Patent Application
20080216956
The present application is based on and claims priority of Japanese patent application No. 2007-057426 filed on Mar. 7, 2007, the entire contents of which are hereby incorporated by reference.
1. Field of the Invention
The present invention relates to a film thickness and etching depth measuring method for detecting an etching amount of a material to be processed by an emission spectroscopy method in the manufacture of a semiconductor integrated circuit or the like, and to a processing method of the material to be processed by using the measuring method. More particularly, the present invention relates to a measuring method and device of a depth and a film thickness of a material to be processed, suitable for accurately measuring etching amounts of respective layers provided on a substrate in an etching processing using plasma discharge, and for obtaining a desired film thickness and an etching depth of the layers, and relates to a processing method and device of the material to be processed, using the measuring method.
2. Description of the Related Art
In the manufacture of a semiconductor wafer, dry etching is widely used for removing a layer of various materials formed on the surface of a wafer, and particularly for removing a layer of a dielectric material or forming a pattern of the dielectric material. For the control of process parameters, it is most important to accurately determine an etching end point to stop the etching at a desired film thickness and etching depth during the processing of the layers.
During the dry etching processing of the semiconductor wafer, the emission intensity of a specific wavelength in plasma light is changed in accordance with the progress of etching of a specific film. Thus, conventionally, as one of the methods for detecting the etching end point of the semiconductor wafer, there is a method for detecting the change in the emission intensity of the specific wavelength from plasma during the dry etching processing, and detecting the etching end point of the specific film on the basis of the detection result. In this case, it is necessary to prevent an erroneous detection based on fluctuation of a detected waveform due to a noise. As a method for accurately detecting the change in the emission intensity, there are known a detecting method based on a moving average method (see, for example, Japanese Patent Laid-Open Publication No. 61-53728 (Patent Document 1)), a method for reducing a noise by an approximate processing based on a primary least-squares method (see, for example, Japanese Patent Laid-Open Publication No. 63-200533 (Patent Document 2)), and the like.
In accordance with the miniaturization and high integration of a semiconductor in recent years, an opening ratio (area to be etched in a semiconductor wafer) is reduced, whereby the emission intensity of the specific wavelength taken into a photodetector from a photosensor is made weak. As a result, the level of a sampling signal from the photodetector is lowered to make it difficult for an end point determining device to surely detect the etching end point on the basis of the sampling signal from the photodetector.
Further, in the case of detecting the etching end point and stopping the etching processing, it is actually important that the remaining thickness of the dielectric layer is equal to a predetermined value. In the conventional process, the whole process is monitored by using a time and thickness control technique based on a premise that the etching rate of each layer is constant. The values of the etching rate are obtained by, for example, processing a sample wafer in advance. In this method, the etching process is stopped once time period corresponding to a predetermined etching film thickness has lapsed, on the basis of a time monitoring method.
However, it is known that an actual film, for example, an SiO2 layer formed by a LPCVD (low pressure chemical vapor deposition) technique has a low reproducibility of thickness. The allowable thickness error due to the process fluctuation during the LPCVD processing corresponds to about 10% of the initial thickness of the SiO2 layer. Therefore, the actual final thickness of the SiO2 layer remaining on a silicon substrate cannot be accurately measured by the method based on the time monitoring. Thus, the actual thickness of the remaining layer is finally measured by a technique using a standard spectroscopic interferometer. When it is found that the layer is excessively etched, the wafer is judged as unacceptable and discarded.
Further, in an insulating film etching device, there are known time-based changes such as a decrease in the etching rate in accordance with repetition of etching. This may result in a case where the etching is stopped in the midway, and hence is a problem which must be solved. In addition, for the stable process operation, it is important to monitor the time-based fluctuation of the etching rate. In the conventional method, however, only the time for determining the etching end point is monitored, and any suitable method is not used to cope with the time-based change and fluctuation of the etching rate. Further, for determining the etching end point in the case where the etching time is as short as about ten seconds, it is necessary to adapt the end point determining method to shorten the determination preparation time period, and also to sufficiently shorten the interval of determination time. However, these measures are not always enough. Further, in an insulating film, the area to be etched is 1% or less in many cases, and the change in the plasma emission intensity from a reaction product generated in accordance with the etching is small. Therefore, an end point determining system capable of detecting even a slight change is necessary, but there is no such system that is practical and inexpensive.
On the other hand, there is also known various methods using an interferometer as the other methods for detecting the etching end point of the semiconductor wafer. That is, a first method is a method in which the etching end point detection is performed by detecting interference light (plasma light) by the use of three kinds of color filters of red, green and blue (see, for example, Japanese Patent Laid-Open Publication No. 5-179467 (Patent Document 3)). A second method is a method in which extreme values of an interference waveform (maximum and minimum values: zero-crossing points of a differential waveform) are counted by using the time change of an interference waveform of two wavelengths and the differential waveform of the interference waveform, and in which an etching rate is calculated by measuring the time until the counted value reaches a predetermined value, and the remaining etching time until reaching a predetermined film thickness is obtained on the basis of the calculated etching rate, to stop the etching processing on the basis of the remaining etching time (see, for example, Japanese Patent Laid-Open Publication No. 8-274082 (U.S. Pat. No. 5,658,418 Specification) (Patent Document 4)). A third method is a method in which a waveform (with the wavelength taken as a parameter) of difference between a light intensity pattern (with the wavelength taken as a parameter) of interference light before processing and a light intensity pattern after the processing or during the processing is obtained, and in which a level difference (film thickness) is measured by the comparison between the obtained difference waveform and difference waveforms stored in a database (see, for example, Japanese Patent Laid-Open Publication No. 2000-97648 (Patent Document 5)). A fourth method is a method which relates to a rotary coating device, and in which a film thickness is obtained by measuring the time change in interference light over multiple wavelengths (see, for example, Japanese Patent Laid-Open Publication No. 2000-106356 (Patent Document 6)). A fifth method is a method in which a characteristic behavior of the time change of interference light is obtained by measurement and stored in a database, and the etching end determination is performed by the comparison between the database and a measured interference waveform, and prompts to change etching process conditions on the basis of the determination (see, for example, U.S. Pat. No. 6,081,334 Specification (Patent Document 7)).
