WO1993025893A1

WO1993025893A1 - Endpoint detection technique using signal slope determinations

Info

Publication number: WO1993025893A1
Application number: PCT/US1993/004936
Authority: WO
Inventors: Herbert E. Litvak
Original assignee: Luxtron Corporation
Priority date: 1992-06-09
Filing date: 1993-05-24
Publication date: 1993-12-23
Also published as: JP3375338B2; EP0645008A1; JPH07507904A; KR950702029A; TW223176B

Abstract

A predetermined stage of a changing condition is detected by monitoring a change in a signal relating to the changing condition. For example, the existence of a breakthrough that results from photoresist development, material etching, and the like, is detected by optically monitoring a semiconductor wafer, printed circuit board, and the like, that is being processed. An optimum end of the processing can then be determined in response to a breakthrough being detected, either for the purpose of monitoring the process or in order to automatically terminate the processing at that determined time. As part of the processing that determines the existence of a breakthrough, the optical signal is digitized and individual slope values calculated from groups of consecutive digital signal values.

Description

ENDPOINT DETECTION TECHNIQUE USING SIGNAL SLOPE DETERMINATIONS

Background of the Invention

This invention relates generally to techniques of signal processing that result in determining when a particular stage of a changing condition has been reached, either for the purpose of observation or for automatically controlling the events causing the condition to change, an example being the monitoring of an electrical signal in order to determine when an endpoint has been reached in a semiconductor processing operation such as photoresist development or etching operations.

There are many situations where an electrical signal exists, or can be obtained, that varies in a manner related to a changing condition desired to be monitored. One class of such situations is during the processing of materials where its desired to determine when a particular processing operation has been completed; ie, to detect an endpoint of the processing operation. The endpoint determination is then used to monitor the progress of the process and/or to control the process, such as by terminating the specific processing operation being monitored. The semiconductor industry is one example of where such process monitoring occurs. At several stages in the manufacture of circuits on a semiconductor wafer, a mask is formed of photoresist material. The resulting mask is used to limit processing of a layer covered by the mask to a patterned area. The mask is formed by exposing the photoresist layer to light in the desired pattern, followed by developing the photoresist layer through application of a developer solution to it. With the usual photoresist material, the exposed regions are removed during the development process to expose the layer below. The time at which the underlying layer first becomes exposed by removal of photoresist material is termed the "breakthrough" or "endpoint." The development process is allowed to continue for a period of time after breakthrough is first detected, the end of that period of time being the end of the development process, termed its "process end."

Because of various processing and environmental variations that exist from batch to batch of semiconductor wafers, the development process is monitored in order to determine when breakthrough occurs.- A beam of light of finite bandwidth is directed against the photoresist layer of one wafer of a batch and a reflected or transmitted light signal is then detected and the resulting electrical signal processed to determine when breakthrough occurs. In one form, light reflected from both the top and bottom surfaces of the substantially transparent photoresist layer interferes at a photodetector. As a portion of the photoresist layer is removed during the development process, the detected intensity of the reflected light cycles between a maximum and minimum as the material removal alters the relative phase between the two interfering beams. At breakthrough, however, this variation in signal ends, a condition which is detected by analyzing the photodetector output signal. Development is then usually allowed to proceed for a fixed time after detection of breakthrough, at which point the development is terminated by rinsing away the development solution or by some other means. Wet etching processes, wherein substantially transparent material layers other than photoresist material are etched away, also use a similar breakthrough detection process. In the case of a dry etching process, wherein material is removed by bombardment in a plasma chamber, a species within the plasma is monitored by detecting the emission of light in a limited wavelength band that shows material is being removed from the layer being etched. The detected signal in such applications usually drops off considerably at breakthrough and this is electronically detected.

It is an object of the present invention to provide an improved technique for processing such a detected electrical signal being monitored during semiconductor wafer processing.

It is a more general object of the present invention to provide an improved signal processing technique for use in a wide variety of applications wherein an electrical signal exists or may be obtained that is related to a changing condition being monitored.

