I What is the temperature dependence of the NIRCam sensor's responsivity?

[Devin-M]

You are loosing also 50% more of the photons emitted by the sensor which otherwise would have been reflected back onto the sensor or onto some part in thermal equilibrium with the detector.

I calculated the surface area for one of the JWST sensors is approximately 12.9cm^2.
http://www.teledyne-si.com/products/Documents/TSI-0855 H2RG Brochure-25Feb2022.pdf

I used the Planck radiation formula for a 300k black body, 0.95 emissivity, 12.9cm^2 and a radiation wavelength range from 3400nm to 3600nm and get 0.00019W of radiated power for that wavelength range which is 0.034% of the total radiated power.
http://hyperphysics.phy-astr.gsu.edu/hbase/quantum/radfrac.html

I calculated the spectral responsivity of a 300k HgCdTe panel from figure 12 in this paper is 0.0003 amps from 0.00019W of 3.5 micron light.
https://www.researchgate.net/profile/Pawel-Madejczyk-2/publication/343856156_Higher_Operating_Temperature_IR_Detectors_of_the_MOCVD_Grown_HgCdTe_Heterostructures/links/6086b5a62fb9097c0c0d3442/Higher-Operating-Temperature-IR-Detectors-of-the-MOCVD-Grown-HgCdTe-Heterostructures.pdf?origin=publication_detail

I found the typical reverse leakage current of a particular Schottky diode is 0.00000065 amps with 5 reverse volts at 298K from this datasheet:
https://www.st.com/resource/en/datasheet/bat20j.pdf

When I subtract the max expected 0.00000065 amps reverse current through the Schottky diode from the 0.0003 amps forward current into the capacitor from the spectral responsivity of the HgCdTe panel from the 0.00019W of 3.5 micron photons from the 12.9cm^2 300k black body in close proximity of the 300k 12.9cm^2 HgCdTe photovoltaic panel, I calculate a net current into the capacitor of 0.00029 amps.

I assume this is a mistake since the Carnot efficiency for the setup is 0%. On which step did I mess up the calculations?

Circuit diagram for reference:

 

[WernerQH]

I’m a bit surprised by these answers as I am fully assuming the 2nd Law of Thermodynamic is inviolable and also that it forbids obtaining “useful work” by transferring energy from a cold reservoir to a hot one.

It is not helpful to think of the JWST detectors as heat engines, as photovoltaic cells designed to harvest energy from distant galaxies in the form of infrared radiation. On the contrary, much free energy is expended to amplify the weak signals. The second graph in your post #1 is quite remarkable: For a source flux of ${\rm 10~nJy = 10^{-34}~W~m^{-2}~Hz^{-1} }$ and a mirror size of $25~{\rm m}^2$ I estimate a photon flux of only $0.38~{\rm s^{-1}}$ (at frequency $10^{14}~{\rm Hz}$ and bandwith $10^{13}~{\rm Hz}$). This corresponds to a rather small photocurrent of only $6×10^{-20}~{\rm A}$. This requires a lot of amplification and an integration time of 10000 seconds to turn it into a signal!

Well I know when the photodiode detector is 300k, then the Carnot efficiency is 0% for a 300k emitting object and 0.33% for a 301k emitting object.

I find this a very strange argument. Do you think that radio transmission requires that the sender's antenna must be hotter than the receiver's?

 

[Devin-M]

It is not helpful to think of the JWST detectors as heat engines

Do you think that radio transmission requires that the sender's antenna must be hotter than the receiver's?

Well when I read about photodiodes & the internal photoemission process I learned that individual photons with the appropriate wavelength/energy kick individual electrons from valence to conduction, one by one, and that’s how the photocurrent is produced in photodiodes operating in zero bias photovoltaic mode.

When I used the Planck radiation formula for a 300k black body, 0.95 emissivity, 12.9cm^2 (same size as the JWST sensor) and a radiation wavelength range from 3400nm to 3600nm, I get 0.00019W.

