JP5425494B2 - Improvement of defect detection system - Google Patents

Description

  The present invention relates generally to defect detection, and more particularly to detecting surface anomalies such as particles, as well as defects due to surfaces such as, for example, crystal-induced particles (COP), surface roughness, microscratches, etc. It relates to the improvement of the system.

The SP1 ^TBI ™ detection system available from KLA-Tenker Corporation of San Jose, Calif., The assignee of this application, is particularly effective in detecting defects in semiconductor wafers that are not patterned. is there. The SP1 ^TBI system is superior to defects when inspecting on a bare wafer or an unpatterned wafer than when used to inspect a wafer having a pattern on its surface, such as a wafer having a memory array. Exhibits sensitivity. In this system, all light is collected by a lens or elliptical mirror and directed to a detector to produce a single output. Thus, if these patterns are collected and sent to a detector in order for the pattern on the wafer to generate Fourier and / or other strong scattered signals, the output of the signal detector will saturate and Therefore, it becomes impossible to provide effective information for detecting a defect of the device.

  Although conventional techniques for detecting defects on a wafer are configured to fit either for inspection of patterned wafers, or for inspection of non-patterned wafers or bare wafers Not suitable for doing both. Although a detection system for patterned wafers can be used to inspect non-patterned wafers, this system is generally not optimal for that purpose. On the other hand, systems designed for inspecting unpatterned wafers or bare wafers handle diffraction and other scattering caused by pattern structures on patterned wafers for the reasons described above. Is difficult.

  Inspecting a patterned wafer has used a completely different inspection system. A commercially available system, known as the AIT ™ inspection system, is available from KLA-Tenker Corporation of San Jose, California, the assignee of the present application. This system is described in various patents, such as US Pat. No. 5,864,394. In the AIT system, a spatial filter is employed to shield the detector from diffraction and scattering due to the pattern structure on the wafer. The design of this spatial filter may be based on conventional knowledge about the pattern structure, but can be very complex. In addition, the system employs a die comparison process to more accurately identify the presence or absence of defects.

US Pat. No. 5,864,394 US patent application Ser. No. 08 / 770,491 US Pat. No. 6,201,601

  None of the above devices were completely satisfactory in inspecting the wafer on which the pattern was formed. Therefore, it is desired to improve a defect detection system that can alleviate the above-described problem with respect to a wafer on which a pattern is formed. Furthermore, in order to save the space required for in-line inspection, it is desired to provide an apparatus that is optimal for inspection of both a wafer having a pattern formed thereon and a wafer having no pattern formed thereon.

  Chemical mechanical planarization (CMP) technology is widely accepted in the semiconductor industry. However, even in the CMP process, if the defect cannot be properly controlled, various defects are generated that can have a large impact on the production amount of the integrated circuit (IC). In the CMP failure, the micro scratch has a great influence on the IC production amount. Therefore, it is desirable to be able to detect microscratches and other CMP defects and differentiate them from particles.

  An important parameter for monitoring the quality of unpatterned or bare films on a silicon wafer is surface roughness. The roughness of the surface is generally determined by the HRP (registered trademark) equipment of the KLA-Tenker Corporation, the assignee of the present application, or other equipment such as an atomic force microscope or other types such as scanning tunneling microscopes. Measured by a scanning probe microscope. One drawback of such devices is that their operating speed is slow. Therefore, it would be desirable to provide an alternative system that can measure surface roughness at a significantly faster rate than the equipment described above.

One aspect of the present invention is based on the viewpoint that the concentrator in the SP1 ^TBI instrument stores information on the azimuth angle of scattered light from the surface to be inspected. Therefore, if the focused path is separated by guiding the scattered light focused by the type of concentrator used in the SP1 ^TBI device into different azimuth angles, the above-mentioned problem is solved and the device forms a pattern. It can be configured to be optimal for wafer detection. Thereby, the small apparatus which can measure the defect of the wafer in which the pattern was formed can be obtained. In addition to the elliptical mirror used in the SP1 ^TBI instrument, other azimuthally symmetric concentrators may be used, such as a parabolic mirror used with one or more lenses.

Like the SP1 ^TBI system, a surface inspection system according to one aspect of the present invention focuses light scattered from a surface by a concentrator for focusing scattered light substantially symmetrically with respect to a line perpendicular to the surface. To do. By focusing the scattered light in different azimuth angles or in different directions with respect to the vertical line and directing it to different paths, these paths relate to the scattered light at a position corresponding to the relative azimuth of the scattered light. Can carry information. Preferably, these paths are separated by a separator to reduce crosstalk. Focused scattered light carried through at least a portion of the path can be used to determine the presence or character of anomalies within or on the surface. Furthermore, since the same event can be considered from multiple directions, the real-time defect classification (RTDC) process can be greatly simplified.

In the above-described mechanism, it leads to only a portion of the focused beam to different routes, by directing the remaining portion of the focused beam at different azimuth angles in a single detector, single as in the conventional SP1 ^TBI works Can be used to inspect both unpatterned and patterned wafers using the system. In other words, if the SP1 ^TBI mechanism is changed by redirecting a part of the light beam focused by the above-described method to a different path while preserving the azimuth information, the wafer having no pattern and the pattern can be changed. A multi-purpose tool capable of optimally inspecting both formed wafers is realized. In this way, the semiconductor manufacturer does not need to prepare an optimum tool for detecting a wafer on which a pattern is formed and a wafer on which a pattern is not formed separately.

  In the mechanism described above, rays focused at different azimuth angles with respect to a line perpendicular to the surface are directed to different focusing paths and converted to another signal, thus discarding the signal including pattern diffraction. Then, the remaining signals that do not include pattern scattering can be used to detect and classify anomalies within or on the surface of the wafer. The system described above is particularly useful for inspection of semiconductor wafers, but is also used for inspection of abnormalities in other surfaces and other applications such as flat panel display systems, magnetic heads, magnetic and optical recording media, etc. can do.

