US 3740563 A
An electroptical system for controlling the diameter of crystals pulled from a melt wherein separate process control loops simultaneously provide: (1) closed loop melt tracking for the system optics, (2) direct crystal diameter closed loop control, and (3) normalizing brightness control for a novel photodetector and associated novel optical geometry of the system. This photodetector is uniquely constructed to simultaneously provide the above three functions and is further especially adapted for use as an integral portion of various feedback loops which comprise the electroptical system.
Description (OCR text may contain errors)
United States Patent 1 Reichard June 19, 1973  Filed:
[ 1 ELECTROPTICAL SYSTEM AND METHOD FOR SENSING AND CONTROLLING THE DIAMETER AND MELT LEVEL OF PULLED CRYSTALS  Inventor: Thomas E. Reichard, Kirkwood,
[ Assignee: Monsanto Company, St. Louis, Mo.
June 25, 1971  Appl. No.: 156,760
 U.S. Cl. 250/222 R, 250/211 R, 23/273 SP,
 Int. Cl B01d 9/00  Field of Search 250/217 R, 216, 222 R,
250/211 R, 211 K, 211 J, 83.3 H, 219 S, 218;
23/301 SP, 273 SP; 356/199, 159, 160
 References Cited UNITED STATES PATENTS 3,493,770 2/1970 Dessauer et al. 250/222 X 3,459,949 8/1969 Poncet 250/222 R 3,244,889 4/1966 Preston et al.... 250/211 R 3,028,500 4/1962 Wallmark 250/211 .1 3,621,213 11/1971 Jen et al. 23/302 SP 6/1971 Baker 250/216 4/1971 House 250/216 X OTHER PUBLICATIONS Patzner et al: Automatic Diameter Control of C20- chraski Crystals; SCP and Solid State Technology; Oct. 1967; pp. 25-30.
Primary ExaminerWalter Stolwein Attorney-H. R. Patton  ABSTRACT An electroptical system for controlling the diameter of crystals pulled from a melt wherein separate process control loops simultaneously provide: (1) closed loop melt tracking for the system optics, (2) direct crystal diameter closed loop control, and (3) normalizing brightness control for a novel photodetector and associated novel optical geometry of the system. This photodetector is uniquely constructed to simultaneously provide the above three functions and is further especially adapted for use as an integral portion of various feedback loops which comprise the electroptical systern.
20 Claims, 7 Drawing Figures P TO PRE- AMPLIFIERS Patented June 19, 1973 3 Sheeta-$heet 1 FIG. IB
M a 8 1 w 5 1 0 J/ 4 4 4 4 a f z m m 2 M CS; \fl n y u$$ W F 2 h. m T F w W INVENTOR THOMAS E. REICHARD BY M4 4 9 M ATTORNEY Patented Jurge 19, 1973 3,740,563
3 Sheets-Sheet 2 TO BULB POWER SUPPLY FIG. 3 T0 PULL RATE SERVO MOpOR TO MELT TRACK- ING PROCESS CONTROLLER TO TEMPERATURE CONTROL SYSTEM INVENTOR THOMAS E. REICHARD BY I Judah ab.
. ATTORNEY ELECTROPTICAL SYSTEM AND METHOD FOR SENSING AND CONTROLLING THE DIAMETER AND MELT LEVEL OF PULLED CRYSTALS FIELD OF THE INVENTION This invention relates generally to pulling single crystals from a melt and more particularly to an improved electroptical system and method for precisely sensing and controlling the diameters of pulled single crystal semiconductor rods.
BACKGROUND OF THE INVENTION Pulling single crystals of various types of materials, e.g. semiconductive materials, to close diameter tolerances is an important phase of the mass production of a variety of electronic devices. If a uniform crystal diameter can be maintained throughout a crystal pulling operation, manufacturing time and material waste will be greatly reduced, and tooling and labor cost savings will also be realized in the crystal growing as well as subsequent device manufacturing processes. In the absence of some form of crystal diameter control, single crystals pulled by the Czochralski method, for example, will have widely varying diameters which must be shaved down to a uniform size in preparation for subsequent wafer processing steps.
DESCRIPTION OF THE PRIOR ART Presently, the diameter of semiconductor crystals pulled by the Czochralski method are frequently controlled by skilled operators who continuously make small adjustments to the semiconductor melt temperature. These adjustments have the effect of controlling the crystallization rate of the semiconductor melt at its solid-liquid interface and thus provide a strategic control of the pulled crystal diameter. However, these manually controlled open loop systems cannot provide the high degree of crystal diameter control that automatic closed-loop process control systems are potentially capable of providing.
