Infrared technique: Scanner to check quality of photo detectors (1971)

 

Infrared: scanning the surface of photo-detectors (1971)

 
Original publication: “An On-line Display Method for Scanners Applied to Photo-detectors” 

Synopsis

During the development of solid-state devices, measurements are made to improve manufacturing methods. If a certain parameter varies along the surface of the device, a scanning system may be used for its examination. In the course of the manufacturing of infrared detectors of InSb in our laboratory, a flying-spot scanner with a display system was developed. These have been used to examine the sensor responsivity along the surface of the detectors. Below, a description is given of the scanner and the display system. Typical results obtained with the display system are shown.

A modern photovoltaic infrared sensor (2 x 2 mm)
A modern photovoltaic infrared sensor (2 x 2 mm)

Introduction

The number and type of imperfections on the surface of a single crystal of semi-conducting material will depend on manufacturing material treatments like sawing, lapping, etching and (chemical) polishing. Besides these, other sources of imperfections like dust particles and drying residuals may occur. In the following, we explain how information can be acquired about the inhomogeneities in the sensitivity of a photodetector over its surface.
If a photodetector is scanned with a dot of light, the output of the detector will correspond to the photo efficiency at the place of the spot on its surface. The characteristics of the inhomogeneities in the photovoltaic response along the surface should give indications about their causes and in some respect, of the quality of the detector.

On-line display methods

Two methods are in use for displaying the output of a scanning system online on an X-Y oscilloscope. The principles are as follows:

  1. The voltage corresponding to the X and Y coordinates of the flying spot measure on the surface of the device relates to the horizontal and vertical inputs respectively of the oscilloscope. From the output signal of the device, a signal Z is derived, which modulates the intensity of the spot on the screen.
  2. From the three signals, X, Y and Z (Z being the output of the device) the following two signals are derived: Z + A*X and Y+ B*X, where A and B are properly chosen constants.

These signals are connected to the vertical and horizontal inputs of an oscilloscope. The number of scans, which are displayed will be kept rather small in practice, to restrict overlapping of parts of the successive scans.
A new design for the online display, comparable with the second method, is described hereafter. The method has the advantages, that overlapping of areas in the pictures is avoided and that the pictures are very detailed. Experience has been obtained with this method by application to the scanning of photovoltaic detectors.

Photovoltaic energy is the direct transfer of (sun)light in electricity. Some materials have the property to absorb light-photons and along this way to free tighten electrons. This is called the photovoltaic effect if an electric voltage is generated.

Description of the applied method

The Cartesian coordinates of a moving spot on the scanned surface will be denoted by X and Y, measured on the surface. Y stands for the coordinate of the fast scan. The output of the scanned device will be denoted by Z. X, Y, and Z are represented by voltages (Volts). The single-valued function Z(X, Y) can be represented by the curved surface of an “object” in our visual three-dimensional space. The situation is depicted in Fig. 1, where the section with the Z-Y plane is marked by dots. This section is sketched separately in Fig. 2. Let (Y, Z) be a point of the boundary of the section, and let Yp be a function of Y and Z, defined by Yp = Y + Z * cotg(alpha), where Yp can be considered as the point (Y, Z) projected on the Y-axis from a direction of observation, given by the angle alpha.

Let the point (Y1, Z1) be a point on the boundary, which is projected in Y1p. Then from Fig. 2, we may conclude that (Y, Z) is visible to the observer when no point (Y1, Z1) can be found, for which Y1p > Yp while Y1 < Y.

Fig 1: The single-valued function Z(X, Y) can be represented by the curved surface of an “object” in our visual three-dimensional space
Fig 1: The single-valued function Z(X, Y) can be represented by the curved surface of an “object” in our visual three-dimensional space

 

 Fig. 2: Visibility to the observer
Fig. 2: Visibility to the observer

By placing the observer far from the object we obtain the simple relation Yp=Y+Z. A where A stands for cotg(alpha) which is a constant now. The process of making a section with the Y-Z plane will be called a scan. At every scan during which Y increases we find that, for the visibility of the corresponding point (Y, Z), Yp has to be larger than all previous values arising during the scan.