  Another aspect of the invention is focusing by a concentrator (as described above) using a spatial filter having an angled gap that corresponds to the angular difference of the ray components expected to be scattered by the pattern on the surface. It is based on the viewpoint that it can filter the light beam. Thus, filtered rays at several relative positions with respect to the filter on the surface contain information about the surface defects exposed by the scattering of patterns that can interfere with the measurement. By detecting such a light beam with a detector, the output of the detector can be used to detect the presence and / or characteristics of an abnormality within or on the surface.

Using the SP1 ^TBI tool or the system described above, particles from CMP and microscratches can be distinguished. Scattered light along a direction close to the vertical direction is focused by the first detector, and scattered light along a direction away from the vertical direction is focused by the second detector. By determining the ratio from the outputs of the two detectors, it is determined whether the surface abnormality is a micro scratch or a particle.

  CMP microscratches tend to scatter forward light from obliquely incident light. On the other hand, the particles scatter such light rays more uniformly. Light rays scattered in the forward scattering direction by the surface are focused separately from the scattered light in the other scattering directions. Two different signals are derived from the separately focused scattered light and compared to determine whether the anomalies on the surface are micro scratches or particles.

  In another aspect of the present invention, S-polarized light and P-polarized light are incident obliquely on the surface in order during two different types of surface scanning. Light rays scattered by defects during the first and second scans are focused and output a pair of signals representing scattered light of two different incident polarizations. The pair of signals is compared to a reference to determine if the anomaly on the surface is a microscratch or a particle.

In order to perform the process of determining the surface roughness of the thin film more quickly, a database that associates the surface roughness of the thin film with the haze value is created. The surface haze value is measured with a tool such as SP1 ^TBI or one of the systems described above, and the surface roughness value can be determined from the measured haze value and database. For example, a database may be a tool such as SP1 ^TBI for measuring haze values of a typical thin film or one of the systems described above and an HRP (registered) for measuring the surface roughness of such films. Collected by a trademark profiler or other type of profilometer or scanning probe microscope.

  Aspects of the invention as described above can be used individually or in combination to achieve the advantages described herein.

  In addition, in order to make the explanation easy to understand, the same parts are denoted by the same reference numerals in the present application.

1 is a schematic diagram of an SP1 ^TBI ™ system for illustrating the present invention. 1 is a schematic diagram illustrating a converging hollow cone of light to illustrate one aspect of the present invention. 2 is a schematic diagram illustrating a possible arrangement of multi-fiber paths for carrying scattered light focused by an elliptical concentrator of the system of FIG. 1 to illustrate one aspect of the present invention. 3B is a schematic illustration of a multi-anode phototube (PMT) that can be used in connection with the placement of a multi-fiber path as shown in FIG. 3A to illustrate one aspect of the present invention. 2 is a schematic diagram of a fiber path / multiple detector arrangement for carrying scattered light focused by a lens concentrator in a narrow angle path of the system of FIG. 1 to illustrate one aspect of the present invention. 1 is a cross-sectional view of a defect inspection system for illustrating a preferred embodiment of the present invention. It is sectional drawing which shows arrangement | positioning of the separate optical path used in embodiment shown to FIG. 5A. It is sectional drawing of the defect inspection system for demonstrating another embodiment of this invention. FIG. 6B is a cross-sectional view showing the arrangement of the divided optical paths used in the embodiment shown in FIG. 6A. FIG. 6 is a top view of a portion of a defect inspection system to illustrate another alternative embodiment of the present invention. 8 is a schematic diagram of a multi-element detector in the embodiment of FIG. 8 is a schematic diagram of a multi-element detector for use in the embodiment of FIG. FIG. 6 is a partial cross-sectional partial perspective view of a defect inspection system for illustrating yet another alternative embodiment of the present invention. 9B is a schematic diagram of a filter wheel useful in the embodiment shown in FIG. 9A. 9B is a schematic diagram of a filter wheel useful in the embodiment shown in FIG. 9A. 1 is a schematic diagram illustrating a two-dimensional diffraction component from a pattern on a surface to be inspected to illustrate one aspect of the present invention. 6 is a schematic diagram of a failure detection system for illustrating yet another alternative embodiment of the present invention. 2 is a schematic illustration of an asymmetric mask for use in different embodiments of the present invention. 2 is a schematic diagram of two masks used in different systems of the present application to illustrate another aspect of the present invention. 2 is a schematic diagram of two masks used in different systems of the present application to illustrate another aspect of the present invention. It is a graph which shows the interference intensity | strength of the surface of a thin film when irradiated with the light of three different polarization | polarized-light for illustrating another situation of this invention. It is a graph of the haze and surface roughness for illustrating another situation of this invention. FIG. 16 is a block diagram showing a system for measuring surface roughness and haze of a representative film for collecting the database useful in the present invention shown in FIG. 15.

FIG. 1 is a schematic diagram of an SP1 ^TBI ™ system 10 available from the KLA-Tenker Corporation of San Jose, California, the assignee of the present application. Aspects of the SP1 ^TBI system 10 are described in US patent application Ser. No. 08 / 770,491 filed on Dec. 20, 1996 and US Pat. No. 6,201,601. Yes. These patent applications and patents are hereby expressly incorporated by this reference. To simplify the drawing, some of the optical elements of the system, such as components that direct illumination light to the wafer, are omitted. The wafer 20 to be inspected is irradiated with normal incident light 22 and / or inclined incident light 24. The wafer 20 is supported on a chuck 26 that is rotated by a motor 28 and redirected by a gear 30, and light 22 and / or 24 travels through a helical path on the surface of the wafer 20 to inspect the surface of the wafer; The area or spot 20a to be tracked is illuminated. The motor 28 and gear 30 are controlled by a controller 32 as is well known to those skilled in the art. Alternatively, the light 22, 24 travels in a manner well known to those skilled in the art and tracks a spiral path or another type of scanning path.

  On the wafer 20, the region or spot 20a illuminated by one or both lights scatters light from these lights. Rays scattered by the region 20a along the direction perpendicular to the surface of the wafer and close to the line 36 passing through the region 20a are focused by the lens concentrator 38, collected and directed to the PMT 40. The lens 38 focuses the scattered light along a direction close to the vertical direction, and such a focusing path is referred to herein as a narrow angle path, and the PMT 40 is referred to as a dark field narrow angle PMT. If desired, one or more polarizers 42 may be placed in the optical path of the focused light in the narrow angle path.

