US RE41924 E1
A system and method for inspecting machine readable marks on one side of a wafer without requiring transmission of radiant energy from another side of the wafer and through the wafer. The wafer has articles which may include die, chip scale packages, circuit patterns and the like. The marking occurs in a wafer marking system and within a designated region relative to an article position. The articles have a pattern on a first side. The method includes the steps of imaging a first side of the wafer, imaging a second side of the wafer, establishing correspondence between a portion of first side image and a portion of a second side image, and superimposing image data from the first and second sides to determine at least the position of a mark relative to an article.
1. A method for inspecting machine readable marks on one side of a wafer without requiring transmission of radiant energy from another side of the wafer and through the wafer, the wafer having articles which may include die, chip scale packages, circuit patterns and the like, the marking occuring in a wafer marking system and within a designated region relative to an article position, the articles having a pattern on a first side, the method comprising:
imaging a first side of the wafer;
imaging a second side of the wafer;
establishing correspondence between a portion of first side image and a portion of a second side image; and
superimposing image data from the first and second sides to determine at least the position of a mark relative to an article.
2. The method of
3. The method of
4. A system for inspecting machine readable marks on one side of a wafer without requiring transmission of radiant energy from another side of the wafer and through the wafer, the wafer having articles which may include die, chip scale packages, circuit patterns and the like, the marking occurring in a wafer marking system and within a designated region relative to an article position, the articles having a pattern on a first side, the system comprising:
means for imaging the first side of the wafer to obtain an image;
means for imaging a mark on the second side of the wafer to obtain an image;
means for establishing correspondence between a portion of a first side image and a portion of a second side image; and
means for superimposing image data from the first and second sides to determine at least the position of the mark relative to an article.
5. The system as claimed in
This application claims the benefit of U.S. provisional application Ser. No. 60/381,602, filed May 17, 2002.
1. Field of the Invention
The invention relates to laser marking of workpieces, including semiconductor substrates, wafers, packages and the like. The invention is particularly adapted for, but not limited to, marking machine readable codes on a second side of semiconductor wafers which have a high density circuit patterns on a first side, for instance chip scale packages having a high density of interconnects which could be damaged by a marking beam, or where space for codes is limited.
2. Background Art
The following representative patent references relate to various aspects of laser marking of wafers and electronic assemblies, illumination, and inspection/reading marks: U.S. Pat. Nos. 4,522,656; 4,945,204; 5,329,090; 6,309,943; 6,262,388; 5,929,997; 5,690,846; 5,894,530; 5,737,122; and Japanese Patent Abstract 11135390.
The following representative references provide general information on various laser marking methods and system configurations and components: “Galvanometric and Resonant Low Inertia Scanners”, Montagu, in Laser Beam Scanning, Marcel-Dekker, 1985, pp. 214-216; “Marking Applications now Encompass Many Materials”, Hayes, in Laser Focus World, February 1997, pp. 153-160; “Commercial Fiber Lasers Take on Industrial Markets”, Laser Focus World, May 1997, pp. 143-150. Patent Publications: WO 96/16767, WO 98/53949, U.S. Pat. Nos. 5,965,042; 5,942,137; 5,932,119; 5,719,372; 5,635,976; 5,600,478; 5,521,628; 5,357,077; 4,985,780; 4,945,204; 4,922,077; 4,758,848; 4,734,558; 4,856,053; 4,323,755; 4,220,842; 4,156,124.
Published Patent Applications WO0154854, publication date 2 Aug. 2001, entitled “Laser Scanning Method and System for Marking Articles such as Printed Circuit Boards, Integrated Circuits, and the Like” and WO0161275, published on 23 Aug. 2001, entitled “Method and System for Automatically Generating Reference Height Data for use in a Three-Dimensional Inspection System” are both assigned to the assignee of the present invention. Both applications are hereby incorporated by reference in their entirety.
