JP2013252613A - End point detection device and polishing apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an end point detection device capable of determining a polishing end point with high accuracy during polishing.SOLUTION: An end point detection device includes: an initial condition setting unit 62 for setting a permissible time determined with a constant time width before and behind a prediction polishing time; a detection condition determination unit 64 for plurally selecting at least two combinations of spectrum wavelength ranges from a plurality of spectrum wavelength ranges detected by polishing a first wafer beforehand to calculate spectrum composite signals of at least two wavelength ranges, respectively, and for determining combinations of at least two specific wavelength ranges in which an output value of the composite signal generated during the permissible time represents such a feature point as being coincident with a predetermined condition from among the plurality of at least two wavelength ranges; and an end point determination unit 65 for determining the end of polishing by detecting a polishing end point at a time point when the composite signal obtained on the basis of the spectrum of at least two specific wavelength ranges detected during the polishing of another wafer represents the feature point.

Description

  The present invention relates to an end point detection device for detecting an end point of polishing in polishing processing of an object to be polished such as a semiconductor wafer, and a polishing apparatus using the end point detection device.

  The progress of high-density semiconductor devices has continued without limit, and various technologies and methods are being developed to achieve high density. One of them is multilayer wiring, and technical problems associated with this include global device surface flattening (in a relatively large area) and wiring between upper and lower layers. Considering the reduction of the depth of focus during exposure accompanying the shortening of the lithography wavelength, there is a great demand for accuracy in flattening the interlayer layer at least in the range of the exposure area. In addition, a so-called image fitting (plug, damascene) requirement for embedding a metal electrode layer is also great for realizing a multilayer wiring. In this case, removal and planarization of an extra metal layer after lamination must be performed.

  A polishing process called CMP (Chemical Mechanical Polishing) is drawing attention as an efficient planarization technique for these large areas (at the die size level). CMP is the most promising candidate for global planarization and electrode formation technology in the process of removing the surface layer of the wafer by using a chemical action in combination with physical polishing. Specifically, using an abrasive called slurry in which abrasive grains (silica, alumina, cerium oxide, etc.) are dispersed in a soluble solvent of an abrasive such as acid, alkali, or oxidizer, suitable polishing is performed. Polishing is advanced by pressing the wafer surface with a cloth and rubbing with relative motion.

  This process has many problems as a device process technology, and among them, the polishing process speed is not constant due to various reasons, so it is a problem to know the timing of the appropriate polishing process completion or change. Yes. For this reason, there is a need for a polishing process monitoring technique called end point detection (EPD) that performs measurement of the wafer surface state by in-site polishing (hereinafter referred to as “in-situ polishing measurement”). As an end point detection device conventionally used, there is known an end point detection device that optically measures the surface state of a wafer (see, for example, Patent Document 1). In such an apparatus, the surface of the wafer (surface to be polished) is irradiated with probe light, reflected light from the wafer irradiated with the probe light is detected, and end point detection (EDP) is performed based on the intensity change and spectrum distribution. ).

Japanese Patent No. 3360610 Japanese Patent Laid-Open No. 11-33901

  In general, the CMP process includes not only a metal CMP process but also a shallow trench isolation (STI) process for polishing a silicon oxide film, an interlayer dielectric film (ILD) process, and the like. As an example of what is called a metal CMP process, the point at which the film switches from the copper layer to the barrier layer (such as tantalum) must be detected, but the exposure of the film from the copper layer to the barrier layer greatly changes its reflectivity. Therefore, the end point can be detected relatively easily using the conventional optical end point detecting device. On the other hand, in the STI process, detection of each state such as detection of a state where the layer is switched from the oxide film to the nitride film, detection of a state where the oxide film is slightly left, or a state where the nitride film is slightly processed is required. Similarly, in the ILD process, it is required to detect the film thickness of an arbitrary dielectric film. However, in the STI process, it is not easy to detect the end point stably because the nitride film exposure point has a low reflectivity. Further, as represented by the ILD process, there is a problem that it is very difficult to detect the end point at an arbitrary film thickness in the layer. This means that it is difficult to match the determination of the end point detection with the ideal polishing end point (near the vicinity), and improving the detection accuracy of the polishing end point is effective management in the CMP process and of the CMP apparatus. It has become an important technical issue in improving throughput.

  The present invention has been made in view of such a problem, and provides an end point detection device capable of determining a polishing end point with high accuracy during polishing, and a polishing device using the end point detection device. For the purpose.

