existing state of affairs, namely that:

Those three elements of the proposed common but mistaken shared assumption are broadly the same as the representations alleged to have been made by BNFL as part of the defendants’ misrepresentation case. I shall have to analyse them in much more detail in due course. As will appear, by the time of closing speeches the defendants had chosen to make certain adjustments to those formulations. But even supposing that those representations were made, and that the BLCs reasonably believed them, the case in common mistake is in my judgment unsustainable when the facts are tested against the legal principles set out above. My reasons follow.

First, even though THORP’s failure to reprocess 7,000tU within 10 years may have been unexpected by the parties in 1989 when making the Side Letters, that failure could hardly sensibly be described as a “wholly exceptional circumstance” within the meaning of the second of Steyn J’s principles. Nothing in the Side Letters elevated the expectation as to THORP’s 7,000tU in ten years capacity above the level which had underpinned the expectation of its 6,000tU capacity in contemplation at the time of the making of the Service Agreements. Those agreements manifestly contemplated the risk that THORP would not reprocess the 6,000tU baseload within 10 years, and made detailed provision against that eventuality, all of which remained in place in and after 1989.

Furthermore, in discussions at the 18^th meeting of the TEC in March 1987 there had been a discussion initiated by Mr Schleef of Preussen Elektra as to what would happen if a 7,000tU baseload was not achieved within 10 years, and a reply from Mr Tindle which by no means suggested that this was an inconceivable or wholly exceptional circumstance: see paragraph 247 above.

Secondly, the facts do not permit a conclusion that BNFL laboured under any mistake as to the nature of the design of THORP as at 1989, or as to the process of review which had been applied to it, even though BNFL may have been over-optimistic as to THORP’s capacity (as it turned out). It would in my judgment be a wholly extraordinary case in which common mistake could be established where the subject matter of the mistake lay wholly within the detailed knowledge of one of the parties to the contract. It follows that Steyn J’s third principle is not satisfied.

Thirdly, it does not seem to me that the defendants have begun to establish that, regardless of the extent of the alleged misrepresentation, the subject matter of the Side Letters was essentially and radically different from the subject matter which the parties to those Letters believed to exist. The subject matter of the Side Letters was a reprocessing service for an additional 1,000tU of spent fuel, at the end of an existing, already contracted, service for the reprocessing of an aggregate 6,000tU, on terms which expressly contemplated that the 10 year target for completion might be missed. Steyn J’s fourth principle is therefore also unsatisfied.

It follows that this is a case when the first imperative of the law, namely to uphold rather than to destroy apparent contracts is, subject to the separate claim for rescission, to be applied, with the consequence that the claim in common mistake fails.

Misrepresentation

At the beginning of this judgment I summarised the defendants’ misrepresentation case as being firstly, that in December 1986 BNFL had stated that its expectation that THORP was capable of reprocessing 7,000tU during its first 10 years of operation was the result of “a rigorous technical review”, and secondly, by implication, that the 7,000tU throughput expectation was based on reasonable grounds, and that both those representations had been false and remained false when the Side Letters were made, nearly three years later. I also noted that it was not seriously disputed that those two representations had been made, BNFL’s defence being that they were both true, in 1986 and 1989.

The precise nature and detailed meaning of the innocent sounding phrase “rigorous technical review” has however been a matter of the most intense debate. Further, since the grounds requisite to justify as reasonable an expectation of THORP’s ability to reprocess a 7,000 tonne baseload within 10 years would vary according to the level of confidence at which that expectation was expressed, there has also been intense debate as to the level of confidence expressed or implied by BNFL’s statements as to its throughput expectations for THORP.

The fact that these representations were first made nearly three years before the making of the Side Letters creates additional difficulties. In Hagen v ICI Chemicals and Polymers Ltd [2002] IRLR 31 Elias J said this at paragraph 172:

“When analysing the representations made and determining the potential liability to which they might give rise, it is important to bear in mind certain principles (which were not, as I understand it disputed):

i)

Representations must not be taken in isolation but must be construed in context.

ii)

The question is what is being represented at the time of [contract]…..In so far as statements were made at an earlier date but were later corrected they will have become spent.

iii)

However, in circumstances where a representation is being corrected or revoked, it is important for it to be made in unambiguously clear terms to those to whom the original representation was made that it is no longer accurate or being relied upon: see Arnison v Smith (1889) 41ChD 348 (CA).

iv)

In determining whether a representation is true or not, it is necessary to focus on the substance of the complaint and to ask whether the representation is substantially true. In Avon Insurance v Swire Fraser [2000] 1ALL ER Comm 573 at 579 Rix J said this:

“… a representation may be true without being entirely correct, provided it is substantially correct and the difference between what is represented and what is actually correct would not have been likely to induce a reasonable person in the position of the claimants to enter into the contracts. ””

A particular issue within Elias J’s point (iv) arises in the present case, because when the rigorous technical review representation was repeated by way of recital in paragraph 2 of the Side Letters, it was repeated as a statement as to what had been originally reported to BLCs in December 1986. Thus, assuming that there was no rigorous technical review by December 1986, but that there was one by 1989, the representation repeated in the Side Letters would still have been untrue, but only in terms of the date by which a rigorous technical review had been completed. The question would then arise whether the misstatement of the date was of sufficient materiality to the BLCs in deciding whether or not to enter into the Side Letters.

During trial, BNFL abandoned a pleaded allegation that rescission was barred by reason of the defendants’ affirmation of the Side Letters, and neither the defendants nor BNFL invite me to consider damages in lieu as an alternative to rescission, if a case of misrepresentation sufficient to justify rescission is made out. Accordingly, I shall address the misrepresentation issues in the following order:

Representations in 1986

There is no doubt what Mr Tindle told the BLCs at the Joint Committee meeting in December 1986. The question is what his statement meant, set in its proper context. This is not a subjective question as to the meaning which he intended to convey, or even as to the meaning which the recipients (including the defendants, who, being absent from the meeting, only saw the minutes), thought he meant. The true meaning of a representation is in my judgment to be ascertained objectively, in much the same way as a written or oral statement forming part of a contract. Adapting the words of Lord Hoffmann in ICS Ltd v West Bromwich Building Society [1998] 1WLR 896, at 912H, the true interpretation of what Mr Tindle said is the ascertainment of the meaning which his statement would convey to a reasonable person having all the background knowledge which would reasonably have been available to the defendants as at the time of the Joint Committee meeting.

The immediate background to the December 1986 meeting is the 17^th meeting of the TEC held on 7^th November 1986 in Hamburg, at which Mr Tindle had told the BLCs of BNFL’s confidence that THORP could reprocess more than 6,000tU in its first 10 years of operation, that BNFL hoped to issue a business plan for 7,000tU in January 1987 for discussion at a further TEC meeting in early 1987, that “careful studies” within BNFL had indicated that THORP was capable of meeting the 7,000tU target, but that checks were currently being undertaken on peripheral items of plant, and that BNFL hoped to make a more positive statement on this subject at the Joint Committee meeting. In response to a question from Mr Schleef, Mr Tindle had said that technical data justifying the higher throughput would be provided for BLCs. He mentioned the stochastic modelling which had been done, and continued:

“As design work progressed, the information used in the model became more detailed and throughputs permitting the reprocessing of 7,000tU in ten years had been demonstrated. In fact, throughputs as high as 1,000t pa were being indicated. It had never been stated that THORP would be scaled down to a capacity of 6,000t over a ten year period. BNFL’s contracts saw the reprocessing of 6,000t in ten years and it was now believed that 7,000t could be achieved over the same period. The slow build up in the reprocess programme now assumed could be seen as indicative of some caution in this respect. ”

At the December meeting of the Joint Committee, Mr Tindle said that:

“A rigorous technical review of the peripheral facilities as well as the THORP Head End and Chemical Plant had now been completed”

The two critical issues about the meaning of Mr Tindle’s statement to the Joint Committee in December 1986 are first, as to the precise meaning of the phrase “rigorous technical review”, and second as to the degree of confidence being expressed as to THORP’s capacity to achieve a 7,000 tU throughput in ten years. As to the first issue, the defendants’ primary case was that the phrase “technical review” was a term of art with a settled international meaning. Professor Von Collani supported that case, by reference to the US Military Standard “MIL/STD/1521A on Technical Review” issued in June 1976 and again, with modifications, in 1985, and by reference to British Standard 5882(1980) in relation to audit, which he described as a type of technical review.

Professor Von Collani’s opinion was that the phrase technical review imported, as a minimum, the following requirements:

The defendants’ alternative case, upon which they mainly relied in closing submissions, was that even if not a term of art, the phrase “rigorous technical review” imported by common consent of all the experts, at least the following minimum characteristics:

Those requirements were specified by Mr McGuigan in paragraph 130 of his report. When these criteria were put to Mr Donaldson in cross examination, he agreed broadly with the first four. As to the third, his view was not that the review should be insensitive to minor variations, but that sensitivity to variations was an aspect of risk and uncertainty calling for study, having regard to the subject matter of the review. As to independence, his opinion was that the review needed to be able to take a dispassionate view, but did not necessarily need to be conducted by a person with no involvement in the subject matter. Where the subject matter called for specialist knowledge, the more realistic criterion might be objectivity rather than independence.

Mr Donaldson did not agree that by definition the phrase rigorous technical review called for the production of its findings in a single written report, nor that the concept necessarily required a single separate exercise dedicated, in the present case, to the issue as to THORP’s capacity to process 7,000 rather than 6,000tU over 10 years. Although the phrase must mean just that in certain circumstances and to certain people, it could in his opinion equally accommodate an on-going process of review, in parallel with a design process.

In this last respect, Mr Donaldson’s opinion reflected BNFL’s case, namely that Mr Tindle’s reference to a rigorous technical review was to the process which had been on-going in relation to the design of THORP since at least 1983, rather than to a separate and distinct process triggered by BNFL’s perception in mid 1986 that the offer of an increased throughput, from 6,000 to 7,000tU over ten years would be of real assistance in the preservation of a cooperative relationship between BNFL and the BLCs.

In my judgment, the phrase rigorous technical review when first used by Mr Tindle at the December 1986 Joint Committee meeting was neither itself a term of art, nor was the phrase technical review a term of art to which he applied the adjective rigorous. Interesting though the debate between the experts was, I do not consider that the precise meaning of the phrase, in the context in which it was used, is significantly assisted by the evidence of experts, whose opinions are necessarily based upon its meaning in other contexts, different places and different times, and in relation to different subject matter.

In considering the meaning which Mr Tindle’s phrase would have conveyed to a reasonable person in the position of the BLCs it would not in my judgment be appropriate to seek to put too precise or finely detailed a meaning to it. At the November 1986 meeting of the TEC, Mr Tindle had described BNFL’s confidence in a 7,000tU throughput as made on the basis of “careful studies”, and as having arisen from the input of more detailed information in the stochastic models, “as design work progressed”. Furthermore, as I have noted, he had promised a technical justification for the increased throughput which was still outstanding as at December 1986, and which the BLCs did not expect to receive until later. In my judgment a reasonable BLC listening to Mr Tindle in December, or reading the minutes of the Joint Committee meeting, would have wanted to wait to receive and study the promised technical justification before forming any precise view as to what Mr Tindle meant, not least because no reliant decision was sought from the BLCs in the December meeting, it having been proposed by Mr Tindle in November that discussion of the 7,000tU Business Plan and any consequential decision making should be adjourned to the next meeting of the TEC early in 1987.

Subject to those caveats I consider that the meaning reasonably to be attributed to Mr Tindle’s statement in December was as follows:

As to the last point in the above list, this is a conclusion fairly to be derived from the use of the word rigorous, the reference to designers and plant operators, and from the general law, encapsulated for example in Brown v Raphael [1958] Ch 636, in circumstances where the maker of a statement or opinion of importance in relation to the subject matter of a proposed contract is in a better position than the recipient to know whether or not it is correct.

As originally presented, the defendants’ case was that Mr Tindle’s statement necessarily implied that BNFL already had a fully worked design for THORP. In closing submissions, Mr Darling invited me to adjust that allegation to an implied representation that the design of THORP was sufficiently worked to enable BNFL to make its 7,000tU throughput prediction with confidence. I accept that narrower implication, but it is for practical purposes of no particular assistance. Some parts of a complex industrial process may need to be comprehensively designed before a review can ascertain their likely throughput. An example well appreciated by BNFL was the Highly Active Cycle in the SEP. By contrast, other parts such as pipe-work, buffer tanks, or even the Highly Active Storage Tanks in the HALES building were based upon existing proven technology, such that the only question to be resolved was how many to build, and of what size. The result is that the implied statement as to the stage of design which had been reached by December 1986 adds nothing of substance to the implication that the throughput prediction was based on reasonable grounds. It would be a matter of judgment in relation to each item of plant whether the stage of its design reached by December 1986 was sufficient to enable the overall throughput prediction to be made.

I turn to the second issue, namely the level of BNFL’s confidence which may be taken to have been implied by Mr Tindle’s statement. Again, this needs to be seen in context. At the November 1986 TEC, Mr Tindle has spoken in terms of confidence that THORP could reprocess more than 6,000tU in 10 years, and said that it was “now believed” that 7,000tU could be achieved over the same period. He stopped short of expressing any greater level of confidence in relation to the 7,000tU, pending the completion of studies of the ancillary plant.

By December, the impediment imposed by the incomplete state of the studies of the ancillary plant had, according to Mr Tindle, been removed. The level of confidence entertained within BNFL by December is to be found in the following part of Mr Tindle’s statement:

“he was able to confirm officially that it had been accepted at the highest level in BNFL, by both the designers and the Sellafield operators, that a throughput of 7,000tU could be achieved in the first ten years.”

The defendants characterised BNFL’s level of confidence by alleging that Mr Tindle was asserting that there was no foreseeable reason why it would not be possible to reprocess 7,000tU in ten years. Again, it seems to me that it is inappropriate to ascribe any very specific level of confidence to Mr Tindle’s December statement, pending the provision to BLCs of BNFL’s outstanding technical justification for its throughput prediction. The discussion which took place at the March 1987 TEC makes an assessment of the level of BNFL’s confidence in this respect much more meaningful. Plainly however, whatever may be the meaning of “could” in other contexts, Mr Tindle was implying that a 7,000tU throughput was more than possible in BNFL’s opinion. He meant at least that it was probable, but attempted no assessment whether by way of percentage or otherwise of the degree of probability.

The 18^th TEC in March 1987

I have described what happened at the March 1987 TEC meeting in some detail, in paragraphs 230 - 257 above. In summary, the BLCs received the promised technical justification for BNFL’s increased throughput prediction in the form of Mr Moulding’s report, and the additional comments which he is minuted as having made when presenting it. In addition, Mr Grimoldby repeated that BNFL had conducted a rigorous technical review prior to the December 1986 Joint Committee meeting.

Mr Moulding’s paper and oral explanation provided the BLCs with considerable information about the nature and extent of the process which had led to BNFL’s revised throughput prediction. Mr Moulding explained for example that the stochastic modelling process had been extensively applied to the HEP and SEP over an extended period of time, and it was a reasonable inference for BLCs to draw from the absence of any reference to stochastic modelling in relation to the ancillary plant, that the review had not involved such modelling, save possibly in relation to the WVP. Furthermore, it is apparent from paragraph 14 of Mr Moulding’s report that at least one part of the ancillary plant was still at a stage of flow-sheet design, and that alterations to the design necessary to accommodate the increased throughput had yet to be made.

I consider it also reasonably apparent from the way in which Mr Moulding dealt with ramp-up and contingency margin that he was not suggesting that either of these two important aspects of a throughput prediction had been the subject of stochastic modelling. His description of the stochastic modelling process made it clear that the essence of the modelling was directed to establishing normal throughput per day or per campaign, rather than throughput over 10 years, and that the 10 year throughput was to be achieved by discounting an assumed annual throughput by reference both to ramp-up and contingency.

I shall have to deal in due course with the defendants’ criticism of the thoroughness and accuracy of Mr Moulding’s presentation, but the present analysis is concerned with the question what impact its presentation, and the March 1987 TEC meeting as a whole, should be taken as having made upon Mr Tindle’s December 1986 statement, in particular in relation to the nature and extent of the rigorous technical review, and as to the level of BNFL’s confidence in the revised throughput. As to the first of those questions, it cannot in my judgment safely have been assumed by the BLCs that the rigorous technical review had been a single, one-off self contained exercise (as they contend) rather than an on-going process of review running in parallel with the design process, (as BNFL contends). By the end of the March 1987 meeting, BNFL had said things which suggested that it might have been one or the other, and it is evident that none of the BLCs troubled to ask which.

On the other hand, nothing in Mr Moulding’s presentation suggested that the review had not been “rigorous” i.e. thorough, detailed and objective. Plainly, substantial parts of the throughput analysis had involved the exercise of broad judgment rather than minute calculation but, as is common ground, the BLCs were entitled to assume that all the judgmental inputs had been within reasonable boundaries, in particular in relation to the period required for ramp up and the amount (whether in terms of time or tonnage) required for general contingency margin.

As to the second issue, Mr Moulding said in terms that 7,000tU “was a throughput which BNFL had confidence that it could achieve within a 10 year period”. Nonetheless it is apparent from the minuted conversation between Mr Schleef and Mr Tindle about what would happen if 7,000tU throughput was not achieved within 10 years that BNFL’s confidence fell well short of certainty, and indeed short of a level sufficient to enable BNFL to offer a contractual obligation to achieve that result, any more than it had done in relation to the originally predicted 6,000tU throughput, in the Service Agreements. Again, none of the BLCs represented at the meeting questioned BNFL as to its level of confidence in its revised throughput prediction and Mr Passig’s statement that Mr Moulding’s report was “reassuring” suggests that the BLCs were not anxious to form a precise view on this question. I do not find it either possible or helpful to put a gloss on Mr Moulding’s statement of “confidence” by ascribing any particular level to it, beyond saying that it plainly fell short of a certainty, and that the risk that it might not be fulfilled was sufficiently large for BNFL to be determined to place that risk on the BLCs, rather than to assume it itself. Plainly however, the confidence being expressed was well in excess of a bare 51% probability.