  The light beam scattered along the direction away from the vertical direction 36 at the spot 20a of the wafer 20 irradiated by one or both of the light beams 22 and 24 is focused by the elliptical concentrator 52, and passes through the stop 54 and an optional polarizer 56. Then, the light is condensed on the dark field narrow angle PMT60. Compared to the lens 38, the elliptical concentrator 52 can focus scattered light along a direction at a larger angle with respect to the vertical direction 36, and such a focusing path is called a wide-angle path. The outputs of the detectors 40 and 60 are sent to the computer 62, where the signals are processed to determine the presence or absence of abnormalities and their characteristics.

The SP1 ^TBI system is advantageous for inspecting an unpatterned wafer because the focusing optics (lens 38 and mirror 52) are rotationally symmetric with respect to the vertical direction 36, and defects on the surface of the wafer 20 The direction of the system shown in FIG. Further, the angular range of the scattering space targeted by these concentrators coincides with the angular range necessary for detecting the abnormality in the wafer inspection where the pattern is not formed.

However, in addition to the above features, an important feature of the SP1 ^TBI system 10 is that both its lens concentrator 38 and elliptical mirror concentrator 52 are oriented in the light scattered by defects on the surface of the wafer 20. The point is to save the corner information. Thus, certain defects and / or patterns on the wafer can cause light rays to scatter in a certain azimuthal direction. By utilizing the azimuth information stored in the rays focused by the concentrators 38 and 52, the system 10 can be advantageously adapted and modified for the detection of defects on the patterned wafer.

  One aspect of the present invention can separately detect light scattered in different azimuthal directions by splitting the light focused by lens 38 and / or elliptical mirror 52. In this method, the detector that detects the light diffracted or scattered by the pattern may be saturated, but other detectors that do not detect the diffraction or scattering are used to detect and classify defects on the wafer 20. An effective signal can be generated. Since the lens 38 and the elliptical mirror 52 store the information on the azimuth angle of the scattered light, the knowledge on the pattern and defect type existing on the wafer 20 is used for the design arrangement of a plurality of detectors on the wafer. Can be advantageously detected and classified. This is particularly true for regular patterns such as storage structures formed on the wafer 20 described below. This is because the light diffracted by such a regular pattern is regular.

  FIG. 2 is a schematic diagram showing a converging hollow conical beam that can be focused by lens 38 or mirror 52. In the case of the lens 38 in FIG. 1, a spatial filter (not shown in FIG. 1) is employed to prevent regular reflection of the normal incident light 22 from reaching the detector 40. As a result, the light beam condensed on the PMT 40 has a converging hollow cone shape as shown in FIG. In the case of the elliptical mirror 52, since the mirror is not perfectly elliptical, only the light scattered at a large angle with respect to the vertical direction 36 is focused without focusing the light scattered near the vertical direction. The light beam condensed on the detector 60 also has a converging hollow cone shape as shown in FIG.

  FIG. 3A is a schematic diagram illustrating a possible arrangement of a multi-fiber path for receiving the conical light beam shown in FIG. 2, such as focused on a mirror 52, and illustrates a preferred embodiment of the present invention. The configuration of FIG. 3A includes two generally concentric annularly arranged optical fiber paths 72 that are used to carry the focused scattered light in the focusing hollow cone shown in FIG. A Fourier component or other pattern scattered from the pattern on the wafer 20 reaches a portion of the fiber 72, thereby saturating the detector that detects the light from that path. However, there are other fiber optic paths that do not accept such undesirable pattern scattering. By using the multi-fiber path 72, the focused scattered light is effectively split into different sectors or segments so that only a portion of the fiber path receives a strong signal and is saturated by Fourier or other pattern scattering. Also, it can carry information that can be analyzed to detect anomalies from the remaining paths. As will be described below, since the information on the azimuth angle of the focused scattered light in the conical shape of FIG. 2 is stored, various patterns are adopted and the pattern when the division method of FIG. 3A is adopted. The influence of scattering can be minimized.

  Different types of detectors can be used to detect the light beam carried by the fiber path 72, such as the multi-anode PMT shown in FIG. 3B. However, when using a multi-anode PMT, there is 3 percent nominal crosstalk between two adjacent paths. In order to avoid such crosstalk, fibers 72 are aligned with every other PMT anode, as shown in FIG. 3B. FIG. 3B is a schematic diagram of a multi-anode PMT. As shown in FIG. 3B, only the shaded anode 74 is aligned with the fiber 72, and the anode 76 is not aligned with any of the fibers 72. This avoids the 3% crosstalk that occurs when all of the anodes shown in FIG.

  FIG. 4 is a diagram showing an arrangement 80 of fiber paths or multiple detectors 82 for narrow-angle paths. As shown, the fiber or detector 82, along with the focused scattered light shown in FIG. 2 in the narrow angle path focused by the lens 38, splits the light beam in the same manner as in the wide angle path.

FIG. 5A is a partial cross-sectional view and a partial schematic diagram of the defect detection system, and shows an embodiment of the present invention. To simplify FIG. 5A, the two illumination rays 22 and 24, the computer 62, and the mechanism for moving the wafer are not shown. Light rays scattered by the spot 20 a on the wafer 20 and focused by the lens 38 are reflected by the mirror 102 to the detector 40. Since the stopper 104 prevents regular reflection of the normal incident light 22 from the detector 40, conical focused light as shown in FIG. 2 is generated. The light focused and condensed by the lens 38 and reflected by the mirror 102 passes through the beam splitter 106, and the portion of the focused light that has passed through the beam splitter is collected on the detector 40 and is subjected to normal SP1 ^TBI operation. Produces a single output. The beam splitter 106 reflects a portion of the focused light from the lens 38 and redirects it to the optical fiber array 80 of FIG. Preferably, the dimensions of the optical fiber 82 and the dimensions of the hollow cone reflected by the beam splitter 106 are set so that the fiber 82 focuses and carries most of the hollow cone-shaped rays. Each fiber 82 is then connected to a corresponding detector, or a detection unit of a multi-unit or multi-element detector. Similarly, the beam splitter 112 redirects a small amount of the light beam focused by the elliptical mirror 52 toward the array 70 'of optical fiber paths 72 shown in more detail in FIG. 5B, where each path 72 is a separate entity. Or a separate detection unit of a multi-element detection system (not shown). As shown in FIG. 5A, the beam splitter 112 is configured to redirect light toward the array 70 ′ only within the small diameter ring 114. Most of the light beam focused by mirror 52 passes through beam splitter 112 and is collected at detector 60 to produce a single output, as in normal SP1 ^TBI operation. In FIG. 5A, the irradiation light beams 22 and 24 and the mechanism for moving the wafer are omitted for the purpose of simplifying the drawing.