In the method using the interferometer, monochromatic radiation emitted from a laser is made incident on a wafer including a laminated structure of different kinds of materials, at the vertical incident angle. For example, in a wafer in which an SiO2 layer is laminated on a Si3N4 layer, an interference fringe is formed by the emitted light reflected by the upper surface of the SiO2 layer, and the emitted light reflected by the interface formed between the SiO2 layer and the Si3N4 layer. The reflected emitted light is irradiated on a suitable detector, to generate a signal whose intensity is changed with the thickness of the SiO2 layer during the etching process. As soon as the upper surface of the SiO2 layer is exposed during the etching process, the etching rate and the current etching thickness can be continuously and accurately monitored. There is also known a method for measuring a predetermined emitted light emitted from plasma instead of the laser by a spectrometer.
The above described known techniques cause the following problems.
A. In the case of film thickness determination in a thick film processing process (resist etch-back with film thickness of several μm and the like), the interference light changes with time in a complicated manner for several tens periods or more, which causes the film thickness determination to be influenced by a slight disturbance.
B. In the case of film thickness determination in a thin film processing process (etch-back of a gate oxide film, an oxide film or the like), a weak change in interference light needs to be measured, which causes the film thickness determination to be influenced by a slight disturbance. That is, the time change in the interference light during the thin film processing is not more than ½ to ¼ periods, and the interference fringe is slightly changed, so that it is necessary to eliminate the effect of a noise in the film thickness determination.
C. In a wafer for processing in the mass production process, peripheral circuits or the like are mixedly provided, and various kinds of materials (a mask material, a material to be etched, the other materials in the peripheral circuits) are simultaneously subjected to etching processing, and hence interference lights from different materials are superimposed with each other in a complicated manner. In addition, there is a variation in the film thickness of various materials of the wafer for processing in a lot or between the lots, and hence the time change of the interference light during etching processing is changed in a lot or between the lots.
D. In the mass production processing of small amount and many kinds, since various etching processes are mixedly performed, the etching device is liable to change with time, so that an abnormal discharge or the like is caused to make the plasma fluctuate. Thereby, plasma light emission is changed and a disturbance is superimposed on the interference light, to influence the determination.
For the above described reasons, it has been difficult to accurately measure and control, in a required precision, a residual amount of film and an etching depth of a processed layer, and especially of a processed layer in the plasma etching processing.
An object of the present invention is to provide an etching end point determining method using a measuring method of a film thickness or an etching depth of a material to be processed, which is capable of accurately measuring online an actual residual amount of film and an etching depth of a layer to be processed in a plasma etching process in a semiconductor element manufacturing process, and to provide a plasma processing apparatus adapted to perform the etching end point determining method, a plasma processing method of a material to be processed, using the plasma processing apparatus, and a plasma processing apparatus using the plasma processing method.
Further object of the present invention is to provide an etching processing method which is capable of performing highly precise online control to make each layer of a semiconductor element processed into a predetermined film thickness and etching depth in a semiconductor element manufacturing process.
Further object of the present invention is to provide a measuring device of a film thickness and an etching depth of a material to be processed, which is capable of accurately measuring online an actual film thickness and etching depth of a material to be processed in a semiconductor element manufacturing process.
In order to solve the above described problems of the prior art, and to attain the above described objects of the present invention, the present inventors have obtained a time differential waveform of interference waveform for each of a plurality of wave lengths, and obtained a pattern showing a wavelength dependence of a differential value of the interference waveform (that is, a pattern of the differential value of the interference waveform, with the wavelength taken as a parameter), thereby effecting following malfunction operation prevention measures at the time of measuring a film thickness by using the pattern.
1) In a standard pattern database of an interference waveform corresponding to an etching amount (film thickness and depth) of a material to be etched, a comparison with a standard pattern whose etching amount is not more than a target etching amount, is not performed.
2) In the pattern matching between an interference waveform pattern measured at the time of etching and the standard pattern, a standard deviation value is monitored, and when the standard deviation value is large, the etching amount at the time point is estimated on the basis of the past etching amount transition.
3) When the etching amount obtained from the pattern matching with the standard pattern is greatly different from the amount estimated from the past etching amount transition, the etching amount at the time point is estimated on the basis of the past etching amount transition.
4) When the etching rate obtained from the past etching amount transition is compared with the etching rate of the standard pattern database and the obtained etching rate is greatly different from the standard pattern, the etching amount at the time point is estimated on the basis of the past etching amount transition.
5) When the etching rate obtained from the past etching amount transition is compared with the etching rate of the standard pattern database and the obtained etching rate is greatly different from the standard pattern, the etching amount which can be obtained by the pattern matching with the standard pattern is replaced by the etching amount which can be obtained from the etching rate of the standard pattern at the time point, and the etching amount at the time point is estimated on the basis of the replaced etching amount and the past etching amount transition.
The reason for using the pattern showing the wavelength dependence of a time differential value of an interference waveform in the present invention is that the measurement is based on the in-situ (real time) measurement during etching, and hence the film thickness of a processed film is changed every moment. Therefore, it is possible to perform time differential processing of the interference waveform, in order to reduce the influences of contamination and scraping of a measuring window which cause a problem in the interference light intensity measurement, but the time differential processing of the interference waveform need not always be performed.
Further, when plasma light emission is abruptly changed by abnormal discharge associated with the time-based change of a device, or the like, the etching amount measurement based on the interference light and the etching end point determination based on plasma light emission are performed in such a manner that a change quantity (ratio: correction coefficient) of the light emission is obtained by comparing a current light emission waveform with past light emission waveforms, and a current and future light emission waveforms are corrected on the basis of the correction coefficient.