Other prior art, closed-loop electroptical systems, including pyrometer systems, have been used wherein the light emitted (including reflections) from the liquid meniscus portion of the semiconductor crystal being pulled is used to generate an electrical control signal which varies as a function of crystal diameter. Such electrical signal is in turn used to alter system parameters, such as crystal pull rate or the semiconductor melt temperature, or both, to thereby maintain the semiconductor crystal size within certain tolerances. However, these prior art closed-loop control systems have been subject to control errors which are introduced into the system as a result ofinherent and environmental system changes. These changes include: (1) effective shifts in diameter set-point and control-loop gain resulting from variations in light transmission efficiency along the optical path of the system e.g. by window fogging, (2) a varying pattern of brightness gradient near the crystal due to changing reflection patterns from the susceptor walls during a crystal pulling operation (run), and (3) a rather limited range of crystal radius variation within which the light sensor response is linear. As a result, changes due to crystal eccentricities, flats and other departures of the crystal from a perfectly cylindrical shape are not necessarily correctly averaged to a true diameter signal.
Also, the mechanical structure of most crystal pullers is such that any optical viewing path must be tilted. slightly away from the axis of the crystal, and therefore any variation in the distance from sensor optics to melt surface creates a systematic error in the effective diameter setting. In prior-art systems this distance is determined and controlled only by indirect open-loop means, and is subject to several sources of error within a run and between runs.
The dependence of these prior art closed-loop electroptical control systems on such environmental changes limit the crystal diameter control capability of the system. Various means for eliminating or minimizing these sources of errors and uncertainties are provided in accordance with the teachings of the present invention.
SUMMARY OF THE INVENTION The general purpose of this invention is to provide a new and improved closed-loop electroptical system and method for precisely controlling the diameter of pulled crystals and the present system has been designed to overcome the aforedescribed disadvantages. To attain this improved crystal diameter control, a novel closedloop electroptical process control system has been constructed, and the present electroptical system operates substantially independent of inherent environmental changes in a semiconductor crystal puller with which it operates. The present system simultaneously converts variations in semiconductor crystal melt level and semiconductor meniscus lateral position into separate electrical control signals which are continuously processed in separate closed feedback loops to provide good closed-loop control of the crystal diameter throughout a crystal pulling operation. The electroptical system according to the present invention also features a novel photodetection means and novel optical system geometry for simultaneously generating melt level and meniscus position controls signals independently of each other. And the voltage amplitude-vs-reflected image displacement response characteristics of the signal generating photosensitive elements of this photodetection means make this detection means uniquely suited for optimum response to melt level and crystal diameter changes which occur during a normal crystal pulling operation.
Accordingly, an object of the present invention is to provide a new and improved electroptical system for precisely controlling the diameter of a pulled crystal.
Another object is to provide an electroptical system of the type described which is relatively insensitive to variations in light levels within a crystal puller assembly with which it operates.
Another object of this invention is to provide a new and improved crystal diameter control system and method of the type described herein which is operable to greatly reduce semiconductor material waste, operator time, and the tooling and labor costs presently involved in large scale semiconductor crystal pulling operations.
A feature of the present invention is the provision of a closed-loop electroptical feedback system which includes frequency, phase, and/or optical discrimination of the reflected optical signals therein.
Another feature of this invention is the provision of differential detection within the system to optimize the closed-loop feedback control of the diameter of crystals being pulled; this detection insures good common mode rejection for the system.
Another feature of this invention is the provision of normalizing feedback control for compensating for environmentally induced brightness variations within the system.
Another feature of this invention is the provision of novel photodetection means and optical geometry therefor for insuring optimum sensitivity to the reflected diameter control image of the system.
The above as well as other objects and features of this invention will become more fully apparent in the following description of the accompanying drawings wherein:
DRAWINGS FIG. 1A is a front elevation view, shown partially in cross-section, of the optics portion of the control system according to the invention. FIG. 1A illustrates the shift of the image reflected by the melt meniscus and received by the photodetector with a drop in melt level within the crucible.
FIG. 1B is a plan view of the novel optical geometry of the invention and illustrates the layback angle, -y, to be further described;
FIG. 2A is a perspective view of the novel photodetector of the present invention;
FIG. 2B is a plan view of the photodetector shown in FIG. 2A and illustrates the exact shape of the photosensitive surfaces thereof and the reflected image received thereby;
FIG. 2C is an end elevational view of the photodetector in FIGS. 2A and 2B and further includes an associated focusing lens for providing a desired beam refraction for the image projected onto the photodetector;
FIG. 3 is a partially cut-away elevational view of a crystal puller with which the present invention operates, and FIG. 3 also illustrates the optics portion (and mechanical support therefor) of the closed-loop system embodying the invention; and
FIG. 4 is a functional block diagram of the multiple closed-loop electroptical process control system embodying the present invention.