  In order to achieve such an object, an end point detection apparatus according to the first aspect of the present invention is an end point detection apparatus that determines the end of polishing processing in polishing of an object to be polished, and is provided on a surface to be polished of the object to be polished. An illumination unit that irradiates probe light, a light detection unit that detects light from a surface to be polished irradiated with probe light as a plurality of spectra for each wavelength region, and before and after the scheduled polishing end time calculated according to predetermined conditions A condition setting unit in which an allowable time zone determined with a certain time width is set, and at least from a plurality of spectral wavelength ranges detected by the light detection unit by polishing the first object to be polished in advance. Select a plurality of combinations of two or more spectral wavelength ranges, synthesize each light intensity signal of two or more spectral ranges, and calculate a combined signal, respectively, from among a plurality of combinations of two or more wavelength ranges Synthetic signal output that occurs in the allowable time zone A detection condition determination unit that determines a combination of two or more specific wavelength ranges that represent feature points whose values match a predetermined condition, and is performed after the polishing of the first workpiece. During polishing processing of other objects to be polished, the polishing end point is detected when the composite signal obtained based on the spectrum of two or more specific wavelength ranges detected by the light detection unit represents the feature point, and polishing processing is performed. And an end point determination unit that determines the end of the process.

  An end point detection apparatus according to a second aspect of the present invention is an end point detection apparatus that determines the end of polishing processing in polishing of an object to be polished, and illuminates a surface of the object to be polished with probe light. A light detection unit that detects light from the surface to be polished irradiated with the probe light as a plurality of spectra for each wavelength region, and a fixed time width before and after the scheduled polishing end time calculated according to a predetermined condition. Based on a condition setting unit in which an allowable time zone is determined and a light intensity signal for each wavelength region of a plurality of spectra detected by the light detection unit by polishing the first object to be polished in advance. A detection condition determining unit that determines a specific wavelength range representing a feature point such that an output value of a light intensity signal generated in an allowable time zone from a plurality of wavelength ranges matches a predetermined condition, Others performed after the polishing of 1 An end point determination unit that detects a polishing end point and determines the end of the polishing process when a light intensity signal of a spectrum in a specific wavelength range detected by the light detection unit represents a feature point during polishing of the workpiece; It is configured with.

  Furthermore, the polishing apparatus according to the present invention includes a holding mechanism that holds an object to be polished, and a polishing member that polishes the object to be polished held by the holding mechanism, and the object to be polished held by the holding mechanism is polished. A polishing apparatus configured to perform polishing processing of an object to be polished by moving the polishing member while abutting a polishing member on a surface, further comprising an end point detection apparatus having the above-described configuration. .

  According to the present invention, it is possible to determine the polishing end point with high accuracy during the polishing process, thereby efficiently managing the polishing process and improving the throughput of the polishing apparatus.

It is the schematic which shows the CMP apparatus which is an example of the grinding | polishing apparatus which concerns on this embodiment. It is the schematic which shows the structure of an optical EPD part. It is a graph which shows the relationship between polishing time and the intensity | strength of reflected light, (a) is when an initial film thickness is thick, (b) is when an initial film thickness is thin. It is explanatory drawing which shows the relationship between prediction grinding | polishing time and permissible time. It is a block diagram which shows the structure of a control apparatus. It is a flowchart which shows the control flow in the case of calculating | requiring the optimal single wavelength range in a control apparatus. It is a flowchart which shows the control flow in the case of calculating | requiring the optimal combination of a some wavelength range in a control apparatus. 1 is a schematic view of a laminated structure of a semiconductor wafer used in Example 1. FIG. 6 is a graph showing temporal changes in spectral intensity at observation wavelengths of 450 and 750 nm in Example 1. FIG. 6 is a graph showing temporal changes in spectral intensity in the optimum wavelength region (553 nm) in Example 1. 6 is a graph showing temporal changes in spectral intensity in an optimum wavelength region (648 nm) in Example 1. It is a figure which shows the screen displayed on an input / output device in Example 1. FIG. 6 is a graph showing a temporal change in spectral intensity in a combination of optimum wavelength ranges in Example 2. It is a graph which shows the time change of the spectral intensity at the time of changing polishing pressure in Example 2. FIG. It is a graph which shows the time change of the spectral intensity in the combination of the single wavelength range in Example 3, and a several wavelength range.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings. Here, as an example, a case where a polishing end point is determined in a CMP process (STI process or ILD (interlayer insulating film) process) for polishing a silicon oxide film will be described in detail. First, FIG. 1 shows a schematic configuration of a CMP apparatus (chemical mechanical polishing apparatus) 1 which is a typical example of a polishing apparatus to which the present invention is applied.

  The CMP apparatus 1 includes a holding mechanism 10 that rotatably holds a semiconductor wafer (hereinafter referred to as a wafer) W on the upper surface side, and a polishing pad 21 that is provided above the holding mechanism 10 and that is provided on the lower surface side. A polishing head 20, a slurry supply unit 30 for supplying a slurry (abrasive) 31 from the center of the polishing pad 21, an EPD unit (optical measurement unit) 40 for detecting an end point, and a controller 60 for controlling these operations. It is mainly composed.

  The holding mechanism 10 includes a disk-shaped wafer chuck 11 and a spindle 12 that supports the wafer chuck 11 and extends vertically downward. The holding mechanism 10 rotates at right angles to the upper surface of the wafer chuck 11 by a rotation driving mechanism (not shown). The wafer chuck 11 is rotationally driven in the horizontal plane around the central axis O1. Inside the wafer chuck 11, a vacuum chuck structure that vacuum-sucks the lower surface (held surface) of the wafer W is provided so that the wafer W can be attached and detached, and the wafer W held by the wafer chuck 11 is sucked and held. The surface to be polished (that is, the surface to be polished) is held in an upward horizontal posture.

