I had heard that when an LED is submerged in liquid nitrogen (LN2) it changes color. After the PTSOS workshop last January I had some leftover LN2 and decided to see for myself. I replaced the incandescent bulb in a Mini-Maglight with a green LED and submerged it in a Dewar of LN2. This was the result:
I was pleased when the color change was very apparent as the audio track confirms. My expectation was that the green light would change to a lower energy wavelength like yellow. I had a little familiarity with how LEDs work and knew the energy of the emitted photons was based on the band gap energy. I thought a colder material would have a lower band gap energy. My observation confirmed what I would find out later to be an erroneous expectation. Fortunately, after I made the video I got out my Red Tide spectrometer and recorded the spectrum of the LED at room temperature (22 degrees C) and in LN2 (-196 degrees C). I took a quick look but didn’t study it because it was a Friday and I had spent enough time fooling around, er, I mean experimenting, in my class room. I tweeted out the video and went home.
 |
| Still image of original Tweet, link to entire thread |
The next day the tweet was getting a lot of attention. Several people responded that the color should blue shift when submerged in LN2. Others sent me related links and I started reading up on the subject. These confirmed that the band gap energy should increase as the temperature decreases although one source indicated that it could go the other way for some materials. Most of these sources invoked quantum mechanics in their explanations, much of it went over my head. After reading many sources I found an explanation that would satisfy a high school student (or teacher!). But first they would need to understand how p-n junctions and LEDs work. If you want to wade through all the replies to my tweet, here is the link: https://twitter.com/kilroi22/status/962145895805476864
https://twitter.com/kilroi22/status/962145895805476A p-n junction is a piece of semiconductor divided into two regions. The p-type side (p for positive) has impurities added that cause the semiconductor to accept electrons. The n-type side (n for negative) impurities cause the semiconductor to give up electrons. At the junction between the two materials, the electrons from the n-type side are attracted by the p-type side and drift into it. This leaves the region on the n-type of the junction side lacking electrons and the region on the p-type side with extra electrons. A region lacking electrons can be described as having positive “holes” (see diagram above). Now the remaining electrons on the n-type side face a potential barrier from the electrons on the other side of the junction. An applied voltage can get them to flow to the p-type side if it is large enough to overcome the potential of the barrier. The energy required for an electron to overcome the potential is the “band gap energy”. An LED is a semiconductor with a p-n junction. If enough voltage is applied to the LED, the electrons receive enough energy to cross the junction, then drop back down in energy, giving off light. This is analogous to an electron being excited to a higher energy level in an atom, then dropping back down, giving off a photon of light having a very specific energy or wavelength. An LED differs because there is a spread of possible photon energies that can be emitted, centered on what is most probable, the band gap energy. The color or wavelength of an LED is often, but as we will see, not always the same as the wavelength corresponding to the bang gap energy. The figure below shows the spectrum for a green LED at room temperature. The peak wavelength is 571 nanometers (nm). This is right on the boundary between what is perceived by the human eye as green or yellow. The overall green color that is seen is a result of all the photons coming from the LED. Since the human eye peaks in color sensitivity at about 555 nm, the green photons to the left of the peak have more of an effect on what is seen than those on the right.
 |
| Spectrum of green LED at room temperature using Ocean Optic’s Red Tide Spectrometer |
I naively thought that cooling down the LED would “shrink” the band gap energy, causing the most probable wavelength to be of lower energy. This would mean it shifts to toward the red end of the spectrum. When I saw the color change from green to yellow, it confirmed my expectation. After doing some research, I learned that my expectation was wrong and there should be a blue shift. This is because at lower temperatures, the electrons on the n-type side have a lower initial energy from the random thermal motion in the material. This makes it harder for them to overcome the band gap energy much like it is harder to jump over a ditch while standing still versus taking a running start. My video seemed to refute this explanation so I took a closer look at the spectrum of the LED at room temperature and in the LN2. (see figure below) The peak wavelength in LN2 did show a tiny blue shift. It went from 571 nm to 568 nm. The Red Tide spectrometer has a 1 nm resolution but I would say this is essentially unchanged. The main difference between the two is the narrowing of LN2 spectrum. The room temperature spectrum has a larger fraction of green photons coming from it. The LN2 spectrum is depleted of the green photons, allowing yellow to dominate. There also are what appear to be some nitrogen absorption features in the LN2 spectrum.
