Chapter 246: The Cold Threshold
Indian Semiconductor Manufacturing Corporation — Advanced Research DivisionGorakhpur Industrial Complex
October 15, 1976
The anomaly had been sitting in Vasudha Krishnan's data for seven months.
She had not hidden it. She had not suppressed it. She had written it in the lab log in October 1975 with the careful, neutral language that twelve years of materials research had trained into her — anomalous resistance reduction observed at approximately 130K; cause unclear; further measurement warranted — and had then continued with the work that was the reason she had been running that particular experiment, which was the characterisation of heavily doped gallium arsenide variants for high-speed device applications. The anomaly was a footnote in the context of the device application work. It was an interesting footnote. It was an unexplained footnote. But the device application work was on a schedule, and the footnote was not part of that schedule, and so the footnote had sat in the lab log for seven months being periodically looked at and periodically not yet explained.
By October 1976, it had been looked at six times by three different members of the ISMC Advanced Research Division team, and each time it had been looked at, the person looking had concluded the same thing: the anomaly was real, its explanation was not obvious, and it fell outside the scope of the current programme's measurement capability.
The measurement capability limitation was the thing that had kept the anomaly from being investigated properly. To understand what was happening to the resistance at 130 Kelvin, you needed measurements at temperatures below 77 Kelvin — below liquid nitrogen temperature, where the current facility's cryogenic equipment stopped working — and you needed measurement resolution below what the current electronics provided. Both were achievable. Neither had been prioritised, because the current programme's requirements did not extend to those temperatures and that resolution.
This was the technical context in which Karan arrived at the ISMC Electronics Research Building at seven-forty on the morning of October 15th, forty minutes before the meeting he had scheduled, because Karan arrived early at technical meetings for the same reason he arrived early at everything: the fifty minutes before a scheduled meeting were the most useful fifty minutes of any meeting he attended.
The building was a converted warehouse. This was not an aesthetic statement.
The conversion had been done in 1971, when the warehouse's twelve-metre ceiling height was the correct height for the semiconductor deposition chambers that the team needed, and when constructing a purpose-built laboratory to those specifications in the available time would have cost four times as much and taken twice as long. The warehouse had been lined with clean-room panels, fitted with humidity and temperature control systems that ran continuously, and had accumulated over five years of investment the specific quality of a research facility that had been built by people who understood what research required — not what it looked like, but what it actually required.
From the outside, the building looked like a warehouse that had been fitted with some ventilation equipment and air conditioning units.
This was intentional.
Karan walked through the observation corridor of the main clean room. A deposition run was in progress — three technicians in full cleanroom suits managing a process that had been running for approximately four hours, the substrate rotating slowly inside the chamber, the deposition source visible as a diffuse glow. He stopped at the observation window and read the run log on the clipboard.
The duty technician was a young man named Sunil who had been with the process team for fourteen months and who recognized Karan from his previous visits and who therefore answered the questions that followed without either the panic of the first-timer or the performance of the long-accustomed. He answered them accurately and quickly.
The third question Karan asked was about the uniformity variance in the northern quadrant of the substrate, which had been appearing consistently across six runs.
Sunil said: "Candidate hypothesis is a minor misalignment in the source geometry relative to the substrate rotation axis. We've identified the adjustment. Thursday maintenance window."
"What's the adjustment?"
Sunil described it. It was correct. The literature said the same thing. Karan returned the clipboard and moved on.
He found Trivedi coming from the direction of the offices.
Trivedi was fifty-two, from Varanasi, and had spent twenty years in semiconductor research before his three years at Bell Laboratories in the mid-1960s, where he had worked on silicon transistor processing at the specific moment when that was one of the most consequential things happening in the world. He had returned to India in 1968 not because the opportunity at Bell was finished but because the work he had gone to do there — bringing back what Bell Laboratories knew — was finished, and what he had brought back needed to be deployed. He had deployed it. The ISMC Advanced Research Division was the deployment.
"You're early," Trivedi said.
"The run looked interesting from the corridor," Karan said.
"Conference Room B," Trivedi said. "Nine o'clock. We have coffee and we have something you need to hear."
Conference Room B held, at nine o'clock, eight people who represented the most concentrated assembly of semiconductor physics expertise in India and, by Trivedi's private assessment, the most capable basic research team anywhere in Asia outside Japan.
Dr. Anand Kumar Sharma — forty-six, device physicist, done his doctorate at IIT Kharagpur on metal-semiconductor contacts, spent four years at IBM Yorktown Heights before returning in 1971 — was the man who had first noticed the anomaly in the data that Vasudha had generated, had mentioned it to Trivedi, and had suggested that the ISMC electronics team and not the ISMC processes team was the right group to pursue it. He had the compact, focused manner of a physicist who had learned at IBM that the difference between an interesting observation and a productive research direction was the precision of the measurement.
Dr. Vasudha Krishnan — forty-four, materials physicist, twelve years at ISMC building the Materials Library — had generated the anomaly. She was aware that the anomaly was the most interesting thing in her data. She was also aware that twelve years of careful materials characterisation work formed the context without which the anomaly could not be understood, and she had the slightly compressed patience of a researcher who has been generating data of extraordinary quality for a long time and is waiting for the moment when that data's most important content becomes visible.
Dr. Rajesh Kumar Banerjee — forty-one, device physicist, Cambridge doctorate, two years at IBM Zurich — worked on device structures, specifically the geometry of semiconductor junctions and their electrical characteristics. He had the Cambridge-trained habit of thinking in physical principles before thinking in engineering solutions, which made him slow and frequently made him right.