  The polishing head 20 includes a head member 22 to which a polishing pad 21 is attached on the lower surface, and a spindle 23 that supports the head member 22 and extends vertically upward, and is orthogonal to the lower surface of the head member 22 by a rotation drive mechanism (not shown). The polishing pad 21 is driven to rotate in a horizontal plane around the rotation center axis O2. The polishing pad 21 is formed in an annular shape whose outer diameter is smaller than the outer diameter of the wafer W to be polished. For example, the polishing pad 21 is formed using a rigid polyurethane sheet having an independent foam structure, and the lower surface of the head member 22. The polished surface is held in a downward horizontal posture. The polishing head 20 is configured to be horizontally swingable and vertically movable with respect to the wafer chuck 11 using a head moving mechanism (not shown).

  The slurry supply unit 30 passes the slurry 31 to a contact portion between the polishing pad 21 and the wafer W from a hole (not shown) formed through a pipe 32 in the center of the polishing pad 21 and the head member 22 vertically. Supply.

  In order to polish the wafer W, the wafer chuck 11 and the head member 22 are respectively rotated in the same direction or in the opposite direction, and the slurry 31 is supplied from the center (hole) of the polishing pad 21 by the slurry supply unit 30. While the polishing head 20 is lowered and the polishing pad 21 is brought into contact with the surface to be polished of the wafer W, the polishing head 20 is reciprocated (horizontal) in the radial direction between the center portion and the outer peripheral portion of the wafer W in this state. Let As a result, the surface to be polished of the wafer W is subjected to the mechanical and chemical polishing action of the slurry 31 interposed between the polishing pad 21 and the surface to be polished.

  The EPD unit 40 measures the spectral intensity of light in order to determine whether or not the surface to be polished of the wafer W has been polished by a predetermined amount, that is, to determine the polishing end point. The EPD unit 40 includes a light source 41, an optical path dividing member 45, a photodetector 54, and the like. Since the EPD unit 40 is disposed so as to be opposed to the upper side of the wafer W that is an object to be polished, the EPD unit 40 is accommodated in the cover 55 in order to protect the device from the slurry 31 scattered during the polishing process. In addition, the EPD unit 40 accommodated in the cover 55 uses an EPD swing mechanism (not shown) from a measurement position located above the wafer W (wafer chuck 11) and from above the wafer W (wafer chuck 11). It is configured to be able to swing back and forth between the retracted position. The EPD unit 40 reflects the light beam emitted from the light source 41 on the surface to be polished of the wafer W, takes out the reflected light by the optical path dividing member 45, and splits the light beam by a spectroscope or the like to be described later with a photodetector 54. The spectral intensity is measured.

  Here, the EPD unit 40 will be described in more detail with reference to FIG. The light source 41 is a white light source having a multi-wavelength component, and a white LED, a xenon lamp, or the like can be used. The light beam emitted from the light source 41 is converted into a parallel light beam by the lens 42, passes through the illumination area control slit 43, and then collected by the lens 44 onto a beam splitter (light path dividing member) 45. The light beam that has passed through the beam splitter 45 is converted into a parallel light beam again by the lens 46 and is irradiated onto the surface (surface to be polished) of the wafer W to form a beam spot having a predetermined spot diameter on the wafer surface. The reflected light reflected by the surface to be polished of the wafer W is condensed again on the beam splitter 45 through the lens 46. Reflected light from the wafer W is changed by 90 ° in the beam splitter 45 and is converted into a parallel light beam by the lens 48. Then, the light is reflected by the mirror 49 and condensed by the lens 40 on the slit 51 which is a light shielding means having an opening for selecting only the 0th order light (regular reflection light). Noise components such as scattered light and diffracted light are removed by the slit 51, and the light is projected onto the diffraction grating 53, which is a spectroscopic means, through the lens 52 and dispersed. The light beam dispersed by the diffraction grating 53 enters a photodetector (linear sensor) 54 that is a photoelectric conversion element, and the spectral intensity is measured.

  In this optical system, the image of the light source 41 is formed on the beam splitter 45, and the image formation position is located at the front focal point of the lens 46. Therefore, the surface to be polished of the wafer W is irradiated with a uniform and substantially vertical parallel light beam. Further, the slit 51 is provided at a condensing position of vertically reflected light from the wafer W. Therefore, by adjusting the diameter of the opening, the numerical aperture (NA) of the reflected light, that is, the reflection angle of the light beam passing through the opening can be adjusted. Therefore, by selecting the numerical aperture so that the first-order or higher-order diffracted light generated from the assumed pattern on the polished surface of the wafer W cannot pass through the opening of the slit 51, zero-order light (regular reflection light) The other reflected light is shielded by the slit 51, and only the regular reflected light enters the diffraction grating 53, and is diffracted at a diffraction angle corresponding to the wavelength.