In order to solve the above described problems of the prior art and further to achieve the above described objects of the present invention, the present inventors have devised a method wherein a waveform is obtained by arranging in time series time differentials of interference waveforms for each of a plurality of wavelengths of a reflection wave from a sample (semiconductor element) during plasma processing, wherein on the basis of the obtained waveform, a pattern showing a wavelength dependence of a differential value of the interference waveform, that is, a pattern in which the differential values of the interference waveforms, with the wavelength taken as a parameter, are arranged in time series, is obtained, and wherein a film thickness measurement is performed by using a pattern which is obtained by arranging in time series differential values of changes in the intensity of interference light of a plurality of wavelengths relating to the processing of another sample, and obtained before the processing of the sample currently processed, and which is formed by a plurality of standard differential waveforms respectively corresponding to a plurality of thicknesses of the film to be processed of the sample.
The reasons for using the pattern showing the wavelength dependence of a time differential value of an interference waveform in the present invention are as follows.
A. In the present invention, since the measurement is based on the in-situ (real time) measurement during etching, a residual film thickness of a processed film which is changed every moment, can be subjected to a time differential processing by using an interference waveform, and further a noise of the interference waveform can be removed by the differential processing.
B. Due to the fact that refractive indices of materials to be etched (for example, silicon and a nitride film of a mask material) are different from each other with respect to the wavelength, a characteristic change (film thickness dependence) of the each material can be measured by the interference light measurement over multiple wavelengths.
According to an aspect of the present invention, a residual film thickness measuring method and an etching depth measuring method which are an etching amount measuring method of a material to be processed, includes:
A. a step of setting, with the wavelength taken as a parameter, a standard differential pattern PS of a differential value of interference light with respect to a predetermined etching amount of a first (sample) material to be processed;
B. a step of setting, with the wavelength taken as a parameter, a standard differential pattern PM of a differential value of interference light with respect to a predetermined etching amount of a mask material which prevents the first material to be processed from being scraped;
C. a step of measuring, respectively for a plurality of wavelengths, the interference light intensity of a second material to be processed, which has the same constitution as that of the first material to be processed and is to be actually subjected to etching processing, and of obtaining an actual differential pattern (Pr) of a differential value of the measured interference light intensity, with the wavelength taken as a parameter; and
D. a step of obtaining an etching amount of the second material to be processed on the basis of the standard differential patterns (PS and PM) and the actual differential pattern (Pr) of the differential value.
According to the present invention, in the plasma processing, especially in the plasma etching processing, it is possible to provide a measuring method of a residual film thickness or an etching depth of a material to be processed, which is capable of accurately measuring online an actual etching amount of the material to be processed, and to provide a processing method of a sample of the material to be processed by using the measuring method.
Further, it is possible to provide an etching process capable of performing highly precise online control to make each layer of a semiconductor element (semiconductor device) processed to a predetermined etching amount. Further, it is possible to provide a measuring device of a residual thickness or a measuring device of an etching-depth of a material to be processed, which is capable of accurately measuring online an actual etching amount of a layer to be processed.
In the following, respective embodiments according to the present invention will be described. Note that in the following respective embodiments, an element having the same function as that in the first embodiment is denoted by the same reference numeral as that in the first embodiment, and the detailed explanation thereof is omitted.
In the following, a first embodiment according to the present invention will be described with reference to
First, with reference to
The etching amount measuring device 10 includes a spectroscope 11, a first digital filter circuit 12, a differentiator 13, a second digital filter circuit 14, a differential waveform comparator 15, a differential waveform pattern database 16, a pattern matching deviation comparator 115, a deviation value setting device 116, a residual film thickness time series data recorder 18, a regression analyzer 19, an end point determining device 230, and an indicator 17 which displays a result from the end point determining device. Note that
In the spectroscope 11, an emitted light of multiple wavelengths from a measuring light source (for example, halogen light source) is introduced into the inside of the vacuum container 2 by the optical fibre 8, and made incident at the vertical incident angle onto the material to be processed 4. An interference light from the material to be processed 4 is introduced into the spectroscope 11 of the etching amount measuring device 10 via the optical fibre 8, and on the basis of the state of the interference light, an etching depth measurement of silicon or a residual film thickness measurement of a mask material, and an end point determination processing of etching processing are performed.
The emission intensity of the interference light of multiple wavelengths received by the spectroscope 11 are made into current detecting signals, each of which corresponds to emission intensity for each specific wavelength, and converted to voltage signals. The signals of a plurality of specific wavelengths (j denotes the number of wavelength) outputted as sampling signals by the spectroscope 11 are stored as time series data yi,j in a storage device such as a RAM (not shown). Next, the time series data yi,j at time i is subjected to a smoothing processing by the first digital filter circuit 12, and stored as smoothed time series data Yi,j in a storage device such as a RAM (not shown). On the basis of the smoothed time series data Yi,j, time series data di,j of differential coefficient values (first-order differential values or second-order differential values) is calculated by the differentiator 13, and stored in a storage device such as a RAM (not shown). The time series data di,j of the differential coefficient values is subjected to a smoothing processing by the second digital filter circuit 14, and stored as smoothed differential coefficient time series data Di,j in a storage device such as a RAM (not shown). Then, an actual differential pattern (Prj)=Σj (Di,j) (wavelength j taken as a parameter) which shows the wavelength dependence of the differential value of the interference light intensity can be obtained from the smoothed differential coefficient time series data Di,j.
On the other hand, in the differential waveform pattern database 16, there are set differential waveform pattern data values PSj of the interference light intensity with respect to the each wavelength corresponding to an etching depth which is obtained beforehand by using the first (sample) material to be processed, and expressed by a residual film thickness s of the material to be processed, so as to serve as a target of the etching amount measurement. In the differential waveform comparator 15, the actual differential pattern (Prj)=Σ(Di,j) and the differential waveform pattern data value PSj of the film thickness s are compared with each other. In the pattern matching deviation comparator 115, a pattern matching (minimum) deviation value σs at which a pattern matching deviation (σs=√(Σj(Di,j−PSj)×(Di,j−PSj)/j)) is minimized, is obtained and compared with a pattern matching (set) deviation value σ0 set beforehand in the deviation value setting device 116. When the pattern matching (minimum) deviation value σs is not more than the pattern matching (set) deviation value σ0, the film thickness value s is stored as an instantaneous film thickness Zi at time i in the film thickness time series data recorder 18. When the pattern matching (minimum) deviation value σs is not less than the pattern matching (set) deviation value σ0, the instantaneous film thickness Zi at time i is not stored.
