BACKGROUND OF THE INVENTIONThis invention relates in general to the inspection of surfaces to detect anomalies, and in particular, to an improved system that illuminates the surface inspected at the plurality of spots simultaneously for anomaly detection.[0001]
Conventional optical inspection methods employing scanning techniques typically causes a single spot on the surface inspected to be illuminated where the spot is scanned over the entire surface for anomaly detection. For improved signal-to-noise ratio caused by background scattering, the size of the illuminated spot has been continually reduced. This means that the amount of time required for the spot to scan over the entire surface is increased which is undesirable.[0002]
One solution to the above dilemma is proposed in U.S. Pat. No. 6,208,411 which is incorporated herein by reference in its entirety. This patent proposes a massively parallel inspection and imaging system which illuminates the surface at a plurality of spots where scattered light from the spots are imaged onto corresponding detectors in a detector array.[0003]
While the system in U.S. Pat. No. 6,208,411 provides a major enhancement in the total inspection throughput, it may be further improved for enhanced performance in certain applications. It is, therefore, desirable to provide an improved multi-spot inspection and imaging system with enhanced characteristics.[0004]
SUMMARY OF THE INVENTIONWhile the system described in U.S. Pat. No. 6,208,411 provides a major enhancement in the total inspection throughput, for some applications, it may be desirable for the system to be compact and have a smaller footprint. In such event, it may be desirable for the focusing optics focusing multiple beams of radiation to an array of spots and the imaging optics imaging scattered radiation from the spots to an array of receivers or detectors to employ different objectives. In one embodiment of the invention, the objective used for imaging has a larger numerical aperture than the objective use for focusing. This enhances detection sensitivity.[0005]
In another embodiment, radiation reflected from an array of illuminated spots on the surface may be imaged onto a first array of receivers or detectors in a bright field detection configuration and radiation scattered from the spots may be imaged onto a second array of receivers or detectors in a dark field detection configuration. The use of both bright and dark field detection provides more information for anomaly detection.[0006]
In yet another embodiment, the multiple beams of radiation are focused to an array of spots on the surface where the radiation comprises at least one wavelength component in the ultraviolet (“UV”) or deep ultraviolet range of wavelengths. Scattered radiation from the spots are imaged by means of optics that comprises a reflective objective to reduce chromatic aberration.[0007]
In still another embodiment in a compact and modular approach, an optical head for anomaly detection includes illumination optics focusing illumination beams of radiation to an array of spots on a surface and imaging optics that images scattered radiation from the spots onto an array of optical fibers. The signals carried by the fibers contain information on scattered radiation from corresponding spots. Such information may be supplied to detectors outside the optical head for processing and anomaly detection. Instead of optical fibers, other types of receivers may also be used.[0008]
In still another embodiment of the invention, in addition to the optical head described immediately above, a plurality of detectors generate signals in response to the information from the receivers or fibers and rotational motion is caused between the surface and the illumination beams so that the beams are scanned over substantially the entire area of the surface. Where the surface inspected is that of a semiconductor sample having multiple dice thereon, signals from the detectors from at least two dice of the surface are stored as the beams scanned over the surface. Preferably, the scattered radiation from the two dice may be compared in a die-to-die comparison for improved signal-to-noise ratio in anomaly detection.[0009]
The surface inspected sometimes has diffraction patterns thereon. In such event, scattered radiation from the array of illuminated spots on the surface may be masked by diffraction from the pattern. Thus, in another embodiment of the invention, when relative rotational motion is caused between the surface inspected and the illumination beams, a filter having an aperture that is caused to move substantially in synchronism with the rotational motion to reduce diffraction from the pattern that is passed to the array of receivers or detectors. Alternatively, as rotational motion is caused between the surface and the beams, a stationary filter in the shape of an annular aperture may be employed to shield the detectors from pattern diffraction. For some applications, both types of filters may be used at the same time during inspection.[0010]
In still one more embodiment, beams of radiation are focused to an array of elongated spots on the surface at oblique angle(s) of incidence to the surface where the centers of the spots are arranged along a substantially straight line. Scattered radiation from the spots are imaged onto corresponding receivers or detectors in an array by means of imaging optics with a focal plane that substantially contains all of the spots.[0011]