U.S. Pat. No. 6,309,943 relates to identifying and determining a position of a scribe grid on a front-side surface of a wafer with a camera. Based on this information, a laser is fired to form an alignment mark on the back-side surface of the wafer.
U.S. Pat. No. 6,496,270, assigned to the assignee of the present invention, describes a method and system for automatically generating reference height data for use in a 3D inspection system wherein local reference areas on an object are initially determined and then the height of these local reference areas are determined to generate the reference height data.
The WH-4100 Laser Marking System is a commercially available backside laser marking system produced by the assignee of the present invention. A fine alignment vision subsystem corrects rotational or offset errors (X, Y, Angle) which are introduced when a wafer is placed in the marking station. A manual “teach tool” allows the user to train the system to recognize three non-collinear points on the wafer that is to be used for the correction. An iterative trial and error process with various adjustments, and manual evaluation of the results is required with the system. The information is then used to determine mark locations on the bottom side of the wafer.
U.S. Pat. Nos. 5,894,530 and 5,929,997 relate to viewing systems used for inspection and/or alignment operations in microelectronics. In the '530 patent, optical elements are selectively positioned such that images of indicia fields disposed on either side of a substrate can be viewed (at the same magnification) whenever the substrate is in a given orientation, or such that images of indicia fields disposed on both sides of the substrate may be viewed at the same magnification, simultaneously.
With increasing density and complexity of circuitry on semiconductor devices, (e.g., exceeding ten-thousand die on a 300 mm wafer), multiple designs and layout of circuitry, there is an on-going need to provide advanced interactive tools for improving throughput and yield.
An object of the present invention is to provide an improved method and system for machine vision-based feature detection and mark verification in a workpiece or wafer marking system.
In carrying out the above object and other objects of the present invention, a precision laser based method of marking a semiconductor wafer having articles which may include die, chip scale packages, circuit patterns and the like. The marking occurs in a wafer marking system and within a designated region relative to an article position. The method includes determining at least one location from which reference data is to be obtained using (a) information from which a location of an article is defined, and (b) a vision model of at least a portion of at least one article. Reference data is obtained to locate a feature on a first side of the wafer using at least one signal from a first sensor. The method further includes positioning a marking field relative to the wafer so as to position a laser beam at a marking location on a second side of the wafer. The positioning is based on the feature location. A predetermined pattern is marked on the second side of the wafer using a laser marking output beam. The step of determining includes: measuring at least one feature in an image obtained from a first wafer portion; relating the measured feature to a wafer map; and storing the data for use when marking wafers substantially identical to the first wafer. The steps of measuring, relating, and storing are performed automatically.
The step of measuring may include measuring the average pitch of a plurality of articles and relating the average pitch to a wafer map.
The articles may comprise a row-column pattern of die, and the step of determining may further include: locating a pair of orthogonal edges of the row-column pattern; forming bounding boxes from the edges and; defining a die pattern coordinate system from the bounding boxes.
The relative positioning of the wafer may be carried out in a primary coordinate system substantially aligned with the movement of at least one positioner. The method may further include transforming coordinates to relate the primary coordinate system with the die pattern coordinate system.
The step of determining may still further include: obtaining a coordinate using a wafer map to provide the information from which a location of the article is defined; and imaging at least a portion of an article on a first wafer to generate the vision model.
Still further in carrying out the above object and other objects of the present invention, a method is provided for inspecting machine readable marks on one side of a wafer without requiring transmission of radiant energy from another side of the wafer and through the wafer. The wafer has articles which may include die, chip scale packages, circuit patterns and the like. The marking occurs in a wafer marking system and within a designated region relative to an article position. The articles have a pattern on a first side. The method includes imaging a first side of the wafer, imaging a second side of the wafer, establishing correspondence between a portion of first side image and a portion of a second side image, and superimposing image data from the first and second sides to determine at least the position of a mark relative to an article.
The system may further include substantially matching images obtained from the first and second sides so that the superimposed image portions correspond. The step of substantially matching may be carried out using a calibration target and a matching algorithm.
