Platform
Replacing trial and error with computation.
Bring the device and the constraint. The platform searches the material, what it touches and how it is made at the same time, returns the routes that clear your constraint with the numbers under each, and runs the programs that prove them. Paid search campaigns are running on it now.
- Paid search campaigns running with fusion and quantum hardware teams
- Live with CAN Superconductors today
- Two never-made superconductors in synthesis, Georgia Tech runs in final stages
A representative run of the engine: an HTS-cable brief is read with its metrics, the space of materials, interfaces and routes is swept at once, what runs today is scored first, the routes are gated on current, anneal window, cost and length, five routes to a working cable clear, and a record is assembled with provenance on every row. The bench result returns to the model.
01 · Your stack, solved
Bring the stack you run. We search every one you could.
A team that runs on REBCO, Nb₃Sn or a nitride film brings the stack it runs: the conductor, the layers against it and the process at every step that makes it. The search covers every point on all of that at once, so the loop that used to cost a furnace run per answer costs a sweep, and a bounded question about your own stack runs the same machinery as a search for something new. One route, every step, below.
The stack you hand us stays yours: one workspace per program, encrypted in transit and at rest, and who keeps what is written down before the first run. What each side keeps
Where this lands
None of the 19 fields this work reaches is a market anybody has to create: 17 of them already run on a superconductor, and every one of them is capped by how cold it has to be kept. The nineteen fields, one by one
One process, and every route through it
31,104routes through these nine steps
SubstrateHastelloy C-276This route· 3 published alternatives at this step
- What it meets
- Hastelloy C-276 tape, about 50 µm thick, electropolished; room temperature, in air.
- What the search decides
- The alloy and the finish under the whole stack: what the buffer can grow on, and the piece length a vendor can supply.
- What fails here
- A rough or contaminated surface seeds every defect in the layers above it.
Buffer stackIBAD-MgOThis route· 3 published alternatives at this step
- What it meets
- Sputtered oxide layers a few hundred nanometres thick, in vacuum, at several hundred degrees.
- What the search decides
- The texture the superconductor inherits, grain by grain, and the barrier that keeps the substrate's nickel out of it. The three routes to that texture are ion-beam-assisted deposition, inclined-substrate deposition and a rolling-assisted textured substrate.
- What fails here
- Misaligned grains cut the critical current by an order of magnitude.
REBCO growthMOCVDThis route· 4 published alternatives at this step
- What it meets
- The film itself, one to two microns, by MOCVD near 800 °C in oxygen.
- What the search decides
- The composition, the pinning centres that hold current in field, and the rate a fab can run without losing them. Four deposition routes are in production or pilot production: metal-organic chemical vapour deposition, pulsed laser deposition, reactive co-evaporation and metal-organic deposition.
- What fails here
- The wrong phase forms, or the pinning is not there at 77 K.
Oxygen anneal450 to 500 °CThis route· 3 published alternatives at this step
- What it meets
- Hours in flowing oxygen between roughly 400 and 550 °C, the window the fab has to hold every run.
- What the search decides
- The oxygen content that makes the film superconduct at all, and the window the route asks of the fab.
- What fails here
- An under-oxygenated film never reaches its critical current.
Silver cap2 µm silverThis route· 3 published alternatives at this step
- What it meets
- About two microns of sputtered silver, then a short anneal to bond it to the film.
- What the search decides
- The electrical contact into the film, and how every joint on the cable will solder.
- What fails here
- A poor interface adds resistance at every joint and every termination.
Copper stabiliser20 µm copperThis route· 4 published alternatives at this step
- What it meets
- Electroplated copper, tens of microns a side, at room temperature in a plating bath.
- What the search decides
- Where the current goes when the film quenches, and the thermal mass that decides how fast. The copper is electroplated; the laminates are soldered on, in copper for conduction or in steel for strength.
- What fails here
- Too thin and a quench burns the tape; too thick and the engineering current density falls.
Slitting4 mmThis route· 4 published alternatives at this step
- What it meets
- The wide tape cut into strips a few millimetres wide, at speed, in air.
- What the search decides
- The edge quality, because a crack at the edge propagates under bending.
- What fails here
- Edge damage that shows up only at 77 K, as lost current.