In the regression analyzer 19, a calculated film thickness F at time i is obtained on the basis of the regression straight line approximation using the instantaneous film thickness data before the time i. Whether or not the calculated film thickness F is not more than the target film thickness set beforehand is determined by the end point determining device 230. The result of etching amount of the material to be processed which is obtained by the above described processing is displayed by the result indicator 17.
Note that there is shown a case where only one spectroscope 11 is provided in the first embodiment, but when it is desired that the inside of the surface of a material to be processed is measured in an expanded state and controlled, a plurality of spectroscopes 11 may be provided.
Next, by using a flow chart shown in
First, there are performed a setting of a target etching amount (target residual film thickness value), and a setting of differential patterns (residual film thickness standard differential patterns) PSj of wavelength regions (at least three wavelength regions) extracted from the differential waveform pattern database 16, and a setting of a pattern matching (set) deviation value σ0 (step 600). That is, a standard differential pattern PSj corresponding to an etching amount (residual film thickness) s which is needed in accordance with processing conditions of a material to be processed is set beforehand in the differential waveform pattern database 16.
In the subsequent step, a sampling (at every 0.25 to 0.5 seconds) of interference light from an object to be processed is started (step 601). That is, a sampling start instruction is issued in accordance with the start of etching processing. The emission intensity of multiple wavelengths which changes in accordance with the progress of etching is detected by a photodetector (spectroscope 11) as a photodetection signal of a voltage corresponding to the emission intensity. The photodetection signal of each wavelength j from the spectroscope 11 is digitally converted so that a sampling signals yi,j are obtained.
Next, multiple wavelength output signals yi,j from the spectroscope 11 are smoothed by the first stage digital filter 12, so that smoothed time series data Yi,j are calculated (step 602). That is, a noise is reduced by the first stage digital filter, so that smoothed time series data Yi,j are obtained.
Next, differential coefficients di,j for each wavelength are calculated by differentiating the smoothed time series data Yi,j by the S-G method in the differentiator 13 (step 603). That is, (first-order or second-order) differential coefficients di,j of signal waveforms for the respective wavelengths are obtained by the differential processing (the S-G method). Further, the smoothed differential coefficient time series data Di,j are calculated by the second stage digital filter 14 (step 604). Then, in the differential waveform comparator 15, a matching pattern (minimum) deviation value σs=√(Σ(Di,j−PSj)2/j) is calculated to obtain a minimum value σ of the matching pattern (minimum) deviation value σs with respect to the residual film thickness s (step 605).
Next, in the pattern matching deviation comparator 115, the determination whether σ≦σ0 is performed by comparing the calculated matching pattern (minimum) deviation value σ with the matching pattern (set) deviation value σ0 (step 606). When σ≦σ0, it is determined that the film thickness of the material to be processed is made into a residual film thickness s, and an instantaneous film thickness Zi at a time point i is stored in the residual film thickness time series data recorder 18 (step 607). Except when σ≦σ0, the instantaneous film thickness Zi at the time point i cannot be obtained from the standard differential pattern database, and the instantaneous film thickness is not stored in the residual film thickness time series data recorder 18 (step 608). These smoothed differential coefficient time series data Di,j and the differential patterns PSj set beforehand in the differential waveform comparator 15 are compared with each other, so that a residual film thickness value Zi at the time point is calculated (step 615).
Next, by the use of the stored past time series data Zi, a first-order regression straight line Y=Xa×t+Xb (Y: residual film amount, t: etching time, absolute value of Xa: etching rate, and Xb: initial film thickness) is obtained by the regression analyzer 19, so that a calculated residual film amount F at time point i (present time) is calculated on the basis of the regression straight line (step 609). Next, in the end point determining device 230, the calculated residual film amount F and the target residual film thickness value are compared with each other to determine the etching amount (residual film thickness value). When the calculated residual film amount F is not more than the target residual film thickness value, it is determined that the etching amount of the material to be processed is made into a predetermined value, and the comparison result is displayed in the indicator 17 (step 610). When the calculated residual film amount F is not less than the target residual film thickness value, the process returns to step 602 and these steps are repeated. Finally, when the calculated residual film amount F is not more than the target residual film thickness value in step 610, the setting of the end of the sampling is performed (step 611).
Here, the calculation of the smoothed differential coefficient time series data Di relating to a certain wavelength j at time point i is explained. As the first digital filter circuit 12, for example, a second-order Butterworth type low pass filter is used. The smoothed time series data Yi can be obtained by the second-order Butterworth type low pass filter on the basis of the following formula (1).
Yi=b1yi+b2yi−1+b3yi−2−[a2Yi−1+a3Yi−2] (1)
Here, numerical values of the coefficients b and a are different in dependence upon a sampling frequency and a cut-off frequency. For example, there are used a2=−1.143, a3=0.4128, b1=0.067455, b2=0.13491, b3=0.067455 (sampling frequency: 10 Hz, cut-off frequency: 2.5 Hz), or a2=−0.00073612, a3=0.17157, b1=0.29271, b2=0.58542, b3=0.29271 (cut-off frequency: 2.5 Hz), or the like.
The time series data di of second-order differential coefficient values are calculated as follows on the basis of the following formula (2) using the polynomial fitting smoothing differential method of the time series data Yi of five points by the differentiator (differential coefficient operation circuit) 13.
Here, w−2=2, w−1=−1, w0=−2, w1=−1, and w2=2.