The method may further include providing an input using the user interface so as to cause a region of interest to be defined within at least a portion of an image of an article.
The region of interest may be operator adjustable.
The superimposed data may be used to determine the position of a mark relative to the article.
The method may further include providing an inspection station having a wafer positioning subsystem separated from a positioning subsystem used for marking.
Yet still further in carrying out the above object and other objects of the present invention, a precision laser based system of marking semiconductor wafers, the wafer having articles which may include die, chip scale packages, circuit patterns and the like. The marking occurs in a wafer marking system and within a designated region relative to an article position. The system includes means for determining at least one location from which reference data is to be obtained using (a) information from which a location of an article is defined and (b) a vision model of at least a portion of at least one article. The system further includes means for obtaining reference data to locate a feature on a first side of a wafer using at least one signal from a first sensor. The system further includes means for positioning a marking field relative to the wafer so as to position a laser beam at a marking location on a second side of the wafer. The positioning is based on the feature location. The system still further includes means for marking a predetermined pattern on the second side of the wafer using a laser marking output beam. The means for determining measures at least one feature in an image obtained from a first wafer portion, relates the measured feature to a wafer map, and stores the data for use when marking wafers substantially identical to the first wafer. The measuring, relating, and storing are performed automatically by the means for determining.
Still further in carrying out the above object and other objects of the present invention, a system is provided for inspecting machine readable marks on one side of a wafer without requiring transmission of radiant energy from another side of the wafer and through the wafer. The wafer has articles which may include die, chip scale packages, circuit patterns and the like. The marking occurs in a wafer marking system and within a designated region relative to an article position. The articles have a pattern on a first side. The system includes means for imaging the first side of the wafer to obtain an image, means for imaging the mark on the second side of the wafer to obtain an image, means for establishing correspondence between a portion of a first side image and a portion of a second side image, and means for superimposing image data from the first and second sides to determine at least the position of a mark relative to an article.
At least one of the means for imaging may include a zoom lens.
The means for establishing correspondence may include a calibration target and an algorithm.
Yet still further in carrying out the above object and other objects of the present invention, a laser based system is provided for laser marking of substrates such as semiconductor wafers or similar substrates with a laser marking beam. The substrates have a repetitive pattern of articles arranged in rows and columns. Each of the articles have a feature detectable with an imaging subsystem. The system has a laser marking head, the imaging subsystem for imaging and measurement, a motion subsystem having a stage for positioning at least the substrate relative to the imaging subsystem, and a user interface connected at least to the imaging subsystem and motion subsystem. Laser marks are to be placed at predetermined locations relative to the articles. A method of laser marking with beam position control using predetermined pattern features is further provided. The method includes providing, through the user interface, an input so as to cause a portion of the pattern to be identified for automatic feature detection and measurement with a machine vision algorithm. The method further includes positioning a first substrate relative to the imaging subsystem automatically to traverse the pattern along at least one of a row or column of the pattern so as to acquire image data at a first set of feature locations. A dimension is measured using at least one detectable feature of a plurality of articles, the algorithm, and the image data. Dimensional data is stored based on the measurement. At least three feature locations of a second set of feature locations are determined relative to the pattern using the dimensional data. The feature locations of the second set suitably define a relationship between a pattern coordinate system and a stage coordinate system. The first substrate is removed and a second substrate is positioned to be marked relative to the imaging subsystem. At least three corresponding feature locations of the second set of feature locations are located in image data obtained from the corresponding pattern on the second substrate. Coordinates of the pattern on the first substrate are related to the corresponding pattern on the second substrate. The substrate is positioned relative to the marking beam based on at least the three feature locations of the second set to mark the substrate.
A first estimate of the dimension may be obtained by semiautomatic relative positioning of the substrate and the imaging subsystem over a substantially small area of the pattern, and further includes identifying a feature in a displayed image, and communicating the image location of the feature using the user interface.
