Fabrication of a Silicon “Proto-retina” Chip for Artificial Vision Application—An Experiment To Be Performed
February 24th, 2003 | No Comments »BACKGROUND AND DISCUSSION OF THE PROPOSED EXPERIMENT
The basis for this new model of the eye is that: a.) light is detected via spatial dimensionalities (”optical antennas”) and that, when realizing this, b.) the specific array of two sizes of retinal receptors that have evolved did so as a logical consequence of the distribution of wavelengths falling onto it resulting from what has been termed, the “longitudinal chromatic aberration” of the lens and optical refraction properties of the eye. Everything is contained in those two statements that explain so much about vision!
In 1991 a discovery was made in the semiconductor field by Canham of a visible light interaction effect with silicon surface etched in an unusual manner. It has been historic dogma in the semiconductor field that hydrofluoric acid (HF) etches the oxide from a silicon surface but does not etch the silicon substrate itself. This was overturned in Canham’s experiments wherein he found that electro-etching in HF produced a highly unusual surface that surprisingly absorbed and emitted intense visible light. Examination of this surface revealed that it was composed of exceedingly deep pores etched into the silicon (where no etching at all was expected).The surface nanostructure was then seen to be composed of an array of silicon “pillars and intervening pores” that interacted with visible light. It also must be noted that the wavelength of light could be made to vary..although there were no “pigment molecules” present…only “inert” silicon pillars and intervening pores. This phenomenon has been termed “porous silicon” (or “PS”) and has elicited a great deal of interest in the semiconductor community worldwide.
I would note here of importance that the aspect ratio (i.e., pore depth to diameter) of the pillar-pore PS nanostructure is very high… approximately 20:1. This gives the appearance of a “dense forest of tall pine trees” of silicon pillars. This is startlingly like the appearance of the array of retinal receptors only they have an even higher aspect ratio of approximately 50:1. If one can get past the pigment arguments for light interaction with biological retinal receptors the nanostructures of the retina and PS are physically and dramatically very similar!
A high aspect ratio silicon nanostructure composed of inert silicon material that strongly interacts with visible light.
As primarily a solid state physicist interested in radiation detection I became interested in the PS phenomenon. I remember calling Canham in England and in the conversation he stated that the effect occurred only when the silicon pillar diameter was reduced to electron quantum confinement dimensions. In physics this dimension is termed the “electron in a box” approximation with the dimension in silicon at room temperature being of the order of 10-100 nanometers. If the silicon pillars were not on average reduced to this dimension there was no visible light interaction effect.
My experience with silicon technology led me to ask the question : accepting the fact that deep pores are electro-etched into the silicon surface, where on the surface, are these pores nucleated? What surface artifact or energy anomaly nucleates a pore? A logical proposition seemed to me to be the energy “sink” provided by dopant atoms that happened to appear at the silicon surface. I’ll not go into the quantitative calculation here but the density of such dopant atoms is such as to produce the PS pore array.
I became sufficiently interested that I was motivated to conduct an experimental effort to study the effect of dopant density on luminescent wavelength of PS surfaces would note that electro-etching of PS surfaces is not a difficult undertaking at all .. one needs only a power supply, acid, and a simply designed etch resistant cell (of Teflon) to hold the silicon substrate. ….and voila! This experimental effort is described elsewhere on this webpage but, to summarize, we found with the requisite that silicon pillars were reduced to EQC dimension, the visible luminescent emission was correlated with the density of dopant atoms that we felt determined pore separation. Thus we found that pore separation determined wavelength in exactly the same manner that retina receptor separation determines wavelength in the eye. In the series of experiments we could vary emitted wavelength from deep red across the visible spectrum to orange and then green blue. This experiment and the logic assumed within it was subsequently replicated by a laboratory in Japan in an effort conducted under my direction.
Now… one must note that the distribution of dopant atoms on the the silicon surface is statistical (or random). This results in the PS electro-etching process in an array of pores that can best be described as chaotic… the “web” of silicon between some adjacent pores “etching completely through” while in other areas this web would be too thick to support the EQC requirement. The result should be a low value for overall visible light conversion efficiency. Such low values (of only a few percent) are the norm for PS etched surfaces (although the observed brightness of PS surfaces can only be described as intense - but this requires another discussion).
Thus, dopant density, although chaotically distributed, will control the wavelength of the PS light conversion process.
At this point we began to consider the light interactive surface effect described under the link “How This Idea Developed - a Brief History” - which I commend to the reader. I specifically refer to the experiments of Dvorschak et al, and Young et a, wherein the wavelength of light incident onto a surface is orthogonally expressed on that surface as features having the same dimensionality as the incident wave.These experiments have been replicated in a number of ways as described on the link.Thus we reasoned that if a discrete wavelength of light was made incident onto the silicon surface during the PS etching process such light might tend to “pull” the character of the surface (the distribution of pores) towards that wavelength. In summary we found this to be so - using, as I remember green light and achieving green emission in that experiment. I believe that this effect has now been achieved by others although I have not kept up with developments in the field.
Thus, light incident on the silicon surface during the PS etching process can effect (or control) the resulting distribution of pores and thus the wavelength of interaction. To achieve this the competing dopant density pore generation effect must be suppressed, i.e., the silicon must be purified to the greatest extent.
To describe the experiment that I think can be done which I think will, in reality, replicate the evolution of the light interactive retinal structure of the eye, would utilize a converging optical lens (simulating the lens of the eye) impressing the solar spectrum at some reasonable intensity onto a silicon surface during the PS electro-etching process. What should result in my view would be a circularly symmetric diffractometric pattern of the same character that I have predicted for the biologically evolved retina. I think that this would constitute a very fundamental result indeed.