Dr. Sundaramurthy Iyer — thirty-eight, electrical engineer, MIT doctorate on thin-film transistor fabrication — was the man who could take research directions and materials data and device physics and translate all three into a process sequence that could actually be executed in the clean room. Without him, the theoretical and materials work produced knowledge. With him, the knowledge became capability.
Dr. Pradeep Nair — forty-three, surface physicist, IISc doctorate, six years at ISMC — had spent his career studying what happened at boundaries between materials. He thought about the world primarily in terms of interfaces, which was the kind of thinking that frequently produced insights that bulk-focused researchers missed.
Dr. Mohammed Akhtar Hussain — forty-seven, from Lucknow, measurement specialist — had spent fifteen years developing characterisation techniques. His measurement capabilities at ISMC were used daily by every other member of the team, and periodically stared at by visiting researchers from foreign institutions who had not expected to find that level of precision in India.
Dr. Krishanu Chakraborty — thirty-three, theoretical physicist, Jadavpur University doctorate — worked on theoretical modelling of semiconductor properties. He had the specific quality of theorists who had committed to a research programme without yet knowing where the most important problem in that programme was, and who were therefore perpetually attentive to anomalies.
Karan sat at the end of the conference table. Trivedi sat to his right.
"I want to start with a question," Karan said. "Not a presentation. What is the most important problem in semiconductor physics that this team is not currently working on?"
This produced a specific silence. Not the silence of people without answers — the silence of eight people with different answers calculating which to give.
Trivedi spoke first. "Room-temperature operation of high-electron-mobility compound semiconductor devices. We have preliminary work on gallium arsenide that shows mobility characteristics substantially better than silicon. The application development for what you do with that mobility advantage is not yet a programme."
"Noted," Karan said. He looked at Vasudha Krishnan.
She said: "The systematic study of semiconductor properties at very low temperatures. I have characterisation data on germanium and silicon down to liquid nitrogen temperature — seventy-seven Kelvin — and a smaller dataset down to liquid helium, four Kelvin. But the work has been opportunistic rather than programmatic. Most of the compound materials in our library have not been characterised below room temperature."
"Why not?" Karan said.
"Cryogenic access is limited. The liquid nitrogen system is adequate for some measurements. The temperatures below seventy-seven Kelvin require liquid helium, and liquid helium is not available continuously in this facility. We get access to it for specific experiments when we arrange supply from the National Physical Laboratory in Delhi."
"Sharma," Karan said.
"The interface chemistry between dissimilar semiconductor layers," Sharma said immediately. "We know what the bulk properties of our compound materials are. We understand bulk junctions reasonably well. What we don't understand adequately is the electronic states at the interface between two different compound semiconductors — states that are different from the bulk states of either material and that affect device performance in ways that our current models can't fully predict."
He had said this with the directness of someone who had been waiting for an occasion to say it.
"What measurements would resolve the uncertainty?" Karan asked.
"Angle-resolved photoemission spectroscopy to map the electronic states at the interface directly. That requires a synchrotron X-ray source. The nearest one is Photon Factory at the High Energy Accelerator Research Organisation in Japan, which became operational earlier this year."
"Banerjee," Karan said.
Banerjee said: "Quantum effects in semiconductor junctions. The regime where classical device physics models break down and quantum mechanical effects dominate the device behaviour. The theory is there. The experimental techniques to probe it at the single-junction level are not in this facility. The resolution is below what our current instrumentation can achieve."
"Why not improve the instrumentation?" Karan asked.
"Equipment and capability," Banerjee said. "The measurements that probe quantum effects in junctions require current measurement in the picoampere to femtoampere range and voltage measurement in the microvolt range. Those require low-noise amplifier systems and cryogenic measurement setups where you cool the electronics as well as the sample. Hussain."
"The measurement resolution gap," Hussain said. "The thermal noise floor of conventional measurement electronics at room temperature buries the signal we're trying to measure. You have to cool the electronics below liquid nitrogen temperature to reduce the thermal noise sufficiently. We have a liquid nitrogen flow cryostat that we use for Vasudha's low-temperature materials work. Extending to liquid helium and adding low-noise electronics is achievable. Investment and two to three months of setup work."
"Iyer," Karan said.
"Thin-film superconductors on semiconductor substrates," Iyer said. He said it with the careful neutrality of someone who has been thinking about something adjacent to his current work and is uncertain whether the adjacency is worth raising.
"Say more about that," Karan said.
"The deposition work we do for semiconductor devices — sputtering, evaporation, chemical vapour deposition — uses the same physical techniques that would be used to deposit a thin superconducting film. Niobium, lead, tin in the right phase are materials that can be deposited as thin films. Whether a thin-film superconductor behaves the same as a bulk superconductor is a materials question I don't have a complete answer to. The literature suggests that in many cases the superconducting transition is preserved in thin films."
"What would you do with a thin-film superconductor on a semiconductor substrate?"
"The device structure that interests me," Iyer said, "is the Josephson junction. A Josephson junction is two superconductors separated by a thin barrier — a semiconductor or insulating layer — through which the superconducting electrons tunnel quantum mechanically. The junction switches states in response to electromagnetic signals at speeds that are ten to one hundred times faster than silicon transistors at equivalent geometries. IBM has a programme building Josephson junction logic at liquid helium temperatures."
"The IBM programme's limitation?" Karan said.
"Operating temperature," Iyer said. "Four Kelvin. Liquid helium. Expensive, scarce, complex cryogenic infrastructure. A computer that runs at four Kelvin is a laboratory instrument, not a deployable technology. If there were a superconducting material that operated at significantly higher temperatures — even at liquid nitrogen temperature, seventy-seven Kelvin — the economics change completely."