  Since the specularly reflected light incident on the photodetector 54 is diffracted and dispersed in accordance with the wavelength by the diffraction grating 53, the photodetector 54 detects the spectral intensity for each wavelength component in the regular reflected light, and this spectral intensity information. Is input to the control device 60.

  Thus, the spectral intensity for each wavelength component of the reflected light detected by the photodetector 54, for example, the spectral intensity at an arbitrary wavelength λ detected with respect to the reflected light of the interlayer insulating film (ILD) which is a transparent film, The fluctuation characteristics as shown in FIG. 3 are shown. At this time, assuming that the refractive index of the film to be polished is n, a maximum point appears as shown in FIG. 3 every time the polishing process proceeds and the film thickness changes approximately λ / 2n. In an algorithm that automatically tracks the spectral intensity and knows the film thickness variation by the number of occurrences of the maximum point, the maximum number of points is required to reach the target film thickness in the time variation of the reflected light intensity (reflectance). Whether the polishing end point is determined is calculated in a timely manner, and the polishing end point is determined by counting the number of appearances of the maximum point. The variation characteristic of the reflected light intensity is the same even when the minimum point of the reflected light intensity is measured. In the present embodiment, as shown in FIG. 4, the control device 60 described later uses a polishing time until a target film thickness calculated from a predetermined polishing rate or the like (in the following description, “predicted polishing time”). And an allowable time ΔT (for example, ± 5% of the predicted polishing time Te) determined with a certain width before and after the predicted polishing time Te, and a maximum and minimum point ( It is configured to determine an optimum value of a parameter (a specific wavelength region for determining the end point or a combination thereof) such that a determination point) appears.

  As shown in FIG. 5, the control device 60 is configured to hold the holding mechanism 10, the polishing head 20, the slurry supply unit 30, and the EPD unit 40 based on a control program that is set and stored in advance and a processing program that is read according to the polishing target. Wavelength components input from the operation control unit 61 that controls the operation of the above, an initial condition setting unit 62 that sets initial conditions such as a predetermined algorithm for determining the polishing end point, and the EPD unit 40 (photodetector 54) Based on the setting information from the initial stage setting unit 62, the spectral intensity information stored in the storage unit 63, and the like, parameters for determining the polishing end point (end point detection condition, That is, a detection condition determining unit 64 that determines an optimum value of a specific wavelength range for determining the end point or a combination thereof, or the count value of the maximum and minimum points serving as a determination point; Constructed and a end point determination unit 65 to detect the polishing endpoint of the wafer W based such as parameters determined by the detection condition determining unit 64 controls the PD unit 40.

  Now, with regard to the parameters for determining the polishing end point, the procedure for determining the optimum value using the control device 60 will be described with reference to the flowcharts shown in FIGS. In this procedure, two cases will be described, in which a single wavelength region is used as the parameter and a combination of a plurality of wavelength regions is used.

  First, a predetermined initial condition is input using an input / output device 66 (a human interface including a keyboard for inputting keys of various conditions and a display for displaying an image) connected to the control device 60 (steps S101 and 201). The initial condition is set in the initial condition setting unit 62. As the initial condition, in addition to a predetermined algorithm for determining an end point, an estimated polishing time Te calculated based on a polishing amount until a target film thickness is reached and a predetermined polishing rate, and an estimated polishing time Te For example, an allowable time ΔT is set. In addition, from a number of channels ch (for example, 256 channels (ch1 to ch256)) possessed by the spectroscope (consisting of the diffraction grating 43 and the photodetector 54), the start wavelength (channel ch1) in the wavelength region to be observed. ) And the end wavelength (channel ch 256). When the light source 41 is a white LED, the general value is 450 nm for the start wavelength (channel ch1) and 750 nm for the end wavelength (channel ch256).

Here, with reference to FIG. 6, a description will be given of the case where the parameter for determining the end point among the wavelength components detected by the photodetector 54 is set to a single wavelength region. At this time, the detection condition determining unit 64 uses the spectral intensity information stored in the storage unit 63 to set the channel ch _A to be observed as the channel ch1 (starting wavelength 450 nm) as its initial value (step S102). Then, a smoothing process is performed as a pre-process on the signal waveform of the channel ch _A (step S103), and spectral intensity data D (0) and D (1) for each sampled time zone (sampling period interval). , D (2)... D (m) are acquired (step S104). Here, data D (0) is spectral intensity data when the sampling time T = 0, data D (1) is spectral intensity data when the sampling time T = 1, and data T (m) is sampling time T = m. Represents the spectral intensity data at.

  Then, only maximum values are extracted from these data D (0) to D (m) to create a peak value list, and only minimum values are extracted to create a bottom value list (step S105). The respective sampling times at which the maximum value and the minimum value appear are obtained from the peak value and bottom value lists, and the maximum value appearing within the allowable time ΔT before and after the predicted polishing time Te set in the initial condition setting unit 62, and Only the minimum value is extracted and temporarily stored in the storage unit 63 (step S106).

