Notes on Do-It-Yourself Microscopy

Roger Baker

I have published directions on how to make homemade water immersion, short-focus doublet lenses for optical microscopes with motor-driven tubular grinding and polishing tools. These enable one to see common bacteria. Complete directions on how amateurs may make lenses using a variety of techniques to give ground, melted, and fluid lenses was published by the author in a piece titled "The Homemade Microscope," in Science Probe , April, 1991, p. 53.

These studies revealed to me something important about Leeuwenhoek's jealously guarded methods. The theoretically possible resolution of the optimum dry single-lens microscope is directly related to its numerical aperture (which is the maximum angle of the cone of light that the lens can receive from any point on the specimen). A dry lens cannot account for the sub-micron detail of Leeuwenhoek's best discoveries. Not unless the front lens is in optical contact with a fluid like water or oil surrounding the immersed specimen. This seems to me to be de facto proof that Leeuwenhoek's secret method must necessarily have been based on the discovery of the compound fluid immersion lens. A full account of this conclusion was published in the journal, Microscope , 1993, p. 1-6, vol. 41.

Another tempting goal for amateur microscopists nowadays from my point of view is to build a home-brew scanning probe microscope. This has been one of my back burner projects for a decade or so. (I so recall seeing a web link to a site devoted to developing an amateur scanning tunneling microscope that proposed some details a year or two ago, but the site became inactive without getting one to work, so far as I know).

I have not constructed a fully working model either, with the tough part to me being mainly in the computer software and digital part of the design, where I am weakest. This should not be a major obstacle to further progress, because I think the hardest part is the mechanical design, where the results look pretty good. Following are some details of my work so far. The most important design problems are probably the scanner design and the detection and feedback control of the vertical axis. The feedback loop between the vertical positioning of the vibrating probe tip initially closes in on the sample-and then backs off just enough so as to have the minimum vibratory contact that can be detected. It must then maintain this condition while "feeling" its way back and forth across the sample surface, all the while capturing digital information corresponding to the height of the specimen at every point. This gives the data to build an image from the bytes that comprise the data stream for all the rows and columns of the image, much as a TV picture is generated.

The sensitive detection of this initial vibratory contact between the point of the probe and the specimen is a key requirement. This step is accomplished by amplifying the light modulation due to the reflection of light from the mica cantilever. A detection circuit (full wave rectifier) converts this into a variable DC voltage that is fed back into the vertical peizo. The piezo causes a mechanical post to the probe tip mounted on the post just enough to maintain a constant low level of vibration. The detection and feedback part of the circuit is analog whereas the rest is mostly digital.

A laser pointer from K-mart, etc., is a bargain at less than $10. It will be a somewhat noisy light source with its own internal feedback loop. Split beam detectors are not hard to hack; such a detector is made from a pair of sensors facing each other at 90 degrees to the beam with an aluminum reflector mounted between them. The pair are connected in series and the midpoint connection tapped as the output signal leading into the detection circuit. However for development, I used a phototransistor in series with a resistor. Phototransistors are photodiodes with one transistor stage of amplification built in. Connected with the right polarity, they act much like light-variable resistors.

My probes are either vibrating tungsten wire tips made from light bulb filaments or else dagger-like slivers of broken glass. I suspect glass may be superior. Since tungsten is a very heavy metal, and I have preferred not to weight down the cantilever and reduce its sensitivity and vibration frequency. For this reason, I mount the specimen rather than the probe on the reflective cantilever. The probe does almost all the moving, both scanning and vibrating. However I think a tiny glass probe glued onto the mica cantilever, in accord with conventional practice, might ultimately be best.

I employ a tiny reflective cantilever made from a tiny elongated (perhaps a micron or less thick and thus iridescent) flake of mica, compression-bonded at one end to an aluminum base with beeswax. Very thin sections of mica can be peeled off of a larger piece with adhesive tape and then trimmed by pressing them against a flat surface with a razor blade. The sample to be viewed is a tiny speck of almost anything solid smeared, glued, or deposited onto the underside of the tiny mica sliver, which probably reflects 5 percent-10 percent of the light that hits it. This will typically take a needle and binocular microscope and a steady hand and a trace of some adhesive like silicone rubber applied with a hair.