I calculated the spectral responsivity of a 300k HgCdTe panel in zero bias photovoltaic mode from figure 12 in this paper is 0.0003 amps from 0.00019W of 3.5 micron light.
https://www.researchgate.net/profile/Pawel-Madejczyk-2/publication/343856156_Higher_Operating_Temperature_IR_Detectors_of_the_MOCVD_Grown_HgCdTe_Heterostructures/links/6086b5a62fb9097c0c0d3442/Higher-Operating-Temperature-IR-Detectors-of-the-MOCVD-Grown-HgCdTe-Heterostructures.pdf?origin=publication_detail

But when I recently sat down with the head of the physics department at one of the California state universities and asked if any photocurrent would be produced in a 300k detector from photons from a 300k black body, he assured me it wouldn’t, because the Carnot efficiency of a 300k black body and a 300k detector is 0%… which led me to wonder what is different about the individual 3.5 micron photons from a 299k object vs a 301k object, such that the 301k sourced 3.5 micron photons can produce some useful photocurrent in a 300k detector while the 299k sourced 3.5 micron photons can’t.

He was very busy at the time and refused to go into a discussion about quantum mechanics so I was left with some unanswered questions.

 

[Devin-M]

Mod note: posts regarding sensors and the second law of thermo split -- again.

The noise from the telescope's internal IR swamped these long wavelength observations with too much noise to get useful images.

I’m confused on this point. Useful images aside, my understanding is the 2nd Law of Thermodynamics (which I assume is inviolable) forbids obtaining useful energy from a single temperature reservoir. The “noise” prior to readout consists of a net charge forming in the in-pixel capacitor. Useful work can be done from a net charge in a capacitor. If one considers the telescope that has run out of coolant to be a single temperature reservoir, how does the net “noise” charge form in the in-pixel capacitor considering the 2nd Law and the assumed impossibility of obtaining useful work from said single temperature reservoir?

 

[Drakkith]

I’m confused on this point. Useful images aside, my understanding is the 2nd Law of Thermodynamics (which I assume is inviolable) forbids obtaining useful energy from a single temperature reservoir.

I believe the target is one reservoir, the rest of the universe is the other reservoir.

 

[Devin-M]

…Spitzer telescope had to stop long wavelength observations once it ran out of liquid helium coolant. The noise from the telescope's internal IR swamped these long wavelength observations with too much noise to get useful images.

It’s pretty easy to observe the effect even with an ordinary DSLR. I took this sequence of 5 minute exposures while my Nikon DSLR camera heated up from 35f to 70f with the lens cap on (no visible light is hitting the sensor). As the temp increases so does the noise. The difference between the Nikon sensor and the HgCdTe material that is used in NIRCam is the Nikon uses bias voltage across the photodiodes (and a has a different band gap) whereas the HgCdTe material can photo-detect infrared without bias voltage in photovoltaic mode (not necessarily in ordinary operation). I have no idea whether the HgCdTe photodiode material in NIRCam would similarly display increasing dark current with increasing telescope temperature in photovoltaic mode with no bias voltage.

https://www.speakev.com/attachments/13c208ea-b300-4a24-a496-9e9f9fae4c25-gif.156383/

 

[FactChecker]

It doesn't. That's why IR telescopes are usually cooled to a temperature well below whatever objects they will be observing and why the Spitzer telescope had to stop long wavelength observations once it ran out of liquid helium coolant. The noise from the telescope's internal IR swamped these long wavelength observations with too much noise to get useful images.

This is also why the Webb telescope shields itself not only from the Sun and Earth but also from its own heat/energy sources. It also has active cryocooling of the detectors to -448 F (I think that includes the mirrors).
The Spitzer telescope was passively cooled to -408 F (-244 C) and actively cooled to -450 F (-267 C).

 

[hutchphd]

Useful work can be done from a net charge in a capacitor.

Yes but you are creating a Maxwell's demon scenario: half the time the capacitor fluctuation goes the other way. The demon is in the detail. This has been tried in many forms without success. Usually the demon overheats and quits...

The difference between the Nikon sensor and the HgCdTe material that is used in NIRCam is the Nikon uses bias voltage across the photodiodes (and a has a different band gap) whereas the HgCdTe material can photo-detect infrared without bias voltage in photovoltaic mode (not necessarily in ordinary operation).

FYI I think the photodiode in the camera array is probably biased to improve its electronic response time (for fast motion). The silicon photodiode also works without bias.

 

[FactChecker]

Summary: Non heat IR?

