US 3790287 A
A surface inspection system uses a highly focused spot of light moved rapidly over the surface to perform an inspection. This permits a ready location of the defects, gives some indication as to their size, and permits a high resolution inspection to be performed in a relatively short time. An optical arrangement is employed in which a scanned beam is brought back through the same lens system used in generating the scan to produce an immobilized return signal beam. A stop or spatial filter is used on the return beam to permit only the light scattered from defects to reach a detector.
Description (OCR text may contain errors)
nited States Patent "1 1 Cut'hbert et a]. v I
, FOCUSED LIGHT BEAMS SURFACE INSPECTION WITH SCANNED Inventors: John David Cuthbert, Bethlehem;
Richard George McMahon,
Allentown; David Farnham Munro,
Wescoesville, all of Pa.
Filed: Mar. 31, 1972 Appl. No.: 239,900
US. Cl. 356/120, 356/237 Int. Cl. G0ln 21/16 Field of Search... 350/162 SF, 6, 285; 356/120, 356/167, 200, 209, 210, 237, 129; 250/219 DF References Cited UNITED STATES PATENTS l0/l961 Lentze 250/2l9 DF 1451 Feb. 5, 1974 3,062,965 11/1962 Sick 356/237 3,460,880 8/1969 Henderson 350/6 3,565,568 2/l97l Hock 356/209 ABSTRACT A surfaceinspection system uses a highly focused spot of light moved rapidly over the surface to perform an inspection. This permits a ready location of the defects, gives some indication as to their size, and permits a high resolution inspection to be performed in a relatively short time. An optical arrangement is employed in which a scanned beam is brought back through the same lens system used in generating the scan to produce an immobilized return signal beam. A stop or spatial filter is used on the return beam to permit onlyv the light scattered from defects to reach a detector.
17 Claims, 11 Drawing Figures PAIENIED 51974 sum 1 or 6 (Km/mm (((WWWW PAIENTEDFEB SIIII BJSOZB? SHEU 3 0F 6 44 84 86 REFERENCE VARIABLE INVERSION PHOTODIODE GAIN AND OFFSET PREAMFLIFIER AMPLIFIER AGG V S'GNAL PREAMPLIFIER HIGH PASS PHOTODIODE AMPLIFIER AMPLIFIER,
F/GIKS GATE T VOLTAGE IN .01 F\ SIGNAL I SIGNAL 0 I I OUT IN I I PATENTEDFEB W 3.790.287
saw u or 6 IOOO- ON RESISTANCE (DRAIN To SOURCE) GAlN R ER 3R 4R FEEDBACK RESISTANCE PAIENTED 5W4 3 790.287 sum 5 or 6 DC GAIN GATE VOLTAGE SIGNAL RESPONSE LIGHT lNTENSITY PAIENIEII 3.790.287
' sum 5 BF 6 INGREASING BIAS SIGNAL RESPONSE LIGHT INTENSITY INCREASING GAIN ' SIG NAL RESPONSE LIGHT INTENSITY I SURFACE INSPECTION WITH SCANNED FOCUSED LIGHT BEAMS BACKGROUND OF THE INVENTION 1. Field of the Invention The invention relates to methods and apparatus for optically detecting defects in the surface of an article. The methods and apparatus are particularly usefulin the detection of crystallographic defects in epitaxially formed layers of silicon used in producing semiconductor devices.
2. Description of the Prior Art In the thin film and silicon integrated circuit technologies there has long been a need for automated inspection techniques.
Silicon wafers, either immediately before or immediately after growth of an epitaxial layer, require extensive inspection. Before epitaxial deposition, the smoothly polished surface must be free of all dirt and particulate contamination. After the epitaxy step, the surface must remain highly smooth and free of inclusions, stacking faults, and other disruptions. If the density of defects on the surface is too great either before or after epitaxy, later yields will be very low, and overall processing costs high unless the poor wafers are rejected.