Did BNFL’s representations change by 1989?

In my judgment, nothing of significance changed in the statements being made by BNFL to the BLCs as to its revised throughput prediction, or as to the basis upon which it had been reached with confidence, during the period following the March 1987 TEC meeting until the making of the Side Letters. On various occasions, BNFL repeated its confidence in its 7,000tU throughput prediction, but there was no further express reference to the rigorous technical review until Mr Tindle’s December 1986 representation was expressly confirmed by way of recital in the Side Letters.

There was however a relevant discussion at the 22nd meeting of the Joint Committee on 27^th June 1989. BNFL was insisting on a deadline for acceptance of its offers of the aggregate 1,000tU by 1^st November 1989, and a representative of one of the BLCs suggested that it would be more appropriate for them to make a decision when the plant was operational and the throughput was established. Mr Tindle replied that “waiting until THORP became operational would add little to the knowledge of the plant’s capability over a ten year period…..”

I do not regard that exchange as showing that BNFL was suggesting that there was no risk that the 7,000 tU throughput might not be achieved in 10 years, but rather that the degree of risk that it might not depended on factors which would be unlikely to emerge at the beginning of the 10 year period. Such factors would of course include the questions whether BNFL’s ramp-up allowance and general contingency margin were sufficient and, since the plant would only be pressed to its full operating capacity at the end of ramp-up, BNFL’s normal throughput prediction of 1,000tU per annum would only be subjected to a practical test after several years’ operation.

To what extent were BNFL’s Representations untrue?

1986/7

For reasons already explained, I consider this question first as at the date of the March 1987 TEC meeting. I address first the question whether a rigorous technical review of both the THORP main plant and the ancillary plant had by then been completed.

I have already concluded, contrary to the defendants’ case, that Mr Tindle’s statement, taken in context and repeated by Mr Grimoldby in March 1987, did not unambiguously state or imply that the review had been in the nature of a single one-off exercise aimed specifically at the question whether a 7,000tU throughput could be predicted, rather than at an on-going review process running in parallel with the development of the design. In case a higher court were to take a different view, I shall briefly address the question whether any such one-off process qualifying as a rigorous technical review had been completed, either by December 1986, or by March 1987.

The only process which had in my judgment been completed was that initiated by the main board of BNFL on 28^th August 1986, and which led to Dr Wilkinson satisfying the Executive on 3^rd December 1986 that it would be possible to reprocess 7,000tU during the first 10 year operation of THORP. I have described the process in detail in paragraphs 171 – 218 above. In summary, it consisted of a personal but undocumented judgmental process undertaken by Dr Wilkinson himself, Dr Jeal’s review of the progress of the stochastic modelling reflected in his Memorandum, the undocumented meeting in Mr Smith’s office between designers and plant operators, and the formation by Mr Smith of his own judgment as to the suggested increased throughput reflected in his Memorandum but, as I have found, orally explained by him to Dr Wilkinson by 3^rd December 1986. For present purposes, I assume that there were also a number of undocumented informal discussions involving at least Dr Wilkinson, Mr Smith and Mr Bennett, the detail of which cannot now be recalled.

In my judgment, that process, although completed, did not amount to a rigorous technical review of the main THORP plant and the ancillary plant. Although it was a review, and in part technical, it was simply not rigorous. My reasons follow. First, as Mr Smith himself readily admitted, the brief appraisal of the question whether the ancillary plant would constitute an obstacle to an increased throughput of 7,000tU, limited so far as I can see to the meeting at his office, and his own personal consideration of the question and report to Dr Wilkinson, came nowhere near a thorough or detailed review, albeit that I do not question Mr Smith’s objectivity.

Secondly, Dr Jeal’s review of the stochastic model was in my judgment, although obviously technical, not rigorous in a sense that the circumstances and time-frame in which he was requested to report permitted only a provisional rather than finally concluded exercise. In particular, his attempt to update the current HEP model by reference to potentially adverse design changes had been compromised by an inadvertent input error, leading him to base his judgment on conclusions to be drawn from the report P275 which was by then almost 3 years old. Further, in relation to the SEP, the existing model of the HAC process was about to be replaced and, as Dr Jeal acknowledged in his Memorandum, his attempt to update the existing model was compromised in terms of statistical reliability by his having time only to run 10 sample campaigns.

By none of this do I imply criticism either of Dr Jeal or Mr Smith. Neither of them had been instructed to carry out, from start to finish, a self-contained rigorous technical review. They had merely been asked to provide opinions from their own respective positions of expertise and experience upon the question whether a 7,000tU throughput in 10 years could be predicted, or to use Mr Bennett’s and Dr Jeal’s private and, to an outsider, rather misleading vocabulary, “guaranteed”.

Thirdly, nothing in Mr Moulding’s presentation to the TEC in March 1987 revealed how informal the process which had begun in mid 1986 and ended in December had been, nor how qualified and provisional had been both Mr Smith’s and Dr Jeal’s conclusions, as expressed in their Memoranda. On the hypothesis (which of course I have not accepted) that Mr Tindle’s reference to a rigorous technical review was to be understood by BLCs as meaning a one-off self-contained exercise, I would have expected the BLCs attending the March 1987 TEC to have been very surprised if they had been told that the process was as informal and provisional as I have described.

I reach the conclusion that the process confined to the second half of 1986 was not a rigorous technical review without difficulty, since it was not BNFL’s case that Mr Tindle’s statement referred to anything other than the continuous process of review which had been going on from at least 1983 in parallel with the development of the design of THORP. I turn therefore to the question whether that longer process itself satisfied Mr Tindle’s description (and Mr Grimoldby’s repetition of it), either in December 1986 or March 1987.

Leaving aside for the moment the question whether the throughput prediction was based on reasonable grounds, in my judgment that longer process could properly be described by December 1986 and a fortiori therefore by March 1987 as a rigorous technical review, but it could not on either of those dates possibly be described as having been completed, for the following reasons. First, the design of the ancillary plant (whether by way of new plant or modifications to existing plant) had not by March 1987 reached the stage where it was capable of being reviewed. It had not been modelled, for the purpose of ascertaining its normal throughput, save possibly in relation to the WVP. All that had happened was that a judgment had been formed that the design of the new or modified ancillary plant sufficient to accommodate a 7,000tU throughput prevented no technical challenges which could not be overcome as easily as a design to accommodate 6,000tU.

Secondly, the process of review of the main THORP plant by stochastic modelling was by no means complete, for reasons already given. The HEP model needed much more thorough updating to respond to design changes than Dr Jeal had carried out, and the HAC model, although available for the first time by December 1986, was still in a far from complete or reliable state, as is reflected in the report P413 which contained the express warning that the data derived from its use should be treated with care since they represented only a stage in the development of the study.

The defendants fundamentally challenged the proposition that, even if Mr Tindle’s December 1986 statement could be construed as referring to an ongoing process of technical review, it was rigorous. Since the incompleteness of a review is in itself something which detracts from its rigour, and since I must in any event address again the question whether there had been a completed rigorous review by 1989, I shall not overburden this long judgment by addressing the defendants’ challenges in detail at this stage. It is sufficient to say that, save for its incompleteness, I reject the defendants’ challenges to the rigour of the technical review which had been ongoing since 1983, for the reasons set out in detail below, in relation to the position as at 1989. In my judgment the ongoing technical review of the plant was both throrough, detailed and objective.

I intend to take the same course in relation to the related question whether the throughput prediction made in December 1986 and repeated in March 1987 was based upon reasonable grounds. There is obviously a large overlap between the question whether a technical review is rigorous and the question whether the conclusions derived from it are based upon reasonable grounds. Nonetheless, a judgment may be unassailable as to its reasonableness even in cases where there are demonstrable short comings in the thoroughness, detail or objectivity of the process on which it is based. Similarly, a thorough, detailed and objective process may nonetheless be followed or vitiated by an unreasonable judgment. In the present case, as will appear, what matters is whether there were reasonable grounds for BNFL’s continued confidence in a 7,000tU throughput prediction by 1989. For present purposes, for reasons which will appear in detail when I consider that question directly, I have concluded that BNFL’s throughput prediction was not based on reasonable grounds in December 1986, or March 1987.

1989

Rigorous Technical Review

The principal change which occurred between March 1987 and the making of the Side Letters in late 1989 was that the process of review by stochastic modelling had largely been completed in the meantime. In summary, by the end of 1988 the ancillary plant had all been modelled, the HEP and HAC models had been further developed to a point of reliability, and all three models (that is HEP, HAC, and the THORP purification, finishing and effluent model) had been run both separately and together, so as to produce data as to the interaction between the various elements of the plant, and to identify their throughput in the aggregate.

The main thrust of the defendants’ case as to the lack of rigour in the process of technical review was focussed upon the position as at December 1986, and predicated on the assumption that Mr Tindle’s statement had identified a one-off separate review conducted in late 1986, rather than an ongoing process running in parallel with the development of the design. To that extent therefore, the challenges were misdirected, for reasons which I have already given. In particular, the defendants’ detailed critique of the inadequacies arising from the incomplete state of the stochastic modelling as at December 1986, and the consequential challenges to the adequacy of Dr Jeal’s attempts to derive a conclusion from incomplete and unreliable data, fall away by 1989.

The defendants’ challenges to the rigour of the review, as at 1989, may be summarised as follows:

A common theme in the defendants’ criticism of the adequacy of BNFL’s stochastic modelling (see grounds 1-5 above) was that the modelling had been carried out for the purpose of assisting the ongoing design, rather than for the purpose of predicting a 10 year throughput, so that the use of the modelling for the latter purpose involved an inherent and fundamental lack of rigour. I shall have to address this criticism in relation to each of grounds 1-5, but it is convenient first to outline BNFL’s general response to this criticism. It began by accepting that the primary purpose of the modelling was indeed to assist in the refinement and improvement of the design. Nonetheless, it was submitted that since throughput is the essential output of the models (whether per campaign or per day) there is nothing wrong in using this output for the separate purpose of a 10 year throughput prediction, provided that any necessary adjustments are made.

No Global Model

Professor Von Collani’s opinion was that no reliable estimate of the throughput of a reprocessing plant can be derived from anything other than a model of the whole plant. BNFL’s response, supported by Professor Robinson, was that although BNFL used the most up to date and sophisticated computer software available at the time, it was then, albeit not now, inadequate for the construction of a useful single model of the entirety of the plant at Sellafield used for the reprocessing of spent Oxide fuel.

The question for me is not whether in an ideal world a global model would have been preferable, but whether by not using one BNFL displayed a lack of rigour by the standards of the time. In my judgment the answer is no. The contemporary documentation provides some evidence that the available computer software was being stretched to its limits by the modelling with BNFL did do. This it will be recalled was the first engineering project which BNFL reviewed by stochastic modelling, and it contributes to an affirmative conclusion about rigour that they used the process at all, and did so with the best available computer software.

A separate question, left open by Professor Robinson, and not pursued in any detail by the defendants, was whether the use of 3 models run together by 1988, rather than 1 single global model produced any distortion or called for any express qualifications in BNFL’s confidence in its throughput prediction. The evidence did not in my judgment suggest that this was so. Of course, the absence of a single model generating throughput information about, say, the THORP main building and HALES plant together gives rise to risks of bottlenecks at the point where THORP discharges into the buffer tanks just upstream of HALES, and in the event a breakdown in evaporator C at HALES caused just such a bottleneck, which the available bufferage was insufficient to prevent. But the requirement for bufferage between the THORP building and HALES was itself studied by reference to the separately modelled throughputs of THORP and HALES, and the breakdown which later brought about the bottleneck was not one which it was unreasonable for BNFL not to have modelled, because it was the result of a crystallised deposit, as to the risk of which there was no reliable input data, and which testing had not revealed. This first ground therefore failed.

No ten year or full year simulation

It is factually correct that BNFL’s stochastic modelling proceeded by way of the repetitive simulation of single half yearly campaigns, rather than by an attempt to construct a yearly simulation, still less a 10 year simulation. With the support, again, of Professor Von Collani the defendants submitted that the failure arose from the design development purpose of the stochastic modelling, and involved a consequential failure to model the maintenance period separating each campaign (to see whether the time allowed was adequate), and a failure to model those aspects of a 10 year process, such as ramp-up and ageing of equipment, necessary for any reliable long term throughput prediction. I shall deal with the ageing and ramp-up issues under separate headings, to the extent that they can be disentangled as separate issues.

Again, BNFL’s defence to this allegation consisted of confession and avoidance. Accepting that the modelling was based on the repeated simulation of single campaigns, BNFL submitted first that the typical campaign was structured so as to take into account the various different types of spent Oxide fuel which would have to be reprocessed from time to time, so that (although a single campaign) it was a deliberately constructed average campaign, covering the entirety of the anticipated baseload. That was, in my judgment, sufficiently demonstrated by the evidence.

Secondly, and supported by Professor Robinson, BNFL admitted but sought to justify not modelling maintenance periods. The essence of modelling, said Professor Robinson, was to capture a prediction of throughput. Since there was no throughput in a maintenance period, it did not need to be modelled. He justified that by reference to what he described as the standard practice not to model weekends and shutdown periods in industrial production lines, for the same reasons. Furthermore, there was no suggestion that, although not modelled, the time required for maintenance was not subjected by BNFL to careful study. The point was simply not explored.

To some extent, questions of single campaign modelling and ramp-up cannot be disentangled. Since BNFL’s view at the time was that ramp-up could not itself usefully be modelled due to the absence of any adequate input data, the obvious shortfalls below the normal annual throughput in ramp-up years could not be input into a reliable 10 year simulation, which was a separate and independent reason for not even attempting a 10 year model.

Whilst I accept Professor Von Collani’s opinion that, in abstract theory, anything can be modelled, I prefer Professor Robinson’s opinion that modelling ceases to be of practical value at the point where reliable input data is unavailable, and I accept as plainly correct the view of BNFL, supported by Professor Robinson, that ramp-up was not a phenomenon as to which reliable input data could be assembled in relation to THORP. The duration of, and throughput discounts to be made during, a ramp-up period were a matter for experienced and informed judgment.

It follows that I reject the substance of the defendants’ second ground of challenge, save that they are in my judgment right to point to the need to have regard to the fact that single campaign modelling is inherently unlikely to deal adequately with the ageing of the plant over 10 years, unless some technique is adopted in relation to the output data from the repeated single campaign simulations which sufficiently takes ageing into account.

Ageing

The defendants’ criticism here was that, by failing to use a Weibull distribution technique, BNFL failed to take account of the increased propensity for breakdown of equipment as it gets older, and the consequential effect on time between failures in the stochastic model. Dr Jeal’s evidence was that, at an early stage, a Weibull distribution was used for this purpose, but that it was abandoned in favour of an exponential distribution applied to input data which, because it was derived from existing plant already in service, already reflected the failure rate of aged rather than brand new plant.

Professors Von Collani and Robinson both agreed that, in an ideal world, it was better to use a Weibull distribution technique in a simulation used (whether or not designed) to predict throughput over 10 years. But there was a sharp, highly technical and lengthy disagreement between them as to the question whether in this particular case the use of a Weibull distribution technique throughout the stochastic modelling process would have made any significant difference to the output data, and therefore to the 10 year throughput prediction. Professor Robinson’s view, under cross examination, was that while it was possible that slight differences in the output data might have resulted from the use of a Weibull distribution technique throughout, they would be unlikely to be large, and it would be impossible to tell whether that would have led to a more pessimistic or optimistic throughput outcome.

I shall not burden this judgment with a detailed description of this technical debate. It is, like many differences between Professors Von Collani and Robinson, one where Professor Von Collani may be said to have had mathematical or academic purity on his side, but Professor Robinson had practical realism on his, together with greater experience in the practical application of stochastic simulation modelling to industrial plant. On balance I conclude that Professor Robinson’s view is to be preferred over that of Professor Von Collani on this issue. The use of a Weibull distribution technique would, in an ideal world, have be a better approach, but the use of an exponential distribution based upon data from already aged equipment was in my judgment a substitute which fails to justify a finding of lack of rigour, and I am not satisfied that any small differences in the outcome would be sufficient to undermine the reasonableness of any throughput judgment based upon the stochastic modelling used by BNFL. It follows that this third ground of challenge also failed.

No modelling of the human factor

This issue throws up another example of the competition between mathematical purity and practical experience. The criticism is that BNFL failed to model stochastically the uncertainties arising from the interaction between man and machine in a complicated industrial plant. Again, BNFL responded by confession and avoidance. The uncertainty inherent in human conduct was not itself modelled, but the input data which BNFL used in its stochastic modelling took account of human behaviour in two respects. First, the reliability data based upon BNFL’s experience with existing similar plant had human behaviour input into it by definition, since the frequency and duration of breakdowns was, to a greater or lesser extent in any particular case, itself the consequence of the operation of the relevant plant by its staff. Secondly, BNFL applied an operator performance index to certain of the systems being modelled, calling for a 30 % efficiency discount for mixed man/machine operations and a 40% discount for purely manual operations.

Again, there was an interesting debate between the two professors about the usefulness of stochastic modelling of human interaction with machines. Professor Von Collani had never carried it out in practice, but considered it valuable in principle, whereas Professor Robinson had attempted to carry it out in practice when modelling plant for the Ford Motor Company, but had found it to be of no practical use.

Again, I reject this challenge as either an indication or an example of a lack of rigour on BNFL’s part, or as something which undermined the reasonableness of the judgment based upon the throughput output from the stochastic modelling. The evidence simply failed to persuade me that the stochastic modelling of human behaviour was during the 1980s something which it would be unreasonable for a major designer or user of sophisticated industrial plant not to use, and the method by which BNFL did input the inherent fallibility of human behaviour into its input data seems to me in any event to have been an adequate and properly thought-out substitute.

No modelling of ramp-up or major outage

I have already explained why, in my judgment, BNFL were correct to submit that ramp-up was not susceptible to useful stochastic modelling, notwithstanding Professor Von Collani’s opinion that it could at least in theory be done. As for major outage, I have also explained, at paragraphs 139-140 above, how the availability of reliable input data limited BNFL to modelling breakdowns giving rise only to outages measured in hours or days, rather than in months or years. I have summarised Mr Donaldson’s evidence to the effect that, in relation to longer outages which could not be modelled, BNFL’s focus was upon designing the plant in such a way as to avoid, or by duplication to cope with, such outages as could be foreseen, and to build in a contingency margin for the irreducible risk of the occurrence of those which could not be foreseen.