"Why?" Karan said.
"Liquid nitrogen costs approximately one-fiftieth of liquid helium per litre," Hussain said. "And it can be produced on-site with standard industrial equipment rather than from the global helium supply chain. A Josephson junction computer that runs at liquid nitrogen temperature is potentially economically viable. A Josephson junction computer at liquid helium temperature is not, outside a laboratory."
Karan was quiet for a moment.
"Nair," he said.
"Interface chemistry in semiconductor processing," Nair said. "Specifically what happens at grain boundaries and internal interfaces in compound semiconductor materials when you process them at low temperatures. The low-temperature annealing processes we use for surface passivation modify the surface chemistry as we expect — but they also modify the internal interface chemistry in ways we have not fully characterised. The impurity distribution around grain boundaries changes during low-temperature annealing in ways that the standard process models don't predict."
"In which materials?" Karan said.
"Primarily the heavily doped gallium arsenide variants," Nair said. He said this with a slight careful quality in his voice, as if he were approaching something from one side. "The variants that Vasudha has been characterising for the high-speed device programme."
Karan looked at Vasudha Krishnan.
"The anomaly," he said.
The room became very still.
Vasudha said: "You read the lab log."
"I read all the lab logs," Karan said. "Walk me through it."
The anomaly had appeared in a batch of six gallium arsenide samples processed in March 1975.
The samples were variants of ISMC's standard semi-insulating gallium arsenide, heavily doped with a specific combination of impurities to achieve a carrier density appropriate for high-speed transistor applications. They were part of a programme to optimise the GaAs material for the specific device geometry that Sharma and Banerjee's team was working on.
The non-standard element was an annealing step that Nair's team had added to the process sequence for surface passivation reasons. The anneal was performed at 150 degrees Celsius for two hours under a specific ambient. It was not expected to affect the bulk material — surface passivation treatments were expected to affect only the first few nanometres of material at the surface.
The anomaly appeared when Vasudha's team ran the standard four-point electrical resistance measurement on the processed samples and extended the temperature range downward to liquid nitrogen temperature as part of the low-temperature characterisation programme.
At approximately 130 Kelvin, the resistance dropped.
Not to zero. Not by a small amount. The resistance dropped by approximately thirty-five percent — from the value it had been holding from room temperature down to that point, to a significantly lower value that then remained stable all the way to 77 Kelvin, the measurement limit of the current equipment.
Vasudha put the data on the table. The graph was clear. There was a kink at approximately 130 Kelvin that was not a measurement artifact — she had checked the instrument calibration three times and run the measurement in two different instruments with the same result. Below the kink, the resistance was lower. The kink was sharp — it happened over a temperature range of approximately five Kelvin, which was the characteristic behaviour of a phase transition rather than a gradual change in material properties.
"Phase transition at 130 Kelvin," Sharma said. "Reproducible across all six samples of this batch. Not present in any other batch of GaAs we have processed. The distinguishing feature of this batch is the non-standard anneal step."
"What is your hypothesis?" Karan said.
"We have three," Vasudha said. "First hypothesis: a structural phase transition in the impurity sublattice at low temperature — the impurities rearrange into a more ordered configuration below 130 Kelvin, reducing scattering and therefore reducing resistance. This is physically plausible. We have not proven it." She paused. "Second hypothesis: a defect trapping mechanism — certain defects that scatter carriers at higher temperatures become inactive below a threshold temperature. This is less physically compelling for a sharp transition but not impossible." She paused again. "Third hypothesis—"
She stopped.
The room was very quiet.
"Third hypothesis," Karan said.
"Superconductivity," Vasudha said. "Partial superconductivity — not of the entire sample, but of a fraction of the material volume, possibly associated with the internal interfaces that the anneal step creates around the impurity clusters." She said it carefully. "I want to be explicit that we have not claimed this hypothesis. The reason we have not claimed it is that it would require a critical temperature above 100 Kelvin, and the current theoretical consensus is that this is not achievable in any known material."
"The current theoretical consensus," Karan said, "is based on the McMillan formula's upper limit for phonon-mediated superconductivity in bulk materials that have been studied." He looked at Chakraborty. "Is that correct?"
Chakraborty said: "That is correct. The McMillan formula applies to phonon-mediated pairing in materials whose phonon spectra and electronic structures are consistent with the BCS theory's assumptions. In those materials, the formula predicts a maximum achievable critical temperature in the range of thirty to forty Kelvin. Twenty-three point two Kelvin in niobium-tin is the current experimental record."
"Has the McMillan formula been applied to heavily doped compound semiconductors with anomalous impurity cluster structures and modified internal interface chemistry from non-standard processing?" Karan said.
The silence that followed had a specific quality — the quality of a room in which several people had simultaneously thought something and were deciding whether to say it.
"No," Chakraborty said.
"Why not?" Karan said.
"Because the material system has not previously shown any superconducting behaviour that would motivate applying the formula," Chakraborty said. "The formula is applied to materials that are candidate superconductors. Gallium arsenide is not a candidate superconductor. The formula has not been applied because there was no reason to apply it." He paused. "Until now."
"There is now a reason," Karan said. "Because the anomaly exists."
The formal presentations began at ten o'clock. Karan had given the team fifty minutes of questions before the scheduled discussion, and those fifty minutes had distributed the morning's most important content across the room in a way that changed the character of what the presentations would be.