DESCRIPTION OF THE PREFERRED EMBODIMENT Referring to FIGS. 1A and 13, there is shown a semiconductor crystal which is pulled from a melt 12 within a quartz crucible 14, and the crucible 14 is heated, such as by RF or resistance heating, in a wellknown manner in a graphite susceptor 16. The crystal puller apparatus within which the susceptor 16 is located is of the conventional Czochralski type of crystal puller and is therefore not described or shown in further detail in FIG. 1. The optics chamber 17 of the system includes a tungsten light bulb 80 which is the single light source for the electroptical system. The bulb 80 is positioned adjacent a condenser lens 82 which coverges and focuses the light as shown toward and through an optical slit 84. The light passing through the optical slit 84 is chopped by a conventional mechanical light chopper comprising a motor 86 driving a rotating shutter 88 at a desired light chopping frequency. The chopping frequency is typically in the order of 270 Hertz and this chopped light permits the light dependent electrical signals processed in the system to be frequency and/or phase discriminated as will be described herein.
The chopped light beam passes to a projection lens 92 from which it is projected through an optical filter 31b and onto the liquid meniscus 19 of the semiconductor crystal 10 at a chosen angle of incidence a with respect to the longitudinal axis of the crystal 10. A portion of this incident light is reflected from the liquid meniscus and to a photodetector 21 to be described at an angle of reflection B, where ,8 a. The total angle a B is referred to as the binocular angle of the system.
A portion of the reflected light beam passes through an optical filter 31 and through a receiving lens which focuses the beam onto the photosensitive surfaces of the photodetector 21 to be further described below with reference to FIG. 2. As the surface of the melt l2 recedes in the crucible 14 from a level 97 to a lower level 98, the location of the light pattern focused on the liquid meniscus 1Q shifts from position 96 to a lower position 99 and thereby changes the reflected optical path from the solid line 101 to the dotted line 103 as shown. This change causes a shifting of the reflected image pattern on the photosensitive surfaces of photodetector 21 from a previously held position 102 to a new position 104 which is slightly counter-clockwise with respect to position 102. This slight shifting in positionin the X direction as shown in FIG. 2B generates an unbalance in the A and B signal voltages at the output of the photodetector 21, and these voltages are processed in a first or melt tracking feedback loop to be described.
If a change in the diameter of the crystal 10 causes the liquid meniscus 19 upon which the light beam 18 is focused to move laterally outward or inward during the crystal pulling operation, such a movement will cause the image pattern 116 received on the photosensitive surfaces of the photodetector 21 to be shifted in a second coordinate (Y) direction in FIG. 2B. As viewed in FIG. 1A such image movement will be into or out of the sheet of drawing, depending upon whether the diameter of the crystal 10 is increased or decreased. This shifting of the image pattern 116 on the surfaces of the photodetector 21 generates an unbalance in the C and D signal voltages at the output of photodetector 21 which are electrically processed in feedback loops to be described.
The plan view of FIG. 1B (and the side elevational view of FIG. 3A) illustrate the layback angle 7 which is the angle between a vertical line parallel to the longitudinal axis of the crystal 10 and a plane containing the centerlines of the projected and reflected light beams. FIG. 1B also illustrates the illuminated area 85 on the meniscus 19. In a preferred embodiment of this invention, this area is approximately 0.52 inch long by 0.04 inch wide, being a sharply focused, uniformly illuminated image of the optical slit 84. The longer dimension of this illuminated area is aligned radially outward from the crystal axis and essentially spans the entire curved portion of the liquid meniscus 19. However since this meniscus has a continuous concave curvature upward from the horizontal melt surface to essentially vertical at the crystal'edge interface (as shown on FIG. 3) most of the light is reflected at angles which do not enter the receiving lens 100. Reflected light enters the receiving lens 100 and thereby impinges on the photosensitive surfaces of the photodetector 21 only from a relatively short segment of the illuminated strip 85 where the upward tilt of the meniscus 19 from the horizontal (angle y on FIG. 3A) approximately equals the layback angle 'y on FIG. 3. The meniscus curvature remains essentially constant throughout the crystal pulling operation so that the meniscus spot having upward tilt equal to 7' occurs always at an essentially constant distance out from the edge of the crystal rod 10. This spot is effectively locked onto the periphery of the crystal at a fixed distance therefrom and moves with it, following any changes in overall crystal size and variations due to rotational eccentricity and/or any non-circular shape effects.
The entire illuminated area 85 on the melt (FIG. 1B) is imaged (FIG. 28) onto the photosensitive surfaces of photodetector 21 so that its long axis approximately covers and matches the entire photosensitive length in the yaxis direction. The layback angle -y remains fixed throughout a run, and the entire optics chamber 17 remains at a constant horizontal distance from the crystal rotation axis throughout a given run. The optics chamber 17 only moves vertically during a run to track the level 97 of the melt 12, and the reflected spot 116 (FIG. 2B) seen by the photodetector 21 at any given 7 moment is at a position along the y axis of the photodiode array which is directly proportional to the crystal radius at that particular moment. Variations in crystal radius cause the detected spot to move correspondingly along the y axis of the detector, tracking all variations in crystal radius, shape, and rotational eccentricity.