There are different ranges of IR. Are they all represented as heat? If not maybe Webb telescope can read those only so our own sun isn't as much of an obstacle? Maybe xray instead?

One important point is that the more distant the stars are, the more their radiation is shifted to the infrared. The expanding universe does that. That is why the JWST is tuned to see infrared that the Hubble telescope was not able to see. x-rays are in the opposite direction.

 

[Devin-M]

Yes but you are creating a Maxwell's demon scenario: half the time the capacitor fluctuation goes the other way. The demon is in the detail. This has been tried in many forms without success. Usually the demon overheats and quits...

FYI I think the photodiode in the camera array is probably biased to improve its electronic response time (for fast motion). The silicon photodiode also works without bias.

Can you spot which step I made the math error in the following post? I know from the Carnot efficiency the answer is zero, but when I do the calculation based on the quantum efficiency of the photodiode and the leakage current of the Schottky diode, it doesn’t zero out so I know I missed a step... The circuit features a 300k HgCdTe photodiode similar to the NIRCam active material facing a 300k same-surface-area black body, a capacitor and a Schottky diode to prevent back-flow out of the capacitor.

I calculated the surface area for one of the JWST sensors is approximately 12.9cm^2.
http://www.teledyne-si.com/products/Documents/TSI-0855 H2RG Brochure-25Feb2022.pdf

I used the Planck radiation formula for a 300k black body, 0.95 emissivity, 12.9cm^2 and a radiation wavelength range from 3400nm to 3600nm and get 0.00019W of radiated power for that wavelength range which is 0.034% of the total radiated power.
http://hyperphysics.phy-astr.gsu.edu/hbase/quantum/radfrac.html

I calculated the spectral responsivity of a 300k HgCdTe panel from figure 12 in this paper is 0.0003 amps from 0.00019W of 3.5 micron light.
https://www.researchgate.net/profile/Pawel-Madejczyk-2/publication/343856156_Higher_Operating_Temperature_IR_Detectors_of_the_MOCVD_Grown_HgCdTe_Heterostructures/links/6086b5a62fb9097c0c0d3442/Higher-Operating-Temperature-IR-Detectors-of-the-MOCVD-Grown-HgCdTe-Heterostructures.pdf?origin=publication_detail

I found the typical reverse leakage current of a particular Schottky diode is 0.00000065 amps with 5 reverse volts at 298K from this datasheet:
https://www.st.com/resource/en/datasheet/bat20j.pdf

When I subtract the max expected 0.00000065 amps reverse current through the Schottky diode from the 0.0003 amps forward current into the capacitor from the spectral responsivity of the HgCdTe panel from the 0.00019W of 3.5 micron photons from the 12.9cm^2 300k black body in close proximity of the 300k 12.9cm^2 HgCdTe photovoltaic panel, I calculate a net current into the capacitor of 0.00029 amps.

I assume this is a mistake since the Carnot efficiency for the setup is 0%. On which step did I mess up the calculations?

Circuit diagram for reference:

 

[Drakkith]

Can you spot which step I made the math error in the following post? I know from the Carnot efficiency the answer is zero, but when I do the calculation based on the quantum efficiency of the photodiode and the leakage current of the Schottky diode, it doesn’t zero out so I know I missed a step...

Best guess is that you are assuming that the two systems are in thermodynamic equilibrium when they are merely in thermal equilibrium. The incoming radiation changes the internal energy of the circuit system, but does zero work (since the system isn't connected to anything to do work on). This change of internal energy is perfectly allowable. Consider a setup where you surround a piece of ice (water ice) by a container at 0C. The ice slowly heats until it reaches thermal equilibrium at its melting point. But does it melt? I think that it partially melts until thermodynamic equilibrium is reached. The system (ice) can increase its internal energy while remaining in thermal equilibrium with the container.

But I stress that this is a guess that comes from about an hour of googling and thinking and not any professional experience or education, so take it with a grain of salt.

 

[hutchphd]

Can you spot which step I made the math error in the following post? I know from the Carnot efficiency the answer is zero,

I have no idea what the question is.
The sources of the light being observed are stars at thousands K and the so first carefully define your system.
A photodiode inside a black body at temperature T will not produce photocurrent because of stimulated emission by the photodiode. This seems to be what you are modeling and Carnot is correct.