Trivedi presented the division's programme structure. The 1-micron fabrication line — the line that ISMC had achieved in August 1976, four years ahead of Intel's roadmap and substantially ahead of any European or Soviet programme — was producing the Brahma-32 processors and the first generation of compound semiconductor devices for military avionics applications. The compound semiconductor research programme was investigating gallium arsenide and gallium phosphide device structures for high-speed switching above two gigahertz. The process development programme was working toward 0.8-micron in the research phase.
He presented these achievements with accuracy and without the inflation that programme reviews sometimes accumulated.
Karan listened for the details rather than the summary. The 1-micron line's yield at thirty-nine percent — still improving from the initial forty-three percent when it had been confirmed in August, because the Surya yield analysis system had found and corrected two additional process defects in the intervening weeks. The gallium arsenide transistors operating above two gigahertz — the specific frequency performance that made them relevant to the radar processing and communications applications that the Airavat-II and the Gajendra-II needed, and that silicon at any current geometry could not deliver. The 0.8-micron research programme's use of the Zeiss 193-nanometre stepper that had arrived from Germany in September.
"The 0.8-micron timeline," Karan said.
"Laboratory demonstration by end of 1977," Trivedi said. "Production process by 1979 if the resist chemistry work proceeds as Seshadri and Deshpande expect. The acid-catalyzed resist that they developed for the 1-micron process extends to 193 nanometres with modifications they have been working on since July."
"And below 0.8," Karan said.
Trivedi looked at him. "The 0.5-micron regime requires extreme ultraviolet lithography — wavelengths below 100 nanometres — or electron beam lithography for direct writing of patterns. The electron beam approach is feasible at laboratory scale. The production-scale implementation is a different problem."
"Note it in the planning document," Karan said. "The 0.5-micron programme should be on the research agenda for 1978." He paused. "The process roadmap we set in 1974 was: 1-micron by 1976, 0.8-micron by 1979. We are on schedule at the first milestone. The 1980 target on that roadmap was 0.5-micron."
"The 1980 target is achievable if the EUV work begins now," Trivedi said. "It is not achievable if the EUV work begins in 1979."
"Then the EUV work begins now," Karan said.
After Trivedi's overview, Vasudha Krishnan presented the Materials Library programme in full.
Twelve years. More than three hundred semiconductor and compound semiconductor materials characterised. The largest experimental dataset on III-V and II-VI semiconductor properties in Asia, almost certainly.
And then the anomaly.
She put the data on the screen — a simple X-Y plot, temperature on the horizontal axis from room temperature to 77 Kelvin, normalised resistance on the vertical axis. The six samples shown as six lines, all closely overlapping, all showing the same thing: a smooth, slowly decreasing resistance from room temperature down to approximately 135 Kelvin, and then a sharp, discontinuous drop, and then a stable lower resistance from approximately 125 Kelvin to the measurement limit at 77 Kelvin.
The kink was sharp. The repeatability was exact. The six lines were essentially identical.
Sharma said: "The instrument calibration has been checked. The cryostat thermal stability at the transition temperature has been measured. The contact geometry has been varied. The result is not an artifact."
Karan looked at the graph for a long time.
"The drop is thirty-five percent," he said. "Not to zero."
"Not to zero," Vasudha confirmed. "Which is why the superconductivity hypothesis is qualified. A complete superconductor shows zero resistance below the critical temperature. This shows a thirty-five percent reduction."
"In a bulk measurement," Nair said.
Karan looked at him.
"The resistance measurement we are making," Nair said, "is a bulk measurement — it measures the aggregate resistance of the entire sample including the matrix material that is not modified by the impurity cluster anneal. If the superconducting behaviour — if that is what it is — is occurring in a fraction of the material volume, the bulk measurement would show a partial resistance reduction rather than a total reduction. The magnitude of the reduction would reflect the volume fraction of superconducting material." He paused. "Thirty-five percent volume fraction is not unreasonable for the impurity cluster density in this material."
The room absorbed this.
Chakraborty said, from the end of the table: "If the Ginzburg interface superconductivity mechanism is operative — if the Cooper pair formation is occurring at the internal interfaces between the impurity-rich cluster regions and the matrix material — the volume fraction of material participating in the superconductivity would be determined by the interface area per unit volume, not by the volume of the clusters. Thin interfaces can have a significant effective volume fraction in a highly clustered material."
"You've thought about the Ginzburg proposal in this context," Karan said. It was not quite a question.
"I have been thinking about it since Vasudha showed me the data in April," Chakraborty said. He said it without apology, as a fact. "I have not published anything about it because the data does not support publication at the current measurement precision. The measurements that would support or refute the hypothesis require capabilities we do not have."
"Describe the capabilities," Karan said.
Hussain answered. "Specific heat versus temperature, to detect the anomaly in thermodynamic properties associated with the superconducting transition. Magnetic susceptibility versus temperature, to detect the Meissner effect — the expulsion of magnetic field from a superconducting region below the critical temperature. And magnetotransport measurements — the resistance behaviour in an applied magnetic field. Superconductivity is suppressed by applied magnetic fields in a characteristic way. If we apply a magnetic field and the transition temperature shifts in the expected manner, that is strong evidence for superconductivity." He paused. "None of these measurements are possible with the current setup. All three require temperatures below 77 Kelvin, extending to 4 Kelvin, and the current system stops at 77."
"And magnetic field capability," Karan said.
"Yes. We have no superconducting magnet in the facility. A conventional electromagnet would be adequate for initial measurements."
The lunch discussion was where the programme structure actually emerged, because the formal session had opened the direction and the informal conversation during lunch was where the team worked out what following the direction required.
