Making an Ultra-Efficient Yellow-Green LED Lamp

Updated 2/16/2023.

CAUTION - This involves hacking/homebrewing with possible hazards. I disclaim adequacy/completeness/etc. of hazards/safety information for the following hacking/homebrewing. Do anything mentioned below only at your own risk!

Here are ways to make yellow-green LED lamps that have an overall luminous efficacy of somewhere around 160 to 230 possibly 270 lumens/watt by surrounding (or in some cases covering) a suitable white or blue LEDs!

UPDATE 2/16/2023: The most efficient modern blue LEDs good for this may achieve 265-270 possibly 300 lumens/watt with one of the schemes below, maybe with modifications. The most efficient white LED that I am aware of with an efficiency claim that I consider reasonably credible has claimed luminous efficiency / luminous efficacy of 229 lumens out per watt in, and I expect slight improvement to be achievable by using a green-fluorescing yellow plastic or liquid.

The Plastic Method
The Liquid Method

The Plastic Method

You will need some of this plastic. There are various kinds that are yellow and glow yellow-green in a manner resembling that of alkaline solutions of fluorescein (or of salts thereof). You can get acrylic with such a tint at some of those plastics shops that artists and signmakers go to.

(Stuff on using other plastics deleted on 9/23/2001 due to repeated unsuccessful results along with high risk of fire/burns/fumes.)

UPDATE 2/14/2023: The below plastic rod hacking is described in a way applicable for low power LEDs. The modern LEDs with best prospects for extreme efficiency are not low power ones.

UPDATE 5/11/2009: Beware that I now know of two visibly discernable (at least to me) variants of green-fluorescing yellow acrylic. One has fluorescence being "lime green", while the other has its fluorescence being a more yellowish shade of green like the color of most "old tech"/"cheap" green LEDs. I recommend the one with the more greenish fluorescence.

My best results so far were achieved by cutting a piece of 3/8 inch or 1/2 inch acrylic rod into a piece maybe 3/4 inch long and drilling a hole into one end of this piece and gluing the LED into the hole. You may want to use a smaller size LED such as T1 (3 mm) and/or sand/grind down the LED to a smaller size to enable use of a smaller hole. Acrylic cracks easily when drilled. Drill gently with a higher speed drill to minimize the chance of breaking the rod.


Smooth transparent fluorescent plastic will trap much of the fluorescence by total internal reflection. Outgoing photons that hit the surface at an angle more parallel to the surface than some critical angle around 45 degrees have a 100 percent rate of being reflected back in and have a high risk of getting absorbed before finding a way out. So you need to roughen up the outer surface with sandpaper to allow more of the fluorescence to escape. Also, the light from the fluorescence and the light from the LED chip will take different paths and cause an uneven color unless you make the LED lamp diffused enough to make the light from both mechanisms go everywhere.

I would like to add/clarify on 5/11/2009 that I recommend filling the airspace between the LED and the drilled-out acrylic rod with epoxy or with acrylic casting resin. This reduces internal reflections of kinds that tend to favor losses.

UPDATE 9/16/2001 - I actually made an "LED fluorescent lamp" by drilling a 3/8 inch acrylic rod and gluing a small LED into it! Reddish magenta plastic was disappointing for ability to utilize blue light for purposes of creating a "purple LED lamp". But the yellow stuff that fluoresces green looks especially promising.

Note that fluorescent acrylic ("Plexiglass" and the like) is often not good at efficiently fluorescing from UV the way it sometimes is good at fluorescing from some visible wavelengths!


1. Use blue LEDs instead of white ones to get more of a lime-yellow-green instead of a chartreuse-greenish-yellow.

2. Use white or blue LEDs with shorter peak wavelength around 450-460 nm as opposed to longer around 470 nm.

How an LED Lamp Manufacturer Would Do This:

I think the real way to do this is to tint the epoxy used to make the LED lamp body with a suitable fluorescent dye as well as a diffusing agent to make the LED lamp translucent ("milky") instead of transparent.

Another method would be putting a small blob of *translucent* (as opposed to transparent) fluorescent plastic (or powdered fluorescent plastic or a mixture of powdered fluorescent plastic and epoxy) into the die cup of the LED lamp and over the semiconductor die before molding the main epoxy body over all of this. Just don't cook the semiconductor die molding on molten fluorescent plastic! WARNING - the fluorescent dye may degrade in unacceptably short time if it is close to the LED die and is illuminated very intensely.

