Hello everybody!
Today's situation about letting semiconductors superconduct (please correct if this is false, as I don't read news regularly):
- For a long time, it didn't work, even at seriously low temperatures
- Maybe 2 years ago, people from Orsay (hi there) achieved it at 0.1K with a boron doping of about 10% of all silicon atoms.
That is, silicon became a superconductor only as the carrier concentration (holes in this case) became comparable with a metal - a doping level never used in normal semiconductor technology.
Now, I wonder whether another way can achieve the same carrier concentration.
Make a heterojunction. Dope P+ (just normally P+) the wide gap material, and the holes will fall in the narrower gap material nearby. This narrow gap layer can be thin and can also be topped by another P+ wide gap layer.
The carrier concentration in the narrow gap layer will be huge, because the step in the bands concentrates the carriers from the wide gap. Even more so at low temperatures! In fact, only the band bending limits the number of carriers transferred to the narrow gap; use a thin layer, and it will be flooded.
A few advantages:
- A second experiment to compare with the first one
- Looks simpler than the alloying technique developed to preserve the monocrystal at such dopings
- Hence it could be tried at various carrier concentrations. Measure it by an equivalent of plasma resonance?
- The crystal is sound and its properties better known, as doping levels are civilized
- Could we switch the superconductivity just by adjusting the carrier concentration with some junction or gate?
- Similar to some theories about YBaCuO
Comments? Impossible? Already done?
Marc Schaefer, aka Enthalpy
And here a few figures. As usual, ideas look less favourable once you put figures one them than with a qualitative approach... This is why readers are so wary about proposals without figures, aren't they? Anyway, this idea has survived the computation.
My special thanks to the highly helpful
http://www.ioffe.ru/SVA/NSM/Semicond/Si
http://www.ioffe.ru/SVA/NSM/Semicond/Ge
http://www.ioffe.ru/SVA/NSM/Semicond/GaAs
http://www.ioffe.ru/SVA/NSM/Semicond/AlGaAs
very nice of them, I really appreciate.
In the first case, I take a (really) thin layer of GaAs between two AlAs (100% Al and 0% Ga). The step is 0.46eV between the valence bands.
AlAs is doped P+ at 1e20/cm3=1e26/m3 (this exists, do it as you can, more is better). With a permittivity of 10, the junction depletes 2.2nm of each AlAs side.
Then, a 0.56nm layer of GaAs gets 8e20/cm3=8e26/m3 holes concentration.
Comments:
- Yes, I used band parameters from low dopant and carriers concentrations and from thick layers. I somebody does it better, welcome.
- Dopant levels use to be 30meV above the valence band, which is much at low temperatures. I believe it doesn't matter, because the dopant level is 0.46eV below the GaAs valence band, so the holes won't go back to AlAs when it's cold.
- The GaAs layer is really thin. But the hyperdoped Si layer that superconducted was also 2 to 10 atoms thin.
- The hole concentration is 6 times less than the B concentration in Si, which isn't the hole concentration. And the dopant concentration in AlAs can hopefully be increased.
- In case somebody believes it matters: the effective valence band density of states of GaAs is half that of Si; Ge is one-fourth.
- Holes are about as heavy in GaAs as in Si and a bit lighter in Ge.
- One could try to dope N+; many people believe holes are more favourable to superconductivity.
In a second case, use Ge instead of GaAs - but still dope AlAs P+, do it as you can. The step between the valence bands is now 1.01eV. With the same parameters as above, Ge gets 1.2e27/m3 = 1.2e21/cm3 holes, or 4 times less than the B concentration in Si.
My special thanks to the highly helpful
http://www.ioffe.ru/SVA/NSM/Semicond/Si
http://www.ioffe.ru/SVA/NSM/Semicond/Ge
http://www.ioffe.ru/SVA/NSM/Semicond/GaAs
http://www.ioffe.ru/SVA/NSM/Semicond/AlGaAs
very nice of them, I really appreciate.
In the first case, I take a (really) thin layer of GaAs between two AlAs (100% Al and 0% Ga). The step is 0.46eV between the valence bands.
AlAs is doped P+ at 1e20/cm3=1e26/m3 (this exists, do it as you can, more is better). With a permittivity of 10, the junction depletes 2.2nm of each AlAs side.
Then, a 0.56nm layer of GaAs gets 8e20/cm3=8e26/m3 holes concentration.
Comments:
- Yes, I used band parameters from low dopant and carriers concentrations and from thick layers. I somebody does it better, welcome.