Inspection is currently done by a human operator. A microscope illuminator light is directed down toward the shiny surface of the wafer, which is held in such a fashion that the specurlarly reflected light just misses the operators eye. Under these conditions, small defects show up as glints of light on the otherwise dark surface. The sensitivity of this technique is perhaps somewhat amazing, for it appears that under good conditions on a favorable wafer, defects as small as 2 microns can be seen with the unaided eye. The results are highly dependent on a number of factors, however, not all of which can readily controlled. The physical parameters such as the intensity and type of light used, background lighting, etc., can probably be adequately set and monitored, but the subjective human judgment as to which wafers are .good" and which are bad remains difficult to put under rigorous control.
Several attempts have been made to augment or supplant visual inspection of silicon wafers by automated techniques. One proposed system used board area white light illumination of a rotating wafer is conjunction with a photomultiplier to receive some of the scattered light to obtain information relative to the orientation and severity of crystallographic defects on epitaxial silicon. Information on the total defect area was derived, but not the more important and useful data on the number of defects, their approximate size, and lo cations on the wafer.
Another system used laser illumination, a spatial filter, and a vidicon/TV system for viewing defects. While this system provided an excellent picture of the defect locations, it too had some drawbacks. Lack of perfect flatness of the wafers introduced difficulties, and thelimited spatial resolution of the vidicon/TV system necessitated observation of only a small area of the wafer at any one time, so that a stepped field of view was required to inspect a whole wafer.
SUMMARY OF THE INVENTION- It is an object of the invention, therefore, to provide an improved system of inspection which will generate a quantitative measure of the defects present on the surface of an article.
It is a further object of the invention, to provide such an improved system which will measure extremely small defects.
It is still a further object of the invention, to provide a system which will yield graphic as well as quantitative results from a defect analysis.
These and other objects are achieved by providing a system for inspecting the surface of an article wherein a light beam is scanned across the surface of the article. A reflected main portion of the beam is spatially filtered so that only reflected portions of the beam, which are scattered by defects, are detected beyond the spatial filter. A beam focusing lens is positioned at a point that is midway between a beam scanning device and the article surface, which is undergoing inspection. The lens has a focal length equal to one-half the distance between the scanning device and the surface. The result of the combination of elements is that the returned beam is stationary and can be precisely controlled to disclose extremely small defects. A counter and display system are triggered by the detector to provide both a quantitative and graphic display of the defects which are found.
BRIEF DESCRIPTION OF THE DRAWINGS Other objects and features of the present invention will be more readily understood from the following detailed description of specific embodiments thereof, when read in conjunction with the appended drawings in which:
FIG. 1 is a perspective overall view of an inventive inspection apparatus;
FIG. 2 is an illustration of an alternate optical arrangement of the apparatus of FIG. 1;
FIG. 3 is a schematic diagram of a control system for operating the apparatus of FIG. 1;
FIG. 4 is a schematic diagram of an automatic gain control system of the apparatus of FIG. 1;
FIG. 5 is a circuit diagramof a non-inverting amplifier used in the automatic gain control system of FIG. 4; and
FIGS. 6-11 are various graphical representations of interactions of parameters within the automatic gain control system.
DETAILED DESCRIPTION The invention will be described in connection with a technique for inspecting a slice of single crystal silicon on which an epitaxial layer of silicon has been formed. The silicon is used in the manufacture of semiconductor devices. It is to be understood, however, that the invention is broader than that disclosed in the examples and has applicability to the inspection of many types of surfaces.
GENERAL OPERATION Referring now to FIG. 1, an expanded beam of laser light 20 emanates from a laser 22 and is deflected from a torsionally oscillating mirror 24 toward a large lens 26, placed one focal length away. A silicon wafer 28 to be inspected is placed one focal length away from the lens 26. The oscillating mirror 24 and lens 26 combination serves to produce a normally incident, highly focused spot of light which travels in a line scan back and forth over the wafer 28. Movement of the wafer 28 in a direction perpendicular to the line scan by a drive table 30, results in a complete coverage of the surface to be inspected by an effective two-dimensional raster of the focused spot.