The Liquid Method

More work such as optical configuration details needs to be done here as of 2/16/2023.

A high power or mid power white or blue LED can be surrounded by a fluorescent dyed liquid that absorbs blue light and fluoresces green or yellowish green light, to achieve increase of lumens out per watt in. I expect small improvement of white LEDs by doing this. I expect even more lumens/watt from doing this with a royal blue LED. I consider 270 lumens out per watt in being achievable if the LED is Cree XPGDRY-L1-0000-00601.

As for what liquid and what dye to use: As of 2/15/2023: I favor one of the lubricants that get mixed with refrigerants in refrigeration systems such as air conditioning systems as the liquid base. These lubricant liquids are typically PAG (which comes in at least three viscosity grades) or POE. The dye that I favor using with PAG or POE is a naphthalimide type "UV dye".

As of 2/15/2023, I have done some testing of a PAG lubricant and a POE lubricant with two naphthalimide "UV dyes" that are slightly different from each other.

Fluorescein (or its sodium salt uranine) in water with a small amount of alkaline is well known to fluoresce a yellowish green from blue light. I have yet to do testing here with this, because fluorescein has absorption that is not very great at wavelengths much shorter than its absorption peak around 490 nm.

As for some statistics of test results I achieved so far with naphthalimide type UV dyes in PAG and POE type refrigeration system lubricants:
Lumens out per watt out has varied from 420 to 499, with the exciting light source being a Cree XPGDRY-L1-0000-00601 LED with 350 mA of current. And, I noticed using greater effective dye concentration does not increase light meter readings as much as it increases lumens out per watt out. More specifically, as of 2/15/2023 I experienced using more dye (and/or effectively more dye) getting almost or possibly counterproductive at dye concentration or effectiveness getting past that which achieves 475 lumens out per watt out. I consider 465 lumens out per watt out as a result when the optimum amount of dye concentration (and liquid layer thickness) is achieved.

This 465 lumens/watt needs to be multiplied by the LED's efficiency and also by the Stokes efficiency. The Cree XPGDRY-L1-0000-00601 has efficiency of 73% minimum and (I expect) 75% typical at 350 mA. I expect slightly higher at somewhat lower currents, and I expect best overall result at about 200-250 mA.

As for Stokes efficiency, which is (approximately) the centroid wavelength of the absorbed radiation divided by the centroid wavelength of the output radiation: I have used suitable equipment to determine the centroid wavelength being very close to 540 nm when the amount of naphthalimide dye used achieves 465-475 lumens out per watt out. As for centroid wavelength of the Cree XPGDRY-L1-0000-00601 LED, I have yet to measure this but on 2/15/2023 I estimate this as 450 nm. 450/540 is .8333 or 83.33 %.

75% of 83.33% is 62.5% or .625 . I figure for an optimistic figure here, .625 x 465 lumens/watt, which is 290.6 lumens/watt. I also have a from-my-gut expectation of achieving 92-93 % of this, which means 267-270 lumens per watt from optimum usage of refrigeration lubricant dyed with a naphthalimide type UV due with a Cree XPGDRY-L1-0000-00601 with 350 mA of current. I figure 270-275 lumens out per watt in is achievable at 200-250 mA.

As of 2/15/2023, I have yet to test or describe an optical arrangement that minimizes optical losses of light produced by a fluorescent liquid being irradiated by a Cree XPGDRY-L1-0000-00601 LED.

Update 2/16/2023: I plan to test using water dyed with a perylene dye that is probably similar to, and possibly the same as, the dye in green fluorescing antifreeze. I also plan to test water dyed with fluorescein (more specifically its sodium salt, uranine) with a small amount of sodium hydroxide. (Fluorescein works best in alkaline solutions.) I might over 500 lumens per watt of output radiation. However, these dyes may not have good absorption at shorter blue wavelengths (and violet wavelengths) of 400-455 nm.

Written by Don Klipstein.

Copyright (C) Donald L. Klipstein 2007, 2009.

Please read my Copyright and authorship info.
Please read my Disclaimer.