Here, the contest was between Professor Von Collani’s academic approach, and the practical realism of Mr Donaldson and Dr Jeal. Again, I prefer the latter. It seems to me, and in this Professor Robinson agreed, that an essential feature of reliable or useful stochastic modelling is to recognise the point at which the unreliability or the element of judgment which requires to be incorporated into input data undermines the reliability of the model’s output or, as Mr Donaldson put it, produces a model which tells you nothing that you do not already know. Accordingly, I reject this challenge as one which goes to the rigour of the technical review. In my judgment, BNFL was right not to seek to model outage for which it had no reliable input data. The question whether BNFL made a sufficient allowance in its general contingency margin for outages longer than those modelled is one to which I shall return.

No thorough judgment in relation to ramp-up or major outage

Taking ramp-up first, it seems to me that Mr Smith’s judgment that allowance should be made for a ramp-up period of 5 years, including within it a commissioning period of 21 months, simply cannot be challenged, for the simple reason that hindsight has shown that it was correct. In this respect, it is rather like a property valuation shortly followed by an open market sale at precisely the price indicated by the valuer.

While it is fair comment that Mr Smith appears to have formed his judgment as it were, within his own head rather than on paper or on a computer, using spread sheets and sensitivity tables, he was on the evidence the person best experienced and qualified within BNFL to make that judgment, and I am not persuaded that the manner in which he made it displayed any relevant lack of rigour. On the contrary, he seems to me to have adopted a more cautious approach than his colleagues, but nonetheless to have persuaded them to his point of view.

The evidence has not however, despite Mr Donaldson’s best endeavours, persuaded me to the same conclusion in relation to major outage, by which I mean all outages for which there was no reliable input data sufficient to enable them to be modelled. I have already described the process by which an initial 4 year period of major outage was reduced to 3 years and then, possibly, 2 years before being replaced by a general contingency within which major outage was to be accommodated, which I have concluded amounted only to one year by the time of the Smith Memorandum in late 1986, before being described as “excess capacity over nominal programme…to allow for recovery from a postulated shutdown of, say, up to 2 years”, by Mr Moulding in his oral presentation to the TEC in March 1987.

It inspires no confidence in the quality of BNFL’s approach to this important issue that it has been so difficult to ascertain, purely as a matter of fact, what the provision for general contingency including major outage was, by March 1987. The surviving documents do not speak with one voice as to this issue. In Mr Smith’s Memorandum he described the concept of 2 years’ outage as having been “dropped”, and in his witness statement he describes this as an uncontroversial statement of fact. For reasons which I have explained in full, both Mr Smith’s and Mr Moulding’s yearly programmes actually assume a general contingency margin of one year at normal capacity, i.e. 1,000tU, but the minutes of the March 1987 TEC shows Mr Moulding contemplating a general contingency margin of up to 2 years, based on an assumption that the normal capacity of 1,000tU itself left a margin as against the throughputs emerging from the model. But BNFL’s real judgment at the time was, as I have held, that a one year margin was sufficient.

While I have no reason to doubt the objectivity with which that judgment was exercised, I consider that, viewed separately, it failed to meet the other 2 requirements for having been rigorous, namely thoroughness and detail. I consider it likely, having made due allowance for the risk that relevant documents have been lost or destroyed, that the judgment was also entirely undocumented, save for the inaccurate and unsatisfactory extent to which it makes a brief appearance in the minutes of Mr Moulding’s oral presentation to the TEC in March 1987. As I have already held, his reference to “up to 2 years” was no more than an off-the-cuff explanation of why THORP could not be expected to achieve 9,000tU in 10 years, rather than a careful or considered expression of BNFL’s then thinking as to the contingency margin, which did not in my judgment extend beyond one year at normal capacity.

Bottlenecks

In my judgment this objection simply fails on the facts, for the reasons set out in paragraph 357 above. Bottlenecks were, albeit not always stochastically modelled, rigorously reviewed.

In conclusion therefore, as at 1989, I consider that BNFL had completed a rigorous technical review of both the main THORP plant and the ancillary plant. The rigour of its throughput prediction was however subject to one reservation, namely in relation to the contingency margin. Subject to the question of materiality, there were therefore two elements of potential untruth in the ‘rigorous technical review’ representation repeated by way of recital in the Side Letters. The first was that a rigorous technical review had been completed by December 1986, rather than by 1989. The second was, in effect, that the rigorous technical review had extended beyond questions of normal capacity and ramp-up to the issue of contingency margin. The question is whether those two elements of untruth are, separately or together, sufficiently material to render the rigorous technical review representation, as at the date when the Side Letters were made, in substance untrue as a whole. To that question I apply the test of Rix J approved in the Hagen v ICI case: whether the difference between what is represented and that which is actually correct is likely to induce a reasonable person in the position of the representee to enter into the contract.

As to the misstatement of the date by when the review had been completed, that was not in my judgment material. The question for the BLCs was whether there appeared to be a sufficient basis behind BNFL’s confidence in a 7,000 tU ten year throughput to make it a prudent business decision to accept their shares of the extra 1,000tU so as to obtain the benefit of a lower unit cost, notwithstanding the risk of increased cost if the throughput estimate was not achieved. In my judgment it would have made no significant difference to a BLC in 1989 if BNFL had simply stated that its confidence was based upon having by then, rather than by December 1986, completed a rigorous technical review of the relevant plant. I could find nothing in the evidence of the defendants’ witnesses to the contrary.

By contrast with ramp up, the benefit of hindsight does not come to the assistance of BNFL in relation to the second aspect of potential untruth, namely that there was some lack of rigour in the formation of the contingency margin. The 1 year margin created by a rigorously modelled throughput of 1,000tU per annum (after allowing for ramp-up) was in fact insufficient, and its insufficiency was the main reason why THORP failed to achieve either the 7,000tU or even the original 6,000tU baseload in ten years.

I have however come on balance to the conclusion that this element of untruth did not deprive the representation that a rigorous technical review of the plant had been completed of its substantial truth by 1989. Put shortly, its element of untruth was not sufficiently material. I have two reasons for this. The first is, as with ramp-up, that BNFL’s representation was that its confidence followed the completion of a rigorous technical review of the plant. BNFL did not represent in terms that every aspect of the process whereby it had arrived at its revised throughput estimate had itself been subjected to rigorous technical review.

Secondly, whereas the ascertainment of a reliable normal throughput was pre-eminently suitable for rigorous technical review, for example by stochastic modelling, judgment as to likely ramp-up period and the amount required for a contingency margin were not. There the primary requirements were for objectivity, experience and skill.

My conclusion that this aspect of untruth does not undermine the substantial truth of the representation that there has been a rigorous technical review of the relevant plant may be tested in this way. Suppose that BNFL had either in March 1987 or in 1989 told the BLCs that, although it had rigorously reviewed the plant for the purpose of determining its normal throughput, it had applied discount referable to contingency margin on the basis of the collective judgment of the most highly qualified, senior and experienced engineers in charge of the design and operation of THORP at Sellafield, would a reasonable person in the position of the BLCs have been less likely to enter into the Side Letters? In my judgment the answer is on balance no, but of course only on the assumption that the contingency margin discount, and therefore BNFL’s confidence in its 7,000tU prediction, were based on reasonable grounds.

It follows that the first limb of the defendants’ misrepresentation case, namely that there had been no rigorous technical review, fails, essentially because the respects in which it was strictly untrue are in my judgment immaterial. In substance, there had been a rigorous technical review of the relevant plant by 1989.

1989 - Reasonable Grounds?

In this respect, the grounds for BNFL’s continuing confidence that the THORP design was capable of reprocessing 7,000tU in 10 years did not significantly change between December 1986 and 1989. All that changed was that a provisional assumption as to normal throughput derived from a stochastic modelling process which was incomplete in 1986 had been confirmed by the completion of the modelling process and associated testing, by 1989, and by the substantial further progress during that period of the design of the ancillary plant. By contrast, nothing at all changed as to the basis for BNFL’s ramp-up and contingency margin judgments.

The question whether BNFL’s confidence as to the increased throughput was based on reasonable grounds by 1989 is, in one sense, a single question, but it is necessary first to address the reasonableness or otherwise of the grounds for each of the three constituent elements upon which that assumption was based. I shall therefore take the assumed normal throughput (post ramp-up) of 1,000tU first.

The defendants’ case that the 1,000tU per annum throughput assumption was not based on reasonable grounds depended upon many, but not all, of the grounds which they advanced, but which I have rejected, for their challenge to the rigour of the technical review. For that reason, Mr Darling did not place any extensive separate emphasis in his closing submissions upon this question, preferring to focus primarily on the rigorous technical review aspect of his case, which I have just rejected. Nonetheless, he made it clear, rightly in my judgment, that a negative conclusion on the defendants’ case of the rigorous technical review did not of itself necessarily lead to the conclusion that BNFL’s judgment was based on reasonable grounds, either in relation to the normal throughput, or to ramp-up and contingency margin.

In my judgment BNFL’s confidence in a normal annual throughput of 1,000tU was based on reasonable grounds, even though the daily rate necessary to achieve that annual throughput was not consistently achieved in fact. Although for the most part the interferences with the achievement of that normal rate were, as Mr Donaldson demonstrated in evidence, caused by matters which were either unforeseeable or incapable of being modelled (or both), there were certain instances such as the disruption caused by running a series of short campaigns for the reprocessing of different types of spent fuel which were neither unforeseeable nor, so far as I can see, incapable of being modelled. Nonetheless, those aspects of THORP’s operational experience were only a minor contributor to its failure to achieve the predicted throughput. Generally, as I have held in setting out the relevant facts about THORP’s actual operation, its normal performance, unconstrained by unforeseeable or lengthy outages, was close to the rate necessary to achieve 1,000tU per annum. Putting it another way, the departure was not so great as to suggest that there was anything fundamentally or unreasonably wrong about the modelling process upon which the normal throughput had been assessed.

Furthermore, the detailed grounds upon which I have rejected each one of the defendant’s specific criticisms of the modelling process involved conclusions that in each respect BNFL’s acts or omissions were not merely consistent with rigour but reasonably based. More generally, it is inherently unlikely that the ascertainment of a normal throughput arrived at by a rigorous technical review of all the relevant plant will be based on unreasonable grounds. The rigorous technical review which I have held BNFL carried out in fact constituted a reasonable basis for its view as to normal throughput, provided that it is understood that by “normal” a confident prediction was only that 1,000tU could be achieved post ramp-up, rather than the additional 90 tonnes per annum which represented the average of modelling results referred to by Mr Moulding in March 1987. Such a higher throughput was possible in one or more years, and even possible as an average, but it could not in my judgment reasonably have been predicted with any confidence, as Mr Donaldson implicitly acknowledged in the passage in his cross examination which I have quoted at paragraph 251 above.

There is in my judgment no need to probe as to the reasonableness of the grounds for BNFL’s conclusion as to the appropriate ramp-up period because, as I have held, it was correct. The analogy of a correct but carelessly prepared valuation or of a correct prediction based on two careless but self-cancelling mathematical errors is apt. But if it matters, I consider that BNFL’s approach to ramp-up was based upon reasonable grounds, encapsulated in the cautious, prudent attitude of Mr Smith to that question which, as I have held, prevailed over the more optimistic assumptions of some of his colleagues, including Dr Jeal. The defendants offered no persuasive evidence or submissions to the contrary.

The final question therefore is whether BNFL’s attitude to the contingency margin was based on reasonable grounds. In this respect it is necessary to compare the expert evidence of Mr Donaldson and Mr McGuigan. I have already shown how Mr Donaldson was unable to explain Mr Moulding’s reference to “say, up to 2 years” as the contingency margin, or to identify by reference to the materials and evidence available to him, what BNFL’s attitude was to contingency margin as the relevant times, in terms of amount.

Mr Donaldson’s expert opinion was that it is impossible for engineers to identify a specific period or amount of contingency margin required as part of the basis for a confident 10 year throughput estimate, essentially because the margin served to provide a cushion against the risk of unforeseeable events, which could by definition not be subjected to quantitative analysis. In consequence, his opinion was that the absence of any process within BNFL by which the required margin was quantified, either by 1987 or 1989, was not unreasonable. A broad objective judgment, based upon skill and experience simply had to be made as to whether the difference between the normal throughput discounted by ramp-up and the required 7,000tU 10 year throughput left an adequate margin for unforeseeable disruptions of all kinds. Since the notionally correct or reasonable margin could not be quantified, it followed in Mr Donaldson’s opinion, and Mr Barnes’ submission, that the margin actually identifiable (which I have held to have been one rather than two years at post ramp-up capacity) cannot be categorised as unreasonable, not least because it was regarded as reasonable by a team of senior engineers of the highest skill and relevant experience.

Mr McGuigan’s opinion was that, as at 1983, the then state of the design of THORP, and the uncertainties implicit in it, called for a reasonable contingency margin of between 2 and 4 years. Although he did not volunteer a quantified opinion as at 1986, he answered my question to that effect by saying that, in his opinion, the minimum contingency margin which it was reasonable to accept at that stage of the development of the design was 2 years. Mr McGuigan did not offer any detailed justification for those opinions, but in fairness to him, it does not appear that he was asked to when preparing his report, and he was not in fact asked to do so while giving oral evidence.

I have already referred to the fact that BNFL itself began with a contingency margin of 4 years, reduced it to 3 years in 1983 and continued to reduce it thereafter, and to the fact that, in March 1987, Mr Moulding told the BLCs that BNFL’s calculations left room for a major outage of “say, up to 2 years”. Although a careful forensic analysis of Mr Moulding’s March 1987 presentation might have led a sceptical BLC to wonder where Mr Moulding got his 2 years margin from, it would not in my judgment be fair to impose upon the BLCs a duty, in the protection of their own interests, to conduct any such analysis, or to second-guess what they were being told by BNFL’s appointed representative on that very question. It is therefore a relevant consideration that the BLCs were in 1987 being told that BNFL’s approach to its confident throughput estimate involved a contingency margin of up to 2 years, and that this was both incorrect, as a matter of analysis of the available material, (as I have held), and was never corrected. Nonetheless, the fact that Mr Moulding may have got the quantification of the contingency margin wrong can not of itself conclude the question whether a one year margin ( at post ramp-up capacity) was itself unreasonably low.

Looking at the matter in the round, I am therefore faced, on the one hand, with a judgment arrived at by experienced engineers of undoubted skill and competence, supported by a most impressive expert, albeit not of independent status, that a one year margin was reasonable and sufficient, and, on the other hand, with the opinion of a less impressive but wholly independent expert that it was too short, coupled with BNFL’s own incorrect statement to BLCs that allocated a greater margin than was the reality, and the undoubted fact that, with the benefit of hindsight, the margin actually allowed proved to be wholly insufficient to accommodate the interruptions and difficulties with which it had to cope. Of course, the benefit of hindsight is of limited relevance to the reasonableness of a judgment as to future performance formed at the outset. If actual performance justified the judgment, that either obviates or concludes the analysis of reasonableness. It is only if actual performance fails to justify the judgment that the need to address the reasonableness of the grounds upon which it is based arises.

I have, after long and anxious thought, come to the conclusion on balance that BNFL’s judgment as to the adequacy of what was in truth only a 1 year contingency margin was not based on reasonable grounds. My reasons follow. First, it is necessary to focus more precisely than either BNFL or the experts did upon the type of interruption for which the contingency margin represented BNFL’s only protection against a failure to achieve the revised 10 year throughput. A single major outage (whether of 1 year, 2 years or more) was by no means the only such event. In truth, the contingency margin had to accommodate everything which had not been modelled, other than ramp-up. As I have held, that included all types of outage (whether foreseeable or not), in respect of which there existed during the 1980s no sufficient input data based upon BNFL’s experience or the availability of data from other operators, such as COGEMA. In practice, this included outages of more than about a week, since these were the longest breakdowns for which adequate input data was available. It therefore included outages caused for example by longer than expected maintenance periods, outages caused by the need for a safety enquiry or inspection following a breakdown and extending well beyond the relevant repair period, outages caused by longer than expected shut downs in down-stream ancillary plant while necessary modifications were made and tested, and outages caused (for example at evaporator C) by the occurrence of a defect which, although it had been suspected as a possibility, could not be discounted as a risk by testing with the use of non-radioactive chemical near-equivalents.

Although the evidence persuaded me that BNFL had carefully considered and concluded that improved plant design meant that it was no longer necessary to include within a contingency the whole of a serious major outage of the type likely to necessitate a 2 year or longer shut down, the evidence does not show that BNFL addressed with any or any sufficient care the question whether an aggregation of much shorter outages of types which, although not suitable for modelling, could not be said to be unforeseeable as a class, necessitated the allocation of a significant contingency margin of their own.

Furthermore, the immediate background to the conclusion that design improvements had reduced the risk of a single serious outage of a 2 or more year type to a level of risk which could be excluded from the 10 year throughput calculation was that the only 2 reprocessing plants of which BNFL had any experience had both suffered serious malfunctions of a type which, were it to happen at THORP, would be likely to cause a long period of shut down. The first of those plants was the previous Oxide reprocessing plant which had been permanently shut down as the result of a major incident. The second was the MAGNOX plant at building B205, at which a major breakdown had caused a long term closure of one of its parallel dissolver units, where the combination of the time necessary for decontamination, and the time necessary for repair, amounted to 3 years in aggregate. I infer that the events which caused those two breakdowns were no more foreseeable than the event which, for example, caused a more than 2 year breakdown at THORP in 2005. THORP was designed with some, but less, built-in duplication than the MAGNOX plant, so that although the event at the MAGNOX south side dissolver did not in fact cause a 3 year shut down of the MAGNOX reprocessing plant, there was greater risk that the occurrence of a similar event at THORP would occur in connection with plant that was not duplicated, than the equivalent risk at MAGNOX. It follows that although design improvements may properly have led to the conclusion that the risk of such an event was lower at THORP than at either of the two older plants, the risk that such an event if it occurred would cause a major outage was greater at THORP than at the MAGNOX plant.