Next, channel ch _A is counted up, and steps S103 to S106 are performed for all channels (1 to 256 ch) of the spectroscope with respect to channel ch _A , that is, under the wavelength range of 450 nm to 750 nm (step S107). , S110). Therefore, the channel ch _A is counted up from 450 nm to 750 nm in increments of the wavelength range corresponding to the number of channels of the spectrometer, and thus appears within the allowable time ΔT in the wavelength range corresponding to the number of channels of the spectrometer. The maximum value and the minimum value to be extracted are extracted.

When the loop (steps S103 to S106) is completed for all the wavelength ranges, the maximum value appearing within the allowable time ΔT from the peak value list for each wavelength range is compared with the minimum value appearing immediately before it, or the bottom From the value list, the local minimum value that appears within the allowable time ΔT is compared with the local maximum value that appears immediately before. Then, as a result of the comparison, a channel ch _A (specific wavelength range) in which the difference between the local maximum value and the local minimum value just before it, or the difference between the local minimum value and the local maximum value just before the maximum value is obtained (step S108). . This is because, as the difference between the maximum value and the minimum value increases, the fluctuation of the spectral intensity that should be captured for the end point determination becomes more prominent, so that stable detection can be performed regardless of the influence of noise elements. Thus, when priority is given to the S / N in the end point determination, the channel ch _A (wavelength range) in which the difference between the maximum value and the minimum value is the maximum is the optimum parameter value for the end point determination. At the same time, the count number of determination points (maximum / minimum points) appearing within the allowable time ΔT in the waveform of the spectral intensity of the channel ch _A set to the optimum value is also calculated (step S109).

On the other hand, when priority is given to the target time (predicted polishing time Te), instead of step S108, a channel ch _A (a specific wavelength region) in which the maximum and minimum points appear at the point closest to the predicted polishing time Te. Is determined as the optimum parameter value for determining the end point (step S108 ').

Next, a case where the end point determination parameter is a combination of a plurality of wavelength ranges will be described with reference to FIG. At this time, by using the spectral intensity information stored in the storage unit 63 by the detection condition determination unit 64, the channel ch1 (starting wavelength 450 nm) and the channel are set with the two channels ch _A and ch _B to be observed as their initial values. ch256 (end wavelength 750 nm) is set (step S202), and the absolute value of the time series signal based on the channel ch _A (the wavelength range of ch1) and the absolute value of the time series signal based on the channel ch _B (the wavelength range of ch256) _A signal waveform (composite waveform) based on a ratio to the value (time series signal of ch _A / time series signal of ch _B ) is calculated (step S203). This is because the signal waveform of each spectral component includes many noise elements caused by scattering due to fluctuations in the slurry 32, and therefore, each spectral intensity measured under the same condition is calculated as a ratio. This is because the noise element can be eliminated, and the S / N can be improved as compared with the case where the above-mentioned parameter is measured as a single wavelength.

  Then, a smoothing process is performed as a pre-process on the synthesized waveform (step S204), and spectral intensity data D (0), D (1), D (for each sampled time zone (sampling period interval). 2)... D (m) is acquired (step S205).

  Then, only maximum values are extracted from these data D (0) to D (m) to create a peak value list, and only minimum values are extracted to create a bottom value list (step S206). At this time, respective time zones in which the maximum value and the minimum value appear are obtained from the peak value and bottom value lists, and appear within the allowable time ΔT before and after the predicted polishing time Te set in the initial condition setting unit 62 from among them. Only the local maximum value and the local minimum value are extracted and temporarily stored in the storage unit 63 (step S207).

Next, the channel ch1 is counted up and the channel ch2 is counted down (steps S208 and 211), and steps S203 to S207 are performed under the combination of all channels (1 to 256 ch) for the channels ch _A and ch _B. Done. That is, channel chA is counted up from 450 nm to 750 nm in increments of wavelengths according to the number of channels of the spectrometer, and conversely, channel chB is counted down from 750 nm to 450 nm in increments of wavelengths according to the number of channels of the spectrometer. Then, the maximum value and the minimum value appearing within the allowable time ΔT under the combination of the wavelength ranges corresponding to the number of channels of the spectroscope are extracted.

Then, when the loop (steps S203 to 207) is completed for all combinations of wavelength ranges, for each combination of wavelength ranges, the maximum value that appears within the allowable time ΔT is compared with the minimum value that appears immediately before it, Alternatively, the local minimum value that appears within the allowable time ΔT is compared with the local maximum value that appears just before it. Then, as a result of the comparison, a combination of channels ch _A and ch _B in which the difference between the local maximum value and the local minimum value just before or the difference between the local minimum value and the local maximum value just before the maximum value is obtained (step S209). . When priority is given to S / N in the end point detection, the combination of the channels ch _A and ch _B (its wavelength range) becomes the optimum parameter value for end point determination. At the same time, a predetermined count number of determination points (maximum / minimum points) appearing within the allowable time ΔT in the waveform of the light intensity of the combination of the channels ch _A and ch _B (wavelength ranges thereof) is also calculated (step S210). .

