Fall, 1986, p. 161-175
In 1974, India became the first and only country in the world to explode an atomic bomb made from materials imported for peaceful nuclear purposes. India made the bomb with plutonium extracted from spent reactor fuel. Canada supplied the reactor and the United States provided the heavy water needed to run the reactor. India had promised to use the reactor and the heavy water for peaceful purposes only; thus it insisted on calling its bomb a peaceful nuclear device.
Heavy water was essential to the Indian bomb then; it is just as essential now. It is an ingredient that most of the country’s reactors—and all of its newest reactors—need to operate. India’s chances of becoming a genuine nuclear weapons state and of operating a successful nuclear power program depend upon an adequate and steady supply of heavy water.
A close study of India’s heavy water inventories reveals a large gap between this supply and India’s demand. This gap leads to one of two conclusions. Either India is illegally diverting heavy water from international safe-guards, or India has received a secret import—probably from China. New Delhi could also be pursuing both courses. India is using this illegally or secretly acquired water to run three new reactors outside international inspection.
These three new nuclear reactors will in-crease dramatically India’s nuclear-weaponsbuilding capability. Up to now, India has accumulated only small amounts of plutonium—enough for between 5 and 10 bombs but this material is restricted to peaceful use by Indian pledges to nuclear supplier countries. The new reactors, however, may pro-duce enough plutonium for 15 bombs per year, and the plutonium will not be restricted to peaceful use.
India has either secretly imported heavy water from China or diverted it illegally from international inspection to run its new nuclear facilities.
Both India and China have denied the contention that India covertly imported unsafeguarded heavy water from China, and India has denied any diversion from safe-guards. Yet despite invitations to do so, neither country has provided any information to support its denial. Neither country has signed the 1968 Treaty on the Non-Proliferation of Nuclear Weapons or agreed to adhere to the export control guidelines of the Nuclear Sup-pliers Group. Thus neither was breaking any international obligations by secretly trafficking in heavy water. Both, however, are members of the International Atomic Energy Agency (IAEA). The agency’s central purpose is the safeguarding of all important nuclear exports. China, moreover, has now specifically pledged to the United States that it will not use its nuclear exports to help other countries develop nuclear weapons. Both India and China will continue to need nuclear imports for their civilian programs, and both wish to import other high technology from the West. Until these large discrepancies in India’s heavy water balances are explained, nuclear supplier states should reconsider not only nuclear trade with India and China, but also any sensitive high-technology commerce with these countries.
India has manufactured its own heavy water since 1962, and though it can do what it pleases with the domestic product, India has never come close to meeting its needs. The country has been forced to import heavy water and to promise that all plutonium made by reactors using this water will be restricted to peaceful purposes.
Heavy water, scientifically known as deuterium oxide, is found in minute concentrations in ordinary water. To produce it, one must separate it physically from ordinary water in an expensive process that uses large amounts of energy. Its main advantage is that it allows a country to run reactors on natural uranium, which is widely available, rather than .on enriched uranium fuel, and thereby avoids the peaceful-use and safeguard restrictions that suppliers of enriched uranium require.
India’s disturbing shortages exist because domestic production has been very low, need has been high, and publicly acknowledged imports have been inadequate to make up the difference. In fact, an examination of India’s production, need, and imports reveals a gap of 68 metric tons (t) of unsafeguarded heavy water in 1983 and 293t in 1985. India in effect has been running reactors on water it does not admit having.
Domestic Production. India’s domestic production is the first factor in the heavy water equation. India produces heavy water at four principal facilities: Nangal, Baroda, Tuticorin, and Kota. Minute quantities may have been produced at a fifth plant called Talcher. Overwhelming operating problems and de-sign flaws have prevented these plants from producing more than a small part of their capacities. In testimony to the Indian Parliament in spring 1986, Shri Srinivasan, chief executive of Heavy Water Projects of India’s Department of Atomic Energy, reported that all domestic heavy water facilities have operated far below capacity and expectations. Despite a total annual production capacity of more than 300 metric tons (a unit of weight roughly 10 per cent more than a U.S. short ton of 2,000 pounds), the public record reveals that India never produced more than 50t of heavy water annually until fiscal year 1984–85, the last year for which figures are available. (India’s fiscal year runs from April to April.)