Fig. 3 Peak-hold circuit
Fig. 3: Peak-hold circuit

By the use of a peak-hold circuit, of which a simple example is shown in Fig. 3, the visibility of a point may be determined. Yp is used as the input signal for the peak-hold circuit, the output will be denoted by Ym. The diode in the circuit will be considered ideal. Obviously, during the periods that Ym is increasing it will follow Yp, at other moments it is constant and larger than Yp. The value of the threshold Yp has to be chosen such that for the initial points of all scans Yp is larger than Y0. A complete image is built up by combining Ym-traces of successive numbers of scans, at gradually increasing values of X, into one picture. The X-scan will be called the slow scan. The capacitor has to be discharged after every fast scan.

Track highest value with increasing X
Fig 4: Track highest value with increasing X

In Fig. 4, Y, Yp and Ym are depicted for an arbitrary case. A characteristic of the peak-hold circuit is that the visibility of a point is determined by the diode, which is either conducting or non-conducting. The state of the diode is also indicated by Ym which is either increasing or constant. It will be obvious that points are visible only during the conducting periods of the diode. By using Yp for the vertical and X for the horizontal deflection of an oscilloscope and suppressing the brightness of the spot on the screen during the nonconducting periods of the diode we obtain pictures showing the curved surface without overlap, provided there is also suppression during the back-sweep of the fast scan. From Fig. 4 we may conclude that by using Ym instead of Yp overlapping will not occur at all. Thus there is no necessity to decide on suppressing depending on the state of the diode. The best results have been obtained by using Ym and suppression, the latter only during relatively long periods of constant Ym. This results in a contour intensification in the general pictures.
Successive scans should be close together to obtain coherent pictures, a typical number being 2000 scans per picture. With a frequency of 100 Hz for the fast scan, the generation of a picture will take 20 seconds. During that period a photographic record is made of the display on the screen.

Experimental set-up

An outline of the flying-spot scanner is shown in Fig. 5. In the experimental set-up, a He-Ne laser Beam (0.63 µm) was attenuated, expanded with a telescope, reflected by two mirrors and focused on the surface of the device to be tested. A mirror galvanometer has been used for slow scanning. The fast scanning was obtained from a circuit of which an outline is shown in Fig. 6. All frequencies, except that of the saw-tooth generator providing the X-signal, were derived from a 100 kHz crystal oscillator. Two square waves of 100 Hz were obtained by division, and a manually adjustable phase shift between these square waves was introduced with a univibrator which in Fig. 6 is denoted by “variable delay”. From these waves, a triangular and a sine-wave were derived by integrating and low-pass filtering respectively.

Fig.5: Optics and mechanics of the flying-spot scanner
Fig.5: Optics and mechanics of the flying-spot scanner

 

Principle of a mirror galvanometer - mirror moves due to the generated magnetic field by the flow of current
Principle of a mirror galvanometer – mirror moves due to the generated magnetic field by the flow of current

 

Four modes of the torque scanner
Four modes of the torque scanner

 

Fig.6: The circuit which generated the steering signals for the mirrors and oscilloscope
Fig.6: The circuit that generated the steering signals for the mirrors and oscilloscope

The sine-wave was fed to the torsional scanner. Due to the small mismatch between the resonance frequency of the torsional scanner and the sine-wave frequency a phase-shift would be introduced, which then was compensated by using the above-mentioned adjustable delay. The amplitude of the sine-wave was chosen such that the dot of light was on the sensitive area of the device only during the almost linear parts of the sine-wave. The output (Z) was amplified (frequency: D.C. to 100 kHz) and added to the triangular wave, which represented the Y coordinate. The result was Yp=Y+Z.A. Signal Yp was fed to the peak-hold circuit, which gave Ym as output. This signal was connected to the vertical input of the oscilloscope. The saw-tooth generator had an adjustable period of up to 20 seconds. This signal was connected to the horizontal input of the oscilloscope and the X-mirror.