Karan ate with Trivedi, Chakraborty, and Sharma at one end of the canteen table.
He had brought a folder. He placed it on the table between himself and Chakraborty.
Five papers.
The first: McMillan's 1968 paper. "Read the specific assumptions on page seven," Karan said to Chakraborty. "The formula is derived assuming specific things about the electron-phonon coupling spectrum. Those assumptions have been tested for simple metals and intermetallic compounds. They have not been tested for compound semiconductors with anomalous interface structures."
Chakraborty found the page. He read in silence for a few minutes.
The second paper: Ginzburg's 1970 proposal for surface and interface superconductivity, published in the Soviet physics literature and translated by the American Institute of Physics in 1971. "The mechanism Ginzburg proposes requires a thin layer at a surface or interface where the dimensionality changes the electron-phonon coupling."
The third: Little's 1964 paper on excitonic superconductivity in organic systems. "Little argues that non-phonon mediation — pairing mediated by electronic excitations rather than lattice vibrations — removes the McMillan upper limit entirely. His specific material proposal has not been realised. The physical principle has not been refuted."
The fourth: Matthias's 1975 empirical survey of intermetallic superconductors. "Read the concluding section. Matthias acknowledges directly that the theoretical tools are insufficient to predict which untested materials will be superconductors and at what temperatures."
The fifth: a paper that none of them had read, from a Swiss research group in the Journal of Physics, published in 1974, describing anomalous low-temperature resistance behaviour in heavily doped strontium titanate. "Bednorz was a co-author on this paper," Karan said. "He is currently at IBM Zurich pursuing perovskite oxides. His direction is different from ours. But the observation of anomalous resistance behaviour in an oxide perovskite — which is a material class nobody was looking at as a superconductor candidate — is relevant. Anomalous resistance behaviour in unexpected material classes is apparently more common than the current mainstream programme acknowledges."
Chakraborty had been reading the McMillan paper carefully. He looked up.
"The formula assumes weak to moderate coupling," he said. "McMillan's own discussion section acknowledges that in the strong-coupling regime, the formula's predictive validity becomes uncertain. The formula extrapolates beyond its validated range when applied to strongly coupled systems." He paused. "Heavily doped compound semiconductors with impurity cluster interfaces — if the coupling at those interfaces is strong in the sense the Eliashberg theory describes — would be in exactly the regime where McMillan's formula is an extrapolation rather than a validated prediction."
"Yes," Karan said.
"I want to calculate the transition temperature prediction for the specific geometry that Nair is describing," Chakraborty said. He said it with the focused directness of a theorist who has identified the problem he wants to work on. "The two-dimensional interface case. The Eliashberg equations can be extended to a two-dimensional geometry — this has been done for flat surfaces in the literature. For the specific geometry of impurity cluster interfaces in a doped compound semiconductor, the calculation requires knowing the interface phonon spectrum, which we don't have. But I can do the calculation for a range of coupling constant values and interface geometries and produce a prediction envelope. If the prediction envelope for physically reasonable parameters includes temperatures above 77 Kelvin, the hypothesis is worth pursuing experimentally."
"How long for the calculation?" Karan said.
"Three days," Chakraborty said. "A week if I want to be careful about every approximation."
"Be careful," Karan said. "This is the theoretical basis for an experimental programme. Get it right."
The afternoon session was working rather than formal. Karan, Trivedi, Chakraborty, Sharma, Vasudha, Nair, and Hussain — the team whose work was directly relevant to the programme direction — sat in Conference Room B with the morning's data and the five papers and built the programme structure from the ground up.
Hussain presented the cryogenic capability extension in detail.
The liquid helium system: a closed-cycle helium recovery and re-liquefaction system, which could be installed in the facility within three months and would provide continuous liquid helium production from recovered helium gas rather than requiring delivery of liquid helium from the NPL in Delhi. The upfront cost was sixty lakhs. The ongoing operating cost after installation was approximately eighty percent lower than the current arrangement, and the supply chain vulnerability — the dependence on NPL's liquefier schedule and the Delhi-to-Gorakhpur transport logistics — was eliminated.
The low-noise measurement electronics: a custom electronics package that Hussain had been designing for eighteen months, initially for the quantum junction measurements that Banerjee wanted to do, which could be adapted for the superconductor detection measurements. The package operated at 4 Kelvin, reducing the thermal noise floor to below ten picovolts per root hertz — adequate for the measurements the programme required.
The superconducting magnet: a small solenoid magnet operating at liquid helium temperature, capable of producing fields up to eight Tesla. This was the standard tool for magnetotransport measurements of superconducting samples. It was commercially available from several suppliers. Delivery time approximately six months.
"The magnet is the critical path item," Hussain said. "The helium recovery system and the electronics can be ready in three months. The measurements that require a magnetic field have to wait for the magnet."
"Who is the vendor?" Karan said.
"Oxford Instruments in England," Hussain said. "They have the best lead time — six months is their estimate. The American and German options are eight to ten months."
"Order it today," Karan said. "Offer them double the list price for a four-week delivery. If they refuse, offer triple. Make it a commercial proposition they cannot justify declining."
Hussain nodded, writing it down.
Karan looked at the table. "What is the Oxford magnet coil actually made of?"
"Niobium-titanium alloy," Hussain said. "Drawn into fine filaments and embedded in a copper matrix for thermal stability."
Karan sat with that for a moment. He ran the arithmetic of the industrial empire he had just reviewed in Lucknow.
"We mine niobium in our rare earth operations," Karan said, his voice dropping into the flat, absolute register he used when making structural decisions. "We process aerospace-grade titanium at the Ganjam complex in Odisha. We refine copper in Gilgit. The Kanpur plant specializes in extreme-tolerance metallurgy."