 

[collinsmark]

Can you spot which step I made the math error in the following post? I know from the Carnot efficiency the answer is zero, but when I do the calculation based on the quantum efficiency of the photodiode and the leakage current of the Schottky diode, it doesn’t zero out so I know I missed a step... The circuit features a 300k HgCdTe photodiode similar to the NIRCam active material facing a 300k same-surface-area black body, a capacitor and a Schottky diode to prevent back-flow out of the capacitor.

I believe the mistake is that you are inferring applicability from figure 12 of that paper beyond what figure 12 of that paper was meant to convey. (from you reference, https://www.researchgate.net/profile/Pawel-Madejczyk-2/publication/343856156_Higher_Operating_Temperature_IR_Detectors_of_the_MOCVD_Grown_HgCdTe_Heterostructures/links/6086b5a62fb9097c0c0d3442/Higher-Operating-Temperature-IR-Detectors-of-the-MOCVD-Grown-HgCdTe-Heterostructures.pdf?origin=publication_detail)

When the data for the figure was created, it was done so with a bright source combined with a diffraction grating such that you can shine a bright beam of light on the detector, all within the wavelength range of interest.

In other words, the irradiance reaching the surface of the photodiode was large. It was significantly larger than the radiant emittance of the photodiode.

There is an inherent assumption about interpreting figure 12 that in order to interpret it correctly, it's understood that the irradiance >> emittance of the sensor. Using slightly other terms, it's assumed that the photon flux density striking the sensor >> the photon flux density leaving the sensor.

This assumption fails completely when both the sensor, environment, and the source are all at 300 K. As a matter of fact, when both the detector, source, and everything else are at 300 K, no energy transfer takes place at all. Figure 12 does not apply. The (figure-12-like) curves that would be generated from such a situation are flat. They are drown in thermal noise.

--------------------------------

I hate to beat a dead horse here, but here we go.

If one were to ask, "Well how does the photodetector know what photons came from the source and what photons came from thermal noise?" It doesn't. It's a matter of quantity here, not quality.

When the photodetector, source, environment, etc., all at 300 K, suppose, say for example [hypothetically], 17 photons leave the detector and 17 photons arrive at the detector.

Comparatively, when a particular data point for Figure 12 was generated (with the hot source and the diffraction grating), 17 photons leave the detector but 504,237,620,347,656 photons arrive at the detector.
[Edit: disclaimer: The actual numbers here are made up. My only point is one number is much, much larger than the other.]

That's what I'm talking about. Quantity. Figure 12 from the paper doesn't have enough information in it to examine its properties at such low irradiance levels [in the case where no temperature differences exist].

 

[Devin-M]

If that's the case, the photocurrent per watt @ 3.5 microns & quantum efficiency would need to be at least a factor of ~461x times lower than predicted from Fig. 12 to be equal in magnitude to the max expected possible reverse current through the Schottky diode.

 

[Devin-M]

The second graph in your post #1 is quite remarkable: For a source flux of 10 nJy=10−34 W m−2 Hz−1 and a mirror size of 25 m2 I estimate a photon flux of only 0.38 s−1

& yet the JWST is able to image objects and charge its on pixel capacitors at less than 1 photon per second across the 25m^2 mirror over long integration times…

 

[collinsmark]

If that's the case, the photocurrent per watt @ 3.5 microns & quantum efficiency would need to be at least a factor of ~461x times lower than predicted from Fig. 12 to be equal in magnitude to the max expected possible reverse current through the Schottky diode.

Huh?

 

[Devin-M]

Huh?

The Schottky diode that prevents the capacitor from discharging has a typical leakage current of 0.00000065 amps with 5 reverse volts.

I found the typical reverse leakage current of a particular Schottky diode is 0.00000065 amps with 5 reverse volts at 298K from this datasheet:
https://www.st.com/resource/en/datasheet/bat20j.pdf

The 0.0003 amps of photocurrent predicted from Fig 12 is 461x higher than that, so Fig 12 would need to be off by a factor of 461x at low irradiance.