The substrate may be a semiconductor wafer, the articles are die, and a feature is a corner of the die.
The dimensional measurement may be the average die pitch measured over a substantial number of die along at least one of a row and column.
The average die pitch may be related with a wafer map.
The pattern coordinate system may have an origin defined relative to a boundary of the pattern.
The step of determining may include searching for pattern locations, and searching may be carried out by controlling the stage based on pattern system coordinates.
The step of providing may further include generating a vision model using an image of a portion of the pattern.
The above object and other objects, features, and advantages of the present invention are readily apparent from the following detailed description of the best mode for carrying out the invention when taken in connection with the accompanying drawings.
FIG, 5A is a schematic diagram showing certain sub-systems of a laser marking system for semiconductor wafers for use in a production system;
Several components of a system 100 for laser marking and inspection of wafers, for instance 300 mm wafers, is schematically illustrated in
A marking sequence, following opening of a FOUP, includes:
An exemplary 300 mm wafer may have several thousand articles (e.g.: chip scale packages, integrated circuits). The density of the circuitry on each article can lead to difficulty in placing machine readable marks, such as 1-dimensional or 2-dimensional codes, in restricted areas. For instance, the die size on a 300 mm wafer may vary from about 25 mm to 0.5 mm or smaller, with dense, complex circuit patterns. Further, damage to circuitry which might be caused by a high energy marking beam is to be avoided. WO0154854, assigned to the assignee of the present invention and hereby incorporated by reference in its entirety, discloses a method of high resolution marking of electronic devices. Laser mark registration is obtained from circuit features measured with a sensor, and in one embodiment the sensor is located disjoint from a marking head. Examples are included in '854 for marking of PCB multi-ups and packages such as chip scale packages and die in a tray. Sections of the '854 disclosure, including: page 4, lines 9-16, page 6, lines 1-5 and 22-29, Page 8, lines 10-17, page 9, line 15-page 10, line 30, page 11, lines 14-20 and the sections in the detailed description entitled “scan head”, “marking operation”, and “registration” and the associated drawings of the sections are related to the present disclosure and provide additional support for various aspects of precision marking methods and systems disclosed herein.
In a preferred embodiment for marking large (e.g.: 300 mm) wafers a calibration process will be used to relate an alignment vision system coordinate (e.g., a “first side” position, for instance at the sensor center position, and at best focus) and beam positioning sub-system coordinate (e.g.: laser beam waist position at the center of a marking field). Preferably, the calibration will provide three-dimensional correction. The increasing demand for precision placement of marks in localized areas over a large field lead to increasing beam positioning accuracy and decreasing spot size requirements for obtaining finer line widths or character sizes. Over a large workpiece “sag” and warpage may be significant relative to the depth of focus, which introduces conflicting design parameters. Preferably, a laser spot size can be adjusted during system operation while maintaining spot placement accuracy. With reference to the arrangement of
The three-dimensional calibration process of SECTION 1 of the Appendix, with suitable height measurements of the wafer, may be used to determine a correction to be applied to the beam positioner.
In a preferred telecentric system, the error is reduced to about 1 spot or finer with a lens (see SECTION 5 which follows entitled “Precision Telecentric Lens”) of low to moderate cost. Preferably, the telecentric design compensates for the worst case wafer warpage and additional system “stackup” errors. With the preferred arrangement a field size supporting relatively high marking speeds is maintained. In the telecentric case the calibration process may be streamlined, but multiple calibration files used to at least control and maintain the laser spot size over the working volume are preferred. This provides for consistent marks and for mark contrast control.
Three-dimensional tolerances are to be considered for the alignment and marking sub-systems in view of the workpiece variations relative to the depth of focus of the optical systems. Increasing the alignment system magnification to improve feature location accuracy decreases the depth of focus. Various focusing methods are useful to position the entire sub-system 14 and/or lens system 15 (shown as a telecentric lens but not so restricted) relative to the workpiece along the Z-axis. For example, the Z-axis position corresponding to the maximum edge contrast at a die location is a possible measure. A measurement of the maximum intensity of a “point” or small target (one the order of a pixel) may provide more sensitivity to depth changes.