In yet another embodiment, the same optics is used for focusing illumination beams of radiation to an array of spots on a surface and for imaging scattered radiation from the spots onto corresponding receivers or detectors in an array. The optics has an aperture where the illumination beams are focused through a first portion of the aperture and the imaging occurs through a second portion of the aperture. Preferably, the second portion is larger than the first portion, which enhances the sensitivity of detection.[0012]
It should be noted that any one or more features in the above-described embodiments may be employed in combination with one or more features of a different embodiment for enhanced performance.[0013]
BRIEF DESCRIPTION OF THE DRAWINGSFIG. 1 is a schematic view of a multi-spot dark-field/bright-field inspection and imaging system to illustrate an embodiment of the invention.[0014]
FIG. 2 is a schematic view of a two-dimensional arrangement of multiple illuminated spots on the surface inspected to illustrate the embodiment of FIG. 1.[0015]
FIG. 3 is a schematic view of the multiple spots of FIG. 2 and their scan paths across the surface inspected to illustrate the embodiment of FIG. 1.[0016]
FIG. 4 is a schematic view illustrating the scan paths of two adjacent spots to illustrate the embodiment of FIG. 1.[0017]
FIG. 5 is a schematic view of a spatial filter in the collection path of the embodiment of FIG. 1 to further illustrate the embodiment.[0018]
FIG. 6 is a schematic view of an annular-shaped spatial filter that may be used in the in the collection path of the embodiment of FIG. 1 to illustrate the invention.[0019]
FIG. 7 is a schematic view of an annular-shaped illuminated region of the surface inspected containing two dice to illustrate one aspect of the invention in the embodiment of FIG. 1.[0020]
FIG. 8 is a schematic view of an optical inspection and imaging system to illustrate another embodiment of the invention.[0021]
FIG. 9 is a top schematic view of an optical inspection and imaging system to illustrate yet another embodiment of the invention.[0022]
FIG. 10 is a schematic side view of an optical inspection and imaging system to illustrate still another embodiment of the invention in a single dark field configuration.[0023]
For simplicity in description, identical components are labeled by the same numerals in this application.[0024]
DETAILED DESCRIPTION OF THE EMBODIMENTSThe costs associated with dark-field pattern inspection has increased steadily with enhanced performance. As semiconductor fabrication approaches finer design rule and resolution, the complexity of inspection tasks has increased dramatically, which, in turn, increases the complexity and costs of the optical front end of the inspection tool and of detection electronics. Furthermore, the variety of situations calling for optical inspection means that a versatile optical inspection tool should be compact, have a small foot print and be rugged so that it is less sensitive to vibrations, and integratable with semiconductor processing equipment. Preferably, the system can be used for inspecting surfaces with diffracting patterns thereon such as patterned wafers, as well as surfaces without such patterns such as unpatterned semiconductor wafers. The embodiments of this invention also enable faster and more sensitive inspection to be performed at a reasonable cost.[0025]
The elements of the optical front-end design (such as those in an optical head) of the proposed[0026]system20 are shown in FIG. 1. The radiation from alaser22 is first split into anarray24 of beams, preferably a two-dimensional array, by the action of a diffractiveoptical element26aonsubstrate26. These beams are simultaneously focused onto the surface of a sample such as asemiconductor wafer28, placed on a spinning stage, preferably a precision spinning stage, by a lens such as a simpledoublet lens30. Preferablylens30 has a numerical aperture of not more than 0.8. The radiation scattered off each spot is collected by areflective lens32, and imaged by an objective38 onto a corresponding fiber in an M×N array34 of optical fibers arranged to correspond to the distribution of the spots on the wafer. These fibers carry the radiation to anarray36 of avalanche photodiodes (APD), amplifiers, and digitizers. Other types of detectors are possible and may be used as described below. Alternatively, instead of imaging the scattered radiation collected from each spot on the wafer to an optical fiber, it may be imaged onto a detector in an detector array. In the embodiment of FIG. 1, the illumination beams24 are directed towards the wafer surface in directions that are substantially normal to the surface of the wafer. Preferably the beams illuminate on the wafer surface spots that are substantially circular in shape
The orientation of the[0027]spots42 illuminated by thearray24 of beams is slightly rotated with respect to the tangential direction x of the wafer as the wafer is rotated as shown in FIG. 2, resulting in the “painting” of the spacing alongpaths44 between any two adjacent spots along a row with the path taken by the spots along the columns as shown in FIG. 3. In a xy coordinatesystem41, thethick arrow46 illustrates the y direction of the image obtained. The separation between the adjacent spots is chosen so as to satisfy a desired sampling level (e.g. 3×3 or 4×4 samples per point spread function, PSF). This is illustrated in FIG. 4, which shows the paths of two adjacent spots, such asspots42aand42bin FIG. 2 travelling alongpaths44aand44brespectively. The two paths may be offset by a separation d substantially equal to one-third or one-quarter of the spot size to achieve the 3×3 or 4×4 samples per point spread function, so that thespots42aand42bwould overlap by two-thirds or three-quarters of the spot size. Thus, a one-dimensional scan of the wafer produces a two dimensional image, as illustrated in FIG. 3.