Nor is it correct to assume that the existence of duplicate plant (whether at THORP or in any other plant operating in a radioactive environment) of itself means that a failure in something which is duplicated will not cause a significant outage, well beyond the type susceptible for modelling, even if less than one or more years. For example, the piping carrying coarse fines which developed a serious leak in the 1998/9 year, and the piping in which needle-like fragments of fuel casing caused a birds-nest type blockage, was all duplicated, but the outages necessary to examine the breakdowns and bring into operation parallel duplicated systems bit very substantially into the available contingency margin, when they occurred: see paragraphs 274-278 above.

Furthermore, there were still both within the main THORP plant and within the ancillary plant, important items which were not duplicated for the purpose of Oxide reprocessing, such as evaporator C at HALES, in respect of which a serious breakdown (whether due to a foreseeable or unforeseeable risk) threatened a significant outage at THORP in terms of months rather than weeks, for which bufferage was a wholly inadequate protection. In referring to these three significant outages which occurred during the operation of THORP, I have availed myself of the benefit of hindsight only for the purpose of referring to examples, rather than because I regard the specific breakdowns which occurred as ones which ought to have been foreseen. The point is that they are examples of a class of breakdown likely to cause outages in terms of months rather than weeks which are foreseeable as a class, even if the individual members of that class cannot be predicted. As was ruefully acknowledged in evidence, the only thing that can be said about unforeseeable events as a class, is that one or more of them will probably occur, given enough time.

The next point is that a 1 year contingency margin allows only for 300 odd days of lost production. The evidence did not suggest that any outages could be made good during maintenance periods, which occupied the rest of any relevant year. It follows that a judgment that a one year contingency margin would be sufficient involved the conclusion that there was no significant risk that the aggregate of all non–modelled types of breakdown would cause closures amounting in aggregate to more than 10 months in 10 years. While such a margin in a normal complex industrial plant, to the whole of which investigation and repair staff have ready access, may appear reasonable and appropriate, the same cannot in my judgment be said of a nuclear plant, still less of a state of the art integrated reprocessing plant for which there existed no near or precise precedent of which BNFL had, or could obtain, experience or detailed data.

Mr Barnes attempted to reduce the implications of this calculation by submitting that outages during the commissioning or ramp-up period would be more easily redressed than outages at a time when the plant is expected to be in unconstrained normal operation, because a lower tonnage would be lost per week or month during the ramp-up period. In my judgment that initially attractive submission was flawed. For example, a 6 month outage during the ramp-up period would, (absent special factors) simply delay the conclusion of the ramp-up by the same period and therefore reduce the period during which, within its first 10 years, THORP could be expected to reprocess at full normal capacity.

There is in my judgment an equivalence between Mr McGuigan’s minimum 2 year contingency margin as at 1986 and the “up to 2 year” contingency margin suggested as incorporated by Mr Moulding in March 1987 which goes beyond mere coincidence. The BLCs were all sophisticated operators of nuclear plant, for whom the concept of a contingency margin must have been as real as it was (or needed to be) at THORP. I consider that the apparent consensus at the March 1987 TEC meeting as between BNFL and the BLCs as to the wisdom of including an up to 2 year contingency margin significantly supports Mr McGuigan’s opinion, even if the fair interpretation of the minutes of that meeting suggest that those attending were looking more at single major outages than the aggregate of a series of intermediate outages.

I have also been significantly influenced in reaching my conclusion that the one year contingency (which was all that was available between the normal annual throughput in which BNFL had confidence and the required 7,000tU throughput over 10 years) was not adjudged sufficient on reasonable grounds, by the absence of any clear and consistent documentary or other evidence as to how that judgment was reached, the absence of any contemporaneous documentation upon which it was recorded, and the fact that it was actually misstated by Mr Moulding to BLCs at the relevant time. Put another way, the apparent lack of rigour with which this important component in the 10 year throughput prediction was addressed goes a considerable distance towards the conclusion that it was not based on reasonable grounds, even though it is not on its own conclusive proof of it.

I was much pressed by Mr Barnes with the submission that final responsibility for this judgment rested with Dr Wilkinson, a man of the highest eminence, skill and experience in the field. While I acknowledge Mr Barnes’ description of Dr Wilkinson as accurate, it seems to me, as I have already held, that his approach to the question of THORP’s capacity to achieve a 7,000tU throughput in 10 years involved a dangerously high level of generality, at least at the outset . His original statement to the main board on 28^th August 1986 that a 7,000tU throughput was judged to be realistic by BNFL’s designers and plant operators was not merely given in advance of any formal consultation with designers and plant operators of the type which later occurred, but was based upon his view that a 6,000tU baseload assumed only 50% utilisation of the plant, a conclusion which I have described as having involved the highest conceivable level of generality. It involved no separation out of the different discounts appropriate for ramp-up and contingency, and no considered discount from the plant’s design capacity to its normal annual capacity. Making due allowance for Dr Wilkinson’s understandable loss of detailed recollection over a substantial period of years, he did not in his evidence provide any explanation of the basis upon which the development of his views between then and the end of 1986 involved any specific or detailed consideration of the minimum contingency margin sufficient to justify a confident assertion of a 7,000tU 10 year throughput.

It follows therefore that in my judgment, BNFL’s expressed confidence, by 1989, that THORP could achieve the 7,000tU throughput in 10 years was based on reasonable grounds as to the first and second of the three constituent elements but are not in relation to the third. The question therefore arises whether the absence of reasonable grounds for the perceived adequacy of a one year contingency margin is of sufficient materiality to lead to the conclusion that, in this second respect, the representation made and persisted in until the making of the Side Letters was materially false. In my judgment it was of sufficient materiality.

By contrast with the representation as to the rigorous technical review, which applied to the plant rather than to the formulation of the revised throughput estimate, the implication that a confident belief as to THORP’s 7,000tU in 10 years throughput was based upon reasonable grounds plainly extended to the whole process which underlay that estimate. The estimate was unqualified by any specific reservations or health warnings, other than those implicit in Mr Moulding’s presentation in March 1987, and in BNFL’s reluctance to undertake a contractual commitment to achieve its estimate. The throughput was estimated with confidence, rather than as a possibility or bare probability, and although BLCs may be taken to have been experienced in the nuclear industry, BNFL plainly had the critical experience both of reprocessing and of the design of THORP which the BLCs lacked, for the purpose of formulating any throughput estimate. The consequence is that a lack of reasonable grounds for the formulating of one of the three important elements of the throughput estimate materially detracts from the truthfulness of the implied representation that the estimate was, as a whole, based on reasonable grounds.

There is in my view no inconsistency with the contrary conclusion which I have reached in relation to materiality of the lack of rigour applied to the formulation of the general contingency margin. As a matter depending primarily on judgment, experience and skill, I consider that the existence of reasonable grounds for a throughput estimate is far more central to the validity of the judgment than is the deployment of rigour to an estimate based on, but not forming part of, a rigorous technical review of the plant.

Accordingly, and solely upon the basis that the judgment that a one year general contingency margin was adequate for the confident 7,000tU throughput estimate was not based upon reasonable grounds, I have concluded that there was a material misrepresentation both in 1986 and in 1989, which continued until the making of the Side Letters.

Reliance

Whereas BNFL challenged the allegation that the defendants relied upon the statement that there had been a rigorous technical review, it was not seriously disputed that the BLCs relied upon the implied representation that there were reasonable grounds for BNFL’s confident throughput estimate. The challenge as to reliance was primarily based upon the fact that the defendants were not represented in the December 1986 Joint Committee Meeting, although it is accepted that their representatives read the minutes of it.

That challenge is of no account in my judgment in relation to the implied representation that the assertion of confidence in the 7,000tU throughput estimate was based upon reasonable grounds. The representation was made at several meetings, as I have described, and repeated in writing by way of recital to the Side Letters, in unmistakable terms.

Accordingly all the necessary constituent elements of a claim for a rescission for misrepresentation have been made out by the defendants. It was not submitted for BNFL that, if there was a material misrepresentation, rescission was either unavailable or unjust in the circumstance. Accordingly, in my judgment, the defendants are entitled to, and have, rescinded the Side Letters.

Consequences

Since no part of the defendants’ share of the additional 1,000tU added to the baseload by the Side Letters has yet even been delivered for reprocessing, the consequence that, in the event of rescission, there must be a restitutio in integrum has no implications in terms of redelivery of fuel assemblies, uranium, plutonium or glass blocks. It is common ground however that recission gives rise to a requirement for financial restitution, because at least some of the invoices delivered to and paid by the defendants have included elements which, or the amount of which, are attributable to work done or expenses and overheads incurred in or about the preparation for reprocessing of the additional 1,000tU. It is agreed that the extent of any such financial restitution is not a matter which, on the evidence and submissions so far presented, I am able to form any reliable judgment. I will hear submissions as to the directions necessary and appropriate to enable that question to be determined, if it cannot be agreed.

MAGNOX PRIORITISATION

The Evidence

Documents

Focussing as it does on the operation rather than the design and modelling of THORP and its ancillary plant, this issue called for reference to more recent documents than those relevant to the RTR issue. As a result, the fact-finding process is not significantly affected by gaps in the documentation caused by the passage of time. For the same reason, there were less significant factual disputes between the parties. The argument raged over what should have been done, rather than what was or was not done.

Generally, the documents consisted of records of the performance and monitoring of THORP and its ancillary plant, records including correspondence, attendance notes and internal memoranda concerning BNFL’s negotiations with NII in 2000, and internal BNFL documents recording its developing thinking about the shortcomings in the WVP, and the improvements which might be made to it. Minutes and other records evidencing discussion about the matter between BNFL and BLCs were available, but of limited relevance, because BNFL does not advance a case that any prima facie breach of the Service Agreements was waived by the BLCs’ consent.

Witnesses

For the same reason, it was unnecessary for the parties to call purely factual witnesses on this issue. A little of the relevant factual detail was provided by Mr Donaldson on behalf of BNFL, but most of it was provided by another BNFL employee, Dr Paul Robson who, like Mr Donaldson, gave both factual and expert evidence.

Dr Robson was first employed by BNFL in September 1979 as an Assistant Scientific Officer. With BNFL sponsorship he obtained a degree in chemistry at the Huddersfield Polytechnic in June 1983. In his own time but using BNFL facilities he obtained a Ph.D in the Chemistry of Highly Active Liquid Waste at Bradford University in 1987, and an MBA in Business Administration from the Open University, again in his own time, in 1992. He led a research team examining waste treatment processes for BNFL between 1983 and 1986 and was Commissioning Area Manager for the WVP between 1986 and 1989. In 1993 he became, for a year, the Vitrification Technical Support Manager, and in 1994 the Plant Manager at HALES. For 3 years from November 1999 he was the Senior Technical Manager for Waste Management at Sellafield and has been the Head of Manufacturing at HALES from 2001 until now.

Mr Robson’s career has therefore consisted entirely of working for BNFL, and has focussed upon the treatment, storage and vitrification of Highly Active Waste. He has experience both in relevant analytical research, plant commissioning and operation, and in particular was part of the BNFL negotiating team charged with responding to the NII February 2000 Report, having been involved both as author and recipient of a significant part of the relevant internal correspondence.

That experience naturally qualified Dr Robson both as a witness of fact and as an expert witness, although of course, like Mr Donaldson, his full time employment by BNFL which continues to this day makes it necessary for me to give the most anxious consideration to his independence.

Dr Robson demonstrated during his cross examination that he had applied a high degree of careful preparation to the task of acting as a factual and expert witness and, like Mr Donaldson, he was able to supplement from preparation or original recollection a significant amount of additional detail beyond that included in his 3 witness statements, when called upon to do so in cross examination.

It seemed to me that, initially, Dr Robson demonstrated a certain defensiveness in his answers to cross examination, suggestive at least potentially of an agenda to do nothing which might damage BNFL’s interests, if that could be avoided. It manifested itself in what I can best describe as a slight tendency to play a straight bat to a difficult ball, rather than to engage fully with the question. Nonetheless, this early tendency soon disappeared and for most of his cross examination Dr Robson gave straightforward and forthright answers, making short and unvarnished concessions where appropriate, in a manner which suggested to me that his perception of his duties as an expert overcame his initial defensiveness as an employee.

In the result therefore, although the range and depth of his factual knowledge was perhaps slightly less than that of Mr Donaldson, he was nonetheless a very impressive and reliable witness of fact. Similarly, although there were moments when his slight defensiveness caused me to wonder about his independence, his expert evidence taken as a whole was, in my judgment, independently given in accordance with his duty to the court, and consequently of real value notwithstanding his status as the claimant’s employee throughout.

The defendants relied upon Mr McGuigan as their expert witness, just as they had done in relation to the RTR issue. I have already described his quality, at paragraphs 79 - 83 above. The defendants called no witnesses of fact in relation to the MAGNOX prioritisation issue.

The Facts

Technical Introduction

I have already briefly described in the introduction to this judgment the bare essentials of the process whereby Highly Active Raffinate emanating from both the MAGNOX plant and THORP are evaporated, stored and then vitrified, so as to incorporate what is in its raw form a highly radioactive and dangerous liquid into an inert, stable, still dangerous but much safer form of glass blocks encased in stainless steel canisters, so as to make the radioactive waste suitable (or as some would say less unsuitable) for long term permanent storage. It is now necessary to describe in a little more detail the different chemical constitution and properties of MAGNOX and Oxide HAR and HAL, so that BNFL’s approach to the NII 2000 Report and its decision making in relation to the Specification can be examined in its proper context.

I have already mentioned in my technical introduction to the RTR issue that a fundamental difference between MAGNOX and Oxide spent fuel assemblies is that the latter is more highly irradiated than the former, and that the stainless steel construction of their containers make it suitable for long term storage underwater prior to reprocessing, whereas MAGNOX spent fuel rods are liable to suffer serious erosion whilst stored under water, such that they cannot safely be stored under water for more than limited periods. For completeness I should add that (as emerged during cross examination) attempts have been made to provide dry storage for spent MAGNOX fuel rods so as to avoid that corrosion problem. Dr Robson told me (and I accept) that BNFL constructed a dry storage plant for spent MAGNOX fuel rods at its Wylfa MAGNOX power station, but had to discontinue use of it in 2002 after a security review led to an instruction to that effect from the Office of Civil Nuclear Security.

Two consequences flow from those two differences between spent MAGNOX and Oxide fuel. The first is that the more highly irradiated state of Oxide fuel means that a volume of approximately 4 times more HAL is produced from Oxide fuel than from MAGNOX fuel as a result of its reprocessing. Dr Robson’s evidence was that 1tU of spent Oxide fuel results in 160-220 litres of Oxide HAL, whereas 1tU of MAGNOX spent fuel produces on average 50 litres of MAGNOX HAL. There are detailed reasons for this which do not matter, but the effect of them is to limit the extent to which Oxide HAR can be concentrated by evaporation, by comparison with MAGNOX HAR. The consequence is that for 1tU of spent fuel, Oxide HAL takes up approximately four times as much storage space at HALES as MAGNOX HAL. The range of between 160 and 220 litres of Oxide HAL per tU of spent fuel largely derives from the various different types of spent fuel submitted for reprocessing at THORP, by comparison with the broadly uniform type of MAGNOX spent fuel derived from the UK’s MAGNOX power stations. Dr Robson told me that for practical purposes BNFL worked upon a ratio of MAGNOX HAL: 50 litres and Oxide HAL: 200 litres per tU of spent fuel.

The second difference in terms of HAL characteristics is that Oxide HAL is in terms of radioactivity “older” than MAGNOX HAL, because the spent fuel rods from which it is derived will have been in storage for approximately 5 years, whereas the corrosion problems associated with MAGNOX spent fuel will mean that it has been reprocessed within 6 months to 1 year from being withdrawn from active use in the power station. The result of this is that the heat-generation properties of the faster decaying isotopes within waste emanating from the irradiated fuel assemblies will have significantly declined during the longer period of storage of Oxide spent fuel so that, when coupled with the lower level of concentration of Oxide HAL by evaporation, and the initially higher heat generating properties of Oxide as opposed to MAGNOX spent fuel, the time which, if cooling systems were to fail at HALES, would be taken before Oxide HAL boiled and released harmful radiation into the atmosphere is not greatly different than the time taken if MAGNOX HAL boiled in the same circumstances. This is therefore an aspect in which the different treatment of spent fuel in terms of storage of fuel assemblies and evaporation of HAL largely but not completely cancel out their very different heat generation properties.

Nonetheless, other significant differences between MAGNOX and Oxide HAL remain. The first is that, unless blended, the two types of HAL are subject to different vitrification incorporation ratios. Again, the reasons do not matter, but a glass block used for the vitrification of pure Oxide HAL will have only 20% calcined HAL in it, whereas a glass block used for the vitrification of pure MAGNOX HAL will have 25% of calcined HAL in it. By contrast, if Oxide HAL is blended with MAGNOX HAL, it only takes a blending ratio of 75% Oxide to 25% MAGNOX to enable the vitrification incorporation ratio to be raised to the same 25% level as for pure MAGNOX HAL. In terms of numbers of containers produced by the vitrification process, the benefit of blending during vitrification of the whole of THORP’s 7,000tU baseload may be summarised as a saving of 600 containers, and the beneficiaries were the BLCs whose Oxide HAL was to be returned to them after vitrification at BNFL’s option pursuant to the Service Agreements, since no incorporation ratio benefits were associated with the use of blending in the vitrification of MAGNOX HAL.

The second and for present purposes most important difference between MAGNOX and Oxide HAL lies in its potential for causing harm if allowed to boil, with the consequent escape of radioactive vapour into the atmosphere. After considerable debate between Dr Robson and Mr McGuigan, it became reasonably clear to me that nothing in the process of storing spent Oxide fuel rods for 5 years, and evaporating it to lower levels of concentration than those used for MAGNOX HAL significantly detracted from the presence in Oxide HAL of substantially higher levels of the dangerous long half-life isotopes of Plutonium, Uranium, Americium and Curium in storage in Oxide HAL than in MAGNOX HAL, as an enduring consequence of the higher degree of irradiation of Oxide fuel. These are isotopes which, even in very small quantities, can cause serious damage to humans and other living organisms if in differing cases, touched, ingested or inhaled. In this respect, Dr Robson gave graphic examples during cross examination, which Mr McGuigan derided as unscientific and simplistic, but the details of their disagreement do not matter.