On the other hand, when priority is given to the target time (predicted polishing time Te), instead of step S209, the wavelengths of channels ch _A and ch _B (where the maximum and minimum points appear at the point closest to the predicted polishing time Te). (Region) is determined, and this is set as the optimum parameter value for end point detection (step S209 ').

  Then, in the polishing process, the EPD unit 40 for in-situ measurement is operated, and the maximum or minimum point of a predetermined count number is determined based on the wavelength range determined as described above or the optimum value of a combination of a plurality of wavelength ranges. When the end point determination unit 65 outputs the end point detection signal with the detection, the polishing process of the wafer W is completed at the detection timing of the polishing end point.

  As described above, the parameter for determining the end point (determination point) is set so that the maximum / minimum point (determination point) appears in the time region (allowable time ΔT) to be determined as the end point using the spectral signal for each wavelength component acquired in advance. By optimizing the wavelength range, it is possible to accurately grasp the change in the surface state on the surface to be polished of the wafer W, and the end point detection accuracy can be improved. Therefore, it is possible to efficiently manage the CMP process and improve the throughput of the CMP apparatus 1.

  In the present embodiment, the end point detection is determined based on the appearance of the maximum / minimum value. However, the end point detection is determined based on the occurrence of an inflection point or the like when a predetermined threshold is exceeded. May be.

  The CMP process of the wafer W in the CMP apparatus 1 configured as described above will be described. First, after the wafer W is transferred onto the wafer chuck 11 by a transfer device (not shown) and before the polishing process is started, the control device 60 uses the EPD unit swing mechanism (not shown) to move the EPD unit 40 to the above-described retracted position. To the measurement position above the wafer W. The holding mechanism 10, the polishing head 20, the slurry supply unit 30, etc. are actuated to start polishing, and the probe light is irradiated from the light source 41 of the EPD unit 40 to measure the reflected light from the wafer W by the photodetector 54. (In other words, measurement by the EPD unit 40) is started. At this time, as described above, the control device 60 becomes an observation target among the reflected light detected by the photodetector 54 based on the preset predicted polishing time, allowable time, and end point detection condition (optimum). The number of occurrences of the maximum and minimum points of this spectral intensity is counted while sampling the spectral intensity (depending on the specific wavelength range or combination thereof), and is the end point of the polishing process when the maximum and minimum points appear a predetermined number of times. Is determined. When the end point detection signal is output from the end point determination unit 65, the polishing process is controlled at that timing.

  At this time, the waveform of the spectral signal sampled by the EPD unit 40 is always subjected to smoothing processing (smoothing processing) by the control device 60 for noise reduction. And the points before and after the predetermined time. Therefore, in order to perform the smoothing process at the target point, the data of the points sampled after a certain time is also required. Therefore, when performing the smoothing process by detecting the spectral component of the reflected light in real time, As a result, processing delay occurs for a certain period of time, and as a result, the end point detection is delayed, and there is a risk that the target film thickness is greatly exceeded. For example, in the case of a smoothing process at 5 points (a total of 5 points including a target point to be smoothed and two points before and after that) in a waveform of a spectral signal sampled at intervals of t seconds, the target point is corrected ( Smoothing is performed 2t seconds after the point is sampled (at the time when the latter two points are measured), and therefore, a delay occurs after 2t seconds. Therefore, as an optimal parameter value, instead of obtaining a wavelength region (or a combination thereof) in which a local minimum point appears in the vicinity of the predicted polishing time Te, a wavelength region in which a local minimum point appears in the vicinity of 2 tsec before the predicted polishing time Te. (Or a combination thereof) may be obtained. According to this, when the smoothing process is performed on the spectral waveform with the optimum value, the maximum and minimum points appearing in the vicinity of 2 t seconds before the predicted polishing time Te are detected in advance. Even when a delay of 2 tsec due to the smoothing process occurs, the timing for detecting the end point can be matched with the predicted polishing time Te. Therefore, the end point can be detected with higher accuracy without causing a delay due to the smoothing process.

  Hereinafter, the present invention will be described in more detail with reference to examples. However, the present invention is not limited to the following examples.

[Example 1]
First, in a case where a 12-inch silicon wafer (wafer W) subjected to pattern processing of the wafer structure (STI structure, ILD structure) shown in FIG. 8 is polished, the controller 60 sets an optimum value of the polishing end point determination parameter. A result of simulation calculation for calculation will be described. As shown in FIG. 8, a wafer W as an object to be polished has a SiO ₂ layer (silicon thermal oxide film) 71 formed on the upper surface of a silicon substrate 70 to a thickness of 10 nm, and a silicon nitride SiN layer as a barrier layer thereon. 72 is formed to a thickness of 60 nm, and a SiO ₂ layer 73 is formed thereon as a TEOS layer by plasma CVD to a thickness of 400 nm. Note that a trench 74 exists in the silicon substrate 70, and its depth is 200 nm.