As is clear when comparing the system 10 of FIG. 1 to the system 100 of FIG. 5A, the system 100 has substantially all the characteristics of the system 10 of FIG. In addition, the system 100 redirects a portion of the scattered light focused by the lens 38 and mirror 52, respectively, and directs it toward the fibers 82, 72, carrying the split light to a separate detector or detection unit. To do. This system is small in size and requires a separate space as compared with the SP1 ^TBI system 10 shown in FIG. In this way, the integrated coupling device is optimized and can be used for both unpatterned and patterned wafers, and for two types of wafer inspection, two separate No need for equipment.

  When inspecting only the wafer on which the pattern is formed, another defect inspection system 150 shown in FIG. 6A can be used. In FIG. 6A, the irradiation light beams 22 and 24, the computer 62, and the mechanism for moving the wafer are omitted to simplify the drawing. As shown in FIG. 6A, the scattered light focused by lens 38 and mirror 52 is reflected by mirror 112 'and directed toward the system of optical fiber 152, shown in more detail in cross section in FIG. 6B. As shown in FIG. 6B, the system 152 includes an annularly arranged fiber 82 for carrying the scattered light focused by the lens 38 and an annularly arranged fiber for carrying the scattered light focused by the mirror 52. 72. As described above, the fibers 72 and 83 are each connected to a separate detector or a detection unit of a multi-unit detector.

  4 and 5B, the detectors are arranged in a single ring, but a plurality of rings arranged as shown in FIG. 3A can also be used. The optical transmission cores of the optical fibers located adjacent to each other in each of the two arrays 70, 70 ′, 80 are separated from each other by the cladding surrounding the core, reducing crosstalk between the adjacent cores. is there. Obviously, optical paths other than fibers can be used and are within the scope of the present invention. When such a path does not include a separator such as a clad in the case of an optical fiber, crosstalk can be reduced by using another optical separator.

  Refer to FIG. 5A. Although redirecting some of the focused scattered light from detectors 40 and 60 may inspect the particle sensitivity of system 100 somewhat when inspecting an unpatterned wafer, this reduction in sensitivity The system 100 is insignificant due to the high efficiency of the narrow and wide angle focusing paths. If desired, when inspecting an unpatterned wafer, the rays carried by the fibers 72 and 82 are directed to detectors 40 and 60, respectively, to substantially restore the sensitivity of the system 100 and increase its sensitivity. It may be substantially the same as the sensitivity of the system 10 of FIG.

  The systems 100 and 150 of FIGS. 5A and 6A are particularly advantageous in distinguishing microscratches and particles. The scattering pattern due to micro scratching has the highest energy concentration and the highest detection uniformity when captured in a substantially vertical or narrow angle path focused vertically and focused by lens 38. The only indication of a scratch having a longitudinal pattern in the far field allows a simple classification method. Thus, a hollow cone of light collected by the lens 38 toward the fiber 82 when the output of these fibers is redirected by a multi-channel detector or beam splitter 106 directed to an array of individual detectors. If eight or more fibers 82 are arranged in a ring in the optical path, a simple process of comparing a signal obtained through any two diametrically opposed fibers with a signal from the remaining fibers The presence / absence of micro-scratches can be determined by. When illuminated at an angle, the microscratch produces a scattering pattern different from that due to particles by using the multiple detection channels described above for pattern inspection, i.e., multi-fiber units 70, 70 '. It is also possible to place individual detectors or multi-element detection systems directly in the converging hollow conical light path rather than individual optical fibers in both wide and narrow angle paths.

<Array wafer>
When inspecting a wafer having memory cells on its surface using the systems 100, 150, the Fourier components from the memory array spin as the wafer rotates. Thus, these elements rotate and assume different azimuth angles with respect to the vertical direction 36 of FIGS. 1, 5A and 6A. That is, these Fourier components are carried by different fibers 71 and 82 as the wafer rotates. Since the array of memory cells has different dimensions in the X and Y directions of the wafer, the number of detectors saturated by the Fourier component changes as the wafer rotates. As a result, the dimensions of the memory cell in the X and Y directions can be known, and the number of Fourier diffraction elements can be predicted. Alternatively, a learning cycle is performed during the initial initialization step of determining the maximum number of Fourier components that must be removed by recognizing the maximum number of detectors at very strong or saturated outputs. In subsequent measurement steps, only that number of detector outputs is removed, but the removed output is a saturated output or an output having a maximum value. For example, in the case of a multi-anode PMT, if each anode is used connected to a corresponding fiber, crosstalk can be reduced by removing the element adjacent to the detector with the highest power. For example, if the Fourier component is 3 at one location on the wafer and three direct components are deleted at the other two locations together with the two adjacent components, a total of nine detector outputs will be deleted. become. This leaves seven available detector outputs. This number is maintained regardless of whether the wafer is oriented correctly. This allows the user to retain sizing options for the particles.

  Preferably, the fibers 72, 82 are rotationally symmetrical about a direction such as the axes 74, 84 shown in FIGS. 3A, 4, 5B and 6B. With this arrangement, the light scattering direction is located within the same angular segment, and the light scattered within each segment is focused by the corresponding fiber. When the beam splitter or mirror 102, 112, 112 ′ reflects or redirects the portion of the light beam focused by the lens 38 or mirror 52, the azimuthal position of the focused scattered light is such that the reflected or redirected light beam is the fiber 72, When it is guided to 82, it is stored. When the light beam is thus reflected or redirected, the axes 74 and 84 corresponding to the vertical direction 36 and the azimuth of the focused scattered light about the axes 74 and 84 corresponding to the azimuthal position relative to the vertical direction 36. The position is saved.