The Nangal plant has been India’s best producer. The Times of India science reporter Praful Bidwai reported on May 9, 1984, that Nangal has steadily produced between 10t and 12t of heavy water annually. However, according to the February 18, 1982, issue of Nucleonics Week, Nangal produced only 9t in fiscal year 1979—80. Production must have been at about this rate in fiscal year 1980—81 as well, because the plant was closed for at least 3 months.
A wide range of authoritative sources con-firms low output for the three other plants. The biweekly Nuclear Fuel, on June 3, 1985, quoted Raja Ramanna, the chairman of the Indian atomic energy commission, as saying that a combined production of 80t at the Baroda and Tuticorin plants would be “approaching the maximum feasible capacity.” This issue of Nuclear Fuel also cites the Indian government as acknowledging production of 7t at Tuticorin in 1979. Bidwai pegs Tuticorin’s annual production at 14t and 15t in fiscal years 1980—81 and 1981—82, respectively. For fiscal year 1983—84 he reports production at about 30 per cent of capacity, which would yield roughly 23t. In January 1983, the Hindu, a Madras daily, claimed that total production at Tuticorin for 1980 through mid-1982 was even lower. And Chemical Weekly’s August 9, 1983, issue cited a fiscal year 1982—83 figure of only 4t.
Baroda began producing heavy water in 1981, and the private Nuclear Assurance Corporation, which collects production data from nuclear facility operators around the world, reported an output of 12t for that year. Chemical Weekly, in the aforementioned issue, reported Baroda’s fiscal year 1982—83 output at 5t, and the Times of India, on May 8, 1984, reported a 13.6t figure for fiscal year 1983-84.
Precise production figures for fiscal year 1984—85 are not available for either Tuticorin or Baroda. Those in Table 1 are generous estimates that double production at both facilities from the previous year, to a total of 74t in both 1984 and 1985. As previously noted, Ramanna has stated that such levels for both plants would be “approaching the maximum feasible capacity.” Finally, Bidwai reports heavy water production of 5.2t at Kota in fiscal year 1983–84, according to the Times of India of May 7, 1984.
Heavy Water Demand. The second factor in the heavy water equation is demand. India’s demand for heavy water can be documented just as precisely as its production. The country currently operates two research reactors, Cirus and Dhruva, two nuclear power plants in the northwestern desert province of Rajasthan (RAPP-I and RAPP-II), and two near the southern port of Madras (MAPP-I and MAPP-II). All require heavy water, but only the two RAPP reactors are subject to IAEA inspection.
According to official U.S. export records, the Cirus research reactor required 19t of heavy water when it started up in 1960. Its annual losses are probably negligible: The Cirus design anticipates a loss rate of only .3 per cent annually, or t during its first 20 years of operation.
RAPP-I started up in 1972 with an inventory of 216t of heavy water. Following pressurization in 1972, the reactor lost 11.4t of heavy water, raising the facility’s total need in that year to more than 227t. David Hart, then of Imperial College of London, reported in the 1983 book Nuclear Power in India: A Comparative Analysis that the heavy water losses at RAPP-I between 1977 and 1981 totaled 100t. The figures in Table 1 are based on the assumption that no losses took place in 1982 or 1983, when the reactor was shut down. The figures are. also based on the assumption that RAPP-I lost 5t of heavy water in fiscal year 1984–85, when the reactor operated for roughly one-fourth of the year, and 6t in fiscal year 1985–86, when the reactor operated for roughly 4 months. RAPP-II’s beginning inventory was probably the same as RAPP-I’s, but the figures in the table are based on the assumption of smaller loss rates—16t annually—because of likely improvements in design.
MAPP-I’s requirement of 250t of heavy water for its 1983 start-up comes from statements by Indian Minister of State for Science and Technology C. P. N. Singh that appeared in the Hindu on July 29, 1982. He estimated losses of 10t–15t of heavy water annually for the plant.
MAPP-II, which is identical to MAPP-I, also required 250t of heavy water when it began operating in 1985.