The optical components in the design are quite simple. The[0028]multi-beam splitter26amay be one similar to the grating element used for a similar purpose in U.S. Pat. No. 6,208,411, where the element is a specially designed diffractive optical element. In choosing the total number ofspots42, it is desirable to pay attention to the total system complexity including, in particular, costs associated with the electronics. It has been determined that a total number of 128 channels is a reasonable compromise. This is achieved through a 16×8 array of spots. Other combinations are also possible. In some applications, the use of an odd number of spots such as 17×9 may be advantageous. The angular orientation of the spots with respect to the tangential direction of the wafer is chosen such that the spots in the vertical direction traverse the space between any two adjacent horizontally positioned spots (FIG. 2), resulting in a complete coverage of the wafer. In one embodiment, the separation between the spots is chosen such that 4 samples per point spread function (PSF) are attained in each direction. This is a slightly TM denser sampling than in the case of the AIT system available from KLA-Tencor Corporation of San Jose, Calif. But in order to reduce processing time, smaller interpolation kernels are allowed for the same level of residual interpolator truncation error. The fact that the scan is spiral also favors a denser sampling, since the interpolation is inherently more complex than for a rectilinear scanner.
The point spread functions of the spots are Gaussian shaped with a 1/e[0029]2intensity width of 5 microns, for example. At a 4×4 sampling level, where the spot separation is about 20 microns, the total widths (i.e. swath) of the tracks of the 128 (for a 16 by 8 array) spots is about 160 microns. In this context, a track is the locus of a spot as the sample is scanned. The maximum amount of the beam fan out in at the focusinglens30 is so small that only a simple doublet suffices for focusing. Other types of lenses may also be used instead.
The dark-field collector in this design is a reflecting[0030]objective32, placed directly above the illuminated field. While a 0.5 numerical aperture (NA) lens may be used for objective32, lenses of other NA are possible. The reflecting lens performs two tasks: i) it collects the radiation scattered off each point, and ii) it images the field onto a corresponding array of fibers. The separation of the spots on the wafer is such that they can be considered as completely independent, without inter-spot interferences.
The radiation provided by[0031]laser22 may contain one wavelength component or more than one wavelength component. Such radiation may include a wavelength component in the ultraviolet range, deep ultraviolet range, visible or infrared range, or wavelength components in more than one of the four different wavelength ranges. The laser orother radiation source22 may operate in the visible, infrared, ultraviolet or deep ultraviolet range or ranges. An attraction of using a reflecting objective such asmirror32 is that it functions well over a large range of wavelengths. For some applications, a refractive objective may also be used instead of a reflective one for collecting and imaging scattered radiation from thewafer28 to thefiber array34.
[0032]Laser22 may emit radiation of substantially a single wavelength in the system of. Alternatively,laser22 may emit radiation of a plurality of wavelengths, although radiation of only one of the plurality of wavelengths is used at any one time for inspection. In such event, it is possible to alter the wavelength of radiation supplied by the laser for inspection. The diffractingelement26ais preferably placed at the back focal plane oflens30, so that thebeams24 are focused to the surface of thewafer28, where the axes ofbeams24 are substantially parallel to one another and perpendicular to the wafer surface.
Where radiation of a different wavelength is employed in scanning the sample surface (such as where[0033]laser22 contains more than one wavelength components), the spot separation may change if thesame element26ais used to diffract the laser beam, since diffraction byelement26ais wavelength dependent. In such event, a different diffraction element such aselement26bmay be used to compensate for the change in wavelength so that the spot separation remains substantially the same. Beam forming optics (not shown) may be used to change the width of the beam from the laser in order to maintain the same spot size as before, so that the collection optics in the system need not be changed. This switching between diffractingelements26aand26bcan be accomplished readily by moving substrate alongdirection27 using means such as a motor (not shown in FIG. 1). Since phase changes are not of interest and are not detected, there is no requirement to align precisely theelement26bwith respect to the beam. Obviously more than two diffraction elements may be formed on thesame substrate26 in the event the laser beam contains more than two wavelength components.
Instead of changing the diffracting element when radiation of different wavelength is used, the same spot separation and spot size as before can be achieved by altering the focal length of the focusing[0034]lens30 in FIG. 1. Then the collection optics also need not be changed. However, since the diffracting element preferably needs to be placed at the back focal plane oflens30, once the focal length of the lens is altered, the element needs to be moved to a new location which is again at the back focal plane of thelens30. Where it is desirable to change the spot separation and spot size, one can alter the wavelength of the radiation used to inspect the wafer without changing the illumination optics. However, the imaging optics may then need to be altered by changing the magnification of thelens38 so thatlens38 will still focus the collected radiation from the spots and image onto the fibers. To obtain a different spot separation and spot size without changing the wavelength of the radiation used to inspect the wafer, one can alter the focal length oflens30, or alter the diffracting element and beam forming optics. Such and other variations are within the scope of the invention.