  As described above, the azimuthal characteristics of scattered light are preserved for both narrow and wide angle paths. The scattering pattern by micro scratches irradiated by light 22 in a substantially vertical irradiation direction has the highest energy concentration and the highest detection uniformity when captured in a narrow angle path. In addition, the only indication of scratches having the form of a longitudinal pattern in the far field allows a simple classification method. For example, referring to FIG. 4, eight fibers 82 are used in an array 80 to receive and carry scattered light in the hollow conical light of FIG. When connected to individual detectors, the sum of two signals from two diametrically opposed two fibers is compared with the remaining detector output signals to determine the presence or absence of microscratches. Check.

  As explained above, once all the scattered light from the illuminated spot 20a is focused and directed to a single detector, the presence of Fourier or other scattering components saturates the detector and It becomes impossible to provide effective information regarding the abnormality in the irradiated spot. For this reason, the applicant has proposed dividing the focused scattered light into different segments. Dividing the focused scattered light into only a few segments, such as two or three, produces only two or three output signals, and the two or three segments still contain pattern scatter, It is likely that two or three detectors will saturate again and no useful information about the anomalies can be obtained. Therefore, in order to be effective, it is preferable to divide the detection signal sufficiently finely so that no conspicuous pattern scattering exists in a part of the detection signal. Thus, if the lines connecting various Fourier or other scattering components and the vertical direction 36 are not closer to each other than δφ in the angular direction, each detector can receive the scattered light focused within the aperture angle not exceeding δφ. It is preferable to divide into two. In this way, there are at least some detectors that do not receive Fourier or other pattern scattering, and generate an effective output signal that can reliably confirm the presence or absence of features or characteristics of the sample surface. Therefore, when the divided light beams are conveyed to a plurality of optical fibers, it is preferable that at least some of the fibers receive light beams that are focused within an azimuth range of δφ or less.

  Another arrangement for dividing the focus of scattered light is shown in FIG. FIG. 7 is a top view of a rotationally symmetric concentrator such as an elliptical or parabolic mirror 200 having two stops 202,204. The two stops are preferably centered at + 90 ° and −90 ° azimuthal positions with respect to the oblique light 24 shown in FIGS. A multi-element detector or detector array 206, 208 is disposed in each of the two stops. In this case, the detector or array may be a multi-anode PMT or a multi-PIN diode array. FIG. 8A is a schematic side view of a part of the detector or detector array 206, 208 of FIG. 7 as viewed from the direction of arrow 8A. As shown in FIG. 8A, each of the detection units 206a and 208a has a rectangular shape substantially having a width w. Preferably, the detection units 206 a and 208 a are arranged such that their long sides are parallel to the substantially vertical direction 36. Thus, each of the detection units 206a, 208a is centered within a sector of a small angle that is delimited when the angle delimited by the width of the long elements 206a, 208a is less than or equal to δφ. Because the scattered light is focused towards the axis 36, at least a portion of the detector is not blocked by pattern scattering and outputs a valid signal for detecting and characterizing anomalies on the sample surface.

  By disposing two detectors or detector arrays 206 and 208 in the diaphragms 202 and 204, respectively, the detection units 206a and 208a output an effective signal component for detecting an abnormality. The process of either predicting or determining through an immediate learning cycle as described above is applied to the two detectors or detector arrays 206, 208 to determine the maximum number of pattern scatter components that need to be removed. The abnormality can be detected using the remaining detection signals.

  The dimensions of semiconductor circuits are becoming increasingly smaller. Thus, the cell dimensions are similarly reduced, and the number of Fourier or other scattering components is correspondingly reduced. If the cell size is large, each detector unit of the two detectors or detector arrays 206, 208 will be saturated and a valid signal unless the detector unit width w of the detectors or detector arrays 206, 208 is reduced. Cannot be obtained. This can be corrected by the mechanism shown in FIG. 8B.

  As shown in FIG. 8B, the signal collection capability of the detector or detector array 206, 208 can be further enhanced. If the number of pattern scatters increases beyond the design value of the detector or detector array, using the arrangement of FIG. 8B allows anomaly detection despite the increase. As shown in FIG. 8B, the multiple detection units of the detectors or detector arrays 206, 208 are labeled D1, D2,..., D2n, D2n + 1,. The odd number detection units D1, D3, D5..., D2n + 1... Of the multi-unit detector or detector array 206 are covered with a spatial filter 216. Further, the even number of detection units D2, D4, D6... D2n... Of the detector or array 208 are covered with a spatial filter 218 shown in FIG. In this way, even if a relative rotational movement occurs between the sample surface and the detector or array 206, 208, the uncovered detection unit outputs a valid signal.

  FIG. 9A is a cross-sectional view of the concentrator 52 of FIG. 1, modified to include a type of aperture, or detector, or detector array as shown in FIGS. 7, 8A, and 8B. The two stops 202, 204 are preferably sized so that each side is centered at an azimuthal position of ± 90 ° and has an azimuthal gap of about 10 ° to 40 °. The stop is located only towards the bottom of the concentrator, and only the scattered light along the direction close to the surface can be detected by the detector or detector array 206,208. Two scattered lenses 222 and 224 having an appropriate F number are used to focus the scattered light from the irradiated spot 20a and focus it on the respective detectors or detector arrays 206 and 208. Two detectors or detector arrays can be placed on the back focal plane of the two lenses 222,224.