CablingHelical layersThis route· 3 published alternatives at this step
- What it meets
- Tapes wound in layers around the former, at a set pitch, under tension.
- What the search decides
- The pitch and the layer count that share the current and set the AC loss. The three published architectures are helically wound layers, conductor on round core, and a Roebel-transposed cable.
- What fails here
- Bending strain past the tape's limit, or layers that do not share the current.
The cryostat77 K nitrogenThis route· 2 published alternatives at this step
- What it meets
- The cable in its cryostat, liquid nitrogen at 77 K, a joint and a termination at each end.
- What the search decides
- The joints' resistance, the heat load, and the margin at the warmest point of the loop. Nitrogen boils at 77 K at one atmosphere and can be subcooled to about 65 K, which buys margin and costs pumping.
- What fails here
- A joint that heats, or a loss the cryostat cannot lift.
31,104routes through these nine steps
- Hastelloy C-276
- Stainless steel
- Textured Ni-W
- What it meets
- Hastelloy C-276 tape, about 50 µm thick, electropolished; room temperature, in air.
- What the search decides
- The alloy and the finish under the whole stack: what the buffer can grow on, and the piece length a vendor can supply.
- What fails here
- A rough or contaminated surface seeds every defect in the layers above it.
- IBAD-MgO
- ISD-MgO
- RABiTS
- What it meets
- Sputtered oxide layers a few hundred nanometres thick, in vacuum, at several hundred degrees.
- What the search decides
- The texture the superconductor inherits, grain by grain, and the barrier that keeps the substrate's nickel out of it. The three routes to that texture are ion-beam-assisted deposition, inclined-substrate deposition and a rolling-assisted textured substrate.
- What fails here
- Misaligned grains cut the critical current by an order of magnitude.
- MOCVD
- PLD
- Co-evap.
- MOD
- What it meets
- The film itself, one to two microns, by MOCVD near 800 °C in oxygen.
- What the search decides
- The composition, the pinning centres that hold current in field, and the rate a fab can run without losing them. Four deposition routes are in production or pilot production: metal-organic chemical vapour deposition, pulsed laser deposition, reactive co-evaporation and metal-organic deposition.
- What fails here
- The wrong phase forms, or the pinning is not there at 77 K.
- 400 to 450 °C
- 450 to 500 °C
- 500 to 550 °C
- What it meets
- Hours in flowing oxygen between roughly 400 and 550 °C, the window the fab has to hold every run.
- What the search decides
- The oxygen content that makes the film superconduct at all, and the window the route asks of the fab.
- What fails here
- An under-oxygenated film never reaches its critical current.
- 1 µm silver
- 2 µm silver
- 3 µm silver
- What it meets
- About two microns of sputtered silver, then a short anneal to bond it to the film.
- What the search decides
- The electrical contact into the film, and how every joint on the cable will solder.
- What fails here
- A poor interface adds resistance at every joint and every termination.
- 20 µm copper
- 40 µm copper
- Copper laminate
- Steel laminate
- What it meets
- Electroplated copper, tens of microns a side, at room temperature in a plating bath.
- What the search decides
- Where the current goes when the film quenches, and the thermal mass that decides how fast. The copper is electroplated; the laminates are soldered on, in copper for conduction or in steel for strength.
- What fails here
- Too thin and a quench burns the tape; too thick and the engineering current density falls.
- 2 mm
- 4 mm
- 6 mm
- 12 mm
- What it meets
- The wide tape cut into strips a few millimetres wide, at speed, in air.
- What the search decides
- The edge quality, because a crack at the edge propagates under bending.
- What fails here
- Edge damage that shows up only at 77 K, as lost current.
- Helical layers
- CORC
- Roebel
- What it meets
- Tapes wound in layers around the former, at a set pitch, under tension.
- What the search decides
- The pitch and the layer count that share the current and set the AC loss. The three published architectures are helically wound layers, conductor on round core, and a Roebel-transposed cable.
- What fails here
- Bending strain past the tape's limit, or layers that do not share the current.
- 77 K nitrogen
- 65 to 70 K
- What it meets
- The cable in its cryostat, liquid nitrogen at 77 K, a joint and a termination at each end.