Wafer “sag” is somewhat predictable from a specification of wafer thickness. Predictions based on models (fixed edge and simple support) with wafer thickness ranges of about 300 μm to 775 μm indicated about 60 μm of deviation for the latter case. For thinner wafers the deviation increases, and the overall deviations may be further increased by warpage and other stackups. Surface deviations may be estimated and used for certain correction. A telecentric system, for instance as described in SECTION 5, is predicted to yield less than 1 μm of spot placement error over a 4″ marking field. Various sub-systems, including the scan head, alignment vision system, and perhaps inspection system may include at least an option for height sensing. Similarly, a separate sub-system could be added specifically for height measurements at a plurality of locations on the wafer surface. Preferably, any degradation in the cycle time of the machine will be negligible.
In one arrangement the alignment vision system 14 will be relatively positioned at sample points which may include but are not limited to the regions used for feature detection. As mentioned earlier, the focus sensing may be achieved by sampling the image contrast at locations along the z-axis using the alignment vision system. The z-axis locations are recorded. Alternatively, a triangulation or focus sensor, which may be a commercially available module, may be used for measuring surface points which are used with the alignment and calibration algorithms (and the known wafer thickness) to map the surface. Similarly, a direct measurement of the second side may be obtained with a sensor included with the vision inspection module 20. In an alternative arrangement a “full field” system, for instance a commercially available Moire Camera, may be used. In any case, the data will preferably be used to position the marking beam waist at the surface. In accordance with the preferred calibration method of SECTION 1, the desired spot size will be maintained at the marking locations. In one arrangement the marking beam waist may be positioned in discrete steps, for instance at 9 locations within an 80 mm field for center-edge compensation. Non-contact optical sensing is preferred, but capacitance or touch probes may be acceptable.
If the deviations correspond to a simple second order curve and are symmetric, then the wafer surface may be sampled along a diagonal direction using at least three locations (edge region, center, edge region). If warpage is represented with a higher order curve (e.g.: “potato chip”) additional data will be acquired, for instance at least nine locations. If the data is acquired with the first side alignment system, the second side location may be approximated using the thickness of the wafer, which may be measured or specified by the operator.
Similarly, both for calibration and marking, a marking beam focus function may be sampled at a number of locations in the marking field (at reduced power). The system may include a detection system suitable for measuring “featureless” surfaces, for example a bicell or quad-cell arrangement. Alternatively, a projected grid may be used similar to the options provided in commercially available Metrology equipment manufactured by Optical Gaging Products (Rochester, N.Y.). The focusing tool will preferably be used for both alignment and system setup operations in addition to measuring-the working distance during wafer marking.
In the system of
However, other alternatives may be used with appropriate compensation for positioning with depth. For instance, in one embodiment a telecentric lens 15 may be used, but an arrangement similar to 47 of
With reference to
In one embodiment, the marking sub-system 19 includes the system shown in
The fine alignment system provides correction for residual X-Y-angle errors associated with the transfer and pre-aligner. In one embodiment wherein only small variations occur or are specified, the alignment system may correct X, Y, and theta (e.g.: angle) variations with measurements taken at three locations (e.g. fiducials). However, with emerging tight tolerances an increasing density of circuit/wafer, increased accuracy is preferred. The fine alignment system of 14 provides added capability of recognizing and/or measuring features associated with an article 2 of the wafer (e.g.: machine vision/pattern recognition capabilities). A feature location will be determined. An algorithm is used to obtain reference data and to locate a feature associated with at least one article 2 on a first side of the workpiece 3 using at least one signal from the first sensor 13. For example, article 2 of
In a preferred arrangement the matching is automatically performed over all the articles, and without human intervention. It should be noted that many combinations of patterns may be present on a wafer with special marking requirements (e.g.: “binning”) and the preferred algorithm will have substantial flexibility. The “training” may further include a semi-automatic, operator guided teaching phase so as to efficiently program the machine for recognition and matching of complex patterns.