Summary of the Invention

These and additional objects are accomplished by the present invention, wherein, briefly and generally, individual values of change of an electrical signal are calculated as the monitored condition is changing in order to determine when the condition has reached a particular stage. In one form, the signal change being monitored is the slope of the electrical signal. According to a specific aspect of the invention, a digitized electrical signal is acquired and a group of contiguous digital signal values are utilized to calculate one signal slope value. Successive slope values are calculated from successive groups of contiguous digital samples, each group including at least some of the samples of the immediately preceding group but omitting at least the earliest acquired sample of the immediately preceding group. The desired particular stage of a changing condition is determined to exist when the calculated slope values satisfy a preset criteria. No absolute values of the electrical signal are utilized in determining when the change in condition has reached its particular stage; rather only slope or other signal change characteristics are used. The preferred embodiments of the invention described below with respect to the drawings show an application of the improved signal processing technique of the present invention to the monitoring and, optionally, the control of the semiconductor wafer processing. However, it will be recognized that there are numerous other specific applications of the present invention. Other applications include monitoring electro-chemical signals that are proportional to a concentration of a species in a chemical composition, mass spectrometry, electronic spin resonance (ESR) , nuclear magnetic resonance (NMR) , temperature, pressure, acoustical signals, change of index of refraction in response to changing chemical or physical parameters, flowometry, colorimetry, strain gauge signals, light dispersion, crystal frequency changes, fluoro-immuno- assay techniques, and the monitoring of the ph of a solution, to name only a few of the many possible applications.

Brief Description of the Drawings Figure 1 schematically illustrates a semiconductor wafer photoresist layer development process in which the present invention is utilized;

Figure 2 illustrates a modified photoresist development operation in which the present invention is utilized; Figure 3 is an electronic block diagram of an endpoint controller used in connection with either of the photoresist development systems of Figures 1 or 2;

Figure 4 schematically illustrates the optical monitoring of a photoresist layer that results in the electrical signal being processed;

Figure 5A shows an example photodetector output signal obtained during photoresist development;

Figure 5B is a calculated derivative of the photodetector output of Figure 5A;

Figure 6A and 6B illustrate, in enlarged cross-sectional views, two stages of a photoresist development process;

Figure 7A and 7B illustrate, in enlarged cross-sec*:ional views, two stages of an etching process;

Figure 8A shows examples of digital samples of a detector output signal being monitored, and Figure 8B shows the slope values calculated from the samples of Figure 81 Figure 9 illustrates very generally an apparatus for performing a dry etching process through plasma discharge;

Figures 10A and 10B show examples, respectively, of a declining signal level monitored in the plasma system of Figure 9, and its calculated slope;

Figures 11A and 11B show examples, respectively, of a increasing signal level monitored in the plasma system of Figure 9, and its calculated slope;

Figure 12 is a flow diagram that illustrates one signal processing method to obtain endpoint;

Figure 13 is a flow diagram expanding upon one step of the method of the flow diagram of Figure 12, according to one embodiment; and

Figure 14 is a flow diagram expanding upon the same step of the method of the flow diagram of Figure 12, according to an alternative embodiment. Description of the Preferred Embodiments

Referring initially to Figure 1, a commonly used system is described for photoresist development as a step in manufacturing electronic circuits on a semiconductor wafer. A liquid enclosure 11 contains in it a support 13 that is rotated through a shaft 15 by a motor 17 that is outside of the enclosure 11. Positioned on the rotating support 13 is a semiconductor wafer 19 that is being processed. At this point in the processing, the photoresist layer has already been applied to the wafer structure and exposed to a light pattern corresponding to the physical pattern desired to be left after development of the photoresist layer. It is this physical mask pattern that is subsequently used in a further step to etch or otherwise process the layer immediately beneath the photoresist layer.

The developer solution is generally applied to the photoresist layer by spaying, although other techniques can be used as well. A nozzle 21 provides such a spray from a solution in a container 23 that is passed through an electrically controlled valve 25. Various specific techniques of intermittent spray or continuous spray are used by different semiconductor manufacturers. Once the development is complete, the valve 25 is turned off and then another electrically controlled valve is opened to deliver a rinsing solution from a container 29 to another nozzle 31. The developer and rinse spraying usually occur while the wafer 19 is being spun with a uniform velocity. An electronic process controller 33 controls the spinning motor 17, the valves 25 and 27 and other aspects of the processing equipment.

In order to monitor the development process, a source 35 of electro-magnetic radiation in the visible or near visible region directs a beam 37 against the developing photoresist layer on the Wafer 19. A photodetector 39 is positioned to receive a reflection of the beam 37 from the wafer 19. An electronic control system 41 drives the radiation source 35 and processes an electrical signal obtained from the photodetector 39 in a circuit 40. Since the detection of breakthrough is accomplished as a result of light interference, as explained hereinafter, some degree of coherency of the light source 35 is generally employed, a light emitting diode (LED) being satisfactory. An output of the electronic system 41, in a circuit 43, indicates when breakthrough of the photoresist layer has been detected to occur. This signal is sometimes utilized to simply tell the operator w! n breakthrough occurred as a way of monitoring the process variations. Alternatively, the breakthrough signal in the circuit 43 can be applied directly to the process controller 33 to terminate the development process at some specific time after breakthrough has been detected.