Trivedi looked at him carefully. "We have the raw materials. But we have never drawn superconducting wire. The extrusion process and the metallurgical precision required are entirely different from structural alloys."
"Then Kanpur learns," Karan said. "If this anomaly is what Chakraborty's math suggests it is, the global technology architecture is going to pivot. Energy, medical imaging, computing — all of it will run on magnetic fields. I am not building the future of Indian physics dependent on a six-month wait for a British catalog order."
He looked at Trivedi.
"We buy the Oxford magnet to get this specific experiment running now. But you initiate a joint working group with the Kanpur metallurgy division tomorrow morning. Shergill Industries is entering the superconducting magnet business."
Nair presented the microstructural characterisation programme.
To understand what the impurity cluster structure actually looked like — to know whether the internal interfaces that Ginzburg's theory required were actually present — the team needed transmission electron microscopy. TEM could image individual atomic columns in crystalline materials and could therefore show whether impurity clusters had the ordered structure that would produce the interface geometry relevant to the superconductivity hypothesis.
ISMC did not have a TEM.
The nearest capable instrument was at IIT Bombay. A collaboration arrangement was required — IIT Bombay's physics department had a TEM that was used for materials characterisation, and Trivedi suggested a standard joint research agreement that gave ISMC access to the instrument in exchange for academic co-authorship.
Karan stopped him with a sharp gesture.
"No," Karan said. "There will be no joint research agreements. And there will be absolutely no publications."
The room went dead quiet.
"If Chakraborty's calculations are correct," Karan said, leaning forward, "we are looking at a paradigm shift that changes the strategic balance of the planet. If we publish a single paper, the Americans, the Soviets, and the Japanese will have their entire industrial apparatus cloned onto this within six months. We do not announce breakthroughs to the world before we have weaponized and commercialized them."
Trivedi cleared his throat. "Then how do we get access to the IIT Bombay TEM? The academic departments operate on a culture of publication. They won't let us use their primary research asset for classified corporate work without an institutional return."
"We don't ask for a partnership, we buy the asset," Karan said flatly. "What is the replacement cost of that specific TEM model?"
"Approximately twelve lakhs," Trivedi said, taken aback. "Plus customs and installation."
"We write IIT Bombay a check for thirty lakhs tomorrow morning," Karan ordered. "As a direct, unconditional corporate grant from Shergill Industries to build a brand-new, next-generation materials characterization lab. In exchange, they hand over the keys to the existing TEM room to an ISMC engineering team for the next sixty days. Their faculty and students stay out. Our personnel operate the machine under full corporate non-disclosure protocols."
He looked at Nair.
"I want our samples on that TEM by the end of next week, not November. The data comes straight back to Gorakhpur via courier. Nothing is logged on the university network. We do not leave a paper trail for a discovery of this magnitude."
Vasudha presented the systematic material synthesis programme.
Six additional samples, varying the impurity composition and the anneal conditions systematically to map out which processing parameters were associated with the anomaly and which were not. The logic was straightforward: if the anomaly was associated with specific structural features created by the anneal, then systematically varying the anneal should modulate the anomaly in a predictable way. If the anomaly was a random artifact, it would not vary systematically with the anneal parameters.
"Which parameters do you vary?" Karan asked.
"Four axes," Vasudha said. "Anneal temperature, from 100 to 200 degrees Celsius in steps of 25 degrees. Anneal duration, from thirty minutes to four hours. Anneal ambient — inert gas versus reactive ambient. And impurity concentration, spanning the range from the current value down by a factor of two and up by a factor of two." She paused. "The full factorial would be forty-eight samples. I propose a fractional factorial design that captures the main effects and the most important interactions in twelve samples."
"How long?" Karan said.
"Four weeks for synthesis and room-temperature characterisation. The cryogenic measurements below 77 Kelvin wait for the helium system. The liquid nitrogen measurements — down to 77 Kelvin, where we already see the anomaly begin — I can start in two weeks."
"Start in two weeks," Karan said. "I want the nitrogen-temperature data from the systematic series by the end of November."
Chakraborty was at the whiteboard. He had not planned to be at the whiteboard — the whiteboard in Conference Room B was a fixture of the room rather than a planned element of the meeting — but the discussion had reached the point where his thinking needed external representation rather than internal processing, and he had picked up the marker without specifically deciding to.
He wrote the Eliashberg equation in the form relevant to the interface geometry. This was not the standard form that appeared in textbook treatments of BCS theory — it was a two-dimensional extension that Chakraborty had been working on, on paper, since April, and that he now wrote on the whiteboard for the first time.
"The interface is treated as a two-dimensional layer," he said, writing. "The electron density of states in the interface layer is different from the bulk because the dimensionality changes the phase space available for electronic states. The phonon modes at the interface — the interface phonons — are different from the bulk phonon modes because the boundary conditions change the allowed vibration modes." He wrote the interface phonon spectrum term. "The effective coupling constant for the two-dimensional interface is the product of the electron-phonon matrix element, which measures how strongly electrons interact with phonons, and the interface phonon density of states, integrated over the relevant phonon frequency range."
He stopped writing.
"The critical temperature," he said, "in the Eliashberg framework for the two-dimensional interface, depends on the effective coupling constant in the same qualitative way as the bulk critical temperature — exponentially. A small increase in the effective coupling constant produces a large increase in the critical temperature." He looked at the equation. "The question is whether the interface phonon spectrum and the matrix element at the impurity cluster interface are significantly different from the bulk values. If they are — if the interface confinement enhances the coupling — the critical temperature can be substantially higher than the bulk prediction."