  Masks 216 and 218 are arranged between the spot 20 a irradiated by the filter wheels 226 and 228 rotated by the actuators 232 and 234 and the detectors or detector arrays 206 and 208. This method is well known to those skilled in the art, so the connection between these two actuators and the wheel is not shown and a detailed description of its operation is not necessary here. For simplicity, only the mask portions 216 and 218 of the two filter wheels 226 and 228 are shown in FIG. 9A. The characteristics shown in FIGS. 9A, 9B, and 9C can be combined with the systems 100 and 150 of FIGS. 5A and 6A to further enhance their versatility. When the combined device is used for inspection of a wafer on which a pattern is not formed or a bare wafer, for example, the sensitivity reduction due to the two stops 202 and 204 is negligible. Further, when inspecting an unpatterned wafer, the output of the detector or detector array 206, 208 can be added to the output of the detector 60 to at least partially restore the sensitivity of the system. In order to suppress extraneous signals due to film roughness, the characteristics shown in FIGS. Since the roughness of the film scatters the P-polarized light more efficiently than the S-polarized light, under such circumstances, the S-polarized oblique irradiation light 24 is output and only the S-polarized light scattering from the irradiated spot 20a is focused. Is preferred. This can be conveniently achieved by the filter wheels 226,228. Filter wheels 226 and 228 are rotated using actuators 232 and 234, with S polarizer 236 replacing mask 216 in FIG. 9A and another S polarizer replacing mask 218 in FIG. 9A. As can be seen from FIG. 9A, this arrangement is advantageous because the two filters 236, 238 are located near the surface of the wafer 20, so that the focused light is confined at a scattering angle very close to the wafer surface. If the membrane surface is very rough, the upper half of the S-polarizer can be shielded using semi-circular opaque screens 236 ', 238' in the filter wheel to further limit the focusing elevation angle. For example, a semi-circular S-polarizer can limit the focusing elevation angle of the stop within the range of about 55 to 70 ° from the vertical direction 36. This is useful because the amount of scattering due to film roughness increases with the elevation to the wafer surface. FIG. 9C illustrates another filter wheel that can be used to inspect bare wafers or wafers that are not patterned.

  If the direction of the expected pattern scattering surface is known, the spatial filter can be designed to block such scattering, thereby detecting only scattering due to anomalies on the surface. FIG. 10 is a schematic diagram showing a two-dimensional Fourier component of the array structure when illuminated by normal incident light. As the sample rotates, all spots located at the intersection of the XY lines rotate to generate a circle. These circles represent the position of the Fourier component when the wafer is rotated. The dark, dark circle in the center indicates the 0-5 ° blockage of the focusing space by the stopper 104 of FIG. 5A. From FIG. 10, it can be seen that there is a gap where no Fourier component exists between the circles. At least theoretically, it is possible to construct a programmable filter (for example, a liquid crystal filter) in which an annular band having an arbitrary radius is blocked. Many of the objects of the present invention can be achieved by constructing a simple spatial filter. Thus, the cell size of a typical memory array on the wafer is such that, for example, in the 488 nanometer wavelength light used for the illumination light 22, 24, if the XY dimension is about 3.5 microns or less. The Fourier component of 1 is about 8 ° with respect to the vertical direction 36. Thus, when employing a spatial filter, by blocking any focused light in a narrow-angle path that is 8 ° or more relative to the vertical direction 36, approximately 8 ° from the central obstacle rim (ie, 5 or 6 °). An annular gap of 2 or 3 ° extends to the rim of the variable throttle. Under these circumstances, when the wafer spins, the Fourier component cannot pass through the annular gap and scattering from the array is suppressed. In some embodiments, the spatial filter used leaves an annular gap between about 36 and about 5 to 9 degrees in the vertical direction.

  In the example described above, the spatial filter is designed for a narrow angle path. It will be appreciated that similar spatial filters can be designed for wide angle paths. Such and other modifications are within the scope of the present invention.

  As previously explained, pattern scattering in which at least some detector focusing apertures are expected to ensure that at least some detectors receive a valid signal that is not blocked by Fourier or other pattern scattering. Is preferably less than or equal to the angular difference. For this purpose, a spatial filter is configured to block all rays focused in a narrow or wide angle path, except for small aperture angles where the aperture angle is less than or equal to the angle difference between pattern scattering. can do. When such a spatial filter is placed between the irradiated spot 20a and a detector such as detector 40 or 60, the Fourier component spins in and out of this small aperture. If no component passes, the data is valid for detecting anomalies. Otherwise, the signal is very strong or saturated. Thus, at the end of the helical scan, the wafer map is a series of sectors that are both valid and saturated. If the scan is repeated one more time, during the first scan, the center position of the aperture angle deviates from that position by the minimum angle difference of the scattered scattering, and as before, the data is valid and saturated sector A similar map containing is obtained. However, in a region saturated during the first scan, valid data can be obtained this time. Thus, a complete wafer map with valid data can be obtained by combining two data sets using a logical OR operation.

  The above-described process can be simplified by using an asymmetric mask 250 as shown in FIG. As shown in FIG. 11, the two fan-shaped stops 252 and 254 are shifted from the positions facing in the diametrical direction by an angle equal to the predicted minimum angle difference of pattern scattering. When such a filter is placed between the illuminated spot 20a shown in FIG. 1 and the detector 40 or 60, the detectors 40 and 60 produce a complete wafer map as the wafer is scanned.

  FIG. 12 is a schematic diagram of a failure detection system showing another embodiment of the present invention. As shown in FIG. 12, when irradiated with light (not shown) such as light 22 and 24 in FIG. 1, the light is focused by a concentrator 52 (not shown in FIG. 12 for simplicity). The scattered light is collected on a triangular device 262 having two mirrors 262a, 262b on opposite sides of the device. The irradiation beam is also omitted for simplicity. Thus, the scattered light is reflected by the device 262 into two opposing hemispheres. The mirror 262a reflects half of the scattered light to PMT1, the mirror 262b reflects the other half of the scattered light to PMT2, and an asymmetric mask 250 is provided between the mirrors 262a and PMT1 and between the mirrors 262b and PMT2. be able to. Thus, two wafer maps effective for abnormality detection and classification can be obtained by the two PMTs.

<Detection of CMP failure>
One aspect of the present invention includes two algorithms for classifying CMP failures. The first method is based on the spatial distribution of light scattered by defects. Theoretical simulations and experimental results show that light scattered by CMP microscratches first proceeds in the direction of specular reflection, and light scattered by particles (especially small particles) has a different spatial distribution. . As a result, defect classification can be realized by measuring the distribution of scattered light. This can be done by placing multiple detectors at appropriate locations around the scatterer. Alternatively, a single detector is used with multiple spatial filters / masks. Three different ways of implementing this algorithm are described.