Rather than spraying the developer solution onto the photoresist layer, other developing equipment utilizes a tank 45 in which a boat 47 carrying a plurality of wafers, such as wafer 49, is submersed in a developer solution 51. A unit 53 contains the light source and photodetector. A forward positioned wafer 49 is monitored during the development process. When the process end is determined or believed to have occurred, the boat 47 and its wafers is removed from the tank 45 and placed in a rinsing tank to terminate the development process. An endpoint controller substantially the same as the controller 41 is also utilized with the unit 53

A generalized block diagram of the controller 41 is shown in Figure 3. Connected to a system bus 55 are various components found in any general or special purpose computer. A microprocessor 57 is one of these components, as is a system dynamic random access memory δ

(RAM) 59 and a read only memory (ROM) 61. One input- output circuit 63 interfaces over a circuit 65 with a control and display panel, such as a cathode ray tube display and some form of keyboard input. A second input-output circuit 67 receives the photodetector output analog signal in the circuit 40 and digitize it for processing by the controller system of Figure 3. A third input-output circuit 69 provides a control signal output in circuit 43 to indicate that the endpoint has been reached, or to provide a calculated process end time. A controlling software program for operating the controller of Figure 3, including determining when breakthrough occurs, is stored in its ROM 61 and executed by the microprocessor 57. Such an endpoint controller and light source photodetector assembly is commercially available from the Xinix division of Luxtron Corporation, Santa Clara, California. Its Models 2200, 2300 and 2400 Wet Process Endpoint Controllers are widely used for this purpose. Breakthrough of the photoresist layer is currently determined in these instruments by a computer program including a window triggering algorithm. The present invention is directed primarily to providing an alternate, improved technique of determining breakthrough. The improved technique of the present invention may be implemented by loading software of the new technique into a Xinix Controller memory in place of the current window triggering algorithm.

Before describing this improved breakthrough detection signal processing technique, the photoresist development process is described further. Figure 4 shows an exemplary cross-sectional view of a semi¬ conductor structure 71 having a photoresist layer 73 coated on its top surface. A ray 37' of the illuminating beam 37 (Figure 1) is partially reflected by a top surface of the layer 73, as indicated by a ray 75'. Since the material of the layer 73 is substantially transparent, a portion of the incident ray 37' proceeds -through the layer 73 and is reflected from an interface 77 between that layer and the underlying structure, as indicated by a ray 79. A photodetector positioned in the path of the rays 75 and 79 will detect an intensity level that is the result of interference between these two rays, provided that the incident light beam 37 is sufficiently monochromatic and coherent. That intensity is dependent upon the difference in pathlength travelled by the rays 75 and 79. As the layer 73 decreases in thickness, as occurs during a photoresist development process, the pathlength of the ray 75 changes, with the result that the intensity of the interfering beams as seen by the photodetector changes.

An example of the output of the photodetector 35 (Figure 1) during such a process is given in Figure 5A. The signal oscillations occur over time as the exposed regions of the photoresist layer 73 are developed and removed. Once there is a breakthrough through the layer 73, however, there is no further change in light pathlength imparted by further development, and the oscillations cea.se. Thus, at a point 89, where the oscillations of the photodetector output stop, it is inferred that a breakthrough has occurred. The techniques of the present invention are applied to determining when such a breakthrough occurs. Since this specific process takes place in an environment where the electrical signal is quite noisy, however, the photodetector output signal is unfortunately not as pure as that illustrated in Figure 5A for simplicity in explanation. The signal processing techniques of the present invention take into account the possibility of having a noisy signal to analyze. Although the breakthrough detecting techniques described herein are useful when an entire layer of material is removed, it is usually applied to situations where a patterned layer of material is removed. Figure 6A is a cross-sectional view similar to that of Figure 4 except to show when breakthrough first occurs in exposed regions of the photoresist layer 73 during the development process. The reflected signal of Figure 5A is at point 89 when the breakthrough illustrated in Figure 6A first occurs. Development is allowed to proceed for a further period, usually a fixed time after breakthrough is detected, and until the exposed photoresist regions are fully removed, as illustrated in Figure 6B. A similar process also can occur during some wet etching operations. Referring to Figure 7A, a substrate structure 91 has over it a continuous layer of substantially transparent material 93, such as silicon oxide and the like, covered with a mask 95. As etching is permitted to occur through openings in the mask 95, the breakthrough shown in Figure 7A is detectable by monitoring a signal similar to that of Figure 5A to determine when the oscillations in photodetector output signal end. As with photoresist development, the etching is then allowed to proceed for a time after breakthrough is detected, until the regions under the vast openings are fully removed, as shown in Figure 7B.