Where[0035]laser22 emits more than one wavelength component, appropriate wavelength selection optical elements such as filters or beam splitters (not shown) may be employed in the path of the beam fromlaser22 to select the component of the desired wavelength, so that radiation substantially at only one selected wavelength is supplied toelement26aor26bat any one time. In such event,laser22 and the wavelength selection optical elements form an optical source that supplies radiation of a selectable wavelength from a plurality of wavelengths. Obviously, other types of optical source that supplies radiation of a selectable wavelength may be used instead. Thus, alternatively, wherelaser22 emits monochromatic radiation, a different laser emitting radiation of a wavelength different from that emitted bylaser22 may be employed to replacelaser22. As another alternative, separate monochromatic or polychromatic lasers may be combined by means such as dichroic filters to provide radiation of selectable wavelength. Such and other variations are within the scope of the invention.
In a system for enhanced detection sensitivity, it is desirable for the collection optics such as objective[0036]32 to have a large numerical aperture (NA) whereas for the illumination optics such aslens30, a low NA will be sufficient.System20 illustrated in FIG. 1 shows a particularly compact design where the illumination optics and collection optics employ different objectives, that isobjectives30 and32, where thecollection objective32 has a larger NA than that of theillumination objective30. By using low NA illumination optics, it is possible for both the illumination optics and collection optics to fit within the space close to thewafer28 in a particularly compact design of the optical head, as shown in FIG. 1.
Thus, according to another aspect of the invention, the optical head in the embodiment of FIG. 1 is compact and has a particularly small footprint. Thus,[0037]optical head60 within the dotted line box includes alaser22,diffractive element26a,lens30,collection objective32, and the array ofoptical fibers34. In a slightly modified embodiment than that shown in FIG. 1,laser22 may also be located outside theoptical head60 and may be placed so that its output laser beam is directed to thediffractive element26ain theoptical head60, possibly by means of an optical fiber link. Such and other variations are within the scope of the invention.
The[0038]collection objective32 focuses radiation scattered from eachilluminated spot42 on the surface ofwafer28 to a corresponding optical fiber in theoptical fiber array34. Information related to the scattered radiation from each spot is then carried by its corresponding fiber to a two-dimensional diode array where the diodes may be avalanche photodiodes. Alternatively, individual fibers may carry signals to individual avalanche photodiodes photomultipliers, photodiodes or other types of individual detectors. By using anoptical fiber array34, thedetector array36 does not need to be included in theoptical head60 and can be located at a distance from the optical head, thereby further reducing the size of the optical head. Alternatively, for applications where spatial considerations are not as important, theoptical fiber array34 may be omitted and the scattered radiation from each spot is focused directly by objective32 to a corresponding detector in thedetector array36 within the optical head. Such and other variations are within the scope of the invention.Lens38 focuses the scattered radiation from aspot42 to the corresponding fiber within theoptical array34.
In the above description,[0039]element26adiffracts the laser beam fromlaser22 into a two-dimensional array ofbeams24. Instead of diffracting the laser beam into a two-dimensional array of beams,element26amay instead diffract the beam into a one-dimensional array of beams to illuminate a one-dimensional array of illuminated spots on the surface of thewafer28. Such one-dimensional array of illuminated spots may, for example, comprise the five illuminated spots appearing as the rightmost column42′ in FIG. 2. Another example of such one-dimensional array of beams and spots is illustrated in FIG. 9. The paths of illuminated spots incolumn42′ may also overlap as indicated in FIG. 4. Such and other variations are within the scope of the invention.
Bright Field DetectionBright field detection is where specularly reflected radiation is detected, such as that described in S. Stokowski and M. Vaez-Iravani, “Wafer Inspection Technology Challenges for ULSI Technology”, Proceedings of conference on Characterization and Metrology for ULSI Technology, Edited by D. G. Seiler, A. C. Diebold, W. M. Bullis, T. J. Shaffner, R. McDonald, and E. J. Walters, American Institute of Physics, PP. 405-415 (1998)[0040]
In the embodiment of FIG. 1, the array of illumination beams[0041]24 are focused bylens30 to a mirror62 which reflects the beams towardswafer28. Mirror62 also acts as an aperture stop to reduce or prevent specular reflection of the beams from thesurface28 from reaching theoptical fiber array34, so that thecollection mirror32 collects only radiation scattered by the spots along collection paths that are away from the specular reflection direction in a dark field imaging system. Dark field systems are those where the radiation collected and detected is that scattered by the sample and collected along collection paths that are away from the specular reflection direction from the sample surface of the illumination beams. Dark field systems are explained in more detail in the above-referenced article by S. Stokowski and M. Vaez-Iravani.