The photodetector 21 is illustrated in some detail in FIGS. 2A, 2B and 2C and includes a first sensing means comprising first and second rectangular photosensitive areas or surface portions 108 and 110 which generate the A and B signal voltages, respectively. The photodetector 21 also includes a second sensing means comprising third and fourth, tapered wedge photosensitive areas 112 and 114, respectively, which generate the C and D signal voltages.
' Referring in more detail to FIG. 2B, it will be apparent that a shift of the reflected image 116 on the first or X coordinate axis and across the first and second rectangular photosensitive surfaces 108 and 110 will cause more or less light to be received on each of the latter surfaces and these surfaces will generate the A and B signal voltages in proportion to the levels of light impinging thereon. The rectangular geometry of the surfaces 108 and 110 provide a rather steep A-B difference voltage amplitude versus X-direction displacement response characteristic, so that a small shift of the reflected light pattern 106 in the X direction generates a relatively large A-B error difference voltage. This dif ference voltage is advantageously utilized to provide a tight closed-loop control for tracking the receding level of the semiconductor melt 12. Since any variation in the distance from the melt surface 97 to the optics chamber 17 results in a systematic error in the apparent crystal diameter, it is advantageous to utilize this tight closed-loop process control to insure that the vertical position of the optics chamber 17 closely tracks the vertical level of the melt surface 97.
It is desirable that the received light pattern 116 overlap only a relatively small portion of each of the rectangular photosensitive areas 108 and 110. This insures that a slight X direction displacement of the pattern 116 will simultaneously increase substantially the percentage of light impinging on one of these rectangular areas while decreasing the total percentage of light impinging on the other rectangular photo-sensitive area.
At the same time that the melt level 97 is receding in the crucible 14, the semiconductor crystal rod 10 is rotating continuously, typically at about 15 rpm. As a result of variations of the rod 10 from a perfectly cylindrical shape, and eccentricities of rotation which often occur, the rotational motion of the rod 10 produces systematic fluctuations in the exact meniscus location from which the light image 96 is received. The latter fluctuations produce a corresponding shifting in the Y axis position of the detected image pattern 116. Therefore, in order to provide a smooth linear response for the quantitative averaging of these fluctuations, and thereby obtain a signal truly representative of average crystal diameter, it is desirable that the third and fourth photosensitive surfaces 112 and 114 provide a gradual linear C-D difference voltage amplitude versus Y direction displacement response characteristic. In the preferred embodiment, the tapered wedge geometry of the photosensitive surfaces 112 and 114 provides the desired linear response pattern over a range of i. l 25 inch of crystal radius fluctuation, corresponding to i.25 inch of crystal diameter. As shown, the areas 112 and 114 are separated along the y-axis with their adjacent edges forming a serpentine path along the x-axis such that areas 112 and 114 have interdigitated portions. The C-D difference voltage variation produced by crystal eccentricities and flats are averaged to a minimal net error using appropriate low pass filters (not shown) in the demodulator section 28 of the electroptical system. It should be observed here that a shift of the illuminated pattern 116 in the X direction does not affect the total light from the illuminating source which impinges the C and D photosensitive surface areas 112 and 114. Conversely, a shift of the pattern 116 in the Y axis direction does not affect the total light from the illuminating source 80 which impinges the A and B photosensitive surface areas 108 and 110.
In order to minimize the response of the melt level sensing function to the extraneous effects of rotational eccentricity, mechanical misalignments and the like, it is desirable to illuminate only a narrow, sharply defined radial band (see FIG. 1B) of the crystal meniscus 19 by projecting thereon a focused image of a narrow illuminated slit 84. In a preferred embodiment of the invention, this illuminated band width of the image pattern 96 in FIG. 2 is on the order of 0.04 inch wide and 0.52 inch long to accommodate these eccentricities and misalignments. Thus, the rectangular shape of the slit 84 determines the rectangular shape of the dotted pattern 109 which represents the area on the photodetector surfaces which would be illuminated if the image 96 was received from a perfectly flat mirror surface properly oriented and tilted at angle y on FIG. 3.
Referring to FIG. 2C, it is frequently desirable to use a concave cylindrical lens 122 immediately in front of the photodetector 21 in order to refract the light image 116 to a desired width. The lens 122 with its long cylindrical axis oriented along the Y axis direction insures that an image 116 of a desired size is received on the photosensitive surfaces of the photodetector 21 to thereby generate the A, B, C and D signal voltages previously mentioned.