The improved technique of the present invention, which allows the detection of breakthrough from the photodetector output signal, is illustrated generally by the curve of Figure 5B. The photodetector signal output is processed by the controller system of Figure 3 to obtain data of the curve of Figure 5B which is the mathematical derivative of the photodetector output signal of Figure 5A. Breakthrough is determined by use of only the derivative signal of Figure 5B, no direct use of the absolute signal values of Figure 5A being necessary. In a specific implementation, a positive threshold level 97 is set. Once it is determined that the development process has begun, the derivative signal of Figure 5B is monitored to determine when it falls below the threshold 97 for a continuous period of time that is set to be longer than such signal levels may occur from noise in the signal or during the oscillations that occur. Such a fixed duration is shown to end at time 99 in Figure 5, at which point it is known that breakthrough occurred at time 89.

As mentioned, before the signal processing technique being described can look for the breakthrough point, it must determine that the process has actually begun. One way to do this is to determine that the oscillations have begun, such as by determining that a peak or a valley in the derivative signal of Figure 5B has occurred. Another way of determining that the process has begun is to simply note that the derivative signal of Figure 5B has exceeded the threshold 97 for a set period of time, thereby to avoid any false triggering due to noise and the like.

Rather than using a single threshold 97, both it and a corresponding negative threshold 101 can also be utilized. In this case, breakthrough is determined when the derivative signal of Figure 5B lies between the thresholds 97 and 101 for a specific period of time. Similarly, a start of the development process can be confirmed by noting when the derivative signal of Figure 5B exceeds either of the thresholds 97 or 101 for a predetermined period of time.

Figures 8A and 8B illustrate a preferred technique for calculating the derivative signal of Figure 5B from the photodetector analog output signal of Figure 5A. As previously mentioned, the input-output circuit 67 (Figure 2) digitizes the incoming photodetector analog signal in a circuit 40. In order to explain the process being carried out, a few such digital signal samples 103-109 are shown in Figure 8A. From these values are calculated individual values 111- 115 of a changing characteristic of the signal, namely its slope. The points 111-115 of Figure 8B are points of the derivative signal previously discussed with respect to Figure 5B.

The slope value 111 of Figure 8B is determined from the slope of a model equation, in this case a line 117, that is fit to contiguous digital sample points 103, 104 and 105. Similarly, the next slope value 112 is calculated to be that of the same model equation, in this case another straight line 118, that is fit to the contiguous digital samples 104, 105 and 106. Likewise, the slope 113 is that of a line 119, the slope 114 that of a line 120 and the slope 115 that of a line 121.

It can be noted that the first slope value 111 is calculated from a first contiguous group of digital samples 103, 104 and 105. The next slope value 112 is calculated from a group of digital samples that omits the first sample 103 but adds to the group the newest acquired sample 106. Thus, each time a new digital sample of the photodetector analog signal is acquired by the controller system of Figure 3, a new slope value is calculated by fitting the curve to a combination of samples including the newly acquired one and those used during the immediately preceding slope calculation, except that the earliest digital sample is now deleted from the group. That is, each slope calculation is made from a predetermined number of contiguous digital samples that substitutes the most recently acquired sample for the oldest sample in the group.

Although the number of samples in each group has been shown in Figure 8A to be only three, for simplicity of explanation, the group will generally be much larger. The precise number of samples utilized for any particular application will depend primarily on the amount of noise immunity that is desired to be built into the calculation. Also, a straight line curve fit has been shown, also for simplicity in explanation, but it may be desirable in certain applications to use a more complex polynomial fit to the individual groups of digital signal samples. This technique is an application of the known Savitzky-Golay polynomial method. See, for example, A. Savitzky and M. J. E. Golay, Anal. Chem . , Vol. 36, pp. 1627-1639 (1964), or P. A. Gorry, Anal. Chem. , Vol. 62, pp. 570-573 (1990). An advantage of fiting a curve to actual data is that it allows the process control to look ahead by modeling the process.