Dhruva, India’s newest research reactor, required 78t of heavy water for start-up in 1985, reported Nuclear Europe in September 1985. And its annual loss rates are assumed to be similar to those of Cirus. Although Dhruva has not operated for most of 1986, its heavy water was still required for start-up at a time when MAPP-II was already operating.
Heavy Water Imports. Imports are the final factor in India’s heavy water equation. India’s first import came from the United States, which supplied 19t in 1960 to start up Cirus. To start RAPP-I in 1972, India imported 120t of U.S. water through Canada and 80t from the Soviet Union. It is not clear whether the 80t sent by the Kremlin were provided subject to safeguards, but for the purposes of the table they are considered unsafeguarded to give India the benefit of the doubt. These 80t may be subject to a peaceful-use guarantee imposed retroactively when Moscow began to supply the heavy water for RAPP-II in 1976. To keep RAPP-I running and to start RAPP-II, the Soviets agreed to provide 456t from 1976 to 1985. The Soviet import figures are reported by William C. Potter of the Rand/UCLA Center for the Study of Soviet International Behavior. These figures were confirmed by the late Indian Prime Minister Indira Gandhi in March 1983, when she told the Indian Parliament that the country’s heavy water imports had reached 547.6t. This figure includes the 19t from the United States in 1960, the 120t from the United States through Canada in 1962, the 80t from the Soviets in 1972, the 200t from Moscow between 1976 and 1979, and the roughly 12t from the Soviet Union between 1980 and 1982. The remaining 135t were probably imported from the Soviet Union between 1983 and 1985.
Uncertainties. These are the factors in the equation. They show that in 1985 India faced enormous heavy water shortfalls—157t in its overall supply and 293t in its unsafeguarded supply. Can the shortfalls be explained by errors in the table‘s estimates? How far off can the estimates be? First, the actual start-up inventories of MAPP-I and MAPP-II could each have been 240t instead of 250t. The number 240 has been mentioned occasionally. This figure would reduce the total need by 20t. Second, the loss rates could be slightly lower than those shown in the table. How-ever, the only loss figures not based upon Indian government or other actual published reports are the estimated annual 5t and 6t losses in fiscal year 1984-85 and fiscal year 1985-86 for RAPP-I and the estimated annual losses for RAPP-II and MAPP-I. The sum of these loss estimates is 80t. If, by extraordinary good fortune, India had reduced these losses by one-half—unlikely in view of the record—the figures in the table would overestimate the actual loss by 40t. Thus the maximum credible error in the table is roughly 60t, consisting of a possible overestimation of need by 20t and a possible overestimation of loss by 40t. This error cannot explain the 157t and 293t short-ages in 1985. Moreover, an overestimation of the heavy water loss by 40t would overestimate the unsafeguarded loss by only 8t, since the heavy water inventory from which the loss occurred was about 80 per cent safeguarded and 20 per cent unsafeguarded.
How Did India Do It?
India’s supply of heavy water kept up with its need until 1983. But in that year India decided to start MAPP-I without safeguards. Consequently, MAPP-I’s output of plutonium would not be monitored by international inspection and would be available for atomic bombs. This decision also required India to supply MAPP-I’s heavy water itself, presumably from unsafeguarded domestic production.
Yet India’s domestic production and re-serves in 1983 were insufficient to start MAPPI and still run the other operating reactors. In fact, India was 68t short. Its reserve—the heavy water remaining after meeting the needs of the on-line reactors—did contain 73t of heavy water. But that water was safeguarded and could not legally be put into unsafe-guarded MAPP-L So in 1983, India began for the first time to run more reactors than its supply of heavy water appeared to make legally possible.
In August 1985, India started MAPP-II and Dhruva without safeguards. By now, the country’s total public heavy water shortage stood at 157t. India was running four power reactors with only enough heavy water in the public records to run three. And because MAPP-II and Dhruva were started without safeguards, none of their heavy water could legally come from safeguarded imports. Therefore, the 136t of safeguarded reserve in 1985 were not legally available to MAPP-II and Dhruva, and the public shortage of unsafe-guarded heavy water was more than 290t—about one and one-half reactors’ worth.