- What the search decides
- The joints' resistance, the heat load, and the margin at the warmest point of the loop. Nitrogen boils at 77 K at one atmosphere and can be subcooled to about 65 K, which buys margin and costs pumping.
- What fails here
- A joint that heats, or a loss the cryostat cannot lift.
02 · What comes back
Not one answer. Every route that clears your constraint, in full.
A record is every route that clears the brief, ranked, and each one is written at the length the decision takes: the numbers you named with their values and where each came from, the trade-offs on the ones you did not, the stack layer by layer, the first three moves, and a line you can ask anything. One of five, for a real brief, with representative values, below.
Solution 02 of 05
Representative record · not a program result
The brief
A 1 km HTS transmission cable for a city grid: 3 kA per phase at 77 K in liquid nitrogen, tape Ic of at least 400 A per centimetre of width, joint resistance under 20 nΩ, a 1,000 m piece, AC loss inside the cryostat's budget, and cost at or under your ceiling per kA·m.
This route fixes
- REBCO
- on Hastelloy
- by MOCVD
The device
An HTS cable in section: the former, the tape layers, and the tape stack the record is about, drawn heavy. Liquid nitrogen fills the annulus inside the cryostat.
The route's stack
- Copper stabiliser20 µm
- Silver cap2 µm
- REBCO1.5 µm
- Buffer stack0.2 µm
- Hastelloy substrate50 µm
Heights are log-proportional to the published thicknesses. Hover a layer, or a row.
What you asked forFour numbers, each with its value, where it stands and how we know.
Value
Where it stands
How we know
Ic per centimetre of width at 77 K
The number the cable is rated from.
460 A/cm-w · 77 K · self-fieldHow we knowCritical current of commercial REBCO tape at 77 K in self-field is published in the range of roughly 250 to 900 A per centimetre of width, depending on the vendor, the grade and the pinning; most production tape sits between 400 and 600.
Clears, with margin
Measured, short tape
Operating margin at the cable's warmest point
How much headroom the warmest turn has before it quenches.
1.5× at 3 kAHow we knowThe ratio of the cable's critical current, the tapes in parallel with the self-field derate, to the operating current at the warmest point of the nitrogen loop, computed from the row above and the cable's layout. Published cable designs run between about 1.3 and 2.
Clears
Computed
Resistance across a tape joint
Where a cable warms first, and what the cryostat pays for.
14 nΩ, solderedHow we knowSoldered lap joints between REBCO tapes at 77 K are published from a few nanohm to a few tens of nanohm, depending on overlap and solder.
Clears against 20 nΩ
Measured, short tape
AC loss at rated current
The heat the cryostat has to lift, every second the cable runs.
0.7 W/m per phase · 50 HzHow we knowMeasured losses of HTS cable phases at rated current are published from roughly 0.2 to 2 W per metre at power frequency, with the well-optimised cables under 1; this route's number is computed from the tape width, the layer count and the current, and sits in the middle of that range.
Marginal against the cryostat budget
Computed
What we take off your plateThe six steps between a route on paper and a part in your hands.
Every trade-off above is six steps somebody has to run: source the tape, book the furnace, make the joint, measure it, validate it against the spec, ship it. Run it yourself and we plug into the benches you already run. Hand it to us and the waits go: the vendors are ones we already buy from, and the benches are ones we already book.
Run it yourself
Run it with us
Synthesis·The route's tape is made, or bought from the vendor whose published process it is.
Vendors and sourcing·Targets, powders, substrates and tape, from suppliers we already buy from, on one order.
Manufacturing·The tape becomes the part: slit, wound, jointed, on the line the brief names.
Measurement·The numbers the brief asked for, measured on the part, at the conditions it will run at.
Validation·Measured against the spec, row by row, with the misses written down as plainly as the hits.
Shipping·The part ships to you, and the result ships home to the model.
Run on the platform and our network.
The trade-offs on this routeFour numbers you did not ask about, and what this route does to each.
Cost per kA·m at 77 K
1.3× your ceilingHow we knowA ratio to the ceiling the brief states, never a currency. The record prices a route against your number, and this one is over it.
Over the ceiling
Computed
Piece length demonstrated
600 mHow we knowCommercial REBCO tape is supplied in piece lengths of several hundred metres, with a kilometre available from few suppliers; the brief asked for 1,000.