The same signal processing technique finds application in monitoring a wide variety of material or composition processing operations. Another specific example from the semiconductor industry is schematically illustrated in Figure 9. Rather than etching away a layer of material by a chemical solution, as illustrated in Figure 7, etching is often done by a dry plasma process within a vacuum chamber 131. As is well known, a semiconductor wafer 133 is positioned within the chamber under a target structure 135. The target 135 is energized by a direct current or radio frequency power supply 137, depending upon the specific process being implemented. A supply 139 of inert and reactive gases are connected to selectively supply these gases to the interior of the chamber 131. As is well known, a plasma 141 is formed over the semiconductor wafer 133 from the breakup of gaseous molecules under the influence of strong electric fields. The result is that the wafer 133 is bombarded with ions that remove the top layer of material on the semiconductor wafer 133 or, if masked against such removal, only in the regions defined by the mask.

The same goal exists in the plasma process as with the wet etch process, namely the desire to accurately determine when breakthrough first occurs so that the etching process may be terminated at an appropriate point before it proceeds too far. In plasma processes, it is common to view the visible or near visible electromagnetic emissions of the plasma 141 through a quartz window 142. A photodetector 143 is shown in Figure 9 to be positioned immediately outside of the chamber 131 in the path of the plasma emissions that are viewable through the window 142. The emission intensity of the plasma 141 in one or more selective narrow wavelength bands or lines is detected and an electrical signal output of the photodetector 143 is proportional to that intensity. Detection of a limited emission wavelength band is generally accomplished by positioning a narrow bandpass interference or other type of filter in front of the photodetector 143 so that it receives only that band of wavelengths.

The wavelength band chosen to be monitored is one whose intensity varies in a manner related to the progress of the etching process being monitored. Figure 10A illustrates one such narrow band signal that can be monitored. Breakthrough of the layer being etched is determined to occur at about time t_B, where the rate of change of the signal flattens out considerably. Similarly, as illustrated in Figure 11A, a plasma emission bandwidth can be monitored in certain circumstances where the intensity increases as an etching proceeds, an endpoint indicated at time t_B where the increasing signal flattens out.

The preferred signal processing technique for determining when time t_B occurs iri either of the decreasing or increasing signals of Figures 10 and 11, respectively, is the same as that previously described with respect to Figure 8. That is, the analog photodetector output signal of Figure 10A is digitized and processed point by point to obtain a derivative signal shown in Figure 10B. A negative slope threshold level 145 is set at an appropriate level. The processor first determines when the signal of Figure 10B exceeds the threshold 145 for a predetermined time, thus indicating that the etching process has begun. After the beginning of the process has been detected, the signal of Figure 10B is then monitored to determine when it falls below the threshold 145 for a time, indicating that breakthrough has occurred. Separate threshold levels may alternatively be utilized for detecting when the process has begun and when breakthrough has occurred.

Similarly, a derivative signal shown in Figure 11B is obtained from the photodetector output signal of Figure 11A. A threshold 147 is used with the derivative signal. The process is determined to have begun when the derivative signal of Figure 11B exceeds the threshold 147 for a time. Breakthrough is subsequently determined when the derivative signal falls below the threshold level 147. Referring to Figure 12, a preferred digital signal processing technique is given for use in the endpoint controller of Figure 3 for determining breakthrough from any of the signals of Figures 5A, 10A or 10B. In each case, these analog signals are first digitized and their derivatives of respective Figures 5B, 10B and 11B determined by the process described with respect to Figure 8.

A first step 151 of the Figure 12 signal processing technique is to set various parameters and initialize one or more counters that are to be used. One such parameter is denoted as "N", the number of contiguous data points to be utilized at one time for determining the individual slope points. In the example described with respect to Figure 8, N=3. It is desirable to allow the end user to select N since its value is determined primarily by the amount of noise immunity that is necessary in a particular application. The N data points are normally grouped in overlapping intervals, as shown in Figure 8A, but may alternatively be grouped in stepwise intervals, wherein each slope value is determined from a different set of data points. This description is directed to the overlapping case but it will be understood that the stepwise case is handled similarly.

After such initialization, a next step 153 acquires and stores N-l consecutive data points of the type illustrated in Figure 8A. After another data point is acquired as indicated by a step 155, there are enough data points for which a single slope value is calculated by the technique described with respect to Figure 8, indicated by a step 157. Several such calculated slope values are shown in Figure 8B.