Referring now to FIG. 3, there is shown a semiconductor crystal puller housing having sidewalls 12 1, a bottom wall 125, and a top wall 126 in which cylindrical viewing ports 127 are rigidly mounted. The ports 127 include windows 128 through which the projected and reflected light beams pass to and from the semiconductor crystal melt 12. The heat-reflecting optical filters 31a and 12 (see FIG. 1) are also mounted in the viewing port 127 and preferentially block most of the longer-wavelength infrared radiation from the hot melt, crucible, and susceptor 12, thereby shielding the optics chamber 17 from excessive heat and also discriminating preferentially in favor of the shorter wavelength tungsten-filament bulb illumination spectrum. The systems projection and detection optics described above with particular references to FIGS. 1 and 2 are mounted in the optics chamber 17, and the vertical location of this chamber 17 as well as the layback angle 7 of the plane of the center lines of the two optical paths 18 relative to a vertical line parallel to the vertical axis of the crystal 10 can be mechanically adjusted as will be described herein.
The optics chamber 17, micrometer 158, and rack 156 are all mounted on the front movable half 144 of the linear roller bearing assembly 144 and 142 so that an adjustment of the micrometer 158 rotates the entire optics chamber around two spring-loaded ball joints and thereby establishes the diameter setting (optical null). This produces slight unwanted changes in the layback angle y for different sizes of crystals, but is not easily avoided and merely becomes part of the calibration of micrometer setting vs. crystal diameter. The rear stationary half 142 of the linear bearing assembly is mounted on the bracket 132 using bolts 140 and 171. These bolts may be adjusted to align the movement of the bearing precisely parallel to the rotational axis of the crystal l0; i.e., precisely vertical.
A D.C. servomotor drive unit 138 is fixed to the stationary half 142 of the linear bearing assembly, and the drive pinion 154 of the servomotor unit 138 engages the rack 156 which is affixed to the movable half 144 of bearing assembly 142 and 144. This servomotor unit 138 operates to move the optics chamber 17 vertically, up and down, following any changes in the level of the melt l2.
The servomotor unit 138 is rigidly secured to the stationary half 142 of the linear bearing table 142 and 144. The servomotor 138 provides the rotational drive for a small drive-worm gear 150 which intermeshes with a larger worm gear wheel 152. The gear wheel 152 is mounted on a common axis (not shown) with the drive pinion 154, and the drive pinion 154 intermeshes with the rack gear 156. Therefore, the rotation of the driveworm 150 drives the pinion 154 which is intermeshed with the rack gear 156 to drive the entire optics chamber 17 vertically downward at a rate equal to the rate of recession of the melt 12 within the crucible 14.
A manually rotatable wheel 146 is mounted on the same shaft with the gear wheels 152 and 154 and may be used to manually raise and lower the optics chamber 17. The larger wheel 146 may be rotated to directly drive the rack 156 vertically upward or downward, and during this manual operation, a clutch (not shown) allows the gear wheel 152 to slide and be overriden as it is held stationary by its engagement with driveworm 150.
Referring now to FIG. 4, there is shown the complete electroptical feedback control system embodying the present invention. The crystal puller portion 9 and optics chamber 17 have been previously described above,
and the optics chamber 17 houses the light source 80, the light chopper 86, 88, the slit 84, the lenses 82 and associated with the latter optical elements, the receiving lens and the photodetector 21 atwhich the low level A, B, C and D signal voltages are generated.
The desired signals are generated at the photodetector 21 as low-level square-wave voltage pulses corresponding in frequency and phase to the chopping of the illuminating beam. The photodetector 21 receives much larger amounts of light emitted from the hot melt itself and reflected from the incandescent crucible, susceptor, and other parts of the crystal pulling chamber, and the amount of light reaching the photodetector 21 from these other sources may undergo long-term drifts due to changing conditions through the run, as well as short-term fluctuations at various frequencies due to vibrations and rippling of the melt. Furthermore, the electronic circuit components may tend to pick up various voltage signals at still other frequencies from stray R.F. or electromagnetic interferences. Therefore, in accordance with the present invention it is very desirable that the early stages of signal processing discriminate heavily in favor of the desired reflected-light signal and against all other signals. In a preferred embodiment of the invention, this discrimination is done partly by optical filters 31a and b as previously described, and further by using AC coupling tuned precisely to the chopping frequency. Further discrimination is also obtained via phase-sensitive amplification which is locked in phase with the desired chopped light beam.
The raw signals from the four photodetector elements 108, 110, 112 and 114 are processed and demodulated at the system electronics 28 to provide D.C. voltage signals A, B, C and D with respective magnitudes representative of only the amplitudes of the desired chopped-beam light signals, isolated from all other effects which influence the individual photodiode elements 108, 110, 112 and 114 respectively. It is not critical that all of the above modes of discrimination be used, but they combine to provide optimum sensitivity, accuracy, reliability, and stability of the diameter sensing function.