What does matter is that, for good reasons or bad, NII clearly regarded Oxide HAL as presenting a substantially greater risk to health and safety than MAGNOX HAL, when considering how to regulate the phased reduction in storage levels of each type at HALES prior to 2015. Bearing in mind that for the reasons already summarised, it is common ground that the risk of loss of containment (mainly caused by the risk of boiling in the event of a failure in the cooling system) of the two types of HAL were broadly equivalent, I accept Dr Robson’s opinion that the main reason for NII’s greater concerns about Oxide HAL than about MAGNOX HAL must lie in the proposition that, once uncontained, the former is substantially more dangerous than the latter. For present purposes, the question is not whether NII were correct, but whether there were objective reasons behind NII’s greater concerns in relation to the health and safety implications of the storage of Oxide HAL by comparison with MAGNOX HAL. In my judgment there were.

Relevant History Prior to the Publication of the NII’s 2000 Report

The defendants’ case is, in summary, that acts or omissions of BNFL following the publication of the 2000 Report constituted breaches of the Service Agreement, and that they consisted first of the failure to take steps during the negotiations leading to the Specification to obtain a less stringent regime for the reduction, in particular, of Oxide HAL stocks, and secondly, acts and omissions which led to the deliberate restriction of THORP’s throughput in and following the business years 2002/3.

The defendants sought by amendment to introduce a wider case of breach by acts and omissions prior to the publication of the NII 2000 Report, but this was refused by Master Moncaster on 12^th December 2006. It follows that the court is not required to investigate whether acts or omissions of BNFL prior to the publication of the NII 2000 Report constitute breaches of the Service Agreement. Nonetheless, a brief historical account of the progress of the storage and vitrification of HAL forms a necessary part of the background to the consideration of the allegations of breach which remain.

Taking storage first, I have already described how BNFL adapted evaporator C at HALES so that it could accept Oxide HAL, and how, after 5 years from commencing operation in March 1995, that evaporator failed due to unexpected crystallisation, and how its operations and procedures had to be altered thereafter. I turn to describe the history of vitrification, down to 2000.

Prior to 1991, all that BNFL did, or had ever done, with HAL was to store it in HALES, in an ever increasing number of tanks known as HASTs. Following COGEMA’s example, BNFL designed and built a vitrification plant (“WVP”) in Building B355 at Sellafield during the 1980s. Although of course no Oxide HAL had yet been produced, it was intended from the outset that the WVP should be capable of vitrifying Oxide HAL, MAGNOX HAL or blended HAL. Two parallel vitrification lines (numbered 1 and 2) were designed and built. In doing so, BNFL made use to a very large extent of detailed technological design already carried out by COGEMA for its own, earlier, vitrification plant, but did not slavishly copy the COGEMA plant in every respect. Substantial parts of the WVP were not merely based on COGEMA designs, but actually supplied by COGEMA.

The two plants were not by any means identical. For example, the COGEMA plant had 6 lines, 3 of which were dedicated to processing spent fuel from French power stations, and 3 for the reprocessing of spent fuel from customers outside France. Secondly, whereas COGEMA used a single large shielded maintenance cell for the dismantling and cutting up of worn-out melters before removal, and for other maintenance purposes, BNFL designed and built separate, smaller, maintenance cells for each line.

It is common ground that, because a substantial amount of the make-up of the WVP consisted of plant purchased from or based upon proven design by COGEMA, it was not subjected by BNFL to the same degree of design development, testing and modelling that was applied to other parts of the plant requisite for dealing with spent Oxide fuel, or for that matter, MAGNOX fuel.

BNFL’s initial WVP throughput prediction, prior to obtaining any operational experience, was that it could realistically be expected to produce 600 containers of vitrified material (“glass blocks” as Dr Robson described it) per year, after ramp-up. BNFL’s experience after 1991 soon demonstrated that this was a serious over-estimate. The actual throughput statistics for the first 8 years of operation of the WVP are as follows:

In 1993 BNFL decided to build an additional vitrification line (“line 3”), incorporating lessons learnt from the early operation of lines 1 and 2, and with a design throughput of 250 containers per year. Its calculations at the time showed that in order both to maintain vitrification performance in line with the anticipated production of Oxide and MAGNOX HAL, and to start the process of reducing HAL stocks, a combined throughput in excess of 400 containers per year would be necessary. By December 1995, BNFL had decided to adopt the 75 % and 25% blending of OXIDE and MAGNOX HAL, primarily but not exclusively for the reasons already explained above. It was originally planned that WVP line 3 would be operational by the end of 1999.

In November 1995 NII published a report (“the NII 1995 Report”) entitled “Safety of the Storage of Liquid High Level Waste at BNFL Sellafield”. While commending in principle the vitrification of HAL as soon as possible as the best available long term storage solution for high level waste, and confirming that current storage arrangements for HAL were acceptably safe, NII gave 2 relevant warnings which, if converted into enforceable rulings by way of a Specification or Direction, could constrain the rate of Oxide reprocessing at THORP. The first was that it did not regard the building of further HASTs at HALES as an acceptable option for dealing with any excess of HAL production over vitrification capacity. This was not in any event an option favoured by BNFL. It had constructed two extra HASTs to enable HALES to deal with Oxide HAL produced by THORP and did not consider that there was a significant risk of the existing HAST capacity being insufficient, having regard in particular to the forthcoming availability of WVP line 3.

The second warning was that if an unexpected safety issue arose in relation to storage in HASTs necessitating their emptying, then NII would not hesitate to exercise statutory powers to direct the suspension of THORP operations, rather than the suspension of MAGNOX reprocessing, because of the difficulties of storing unprocessed MAGNOX spent fuel for long periods. The NII 1995 Report also noted, with approval, BNFL’s long term plan to reduce HAL stocks to a buffer level by a target date in the region of 2015, the buffer level being the minimum necessary to maintain smooth progress of vitrification, in the event of discontinuities in the supply of HAL from THORP and from the MAGNOX plant at Building B205. The publication of the NII 1995 Report did not cause any change in BNFL’s HAL storage and vitrification policy, which had by then become focused on the design, construction and start up of WVP line 3 as soon as reasonably practicable, to remedy the perceived deficiencies in lines 1 and 2.

Public interest in, and concern about, reprocessing and the issue of the storage of HAL led NII to decide in late 1998 to prepare and publish an update of its 1995 Report. Preparation took place in consultation with (among others) BNFL and with the Nuclear Safety Advisory Committee (“NuSAC”). BNFL’s contribution was to share with NII and NuSAC the results of its modelling of future projected levels of HAL stocks, its plans for improving WVP throughputs, and its objective to achieve by 2015 a buffer stock of HAL in the region of 280/460 cubic metres, sufficient to support ongoing reprocessing operations beyond 2015. By 1999, when these discussions took place, it was apparent to BNFL that the commencement of operation of WVP line 3 was going to be significantly delayed, but that the delay was not expected to give rise to a storage capacity difficulty at HALES, or to a need to reduce the reprocessing rates of either MAGNOX or Oxide fuel.

The NII 2000 Report

The NII 2000 Report was published in February 2000. Its principal conclusions and recommendations may be summarised as follows. First, it reaffirmed the conclusion expressed in 1995 that storage of HAL at HALES was, for the time being, acceptably safe, but that vitrification remained the best means of achieving long term safety. Secondly however NII considered that there had to be a demonstrable reduction from 2000 onwards in the potential hazard and risk represented by the ongoing storage of HAL, by the achievement of a reduction in the amount being stored. Thirdly, it stated that the reduction of HAL stocks to a buffer level (the amount to be the subject of further negotiation) by 2015 should be upgraded from a target to a requirement, and that BNFL should develop a series of HAL stock reduction curves with suitable periodic mile-stones for agreement with NII, so that its progress towards reducing HAL stocks to a buffer level by 2015 could be regularly monitored and if necessary enforced. At page 56 of the 2000 Report, NII set out purely for illustration a notional stock reduction graph, leaving it for the future to decide whether the quantity of HAL should be measured in terms of tonnes, cubic metres, kilowatts or bequerels, but in a form which suggested by implication that reduction should only become mandatory in about 2004 and then reduce steadily from that date until 2015.

Next, NII expressed its concern both at the operational difficulties which BNFL had experienced with WVP lines 1 and 2 and at the expected delay in the commencement of operation of line 3. The report stated that NII remained to be convinced that BNFL would be able to achieve the reduction of HAL stocks to a buffer level by 2015 unless throughput of lines 1 and 2 improved, there was no slippage in the coming on stream of line 3, and line 3 met its planned throughput target. If any of those conditions were not met, NII’s view was that BNFL would have to choose between one or more of the following options:

The 2000 Report contained a clear warning that, once stock reduction parameters had been agreed with or imposed upon BNFL, NII would not hesitate to use its statutory powers, in particular those arising under Licence Condition 32(4) of BNFL’s Sellafield operating licence to enforce compliance, and stated that “such actions may require curtailing the further build-up of HAL stock by halting or slowing reprocessing in THORP”.

It may be asked why that warning was directed specifically at THORP, rather than at spent fuel reprocessing generally at Sellafield, thereby including MAGNOX. There appear to have been two reasons for this. First, NII was aware of the difficulties in increasing the storage period for MAGNOX spent fuel, and the contrasting suitability of Oxide spent fuel for long term storage under water. Secondly, as became clear when negotiations with BNFL continued, NII regarded the storage of Oxide HAL as giving rise to inherently greater risks than the storage of MAGNOX HAL.

BNFL’s reaction – prior to the Specification

The analysis of the defendants’ allegations of breach of the Service Agreements by acts and omission after receipt of the NII 2000 Report do not call for a minutely detailed description of BNFL’s subsequent conduct to be included in this judgment. For the most part, the allegations focussed upon what it is said BNFL omitted to do, rather than what it did. In particular, by the time when BNFL reduced THORP’s throughput in order to comply with the Specification, Mr McGuigan acknowledged in cross examination that BNFL had no realistic alternative. The defendants’ case focuses upon alleged failures of BNFL to do something sooner, so as to avoid being forced subsequently into that position.

I shall nonetheless provide a summary of what BNFL did in response to the NII 2000 Report. It does not involve significantly contested facts and, as I have said, the events occurred sufficiently recently for there to be a reliable documentary trail for the purpose of establishing them.

NII had indicated in its 2000 Report the need to discuss stock reduction curves. On 17^th February 2000 NII stated in a letter its requirement to have agreed with BNFL appropriate HAL stock reduction curves by 18^th August 2000. BNFL’s response was to establish a dedicated project team under the leadership of a Mr Philip Hallington the Head of Technical & Operations Support for BNFL’s Waste Management Services division. He was then Dr Robson’s immediate superior. The objectives and methods of the project team are set out in considerable detail in a Framework Paper dated 28^th April 2000, prepared jointly by Mr Hallington and Dr Robson, and it was supplemented by Dr Robson’s evidence. In a nutshell, it called for an in-depth review of the issue under 4 principal headings: (1) The production of HAL (from THORP and the MAGNOX building at B205); (2) The storage of HAL (within HALES); (3) The vitrification of HAL (in the WVP); (4) Issues relating to the hazard potential of HAL.

Slightly more specifically, the project focused upon the analysis of the storage and vitrification of HAL, including in particular the methods of improving the performance of the WVP, but took for granted the continued production of HAL in accordance with the then current business plans both for Oxide and MAGNOX reprocessing, essentially as a “given”. As the Framework Paper put it: “the range of THORP scenarios has been constrained to those in the current business plan”. As Dr Robson put it: “we treated the THORP business plan essentially as a constant from which we could not and should not deviate”.

The then current business plan for THORP was the October 1999 Business Plan, which (uniquely optimistically) proposed throughput for the reprocessing of the remainder of the 7,000tU baseload at the following rates:

That approach necessarily involved rejecting NII’s alternatives both of early closure of MAGNOX and the voluntary reduction of THORP throughput. Indeed, BNFL as owner/operator of the MAGNOX power stations was in 2000 coming to a decision to commit to a second extension of the operating life of the majority of its power stations, a decision which was to involve the production of a further 1,500tU of spent fuel beyond that which had previously been provided for.

Nor did BNFL seriously consider the construction of a fourth vitrification line at the WVP, NII’s remaining suggested solution if then existing or planned facilities proved inadequate. Rather, it focused its energies on maximising the performance of the existing lines 1 and 2, and upon the completion of line 3, still then under construction, but experiencing serious delay. The effects of that analysis upon HAL stocks, and upon the rate at which they could be reduced to a buffer level by 2015, was subjected to intensive study and modelling, and the results were presented to NII on 31^st July 2000 in the form of a detailed submission, containing both a stock reduction envelope in graph form, with key mile-stones, a proposed buffer stock level, and a detailed justification of BNFL’s belief that, without reducing the production of HAL, or constructing a fourth vitrification line, those reductions and mile-stones could be achieved.

The reduction envelope was based upon current HAL stock of 1407 m^3, and a proposed buffer stock level of 200 m³ . The graph proposed an initial limit of 1500 m³ from 2000 to 2006, a gentle reduction to 1300 m³ by 2013, and much more rapid reduction to the buffer stock of 200 m³ in the final two year period before 2015. The graph expressly contemplated a modest increase from existing stocks for which the initial limit of 1500 m³ was designed to provide head-room. The graph assumed (as had the indicative graph in the NII 2000 Report) that a single aggregate HAL stock envelope was what NII were looking for, rather than any separate limit for Oxide HAL.

BNFL’s principal purpose in proposing a graph (and attendant mile-stones) which permitted a slight increase rather than any reduction of HAL stocks during the period 2000-2006 was to enable it to complete the reprocessing of the THORP baseload within that period, so that reductions could thereafter be achieved by the vitrification of HAL stocks which, from then on, would mainly be coming from the MAGNOX reprocessing at B205. BNFL was at the time in the process of revising its THORP business plan downwards from the throughput levels contemplated by the plan published in October 1999, to levels which contemplated throughputs slightly lower than 900tU per annum, and a programme which would complete the 7,000tU baseload during the 2004/05 business year. Thereafter, BNFL made provision for post-baseload Oxide reprocessing, starting at a reduced rate of approximately 750tU, and declining after 2008/09 to a rate of 234tU in 2013/14. The submission to NII envisaged rather larger annual quantities of MAGNOX reprocessing, in terms of tU per year, starting at 800 in 2000/01, with peaks up to 1250tU above a mean performance of 1050tU until 2009/10, and then falling rapidly to 0 by 2012/13. It will be recalled that MAGNOX HAR could be concentrated by evaporation to a much greater degree than Oxide HAR, so that in every year the proposed production of Oxide HAL in terms of pure volume greatly exceeded that of MAGNOX HAL, in broad generality by a factor of 2 or more.

Various alternative reprocessing cases were considered and modelled, all of which assumed no reduction on the throughput of the THORP baseload below the 2000 Business Plan, but which contemplated various different throughputs of THORP thereafter, and of B205 throughout. I shall refer to the reprocessing throughput which I have already summarised as the “base case”.

BNFL carried out detailed modelling and sensitivity analyses to establish alternative cases for vitrification throughput. Case 1 was the worst, case 2 was described to me as conservative, and case 3 was described as the best. These projected vitrification cases only covered periods though to 2003/04, and a case 4 projecting throughput to 2005/6 was constructed on a basis somewhere between cases 2 and 3 in terms of the balance between conservatism and optimism. This case showed that HAL derived from the base case from reprocessing at THORP and B205 could be constrained within the proposed envelope, designed to produce reductions only after 2006, and the achievement of buffer levels by 2015.

NII’s response to BNFL’s submission took the initial form of a letter dated 4^th August 2000 to Mr Hallington. Although generally appreciative of BNFL’s proposals, it contained warnings under the heading “risk reduction” which suggested that NII intended to focus more closely than it had in the February 2000 Report upon the safety implications of the production of Oxide HAL. This was made much clearer in a letter from NII to Mr Hallington dated 12^th September 2000, in which BNFL was criticised for failing to propose reductions in maximum volumes for 5 years, and warned that NII was “minded to specify in addition to the total volume stocks envelope a further envelope for the Oxide derived HAL” on the grounds that “risk is dominated by the Oxide HAL… it is therefore proposed to apply a fixed limit (of around 400 m³) following a suitable initial reduction period, to the total Oxide derived HAL (THORP plus blend) held in B215. This will have the required effect of driving WVP hard since any shortfall in HAL consumption would then necessitate a concomitant restriction on THORP throughput”.

This hardening of NII’s institutional heart represented a step change in its attitude toward the risks associated with the storage specifically of Oxide HAL, and it is apparent from the passage quoted above that NII regarded blended HAL (75% Oxide 25% MAGNOX) as falling wholly within the Oxide HAL category for the purposes of hazard analysis. At a meeting with BNFL on 14^th September 2000 NII made clear its wish that BNFL should propose a more rigorous stock reduction scheme than that which it had presented at the end of July, failing which NII might decide to impose upon BNFL a stock reduction programme of its own devising.

BNFL embarked upon a further process of review and modelling, with a view to presenting a more rigorous stock reduction programme to NII. Following further meetings between BNFL and NII, and a chasing letter from NII on 8^th November, again threatening to impose stock reduction curves if they could not be agreed, BNFL presented a revised programme for agreement with NII at a meeting on 6^th December 2000. Following an internal meeting of NII’s senior management team on 12^th December, BNFL were told informally that the revised proposal had been accepted and that NII wished to have it submitted formally in writing.