I calculated the spectral responsivity of a 300k HgCdTe panel from figure 12 in this paper is 0.0003 amps from 0.00019W of 3.5 micron light.
https://www.researchgate.net/profile/Pawel-Madejczyk-2/publication/343856156_Higher_Operating_Temperature_IR_Detectors_of_the_MOCVD_Grown_HgCdTe_Heterostructures/links/6086b5a62fb9097c0c0d3442/Higher-Operating-Temperature-IR-Detectors-of-the-MOCVD-Grown-HgCdTe-Heterostructures.pdf?origin=publication_detail

 

[collinsmark]

The Schottky diode that prevents the capacitor from discharging has a typical leakage current of 0.00000065 amps with 5 reverse volts.

Where does the 5 V come from? I'm still not following.

(And please don't base the source of the 5 V on something to do with Fig 12. As I've releated before, Fig 12 doesn't apply when everything is at the same temperature.)

The 0.0003 amps of photocurrent predicted from Fig 12 is 461x higher than that, so Fig 12 would need to be off by a factor of 461x at low irradiance.

Again, you can't use Fig 12 for situations where the irradiance is comparable to the emittance of the sensor (such as when the phtotodetector, the source, and all of the surroundings are ~300 K). Fig. 12 is meaningful only when the irradiance >> emittance.

 

[Devin-M]

Where does the 5 V come from? I'm still not following.

The 5v is the reverse voltage at which the Schottky diode was rated to leak up to 0.00000065 amps. This diode prevents a capacitor which has been charged (up to 5v) from photocurrent from discharging back through the photodiode.

In other words, any time the photocurrent is greater than 0.00000065 amps, an initially 0v capacitor is expected to be building up a charge.

 

[Devin-M]

Quantity. Figure 12 from the paper doesn't have enough information in it to examine its properties at such low irradiance levels [in the case where no temperature differences exist].

Figure 4(c) in this paper shows an increase in quantum efficiency in an HgCdTe detector associated with a decrease in irradiance:
https://pubs.rsc.org/en/content/articlehtml/2018/ra/c8ra07683a#imgfig4

“Fig. 4 The optical characteristics of p+-BLG/n-Hg0.8133Cd0.1867Te photodetector”

“Fig. 4c demonstrates the variation of incident power suggesting increase in Rexti due to the increase in photocurrent which decreases both QEext and NEP.”

 

[Devin-M]

Fig 3(b) shows photocurrent density with zero bias voltage scaling down linearly with decreasing irradiance from 1 W/cm^2 all the way down to 1/1000th W/cm^2 for 20 micron IR…

https://pubs.rsc.org/en/content/articlehtml/2018/ra/c8ra07683a#imgfig4

 

[Devin-M]

Why can’t they avoid the need for cooling the telescope to cryogenic temperatures by operating the detectors with zero bias voltage (photovoltaic mode)? As long as the telescope is the same temperature as the sensor, say 300k telescope and 300k sensor, how can there be any noise-generating dark current? The Carnot efficiency between the temp of the telescope and the temp of the sensor would be 0% if they are both the same temperature. Wouldn’t that reduce the cost and complexity if the need for cryogenic cooling was eliminated by eliminating the dark current?

 

[Devin-M]

Why does Figure 2 show 1mA/cm^2 of DC dark current with 0v bias voltage at 161k? It shows the dark current in amps being directly proportional to the surface area of the imaging sensor & non-zero with 0v bias & no photocurrent. For a sensor 31.7cm x 31.7cm that’s more than a full amp of DC current.

If I have a 0v bias voltage, 161k photovoltaic IR sensor in a 161k telescope with the lens cap on, I shouldn’t be able to generate any useful DC current because the Carnot efficiency is 0%…

https://www.researchgate.net/publication/345174470/figure/fig2/AS:1022731785080842@1620849667815/Dark-current-density-versus-applied-bias-measured-at-several-temperatures-for-the-T2SL.png

Fig. 2: “Dark current density versus applied bias measured at several temperatures for the T2SL MWIR 15 Â 15 detector array of 15 μm pitch and a 4.6 μm cut-off wavelength.”