- Dopant levels use to be 30meV above the valence band, which is much at low temperatures. I believe it doesn't matter, because the dopant level is 0.46eV below the GaAs valence band, so the holes won't go back to AlAs when it's cold.
- The GaAs layer is really thin. But the hyperdoped Si layer that superconducted was also 2 to 10 atoms thin.
- The hole concentration is 6 times less than the B concentration in Si, which isn't the hole concentration. And the dopant concentration in AlAs can hopefully be increased.
- In case somebody believes it matters: the effective valence band density of states of GaAs is half that of Si; Ge is one-fourth.
- Holes are about as heavy in GaAs as in Si and a bit lighter in Ge.
- One could try to dope N+; many people believe holes are more favourable to superconductivity.
In a second case, use Ge instead of GaAs - but still dope AlAs P+, do it as you can. The step between the valence bands is now 1.01eV. With the same parameters as above, Ge gets 1.2e27/m3 = 1.2e21/cm3 holes, or 4 times less than the B concentration in Si.
Superconducting chips never will be faster than usual chips. In sense of energy consumption, not existation of superconducting transistors, long distance between controled usual chip and separeted suporconducted chip, this all makes latency and enourmouse power consumption.
Could a metal be more efficient than a heterojunction to inject holes in a semiconductor?
GaAs has its valence band 5.49eV below vacuum. Ge has it 4.66eV below vacuum. Pt is 5.65eV below vacuum, Ir and Re are to be 5.75eV for some crystal orientations. The Handbook of Chemistry and Physics (and thus Wikipedia) also puts Se at 5.9eV but this looks like a mistake.
So what about putting a layer of Pt over GaAs or Ge, maybe with a monolayer of AlAs buffer between them? The metal would act as an electron acceptor with about 100% concentration.
Ir superconducts at 0.11K, Re at 1.7K (but 1.7 to 7K as a thin film), and I've no data about Pt. Well, if superconductivity disappears in the metal, it would also be a clue.
GaAs has its valence band 5.49eV below vacuum. Ge has it 4.66eV below vacuum. Pt is 5.65eV below vacuum, Ir and Re are to be 5.75eV for some crystal orientations. The Handbook of Chemistry and Physics (and thus Wikipedia) also puts Se at 5.9eV but this looks like a mistake.
So what about putting a layer of Pt over GaAs or Ge, maybe with a monolayer of AlAs buffer between them? The metal would act as an electron acceptor with about 100% concentration.
Ir superconducts at 0.11K, Re at 1.7K (but 1.7 to 7K as a thin film), and I've no data about Pt. Well, if superconductivity disappears in the metal, it would also be a clue.
On Si, GaP (the fcc variant) is a candidate for a heterojunction. 0.89eV step in the valence band, permittivity 11. Again, if you want to dope P+ the GaP layer near Si, do it as you can.
Fun: some papers describe amorphous Si0.7B0.3 as a wide bandgap hole emitter for silicon heterojunction bipolar transistors... Except for "amorphous", this would let the sandwich at Orsay look like a heterojunction injecting holes in Si rather than a heavily doped Si layer.
Superconducting Pt: found a paper (from 1984!) telling it doesn't exist. Only Ru, Os and Ir are to superconduct among the Pt group. But for superconductivity, 1984 is prehistory. More recent papers tell Pt powder, not bulk, superconducts below 1mK and 70µT. This is a good start: if AlAs/GaAs/Pt or AlAs/Ge/Pt or GaP/Si/Pt superconducts at 0.1K, then it's likely the semiconductor.
Fun: some papers describe amorphous Si0.7B0.3 as a wide bandgap hole emitter for silicon heterojunction bipolar transistors... Except for "amorphous", this would let the sandwich at Orsay look like a heterojunction injecting holes in Si rather than a heavily doped Si layer.
Superconducting Pt: found a paper (from 1984!) telling it doesn't exist. Only Ru, Os and Ir are to superconduct among the Pt group. But for superconductivity, 1984 is prehistory. More recent papers tell Pt powder, not bulk, superconducts below 1mK and 70µT. This is a good start: if AlAs/GaAs/Pt or AlAs/Ge/Pt or GaP/Si/Pt superconducts at 0.1K, then it's likely the semiconductor.
OK, I will consider the diamond field emitter for the THz devices. 5 times more current than the competitors, will bring me 25 times more THz power.
Hello Enthalpy, et al.
Here is a paper that may interest you on Electroosmotic Pumps. These are pumps that use charge to move fluids in micro-channels. What if the "fluid" was actually made up of only electrons like those that are present on N-type diamond. Would this make a "superconducting" channel?