These shortages mean that India cannot be running its nuclear power program honestly. The data show that India must be increasing its total supply of heavy water by some secret means, and must also be increasing its supply of unsafeguarded heavy water, probably by the same means. There are only two ways to do this: transfer heavy water from another Indian reactor or import the water secretly.
To bring MAPP-I to full power at the end of 1983, India needed 68t of unsafeguarded heavy water. Heavy water reactors need a full inventory of heavy water to operate, so heavy water could be shifted from another reactor only if that reactor were shut down. RAPP-II was operating in 1983, but RAPP-I was not.
RAPP-I had operated until March 1982, when it was shut down because of leaks. It did not start operating again until January 1985. RAPP-I, therefore, could have been drained to start MAPP-I in 1983. When RAPP-I closed, there was reactor-grade heavy water in its cooling and moderating systems and “degraded” heavy water collected from its leaks. Heavy water is called degraded when it leaks out of a reactor and mixes with ordinary water from the air. All of that heavy water was sitting idle and could have been shifted to MAPP-I. The portion that had leaked out could have been shifted after a process of upgrading, which means reconcentrating the heavy water by removing the ordinary water mixed with it, or the portion from the reactor could have been shifted directly at reactor grade.
India has done a great deal of upgrading. According to Hart, in Nuclear Power in India, from 1974 to 1976 the Indian government collected and upgraded at RAPP-I 60t to 70t of degraded heavy water each year. Bidwai re-ported in the Times of India on July 27, 1983, that India transferred about 100t of degraded heavy water from Rajasthan (RAPP-I) to Madras (MAPP-I) during the 2 years before MAPPI’s start-up. And according to Indira Gandhi, MAPP-I was started up by using 140t of upgraded heavy water.
These statements all show that at least part of MAPP-I’s start-up inventory came from RAPP-I and that the shift occurred after upgrading. It is unclear how much heavy water was involved. If Gandhi’s figure of 140t is accurate, India would have had to remove about 60 per cent of RAPP-I’s inventory and all of RAPP-I’s unsafeguarded heavy water.
The amount of unsafeguarded heavy water in RAPP-I at any particular time depends upon how losses are replenished. Fifty-five per cent of RAPP-I’s original 216t inventory consisted of 120t of safeguarded water exported from the United States through Canada. Therefore, it is logical to assume that 55 per cent of RAPP-I’s yearly loss would be of safeguarded heavy water. As the table shows, RAPP-I lost a total of 176t during the 9 years from 1973 to 1982. Fifty-five per cent of this quantity equals about 100t. This 100t loss reduced the 120t under safeguards in the original inventory to 20t. Yet from 1980 through 1982, India added 46t of safeguarded Soviet heavy water to RAPP-I to make up for annual losses. This addition increased the safeguarded inventory to a total of 66t. Thus there were 66t of safeguarded heavy water and 150t of unsafeguarded heavy water in RAPP-I when it was shut down in 1982. India had to use the rest of the safeguarded Soviet imports received through 1982 in RAPP-II, the only other safeguarded reactor. Thus all the heavy water in RAPP-II is safeguarded.
The alternative to upgrading was simply to shift reactor-grade heavy water directly from RAPP-I to MAPP-I. Based on the previously cited requirement for MAPP-I, previously cited loss estimates for RAPP-II, and India’s domestic production, India needed 68t of unsafeguarded heavy water to start up MAPPI legally in 1983. RAPP-I contained 150t of unsafeguarded water, making that transfer both possible and legal.
Whether India shifted heavy water from RAPP-I to MAPP-I after upgrading, or transferred it directly, RAPP-I’s heavy water inventory would have been cut. If India had shifted only the 68t of unsafeguarded heavy water that MAPP-I needed, RAPP-I would have been left about 70 per cent full. If India had shifted 140t, as Gandhi said, the effect would have been to leave RAPP-I 65 per cent empty—unlikely in light of the fact that RAPP-I began operating again in January 1985.