  Here, assuming that the processing rate of the TEOS layer of the polishing head 20 is 10 nm / second from the existing data, and the target (end point) of polishing is when the uppermost TEOS layer 73 disappears, the predicted polishing time Te, The allowable time ΔT (for example, ± 5% of the predicted polishing time Te) is obtained by the following equation.

Te = 400/10 = 40 seconds ΔT = 40 × 0.05 = 2 seconds (± 2 seconds)
Therefore, in the case of this wafer W, it is predicted that the TEOS layer 73 disappears at a point of 40 seconds from the start of polishing, and the control device 60 sets 40 seconds ± 2 seconds (38 seconds to 42 seconds) as the polishing time. Is done. On the other hand, a white LED (wavelength range: 450 nm to 750 nm) is used as the light source 41 of the EPD unit 40.

  FIG. 9 and FIG. 10 show how the spectral intensity changes with time as a result of the simulation calculation performed by the control device 60 under such conditions. 9 and 10, the vertical axis indicates the spectral intensity, the lower horizontal axis indicates the polishing time elapsed from left to right, and the upper horizontal axis indicates the change in film thickness (residual film thickness) from left to right. Represents. As shown in FIG. 9 showing the temporal change in spectral intensity at an observation wavelength of 450 nm (starting wavelength) and a wavelength of 750 nm (end wavelength), the spectroscopic analysis at a wavelength of 450 nm and a wavelength of 750 nm has a polishing time of 40 seconds ± 2 seconds. Shows that the maximum and minimum points do not occur. On the other hand, a solution (parameter) in which a maximum / minimum point is generated at a point where the polishing time is 40 seconds ± 2 seconds using the control device 60 (detection condition determination unit 64) in the observation wavelength region of 450 nm to 750 nm by the procedure described above. Was determined at a single wavelength, and it was determined that the wavelength of 553 nm was the optimum value. As shown in FIG. 10 showing the temporal change in spectral intensity at a wavelength of 553 nm, it can be seen that a minimum point appears at a point of polishing time (predicted polishing time Te) of 40 seconds at a wavelength of 553 nm. Therefore, in this case, when the third minimum point is detected, it can be determined as the polishing end point.

  In the above case, the polishing end point is a position where the TEOS layer 73 is switched to the barrier layer 72. Next, in order to show that determination with an arbitrary film thickness is possible, an observation when a solution of a parameter (wavelength region) is obtained on the assumption that the film thickness of the TEOS layer 73 is 150 nm as a target of the polishing end point. The results are shown in FIG. At this time, the predicted polishing time Te = 25 seconds. As a result of obtaining this in a single wavelength region, it can be seen that the wavelength 648 nm is determined to be the optimum value, and that the polishing end point can be determined when the second maximum point is detected.

  FIG. 12 shows a screen displayed on the display of the input / output device 66 as a result of the simulation calculation performed by the control device 60 at this time. In the software screen of the input / output device 66, the above-described initial conditions (predetermined algorithm and the like) are set, and as a result of the simulation calculation, the optimum value of the parameter (wavelength) (“Monitored Wavelength = 647 in FIG. 12). .92 "), the maximum / minimum count number (" PeakBottom Counts = 2 "in FIG. 12), the spectral waveform at the optimum wavelength (648 nm) (upper right graph in FIG. 12), and each sampling time at an arbitrary sampling time Spectral intensity (lower right gragg in FIG. 12) and the like for the wavelength region (wavelength region 450 nm to 750 nm) are output (displayed).

[Example 2]
Next, FIG. 13 shows an observation result when a wafer W having an interlayer insulating film (ILD) structure is polished. For polishing the wafer W, an Oscar type CMP apparatus 1 having a polishing pad with a smaller diameter than the wafer W was used. Further, silica-based SS25 (dispersed with an alkali solvent) was used as the slurry, and polishing was performed at a low polishing pressure. The polishing rate was calculated from the data of Sample 1 shown in FIG. 13, and it was determined that the target predetermined film thickness was reached in the predicted polishing time Te = 170 seconds. In this example, when the white LED is used and the optimum value is obtained using a combination of a plurality of wavelength regions as parameters as the start wavelength of 450 nm and the end wavelength of 750 nm, the light of the combination of the wavelength of 549 nm and the wavelength of 694 nm is obtained. In the intensity waveform, the (second) local maximum point was required to appear in the vicinity of the predicted polishing time Te. Therefore, the combination of the wavelength 549 nm and the wavelength 694 nm is determined as the optimum parameter value, and the second (sample 2) and third (sample 3) wafers W are polished based on the end point detection conditions. It was. As shown in FIG. 13, also in the second and third wafer polishing, the second maximum point appears in the vicinity of the predicted polishing time Te (170 seconds), and this is determined as the end point of the polishing process. Thus, the polishing process can be completed at this timing, and it can be seen that the present system operates without any problem.