FIG. 1 also shows the reflected path into the “bright-field”[0042]channels70, which may comprise an optical fiber array similar toarray34. Thebeams24 fromelement26aare first reflected by abeam splitter66 towardslens30 and mirror62. Part of the radiation specularly reflected by the wafer surface is again reflected by mirror62, collimated bylens30 and passes through thebeam splitter66 towards bright-field channels70 and then to an array of detectors (not shown). As in the case of dark field detection, the radiation reflected from each spot is imaged bylens30 onto a corresponding channel inchannels70 and then to a corresponding detector. Also as in the dark field system, the detector array in the bright field system need not be included inoptical head60 for compactness. Where space is not as much a concern,channels70 may be replaced by an array of detectors so thatlens30 and simple optics (not shown) located downstream fromlens30 in the same optical path image radiation reflected by each spot directly to the corresponding detector in the detector array.
The bright-field channels may yield useful information on large defects that can be discerned by detecting the reflectance at various spots on the surface of[0043]wafer28. If bright-field inspection at the proposed resolution is found to be useful, then the appropriate fiber channels can be set up in exactly the same manner as those in the case of the dark-field channels, where a detector array in addition toarray36 is employed. Bright-field and dark-field radiation could also be detected sequentially using the same electronics. Alternatively, they may be used simultaneously using separate electronics.
Wafer Scanning[0044]Wafer28 is supported on a chuck (not shown) which is rotated by means of amotor72 and translated in a direction bygear74 so that the illuminatedspots42 are caused to move and trace a swath of spiral paths on the surface ofwafer28 to inspect the entire surface of the wafer. Both vacuum handling and edge handling of the samples are possible.Motor72 andgear74 are controlled bycontroller76 in a manner known to those skilled in the art. Thus, in the preferred embodiment, theoptical head60 remains stationary and the scanning of thebeams24 across the surface of thewafer28 is accomplished by usingmotor72,gear74 andcontroller76 to move the wafer so that the entire surface of the wafer is scanned. Alternatively, theoptical head60 may be caused to move in a manner known to those skilled in the art to trace the spiral path or another type of scan path for scanningwafer28. The wafer may also be scanned along substantially linear paths using XY stages.
As noted above, the detector in[0045]array36 may be a photodiode such as an avalanche photodiode; alternatively, it may be a photomultiplier tube. The output of each detector in thedetector array36 is supplied toprocessing circuit82 where the circuit may comprise a microprocessor, hardware logic or programmable logic circuits, such as those using FPGA's or dynamic logic.Circuit82 may be a part of or connected to acomputer84 that is in communication withcontroller76, so that scattered radiation from a particular detector inarray36 can be matched with a location on the surface of thewafer28. Where processingcircuit82 is a microprocessor, it can be a co-processor withincomputer84. Processingcircuit82 stores the outputs ofdetector array36 and processes such signals, such as by comparing signals in a die-to-die operation for detecting anomalies. Alternatively, processingcircuit82 may perform certain initial processing of the signals, such as signal amplification and conversion from analog-to-digital form and passes the digital signals tocomputer84 to perform further processing such as die-to-die comparison.
One aspect of the design in[0046]system20 of FIG. 1 is that it is based on a stationary optics as described above, and R/theta spinning of the wafer as described above, in a manner similar to that in the SP1™ tool, also available from KLA-Tencor Corporation of San Jose, Calif. It is preferable forsystem20 to have a rather precise spinning action. For example, the spinner is capable of some +/−15 microns stability in height, and uniform spinning on the micron scale. This performance can be achieved by means of an air bearing stage, for example. By scanning the wafer surface with multiple spots simultaneously, the scanning of the entire wafer surface can be performed in shorter time.
An important consideration that pertains to this spiral scanning action is that it begins to deviate from very closely linear motion as the position of the beams approaches the center of the wafer. However, it can be shown that, by ramping the rotation rate down toward the center, this issue can be resolved. It should also be remembered that at any given time, one has a precise (within a pixel) knowledge of the position of any of the beams, which allows one to correct for such scan deviations.[0047]
Filters for Reducing Diffraction from Manhattan Geometry and from PatternAt any given position on the[0048]wafer28 during the beam scanning and inspection process, each of thespots42 illuminates a number of shapes, which primarily lie along the Manhattan geometry. These shapes all generate a two dimensional Sinc function, but with different phases, giving rise to a “+” speckle pattern. As the wafer rotates, this pattern also rotates. If one were to detect all the available scattered radiation from the wafer, one would also receive parts of this diffraction pattern. In the ensuing die-to-die comparison, the presence of this large background would possibly result in significant errors.