The A, B, C and D output voltages of the discriminator network 28 are fed directly into the differential amplifier network 29 which may include, for example, a plurality of integrated circuit differential operational amplifiers having differential outputs A-B and C-D as shown. The latter differential outputs are connected respectively to the control loops 30 and 32 as indicated. The C and D discriminated signals at the outputs of network 28 are also fed to a summing amplifier 33 for providing a C+D normalizing signal in the control loop 34.
The A-B error difference signal on line 30 is processed in a first or melt tracking feedback control or servo loop which comprises a melt tracking comparator 36 and a D.C. servomotor 40 connected as shown between one output of differential amplifiers 29 and the mechanical support means (see FIG. 3) for the optics chamber 17. The A-B error signal on line 30 is compared in the comparator 36 with a set point reference voltage and as a result of such a comparison an output signal on line 38 is generated and is sued to drive and control the speed and direction of the D.C. servomotor 40. The servomotor 40 operates to continuously move the optics chamber 17 including the photodetector 21 vertically downward toward the melt surface at a rate substantially equal to the rate of recession of the melt 12 in the quartz crucible 14. In this manner, there is no net change in the length of the optical paths l8 and 18 in the electroptical system disclosed, and the recession of the melt 12 during a crystal pulling operation does not therefore introduce errors into the other crystal diameter control signals to be described.
A second feedback control or servo loop means for processing the C-D crystal diameter differential control signal on line 32 includes a first pull rate control or servo loop portion 44 and a melt temperature control loop portion 45 interconnected as shown between the other outputs of amplifiers 29 and the crystal puller apparatus 9. The C-D error difference signal on line 44 is applied to the input of a pull-rate comparator 46 where it is compared to a set point reference voltage to thereby generate a comparator output signal on line 48 which provides control servo of the speed of the crystal pullers vertical pull-rate motor 50. The motor 50 is mechanically linked in a conventional manner (not shown) to the vertical pull-shaft S4 suspending the semiconductor crystal l0, and therefore, such mechanical linkage 52 is not shown or described in detail.
When the diameter of the pulled crystal 10 is nulled at a selected set point diameter, the C-D error difference signal will be equal to the set-point voltage applied to the pull-rate comparator 46, and the output signal of the comparator 46 will be at a predetermined DC level that will not vary the speed of the pull-rate motor 50 from a fixed or set point speed. However, if
the diameter of the semiconductor crystal 10 should increase above a desired set-point diameter, then the lateral movement of the meniscus 19 of the crystal l radially outward from the longitudinal axis of the crystal will cause the reflected image pattern 116 of the light impinging the photosensitive surfaces of the photodetector 21 to shift in the Y direction as indicated in FIG. 28. Such image shifting produces an increase in the OD error difference signal applied to the input of the comparator 46, and this signal voltage increase in turn produces a corresponding increase in the comparator output signal on line 48 which is applied as an input DC control signal to the crystal vertical pull-rate motor 50. The latter action produces a slight increase in the vertical pull-rate of the crystal and thereby tends to null the diameter of the crystal 10 to or toward the desired set-point diameter. The control loop portion 44 of the system for controlling the vertical pullrate of the crystal 10 provides a relatively fast-acting tactical control of the crystal diameter. Simultaneously, the melt temperature feedback loop portion 45 which has a longer time constant than the pull-rate loop portion 44 provides a slower strategic closedloop control action tending to restore the crystal pullrate to its optimal value. I
As indicated above, the crystal diameter-dependent error difference signal, C-D, is fedback via a second feedback control or servo loop portion 45 to a semiconductor melt temperature comparator 56 wherein it is compared to a set point voltage corresponding to a desired semiconductor melt temperature. As a result of this comparison, a difference or comparator output voltage is generated on line 58 and is applied as an input DC control voltage to the R-F power supply 60. This power supply 60 drives the R-F heating coils 64 for the graphite susceptor 16 in a wellknown manner. The above described C-D loop portions 44 and 45 constitute the second process control loop means which are directly responsible for the crystal diameter control. On the other hand, the first or melt tracking closed-loop means 30 for feeding back the A-B signal and the third or normalizing closedloop means 34 for feeding back the C+D normalizing signal provide an indirect control of the diameter of the crystal 10 by insuring that the C-D signal, under all operating conditions, truly and reliably represents the average crystal diameter and that the overall gain of the diameter sensing system (i.e., change in C-D signal vs. actual dimensional change in crystal diameter) remains constant.