By the use of the time series data di of the differential coefficient values, the smoothed differential coefficient time series data Di can be obtained by the second digital filter circuit (second-order Butterworth type low pass filter, but the coefficients a and b of the digital filter circuit may be different) 14 based on the following formula (3).
Di=b1di+b2di−1+b3di−2−[a2−Di−1+a3Di−2] (3)
Here, in order to prevent the abrupt lowering of the instantaneous film thickness, the behavior of the interference waveform was analyzed. Generally, in the interference waveform change, when a material is made into a thin film, the interference light ceases to change in many wavelength regions, and hence first-order differential changes at these wavelengths simultaneously approach zero. Further, even when plasma fluctuation is caused, the corresponding changes are simultaneously caused in many wavelength regions, and the first-order differentials at the wavelengths are simultaneously changed. As the change in plasma is reduced, the first-order differential at the wavelengths approaches zero. Such behavior of the first-order differential is similar to that of the change in the interference light of a thin film. Therefore, in order to prevent such abrupt change, it is necessary to avoid the use of data for thin film thickness as much as possible among the data of standard differential patterns used for the thickness measurement. That is, it is necessary to perform the pattern matching processing with the standard differential pattern, in such a manner that a standard differential pattern of a film thickness less than a target determination film thickness is not used for obtaining an instantaneous film thickness during the etching processing.
Next, there is shown a further embodiment for avoiding plasma fluctuation. Here, there is utilized the fact that when plasma fluctuation is caused, the pattern matching deviation is increased. Generally, during a period of several seconds from the start of the differential processing for film thickness determination after the start of the etching processing, the interference waveform is slightly disturbed due to the influence of plasma ignition, so that the pattern matching deviation value σ deteriorates. On the basis of the pattern matching (set) deviation value σ0 at this time point, the pattern matching deviation value σ after the time point is calculated. When the pattern matching deviation value σ is larger than the pattern matching (set) deviation value σ0, it is determined that the pattern matching with the standard differential pattern is not enough. Then, the instantaneous film thickness Zi is not obtained from the standard differential pattern, but set, for example, to the initial film thickness of the database (standard differential patterns). Thus, the instantaneous film thickness data set as the initial film thickness at this time point is not used for the regression straight line approximate analysis for obtaining the calculated film thickness F.
As a second embodiment,
In the process of mass production adapted to process a semiconductor wafer by using plasma in order to produce a semiconductor device from the semiconductor wafer, a plasma processing apparatus such as the device according to the present invention is continuously operated, to cause fluctuation in conditions in the processing chamber, due to adherence and deposition of products onto the surface of member inside the processing chamber in accordance with the increase in the number of pieces of materials to be processed, and the like. This causes the state of plasma generated in the processing chamber to fluctuate, and causes the surface shape obtained as a result of the processing to be changed. Therefore, it is necessary to manage the process so as to control the fluctuation in the result of processing of the material to be processed in the above described mass production. In the case where such mass production management is performed in the present embodiment, the number of times when the pattern matching deviation value exceeds a predetermined value is monitored in the processing for each wafer as a material to be processed, and the number of times is counted by a recorder, a counter (both not shown), or the like. Such counting may also be performed by the pattern matching deviation comparator 115.
Further, it is possible to grasp a device status and a wafer etching status by comparing transition of the number of times with a predetermined value (for example, a predetermined value about a value of the number of times and an increasing rate). That is, when the number of times is gradually increased, a predetermined value of the number of times is used as a measure to start maintenance work, such as the wet cleaning, in the plasma processing apparatus. When the number of times is abruptly increased and the increasing rate becomes larger than a predetermined value, the need for measures, such as conveying the wafer to be processed to the wafer inspection step, is informed to a user, or a warning is issued to the user. Such notification and warning are displayed, for example, in the indicator 17 shown in
Next, a third embodiment for avoiding plasma fluctuation is explained. Here, instead of stabilizing the instantaneous film thickness transition by the above described pattern matching deviation comparison, the instantaneous film thickness Zi at a time point i during etching processing is obtained. In the case where a calculated film thickness F at the time point i is calculated by the regression straight line approximation on the basis of the time series data of the instantaneous film thickness Zi before the time point i, and where the difference (absolute value) between the calculated film thickness F and the instantaneous film thickness Zi is not less than a film thickness allowable value set beforehand, it is determined that the instantaneous film thickness Zi at the time point i is not an accurate film thickness.
Next, there is explained a fourth embodiment which relates to avoiding erroneous determination in the film thickness measurement, and which utilizes the fact that the etching rate during the mass production processing is almost constant and the fluctuation of the etching rate is in a range of at most ±10%. From the instantaneous film thickness transition shown in
As shown in
First, as shown in
In the subsequent step, the sampling of interference light (for example, every 0.25 to 0.5 seconds) is started (step 1601). That is, a sampling start instruction is issued in correspondence with the start of etching processing. The emission intensity of multiple wavelengths which is changed in accordance with the progress of etching is detected by the photodetector as a photodetection signal of a voltage corresponding to the emission intensity. The photodetection signal of the spectroscope 11 is digitally converted, so that the sampling signals yi,j are acquired.
Next, the multiple wavelength output signals yi,j from the spectroscope 11 are smoothed by the first stage digital filter 12, and the smoothed time series data Yi,j are calculated (step 1602). That is, a noise is reduced by the first stage digital filter, so that the smoothed time series data Yi,j are obtained.
Next, the differential coefficients di,j are calculated by the S-G method in the differentiator 13 (step 1603). That is, (first-order or second-order) differential coefficients di of the signal waveforms are obtained by the differential processing (the S-G method). Further, the smoothed differential coefficient time series data Di,j are calculated by the second stage digital filter 14 (step 1604). Then, in the differential waveform comparator 15, the pattern matching (minimum) deviation value σs=√(Σ(Di,j−PSj)2/j) is calculated to obtain a minimum value σ which is the smallest pattern matching (minimum) deviation value σs with respect to a film thickness s (step 1605).