He wrote a range of critical temperature predictions for a range of coupling constant values.
At a coupling constant of 1.5 — the range of strongly coupled conventional bulk superconductors — the predicted critical temperature was 15-25 Kelvin. Below the McMillan limit as expected, because the formula was being applied to a two-dimensional geometry with an enhanced but not extraordinary coupling.
At a coupling constant of 2.5 — the upper range of what had been measured in any conventional superconductor — the predicted critical temperature was 45-70 Kelvin.
At a coupling constant of 3.5 — above the range of measured conventional values, but within the range that theoretical estimates for strongly coupled interface systems suggested was possible — the predicted critical temperature was 95-140 Kelvin.
The room was very quiet.
"The transition in the anomalous GaAs samples is at approximately 130 Kelvin," Vasudha said. She was looking at the whiteboard.
"Yes," Chakraborty said.
"That is consistent with a coupling constant of approximately 3 to 3.5," Sharma said.
"Consistent with," Chakraborty said carefully. "Not proven by. The calculation makes specific assumptions about the interface geometry that we have not verified experimentally. The coupling constant value of 3 to 3.5 would be extraordinary — significantly higher than any measured bulk superconductor. Whether it is achievable at a compound semiconductor interface with a specific impurity cluster structure — I cannot calculate that without the interface phonon spectrum data."
"Which the experiments will provide," Nair said.
"Which the experiments will provide," Chakraborty confirmed.
Karan was looking at the whiteboard. He had the specific quality of stillness that he maintained when something important had been demonstrated — not excitement, not triumph, the quality of a person who has seen something that is correct and is now reorganising his thinking around it.
"The calculation does not predict that the anomaly is superconductivity," he said.
"No," Chakraborty said.
"The calculation shows that the anomaly is consistent with superconductivity at a coupling constant that is high but not physically prohibited," Karan said.
"Yes."
"And the coupling constant required — 3 to 3.5 — is in the range that the Ginzburg interface mechanism could produce, for specific interface geometries, without invoking any physics outside the established theoretical framework."
"The two-dimensional Eliashberg framework extended to interface geometries is established," Chakraborty said. "It has been applied to flat surfaces. The application to impurity cluster interfaces is new. But the physics is the same physics."
"Then the programme is clear," Karan said. He turned from the whiteboard. "We build the cryogenic capability and we make the measurements that test the hypothesis. Simultaneously, Chakraborty refines the theoretical prediction as the experimental data becomes available. Nair characterises the interface structure by TEM and XPS. Vasudha maps the parameter dependence of the anomaly. Hussain designs the full measurement protocol for the liquid helium regime. Sharma designs the experimental programme for the magnetotransport measurements once the magnet arrives."
He looked at Trivedi.
"Programme structure," he said. "This becomes a dedicated research programme — not a sub-programme of the compound semiconductor work, a standalone programme with its own timeline, its own reporting, and its own personnel. The team working on this does not divide their attention with the production line demands."
"The personnel question," Trivedi said. "The team as constituted is largely committed to existing programme deliverables. A dedicated programme requires either pulling people from those deliverables, which affects them, or hiring additional personnel."
"Both," Karan said. "Pull Vasudha's materials team and Nair's interface team to the dedicated programme — the existing characterisation programme continues but at reduced pace. And hire. I want two postdoctoral positions posted within the month. Experimental physicists with cryogenic measurement experience." He paused. "Where do we find them?"
"IIT Madras has a strong cryogenics programme," Trivedi said. "Several of their recent doctoral graduates have the background. I know Krishnaswami's assessment of the three best from the last two cohorts."
"Contact them this week," Karan said.
The meeting ended at five-thirty.
The team dispersed in the way that teams dispersed after discussions that had produced more work than the agenda had specified — with the forward lean of people who had more to do than they had come in with and who found this state of affairs appropriate rather than burdensome.
Chakraborty stayed at the whiteboard after the others had left, adding to the calculation he had written earlier. He was working on the next layer of the theoretical prediction — the specific parameter dependence that the experimental programme needed to test in order to distinguish the superconductivity hypothesis from the alternative explanations.
Karan watched him for a moment from the doorway.
He thought: this is what the facility is for. Not the production line alone, though the production line matters — it pays for the facility and it produces the capabilities that the aircraft and the computers need. This is what the facility is for. The moments when a careful researcher shows an anomaly in clean data and another researcher derives the equations on a whiteboard and the two meet in a room where someone asks the right questions.
He thought about what the right questions were.
The right question was not: is this superconductivity? That question would be answered by the measurements.
The right question was: if it is superconductivity — if the anomaly in this specific gallium arsenide variant, produced by this specific non-standard anneal process, is the first observation of superconductivity above 100 Kelvin — what does that mean?
He knew what it meant for the physics. The McMillan limit would be broken. The phonon-mediated mechanism in a two-dimensional interface geometry, at sufficiently high coupling constants, could achieve critical temperatures that the bulk theory said were impossible. The entire experimental programme of searching for higher-temperature superconductors — the Matthias programme, the IBM programme, the Soviet programme — would have to reconsider its assumptions.
He knew what it meant for computing. The Josephson junction logic that IBM was building at four Kelvin — the programme that was years away from producing a functional processor and which required liquid helium infrastructure that made it economically impractical at scale — would look completely different if the operating temperature were liquid nitrogen. At 77 Kelvin, the economics of superconducting computing changed from laboratory curiosity to potential commercial technology. The factor-of-one-hundred speed advantage of Josephson junction switching over silicon transistors would become relevant not in a decade but in years.