https://www.researchgate.net/figure/Dark-current-density-versus-applied-bias-measured-at-several-temperatures-for-the-T2SL_fig2_345174470

https://www.researchgate.net/publication/345174470_1f_Noise_and_Dark_Current_Correlation_in_Midwave_InAsGaSb_Type-II_Superlattice_IR_Detectors/fulltext/609c2ee8299bf1259ecd763c/1-f-Noise-and-Dark-Current-Correlation-in-Midwave-InAs-GaSb-Type-II-Superlattice-IR-Detectors.pdf?origin=publication_detail

 

[Devin-M]

Also this paper Fig 4 shows about 0.4 amps dark current at 300k in a 300k enclosure for a 31.6cmx31.6cm sized detector…

“Figure 4(a) shows the dark current density vs applied bias voltage characteristic of the nBn photodetector at different temperatures ranging from 120 to 300 K. During the measurement, the device was covered by a cold shield and cooled by a temperature-controlled stream of nitrogen in the vapor cryostat. The cold shield has the same tempera- ture as the device. “

https://www.researchgate.net/profile/Arash-Dehzangi/publication/335005180_Demonstration_of_mid-wavelength_infrared_nBn_photodetectors_based_on_type-II_InAsInAs_1-x_Sb_x_superlattice_grown_by_metal-organic_chemical_vapor_deposition/links/5d4a53afa6fdcc370a80ec0c/Demonstration-of-mid-wavelength-infrared-nBn-photodetectors-based-on-type-II-InAs-InAs-1-x-Sb-x-superlattice-grown-by-metal-organic-chemical-vapor-deposition.pdf?origin=publication_detail

 

[russ_watters]

Why can’t they avoid the need for cooling the telescope to cryogenic temperatures by operating the detectors with zero bias voltage (photovoltaic mode)? As long as the telescope is the same temperature as the sensor, say 300k telescope and 300k sensor, how can there be any noise-generating dark current?

"the telescope" is a mirror. The camera isn't trying to avoid detecting the mirror it is trying to avoid detecting itself. If it isn't cooled, the whole apparatus is trying to image a background that's colder than it is.

The issue you raise and are hyper-focused on feels a lot like a typical perpetual motion machine problem: people "invent" PMMs that are just complicated enough that they don't understand them well enough to be able find the flaw and instead assume they work.

I don't understand the quantum mechanics of how my cameras work, but I do know that without cooling they operate above ambient temperature, meaning they are net emitters of heat, not absorbers, much less 2nd law violators/spontaneous coolers.

 

[Devin-M]

"the telescope" is a mirror. The camera isn't trying to avoid detecting the mirror it is trying to avoid detecting itself. If it isn't cooled, the whole apparatus is trying to image a background that's colder than it is.

The issue you raise and are hyper-focused on feels a lot like a typical perpetual motion machine problem: people "invent" PMMs that are just complicated enough that they don't understand them well enough to be able find the flaw and instead assume they work.

I don't understand the quantum mechanics of how my cameras work, but I do know that without cooling they operate above ambient temperature, meaning they are net emitters of heat, not absorbers, much less 2nd law violators/spontaneous coolers.

See Fig 4a in post #54.

“During the measurement, the device was covered by a cold shield and cooled by a temperature-controlled stream of nitrogen in the vapor cryostat. The cold shield has the same tempera- ture as the device.”

I also assume the 2nd Law is inviolable, and I also don’t understand the quantum mechanics (I was hoping someone would explain the quantum part), but the current measurement with 0v bias voltage and everything 300k (sensor & surroundings) was roughly 4*10^-3 amps “dark current” per cm^2 of sensor area which works out to about 0.4 amps with a 31.6cm x 31.6cm sensor size. I was hoping someone would explain the quantum mechanical part because the Carnot efficiency with same temp sensor and surroundings should be 0%.

I’m trying to understand the quantum dynamics, not make a perpetual motion machine or violate the 2nd Law.

 

[collinsmark]

Why does Figure 2 show 1mA/cm^2 of DC dark current with 0v bias voltage at 161k? It shows the dark current in amps being directly proportional to the surface area of the imaging sensor & non-zero with 0v bias & no photocurrent. For a sensor 31.7cm x 31.7cm that’s more than a full amp of DC current.