 |
| Spectrum of green LED at room temperature and in LN2 |
I now had a new question, why does the spectrum narrow? The temperature change also affects the number of photons emitted. Being submerged in LN2 selectively suppresses emission of green photons over yellow. Why this occurs I am not sure but it is definitely happening. This experience left me with other questions. What happens with other color LEDs? Will an infrared LED blue shift enough to become visible? Is the amount of blue shift dependent on the energy band gap? What would happen to a laser in LN2? When can I get some more LN2 and have a chance to try this? My chance came during the Fusion/Astrophysics Teacher Research Academy workshops I conduct at Lawrence Livermore National Laboratory. At the end of the first day we had some extra time and I brought out a Dewar of LN2 and we proceeded to immerse various light sources in it. The red laser pointer did not appear to change color but it eventually stopped working. It did recover after warming back up. An infrared LED appeared to show a very dim red light but it was hard to tell through the bubbling LN2. The ICE LED Strip that contained blue, green, yellow, orange, red, and infrared LEDs was more impressive. It survived a lengthy immersion allowing us to see the blue dim, but stay the same color, the green turn to yellow, both the yellow and orange turn to green, the red get slightly red-orange, and no sign from the infrared. None of the video turned out well and the teachers gathered closely around the Dewar made measurements difficult. I saved some LN2 to try again later by myself.
I finally got a chance the week after the LLNL workshops ended. I used ring stands and clamps so I could immerse the LED strip and record video hands-free. This allowed me to carefully measure the spectra. Below is the video in three parts, the initial immersion, after the LED strip reached equilibrium, and a sped-up video of the LED strip warming back up after removing it.
 |
| Frame grabs from video showing LED at room temperature and submerged in LN2 |
The video shows the same color changes that we observed in the workshop. No sign of visible emission from the infrared LED was seen. The infrared blocking filter on my iPhone prevented any infrared emission from showing on the video. The infrared emission does show on the spectrum as seen below. It shows a blue shift when immersed in LN2. The peak emission shifted from 934 nm to 904 nm. It would need to shift to at least 750 nm to be visible to the human eye. Some infrared LEDs peak at 850 nm. It is possible that they would shift enough to be visible but I doubt it.
 |
| Spectrum of infrared LED at room temperature and in LN2 |
The blue LED also shows a blue shift of the peak emission from 445 nm to 422 nm. However, this is due to a change in the relative intensity of the two peaks in each spectrum as shown in the figure below. The room temperature spectrum has a peak wavelength of 445 nm but shows a secondary peak at 425. The LN2 peak is at 422 nm with a secondary at 445 nm. I think the relative change in peak intensity has a similar cause as the green LED appearing yellow. The lower temperature is suppressing emission of certain energy photons more than others. Note: The intensity values on the y-axis of the spectrum graphs are not relevant because they depend mostly on how I aligned the fiber optic cable collecting the light. The blue LED dimmed noticeably so its intensity should be lower if everything else was equal. The small peak at about 570 nm is coming from the green LED that is adjacent the blue.
 |
| Spectrum of blue LED at room temperature and in LN2 |
The remaining 4 spectra are shown below. They all show a significant blue shift as they should according to the references I consulted. This green spectrum is very similar to the one I showed and discussed earlier. They both show a very slight blue shift and a depletion of emission of green photons. The small peaks on the sides are from adjacent LEDs on the strip.
 |
| Spectrum of green LED at room temperature and in LN2 |
 |
| Spectrum of yellow LED at room temperature and in LN2 |
 |
| Spectrum of orange LED at room temperature and in LN2 |
 |
|
| Spectrum of red LED at room temperature and in LN2 |
The red LED spectrum confirms the slight change from red to red-orange that was visible to the eye and in the video. Below is a data table of the wavelengths of the peak emissions as measured by the Red Tide spectrometer and Logger Pro software.
Notice the peak wavelengths for the yellow and orange LEDs are almost the same for both room temperature and in LN2. This is evidence that the color seen by the eye is more dependent on the overall emission from the LED than on the peak. The yellow LED at room temperature is not as narrow as the orange LED but has a little bit more emission from the yellow and green where the human eye is more sensitive. In LN2 their spectra and are almost identical and their visual appearance is almost the same green color.
If you are curious about the effect of temperature on LEDs, I have copied my references below. If you know or learn something relevant to this topic, please leave it as a comment. I am sure the next time I get some LN2 I will have a list of new things to try.
https://ecee.colorado.edu/~bart/book/eband5.htm
https://wiki.brown.edu/confluence/display/PhysicsLabs/7A30.10+LED+in+Liquid+Nitrogen
https://physics.stackexchange.com/questions/80513/how-does-temperature-affect-a-semiconductor-band-gap
https://www.researchgate.net/post/How_are_the_wavelength_of_LEDs_dependent_on_temperature
Why does LED glow brighter in Liquid Nitrogen but only for a moment? from askscience
https://io9.gizmodo.com/watch-an-led-light-change-color-in-liquid-nitrogen-1574982405
https://rebrn.com/re/changing-the-color-of-an-led-by-changing-simply-cooling-it-in-li-2844214/
https://www.osapublishing.org/josk/abstract.cfm?uri=josk-19-3-311