  The second algorithm is based on double polarization. This method uses incident S and P polarized light to compare scattered signals from defects. According to theoretical simulations, it was found that the intensity of scattering is proportional to the intensity of local interference seen by defects. The intensity of this interference is different for S and P polarized light and depends on the height from the surface of the wafer. Thus, there is a large difference between the strength of interference seen by particles (defects on the surface) and the strength of interference seen by microscratches (wafer surface or below the surface). Bad classification can be achieved by comparing the strength of the scattered signal using S and P polarized incident light or rays.

<Operation details>
The following paragraphs describe the execution / operation of the present invention in the Surfscan SP1 ^TBI ™ system. However, the algorithm is not limited to the SP1 ^TBI ™ system and can be implemented with any light scattering tool. All of the algorithms shown below require PSL calibration curves for all paths used. These are very important factors for successful CMP failure classification.

Algorithm # 1, Run # 1, dual channel, oblique incidence, single scan SP1 system has four dark field paths: DWN, DNN, DWO, DNO. DWN represents a path for carrying scattered light focused by an elliptical mirror from vertical illumination light. DNN represents a path for carrying scattered light focused by a lens concentrator from vertical illumination light. DWO represents a path for carrying scattered light focused by an elliptical mirror from oblique irradiation light. DNO represents a path for carrying the scattered light focused by the lens concentrator from the oblique irradiation light. The dual channel method uses two dark field paths, eg, DWO and DNO paths. The principle of this method is based on the fact that the spatial scattering patterns of particles and microscratches are different. The particles scatter light in all directions, which can be focused by both dark field paths. However, microscratches preferentially scatter light in a certain direction, so that the signal captured in one path is much larger than that captured in another path. For example, when using slanted paths DWO and DNO, microscratches are preferentially captured in the DWO path, or the signal in the DWO path is significantly greater than the signal in the DNO path. In order to distinguish micro scratches from particles, the dimensional ratio of each defect caught in the DWO and DNO paths is calculated. If the defect size ratio is close to 1, the defect is classified as a particle. However, if the defective dimension ratio is less than a certain number (for example, 0.8), it is classified as a micro scratch. If a defect is caught only on the DWO path and not on the DNO path, it is classified as a CMP microscratch. If a defect is caught only on the DNO path and not on the DWO path, it is classified as a particle.

Execution in algorithm # 1, execution # 2, dual channel, normal incidence, single scan vertical path is similar to execution in an oblique path. The difference is that the light scattered from the CMP microscratch travels preferentially to the normal incidence narrow angle path (DNN) instead of the wide angle (DWN) path. This is due to the fact that CMP micro scratch scattered light travels in the direction of regular reflection. The defect classification is performed by calculating a dimensional ratio of defects captured by both the DNN and DWN paths. If the defect size ratio is close to 1, the defect is classified as a particle. However, if the defective dimensional ratio is greater than a certain number (eg, 1.6), it is classified as a microscratch. If a defect is caught only on the DNN path and not caught on the DWN path, it is classified as a CMP microscratch. If a defect is caught only on the DWN path and not on the DNN path, it is classified as a particle.

A third method for executing algorithm # 1, run # 3, single channel, oblique incidence, 2-mask, double scan algorithm # 1 uses two masks. One mask (# 1) is designed to preferentially capture the scattered light from the CMP micro scratch. This mask is shown in FIG. 13A. The shaded area indicates an area where light rays are blocked, and the non-shadowed area is an area through which light rays can pass. The other (# 2) is designed to shield light scattered by CMP microscratches. This mask is shown in FIG. 13B, where the shaded area indicates the area where the light beam is blocked, and the unshaded area is the area through which the light beam can pass. Both mask shapes require a calibration curve. The classification of defects is performed by calculating a dimensional ratio of defects captured in both mask shapes. If a dimensional ratio between mask # 1 and mask # 2 is close to 1 for a certain defect, the defect is classified as a particle. However, if the defect size ratio is greater than a certain number (eg, 1.15), it is classified as a micro-scratch. Further, when a defect is caught only by the mask of the form of # 1, and not caught by the mask of the form of # 2, it is classified as a CMP micro scratch. If a defect is caught only with the mask in the form of # 2, and not caught with the mask in the form of # 1, it is classified as a particle.

  Algorithm # 1 can also be executed in a multi-anode PMT. The advantage of this method is that only one scan is required. Essentially the same as using two masks, but only one scan is required for data collection.

Algorithm # 2, execution # 1, single channel, dual polarization, tilted incidence, double scanning algorithm # 2 uses two incident polarizations S and P. This method requires two scans. One is for S-polarized light and the other is for P-polarized light. Use PSL calibration curves for S and P polarizations. The defect classification is performed by calculating a dimensional ratio of defects captured in both P and S scans. If the P and S scan size ratio is close to 1, the defect is classified as a particle. However, if the dimensional ratio of the defect is other than 1 (for example, <0.65 or> 1.85 depending on the film thickness), the defect is classified as a micro scratch. In the case of a dielectric film, the strength of interference with the two polarized lights varies depending on the film thickness. The change in the intensity of the interference between the two polarizations is out of phase. That is, when the P-polarized light interference intensity is maximum, the S-polarized light interference intensity is minimum or vice versa. Therefore, the dimensional ratio of CMP failure is determined to be larger or smaller than 1.0 depending on the thickness of the dielectric film. Similarly, if a defect is captured only in one polarization and not the other, the defect is classified as CMP microscratch or particle depending on the film thickness. This method has been successfully demonstrated using an oxidized CMP wafer. This method seems to give better results for the metal film than for the thick dielectric film, because the change in thickness across the wafer is not a big problem for the metal film having a substantial thickness. It is.

  In one experiment, a PSL sphere is used to calibrate the SP1 ™ instrument, and the particle size ratio of intensity detected during P and S scans is normalized to unity. Thus, if particles are present, the ratio is 1 or about 1. Furthermore, according to the histogram obtained from the instrument, the intensity ratio of the second set is concentrated at a value greater than 1, which is greater when P-polarized light is irradiated than when S-polarized light is irradiated. It represents a set of defects with greater scattering. These are CMP defects such as micro scratches, even when the scattering intensity is detected only during P scan and not during S scan. This is because in that case the ratio is infinite and thus greater than one. The third set of ratios is zero or close to zero. These are likely to represent particles for the following reasons.