In rectilinear scans, one could resolve these problems by means of a stationary spatial filter to filter out the speckle pattern and placing the detectors along the 45 degrees lines with respect to the horizontal-vertical directions. In[0049]system20, the rotation of thewafer28 results in a rotating diffraction pattern. This pattern can be eliminated or reduced by placing a “+” shaped filter90 (i.e. a filter having an aperture that passes radiation except for a “+” shaped area), shown more clearly in FIG. 5. directly above the illuminated field, in the path of the radiation after emergence from thereflective objective32. Thisfilter90 is made to rotate by means of amotor89 in FIG. 1 under the control ofcomputer84 in unison with respect to the rotation ofwafer28 under the control ofcontroller76, resulting in a continuous cancellation of the diffraction pattern. Possible approaches to this issue include the use of ball-bearing based systems, which can be mounted directly at the exit port of the reflecting objective. The use of a programmable liquid crystal filter having an aperture that is changed in synchronism with the rotational motion of the sample surface under the control ofcomputer84 to implementfilter90 instead of a mechanically rotated one as described above may be viable for low rotation rates of the wafer.
In addition to diffraction from the Manhattan Geometry, the presence of any periodic structures such as arrays in DRAMs on the surface of the wafer may also give rise to a two-dimensional Fourier components when illuminated with normal incidence radiation. If the directions of the expected pattern scatter from the surface are known, spatial filters may be designed to block such scattering, thereby detecting only the scatter by anomalies on the surface. FIG. 6 is a schematic view illustrating the two-dimensional Fourier components of an array structure that is periodic in the X and Y directions when illuminated with normal incidence radiation. As the sample rotates, all of the spots at the intersections of the X-Y lines will rotate, thereby generating circles[0050]91. These circles represent the loci of the Fourier components as the wafer is rotated. The dark opaque circle at the center is the blockage of the collection space caused by stop62 in FIG. 1.
From FIG. 6, it is noted that there are gaps in between the circles where there are no Fourier components. At least in theory, it is possible to construct a programmable filter (e.g. a liquid crystal filter) in which annular bands of arbitrary radii are blocked out. A simple spatial filter may be constructed also to achieve many of the objectives herein. Thus, if the cell size of a regular memory array on the wafer is such that its X and Y dimensions are not larger than about 3.5 microns, for example, this means that for 488 nanometers wavelength radiation used in the illumination beams[0051]24, the first Fourier component is at about 8° to thenormal direction36. Therefore, if a spatial filter such as92 of FIG. 1 is employed, blocking all collected radiation in the narrow channel that is at 8° or more to thenormal direction36 will leave an annular gap of 2 or 3° ranging from the rim of the central obscuration (i.e. 5 or 6°) to the rim of the variable aperture at about 8°. Under these conditions, as the wafer spins, no Fourier components can possibly get through the annual gap and scatter from the array is suppressed. In one embodiment, thespatial filter92 in FIG. 1 used leaves an annular gap between about 5 to 9° from thenormal direction64 to thesurface28 of the wafer in FIG. 1. For DRAM structures of smaller cell sizes, such annular aperture type spatial filter may not even be necessary. While bothfilters90 and92 are employed in the embodiment of FIG. 1, it will be understood that for certain applications, the use of only one of the two filters may be adequate and is within the scope of the invention.
It will be noted that even though the[0052]collection objective32 focuses radiation scattered from an array ofspots42, such scattered radiation from the spots are focused towards theoptical fiber array34 through a small area at the focal plane of the objective, so that by placingfilter90 and/or filter92 at or close to the focal plane, the above-described effects can be achieved with respect to the scattered radiation from all of the illuminatedspots42 in the array of spots.
Detection ChannelsIndividual APD's may be used as detectors in[0053]array36 for each of the dark-field channels. These detectors provide close to shot noise limited performance. If bright-field channels are considered important, then a separate APD board may be provided for those, or an array of PIN diodes.
Each APD channel has its own voltage setting, and analog-to-digital converter (ADC), which can be operated at up to 60 MHz. That is, the potential of this system in terms of data rate approaches 5-10 GHz, even though a practical data rate may be somewhat lower. It is important to note that the detection electronics part of the design in this case may be completely separate from the front-end optics, such as the[0054]optical head60, which necessarily results in a simple, compact, and robust design. The optical head may be readily integrated intosemiconductor processing equipment88, so that it is more convenient for anomalies on the wafer surface to be detected during processing or between processing steps by means ofsemiconductor processing equipment88.
Processing Circuit82Preferably the detected signals are directed into a massive bank of random access memory (RAM) in[0055]circuit82, capable of holding up to 85 Gbytes of data. As the wafer is scanned the data are gathered from the various dice at different locations on the wafer. Subsequent image processing is primarily based on a die-to-die comparison process, applied to side-by-side dice, in a rectilinear direction, much in the same way as that in conventional systems, such as the AIT™ systems available from KLA-Tencor Corporation of San Jose, Calif.