If I have a 0v bias voltage, 161k photovoltaic IR sensor in a 161k telescope with the lens cap on, I shouldn’t be able to generate any useful DC current because the Carnot efficiency is 0%…

https://www.researchgate.net/publication/345174470/figure/fig2/AS:1022731785080842@1620849667815/Dark-current-density-versus-applied-bias-measured-at-several-temperatures-for-the-T2SL.png

Fig. 2: “Dark current density versus applied bias measured at several temperatures for the T2SL MWIR 15 Â 15 detector array of 15 μm pitch and a 4.6 μm cut-off wavelength.”

https://www.researchgate.net/figure/Dark-current-density-versus-applied-bias-measured-at-several-temperatures-for-the-T2SL_fig2_345174470

https://www.researchgate.net/publication/345174470_1f_Noise_and_Dark_Current_Correlation_in_Midwave_InAsGaSb_Type-II_Superlattice_IR_Detectors/fulltext/609c2ee8299bf1259ecd763c/1-f-Noise-and-Dark-Current-Correlation-in-Midwave-InAs-GaSb-Type-II-Superlattice-IR-Detectors.pdf?origin=publication_detail

Measurement error given the circuit configuration?

Keep in mind when the data was measured for the plots, the photodetector was placed in an active circuit. The bias voltage was then adjusted, maybe by adjusting a potentiometer in a voltage divider circuit, or perhaps by adjusting the output voltage of a benchtop power supply -- a power supply attached to the circuit specifically for bias voltage adjustments, where the rest of the circuit was powered by a separate supply. Something like that.

By tweaking the bias voltage in the active circuit, one can sort-of simulate a passive circuit when the bias voltage is at 0 V. But any error in the method, or any minor, non-ideal property of the circuit that still provides power to some parts of the circuit, could manifest itself as non-zero measurement when the bias voltage is supposedly "zero."

But what I can all but guarantee is that they didn't yank out the photodetector and place it in a passive circuit just for the special situation of dark current at 0 bias voltage.

You can assume though that if the sensor was actually placed in a passive circuit, such that the photodetector, other components in the circuit, its enclosure, and everything else in the vicinity is all at the same temperature, the dark current density would measure 0.

 

[Grinkle]

It shows the dark current in amps being directly proportional to the surface area of the imaging sensor

I'm not sure this is supported by the graph, it may be, but if the data was collected on a single device (or multiple identical devices) then one cannot see the shape of the curve for the dark current vs active area. Dividing the measured current by the device area all by itself and using this as the Y axis scale does not provide good insight to the area vs dark current shape. Having multiple curves, as we see with the changing temperature does provide this (for temp, and would, for area, if available). Labelling the Y axis this way does imply a linear relationship, and if I were involved in a review of this particular graph, I'd poke at that. As you say, it gives pretty suprising results when one scales it a bit.

I agree with @collinsmark . In my experience reviewing and interpreting semi-conductor measurement data, when one sees a result that looks to be in conflict with physics, one starts from an assumption of measurement non-ideality and tries to identify where its coming from, then one assesses whether it matters enough to go and spend time and $ to attempt an improved measurement. At some point, improved measurements are judged to no longer be value-added, and one calls it a day.

Edit:

Considering the graph a bit more, I speculate that the author was wanting to communicate current density specifically as the quantity of interest in the device being measured, and not intending to imply that it scales linearly with area, and this why the Y axis is reported as it is.

 

[Devin-M]

Here’s another one I found, for which the graph shows more dark current per cm^2 at 300k with 0.0v applied bias voltage than 170k with -0.8v bias voltage:

Fig. 3. (a) Dark current density versus applied bias voltage char- acteristic of the photodetectors as a function of temperature.
https://www.researchgate.net/profile/Arash-Dehzangi/publication/322833826_nBn_extended_short-wavelength_infrared_focal_plane_array/links/5a73e8f1a6fdcc53fe148ed3/nBn-extended-short-wavelength-infrared-focal-plane-array.pdf?origin=publication_detail

 

[Devin-M]

More dark current per cm^2 at 0v applied bias voltage at 300k is indicated in this graph than at 120k, -1.0v bias voltage…

https://www.researchgate.net/profile/Junqi-Liu-3/publication/267340100_Room_temperature_quantum_cascade_detector_operating_at_43_m/links/579709df08aec89db7b86bd8/Room-temperature-quantum-cascade-detector-operating-at-43-m.pdf?origin=publication_detail