The light reflected from the specular surface of the wafer 28 is directed back up through the lens 26 and reflected off the oscillating mirror 24 a second time as a return beam 32. If the mirror 24 and wafer 28 are both at exactly one focal length from the lens 26, this return beam 32 is completely immobilized even though it originates from the moving spot scanning the surface 28. In other words, this beam suffers no movement whatsoever as the focused spot scans over the surface of the wafer 28.
It is extremely important to position the oscillating mirror 24, the wafer 28 and the lens 26 so that the arrangement of focal lengths specified above can be satisfied exactly. In order to achieve accuracy in the desired positioning, the lens 26 is mounted on an adjustable base 31. The vertical position of the lens 26 can thus be very accurately controlled.
if the focused spot strikes a defect on the surface of the wafer 28, the light is scattered. The specular return beam 32 is diminished in intensity, and the region of space immediately surrounding the return beam has light scattered into it. It is more advantageous, from a signal-to-noise point of view, to detect the defect by this scattered light 33 than by a diminishment of the intensity of the return beam 32.
The particular geometry employed in generating the scan beam 20 and return beam 32 permits the incorporation of a novel feature to accomplish this end. A stop or spatial filter 34 is placed to intercept the return beam 32 after reflection from a mirror 36. Some of the scattered light 33 surrounding this beam is imaged onto a small detector 38. An electronic signal resulting from the defect light 33 hitting the detector 38 is then processed to size, count and locate the defect.
The laser light source 22 is a 3 mw He-Ne laser available from Hughes Electron Dynamics Division of Torrence, Calif, as model 3076 l-i/R. A X beam expander is used to permit a more effective utilization of the lens aperture and hence a more finely focused spot. The use of a Carl Meyer F/l.5 135 mm lens as the focusing lens 26 permits surfaces as large as 75 mm in diameter to be inspected with a focused spot of 13 mi-,
crons. This finely focused spot permits defects as small as 2 microns in diameter to be detected.
Most of the beam passes through a hole 42 in the mirror 36 which is aluminized on both front surfaces. The outer portion of the beam 20 which does not go through the hole is reflected at right angles as beam 40 and focused onto a silicon photodiode 44. The signal from this photodiode 44 is used to monitor the intensity of the light from the laser 22, and to provide automatic gain compensation (AGC) for small fluctuations in the laser light output.
The beam 10 is reflected from the oscillating mirror 24, which is driven at approximately 100 hz by galvanometer drive 46 available from General Scanning, Inc.- of Watertown, Mass. as Model G-325. The beam 10 is directed somewhat to one side of the optical axis down into the lens 26;
The wafer 28 is placed flat at the focal plane of the lens 26 where the cone of light is brought down nearly normal (but at a slight angle) to a focus at the surface of the wafer. The specularly reflected light and a portion of the scattered light is collected by the lens 26 and recollimated to produce a return beam 32 which emerges from the lens on the other side of the optic axis in such a fashion as to strike the oscillating mirror 24 nearly adjacent to the input beam 20. The displacement of beam 20 with respect to beam 32 is exaggerated in FIG. 1 for purposes of clarity.
The oscillating mirror 24 then directs the beam 32 back toward the mirror 36, which in turn reflects the beam 32 over toward the detector 38. The specular component of the beam 32 is removed by the physical stop 34 placed just in front of a simple lens 48 which collects and focuses the .scattered light 33 onto a low dark current silicon photodiode detector 38. Voltage pulses from this diode caused by flashes of light from the defects are processed in the electronics package of the instrument.
In order to insure that the entire wafer 28 is scanned, the length of the sweep must be somewhat greater than the diameter of the wafer. As the focused spot sweeps over the edge of the wafer, spurious counts are generated by the wafer edge and the machined surface of the table 30. For a meaningful count to be generated, these signals must be eliminated. This may be accomplished in one of two ways.