The third or normalizing feedback control or servo loop 34 is connected between the C+D output of summing amplifier 28 and the tungsten light bulb 80 with the optics chamber 17, and this feedback loop includes a normalizing signal comparator 66 and bulb power supply 70 connected in the closed-loop 34458-72 as shown. The C+D normalizing signal is proportional to the level of light impinging the C and D photosensitive surfaces of the photodetector 21, and this C+D signal is compared to a set-point reference voltage in the normalizing signal comparator 66 to thereby generate yet another difference signal voltage on line 63. This signal is applied to the input of the bulb power supply 70 and continuously varies the light output of the tungsten light bulb to cause the total light received at the photosensitive surfaces of the photodetector 21 to remain constant at a preset level throughout s crystal pulling operation. This third or normalizing feedback loop 34-68-72 for the electroptical system according to the invention insures that any environmentally induced changes in the reflecting surfaces within the crystal puller chamber or any changes in the optical transmission of any components such as windows or lenses in the optical path of the system do not have any significant effect on the effective set points or gains of the other control loops.
The embodiments of the invention described above may be modified by those skilled in the art without departing from the true scope of this invention. For example, in performing the melt tracking function the crucible 14 can be moved upward rather than the optics chamber 17 downward in order to maintain a fixed distance between the surface of the semiconductor melt 12 and the synthesis optics. Alternatively, this distance can be allowed to change if compensating mechanical adjustments are continually made in the binocular angles a and B and the focal distances from slit 84 to projection lens 94 and from the photodetector 21 to the receiving lens in order to maintain a proper reflecting and focusing geometry.
Another area of modification within the scope of this invention is the performance of the normalizing function. Rather than using the C+D sum signal as a closedloop feedback signal to maintain a constant level of light impinging the photodector 21, the signals may be processed electronically, e.g. via analog divider modules, to produce normalized values (CD)/(C+D) and (AB)/(C+D) which may then be used in place of the C-D and A-B signals described previously.
1. A system for controlling the diameter of a pulled crystal, comprising detection means for receiving a light image reflected from a preselected area on the meniscus of a melt adjacent a crystal which is being pulled from the melt, said detection means including first sensing means responsive to positional variations of said reflected light image along a first axis for providing a first control signal which varies as a function of the level of said melt and second sensing means responsive to positional variations of said reflected light image along a second axis for simultaneously providing a second control signal substantiallyindependent of said first control signal, said second signal varying as a function of variations in the diameter of said pulled crystal, means responsive to said first control signal for maintaining a predetermined distance between said first and second sensing means and said melt and means responsive to said second control signal for controlling the diameter of said crystal being pulled. I
2. The system defined in claim 1 wherein a. said first sensing means includes first and second light-sensitive areas separated from each other and differentially responsive to the positional variations of said image in either direction along said first axis for generating said first control signal, and b. said second sensing means includes third and fourth light-sensitive areas separated from each other and selectively located with respect to said first and second light-sensitive areas so as to be substantially nonresponsive to positional variations of said image along said first axis, said third and fourth light-sensitive areas being differentially responsive to said positional variations of said reflected image along said second axis for generating said second control signal, whereby both the distance between said sensing means and said melt and said crystal diameter are simultaneously and independently controlled during a crystal pulling operation. 3. The system defined in claim 2 wherein: a. said first and second light-sensitive areas are separated from one another along said first axis by a predetermined distance, and b. said third and fourth light-sensitive areas are located between said first and second light-sensitive areas and are separated from one another along said second axis by a predetermined distance, said axes intersecting at substantially a right angle, whereby limited movement of a reflected light image on said first axis across said first and second light-sensitive areas does not change the total light of said image impinging on said third and fourth light-sensitive areas, and limited movement of said light image on said second axis does not change the total light of said image impinging on said first and second light-sensitive areas. 4. The system defined in claim 3 which further includes means for focusing a single light beam on said preselected area of said meniscus, said light beam being reflected from said preselected area to provide said image said light beam bearing focused on said meniscus at a preselected angle with respect to the longitudinal axis of said crystal whereby the reflected light image received by said detection means will cause differential response by said first and second light-sensitive areas and said third and fourth light-sensitive areas, respectively.
5. The system defined in claim 2 wherein said first, second, third and fourth light sensitive areas are each adapted to provide an electrical output and wherein said means responsive to said first control signal comprises a first control loop interconnected with said first and second light-sensitive areas and including a servomotor operable for maintain the predetermined distance between the sensing means and the surface of said melt.
6. The system defined in claim 5 wherein said means responsive to said second control signal comprises a second control loop interconnected with said third and fourth lightsensitive areas and operative for controlling the pull rate of the crystal being pulled so as to maintain a selected crystal diameter.
7. The system defined in claim 6 which further includes a light source whose light output is controllable, said light source providing a light beam which is reflected from said preselected area on the meniscus, and a third control loop interconnected with a plurality of said light-sensitive areas and responsive to said changes in the total light impinging on said plurality of lightsensitive areas for controlling the light output from said light source thereby to maintain a substantially constant total light impinging on said plurality of lightsensitive areas.
8. The system defined in claim 7 wherein said third control loop is a normalizing control loop including comprises a comparator for receiving a signal which is the sum of the electrical outputs of said third and fourth light-sensitive areas and for comparing the last said signal with a reference voltage for controlling the power applied to said light source as a function of vari ations in the total light from said light source impinging on said third and fourth light-sensitive areas.