After each calculation, a next step 159 determines whether or not the semiconductor process has begun. Of course, the criteria for a breakthrough will not be applied until it is determined for sure that the process has actually been started. Otherwise, breakthrough can be determined too early and the manufacturing process terminated before it has even begun. If the process has not been started, then further data values are used to calculate additional slope points.

Once it is determined that the process has begun, however, a next step 161 compares the value of the most recently calculated slope value with a target end slope. If it is less than the target end slope, a slope end counter is incremented by one, as indicated in a step 163. This is one of the counters that was initialized in step 151 to have a beginning zero count. A next step 165 compares the new count of the slope end counter with a target end count. The target end count is another one of the parameters initially set in 151 and preferably made user selectable. This, in effect, sets the duration of time that the signal slope value must remain below its target value before breakthrough is determined to have occurred. This number is also set consistent with the amount of noise on the signal in a particular application. Thus, the target end slope acts as a not-to-exceed threshold value.

Once the slope end counter is determined to have been incremented to a value exceeding the preset target end count, breakthrough is signalled to have occurred, as indicated by a step 167. A process end time is then calculated by addition of an overprocessing period, and the process allowed to continue during such a further period, as indicated by a step 168. If the target count has not been reached by the most recent slope value that is below the target end slope, as determined by the step 165, then the processing returns to the step 155 by acquiring a new data point. If, as indicated by the step 161, the most recent slope value exceeds the target end slope, the slope end counter is reset to zero, as indicated by a step 169.

The particular procedure used as part of the step 159 to determine whether the semiconductor process has begun or not can vary, and will particularly be chosen to match the type of signal that is analyzed. Figures 13 and 14 show alternate ways of determining whether the process has started or not.

The technique of Figure 13 looks to whether the slope value exceeds a predetermined threshold and is applicable to analyzing any of the analog signals of Figures 5A, 10A and 11A. A slope start counter is used in addition to the slope end counter of Figure 12. As indicated by a step 171, the count in the slope start counter is compared with a target start count, also user selectable. If the slope start counter has previously been incremented to exceed the target start count, then processing continues to the step 161 of Figure 12. If not, another step 173 (Figure 13) occurs.

In the step 173, the most recent slope value calculated in step 157 of Figure 12 is compared with a threshold target start slope. If that slope value exceeds the threshold, the slope start counter is incremented by one, as indicated in a step 175. The processing then returns (Figure 12) to step 155, followed by step 157 and then through step 159 on to the step 161. However, if the most recent slope value is less than the target start slope, the slope start counter is reset to zero, as indicated by a step 177, and then the processing of Figure 12 again returns to the steps 155 and 157. In the case of an oscillating signal of the type shown in Figure 5A, an alternative signal processing technique may be advantageously employed. Referring to Figure 14, a step 179 occurs after a new slope value has been calculated in a step 157 of Figure 12. In this step 179, a determination is made as to whether a signal peak has occurred or not. If so, it is known that the semiconductor process has started and the signal processing goes to the next step 161 of Figure 12. But if it is determined that a peak has not been found, then a next step 181 of Figure 14 looks for a signal valley. If such a valley is found, then the next step is also 161 (Figure 12) but, if not, then the signal processing returns to steps 155 and 157 of Figure 12. In each of the steps 179 and 181 of Figure 14, the latest calculated slope value is used with a set number of immediately preceding slope values in order to determine whether a peak or a valley in the derivative signal has occurred. In step 179, a number P slope values are used. It is determined whether a certain number of consecutive slope values of that group initially were increasing, followed by consecutive number of substantially zero slope values, followed by yet another preset number decreasing in value. Similarly, the step 181 utilizes the most recently calculated slope value with a number V slope values to determine whether a valley exists. The valley exists if a first number of slope points is decreasing, followed by another number substantially zero, followed by another number of increasing slope samples. Although the signal processing technique has been explained with respect to specific semiconductor processing examples, it will be appreciated that this technique has wide application wherever analog signals of the type shown in Figures 5A, 10A, 11A, or similar signals, exist. Processing of articles other than semiconductor substrates can certainly be monitored in the same way, whether by use of an interrogating electromagnetic energy beam or by observing some detectable electromagnetic energy that is emitted as part of the material processing operation. The monitoring of chemical reactions also may be accomplished by use of the same technique. In addition, there are numerous applications where the electrical analog signal already exists as a result of, or as a by- product of some operation, such as a signal from a spectrophotometer. Thus, it is understood that the invention is not limited to the monitoring of any specific process or operation but rather is being protected within the full scope of the appended claims.