BNFL’s written proposal took the form of a letter from Mr Hallington to NII dated 14^th December, proposing new aggregate HAL and Oxide HAL stock reduction graphs, together with a list of maximum tracking volumes with key mile-stone commitments. These formed the basis of the Specification issued by NII on 31^st January 2001. The principal features of BNFL’s revised proposal, as reflected in the Specification, are as follows:

BNFL calculated that, albeit not without risk, it could operate within those constraints without reducing THORP’s throughput, as planned in the 2000 Business Plan. It aimed to do so by the use, among others, of the following specific measures, all which were explained to, and approved by, NII:

It will be apparent therefore that BNFL’s response to the NII 2000 Report did not involve adopting any of the 3 alternatives suggested by NII, should the vitrification rate prove to be inadequate. There was to be no fourth vitrification line, no reduction in THORP’s throughput below that contemplated in the 2000 Business Plan, and no early closure of MAGNOX generating capacity. Nonetheless, the decision to propose for agreement with NII a stock reduction programme did involve BNFL taking the risk that if the performance of the WVP fell short of the projections which had been modelled during 2000, to an extent which threatened a breach of either of the stock reduction curves in the Specification, then BNFL would have little alternative other than to reduce the throughput of THORP as its only means of avoiding such a breach. In particular, there was no capacity in the MAGNOX spent fuel storage ponds sufficient to accommodate incoming spent fuel in excess of a voluntarily reduced level of reprocessing in Building B205, and in any event, MAGNOX spent fuel could not safely be stored for extended periods, whereas Oxide spent fuel could be.

That a reduction in THORP throughput was from the date of the Specification BNFL’s only short term solution in the event of further under-performance at the WVP was not lost on NII. In its press release announcing the Specification, the Health and Safety Executive said this:

“The Specification marks a major step in an ongoing regulatory process aimed at securing conversion of the HAL, currently held as liquid in water-cooled tanks, to a passively safe form of storage(i.e. as glass blocks) as soon as reasonably practicable. We will not hesitate to use our regulatory powers to halt THORP reprocessing, should that be necessary, in order to keep BNFL within the Specification…..

The major contributor to the hazard potential comes from HAL derived from…[THORP]. Consequently, NII has also specified controls on the amount of HAL originating from THORP…”

Events Following the Specification

The Specification was imposed during the 2000/01 business year, and its HAL stock control envelopes came in to force with effect from 1^st January 2001. During that year (ending 31^st March 2001) both THORP and the MAGNOX plant at B205 had their production reduced below full capacity due to operating difficulties at HALES. I have already described the difficulties experienced with evaporator C (which treated Oxide HAR) between September 2000 and mid January 2001. Thereafter, HALES was unable either to receive HAR or to transmit HAL for vitrification at the WVP due to an incident which occurred during the connection of a new Caustic Scrubber plant to the HALES ventilation system, which caused a temporary loss of ventilation, and an enquiry into the incident which had threatened plant safety. That shut-down lasted until 25th April 2001.

During the same business year, the WVP suffered its own two month shut down between July and September 2000, due to difficulties with the cables supplying power to the melters. A further planned shut-down occurred between November 2000 and January 2001, to allow for the connection of WVP Line 3. From late January until May 2001 the WVP was closed due to the closure of HALES, as I have already described.

The net effect of all these difficulties was largely self-cancelling, in the sense that although the throughput of the WVP was only 152 containers against a projected 200, that was sufficient to prevent the reduced throughput of THORP and B205 giving rise to any breach in the storage envelopes imposed in January 2001 for the rest of the business year. In fact, there was a little head room under each of the limit curves at the end of the year.

During the financial year 2001/02 the throughput of both THORP and B205 was unconstrained by the Specification. THORP’s throughput of 736tU fell short of an originally projected 895tU first, because it suffered a three week delay in start up while the enquiry into the caustic scrubber incident at HALES continued, losing 80tU of production. Secondly, the balance of the throughput short-fall was caused by the need to accommodate a revised fuel batch reprocessing programme so as to maximise the efficiency of evaporator C.

Throughput at the WVP fell well short of projection, at 120 containers against a projected 330. Three months production was lost due to electrical arcing, and further losses were caused by melters failing earlier than their projected lifetimes, and as the result of maintenance issues relating to cranes operating in the vitrification maintenance cells. WVP Line 3 had been projected to produce 30 containers during the first stage of its commissioning and ramp-up. Although commissioning started, no containers had been produced by the end of the business year.

The business year 2002/03 was the first in which BNFL was forced to reduce THORP’s throughput to avoid exceeding the storage limits set by the Specification. THORP’s throughput was revised down from 838 to 500tU. In fact it produced 502tU.

This had been foreseen before April 2002, as the result of review of the anticipated future performance of the WVP, the results of which were presented to BLCs at a meeting of the JC in February. For convenience I set out below the throughput in Case 4, upon the basis of which the Specification had been issued, and the revised February 2002 predictions:

This led to a planned reduction in THORP’s throughput for the business years 2002/03 onwards (to the extent necessary to complete the 7,000tU baseload). I set out below the throughputs predicted first in November 2001 at the 49^th TEC meeting with BLCs, unconstrained by the Specification, and the constrained throughputs predicted in February 2002:

The throughput of MAGNOX reprocessing at B205 was not deliberately constrained to avoid breach of the Specification during 2002/03. Nonetheless, for other reasons, it fell substantially short of target. By contrast, the WVP slightly out performed its revised target, at 313 containers against a projected 250, but fell short of the original case 4 projection of 390 containers.

The combined effect of all these factors was that aggregate storage of HAL during that year came within 41 m³ of its constraint, and storage of Oxide HAL came within 56 m³ of its constraint. Had THORP’s throughput been maintained at the originally projected level, both constraints would have been breached.

The 2003/04 business year can be briefly described. THORP met its planned (Specification constrained) throughput of 671tU. MAGNOX slightly under performed its anticipated throughput at 1038tU, and the WVP just exceeded its planned production of 340 containers, by one container. As a result, neither the aggregate nor Oxide HAL limits were threatened with breach. The nearest approach to each was 35 m³ and 16 m³ respectively.

The last year in which the Specification made any relevant difference to THORP’s throughput, that is before THORP incurred an unplanned major outage in 2005, was the 2004/05 business year. Both THORP and B205 under-performed their planned throughputs, in the case of THORP due to a suspected defect in its Medium Active Salt Free Evaporator which stopped production during part of December 2004 and January 2005.

The WVP out-performed its planned production of 460 containers by 18 containers. Improvements were experienced in melter life in lines 1 and 2, and a Vitrification Assistance Program which BNFL had agreed with COGEMA came into practical operation, as did the new Vitrification Test Rig.

The combination of under-performance of reprocessing, in both plants, and good performance at the WVP, ensured that the aggregate storage limit was not exceeded, actual storage coming within 18 m³ at its nearest approach to the limit. By contrast, there was a breach of the Oxide storage limit during the period from December 2004 to February 2005, of an amount which at its highest was 23 m^3. As noted above, the occasional breaches of this kind were contemplated by the Specification, subject to limits in number and aggregate duration.

In the result therefore, the programme of deliberately reduced THORP throughput instituted in the 2002/03 business year and notified to BLCs in February 2002 did, by a narrow margin, sufficiently constrain the amount of stored HAL at HALES during the 2002/03 and following business years. This restraint on THORP’s throughput was one which had not been planned at the time of the discussions with NII which led to the Specification, but it became necessary when the foreseeable risk that the WVP would fail to achieve its predicted performance materialised.

It is tempting, but in my judgment over-simplistic, to measure the effect of the constraint imposed by the Specification on THORP’s throughput by the aggregate amount of the year-on-year tonnage reductions revealed by a comparison between the revised February 2002 projections against their predecessors in the then current business plan. But that analysis ignores 3 additional considerations, none of which were explored in any depth in the evidence. The first is that operating difficulties, for example in 2003/04 caused THORP to under-perform even that revised throughput plan. Since those difficulties occurred relatively late in the year, it may well be that THORP was run at less than full capacity earlier in the year to avoid breach of the Specification, so that it is not simply a question of THORP failing, for other reasons, to produce the throughput permitted by the Specification.

The second factor is that even though in 2004/05 THORP produced approximately 120tU less than planned, the Oxide specification was nonetheless still breached. Accordingly, if THORP had not experienced unexpected operational difficulties late in that business year, it might have needed to be slowed down even more than contemplated by the February 2002 programme.

The third consideration is that it is a matter of speculation whether, if unconstrained by the Specification, THORP would in fact have achieved the full throughputs planned for it in the business plan which immediately preceded the February 2002 revision. Two possibilities suggest themselves. The first is that defects subsequently encountered might have had occurred earlier, as equipment wore out under more rigorous operating requirements. The second is that some quite unconnected defect might have occurred if THORP had been operated at full capacity, which did not in fact occur before the major outage which stopped THORP’s further production in 2005. It is of course theoretically possible that, if unconstrained, THORP would have out performed the throughput planned prior to the revision necessitated by the Specification. Again, that is a matter of speculation.

ANALYSIS OF THE MAGNOX PRIORITISATION ISSUES

Construction of the Service Agreements

I have already described the main provisions of the Service Agreements, at paragraphs 157 - 170 above. For present purposes the relevant provisions are clauses 2.7, 6.1, 6.2 and 18.1. In addition, the defendants alleged, but BNFL denied, the following implied term, namely that BNFL:

“shall not give priority to its other operations to the detriment of reprocessing for Baseload Customers”

There is an obvious tension between, on the one hand, BNFL’s obligation to use its reasonable endeavours to undertake storage, reprocessing and conversion of waste into residues in an economical manner having regard to the state of the art and plant requirements, provided in clause 2.7 and repeated in clause 18.1.2 as a “principle of negotiation” and, on the other hand, BNFL’s discretion as to the timing and quantities of reprocessing conferred by clause 6.1.

There is also a question whether the exclusive remedy of fee abatement in the event of failure by BNFL to adhere to the Target Schedule conferred by clause 6.2 is as a matter of construction a remedy for failure resulting from a breach from the Service Agreement, or merely from failure resulting from the exercise of the clause 6.1 discretion but without breach of any other provision of the Agreements.

The parties were broadly agreed (and correctly in my judgment) that the clause 6.1 discretion as to the timing of reprocessing was subject to BNFL’s reasonable endeavours obligation in clause 2.7. There are numerous circumstances in which BNFL might need to adjust the reprocessing rate and quantities out-with that provided in the Target Schedule without being in breach of clause 2.7, which is mainly if not exclusively a qualified best endeavours obligation to achieve economy. Of course, any prolongation of the time taken for the reprocessing of the baseload would, because of the cost-plus payment provisions in the Service Agreement, lead to a reduction in economy, because of the extra time during which the BLCs would be obliged to pay for the operation, manning and upkeep of the plant. But the qualification by reference to “state of the art and plant requirements” plainly envisaged that circumstances might arise in which rigid adherence to the Target Schedule became either impossible or wholly uneconomic, and the “reasonable endeavours” qualification makes clear the parties’ common assumption that such circumstances might arise notwithstanding BNFL’s best efforts to the contrary.

The defendants’ case was that the clause 6.1 discretion was limited to what was described in closing submissions as “micro-management”, such as the necessary decision on a regular basis as to what quantities of particular types of fuel from particular BLCs were to be processed in any given month, campaign or year. By contrast, BNFL submitted that, subject to the clause 2.7 qualified obligation, the discretion was unconstrained.

I have not found it necessary to resolve that issue, precisely because it is common ground that, however construed, clause 6.1 did not derogate from BNFL’s obligation under clause 2.7. If a different analysis of the issue leads a higher court to conclude that this particular question of construction needs to be resolved, then I would have resolved it in favour of BNFL’s broader submission. It seems to me that plant requirements can readily be envisaged which required much more substantial departures from the Target Schedule than may properly be described as micro-management, and in the absence of any obligation to adhere to the Target Schedule it seems to me that BNFL’s discretion as to the timing and quantities submitted for reprocessing would be constrained only by its express and implied contractual obligations in the Service Agreement, even if clause 6.1 was not there.

In my judgment the fee abatement remedy provided in clause 6.2 for a failure to adhere to (or delay in achieving) the Target Schedule is only an exclusive remedy where that failure or delay is the consequence of the legitimate exercise by BNFL of its clause 6.1 discretion. In other words, it does not limit the BLCs’ remedies in respect of a breach of the Service Agreements. Although the 6.2 remedy arises in the event of what is described as failure, (with the implication of some form of wrongdoing by BNFL) that language is by no means sufficiently clear to justify the conclusion that the modest fee abatement remedy was intended to constitute the BLCs’ only recourse in respect of any breach (however serious) of the Service Agreements leading to a non-compliance with, or delay in achieving, the Target Schedule.

I turn to the alleged implied term. BNFL was not at the date of the Service Agreements the owner of the UK MAGNOX power stations. This only occurred in 1996. But BNFL was itself a BLC, as the owner of certain AGR power stations using Oxide fuel, so that it did have “other operations” in terms of a reprocessing requirement lying alongside that of the other BLCs.

It is tempting to imply some sort of term to the effect that BNFL would not as Reprocessor prefer itself as a BLC over the other BLCs. Perhaps more generally, it is tempting to imply a term to the effect that BNFL would act fairly as between all its BLC customers. An implied term of either type would give effect to the perception that a commercial agreement of this type in which a service is being supplied to a number of customers under agreements in identical terms pursuant to which all contribute to the cost would be operated by the supplier, as between them, in good faith.

I have however been persuaded that, at least for the purpose of solving the MAGNOX prioritisation issues, such an implied term fails the necessity test, for the following reasons. Each Service Agreement is at least in form a separate and distinct agreement between BNFL and a particular BLC. To the extent that a preference or prioritisation of one BLC’s reprocessing requirements over those of another BLC caused an increase in the unit cost of reprocessing to that other, then it would prima facie represent a failure by BNFL to use reasonable endeavours to reprocess that BLC’s fuel in an economical manner contrary to clause 2.7, unless the preference was justified by reference to the state of the art or plant requirements. Thus for example a purely voluntary preference by BNFL in favour of its own reprocessing requirements as a BLC which had nothing to do with the state of the art or plant requirements would simply constitute a breach of clause 2.7. So also would a similar preference in favour of one (non BNFL) BLC over another. By contrast, plant requirements might give rise to a reasonable need to process one particular type of fuel ahead of another, so that BLCs submitting that particular type for reprocessing would in fact, although not in intention, be preferred, but without a breach of clause 2.7. I should note that, generally speaking, the cost plus payment arrangements were not such as to reduce the unit cost of reprocessing for a particular BLC over those of another, merely because that BLC had its fuel reprocessed earlier. The cost structure is linked to the reprocessing of the entire baseload, in accordance with the Target Schedule: see Appendix 6 Part A.

In my judgment, the same analysis holds good in relation to any question as to preference as between Oxide fuel reprocessed at THORP, and MAGNOX fuel reprocessed at Building B205. Since both types of reprocessing operation shared HAL storage, evaporation and vitrification facilities, a voluntary preference by BNFL in favour of MAGNOX over THORP in the event of a down-stream bottleneck in evaporation or vitrification would again, prima facie, give rise to diseconomy in the manner in which Oxide fuel was reprocessed and its waste converted into residues, which would constitute a breach of clause 2.7 in the absence of reasonable endeavours by BNFL and/ or a need to prefer MAGNOX consequential upon the state of the art or plant requirements. Leaving aside BNFL’s reliance on the force majeure provisions of clause 18, the effect of the Specification was, in the year 2002/03 to introduce just such a downstream bottleneck, because of the inability of the WVP to convert sufficient HAL to avoid infringing the HAL storage envelopes.

Finally, in relation to issues of contractual construction, I must deal briefly with the parties’ submissions as to the meaning and effect of clause 18. For present purposes, the relevant trigger for its operation is :

“If the performance of this Agreement or of any obligation hereunder by either party is prevented, hindered or delayed by reason of any circumstances beyond such party’s reasonable control, which circumstances include but not be limited to…. restraint of government or any other authority having jurisdiction in respect of the performance of any obligation under this Agreement….”

It is common ground that the Specification was a “restraint of …. any other authority having jurisdiction….” within the meaning of clause 18.1.1, and that, by the 2002/3 business year, the Specification delayed the performance of the Service Agreement by BNFL, since, by then, the deliberate slowing down of reprocessing at THORP was BNFL’s only means of avoiding a breach of the Specification.

But the defendants say that this was not a circumstance beyond BNFL’s “reasonable control”. One of the steps which it is said, lay within BNFL’s control was to reduce, or more importantly, to promise NII that it would reduce, the inflow into Sellafield of spent MAGNOX fuel by early closure of one or more of the MAGNOX power stations. Putting on one side for the moment BNFL’s submission why that proposition is wrong in fact, Mr Barnes QC submitted that in any event the phrase “beyond such party’s reasonable control” in clause 18.1.1 means, in relation to BNFL, its reasonable control as the Reprocessor under the Service Agreements, and not as owner of the MAGNOX power stations, which has nothing to do with the Service Agreements, and the spent fuel from which was not reprocessed pursuant to the Service Agreements.

In my judgment, that submission is correct. In this respect it is of importance that BNFL was not at the time of the making of the Service Agreements the owner of the MAGNOX power stations, but in my judgment Mr Barnes QC’s submission would have been correct even if it had been.

If the contrary construction were correct, it would involve the un-commercial and unreasonable conclusion that BNFL agreed by the Service Agreements to resolve any competition between the requirements of THORP and the requirements of the MAGNOX fuel reprocessing plant for the use of shared facilities not merely by not prioritising the requirements of MAGNOX, but by subjecting them to an overriding priority in favour of THORP, in relation to any of the listed force majeure events giving rise to such competition.

It follows that even if the defendants’ case that, as a matter of fact, a commitment by BNFL to early closure of MAGNOX might have avoided the need to slow down reprocessing at THORP in 2002/3 were correct, a decision within its control not to close the MAGNOX power stations early would not deprive BNFL of a force majeure defence under clause 18.1.1.

The factual issues

By the time of closing submissions, the defendants’ case in support of its counterclaim based on BNFL’s prioritisation of MAGNOX or related breach of the Service Agreements boiled down to the following propositions:

It is necessary briefly to describe why, save only in relation to the suggested option of improving the capacity of lines 1-3, all the other options alleged to have been within BNFL’s control are relied upon only for the purposes of the indirect causation route. The essential reason is that the defendants recognised by the end of trial that none of them would have had a direct effect in sufficient time. I shall explain why that is so, in relation to each of the alleged options, when dealing with them one by one.