At the beginning of 1985, India’s reserve surplus stood at some 103t. With this quantity India could have started MAPP-I in 1983 by draining RAPP-I, and operated MAPP-I through 1984 with the RAPP-I water. Then, in 1984, India could have shipped other water—produced in the interim—back to RAPP-I and restarted it with a full heavy water inventory in January 1985. This means that RAPP-I could have been the sole source of the extra heavy water needed to start MAPP-I in 1983.
In 1985, India started Dhruva and MAPP-II without safeguards. Its apparent shortage of heavy water was vast—some 157t. Its apparent unsafeguarded shortage was even larger — some 293t. Where did the heavy water to start Dhruva and MAPP-II come from? Could RAPP-I have been drained again?
After RAPP-I started up in January 1985, it stopped operating in early summer and remained shut down through the end of 1985. Therefore, its water was in theory available be drained out again—between its shut-down in early summer and the time Dhruva and MAPP-II went critical in August. Yet this means that India planned to bring to criticality two reactors for which it had no water until shortly before they were due to go critical; produced that water by taking off the country’s power grid another reactor that had just started up in another part of India; and transported large amounts of heavy water almost overnight. This scenario seems far-fetched.
In fact, MAPP-II and Dhruva could not have started solely on heavy water from RAPP-I. First, RAPP-I did not have enough heavy water. After RAPP-I shut down in early summer 1985, the residual decay heat from its core still had to be removed continuously by the heavy water in its cooling system. According to the reactor’s specifications, that system required about 70t. The rest of the inventory, about 145t, acted only to moderate the reaction and would have been available for transfer. But these 145t are still some 150t shy of covering India’s 293t shortage.
Second, by 1985, only 140t of RAPP-I’s heavy water were unsafeguarded and legally available for MAPP-II and Dhruva. This 140t is calculated by starting with the 66t of safeguarded heavy water RAPP-I contained when it shut down in 1983, and adding 11t of safeguarded make-up water for its operation during calendar year 1985. This makes a total of 77t under safeguards. The balance of the 216t inventory, which is roughly 140t, was unsafeguarded. Using those 140t would also still leave India some 150t shy of covering its 293t shortage. Thus there is no way India could have covered the shortage in 1985 solely with water from RAPP-I. India was 150t short even after draining all heavy water out of RAPP-I that was legally and physically available.
If RAPP-I’s water was insufficient, what were the other possible sources? RAPP-II, MAPP-II, and Cirus were all running in 1985, making their heavy water unavailable. The only heavy water left in India in 1985 was the 13t of safeguarded reserve. India either diverted it illegally or got a secret import.
Enter the mysterious shipment from Bombay. Bidwai reported in the Times of India on July 27, 1983, that 100t of reactor-grade heavy water were shipped from Bombay to MAPP-I several months before MAPP-I started up in 1983. Bidwai stated that only 70t of reactor grade water were available at MAPP-I in September 1982, in addition to 70t in accumulated domestic production in spring 1983—a total of 140t. This leaves a shortage of about 100t to start MAPP-I in summer 1983. These estimates are quite close to the amounts shown in the table. That total represents all the heavy-water in India, including all the water from upgrading, all the domestic production, and all the previous imports—which at the time were keeping reactors running. Nothing more could be in the system. Then, suddenly, 100t of absolutely pure reactor-grade water reportedly appeared in Bombay from unknown sources. And it arrived in the nick of time to start MAPP-I.
One must conclude that by August 1985 India had either diverted heavy water illegally from safeguards or received a clandestine import. No other explanation suffices. RAPP-i, still possible as the source of MAPP-I’s water in 1983, could not have provided heavy water for RAPP-II and Dhruva in 1985. Even had India used all of RAPP-I’s 140t, it would still have been short 150t of unsafeguarded heavy water. The mysterious shipment from Bombay shows that India probably received an import sometime before mid-1983—in time to use it in MAPP-I.
China is the only country in the world that could legally export this quantity of heavy water without safeguards. The world’s only other significant manufacturers of heavy water are Canada, the Soviet Union, and the United States. Canada cut off nuclear trade with India in response to India’s peaceful nuclear device in 1974. Canada’s decision was public, firm, and has been rigidly adhered to.