Claims

IT IS CLAIMED:

1. A method of detecting a particular stage of a changing condition, comprising the steps of: monitoring the changing condition, acquiring a succession of periodic digital samples representing an electrical signal that varies in a manner related to said changing condition, calculating individual values of signal change from successive groups of said digital samples, and determining when a predetermined number of consecutive ones of said values of signal change fall within a preset limit related to said particular stage, thereby to determine that said changing condition has reached said particular stage.

2. The method according to claim 1 wherein the calculating step includes the steps of: fitting a curve to a first group of a predetermined number of digital samples that are contiguous in time, calculating a single value of signal change from said curve, and repeating in time sequence the foregoing fitting and signal change value calculating steps on a plurality of different groups of said predetermined number of digital samples, the samples of individual groups including at least some of the samples but omitting at least an earliest acquired sample of the immediately preceding group.

3. The method according to claim 1 which comprises the additional step of terminating further change in said condition in response to determining that said condition has reached said particular stage.

4. The method according to claim 1 which comprises the additional step of processing material in a manner that a change in the material as a result of its processing is the changing condition being monitored.

5. The method according to claim 4 wherein the signal monitoring and generating step includes the step of positioning a detector in a path of an electromagnetic radiation signal having a variation related to the changing characteristic of the material being processed.

6. The method according to claim 5 wherein the signal monitoring and generating step includes the step of directing a beam of electromagnetic radiation onto th material and thence to the detector, whereby the detector receives the beam after modification by the changing condition of said material.

7. The method according to claim 5 wherein the detector positioning step includes the step of positioning the detector in the path of electromagnetic radiation being emitted as part of the material processing.

8. The method according to claim 1 which comprises the additional step of altering a thickness of at least a portion of a layer of material across an article surface, and wherein the changing condition being monitored by the monitoring and signal generating step is that of said material layer thickness.

9. The method according to claim 8 wherein the thickness altering step includes increasing the thickness of said material layer.

10. The method according to claim 8 wherein the thickness altering step includes decreasing the thickness of said material layer.

11. The method according to claim 10 wherein the step of decreasing the thickness of said material layer includes decreasing the thickness of said material layer across only a patterned portion thereof.

12. The method according to claim 8 wherein the article upon which the thickness of a material layer is being monitored includes a semiconductor substrate.

13. The method according to claim 8 wherein the signal monitoring and generating step includes the step of directing a beam of electromagnetic radiation onto the material layer and thence to the detector, whereby the detector receives the beam after modification by the changing thickness condition of said layer.

14. The method according to claim 8 wherein the detector positioning step includes the step of positioning the detector in the path of electromagnetic radiation being emitted as part of the material layer thickness changing process.

15. The method according to claim 8 wherein the layer whose thickness is being altered includes a layer of photoresist material.

16. The method according to claim 8 which comprises an additional step of terminating the layer thickness altering in response to determining that the changing thickness condition has reached said particular stage.

17. The method according to claim 1 wherein the particular stage of a changing condition is detected without comparison of the magnitude of the digital signal samples with an absolute reference signal level.

18. A method of detecting a particular stage of an article condition that is changing in response to processing of the article, comprising the steps of: directing onto a detector an electromagnetic radiation signal exhibiting a rate of change that is related to said changing article condition, generating from said detector an electrical signal in a form of a succession of periodic digital values that also exhibits a rate of change that is related to said changing article condition, during the article processing, calculating individual values of signal change from successive groups of said digital values, by the following steps: fitting a curve to a first group of a predetermined number of digital values that are contiguous in time, calculating a single value of signal change from said curve, and repeating in time sequence the foregoing fitting and signal change value calculating steps on a plurality of different groups of said predetermined number of digital values, the values of individual groups including at least some of the values but omitting at least an earliest acquired value of the immediately preceding group, determining from said signal change values that the article processing has begun, and after determining that the article processing has begun, determining when a first predetermined number of consecutive values of signal change fall within a first preset limit related to said article condition stage, thereby to detect that said article condition has reached said stage.

19. The method according to claim 18 wherein the step of determining that the article processing has begun includes determining when a second predetermined number of consecutive values of signal change fall outside of a second preset limit related to said article processing.

20. The method according to claim 19 wherein the first preset limit is a signal change value that is less than a predetermined level and the second preset limit is a signal change value that is greater than said predetermined level.