Because of the fact that the scanning is performed in a spiral rather than rectilinear fashion, the die-die comparison may be performed on a stored version of the 12-bit gray scale data. To achieve this, it will not be necessary to store the data from the entire wafer, rather only a sufficient amount to enable die-die on the present location. Nevertheless, for some applications, it may be desirable to provide sufficient memory to store an entire wafer map. At a pixel size of 1.25×1.25 microns, a 300 mm wafer has approximately 45 gigapixels. To store all pixels as 12-bit values, some 70 GB of memory is preferred. The processing power required must be sufficient to keep up with the pixel rate. A typical pixel rate for some embodiments can be about 1 Gpixels/sec. Higher speeds are also possible.[0056]
In one embodiment where the scanning is non-rectilinear, it may not be possible simply to retain the image data for a single swath in order to do die-to-die comparison, as the AIT and other rectilinear scanning die-to-die machines do. However, by retaining all pixel information on the wafer as it comes in, and by concurrently comparing incoming pixels with those of a reference die which is chosen so that its pixels are acquired a little sooner, each die can be compared with a reference die during the wafer scan; when the spiral scan is complete, the processing will be nearly finished.[0057]
As noted above, a reference die may be chosen so that its pixels are acquired a little sooner, so that each die can be compared with a reference die during the wafer scan. This is illustrated in FIG. 7 which is a schematic illustration of data obtained from an[0058]annular region94 of the wafer. The wafer may be scanned beginning at a point on or near the circumference of the wafer, or at or near the center of the wafer. Assuming that the spiral path scan of the array ofspots42 starts out at the circumference of the wafer and spirals in towards the center of the wafer during the scanning, areference die96 may be defined at or close to the outer circumference of theannular region94. Therefore, when the data from the target die98 is obtained, such data may be compared to the data in the reference die obtained earlier for anomaly detection. Obviously, die-to-die comparison using dice data acquired earlier from a reference dice different fromdice96 may be used instead and is within the scope of the invention.
FIG. 8 is a schematic diagram of an optical inspection and imaging system illustrating an alternative embodiment of the invention. Instead of using two separate objectives, one for illumination and the other for collection, the[0059]embodiment100 of FIG. 8 employs a single objective for this purpose, although different portions of the objective102 may be employed for illumination and for collection. Thus, as shown in FIG. 8, a laser beam fromlaser22 is reflected by a mirror orbeamsplitter66 and is diffracted into an array ofbeams24 by means ofdiffractive element26a.Beams24 are reflected by a centerreflective portion104aofbeamsplitter104 tolens102. In the preferred embodiment, beams24 are focused by acenter portion102aof the aperture oflens102 to the surface ofwafer28. Scattered radiation from the illuminatedspots42 are collected bylens102 and directed towards thebeamsplitter104. The centerreflective portion104aacts as an aperture stop that prevents specular reflection from the surface of the wafer from reaching theoptical fiber array34. Thus, only the scattered radiation collected by thecircumferential portion102bof the aperture oflens102 can pass through thebeamsplitter104 and focused bylens38 towards theoptical fiber array34. Preferably, theportion102bfor collecting scattered radiation is larger than theportion102aused for illumination, which enhances the sensitivity of detection.
To simplify FIG. 8, the components shown in FIG. 1 for moving the wafer, the bright field channels, the processing circuit and computer have been omitted from the figure. The embodiment of FIG. 8 has the advantage that it is even more compact compared to the embodiment of FIG. 1, since a single objective is used for both illumination and collection. Instead of using a lens as shown in FIG. 8, it may also be possible to use a reflective objective to ensure easy operation and a large wavelength range. Instead of using a[0060]center portion102afor focusing the illumination beams and acircumferential portion102bfor collecting the scattered radiation, the arrangement in FIG. 8 can also be modified by directing the illumination beams24 through a side portion of the objective, such as the left side of the objective102 and using the other side, such as the right side, for collection of the scattered radiation. Such and other variations are within the scope of the invention. It will be noted that where the paths of illumination beams are at oblique angles to the surface ofsample28, at least one dimension of the illuminated spots may be such that it is not less than about 5 microns.
In the embodiment of FIG. 1 described above, the illumination beams[0061]24 are directed towards the wafer surface in directions that are substantially normal to the surface of the wafer. This is not required, however. Thus, for some applications, the illumination beams may be directed towards a wafer surface at an oblique angle such as along thepaths24aindicated by the dotted line in FIG. 1, so that at least one dimension of the illuminated spots is not less than about 5 microns. Thus, especially for the inspection of unpatterned surfaces such as unpatterned wafers, illuminating the wafer surface at an oblique angle may be desirable for some applications.