Next, in the pattern matching deviation comparator 115, the determination whether σ≦σ0 is performed for comparing the calculated matching pattern deviation value (minimum value) σ and the matching pattern (set) deviation value σ0 with each other (step 1606). When σ≦σ0, it is determined that the film thickness of the material to be processed is made into the film thickness s, and an instantaneous film thickness Zi at a time point is stored in the residual film thickness time series data recorder 18 (step 1607). Except when σ≦σ0, the instantaneous film thickness Zi at the time point i cannot be obtained from the standard differential pattern database, and the instantaneous film thickness is not stored in the residual film thickness time series data recorder 18 (step 1608).
As for the etching rate during the processing, the calculated film thickness F and the inclination Xa thereof are obtained by the first-order regression straight line approximation in the regression analyzer 19 on the basis of the data in the residual film thickness time series data recorder 18 (1609). Next, in the residual film thickness comparator 20, it is determined whether or not the instantaneous film thickness Zi is a film thickness limited by the calculated film thickness F and the film thickness allowable value z0 (F−z0≦Zi≦F+z0), or in the etching rate comparator 21, it is determined whether or not the inclination of the straight line Xa obtained by the regression approximation is an etching rate limited by the etching rate R at the time of creating the standard differential pattern and the etching rate allowable value R0 (R−R0≦Xa≦R+R0). In the case where (F−z0≦Zi≦F+z0) or where (R−R0≦Xa≦R+R0), the instantaneous film thickness Zi is stored in the residual film thickness time series data recorder 18 (step 1612). In the other cases, the instantaneous film thickness Zi is not stored in the residual film thickness time series data recorder 18 (step 1611).
Subsequently, the film thickness determination is performed on the basis of the calculated film thickness F. When the calculated film thickness F is not more than the target residual film thickness value, it is determined that the etching amount of the material to be processed has reached the predetermined value, and the result is displayed in the indicator 17 (step 1613). The state of the film thickness change during etching can be displayed by displaying the calculated film thickness F. When the calculated film thickness F is not less than the target residual film thickness value, the process returns to step 1602 and these steps are repeated. Finally, the setting of the end of the sampling is performed (step 1614).
When the etching processing is normally performed, the number of data of the instantaneous film thickness which have not been stored in the residual film thickness time series data recorder 18, is almost zero. When the etching characteristic of the etching device is changed due to the time-based change of the device, the matching of the interference differential pattern deteriorates so as to increase the number of data which are not stored. Further, when the specification of the wafer to be processed is changed, the pattern matching deteriorates so as to increase the number of data. Therefore, in the mass production process, it is possible to perform the device management of the etching device and the production management of processed wafers, by displaying in the indicator 17 the number of data of the instantaneous film thickness which have not been stored in the residual film thickness time series data recorder 18.
Next, there is explained a fifth embodiment in which the film thickness determination is performed by correcting the interference light and reference light, which are observed when plasma fluctuation is caused. The interference light is changed by a steep change in plasma emission caused by plasma fluctuation (abnormality), which may make it difficult to obtain an accurate film thickness, as shown, for example, in
Thus, there is explained a method for obtaining a film thickness in a manner of detecting a steep change in plasma, obtaining a change quantity for each wavelength used for measurement, correcting an optical signal of each wavelength in accordance with the change quantity of each wavelength, and performing processings such as the digital filter processing and the polynomial fitting smoothing differential processing to the corrected optical signals.
When collecting the standard pattern data as a database for film thickness determination, a change quantity (difference during a period between time points i and i−1) at a sampling point of the time point i, of the emission data obtained by measuring the time change of the interference light and the reference light, is checked for each wavelength, so that maximum change amounts during etching processing are obtained for the interference light and the reference light. A noise threshold value is set on the basis of the maximum changes per one sampling, so that a steep change in plasma is detected by using the noise threshold value.
Next, when there is a change quantity exceeding the noise threshold value, each correction coefficient for each wavelength (intensity ratio: Si,j=yi−1,j/yi,j) is obtained, to correct the optical signals yi,j on the basis of a formula: y′i,j=Si,j*yi,j. The instantaneous film thickness Zi is obtained by performing processings such as the digital filter processing and the polynomial fitting smoothing differential processing for the corrected optical signals y′i, j, and the determination is performed.
With reference to
The etching amount measuring device 10 includes the spectroscope 11, a sampling data comparator 110, a noise value setting device 111 for setting a noise threshold value, a correction coefficient recorder and indicator 113, a sampling data corrector 112, a first digital filter circuit 12, a differentiator 13, a second digital filter circuit 14, a differential waveform comparator 15, a differential waveform pattern database 16, a pattern matching deviation comparator 115, a deviation value setting device 116, a residual film thickness time series data recorder 18, a regression analyzer 19, an end point determining device 230, and an indicator 17 for displaying a result from the end point determining device 230.
The emission intensities of multiple wavelengths received by the spectroscope 11 are made into current detecting signals, each of which corresponds to emission intensity for each specific wavelength, and converted to voltage signals. The signals of a plurality of specific wavelengths (j pieces) outputted as sampling signals by the spectroscope 11 are compared, in the sampling data comparator 110, with values set beforehand in the noise value setting device 111. When the change value of the signals exceeds the noise value, the time series data yi,j are corrected in the sampling data corrector 112, so as to prevent the signals from being changed. The correction coefficients at this time are stored in the correction coefficient recorder and indicator 113. In this way, the time series data y′i,j obtained by being corrected on the basis of the instantaneously changed signals are stored in a storage devices such as a RAM. Next, the time series data y′i,j at a time point i are smoothed by the first digital filter circuit 12, and stored in a storage device such as a RAM, as a smoothed time series data Yi,j. On the basis of the smoothed time series data Yi,j, time series data di,j of differential coefficient values (first-order differential values or second-order differential values) are calculated by the differentiator 13, and stored in a storage devices such as a RAM. The time series data di,j of the differential coefficient value are subjected to a smoothing processing by the second digital filter circuit 14, and stored as smoothed differential coefficient time series data Di,j in a storage device such as a RAM. Thus, an actual pattern (with wavelength taken as a parameter) which shows the wavelength dependence of the differential value of interference light intensity can be obtained from the smoothed differential coefficient time series data Di,j.