  Due to the interference effect on the surface detected when irradiated with P or S polarized light, the detected scattering intensity is stronger during P scanning compared to that during S scanning, or vice versa. Thus, in the experiment described above, if the scattering intensity detected by the interference effect on the surface is stronger during P scanning than during S scanning, particles of sufficient size in the region where S-polarized light undergoes structural interference. Only there will be. This is illustrated, for example, in FIG. For example, referring to FIG. 14, when the film thickness on the wafer surface is 200 nanometers, the interference intensity on the wafer surface is greater when the P-polarized light beam is irradiated than when the S-polarized light beam is irradiated. Can be expected to be significantly stronger. However, in the case of particles of 300 nanometers or more, the scattering intensity detected during the S scan is significantly stronger than during the P scan.

<Surface roughness determination>
Haze measured from the film when there is almost no change in film thickness in an opaque film such as a metal or a transparent dielectric such as a dielectric having a low k value (both spin on the CVD deposition) Varies depending on the roughness of the film surface. Dielectric film CVD deposition for application in integrated circuits is almost uniform. Thus, the haze measurement provides a quick alternative for measuring film roughness.

Surface roughness can generally be measured by instruments such as the HRP tool of KLA-Tenker Corporation of San Jose, Calif., And can also be used with atomic force microscopes, near-field microscopes, and scanning tunneling microscopes. It can be measured by other types of scanning probe microscopes. This process takes time. Measure film roughness more quickly than conventional methods by utilizing the above relationship between haze and film roughness in uniform dielectric films or in metals with varying uniformity Can do. Thus, according to FIG. 16, the KLA-Tenker Corporation high resolution profiler or AFM is used to create a database of correlations between haze and surface roughness using the computer 310 for films of different thicknesses. Measure surface roughness of representative film 302 of different thickness using type 304 tool to measure SP1 ^TBI system 10 or one of the integrated systems described above (eg, 100) or haze. Databases can be constructed by measuring haze values for these same films using other tools that can. Since the roughness of the surface increases as the film thickness increases, it is preferable to measure similar films having various thicknesses. The database can be shown in a line graph as shown in FIG. Then, when it is desired to determine the roughness of the surface of an unknown film, the roughness is measured by measuring the film haze using a device such as the system 10 shown in FIG. 1 or an integrated device as described above. Can be determined. Then, using the measured value of haze, the corresponding roughness value is selected from the database for the film whose thickness is already known from the graph shown in FIG. This takes about 1 minute to measure the haze value and associate the haze measurement with the RMS roughness calibration curve of FIG. become able to.

  While the invention has been described in terms of various embodiments, it will be understood that modifications and variations can be made without departing from the scope of the invention, which is limited only by the appended claims and equivalents thereof. Like. All references cited in this application are incorporated by reference in their entirety.

Claims (12)

A surface inspection method for detecting anomalies on a surface having a diffraction pattern that scatters light rays,
Scanning the surface with light rays;
A step of focusing the light rays scattered from the surface by concentrator which focuses the scattered light substantially symmetrically with respect to a line perpendicular to the surface, the concentrator is the light scattered by the light and the surface of the optical line or al Have one or more apertures to pass through,
Filtering at least a portion of the light that has passed through the one or more apertures with a spatial filter having an angular gap of an angle corresponding to the angular difference of the light components expected to be scattered by the pattern on the surface;
Determining whether there is an abnormality in or on the surface from at least one of the filtered light and the focused light; and
Including surface inspection methods. The method of claim 1, wherein
The concentrator is to define two aperture through which light scattered by the optical line or these light and said surface,
Steps for the filtering, respectively filters the focused light by two corresponding spatial filter having an angular gap therein, the gap, by an angle corresponding to the angular difference between the components that are expected to scattering by the pattern on the surface from each other Misaligned way. The method of claim 2, wherein
The light beam illuminates the surface at an oblique angle, and the two stops are centered at azimuth positions of +90 degrees and -90 degrees, respectively, with respect to the oblique light beam. The method of claim 2, wherein
A method in which the angular difference of components predicted to be scattered by the pattern on the surface is greater than or equal to a certain value and the gap and displacement are substantially equal to said value. The method of claim 4, wherein
Outputting a signal in response to the filtered first and second portions of the focused light; and combining the signals to detect anomalies within or on the surface. The method of claim 1, wherein
The method in which the component predicted to be scattered by the pattern is a Fourier component. A surface inspection apparatus for detecting anomalies on a surface having a diffraction pattern that scatters light rays,
A source for supplying light for scanning the surface;
A concentrator for focusing a light beam scattered from the surface in a substantially symmetrical with respect to a line perpendicular to the surface, with one or more aperture passing light scattered by the optical line or these light and the surface A concentrator;
A spatial filter for filtering at least a portion of light that has passed through the one or more apertures, wherein the filter has an angular gap of an angle corresponding to an angular difference predicted to be scattered by a pattern on the surface When,
A processor for determining the presence or absence of an abnormality in or on the surface from at least one of the filtered light and the focused light;
A surface inspection apparatus comprising: The apparatus of claim 7.
The concentrator defines two aperture through which light scattered by the optical line or these light and said surface,
The apparatus includes first and second spatial filters for filtering first and second portions of the light, respectively, the two spatial filters each having an angular gap therein, and two An apparatus in which the gaps of the filters are offset from each other by an angle corresponding to the angular difference of the components expected to be scattered by the pattern on the surface relative to the line. The apparatus of claim 8.
An apparatus in which the angular difference of components expected to be scattered by the pattern on the surface is greater than or equal to a certain value and the gap and displacement are substantially equal to said value. The apparatus of claim 9.
An apparatus in which the processor outputs signals in response to the filtered first and second portions of the focused light and combines the signals to detect anomalies within or on the surface. The apparatus of claim 8.
The light beam illuminates the surface at an oblique angle, and the two stops are centered at azimuthal positions of +90 degrees and -90 degrees, respectively, with respect to the oblique light beam. The apparatus of claim 7.
An apparatus in which a component predicted to be scattered by the pattern is a Fourier component.

Images