FIG. 9 is a top schematic view of an optical inspection and imaging system to illustrate yet another embodiment of the invention. As shown in FIG. 9, a single line of illumination beams is supplied at an oblique angle along[0062]direction202 to the surface of awafer28, only a portion of which is shown in FIG. 9. The single line of illumination beams (not shown) illuminate a single file of elongated illuminatedspots204 on the surface of the wafer. Preferably, the beamsplitter (not shown in FIG. 9) that is used to generate the single file of illumination beams is oriented at or near an angle, such as 45° for example, relative to the plane of incidence of the illumination beams, so that the line204aconnecting the centers of thespots204 is also at or near 45° with respect to the plane of incidence. In this context, the plane of incidence is defined by a plane containing theillumination direction202 and a line203 (pointing out of the plane of the paper) intersectingdirection202 and normal to the surface of thewafer28. Thus, ifdirection202 is regarded as an axis of a coordinate system, the line204aconnecting the centers of thespots204 is substantially at +45° to such axis.
Radiation scattered from the[0063]spots204 are collected along directions substantially perpendicular to line204abyobjectives210 and212 located above the plane of the surface inspected and on opposite sides of line204a. Objective210 images the scattered radiation from eachspot204 onto a corresponding forward channel or detector in theoptical fiber array34′ ordetector array36′. Similarly, objective212 images the scattered radiation from eachspot204 to its corresponding backscatter fiber or detector in thefiberoptic array34″ ordetector array36″. It will be noted thatobjectives210 and212 may be situated so that all of thespots204 are substantially within their focal planes. As shown in FIG. 9, objective212 will collect the forward scattered radiation and objective210 will collect the back scattered radiation. Instead of using lenses as shown in FIG. 9,objectives210 and212 may also be reflective objectives.
Alternatively, the beamsplitter that is used to generate the single file of illumination beams may be oriented at −45° with respect to the plane of incidence so that the spots (not shown in FIG. 9) would form a single file oriented at −45° to[0064]direction202, andline204bconnecting the centers of the spots at such new locations is also at substantially −45° with respect to the plane of incidence.
If the beamsplitter for generating the array of illumination beams are oriented at 45° with respect to the plane of incidence, then the[0065]collection objectives210 and212 would also need to be rotated by 90° so that thespots204 arranged with their centers along theline204bwould again be within their focal planes and that these objectives would again collect radiation scattered in directions substantially perpendicular to theline204b.
Instead of collecting and imaging scattered radiation in directions perpendicular to the line joining the centers of the illuminated[0066]spots204 as described above, it is also possible to collect and image the scatter radiation in a double dark field configuration. In such configuration, the two objectives would be at locations indicated indotted lines210′ and212′ where scattered radiation is collected substantially at +90 and −90 degrees azimuthal angle relative to the illumination beams as they reach the surface; the fiber channels or detectors have been omitted in such configuration to simplify the figure. In a double dark field configuration, different spots along theline204aor204bwill be located at different distances from the objectives so that at least some of them will be out of focus. Even though some of thespots204 will be out of focus or somewhat out of focus, this may not have significant adverse effects on some applications, such as unpatterned surface inspection. Obviously only one of the twoobjectives210 and212 (or210′ and212′) may suffice for some applications, so that one of them can be omitted. It is also possible to placecollection optics210″ and adetector array36 or a collection of individual detectors (not shown) directly above the area of the surface of sample (and therefore in the plane of incidence of beams along direction202) inspected in a single dark field configuration to detect surface anomalies, such as in the configuration shown in FIG. 10. When thecollection optics210″ is in such position, it images to the detector array or detector collection the scattered radiation in at least one direction that is substantially normal to the surface. Preferably thecollection optics210″ used has a large numerical aperture for increased sensitivity.
If the illumination beams are polarized, it may be desirable to insert a polarizer between each of the two[0067]objectives210 and210′ and their corresponding fiber or detection channels. Thus, in the presence of a dielectric material such as silicon oxide, circularly polarized radiation in the illumination beam may be preferable. The presence of small defects may cause P-polarized radiation to be more efficiently scattered. If S-polarized radiation is employed in the illumination beams, scattering caused by the presence of roughness on the surface can be further suppressed if only S-polarized light is collected. For this purpose polarizers may be placed in the paths ofbeams24 andpolarizers220 and222 may be placed in the collection path for enabling the detection of polarized radiation components, which may in turn indicate the type of anomalies present on the wafer. Corresponding polarizers may be placed along the collection paths in the double dark field embodiments. Instead of using refractive objectives such aslenses210,210′,212,212′, reflective objectives may be used which can be used for collection over a large wavelength range.
While the invention has been described above by reference to various embodiments, it will be understood that changes and modifications may be made without departing from the scope of the invention, which is to be defined only by the appended claims and their equivalents. For example, while the embodiments are illustrated with respect to wafer anomaly detection, the invention may be used for anomaly detection on other types of surfaces as well, such as flat panel displays, magnetic and optical heads, disks and so on. All of the references referred to above are incorporated herein by reference in their entireties[0068]