I have termed this experiment the quest for a “proto-retina”. because the PS silicon “receptors” will still not possess the capability for detecting light phase (i.e., the giant dipole electrical mechanism along receptor lengths) that human retinal receptors are capable of and that decode direction to satisfy the Fourier equation.
BUT… we will have achieved half of the puzzle.. fabrication of a visible light interactive array of receptors possessing the same architecture as the biological retina.
Importantly, such a silicon array should be more spatially consistent with the neural circuitry that has evolved and is underlying the array of biological receptors and may even at this stage of development have some application in the quest for an “artificial vision chip”.
I had not been optimistic about the further development necessary to electrically “activate” each “silicon pixel”. I think, however, that I am wrong in this considering the phenomenal strides that are being made in the capability for fabricating “multi-mega pixel” densities in the digital imaging field. I think that one might envision an “electrical microcircuitry overlay” for the silicon PS surface that could finally mimic human vision.
What would be the advantages over our current “light intensity sensitive” imaging modalities (film, CCD devices, etc.)? Color constancy characteristic of human vision would be one.
I will append a diagram of the proposed experiment soon as I get to it.
I WILL LEAVE THE FOLLOWING TEXT HERE.. A GREAT DEAL OF IT IS REDUNDANT WITH THE ABOVE BUT… FOR WHAT IT IS WORTH
ADDED 9/06/03
I really have no knowledge of the specific concepts for developing an artifcial silicon retina other than to understand that any such concept must be based on the assumption that the retina is the image plane of the eye … and therefore that current silicon imaging technology (charge coupled devices and the like) would be used. I further assume that such “miniature camera-like” chips would be somehow applied to the outside rear surface of the retina. I guess I would further assume that it would be hoped that retinal nerve endings that were previously attached to the retinal outer segments would somehow grow to be atttached to the silicon chip (I am vaguely aware of some work - I think at Caltech - where such connections were shown to form).
Now… the outer segments of biological retinal receptors “face outward” to incoming light, i.e., incoming light must be transmitted through all of the retinal neuro-circuitry, the inner segments of receptors etc. before they encounter the light interactive outer segments. This has always seemed strange to vision research. The intervening matter is, however, tissue and is therefore fairly transmissive to light.
Now…how would one orient a porous silicon light interactive “proto-retina” (i.e. a circular light interactive surface with a pattern of receptors nearly identical to the human architecture)? Such a chip would perforce have a solid siliocn substrate (upon which the PS surface is etched). Silicon is not transmissive to visible light. Therefore, it would seem logical to orient the chip so that the PS surface was in direct contact wikth the rear of the retinal surface…. hoping that appropriate neural connections would form… and as I have said, I think that their is a greater chance that such useful connections would form as the light ionteractive “pixels” shold be in approximately the proper location. But.. it may be desireable to face the PS in a rearward direction… this would mean that incoming light would have to be transmitted through the silicon substrate before reaching the PS region….silicon is transmissive in the near infrared region (i.e. beyond 700 nm ..perhapos from 700nm to the band edge of silicon at 1,000 nm). This is the region used by the military for “night vision”. These long wavelengths are generally directed to the fovea of the eye…. and therefore, as we propose, contribute to imaging the “outline sketch” of the perceived image. It may be desireable to focus on regaining this capability for the sightless.
I note again that the term “proto-retina” connotes that such a device would possess only the capability of wavelength-tuned visible light interaction, i.e., the “inert pillars” of the silicon nanostructure would not have the ability, as the human receptor does, of decoding light phase (and thus direction) of incoming light.
Just thoughts,
GCH
I believe that the same fundamental mechanism is operative in visible light interaction with the receptors of the retina of the eye and with the nanostructure of recently discovered of “porous silicon”(or “PS”). In the PS situation this involves reduction of the silicon “pillars” of the nanostructure to electron quantum confinement geometry with the wavelength of the visible light interaction then determined solely by the dimensionality of the intervening “pores”.
The biologically evolved retina of the eye, however, possesses an additional capability not now present in the PS light interaction. This is the ability to decode the angle of incident light rays at each “pixel” via (we propose) an electrical “giant dipole” mechanism operative at each pixel. This leads to an image formation capability via the 2-D Fourier transform. PS development does not yet possess this latter capability.
I believe, however, that we can take the first step using PS technology to fabricate on silicon what I term a “proto-retina”, i,e,, an array of visble light receptors that exactly duplicates the receptor array of the biological retina. This will, I believe, demonstrate my thesis as to how the retina of the eye evolved. I cannot help myself from reminding here that the receptors (silicon “pillars”) of the proto-retina will be completely inert, i.e., they will not contain any pigment molecule species.
The experiment can I believe be very simply performed by electro-etching a PS surface with the surface at the focal point of a small converging lens providing illumination with solar quality radiation (an incandescent lamp would probably suffice). The ring shaped diffractometric structure of the biological retina should result - observable by illuminating with UV light. It is my premise that this is how the biological retina evolved!
A few solid state notes - the normal structure of the etched pores of PS are, as we have shown, distributed statistically because the site of their nucleation is determined by the distribution (statistical) of dopant atoms that appear at the PS etched surface. This results in “usual” PS luminescence as a uniform orange color. Pore distribution, however, can be “pulled” or altered by impressing an optical wavlengh onto the surface during the PS etching process (see discussion of this under “How This Idea Developed” on the webpage).
This experiment will provide convincing evidence that the retina evolved along the lines of my thesis. The next step will be to complete the silicon retina by adding electrical contact layers to provide the phase detection for imaging applications.