On the other hand, in the differential waveform pattern database 16, there are set beforehand differential waveform pattern data values PSj of the interference light intensity with respect to respective wavelengths, each of which value corresponds to a film thickness s of the material to be processed, as an object of the etching amount measurement. In the differential waveform comparator 15, an actual pattern and the differential waveform pattern data value PSj of the film thickness s are compared with each other. In the pattern matching deviation comparator 115, a pattern matching (minimum) deviation value σs at which a pattern matching deviation (σs=√(Σj(Di,j−PSj)×(Di,j−PSj)/j)) is minimized, is obtained and compared with a deviation value σ0 set beforehand in the deviation value setting device 116. When the pattern matching (minimum) deviation value σs is not more than the pattern matching (set) deviation value σ0, the film thickness value s is stored as an instantaneous film thickness at the time point i in the film thickness time series data recorder 18. When the pattern matching (minimum) deviation value as is not less than the pattern matching (set) deviation value σ0, the instantaneous film thickness at the time point i is not stored. In the regression analyzer 19, a calculated film thickness F at the time point i is obtained on the basis of the regression straight line approximation using the instantaneous film thickness data Zi before the time point i. Whether or not the calculated film thickness F is not more than the target film thickness set beforehand is determined by the end point determining device 230. The resultant etching amount of the material to be processed, which is obtained by the above described processing, is displayed by the result indicator 17.
In the block diagram shown in
Next, with reference to a flow chart shown in
First, there are set a target etching amount (target residual film thickness value), differential patterns PSj whose wavelength regions (at least three wavelength regions) are extracted from the standard differential pattern database, a deviation value σ0, and a noise value N (step 2600). That is, a standard differential pattern corresponding to an etching amount s needed in accordance with processing conditions of a material to be processed, is set beforehand in the differential waveform pattern databases 15 and 25.
In the subsequent step, the sampling of the interference light (for example, every 0.25 to 0.5 seconds) is started (step 2601). That is, a sampling start instruction is issued in correspondence with the start of the etching processing. The emission intensity of multiple wavelengths which is changed in accordance with the progress of etching is detected by a photodetector as a photodetection signal of a voltage corresponding to the emission intensity. The photodetection signals of the spectroscope 11 are digitally converted, so that the sampling signal yi,j are acquired.
Next, differences between the multiple wavelength output signals yi,j and signals yi−1,j at a time point i−1 are obtained (step 2604). It is determined whether or not the differences (yi,j−yi−1,j) are larger than the value N set beforehand in the noise value setting device 111, by using the sampling data comparator 110 (step 2620). In the cases of the embodiments shown in
Next, the differential coefficients di,j are calculated by the S-G method in the differentiator 13 (step 2603). That is, the (first-order or second-order) differential coefficients di of the signal waveforms are obtained by the differential processing (the S-G method). Further, the smoothed differential coefficient time series data Di,j are calculated by the second stage digital filter 14 (step 2604). Then, in the differential waveform comparator 15, the value σs=√(Σ(Di,j−PSj)2/j) is calculated, to obtain a minimum value σ of the pattern matching (minimum) deviation value σs with respect to the film thickness s (step 2605). Next, in the pattern matching deviation comparator 115, the determination (σs: matching pattern (minimum) deviation value σ, and σ0: matching pattern (set) deviation value) is performed (step 2606). When σs≦σ0, it is determined that the film thickness of the material to be processed is made into the film thickness s, and an instantaneous film thickness at the time point i is stored in the residual film thickness time series data recorder 18 (step 2607). Except when σs≦σ0, the instantaneous film thickness at the time point i cannot be obtained from the standard differential pattern database, and the instantaneous film thickness is not stored in the residual film thickness time series data recorder 18 (step 2608). The smoothed differential coefficient time series data Di,j and differential patterns PZj set beforehand in the differential waveform comparator 15 are compared with each other, so that a residual film value Zi at that time point is calculated (step 2615). Next, by the use of the stored past time series data Zi, a first-order regression straight line Y=Xa×t+Xb (Y: residual film amount, t: etching time, absolute value of Xa: etching rate, Xb: initial film thickness) is obtained by the regression analyzer 19, so that a calculated residual film amount F at the time point i (present time) is calculated on the basis of the regression straight line (step 2609). Next, in the end point determining device 230, the calculated residual film amount F and the target residual film thickness value are compared with each other. When the calculated residual film amount F is not more than the target residual film thickness value, it is determined that the etching amount of the material to be processed is made into the predetermined value, and the comparison result is displayed in the indicator 17 (step 2609). When the calculated residual film amount F is not less than the target residual film thickness value, the process returns to step 2604 and these steps are repeated. Finally, the setting of the end of the sampling is performed (step 2611).
Next, there is explained the interference light measurement when discharge fluctuation is caused as shown in
In this example, the reference light is used as means for detecting a steep change in plasma, but it may be adapted such that values of reflection power and a matching point of electric power for generating plasma, or values of reflection power and a matching point of bias applied to the wafer are monitored, and a change in the values are used as means for detecting the steep change in plasma.
Further, here, the correction coefficients are obtained by Si,j=yi−1,j/yi,j, but an average value of a plurality of waveform data before the time point i−1 may also be used as the correction coefficient. Further, the approximate value at the time point i−1 based on a smooth curve obtained from the past time series data by using an interpolation method, such as the Lagrange interpolation method and a spline interpolation method, may also be used as the correction coefficient. In addition, the emission data at the time point i may be further modified by applying the Lagrange interpolation method and the spline interpolation method to the emission data at the time point i which are modified by the correction coefficient. Further, in the embodiments shown in
When the etching processing is normally performed, the number of times when the noise threshold value is exceeded is zero. However, when the etching characteristic of the etching device is changed due to the time-based change of the device and the plasma state deteriorates, the number of times when the noise threshold value is exceeded is increased. Therefore, in the mass production process, it is possible to perform the device management of the etching device by displaying in the indicator 17 the number of times when the noise threshold value is exceeded.