Here is a paper that may interest you on Electroosmotic Pumps. These are pumps that use charge to move fluids in micro-channels. What if the "fluid" was actually made up of only electrons like those that are present on N-type diamond. Would this make a "superconducting" channel?
Correction to my last post about diamond field emitters.
The linked paper shows an anode current of 160µA but "forgets" to tell it was obtained from 64*64 microspikes. A more complete PhD work from the same team shows that they obtain some 0.5µA from one spike, which is about as little as a (cold) carbon nanotube or a tungsten (cold) microspike deliver. Diamond is only more resilient to vacuum pollution, already a nice advantage.
Still far too little current to make any HF power. And at micrometric source size, hot spikes (W+ZrO, or better LaB6) deliver much more current, up to 20µA per spike.
People wanting to make gate-controlled valves should replace the anode by a photodiode, for instance a few nm aluminium coating a GaAs diode. They would get 100 times more current, or 2mA for a single hot spike, with very little added response time, and the gate-facing photodiode electrode could serve as an electric shield between the control gate and the output electrode. With such a current, the output impedance would be usable for HF.
The linked paper shows an anode current of 160µA but "forgets" to tell it was obtained from 64*64 microspikes. A more complete PhD work from the same team shows that they obtain some 0.5µA from one spike, which is about as little as a (cold) carbon nanotube or a tungsten (cold) microspike deliver. Diamond is only more resilient to vacuum pollution, already a nice advantage.
Still far too little current to make any HF power. And at micrometric source size, hot spikes (W+ZrO, or better LaB6) deliver much more current, up to 20µA per spike.
People wanting to make gate-controlled valves should replace the anode by a photodiode, for instance a few nm aluminium coating a GaAs diode. They would get 100 times more current, or 2mA for a single hot spike, with very little added response time, and the gate-facing photodiode electrode could serve as an electric shield between the control gate and the output electrode. With such a current, the output impedance would be usable for HF.
I've had a look at what vacuum valves with gate control and micrometric size would be capable of.
With 2µm transit between a +50V gate and a +250V anode, the flight time is about 300fs, which would allow some 400GHz operating frequency. No deep THz, but still in the upper limit of transistors and gyrotrons.
Key to extracting a significant HF power is a resonating anode with high Q. I take here a central pillar, 75µm high and 600µm in diameter, between two disks of 1000µm diameter. All voltages, currents, fields having a circular symmetry. Made of gold, the resonator has an estimated impedance of 18 mohm +j15 ohm -j15 ohm. the electron beams hitting the tip of a disk see a parallel impedance of 13 kohm, very favourable.
Now, to get an interesting current, take 628 hot spikes made of W+ZrO or of LaB6, spaced 5µm under the tip of the disk. Each emits 20µA, totalling nice 13mA max or p-p. Put a shared gate on top. Put a second gate over the first one to reduce feedback, this would be very nice.
Then, 6.5mA pk creates 42V pk at the load-matched anode (not bad for 250V supply) and 65mW HF power for 3W supply. Nice to make a small data transmitter, possibly better than transistors in this function.
Alternately, one can make a wave circle around the anode (and the gate of course) so that many operating frequencies are possible, or propagate in a guide to build a distributed amplifier with wide bandwidth.
With 2µm transit between a +50V gate and a +250V anode, the flight time is about 300fs, which would allow some 400GHz operating frequency. No deep THz, but still in the upper limit of transistors and gyrotrons.
Key to extracting a significant HF power is a resonating anode with high Q. I take here a central pillar, 75µm high and 600µm in diameter, between two disks of 1000µm diameter. All voltages, currents, fields having a circular symmetry. Made of gold, the resonator has an estimated impedance of 18 mohm +j15 ohm -j15 ohm. the electron beams hitting the tip of a disk see a parallel impedance of 13 kohm, very favourable.
Now, to get an interesting current, take 628 hot spikes made of W+ZrO or of LaB6, spaced 5µm under the tip of the disk. Each emits 20µA, totalling nice 13mA max or p-p. Put a shared gate on top. Put a second gate over the first one to reduce feedback, this would be very nice.
Then, 6.5mA pk creates 42V pk at the load-matched anode (not bad for 250V supply) and 65mW HF power for 3W supply. Nice to make a small data transmitter, possibly better than transistors in this function.
Alternately, one can make a wave circle around the anode (and the gate of course) so that many operating frequencies are possible, or propagate in a guide to build a distributed amplifier with wide bandwidth.