The idea

With an aspheric thin lens and a spheric reflector it is possible to collect much more light  from a surface emitting LED, compared to standard epoxy encapsulated LED.  Several tests were perormed wich confirmed this. Also a patent application was filed. Further work will be done to confirm the measurments.  ca 50% more light could be collected compared to a standard LED coniguration.  This setup could withstand  almost 3 times current  without fast detoriation. This was done under liquid cooling in paraffine.

A LED die from Cree was chosed. It had a rated power of 12mW. There are dies with more output power, however this Cree die has most output power per area unit. Even thow there are several large LEDs on the market with up to 5W power, the total emmiting area on these  LEDs are much larger compared to  the GaN die from cree. This is very important if one will have low N A and good etendue in an optical system. In  applications as illumination in livingroom this  is not neccesery.  Data from cree

The experiments

This graph display the hugh difference between a cooled and a uncooled  LED. The tests were performed on a Cree megabright 405nm LED. One was standard encapsulated in epoxy the other was directly immersed in circulating paraffine liquid at 20 deg C. Both was run at exactly the same conditions. Appr 3 times more current then specified by cree. One important observation is that the overall efficency will be lower due to the rice of  voltage drop over the LED at higher currents. The absolute light output  is of couse higher, but here is no point running the LED at much higher current then appr 100mA. The power curve is flattend at higher currents.
power as a function of time. cooled/uncooled LED.


This is an image of the test rig with circulating  liquid cooling,   The grey tubes on the bottom and top distirbute the paraffine liquid. A hamamatsu photodiode with well known characteristics were used, the dark square.

This graph shows the increase in optical  efficency and better etendue when a mirror and a aspheric lens is used. Take in account that to get total power the curve must be circular integrated, that is the  power leve/cm2 is  increasing as a square function outside the center. A small change is corresponding to large  total power change.  The little peek outside the center dosent looks much but there are a lot of light energy there. Improvements is much more than the graph shows due to the two dimensional rotational distribution.
The dip showed in the graphs is the contact electrodes
power output comparison Aspheric mirror/ standard epoxy.



This is an image of several  projections on a screen.  Compare the  image up to right with a standard LED. It has a very large  spherical abberation error. The ring around  the center is from a cup reflector inside the standard LED wich is used to collect more light but with the disadvantage that the image will be larger. The image down to right  is from the best type of asphere. The square  image is somewhat larger  but it has no surronding ring and all light is collected in the square. Notice the more intense light in the square.


This is an image of the  projection from the setup.

Possible applications

Some interresting applications is illumination to microdisplays in projectors, both Lcos-, Micromirror-, and transmittive LCD displays.

This is a  word document  describing the theory and setup in more detail.  Some example are shown how a possible setup could be done.
The word document

Other applications where the etendue through the optical system must be good and where the N.A. from the light source must be low, or a corresponding  point source must have a small surface emmiting area.
Some development of the circulating liquid must be realized in compact systems where a separate circulation pump is not allowed.


This is an example of what the LED light source could replace. Ultra high pressure short arc lamps. This is used in optical lithography, and projection displays. This lamp gives an intence light form a very small area (arc) and is containing mercury. Note on the image a large Hg drop that evaporates under high pressure when ignited. The expected life is short, 600-2000 hour typically.

The array

To be able to  get more light out from the LED just one single led isn´t enough.  Becouse of the high light density from light from small dies they are prefered compared to  large dies for example lumiled. But to get a high total flux many leds can be put together in a paralell array.  This is possible becouse of each LED  is already lighting with paralell light.  A condensor lens could be used to collect the light fron the different leds to one point. There is no limit how many  leds that can be placed in an array..  


In this setup 91 GaN leds were placed paralell for some experiments. This gives a good feeling how large the final lamp could be.



The prototype


This is a closeup image where the sperical mirror, the contact electrodes to the LED and the front aspheric lens is shown.
The LED die itself is hard to see becouse it is to small 0.25 mm in side.  The sperical mirror were dielectric coated to  get the best possible reflection at 405nm. The etched aluminum connection were made as thick as possible, several micrometers, to get as low voltage drop as possible. They had to be very narrow  so it did not block the light due to obscuration. Also the LED die it self and its bonding aluminum  island were a small obscuration and was made as small as possible. The bonding were made by  Mandalon AB


The spherical reflector

First  the sperical mirror were made by a  metallic coating of a concave lens. An aluminum surface has only appr 85-90% reflection at 405nm. To be able to get as high reflection as possible, 99% or better, the only choice vere a dielectric reflection coating.  A 13 layer of an adjusted yellow filter adapted to the very strong curved surface was manufactured.  Calculations and measurements was done to  ashure a correct high reflection for this curvature and broad angle of incident. Tomas Näslund performed the experimental coating in a Balzer electron beam coater. Here are a spectra of the transmission at different incident angles.

   
On this image is shown the small substrate mirrors for the LED, attached on a standard holder for the coating plant.


A closeup image of the mirror. The reflection is very weak in the visible wavelengths. As slight yellow surface reflection can be seen.  Notice the reflection of the lamp in the center of the mirror. The substrate were of  polycarbonate. This was chosen becouse of its very good molding properties for optics.


first tests  of the surface bonded LED


An image of the LED mounted on  the 1mm glass surface. Contact electrodes were etched from aluminum coatings. Current is applied through the  handheld electrodes.

Manufacturing of the aspheric lens


Closeup image of the aspheric handpolished lens. This lens were  in optical contact with the backside of the glas with the LED mounted on, through a liquid transperent film. The curvature is corresponding to the simulations made in an optical calculation software (synopsis).  


A piece of PMMA plastic was glued to an aluminum rod. The rod was placed in a lathe and first a rough curvature were turned.  With aid of small cotton tops, soaked first in grinding powder, then in fine polishing powder, a good aspheric  and smooth surface was created. 10-15 different surfaces were made and only good one were picked out for  the experiment.  The plastic piece were afterward cut out from the aluminum bar, polished on back side and made thin, appr 1mm, thickness.