9. A method for controlling the diameter of a crystal pulled from a liquid melt which includes the steps of:
a. focusing a beam of light on the meniscus formed by a crystal being pulled from said melt,
b. receiving a light image reflected from said meniscusand which moves along first and second axes in response to changes in melt level and lateral meniscus position, respectively,
0. generating a first control signal in response to positional variations of said reflected image along said first axis for tracking the level variations of said melt, and
d. generating a second control signal in response to positional variations of said reflected image along said second axis for controlling the diameter of said crystal, whereby said tracking of said melt prevents errors from being introduced into said second control signal by melt level variations during a crystal pulling operation.
10. The method defined in claim 9 which further includes generating a third control signal as a function of a chosen threshold level of light received at a given lo cation relative to said crystal meniscus for varying the light intensity of said beam and thereby maintaining the light intensity at said given location constant.
11. A process for generating an optical signal which is closely proportional to changes in the diameter. of a crystal rod which is pulled from a melt, said process including, in combination:
a. projecting a sharply focused, uniformly illuminated image pattern onto substantially the entire curved portion of the liquid meniscus extending between the melt and the crystal rod at liquid-solid interface therebetween, and
b. sensing the movement of a relatively small spot within said sharply focused pattern and lying in a plane which intersects the horizontal surface of said melt at a predetermined angle, whereby said spot continuously appears at a substantially constant distance from the periphery of the said crystal rod and tracks said periphery and variations in crystal rod size as a result of rotational eccentricities and non-circular shapes in said rod.
12. The process defined in claim 11 wherein:
a. the projecting of the image pattern includes projecting the image pattern such that it has a longer dimension aligned radially outward from the crystal axis and spanning substantially the entire curved portion of said meniscus, and
b. said sensing of the movement of the said spot is effected by photosensing by positioning multiple photosensitive surfaces in the path of the anticipated movement of light reflected from said spot thereby to generate differential signals as a result of the shifting of said spot within said pattern.
13. The system defined in claim wherein said first control loop comprises a differential amplifier providing an output which is the differential of the electrical outputs of said first and second light-sensitive areas, and a comparator for comparing the output signal from said differential amplifier with a reference voltage, said servomotor being controlled by an output from said comparator.
14. The system defined in claim 6 wherein said second control loop comprises a differential amplifier providing an output which is the differential of the electrical outputs of said third and fourth light-sensitive areas, and a comparator for comparing the output signal from said differential amplifier with a reference voltage, the output from said comparator controlling the speed of a motor for pulling said crystal.
15. The system defined in claim 14 further comprising a temperature control loop including a further comparator also for comparing the output from said differential amplifier with a further reference voltage, the output from said further comparator controlling means for heating said melt thereby to control the melt temperature.
16. The system defined in claim 3 wherein said first and second light-sensitive areas are each generally rectangular in shape and of substantially the same area, and said third and fourth light-sensitive areas together define a total area of generally rectangular shape located between the first and second light-sensitive areas, said third and fourth light-sensitive areas being adjacent and separate along said second axis with the adjacent edges thereof defining a serpenture path extending generally along said first axis whereby respective portions of said third and fourth areas are interdigitated and overlie said first axis.
17. The system defined in claim 16 wherein said sensing means is adapted to receive said image with the center of the image generally located at the intersection of said axes with portions of the image overlying respective portions of said first, second, third and fourth light-sensitive areas.
18. A system for controlling the diameter of a pulled crystal comprising:
a. means for directing a beam of light on a preselected area of the meniscus of a melt formed by a crystal being pulled from the melt;
b. first light sensing means for receiving the image caused by reflection of the light beam from the meniscus, said first sensing means being responsive to positional variations of the image along a first axis caused by change in the melt level;
0. second light sensing means for receiving said image and responsive to positional variations of the image along a second axis caused by variations in the diameter the crystal being pulled, the positional variations being substantially independent of positional variations along the first axis;
d. a first servo loop interconnected with said first sensing means and including a sevomotor operable for maintaining a predetermined distance between said first and second sensing means and said preselected area of the meniscus; and
e. a second servo loop interconnected with said second light sensing means and operative to provide servo control of the speed of a motor for pulling the crystal whereby the diameter of the crystal is precisely controlled regardless of changes in the melt level.
19. The system defined in claim 18 further comprising a light source of controllable light output for providing said light beam, and a third servo loop interconnected with at least one of said sensing means and operative to provide servo control of said light output as a function of the amount of light received by the sensing means, whereby compensation is provided for variations in the amount of light received by the sensing means occurring during crystal pulling, and
20. The system defined in claim 18 further comprising a chopper for chopping the light beam and demodulator means interconnected with each of the light sensing means for deriving d.c. signals therefrom.