The fact that, by the end of the trial, the defendants’ main case was based upon the alleged indirect effect of the taking of one or more of the available options means that each of them requires an assessment of whether what I have already described as a hardening of NII’s corporate heart would have been avoided or alleviated, so as either to give rise to no Specification, or to a more lenient one. That analysis necessarily involves a large element of hypothetical speculation as to the attitudes and policies of NII, from which no witnesses were called by either side. Dr Robson had the benefit of having been actively involved in the very negotiation in issue. Although Mr McGuigan’s main specialisation lay in nuclear safety, and he had previous experience of negotiating with NII, he did not claim to have been actively involved in any negotiation with NII comparable with this one. He accepted in evidence that he had no direct experience of a specification being either imposed or withdrawn by NII.

The result is that in addressing the question whether the taking by BNFL of the suggested steps would have had an indirect effect in removing the need to reduce THORP’s capacity, so that the failure to take the step is a breach of contract, the court is required to speculate in something of an evidential vacuum. Difficulties of this kind are occasionally addressed in cases of breach of contract by reference to the “loss of a chance” doctrine, but no such submission was made in this case.

Furthermore, the question what effect the taking of the suggested steps by BNFL would have had upon NII arises in this case not merely at the causation stage of the analysis, but also at the stage when the court has to decide whether by not taking the proposed steps, BNFL committed a breach of its reasonable endeavours obligation in clause 2.7 of the Service Agreements. Strictly therefore, the question is whether it should have been apparent to a reasonable person in the position of BNFL between February 2000 and January 2001 that the taking of one or more of those steps was sufficiently likely to lead to a change of heart by NII, to justify the cost implications inherent in taking, or promising NII that it would take, such a step or steps. Clause 2.7 is, as I have emphasised, a reasonable endeavours obligation to perform the contracted services in an economical manner.

In order to complete the analysis of each of the suggested options, it is also necessary to bear in mind, as I have held, that BNFL’s belief by December 2000 was not that the imposition by way of Specification of the stock reduction curves by then discussed and negotiated would lead to a requirement to reduce THORP’s throughput. On the contrary, BNFL’s belief was that the negotiated stock reduction curves would have no such consequence. There was merely a risk that a Specification along those lines would do so, if BNFL’s then predictions as to the WVP’s future performance turned out (as in due course they did) to be optimistic.

It follows that in deciding whether a decision by BNFL not to take one of the proposed steps constituted a failure to use reasonable endeavours to perform the contracted services in an economical manner within clause 2.7, it is necessary to ask not, “what steps were reasonable to avoid the need to slow down THORP?”, but, “what best endeavours were reasonable to address the risk that such a need might arise in the future?” Asking the question in that way has obvious implications in terms of cost-benefit and practicality, and increases the burden on the defendants of showing that by not taking any of the proposed steps, BNFL failed to use reasonable endeavours.

In cross–examination and in closing submissions, the defendants sought to evade this particular increase in their burden by suggesting that the predictions upon which BNFL based its belief that it could operate within the stock reduction curves proposed for a Specification without reducing THORP’s throughput were themselves unreasonably optimistic, but this was neither pleaded, nor was the basis for that allegation set out, for example, in Mr McGuigan’s expert report with anything like sufficient particularity to enable it to be properly litigated, at a trial where the ambit of the defendants’ case in relation to MAGNOX prioritisation had, after full argument, been precisely limited by Master Moncaster.

Ms Hannaford (who bore the burden of advancing the defendants’ case in relation to this issue) directed my attention to paragraphs 325-329 of Mr McGuigan’s report under the heading “The development and performance of the WVP in the 1990s”, which included a table comparing BNFL’s various WVP throughput predictions with its actual performance. Curiously, that table did not in terms refer to predictions made during 2000, upon the basis of which the Specification was negotiated, and in any event, Mr McGuigan said no more than that it was apparent (as of course it was) that actual performance never reached the predicted throughput. That comes nowhere near to a pleaded or particularised case as to the respect in which, if it be alleged, those predictions were unreasonable. An analysis of that question would have involved examination of the predictions and the underlying stochastic modelling which lay behind them at a similar depth to that which had to be applied to THORP’s throughput predictions under the RTR issue, and no such exercise was attempted. It follows that this means of evasion of the requirement to express the need for further steps, beyond those taken by BNFL, as those necessary to deal with a risk rather than a probability that THORP’s throughput would have to be reduced, fails.

Early Closure of MAGNOX Power Stations

My conclusion that this was not a step within BNFL’s reasonable control as Reprocessor under the Service Agreements is sufficient to dispose of this aspect of the case, in the sense that, even if as a matter of fact such a step would have achieved its suggested effect upon the attitude of NII, BNFL would nonetheless have had available its force majeure defence to any liability for the consequential delay in completion of the baseload greater than fee abatement. Furthermore, it seems to me plain that the early closure of what were by then its own power stations is not on any view within the class of reasonable endeavours pursuant to clause 2.7. The throughput of THORP could be reduced without any requirement of the BLCs to close or reduce their generating capacity, because of the suitability of spent Oxide fuel for long term storage under water.

Against the risk that a higher court may take a different view on construction, I shall address this first option on the hypothetical basis that the early closure of MAGNOX generating capacity was both a potential reasonable endeavour under clause 2.7 and something within BNFL’s reasonable control within clause 18.1.1.

It was clear by the end of the trial, and not disputed, that unless a means of dry storage of MAGNOX fuel could be found, the direct effect of the early closure of one or more MAGNOX power stations would have been to produce a short term increase (or bulge as it was called) rather than decrease in the amount of MAGNOX fuel arriving at Sellafield for reprocessing. This is because, on its decommissioning, a power station has to be stripped both of spent and un-spent fuel, whereas while in operation it is only a much smaller amount of spent fuel that needs to be reprocessed in any given period.

The defendants, supported by Mr McGuigan, submitted that although it appeared that dry storage of MAGNOX fuel was not possible for security reasons outside a power station, there was no reason why it could not simply be left inside a decommissioned power station, and Mr McGuigan stated his belief that this had in fact been done at the Calder Hall power station at Sellafield. This suggestion was made for the first time only during Mr McGuigan’s cross-examination, but he had no answer to the difficulty that storage of fuel requires to be separately licensed, and that there was no evidence that any of the MAGNOX power stations still active in 2000 had a licence for the long term storage of fuel after decommissioning.

In any event, the unlikelihood that an early closure of MAGNOX generating capacity would directly have alleviated BNFL’s difficulties by 2002/3 was compounded by the evidence that the MAGNOX spent fuel storage ponds were already heavily laden by 2000, and that the fuel stored there would in any event had to have been reprocessed relatively quickly to avoid corrosion problems, so that no early closure programme would have impacted in any significant way upon the outflow of MAGNOX HAR in time. I therefore conclude that the early closure of MAGNOX power stations would have been unlikely to have had a significant effect on the output from Building B205 of MAGNOX derived HAR before about 2005.

The next question is therefore whether the offer of such early closure would have softened the heart of NII. As would appear, each of the first three alleged options would, if promised, only have had a delayed effect upon the stock of HAL being stored at Sellafield. The relevant question in relation to the risk of being unable to complete the THORP 7,000tU baseload in time is whether NII could have been persuaded to a more lenient approach to HAL stocks by the then perceived completion date for that baseload, namely 2005.

The parties’ experts canvassed two rival theories in relation both to this option and to the second and third, described above. For the defendants, Mr McGuigan’s view was that the hardening in NII’s attitude during 2000 which I have described was mainly the result of its not being satisfied that BNFL’s plans would achieve a reduction of HAL to buffer stocks by 2015. Accordingly, he said, any proposals which increased the ability of BNFL to reduce its HAL stocks even during the latter part of the period 2000 to 2015 would have led to a more relaxed attitude on the part of NII throughout the relevant period. He went so far as to suggest that if NII’s doubts had been satisfactorily resolved, it would not have imposed a Specification at all. Plainly, said Mr McGuigan, an early closure of the MAGNOX power stations would, regardless of any immediate bulge in fuel requiring reprocessing, have greatly relieved BNFL’s storage difficulties, and its ability to reduce HAL buffer stock by 2015, by impacting upon storage levels with a radical effect from 2005 onwards.

Dr Robson’s view, again in relation to each of the first three options, was that, as a safety regulator operating under the ALARP principle, NII would have accepted with open arms any proposals to achieve a reduction in HAL stocks more quickly or with greater certainty than reflected in the predictions actually used by BNFL when negotiating in 2000, but responded to them by increasing, rather than relaxing, the stringency of the Specification. He said that NII’s starting point was that the storage of HAL as a liquid rather than its component waste in glass blocks was an inherently hazardous operation, and that all steps should be taken to reduce that hazard as quickly as possible, within the parameters of proportionate expense. If BNFL were to propose, for example, the early closure of MAGNOX, or the construction of a fourth vitrification line as hazard reduction measures which were not disproportionately expensive, then NII’s policy would have been to accept the consequential hazard reduction benefits, over and above those actually being negotiated, and build them into its stock reduction curves in the Specification.

My review of the blow by blow conduct of the negotiation between BNFL and NII following the publication of the 2000 Report suggests to me that there are points to be said in favour of both those rival views. If forced to choose, I incline on balance to think that Dr Robson’s view is the more persuasive, both because his close involvement in the negotiations gave him an advantage which was not wholly cancelled out by his lack of independent status, and because his view seems to be more consistent with the regulatory framework encapsulated in the ALARP principle. By contrast, Mr McGuigan’s attitude to the circumstances in which NII might impose a Specification seemed to me to be largely unsupported by objective reasoning, or practical experience, and in reality, more of a hunch.

In any event, the real question is not whether either of those views is definitely right, but whether the promise of an early closure of MAGNOX power stations offered a sufficient prospect of softening NII’s attitude to make it, with its financial consequences, a required reasonable endeavour within clause 2.7 to avoid a risk of the dis-economy of having to slow down THORP’s throughput. Put that way, the clear answer to the question is, in my judgment, in the negative. The essential basis of BNFL’s approach to its negotiations with NII during 2000 was to treat the generation of HAL from both THORP and the MAGNOX plant, in accordance with their current business plans, as a given. In my judgment that approach did not involve BNFL in any breach of clause 2.7 of the Service Agreement.

Improving the vitrification throughput rate of Lines 1-3

The defendants’ original case under this heading was that BNFL had rejected the option of improving the throughput of lines 1-3: see paragraph 19 and 20 (iv)(a)(9) of the Re-Amended Rejoinder and Reply to Defence to Counterclaim. Since it became apparent during trial that BNFL’s negotiations with NII during 2000 did include proposals to improve the throughput of lines 1-3, the defendants’ case in this respect in closing submissions was that BNFL failed to take reasonable steps within its control to institute enough improvements, or improvements in good time.

Mr McGuigan’s report focused upon two aspects of this. First, it said that a technology transfer agreement with COGEMA, the benefits of which were intended to enable BNFL to improve its own similar (and COGEMA derived) vitrification plant, should not have been left to be made as late as 2004. BNFL ought to have sought and obtained assistance from COGEMA sooner. Secondly, he pointed to the fact that the commencement of production of glass blocks on WVP line 3, which had already been significantly delayed by February 2000, was in the event further delayed by slippage in the associated staff training programme, something which was within BNFL’s control to remedy by the application of greater financial and other support.

In addition, Ms Hannaford submitted that (as was the case) BNFL’s draft Vitrification Improvement Plan was not issued until April 2002, and the Capital Expenditure Proposal for the proposed vitrification test rig was not submitted to BNFL’s board until March 2002.

Taking those points in turn, Dr Robson’s evidence was that while BNFL lacked no desire to obtain technical assistance from COGEMA, this was impeded by the imposition, until 2003, of a condition by COGEMA that technical assistance would only be provided in what he described as a “black box” i.e. on terms that BNFL would not have unfettered access to it. It is no criticism of BNFL that it found this condition unacceptable, and Miss Hannaford acknowledged in closing submissions that the timing of the obtaining of technical assistance from COGEMA was not a matter fully within BNFL’s control.

In any event, it seems to me that even if an earlier arrangement with COGEMA had been made, from a standing start in February 2000, it would have been unlikely that the benefits would have flowed through in terms of modified design in time to alleviate BNFL’s problems in 2002/3, although it is possible that some improvement might have been achieved by alteration of working practices on COGEMA’s advice.

The delay to the commencement of WVP line 3 caused by staff training requirements appears from the evidence to have been modest, and its effect was limited to the postponement of an initial commissioning throughput of 30 containers from the end of the 2001/02 year until early in the 2002/3 year. Its direct effect was, as it seems to me, minimal. As to the delay in the implementation of WVP improvements and the construction of the test rig, I consider that neither of these, if implemented sooner, would have led to a significant improvement in throughput in time to alleviate BNFL’s difficulties. Internal reports show that the test rig proposal was expected to give rise to improved plant performance only four years later. As for the other improvements, I bear in mind that the making of significant changes to the design and construction of a plant already operating in a radioactive environment inevitably necessitates outages of its own, and takes substantial time to achieve. While the evidence made it impossible for me to place a date upon which, if implemented sooner, those improvements would have begun to have a direct effect, it did not satisfy me that the effect would have been obtained in time.

Turning to indirect effect, Mr McGuigan said that:

“Had BNFL been able to put a convincing argument to the NII for a prospective increase in WVP performance, the NII might well have taken a view that it was sensible to delay issuing the Specification until it was clearer whether or not WVP performance was improving.”

Construction of a fourth Vitrification Line

It is clear, as a matter of fact, that NII maintained throughout most of 2000 an interest in the question whether BNFL would be prepared to construct a fourth line at the WVP. Equally, it is clear that although BNFL addressed this question, it was from start to finish firmly disinclined to do so, at least at that stage. Dr Robson’s evidence showed that BNFL’s thinking on this issue may be summarised as follows. First, there was no prospect that a fourth line could be put into operation in sufficient time to have any direct effect on vitrification during the period then regarded as sufficient for the completion of THORP’s 7,000tU initial baseload. The defendants did not challenge that conclusion.

Secondly, a fourth line would be very expensive to construct, and since there existed no firm business plan or clear expectation as to the amount of post- baseload Oxide reprocessing which would be carried on thereafter at THORP, and since by that time improvements to lines 1 and 2, and the operation of line 3 was expected to be sufficient to deal with residual MAGNOX reprocessing, with spare capacity for some further Oxide reprocessing, there was no justification in 2000 for very substantial expenditure on a plant which might well never be needed.

The defendants’ case was that, regardless that a fourth line would be of no direct assistance in the reprocessing of the 7,000tU initial baseload, a promise to construct it would have been of substantial comfort to NII, making it virtually certain that the reduction to buffer levels by 2015 would be achieved, and likely that NII would therefore have been much more relaxed about a Specification (if any) as a means of compelling that result. Dr Robson’s answer was that to promise of a fourth line would merely have enabled NII to insist on a more rapid reduction to a buffer stock during the latter part of the period prior to 2015, rather than bring about an alleviation of the proposed curve during the early part.

There are in my judgment insuperable difficulties with this limb of the defendants’ case. The first is that for the reasons which I have already explained in relation to early MAGNOX closure, I have not been persuaded by the basic proposition advanced by Mr McGuigan that the offer of further improvements in vitrificatation by BNFL would have led to an alleviation, rather than a toughening, of the regulatory stock reduction regime. In particular, I consider that a reading of the materials emanating from NII beginning with the 2000 Report and extending during 2000 suggests that it is more likely that NII’s attitude was that although BNFL had the option to achieve the requirements of the proposed Specification by construction a fourth vitrification line if it chose, a promise to do so would by no means have alleviated the requirement for the Specification.

The second difficulty lies in the analysis of the question whether the construction of line four at a very substantial cost but not for the purpose of actually vitrifying HAL derived from the Oxide baseload to be serviced under the Service Agreements, but rather for vitrification of other waste generated subsequently, was capable of being a reasonable endeavour to perform the contracted services in an economical manner. It would have involved spending an enormous amount of money for the indirect purpose of persuading NII not to impose a Specification which created only a perceived risk, rather than a probability, that the throughput of THORP would thereafter have to be reduced. I asked Miss Hannaford whether it was her case that the defendants would have accepted that this expenditure, not directed towards the reprocessing of spent fuel under the Service Agreements other than by very indirect steps, would have been willingly paid for by the BLCs. I received no clear answer to that question, and I consider it highly unlikely that the BLCs would have been prepared to do so, merely to alleviate a perceived risk. If they would not, then I can see no reason why BNFL should have promised NII to construct what they then regarded as a potentially redundant but expensive plant, on the basis that any attempt to charge the BLCs for it would have involved a risk that, not being required for the performance of the contracted services, the expenditure was not recoverable from them.

I therefore conclude that BNFL’s decision not to promise NII to construct a fourth vitrification line, in the hope that it would thereby lead to a less stringent Specification, was not a breach of clause 2.7, or any other obligation, express or implied, in the Service Agreements.

Proposing alternative stock reduction curves

The stock reduction curves imposed by the Specification were all based on the volume of stored HAL. It was common ground between Dr Robson and Mr McGuigan, and recognised in 2000 both by BNFL and NII, that volume was a relatively unsatisfactory measure of hazard, in relation to stored HAL. Put shortly, volume was a crude proxy for risk. In fact, NII expressed its wish during 2000 to agree with BNFL a Safety Index by which the risk posed by the storage of HAL at Sellafield could be better measured. Potential alternatives included measurement by reference to mass, or by reference to radioactive content (normally expressed in becquerels). In fact, no such alternative index was agreed prior to the imposition of the Specification. NII was prepared to proceed on the basis of volume even though, (for example in relation to BNFL’s proposed reduction in the gadolinium content of HAL, and further evaporation of HAL while stored), this enabled BNFL to achieve volume reductions without any corresponding reduction in hazard.

Mr McGuigan’s opinion was that the use of a mass based curve would have provided BNFL with greater flexibility in the management of stored HAL of different types in any stock reduction curves imposed, and that in particular it would have obviated the need for a separate stock reduction curve for Oxide (including blended) HAL. He offered no statistical explanation why (if it be the case) this would in fact have alleviated BNFL’s difficulties in time.

