(Low technobabble, one diagram only, no formulas.)
Human eyes have two types of cells that can detect light: cone cells und rod cells. There are three types of cone cells for different colors but they have a lower sensitivity so they can only detect relatively bright light. Rod cells have a much higher sensitivity (up to 100 times) but they ‘reload’ much slower when ‘used up’ by bright light.
The biological stuff is a bit complex. But we are only interested in the outcome so we can cut several corners here.
The most useful diagram for this is the “Spectral luminous efficacy”. The diagram (taken from Wikipedia) shows the sensitivity of the average human eye for light of different wavelengths.
These curves were generated by experiments and they are standardized by the CIE. There is no way to measure them. Your personal experience may vary depending on the remaining transparency of your eye lenses and other factors.
The diagram shows curves for the sensitivity of the cone cells (lower curve at right) and the rod cells (higher curve at the left) for different wavelengths.
(If you are not familiar with reading y,x-diagrams: Pick a number at the horizontal axis, then go straight up until you hit the curve. From this point go horizontally towards the vertical axis to get the corresponding value. So for 450 nanometer (nm) wavelength on the left curve you get a factor of 800 lm/W. For each value at the horizontal axis there is a value at the vertical axis. You could write this as a large two-column table but the graphical view of a curve gives you a better understanding of how things go together.)
Have a look at the right curve K (λ). It starts at 425 nm indigo blue, almost ultraviolet.
The peak sensitivity is at 555 nm (green). Of all colors green is the one humans can see best.
It ends at 700 nm deep red, almost infrared.
The higher curve K’(λ) on the left shows the sensitivity of the rod system that is used to see at low light conditions.
This curve is much higher because the rods are more sensitive. The curve is also shifted towards lower wavelengths, so more blue, less red.
It starts at 400 nm, almost ultraviolet. The maximum of this curve is at 505 nm (blue-green). It ends at 625 nm, red.
.
How to preserve your low light vision (“night vision”)
If you are using any wavelength of light that the K’(λ) curve covers you will trigger the rod cells. If the light is so bright that it overloads the rod cells they will need a long time to reload. The more sensitive system is OFF until then and you completely lose your ability to see at low light.
There is only one area of wavelength that triggers the cone cells but not the rod cells: Above 625 nm, red light.
(The colors of the rod cell curve K’(λ) are a bit misleading. It shows the color of light but not what you see. There is only on type of rod cells. It is triggered by several different wavelengths but this triggering is only on/off, no information about the color. Humans cannot see colors with rod cells. If you see any color at low light you see it with your cone cells. If it is not red it may be so bright that it already reduces your low light vision.)
.
In theory you would need very little green light to see at low light because the rod cells are so sensitive to it. But this is not how it is usually used.
Bright green light is the best way to temporarily destroy your low light vision.
.
This is how I understand the topic after reading some papers. If I am wrong just tell me.