The simpler approach involves positioning a mask 49 above the wafer 28 and attached to the table 30 to travel along with it. The mask 49 is slightly smaller than the wafer 28, and is at a position in which the beam 20 is not well focused. This prevents spurious counts from being generated.
A second approach involves the use of a third detector (not shown) to sample the unscattered beam 32 reflected from the wafer 28 so that the electronics can be activated only when the beam 20 is on the, wafer.
An alternative optical arrangement for detecting defects which permits great utilization of the aperture of the lens 26 is sketched in FIG. 2. The layout is substantially the same as in FIG. 1, except that the input beam 20 is centered on the optic axis of the scan lens 26. This causes the scanning beam 20 to be incident on the wafer 28 exactly normal, so that the return beam 32 follows exactly the same path as the input beam. Thus, the return beam goes back through the hole 42 in the mirror 36. The hole 42 in the mirror 36 now serves as the spatial filter, since only the scattered light 33 will strike the reflecting surface to be directed to the signal detector 38. This configuration permits scattered light from the entire aperture of the lens 26, minus the area required by the input beam 20, to be directed to the de' fect detector 38.
The arrangement of FIG. 1 sacrifices a second area equal to the beam size in the collection of the scattered light. The primary advantage of the scheme of FIG. 1 is that very little light is directed back into the laser 22. The laser 22, therefore, produces a more uniform output amplitude of the light.
Depending on the exact nature of the defects to be detected, one or the other of the two schemes might enjoy a slight advantage.
ELECTRONICS A block diagram of the electronics is shown in FIG. 3. The x-scan is provided by the motor-driven table of FIG. 1. The electronics is programmed to control the table 30 and provide automatic operation of the inspection procedure.
When a start button 50 is pressed, the table 30 is activated, and the results from the previous run are erased from a large screen storage oscilloscope display 52. A counter 54 is reset to zero. As the wafer 28 is reached, the counting and display functions are turned on, and the count is accumulated as the wafer is scanned. The x-axis signal for the drive on the display is derived from a linear potentiometer 55. At the end of the wafer 28, the table 30 is returned to the initial position for convenient loading of the next wafer, while the display and count are held for leisurely examination and recording.
The y-scan (mirror oscillation is derived from a master oscillator 56 which runs at a constant frequency and generates a sine wave. A power amplifier 58 is used to generate the appropriate signal at the proper impedance to drive the mirror galvanometer 46. The ydeflection for the display 52 is derived from the same master oscillator 56. The position of the electron beam in the display 52 must correspond to the position of the laser spot on the wafer 28. Therefore, the phase shift introduced by the power amplifier 58 and galvanometer 46 must be compensated for in driving the display 52. In order to derive the retrace blanking for the display output, a second phase shifter 59 is required. 1
I The current signal from the photodiode of the detector 38 is transformed into a voltage proportional to the light intensity in a detector preamp 60. This signal is amplified in an automatic gain control (AGC) amplifier 62 by a factor of 1.5 to 2.5. The gain of this stage is adjusted according to the laser intensity in such a manner as to provide for compensation of variations in the laser output intensity. The signal from the AGC amplifier 62 is high-pass filtered and given a further boost in amplitude.
A level detector 64 is used to select those voltage pulses resulting from defects above a certain size. These shaped pulses are then passed through a logic network 66 to perform retrace blanking and pulse width selection. There are two outputs of the logic network, and they are essentially identical. One pulse stream is simply counted in the totalizing counter 54 to provide the estimate of the total number of defects on the wafer. The second group of pulses are used to turn on at the appropriate times the z-axis of a storage oscilloscope display 52, such as that sold by Tektronix under the trade designation 61 1." Therefore, as the wafer is scanned, a picture of the defect locations is presented on the display 52.