By contrast, Dr Robson’s evidence was first that the use of a mass based curve would have aggregated rather than alleviated BNFL’s problems, since the mass of radioactive waste within the stored HAL actually increased between 2000 and 2003, whereas, measured by volume, BNFL were able to achieve the required reduction. Secondly, his opinion remained that regardless of the basis of measurement, NII would have been as firmly committed to a separate identification of Oxide (including blended) HAL reduction, because of its firm view that Oxide HAL presented an inherently greater hazard than MAGNOX HAL.

I have already concluded that NII did indeed hold such a view, and I have more generally been persuaded that Dr Robson’s approach to the pros and cons of use of a different measure of risk are to be preferred to those of Mr McGuigan, in particular because his views appear to me to be more firmly founded upon, and in consistent with, the facts as to the amount of HAL actually stored during the relevant period, when measured by volume or mass as an alternative. In short, I am not persuaded that for BNFL to have proposed an alternative basis of measurement of risk in any stock reduction curve would have alleviated rather than aggravated its difficulties. It follows that BNFL committed no breach of clause 2.7 of the Service Agreements by not doing so.

Persuading NII not to include the MAGNOX element of blended HAL in the Oxide stock reduction curve

I can deal with this alleged option very shortly. The effect of Master Moncaster’s refusal to the defendants of permission to amend to include allegations of breach prior to February 2000 required Mr McGuigan to delete from his expert report a substantial part of his analysis of the reasons why he regarded this last option as one which would have alleviated BNFL’s difficulties. In cross examination, he very fairly volunteered that, being unable to rely upon those deleted passages, he could no longer make good an opinion in support of this final limb of the defendants’ case on MAGNOX prioritisation. I infer that his opinion required an assumption that BNFL would have needed to do things prior to February 2000 which it did not do for that option to have stood any reasonable prospect of achieving its objective. While at a crude level I can envisage that if NII had been persuaded to exclude the 25% MAGNOX element in blended HAL from inclusion within the Oxide storage reduction curve, this would have given some additional head-room to BNFL, it is not for me to speculate why, for more sophisticated reasons, this crude analysis is unsupportable by reference only to things done or not done from and after February 2000. Since it is now unsupported by expert evidence, or other convincing explanation, I need say no more than that I am obliged to reject it.

Conclusion

I have therefore rejected all aspects of the defendants’ case in support of its MAGNOX prioritisation counterclaim. Upon analysis, it discloses no breach of any provision of the Service Agreements. Even if a term specifically prohibiting prioritisation were to be applied, BNFL did nothing which would have amounted to a breach of it. The basis of its approach to the NII 2000 Report was to propose stock reduction curves which, albeit not without some risk, permitted BNFL to continue Oxide reprocessing at THORP in accordance with the then current business plan. I accept that BNFL’s response involved the taking of a risk that, if the WVP performance thereafter fell short of predictions, THORP’s throughput might have to be reduced, but I reject all the submissions that there were reasonable steps which BNFL could have taken prior to the issue of the Specification to maximise the economy of its performance of the contracted services beyond those which it did take, and I also reject the allegation that it failed to use reasonable endeavours in that respect thereafter. When it became necessary to reduce the throughput of THORP, this undoubtedly caused a dis-economy in its performance of the contracted services, attributable to the extension of time thereby rendered necessary to complete the baseload. But there was in my judgment no reasonable alternative by that stage, and the requirement to reduce THORP’s throughput was both a plant requirement, in a sense that the existing backlog of stored MAGNOX fuel could not be left to corrode in storage ponds rather than be reprocessed, and the inevitable response to factors by then beyond BNFL’s reasonable control, namely the restraint constituted by the Specification, within the meaning of clause 18.1.1.

In the circumstances, the case management question whether I should proceed to quantify the delay caused by MAGNOX prioritisation does not arise. I should add that, even if it had, it became common ground that it was not a quantification which I could possibly make at this stage, mainly because, due to the break-down which occurred in 2005, the baseload has not yet been reprocessed even now.

THE ACCOUNTING ISSUE

Construction

This first issue concerns the true construction of, and relationship between, clause 12.2 of, and Appendix 16 to, the Service Agreements. Clause 12.2.1 appears at first sight to define “Costs” as :

“the recorded costs incurred by the Reprocessor in the construction and operation of facilities provided under this Agreement therefor…”

Clause 12.2.2 provides that, in relation to the relevant services:

“charges will comprise:-

1.

64/6000 of the sums corresponding in amount to the COSTS of a facility constructed or used wholly for the purposes of this Agreement and used in common among all BASELOAD CUSTOMERS; or

2.

such other proportion of COSTS as may be appropriate in respect of a facility used partially for the purposes of this Agreement or not used in common among all BASELOAD CUSTOMERS (which proportionate COSTS will be notified by the Reprocessor to the Company on a case by case basis)”

Appendix 16 does not purport to provide exhaustive descriptions of the totality of BNFL’s “normal costing procedures”, but rather a highly compressed summary of them. Even then, it is too long to set out in full in this judgment. I shall therefore summarise its relevant parts. It is headed “Costing and Pricing Procedures”. Its relevant section is part B headed “Costing Procedures”, and this commences as follows: “the Reprocessor operates a process costing system which embodies absorption costing principles”. There is then a description of the main features of that system.

Under Cost Collection, it is stated that costs are analysed subjectively (e.g. wages, salaries and materials) and objectively (e.g repairs and maintenance). All costs are stated to be collected by “cost centres” each of which is a defined area of managerial responsibility. Cost centres include Process centres, Process service centres, Maintenance and supply centres and Overhead centres. Within THORP, the evidence showed that the HEP and SEP were separate Process centres. So were both HALES and the WVP. Overhead centres are described as relating to administration and financial services for the site as a whole. For present purposes, the relevant site is Sellafield.

Two types of costs are allocated to Process centres, namely directly allocated costs and indirectly allocated costs. Overheads are described as allocated only to Process centres, pro rata to direct and indirect manpower.

Under heading “End Levies”, Total Works Process Costs are described as supplemented in respect (among other things) of “Company overheads (ie, proportion of Headquarters Costs)”. Earlier, it is provided that:

“The costs of process service, maintenance and supply and overhead centres are subsequently allocated to process centres so that the total works process centre costs comprise the total expenditure incurred in undertaking the relevant process.”

The particular disputed charges which remain outstanding in this litigation may be summarised as:

The defendants’ case on construction was that the definition of Costs is to be found entirely in clause 12, the reference to Appendix 16 being merely for the purpose of identifying the procedure by which that definition was to be worked out in practice, rather than containing any part of the definition itself. Mr Darling QC submitted that, as a result, the single question in relation to any disputed charge was whether it, or the activity in relation to which the expenditure was incurred, either constituted or had a sufficient proximity with the construction and operation of facilities provided under the Service Agreements. If it did, then an appropriate proportion of it was recoverable from each BLC. If not, it was entirely irrecoverable.

Bearing in mind the express reference to Headquarters Costs in Appendix 16, I invited Mr Darling QC to explain how his test would operate in relation for example to the cost of providing a floral declaration on the main boardroom table at BNFL’s headquarters (which was not at Sellafield). He preferred to answer in relation to the boardroom table itself. He submitted that the necessity test provided the answer. If the construction and operation of THORP required consideration by the main board (as it did) then an appropriate element of the cost of providing a boardroom table was properly recoverable.

Mr Darling QC submitted further that any attempt to have recourse to Appendix 16 as containing the extended definition of Costs would make the apparent definition in clause 12.2.1 redundant. Mr Barnes QC for BNFL submitted that there was no such redundancy, and that the full definition of Costs is be found both in clause 12.2 and in Appendix 16. For practical purposes, he submitted that the best way to understand the application of the definition was to start with Appendix 16, and focus upon the relevant Process centres being, as I have described, first the HEP and the SEP, both being facilities used wholly for the purpose of the Service Agreements within the meaning of 12.2.2.(1), and secondly HALES and the WVP, each of which were Process centres used partially for the purposes of the Agreement, within the meaning of clause 12.2.2 (2).

The operation of BNFL’s “normal costing procedures” would lead to the allocation of Costs (directly and indirectly allocated), Overheads and End Levies to each of those Process centres. Those costs would include Headquarters costs, overheads arising from the administration of the Sellafield site as a whole, maintenance and service costs for each Process centre and the cost of wages, salaries and materials incurred in running those centres.

Out of the total costs allocated in accordance to those procedures to each Process centre, clause 12.2.2 required the whole of the costs allocated to the Process centres within THORP to be distributed among BLCs in accordance with the precise proportions stated in each Service Agreement. A share of the Costs apportioned to HALES and the WVP would fall to be apportioned to the BLCs as a group, and then subdivided in accordance with stated percentages. The apportionment to the BLCs as a group would depend upon the extent to which those shared facilities were used for the performance of the services under the Agreement. In practice, at HALES and WVP they would need to be split between the Oxide service provided under the Service Agreement, and the MAGNOX reprocessing service.

Further, since the activities within those Process centres would at any given moment in time potentially be attributable both to the provision of the service constituted by the reprocessing of the 6,000tU baseload, and to the reprocessing of further (“post baseload”) spent fuel, clause 12.2.1 required a further subdivision, so that only that part of those costs properly proportionate to the construction and operation of those facilities for the purposes of the service agreement (that is the reprocessing of 6,000tU baseload) would fall to be charged to the BLCs. In that way, suggested Mr Barnes, it could be seen that clause 12.2.1, 12.2.2 and Appendix 16 each played a relevant, consistent and harmonious part in identifying, as a matter of contract, the amounts properly chargeable to BLCs.

Standing back for a moment, those two rival submissions are by no means as divergent as, at first sight, they might appear. BNFL’s submission operates in practice as a structured basis for ascertaining, in relation to any particular item of charge, whether it has a sufficient proximity with the construction and operation of the Oxide reprocessing facilities to qualify to be charged to BLCs. Mr Darling’s submission is, in a sense, much simpler, but is inherently likely to produce greater uncertainty, and relegates Appendix 16 to pure procedure, whereas its language seems to me at least as much to be that of summary definition. Furthermore, the use of the words in clause 12.2.1 “shall be determined by the Reprocessor in accordance with its normal costing procedures” strikes me as descriptive both of contractual right (“determined by”), and procedure (“normal costing procedures”).

To the extent that it matters, I prefer Mr Barnes’s submission to that of Mr Darling. In fact, I have tested the particular disputed items against both those submissions, and I consider that, with modest exceptions, they lead to the same conclusions. I consider however that Mr Darling’s necessity test is too strict an interpretation of proximity. The fact that it might be concluded with the benefit of hindsight that a particular piece of expenditure had not been necessary to enable or facilitate the provision of the services under the Agreement seems to me to be altogether too stringent. A company may in good faith reasonably incur costs for a purpose for which they turn out to be unnecessary, but that does not in my judgment lead to their exclusion, in particular in a cost-plus contract where costs are allocated in accordance with the absorption costing principle. The necessity test is one which, had it been agreed, would in my judgment have been stated expressly.

The Disputed Items

The West Cumbria Initiative

As I have said, this was a public benefit exercise carried out by BNFL with the intention of maintaining and improving its standing in the community around Sellafield. The sensitive nature of nuclear activities, in particular those at Sellafield, in the collective mind of the public, and in particular those living in the vicinity of the plant, is a matter of which judicial notice can be taken. It seems to me eminently reasonable for BNFL to have taken steps to maintain and enhance its standing among the local community by measures such as the West Cumbria Initiative. The cost of such a measure seems to me to fall fairly and squarely within overheads associated with the administration of the site as a whole. The uninterrupted pursuit of all the potentially hazardous activities at the Sellafield site, including in particular the Oxide fuel reprocessing service, seems to be likely to be capable of being preserved and enhanced by such an initiative, and on Mr Darling’s proximity test it seems to me also sufficiently proximate to the provision of the Oxide processing service by the operation of the THORP and ancillary plant. I therefore reject the defendants’ challenge to its recoverability.

Headquarters costs

The items in dispute may best be summarised as grouped under two main headings:

As for the first, the evidence satisfies me that these items were incurred by BNFL, mainly at headquarters level, in anticipation of, or by reason of, the Energy Act 2004, pursuant to which assets and liabilities associated with specified nuclear activities (including the whole of BNFL’s assets and liabilities and therefore, of course, Sellafield and THORP) were to be transferred to a new body, following which it was anticipated that BNFL would be appointed as the manager, at least of the Sellafield facilities, by that body.

This process plainly necessitated both a transformation of BNFL from a business which owned its operating plant to a management business, and the company reorganisation also involved, for example, separating out BNFL’s reprocessing activities from its MAGNOX generation activities by the creation of two levels of parent company and a range of subsidiaries of those parents. In the context of BNFL’s continued performance of, and its obligations under, the Service Agreements, I can see no reason why the cost of that transformation and reorganisation should not form part of the Headquarters costs liable to be apportioned as End Levies to Process centres throughout BNFL, including the four relevant Process centres which I have identified. The reorganisation and transformation was no less a proximate activity than the acquisition and maintenance of a boardroom table, so that in my judgment those items are in principle chargeable regardless of the choice between the parties’ differing approaches to the construction of the Service Agreements.

The same goes for the other miscellaneous Headquarters costs, save that the defendants have been able to point to certain very modest items, for example within the general headings “Reorganisation Project” and “Consultancy Legal Fess with Group Finance Costs” which appear to be attributable to specific projects wholly unconnected with THORP, such as a project for the sale of BNFL’s uranium enrichment business, and the repatriation of employees from the USA. To that extent, those activities may be said not to satisfy any test of proximity with the construction and operation of the Oxide reprocessing facilities at Sellafield.

Mr Barnes submitted in response that the operation of an absorption costing system has its inevitable swings and roundabouts. For example, the activities of an individual employed at Headquarters during a particular year may well include, from time to time, activities focused exclusively on one and then on another of the company’s different business projects. Rather than carve up that employee’s salary into small chunks, for the purposes of allocating each chunk to the Process centre in relation to which that person’s activities were focused at a particular time, his salary is apportioned to all the company’s Process centres pro rata their separate head count. The effect of taking that approach to substantially the whole of the company’s Headquarters activities means that, over time, the benefits and burdens of it are self cancelling, quoad the person enjoying and paying for the activities of any particular Process centre. Furthermore, a substantial part of Headquarters time in any particular year is likely to have been taken up directly with consideration of Oxide reprocessing, but the costs attributable to that activity will all have been apportioned pro rata, so that BLCs will have to pay only a part of it.

There is in my judgment a real force in that submission. That appears to have been, as a matter of fact, the normal procedure by which BNFL has allocated Headquarters costs to Process centres as an End Levy. Since End Levies appear to me, in relation to the activities of any particular Process centre, to be part of overheads, at least in substance if not form, it seems to me consistent with the provisions in Appendix 16 to the effect that overheads are allocated to Process centres pro rata to direct and indirect manpower.

I can envisage that a situation might emerge where the activities and expenditure at Head Office become focused to such a substantial extent upon the activities of a particular Process centre that a pro rata manpower apportionment may become unreasonable. But the evidence comes nowhere near suggesting that this has occurred, and in those circumstances I am not persuaded that any of the modest charges so identified ought to be disallowed.

The COGEMA Contract

I have already referred to this contract under the MAGNOX prioritisation issue. It was entered into with COGEMA in 2004 and called for the making of a substantial upfront payment by BNFL. Its purpose was, in due course, to improve the operation of the WVP, and it was entered into at a time when BNFL probably still thought it unlikely that the fruits of the contract would yield practical benefits in terms of the vitrification of HAL derived from the 6,000tU or 7,000tU THORP baseload.

The two year stoppage of THORP which began in 2005 means that it is not now possible to conclude with any confidence that the benefits (if any) derived from the COGEMA contract will not be utilised in the vitrification of HAL derived from that baseload, even if, subject to any appeal, the effect of the rescission of the Side Letters means that the baseload in question is 6,000tU rather than 7,000tU.

Since it is impossible therefore to say that the whole of the benefit to be achieved from the COGEMA contract will apply only to the vitrification of post baseload generated HAL, it is equally impossible to say that no part of the amount payable by BNFL under that contract can properly be attributable to BLCs. The amount attributable will depend upon a proper apportionment of the benefits of the COGEMA contract between baseload and post-baseload fuel.

It is common ground that this last item of challenge includes both a liability and a quantum element. As to liability, it is impossible for me to conclude, as the defendants claim, that no part of those charges are recoverable from them. Equally, until the whole of the 6,000tU baseload has been reprocessed and vitrified, which may be many years hence, in view of the uncertainty when THORP will re-start, and the possibility of relatively long term storage of the relevant HAL at HALES, it would be impossible for me to address the quantum issue at this stage, as the parties accept.

Extrapolation

The final question which I must decide is whether as a result of my determination of the specific accounting issues, and of the issues as to the construction of this accounting provision which I have described, I should direct or permit any wider enquiry into the legitimacy of BNFL’s invoices to date, than is the subject of the specific pleaded objections.

The parties were agreed that this question would be likely to depend upon the extent to which my analysis of the specific issues disclosed grounds for concern that the invoices to date had been prepared on false basis, to a greater extent than had been identified (or than was reasonably capable of being identified) by PWC in its annual audit. The specifically pleaded objections represent the best surviving of the “bullets” provided by PWC to the BLCs for firing at BNFL in connection in any dispute about invoices. Save for the quantum issue in relation to the COGEMA contract, they have all, as I have held, missed their target, and I have concluded that, to the extent that it matters, BNFL’s approach to the construction of the relevant provisions of the Service Agreements is to be preferred to that proposed by the defendants.

In those circumstances, it would seem to me quite wrong to impose upon BNFL an obligation further to justify its invoices beyond litigating, successfully as I have held, the accounting issues already raised in these proceedings. In relation to present invoices, it is in my judgment time to draw a firm line against the litigation of any further objections. Of course, that leaves outstanding the question of quantum in relation to the COGEMA contract, which is in any event by far the largest item in dispute, in terms of its financial consequences.

I shall therefore hear submissions as to the case management of the questions which remain outstanding.