Next, with reference to
First, as shown in
In the subsequent step, the sampling of the interference light (for example, every 0.25 to 0.5 seconds) is started (step 3601). That is, a sampling start instruction is issued in correspondence with the start of the etching processing. The emission intensity of multiple wavelengths which is changed in accordance with the progress of etching is detected as a photodetection signal of a voltage corresponding to the emission intensity by the photodetector. The photodetection signal of the spectroscope 11 is digitally converted, so that the sampling signals yi,j are acquired.
Next, the multiple wavelength output signals yi,j from the spectroscope 11 are smoothed by the first stage digital filter 12, and the smoothed time series data Yi,j are calculated (step 3602). That is, a noise is reduced by the first stage digital filter, so that the smoothed time series data Yi,j are obtained.
Next, the differential coefficients di,j are calculated by the S-G method in the differentiator 13 (step 3603). That is, the (first-order or second-order) differential coefficients di of the signal waveforms are obtained by the differential processing (the S-G method). Further, the smoothed differential coefficient time series data Di,j are calculated by the second stage digital filter 14 (step 3604). Then, in the differential waveform comparator 15, the pattern matching (minimum) deviation value σs=√(Σ(Di,j−PSj)2/j) is calculated to obtain the instantaneous film thickness data Zi from a minimum value σ of the pattern matching (minimum) deviation value σs with respect to the film thickness s (step 3605). The instantaneous film thickness data Zi is stored in the residual film thickness time series data recorder 18.
As for the etching rate during the processing, an inclination Xa is obtained by the etching rate comparator 21 from a regression straight line 1 on the basis of the instantaneous residual film time series data Zi (step 3606). Next, it is determined whether or not the current inclination Xa is an etching rate limited by the etching rate R at the creation time of the standard differential pattern, and the etching rate allowable value R0 (R−R0≦Xa≦R+R0) (step 3607). In the case where (R−R0≦Xa≦R+R0), the instantaneous film thickness Zi is stored in the residual film thickness time series data recorder 18 (step 3608). In other cases, the instantaneous film thickness Zi is not stored in the residual film thickness time series data recorder 18 (step 3609). When the instantaneous film thickness Zi is not stored, correction instantaneous film thickness data Fei and correction instantaneous film thickness auxiliary data Fepi and Femi which are respectively values of Fei ±10%, are obtained from the etching rate R at the creation time of the standard differential pattern, and stored in the residual film thickness time series data recorder 18 instead of the current instantaneous film thickness data Zi (step 3610). Here, the reason for using the values of Fei ±10% is based on the fact that the fluctuation of the etching rate is at most about ±10%, but other values may also be set depending upon the kind of wafer and the state of etching.
Next, in the noise determining device 202, the instantaneous film thickness time series data Zi, the correction instantaneous film thickness data Fei, the correction instantaneous film thickness auxiliary data Fepi and Femi which are respectively the values of Fei ±10%, are compared with the regression straight line 1 calculated in step 3606, so that the instantaneous film thickness data Zi having a difference not less than a fixed value (for example, not less than 10 nm) from the regression straight line 1 is set as noise data (step 3611).
The calculated film thickness F at the time point i is obtained by the regression straight line 2 in the regression analyzer 19 by using the data of the instantaneous film thickness time series data Zi, the correction instantaneous film thickness data Fei and the correction instantaneous film thickness auxiliary data Fepi and Femi, which data are not determined as the noise by the noise determining device 202 (step 3612). It is determined in the end point determining device 230 whether or not the calculated film thickness F is not more than the target residual film thickness value set beforehand (step 3613). Finally, when the calculated residual film amount F is not more than the target residual film thickness value in step 3613, the setting of the end of the sampling is performed (step 3614). The result of etching amount of the material to be processed obtained by the above processing is displayed in the result indicator 17.
In the sixth embodiment, the transition of the calculated film thickness value in the unstable region of plasma during the early stage of etching is stabilized by the film thickness correcting function. In a thin film wafer, the etching time period is short, and the calculated film thickness value stabilized from the early stage of etching is important. When the calculated film thickness value is unstable, the etching processing is ended at a film thickness not less than the target film thickness, which results in a production failure.
In the case where the mass production management is performed in the present embodiment, the totals of the number of data determined as the noise and the data deviation amount (noise amount) are monitored by the processing for each wafer, and the totals of the number of data determined as the noise and the data deviation amount are counted by a recorder, a counter or the like (not shown). Such counting operation may also be performed by the film thickness corrector 201. Further, it is possible to grasp the state of the device and the etching state of the wafer, by comparing the transition of the totals of the number of data determined as the noise and the data deviation amount with predetermined values (for example, predetermined values relating to a value of the number of data and the increasing rate). That is, in the case where the above described number of data is gradually increased, the predetermined values for the totals of the number of data and the data deviation amount are used as measures for starting the maintenance work, such as the wet cleaning, in the plasma processing apparatus. When the totals of the number of data and the data deviation amount are abruptly increased and the increasing rate of the totals exceeds the predetermined value, the need for a measure such as that of conveying the wafer to the inspection process of the wafer to be processed, is informed to a user, and a warning is issued. Such information and warning are displayed, for example, in the indicator 17 shown in
According to the present invention, it is possible to provide a film thickness measuring method in which an etching amount of a material to be processed can be accurately measured online in plasma processing, especially in plasma etching processing, and to provide a process end point determining method using the film thickness measuring method.
Further, it is possible to provide an etching process in which a layer to be etched of a semiconductor device can be highly precisely controlled online, so as to be etched by a predetermined etching amount. Further, it is possible to provide an etching amount measuring device of a material to be processed, in which an actual etching amount of a layer to be processed accurately measured online.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2007-057426 | Mar 2007 | JP | national |