He knew what it meant for everything else that superconductors could do. Magnets. Energy transmission. Measurement. Detection. Medical imaging. The applications that were currently constrained to liquid helium temperatures and therefore to specialised laboratory and industrial use would become available at liquid nitrogen temperatures and therefore potentially to applications that the current constraints made impossible.
He did not know whether the anomaly was superconductivity. The measurements would tell him.
What he knew was that the anomaly was real, that it was reproducible, that it was unexplained, and that the most interesting explanation for it was physically consistent with the established theoretical framework when that framework was extended to the geometry that the material's processing history suggested was present.
That was enough to know. The rest was the work.
He had started many things from less.
He drove back toward the guest house as the October evening settled over Gorakhpur.
The steel complex's lights were coming on to the south. The agricultural processing complex to the northeast. The main Shergill Industries campus with its own power generation and its own water treatment and its own township that housed the families of a hundred and fifty thousand workers and that had not existed in 1965 and now existed so completely that people who had not known Gorakhpur before it could not imagine the city without it.
He thought about what the facility's most important work was.
Not the 1-micron process, which was the best in the world and had cost six years of sustained investment and the specific knowledge of Seshadri and Deshpande and the team they had built. The 1-micron process produced the Brahma-32. The Brahma-32 was not merely the most capable processor in the country; it was the single most advanced microprocessor currently in global existence.
Not the ISMC production facility as a whole, which was the foundation of the country's absolute semiconductor supremacy — the reason India no longer needed to ask permission from an American company or a Japanese company when it wanted to build a capable computing system.
Those things mattered. They mattered enormously.
But the most important work was the work that Chakraborty was doing at the whiteboard at five-thirty in the evening in Conference Room B, extending the Eliashberg equations to a two-dimensional interface geometry and writing down the predicted critical temperature range for a coupling constant that might or might not be achievable in a real material.
Because if the prediction was right — if the anomaly was what the most interesting explanation said it was — then what ISMC had done, accidentally, in a batch of six gallium arsenide samples prepared with a non-standard anneal step for a completely different purpose, was discover the first material in the history of physics in which superconductivity occurred above liquid nitrogen temperature.
That was not a semiconductor achievement. That was not an industrial achievement.
That was a physics discovery.
And the physics discovery — if the measurements confirmed it — would change everything. Not immediately. Not in a way that was visible tomorrow or next year. In the way that the discovery of transistors at Bell Laboratories in 1947 had changed everything — not immediately visible in 1947, apparent within five years, world-altering within twenty.
He did not know yet whether the anomaly was what it might be.
In three months, the cryogenic system would be ready. In four months, the first liquid-helium-temperature measurements would run. In four months, the question would begin to be answered.
Until then: the work.
He arrived at the guest house. He went inside.
There were files on the desk. The infrastructure programme's quarterly review. The constitutional reform commission's latest document. A cable from the Vanguard field director about the Nepal corridor's lethal conclusion. The Taxila board's minutes from its first meeting.
He sat down and opened the first file.
There was still work to do.
There always was.
End of Chapter 241
ISMC Advanced Research Division — Programme Status, October 1976
Current Capability: 1-micron fabrication process (production, 39% yield, improving; current global peak). 0.8-micron process (laboratory research, Zeiss 193nm stepper operational). Brahma-32 processors in production qualification. GaAs high-speed devices above 2 GHz demonstrated. Compound semiconductor materials library: 300+ materials characterised.
Anomaly Under Investigation: Material: Heavily doped GaAs variant with non-standard 150°C anneal treatment. Observation: 35% resistance reduction at approximately 130 Kelvin. Reproducibility: 6 independent samples, 3 independent measurement instruments. Alternative explanations: phase transition in impurity sublattice; defect trapping mechanism; partial superconductivity in interface regions (not yet confirmed).
Theoretical Assessment (Chakraborty, October 15, 1976): 2D Eliashberg calculation for impurity cluster interfaces. Coupling constant 3.0-3.5 required for Tc ~130K. Coupling constant range physically plausible for interface geometry per Ginzburg (1970). McMillan formula does not apply to this material class — derived for bulk phonon-mediated systems, not tested for compound semiconductor interface geometries.
Programme Extensions Authorised: * Cryogenic system: Liquid helium recovery/re-liquefaction installation (60 lakhs, 3-month lead time).
Superconducting magnet (Interim): Oxford Instruments, 8T. Premium buyout authorized for expedited 4-week delivery.
Superconducting magnet (Sovereign): Joint working group initiated with Shergill Metallurgy (Kanpur) to establish internal superconducting wire drawing and magnet production using internal niobium and titanium supply chains.
TEM access: IIT Bombay facility buyout (30-lakh corporate grant in exchange for 60-day exclusive ISMC operational control). Security Protocol: Absolute NDA. Zero academic collaboration or co-authorship permitted. Zero publication.
XPS measurements: Internal capability, immediate.
Postdoctoral hiring: 2 positions, cryogenic measurement specialisation.
Systematic Synthesis Programme (Vasudha): 12-sample fractional factorial design. Parameters: anneal temperature, duration, ambient, impurity concentration. Nitrogen-temperature data target: end November 1976. Helium-temperature measurements: pending system installation (January 1977).
Key Unknown: Whether Tc ~130K anomaly is consistent with superconductivity measurable by Meissner effect (magnetic susceptibility), specific heat anomaly, and magnetotransport characteristics. First liquid-helium-temperature measurements: February 1977.