  Further, FIG. 14 shows an observation result when the polishing pressure is changed and the wafer W is polished in the above case. Here, in FIG. 14, sample 3 is performed at a low polishing pressure (same as the above case), sample 4 is performed at a medium polishing pressure, and sample 5 is performed at a high polishing pressure. As the polishing pressure is increased from low pressure to medium pressure and high pressure, the polishing rate is improved and the predicted polishing time Te is shortened, but if polishing is performed while maintaining other conditions other than the polishing pressure, It can be seen that the calculated optimum value of the parameter and the count number of determination points (maximum / minimum points) are substantially the same. That is, the predicted polishing time at the point where the second maximum point appears under the waveform of the light intensity based on the combination of the wavelength 549 nm and the wavelength 694 nm regardless of whether the polishing pressure is low, medium, or high. It can be seen that Te (Te3, Te4, Te5) is reached. Thus, the film thicknesses of samples 4 and 5 whose end points were detected were substantially equal to the film thickness of sample 3. As described above, even when the polishing pressure is changed, the count value of the maximum and minimum points for determining the polishing end point does not change, so when the polishing pressure is changed as the polishing of the plurality of wafers W progresses, or Even if the polishing pressure varies slightly with the progress of polishing, the end point can be detected with high accuracy when the maximum and minimum points have reached a predetermined count.

[Example 3]
Next, FIG. 15 shows an observation result when the wafer W having the STI structure is polished. For polishing the wafer W, an Oscar type CMP apparatus 1 having a polishing pad with a smaller diameter than the wafer W was used. Also, ceria type was used for the slurry. Since ceria-based slurry is likely to be scattered with respect to silica-based slurry, a large noise element is likely to be generated, and there is a risk of lowering the S / N. At this time, the time change of the spectral intensity is shown in FIG. 15 for a case where a single wavelength region is used as a parameter for determining the end point and a case where a plurality of wavelength regions are used in combination. Here, it is assumed that the optimum value in a single wavelength region is determined to be 519 nm, and the optimum combination of a plurality of wavelength regions is determined to be 641 nm and 733 nm. When a single wavelength region (519 nm) is used as an end point determination parameter, the S / N is poor and the detection accuracy of reflected light is reduced. On the other hand, when a combination of a plurality of wavelength ranges (641 nm, 733 nm) is used as a parameter, the noise element can be excluded from the signal waveform as described above, and the S / S is higher than when a single wavelength range is used. N can be improved. For this reason, it can be seen that the end point can be detected with high accuracy regardless of the influence of the noise element by obtaining the optimum value using a plurality of wavelength regions.

  In the above-described embodiment, as shown in FIG. 1, the holding mechanism 10 that holds the wafer W in an upward horizontal posture, and the polishing pad 21 that faces the lower side of the holding mechanism 10 and faces downward is provided on the lower surface side. Although the CMP apparatus 1 including the polishing head 20 mounted on the wafer has been described, the present invention is not limited to this, and the present invention provides a mechanism for holding the wafer W in a downward horizontal posture and a lower part of the holding mechanism. It can also be used in a CMP apparatus (so-called conventional type) provided with a polishing member that is provided oppositely and on which the polishing pad is mounted upward on the upper surface side.

  In the above-described embodiment, the case where the silicon oxide film is polished (STI process, ILD process) has been described. However, the present invention is not limited to this, and the same effect can be obtained even when used for polishing a metal film or the like. Have.

  Furthermore, in the above-described embodiment, the object to be polished is not limited to the semiconductor wafer W. For example, the present invention can be applied to a case of a glass substrate or the like.

DESCRIPTION OF SYMBOLS 1 CMP apparatus (polishing apparatus) 10 Holding mechanism 21 Polishing pad (polishing member) 40 EPD part (end point detection apparatus)
41 Light source (illumination unit) 54 Photodetector (light detection unit)
60 Control Device (End Point Detection Device) 62 Initial Condition Setting Unit (Condition Setting Unit)
64 Detection condition determination unit 65 End point determination unit W Semiconductor wafer (object to be polished) Te Predicted polishing time (polishing end time)
ΔT Allowable time (allowable time zone)

Claims (1)

In the polishing process of an object to be polished, an end point detection device for determining the end of the polishing process,
An illumination unit for irradiating a surface of the object to be polished with probe light;
A light detection unit that detects light from the surface to be polished irradiated with the probe light as a plurality of spectra for each wavelength region;
A condition setting unit in which an allowable time zone determined with a certain time width is set before and after the scheduled polishing end time calculated according to a predetermined condition;
A plurality of combinations of at least two or more spectral wavelength regions are selected from the plurality of spectral wavelength regions detected by the light detection unit by polishing the first object in advance, and the two or more wavelengths are selected. A combined signal is calculated by synthesizing each light intensity signal of spectrum in the region, and an output value of the combined signal generated in the allowable time zone from among a plurality of combinations of the two or more wavelength regions is predetermined. A detection condition determination unit that determines a combination of two or more specific wavelength ranges that represent feature points that match the conditions of
Obtained based on the spectrum of the two or more specific wavelength ranges detected by the light detection unit during polishing of another object to be polished performed after polishing of the first object to be polished An end point detection apparatus comprising: an end point determining unit that detects a polishing end point when a combined signal represents the feature point and determines the end of the polishing process.

Images