As an alternative to the defocusing mask 49, a spot position monitor (SPM) channel 67 can be used to insure that defect signals are processed only when the focused spot is on the surface of the wafer 28. Signals at the wafer edge may still be counted as defects if the SPM does not turn off fast enough. This problem is solved by making the SPM delibrately slow. Only the first half of the wafer 28 is scanned on the trace portion of the sweep, with the remaining portion being scanned on the retrace. In this way, there is an automatic blanking performed for a short time immediately after the beam comes onto the wafer, eliminating the edge signals.
AUTOMATIC GAIN CONTROL An automatic gain control (AGC) system shown in FIG. 4 is used to cancel out the effect of variations in the output light intensity of the laser 22. A portion of the laser beam is directed to the AGC detector 44 via the mirror 36. The voltage derived from AGC diode 44 is ultimately used to control the overall gain of the signal processing chain, so that accurate countsmay still be obtained even in the presence of laser intensity fluctuations.
The variable gain is achieved in the AGC amplifier 62 (see FIG. 5) by varying the voltage on the gate of a field effect transistor (FET) 70. The FET 70 has an on resistance vs. gate voltage curve which is sketched in FIG. 6. Note that the resistance becomes highly non-linear at low gate voltages, and tends toward very large values. The gain of the AGC amplifier 62 is given by G (R,+ R) /R, where R, is the resistance of the FET 70 and R is the resistance of the feedback resistor 71. The gain as a function of the resistance R, is also a non-linear function, shown in FIG. 7. With the proper feedback resistor 71, these two non-linear effects shown in FIGS. 6 and 7 cancel out over a limited range, giving a region of linear control. An experimental curve of DC gain vs. gate voltage is shown in FIG. 8.
A divider chain consisting of a 3 megohm resistor 72 and a 150,000 ohm resistor 74 on the imput of an operational amplifier 76 produces a constant 0.75 volt offset at both inputs 78 and 80. This acts as a bias voltage across the FET 70. As the signal level changes, this offset level shifts, producing changes in the effective curve of on resistance" vs. gate voltage for the FET 70. A 0.01 ,uf capacitor 82 in series with a 24,000 resistor 84 serves to correct for the non-linearities in the output signal produced by these changes.
The control voltage forthe gate of the FET 70 is derived from a variable gain preamplifier 84, plus an inverting amplifier 86 with a variable offset to control the operating portion of the curve for the FET.
The desired operation of the AGC system can be set forth as follows: For a given size defect, the output signal response should be independent of the laser light intensity, over some finite operating range. This is indicated in FIG. 9. There are a family of curves of the variation of signal response as a function of the gain and bias of the AGC preamplifier. These are shown in FIGS. 10 and 11. Naturally, the gain and bias must be adjusted to give an operating characteristic which is like that shown in FIG. 9. This is generally achieved with a bias voltage on the amplifier 86 (with the amplifier 84 set to zero gain) of 5 to 6 volts, and with the gain of amplifier 84 adjusted to bring the input of the FET 70 gate back down to 3.5 to 4 volts.
Although certain embodiments of the invention have been shown in the drawings and described in the specification, it is to be understood that the invention is not limited thereto, is capable of modification and can be arranged without departing from the spirit and scope of the invention.
What is claimed is:
1. In apparatus for inspecting an article having a reflecting surface comprising:
means for producing a collimated input light beam;
beam focusing means positioned in the path of said input beam between said beam producing means and said article; and Y tiltable light-reflecting means positioned in the path of said beam between said beam producing means and said focusing means for scanning said input beam across said focusing means, the improvement being that:
the reflecting means is positioned such that the axis of rotation of the reflecting means intersects one focal point of the focusing means; and means are included for positioning a portion of the surface to be inspected at the opposite focal point of the focusing means with said portion inclined sufficiently nearly normal to the optical axis of the focusing means that light reflected from the surface of the article passes through the focusing means and is reflected by the reflecting means whereby said light reflected from said portion becomes an immobilized, collimated return beam after reflection from said reflecting means.
2. The apparatus of claim 1 further comprising a spatial filter in the path of said return beam for separating light which is reflected specularly by said surface from light which is reflected scattered by said surface.