The two most difficult parts of this instrument to make are the scanning mechanism component, and secondly the mica cantilevers and probe tips. Physical skill is needed for manipulating sharpened needles and small objects under a binocular microscope, but all these techniques are within the ability an amateur with reasonably steady hands.

The project, under ideal circumstances, may not cost much more than $100 to build, which is far below the cost of any commercial instrument. A single commercial cantilever can cost that much, I have been told. An oscilloscope is probably essential for this project in order to troubleshoot and adjust the circuitry.

The cantilever is mounted horizontally with the sample side down so that a laser beam hits it from a few centimeters away at about a 45 degree angle. I mount a piece of frosted glass a few centimeters away to the other side to catch the reflected light. Then I clamp the photo sensor against this image on the glass near one edge of the reflected spot. The photo sensor can electronically detect the induced vibrations of the mica cantilever up to 5000 hertz or more.

A certain minimum vibration amplitude setting is best since there can be a hysteresis of attraction between the vibrating wire tip and the specimen, making it appear as if the specimen were sticky, particularly if the tip is blunt. Without vibration, and when contact is first made, the specimen will stick to the vibrating wire tip with a force that probably depends on factors, like the fineness of the tip and the humidity and the ambient organic contamination in the air. In order to combat this effect, it is necessary to increase the vibration energy by adjusting the frequency and amplitude, so that the energy of sticking is exceeded by the energy of vibration, whereupon hysteresis of contact disappears.

It also appears that this problem is least when the vibrating tip is positioned near the anchored end of the cantilever, because the mechanical stiffness of the mica flake is greater near its support. The cantilever is seen to bounce against the probe tip if things are not just right. One solution is to move the tip closer to the cantilever support and to arrange things so the detection of and feedback of vibration amplitude is in a linear region, just after initial contact.

Initial probe contact can be seen by a shift in the reflected beam. This helps to adjust the best mechanical setting; as the tip closes in on the mica, the clear red patch of reflected light will be seen to suddenly blur as tip contact vibrates the image on the ground glass.

One way to detect vibration is to put the light sensitive detector in the brightest spot and then to detect the change in light level as the vibration spreads this beam over a larger area. Another way is to put the detector at the edge of the reflection and to sense the change in amplitude at this shoulder.

The vibration energy communicated by the vibrating tip to the cantilever is sensed with the laser beam; sensitive operation requires that both the frequency and amplitude of the vibration signal should be carefully adjusted. Useful frequencies for vibrating the tip appear to lie between 1 and 5 kHz.

It now seems clear that the most sensitive technique is to detect tip position by the amplitude of vibration just after initial contact, and that part of the secret of success is to make sure this initially detected vibration is in a short linear region immediately after contact. This sweet zone depends on an interaction of cantilever stiffness, amplitude and resonant frequency.

My scanner design is pretty unorthodox. In essence the scanner is made from a little open-at-the-top cubical glass box about ten centimeters across bonded with hard, long-curing epoxy. On the inside of the open glass cube, five inexpensive piezo diaphragms about an inch in diameter, from Radio Shack, about the size of black checkers are mounted flat against the five inside walls of the box. Snugly fitting sections of rigid aluminum tubing from the hobby shop are epoxyed like struts to meet at the center of the box so that the opposite diaphragms on each side are mechanically linked in a push-pull arrangement.

The vertical tubular aluminum post rises from the piezo diaphragm at the bottom of the glass box. One uses a DC electrical signal leading to this piezo to induce a vertical movement to the post, on the order of one to ten microns up and down. Then one adds in a little high frequency AC component through a capacitor to cause the post to vibrate a little at the same time. Four other short snug sections of aluminum tubing a millimeter or so in diameter are glued from the centers of the other four diaphragms so they meet the vertical post near its midpoint. When connected properly these opposing piezos cooperate so that the center post is rocked sideways along the two independent horizontal axes.

Thus we have the means to cause the top end of the aluminum tube post, which extends slightly out of the top of the open glass box, to scan independently along three axes. It scans horizontally by rocking sideways in the two horizontal directions, and vertically with a slight superimposed vibration component of motion. These three independently controlled X, Y, and Z piezo motions are each controlled by the digital scanning circuitry at a maximum of about 25 volts DC. The digital circuitry can be operated at a standard five volts and then the 0- 5 volt output from the DACs raised to the higher piezo control voltage with op amps. The piezos should be operated in series with about a l00K resistor.

