How can I properly use PWM to observe hopefully >= 2x normal efficacy, seeing the theoretical max of 1700 lm/W scotopic efficacy?

Question

In a Wikipedia article on "Scotopic Vision", there is a section that states:

The maximum scotopic efficacy is 1700 lm/W at 507 nm (compared with 683 lm/W at 555 nm for maximum photopic efficacy).[11] While the ratio between scotopic and photopic efficacies is only around 2.5 counted at peak sensitivity the ratio increases strongly below 500 nm.

I want to build a device that achieves this astounding 1,700 lumen per watt efficacy.

PWM keeps the luminance above a certain level when it's on, using persistence of vision to cause the appearance of dimness, but I'm not sure that with the scotopic vision, which uses the rods instead of the cones of our sight, that PWM will not cause a problem, preventing attaining the 1,700 figure in practice.

---

EDIT 1: (Feedback from a comment)

The wavelength of 507 nm is in, or very close to, the cyan color, so green light will probably not be very useful in this discussion. (Yes, the other wavelength quoted, 555 nm, is a green, but that was for maximum photopic efficacy, not scotopic efficacy -- we are trying to achieve maximum scotopic efficacy in this question).

---

EDIT 2: (Feedback from an answer)

I may use a small solar cell and a supercapacitor, if the energy is small enough, so the amount of energy really does matter.

Keep in mind that it's also about being undetected, both for spouses, and other needs for being undetected, TBD.

---

EDIT 3: (Answered my own question)

Please see my answer.

I will consider other answers that are better.

---

I have decided that this question is hard to answer because I don't have measurements, and won't, until I order some green and cyan LED's, at which point I'll probably open up a new question anyway.

So, I will implement a flashlight, where I can both dim by changing the PWM duty cycle, and also dim by changing the amplitude (magnitude?) of the PWM peaks. I will then judge that I have achieved success if I can read using cyan 507nm, at half the power of reading using green 555nm, adjusted for the lm/w figures of the individual LED's actually used. I'll update when I find out.

By the way, the white LED's may win in the end simply because they start out with such great lm/W figures.

Thanks all. (Especially Tim.)

---

Question clarification from comments:

I do not want to use a laser, although I heartily thank you because I now understand why a laser would come closer to the 1700 theoretical. (Does an inexpensive semiconductor laser at 507 nm exist?) Here's what I want to do:

Using a Cyan LED light source, having a center wavelength reasonably close to the optimal 507nm, I only want to be able to read using hopefully less than half the power (half the lumens, which should appear brighter.) So this is reflected hopefully in the new title for the question.

Answer 1

It is not possible to actually build a 1700 lm/watt device in practice since power supplies, LEDs and even wires are not 100% efficient. A more reasonable figure given available LED technology would be 500-1000 lm/watt.

However, from your comments I think even aiming for this efficiency is pointless. At the low current you are considering, even a small battery will last for hundreds or thousands of hours, probably far longer than required. Trying to extend battery life from ten times longer than makes sense to 50 times longer than makes sense is unlikely to be worthwhile, especially when it will make your design needlessly complicated and expensive. An inexpensive 100 or 150 lm/watt white LED is likely fine for your application and will be much more pleasant to read with.