Short of your kilometre
Measured, short tape
Substrate you would have to move to
Hastelloy, from stainlessHow we knowCoated conductors are made on Hastelloy C-276; the brief's own stack was specified on stainless steel.
One change
From the published route
Process steps this adds
One, after depositionHow we knowThe route's published sequence: buffer, deposition, then the oxygenation anneal the window row describes.
An oxygenation anneal
From the published route
Every value sits inside a published range. Hover or focus a value for it.
The first three movesThree things to do this month, and why each one.
01Draw the first hundred metres on this route
Every number you asked for on it is measured on tape, and a measured route that clears beats a computed one that clears by more.
02Prove the joint at 77 K on your own rig
It is the line the cryostat budget stands or falls on, and the one your rig can settle in a week.
03Keep the substrate you qualify today
The saving from moving is smaller than the requalification it would trigger, and the record says which of the two it measured.
Ask the platformThree questions this record has been asked, and what it said.
Ask anything about this record. The platform runs on two kinds of model. The physics model, built on the MEL framework, isolates the few variables that decide superconductivity; learned models on top of it rank routes against a brief, and every bench result that comes home, hit or miss, trains them. A founder reads every record today. A seat lets a team change a brief, move a constraint and run again without one. Want a row changed, a constraint moved, a different bench, or the whole brief redone? Ask, and the rerun is what used to be a furnace run.
Why this substrate and not the one we already qualify?
Because the oxygenation anneal this route needs is outside the window your current substrate holds, and every other row here is written against the substrate that holds it. Move back and two of the four rows you asked about stop clearing. That trade is what the cost row is pricing.
What happens to the margin if the phase current goes to 4 kA?
The margin row is computed, so it is the one number here that moves without a new sample: at 4 kA it falls to about 1.1, and the AC loss row rises with it. Ic per width is measured and would need one. Piece length does not move at all. The record tells you which of the three you are actually waiting on before you ask for any of them.
Which of these numbers change if we run subcooled at 65 K?
Ic per width and the margin both rise, and the AC loss falls with them; all three are published trends. Nothing measured in this record was measured below 77 K, so the honest answer is that we do not know until a bench runs it. That is what the program would be for.
03 · Internal today, enterprise seats next
The software we run today ships to your team next.
Today
Paid search campaigns are running with fusion and quantum hardware teams, and so is our own search, on the platform as it stands: a founder reads every record, and the record is the deliverable.
Next
An enterprise package: the same search, the same records and the same ask, run by your team in your own workspace, with less of us in the loop and every rerun yours to start.
Yours stays yours
Encryption in transit and at rest, sign-in through your own directory, one workspace per program. The full ledger, and what nobody has certified, is on one page.
04 · The validation
You do not have to take our word for any of this.
Independent validation
A group at Stanford and SLAC measured independently, across several cuprate families, that superconductivity strengthens the phase coherence of charge order rather than competing with it, and published the result in Physical Review Letters. That effect is what the platform ranks on. We had no part in the work and no stake in its answer.
DOI 10.1103/g41t-8456
Superconductivity reinforces charge-density-wave phase coherence across cuprates
Lee et al. · Phys. Rev. Lett., 2026
Exclusive license
Hyunsung TNC
The Modulated Electron Lattice (MEL) framework, in development at Hyunsung TNC since 2006 and licensed exclusively to SuperMatics.
20+ patents issued and pending
Live today
Paid search campaigns running with fusion and quantum hardware teams. Device teams' own constraints, run on the platform as it stands, and the record is what comes back.
Plug and Play
One of eleven startups in Batch 4 of the NeoCity Semiconductor accelerator, which Plug and Play runs in Florida, with months of access to Plug and Play's mentors and corporate partners.
Underway
Manuscript · in preparation
The lab results are finalized, and the paper is about what they mean for the framework.
Theory
MEL framework
Measurement
Georgia Tech · STM / STS
05 · Send us your question
Run one search against your own problem.
Tell us the device, what it has to do and where it stops. A founder reads it, and within a week you have our first take on whether the platform has anything to offer, at no cost and with no obligation.
It lands in a founder's inbox, and the reply comes from one.
Superconductors are the future.