A glass or aluminum plate perforated with a hole is placed on top of the glass cube so that the post bearing the probe tip sticks up through the hole a centimeter or so. A support made from a strip of brass a few centimeters wide and 10 CM long is doubled up into a skinny U-shape and a nut and screw mounted so as to adjustably spring it apart, raising and lowering the attached cantilever. The cantilever is mounted on the top of this surface and brought into contact with the probe tip by adjusting the screw. Contact between the probe and cantilever is detected by a sudden shift in the position of the reflected laser beam on the ground glass or by the onset of an AC component in the photosensor signal. At this point, the scanning piezos can take control.

There are two methods I use to make sharp probe tips. Initially I used tungsten light bulb filaments. Since tungsten oxide is quite volatile, an easy to way to make an extremely fine point on a fine stretched light bulb filament (I prefer Chinese night light bulbs) is merely to introduce it into a small propane flame. It oxidizes in the flame and the oxide is evaporated by the heat to leave an invisibly fine tapered point at the end of the bare wire, which is then snipped off and mounted at the top of the post. The latter is flattened shut to give a little vertical mounting area where the probe can be glued. This should all be done under magnification so as not to lose the nearly invisible tip.

Another way to make sharp tips is to take a small piece of glass inside aluminum foil and smash it with a hammer. The fragments are put on a smooth black surface and viewed under strong light with the binocular microscope. Sharp pointed slivers of many sizes and shapes are to be found in abundance. I believe that their pointedness, resulting from the intersection of three independent fracture planes, extends down to nearly atomic dimensions.

The laser and the photosensor are both mounted on a thick brass strip which is bent at a 90 degree angle. Its bottom edges ground to rest stably on the flat platform that covers the scanning cube. In the middle of one side, the laser is clamped so the light is projected down at a 45 degree angle where it bounces off the mica flake and up through a window cut in the other side of the brass fixture. Finely ground glass is mounted over this window so that the reflected red laser beam is observed in the glass. The last step is to clamp the photosensor over the spot and to adjust its position until the vibrating probe tip is detected as an electrical output of the photosensor.

The end result is that the probe sequentially probes and gathers data from the entire surface in a microscopic square area, say 10 microns on a side, on the surface of a specimen. This entire surface is scanned to generate data; gray tone pixel values from which a whole image can be reconstructed, in the same way that a television image is transmitted and reconstructed. Such a scanned image might be about 10 microns square.

During operation, the probe is raised or lowered under feedback control to give a constant vibration signal that reveals the height of the surface at each point. The scanned image corresponds to the three-dimensional shape of the surface revealed by converting it into a gray-scale image where high points are white, deep valleys are black, and intermediate points are shades of gray.

The image capture strategy is as follows. For initial design purposes it is reasonable to plan the resolution of the image to be comprised of 256x256 picture elements to give 65,536 pixels total-with 8 bits (or one byte) of vertical information for each of these pixels, giving a total of 524,288 bits per image.

If we presume that the probe takes one millisecond to stabilize vertically, this image could be captured in about a minute, which is probably optimistic, whereas if it took ten milliseconds at each step, the image could be captured in ten minutes, which is still within reason. There are two digital counters corresponding to the two horizontal axis position controls, each capable of counting up to 256 and then resetting, triggered by a clock circuit of some sort. The clock must be set to a timing interval on the order of several milliseconds to allow the vertical feedback position to stabilize before capturing the next vertical position data byte. One of the counters counts along the rows and this count is fed into a digital to analog converter and then into the X-axis piezo until the whole row is counted and the counter resets. This event triggers one count on the second horizontal axis of the image. Each horizontal position is represented by two bytes. With each step along a row, as soon as the analog feedback position is stable, an ADC converter digitizes the analog signal to give a vertical data byte. As the end result, there are three bytes available, whether needed or not, as data corresponding to each point on the image.

This is fed into a computer to be displayed as a bit-mapped image. There are a million ways to arrange the interface details and get similar results. The general procedure described is generic and the various details and data capture options have been published in the Review of Scientific Instruments and in the voluminous literature dealing with scanning probe instruments. Appropriate circuits should not be hard to build with digital conversion chips ($10 or so each) from Digikey. I am not good at computer interfaces, and have to sweat the details more than many readers, probably. Go for it!