The Soviets are bound by the nonproliferation treaty and by their membership in the Nuclear Suppliers Group not to export heavy water without safeguards, and they have adhered strictly to those obligations. Indeed, the Soviet requirements on the heavy water for RAPP-II were the strictest ever imposed on heavy water—stricter than Canada’s requirements on the water for RAPP-I. Moscow presented New Delhi with a set of strict controls and demanded that India accept them. The Soviets have had one of the best export records of any major supplier. They have never been willing to sacrifice nonproliferation goals to gain a political advantage. In fact, the field of nonproliferation is the only one in which the superpowers have maintained a united front against other countries. Therefore, China is the only heavy water source remaining.
Yet China maintains close ties with India’s near-nuclear neighbor and rival, Pakistan. And China and India have long been regional rivals. Would China help a powerful neighbor build a nuclear arsenal?
That is precisely what may have happened. China needs foreign exchange. Beijing was forced recently to scale back its nuclear pro-gram for lack of it. And it has been willing to make just about any nuclear deal to get foreign exchange. China has even offered to take in, for a price, the high-level nuclear waste of other countries. Press reports and congressional statements based on U.S. intelligence information have stated that China has negotiated the transfer of sensitive nuclear technology to Iran, helped Pakistan operate its unsafe-guarded uranium enrichment plant, conducted a nuclear test in the presence of a high-level Pakistani official, provided nuclear weapons design data to Pakistan, sold enriched uranium without safeguards to South Africa, and provided heavy water to Argentina without safe-guards in 1985 and from 1980 to 1982.
Given China’s record and heavy water capabilities, India is obliged to provide some information on the source of the mysterious 293t of heavy water. To date, India has simply said that its heavy water production is adequate, and has refused to supply any figures to back this statement up.
India’s conduct has grave implications for the spread of nuclear weapons. Its unexplained shortages mean either that export controls are being ignored by an important supplier, which appears to be China, or that the IAEA cannot safeguard heavy water. Worse, India could be both circumventing safeguards and importing from China. Until the matter is cleared up, both China and the IAEA safeguards system are suspect.
The remedy—and there must be a remedy if controls mean anything—is to halt nuclear trade with India as long as the public short-ages of heavy water remain. Moscow should not provide New Delhi with any more heavy water. The United States should not sell India anything with a possible nuclear application, such as the supercomputer now being considered. These actions are the minimum necessary. In particular, a Soviet cutoff of heavy water probably would prevent India from operating one or more of the new reactors it is now constructing.
These remedies, however, could all be defeated by more nuclear shipments from China. Washington insists that in mid-1984, China pledged to stop unsafeguarded exports. Beijing must be encouraged in the strongest way possible to comply. China now wants to import nuclear reactors and other high-technology products, and to open its economy to the West. Rather than leaping into the new market, suppliers must first insist that China make internationally binding promises to change its export behavior.
If the United States and the Soviet Union take action, India’s program can be pulled back in the direction of accountability. If the suppliers as a group act against China, China can be pulled in the direction of a responsible export policy. If no one takes any action, India can continue officially to thumb its nose at the world, and the suppliers will have once again failed to get tough on nuclear arms proliferation.
Gary Milhollin is a professor of law at the University of Wisconsin. A consultant for the Nuclear Regulatory Commission from 1976 to 1986, he is currently working on a study of U.S. plutonium policy. The author acknowledges the helpful comments of Professor Frank von Hippel of Princeton University on a previous version of this paper.
 The Patriot (New Delhi), 28 December 1981, 1,7.
 “Commissioning of Vitrification Facility Affirms India’s Self-Sufficiency Claims,” Nuclear Fuel, 3 June 1985, 10.
 “State Says ‘Substitute Clause’ Gets India off of Heavy Water Hook,” Nucleonics Week, 1 July 1976, 6-7.
 William C. Potter, “Soviet Nuclear Export Policy,” in Limiting Nuclear Proliferation, ed. Jed C. Snyder and Samuel F. Wells, Jr. (Cambridge, Mass.: Ballinger Publishing Company, 1985), 213-252.
 Unclassified telegram 13373, from the U.S. embassy, New Delhi, to the secretary of state, July 1982.