21. The method according to claim 18 which comprises an additional step of terminating the processing of said article in response to detecting that the condition of the article has reached said stage.

22. The method according to claim 18 wherein the particular stage of the article condition is detected without comparison of the magnitude of the digital signal values with an absolute reference signal level.

23. A method of detecting a breakthrough in a process of removing material from a substantially transparent layer of a composite structure, comprising the steps of: directing electromagnetic radiation against said layer during the material removal process, thereby to reflect said radiation from both an exposed surface of said layer and a surface beneath said layer, receiving the reflected radiation at a detector, thereby to generate an electrical signal proportional to an intensity of an interference between radiation reflected from the exposed surface of the layer and the surface beneath said layer, said signal changing while material is being removed from said layer, calculating from said electrical signal during the removal process a function representing a time varying slope of said signal, determining from said slope function that the material removal process has started, and after it is determined that the material removal process has started, detecting when the slope function falls in magnitude below a first preselected fixed value for a first preselected period of time, whereby the oscillations of said signal are detected to have ended and the breakthrough is confirmed to have occurred.

24. The method according to claim 23 wherein the step of determining that the process has been started includes detecting the existence of at least a portion of an oscillation of said electrical signal.

25. The method according to claim 23 wherein the step of determining that the process has been started includes detecting when said slope function exceeds a second preselected fixed value for a second preselected period of time.

26. The method according to claim 25 wherein the first and second preselected fixed values of slope are made to be a single value.

27. The method according to claim 23 wherein the slope function calculating step includes the following steps:

(a) acquiring at periodic intervals digital samples of the electrical signal generated by the photodetector,

(b) fitting a curve to a first group of a predetermined number of digital samples that are contiguous in time, (c) calculating at least one value of the slope function from said curve,

(d) repeating in time sequence the foregoing steps (b) and (c) on a plurality of different groups of said predetermined number of digital samples, the samples of individual groups including at least some of the samples but omitting at least an earliest acquired sample of the immediately preceding group.

28. The method according to claim 27 wherein the step of detecting when the slope function falls in magnitude below a first preselected fixed value for a first preselected period of time includes determining when a consecutive number of slope function values calculated from the groups of digital signal data remain below said first preselected fixed value for said first preselected period of time.

29. The method according to claim 23 wherein the material removal process includes developing a photoresist layer that has been made selectively dissolvable according to a pattern.

30. The method according to claim 29 wherein the photoresist layer development includes application of a developer solution to its said exposed surface.

31. The method according to claim 23 wherein the material removal process includes etching the layer of material through a mask.

32. The method according to claim 31 wherein the layer is being etched by contact of an etchant solution with the exposed surface of the layer through said mask.

33. The method according to claim 23 wherein the material removal process includes removing material from said layer carried by a semiconductor wafer.

34. The method according to claim 23 which comprises an additional step of terminating the process of removing material from said layer in response to confirming that breakthrough has occurred.

35. The method according to claim 23 wherein the breakthrough is confirmed to have occurred without comparison of the magnitude of the electrical signal with an absolute reference signal level.

36. A method of detecting a breakthrough in a plasma process of removing material from a layer of a composite structure, comprising the steps of: detecting an electromagnetic radiation component of said plasma, thereby to generate an electrical signal proportional to an intensity of said radiation component, calculating from said electrical signal during the removal process a function representing a time varying slope of said signal, determining from said slope function if the material removal process has started, and after it is determined that the material removal process has started, detecting when the slope function falls in magnitude below a preselected fixed value for a preselected period of time, whereby the breakthrough is confirmed to have occurred.

37. The method according to claim 36 wherein the slope function calculating step includes the following steps:

(a) acquiring at periodic intervals digital samples of the electrical signal generated by the detector,

(b) fitting a curve to a first group of a predetermined number of digital samples that are contiguous in time, (c) calculating a single value of the slope function from said curve,

38. The method according to claim 37 wherein the step of detecting when the slope function falls in magnitude below said preselected fixed value for said preselected period of time includes determining when a consecutive number of slope function values calculated from the groups of digital signal data remain below said preselected fixed value for said preselected period of time.

39. The method according to claim 36 wherein the material removal process includes removing material from said layer carried by a semiconductor wafer.

40. The method according to claim 36 which comprises an additional step of terminating the process of removing material from said layer in response to confirming that breakthrough has occurred.

41. The method according to claim 36 wherein the breakthrough is confirmed to have occurred without comparison of the magnitude of the electrical signal with an absolute reference signal level.