3. The apparatus of claim 2 further comprising means for sensing the scattered light separated by said filter and utilizing said sensing to produce an indication of imperfections in said surface.
4. Apparatus for inspecting an article whose surface reflects light specularly except from imperfections which scatter light, said apparatus comprising:
a source of a collimated input light beam;
a first planar mirror illuminated by said beam, said mirror being placed at an angle to the beam path and having a portion of its beam-illuminated area transmissive to said beam;
a second planar mirror illuminated by the portion of said input beam transmitted by the first mirror, and tiltable with respect to the beam path to produce scanning movement of the inputbeam as reflected from said second mirror;
a focusing lens illuminated by the scanning input beam as reflected from said second mirror, said focusing lens being positioned in the optical path between said second mirror and said article surface;
the second mirror being positioned such that its axis of rotation intersects one focal point of the lens; and
means for positioning a portion of the surface of the article at the opposite focal point of the lens with said portion inclined sufficiently nearly normal to the optical axis of the focusing lens that light reflected from said portion passes through the focusing lens whereby there is produced a specular return beam reflected from said surface and a region immediately surrounding said specular return beam of scattered light reflected from imperfections in said surface, said specular return beam and surrounding region as reflected by said second mirror being collimated and immobilized.
5. The apparatus of claim 4, further comprising:
a third mirror formed on the back of said first mirror and illuminated by said return beam and surrounding region reflected by said second mirror; and
spatial filtering means for selecting from said illumination of said third mirror only said beam surrounding region composed of scattered light reflected from said imperfections.
6. The apparatus of claim 5, wherein said focusing lens is positioned relative to said mirrors so that said specular return beam and surrounding region illuminate the third mirror along a path displaced from the input beam path, and wherein said spatial filter comprises a light stop in the path of said return beam as re flected by said third mirror.
7. The apparatus of claim 5, wherein said focusing lens is positioned relative to said mirrors so that said specular return beam and surrounding region illuminate the third mirror along a path substantially coaxial with the input beam path, and wherein said spatial filter comprises a light reflecting annular portion of said third mirror surrounding said beam transmissive area of said first mirror.
8. The apparatus of claim 5, further comprising:
means for producing a signal whenever said selected portion exceeds a predetermined level; and
means for electronically counting said signals to obtain a numerical value of said imperfections.
9. The apparatus of claim 8 further comprising:
means for producing a second signal from the portion of the input beam reflected by said first mirror; and
means for utilizing said second signal to control the first-mentioned signal so that a uniform sensitivity is maintained in the presence of fluctuations in the input beam.
10. The apparatus of claim 5, further comprising means for producing a signal whenever said selected portion exceeds a predetermined level;
means for scanning a cathode ray tube in a pattern coordinated with the scanning of said article surface by said input beam; and
. means for developing an image on said tube with said signals to obtain a visible pattern of said imperfections. 11. The apparatus of claim 10, further comprising means for electronically counting said signals to obtain a numerical value of said imperfections.
12. The apparatus of claim 5, wherein said article is the epitaxial layer of a single crystal material.
13. The method of inspecting an article having a reflecting surface, comprising the steps of:
producing a collimated input light beam; focusing with focusing means said input beam upon the surface of said article from a predetermined distance while simultaneously defocusing and recollimating with the same focusing means light specularly reflected from the surface; reflectively scanning said beam before it is subjected to said focusing, the center of said scanning being twice said distance from said surface; and
positioning the article such that light reflected from said surface is again subjected to said scanning reflection so as to become thereafter an immobilized, collimated return beam.
14. The method of claim 13 further comprising spatially filtering said return beam to separate light which is specularly reflected by said surface from light which is reflected scattered by imperfections in said surface.
counting said signals to obtain a numerical value of said imperfections.
17. The method of claim 14, further comprising:
producing a signal whenever said separated light exceeds a predetermined level; and
producing a visible pattern of said signals corresponding to the pattern formed by said imperfections.