The following three photographs show the mechanical layout with the electronic circuits of the scanner:

Control-unit with oscilloscope (bottom view)
Control unit with the oscilloscope (bottom view)
Control-unit with oscilloscope (top view)
Control unit with the oscilloscope (top view)
Control-unit (1971)
Control-unit (1971)

The pictures shown in this paper are Ym-X displays. They have the appearance as if several cross-sections being made through the “object” perpendicular to the direction of observation. This is a result of using pulsed intensity modulation in the display with a frequency of 100 kHz. By this presentation, a better display can be obtained of the steep parts.
In the experimental set-up, the intensity modulation was controlled such that it was suppressed during the back-sweep of the X and Y scans as well as during the relatively long periods of constant Ym. The intensity modulation could be accentuated by using pulses of 3 µs width, obtained from the divider circuit. A selection could be made from the repetition rates: 100, 50, 25 and 12,5 kHz.
When an extra pulse is applied to the flip-flop which drives the Y-generator a phase shift of 180 degrees will be introduced in the fast scan, and we obtain due to the symmetrical scanning a similar picture but seen from the “opposite” side.

Preparation of the detectors

The detectors were made of n-type single crystals of InSb. After sawing the original crystal into cubes the damaged surface layers were removed by etching and chemical polishing. A p-type surface layer was obtained by diffusion. A mesa-structure was etched out of the cube by covering a part of the polished surface and etching away the surrounding p-layer. Contacts were made to both sides of the p-n junction, the n-side was soldered to a Covar strip. This strip was attached to the cold finger of a Dewar flask for liquid nitrogen. The mounting place was in front of a window with a low absorption coefficient for both visual and infrared light.

Examples

Sawing and lapping will cause cracks which may increase the recombination of electrons and holes at the surface and, as in the case of a photovoltaic detector, within the depletion layer. Also, the reflection coefficient may be changed. The cracks usually have the structure of tracks and therefore one may expect to find tracks of decreased sensitivity on the surface (Fig. 8).

Fig 8: Sawing and lapping cracks
Fig 8: Sawing and lapping cracks

Dust particles and drying residuals will change the sensitivity according to their optical properties. They may either decrease the sensitivity by absorption or increase it by anti-reflection effects and appear as spikes or holes (Fig. 9).

Fig 9: Dust particles and drying residuals
Fig 9: Dust particles and drying residuals

If an anti-reflection coating is deposited, imperfections like holes will cause spots of altered sensitivity just as dust particles may do. However, the effect depends on the wavelength of the light (Fig. 10).

Fig 10: anti-reflection coating imperfections
Fig 10: Anti-reflection coating imperfections

The resistance of the surface layer of a  photovoltaic detector causes a decrease in efficiency with increasing distance from the contact electrode. The effect of the distance depends on the ratio of the surface impedance of the surface layer and the junction resistance (Fig. 11).

Fig 11: decreasing efficiency with increasing distance from the contact-electrode
Fig 11: Decreasing efficiency with increasing distance from the contact-electrode

At the boundaries of the p-n junction an increase in the sensitivity is often found due to the decreased effect of bulk and surface recombination, especially when the junction lies rather deep (Fig. 11). Of course, there is no sensitivity at the alloyed contact electrode and where the lead covers the surface. The causes of imperfections mentioned above are due to the preparation of the diodes but they will be mixed with the faults in the original crystal. The pictures shown in Fig. 12a, 12b and 13 are examples of diodes which have no striking inhomogeneities.

Fig 12a: diode with homogene surface
Fig 12a: diode with a homogeneous surface
Fig 12b: diode with homogene surface
Fig 12b: Diode with a homogeneous surface

The causes for the inhomogeneities of the responsitivity of this specimen are difficult to determine. However, if the output is clipped at a level which is about 80% of the mean value we see that most of the inhomogeneities are not deep.

Fig 13: diode with homogene surface
Fig 13: diode with a homogeneous surface

The detector in Fig 13. shows a very homogeneous responsitivity along its surface. For a more detailed result a reduced size of the spot and, according to this, a higher cut-off frequency of the electronic circuits will be required.