In WO 0161275, incorporated by reference and assigned to the assignee of the present invention, various detection and recognition algorithms are disclosed for automatic learning of circuit features using grey-scale and/or height information, and subsequent use of the stored information for inspection. For instance, the following sections of the '275 disclosure: page 7, lines 4-26; page 8, lines 1-5 and lines 17-25; page 9, lines 5-10; page 10, lines 24-25; page 11, lines 1-18; page 15, lines 29-30; page 16, lines 1-10; page 17, lines 19-28 and the associated drawings teach the application of various pattern recognition and learning algorithms. Further details of various steps for detection and matching features for obtaining reference locations for precision marking of wafers and similar articles are disclosed in SECTION 2 which follows entitled “Feature Detection and Fine Alignment.”
In one embodiment of a 300 mm marking system, an 80 mm marking field is used for high speed, and an alignment vision field of approximately 16 mm is used to for feature detection. With a 1024×1024 array a 16 μm pixel size will be provided, which is somewhat finer than the spot size of the marking beam. For example, in an embodiment of a backside wafer marking system a spot size of less than 40 μm is preferred, with a most preferred range of about 25-35 μm. The marking field 1 dimension (depicted in FIG. 3 and corresponding to the region 4 of
As shown in
The steps of obtaining reference data, relatively positioning, and marking are repeated so as to locate a feature associated with at least one article on the first side, and to position a marking beam within all the designated regions on the second side based on the feature Iocation(s).
The beam positioning sub-system preferably includes a 2D galvanometer scanner 40, 41, 42, 43 as shown in
Relatively positioning may further include: (i) providing a beam positioner which may include a 2D galvanometer deflector; (ii) adjusting a mirror 42, 43 position (See
In a preferred system for wafer marking at least one workpiece positioner is used in addition to stage 18 (also depicted as 104 in
In view of the aforementioned emerging three-dimensional variations and tolerance requirements, it is preferred that the marking beam focus position shown as 422 in
In an alternative arrangement the wafer may be held in a vertical position. For instance, a suitably modified and automated version of the “Wafer Edge Fixture” produced by Chapman instruments, and configured for a maximum wafer size 300 mm (Chapman Instruments, Rochester, N.Y., and referenced to U.S. Pat. No. 5,986,753) may be used. Six degrees of freedom are included for profiling of wafers. Further description of the tilt stage, wafer chuck, X-Y-Z stage, and controller are found in the article “Wafer Edge Measurements—New Manual Fixture Provides More Features.”
In one embodiment for wafer marking a “split gantry” stage is an alternative with automatic positioning of the horizontal mounted marking head along one direction (e.g.: “X”, horizontal, into the page) and wafer positioning in at least a second direction (e.g.: “Y” vertical and along the page, and “Z” along the optical axis, and preferably including capability for roll and pitch adjustment).
Various combinations of the motion (manual or automatic) of the (1) workpiece positioner 18 and (2) beam positioning sub-system (e.g.: “marking head”) 19 and/or (3) internal components of 19 (e.g.: a dynamic focus sub-system 48 and/or beam expander 49 may be used and coordinated with controller 27. For instance, five axes of motion (e.g.: X, Y, Z and Roll, Pitch) may be implemented for precision positioning in a wafer processing station 100. Further, coarse (possibly manual or semi-automatic) positioning may be implemented in one or more axes, for instance.
The selection of laser pulse characteristics can have a significant effect on the speed, contrast, and overall quality of the marks. For backside marking of Silicon wafers a pulsewidth of about 15 ns, repetition rate of about 25 KHz, and output energy of about 0.23-0.25 millijoules at a wavelength of 532 nm provided favorable results. A short cavity green Vanadate laser was used. Further, marking depth penetration of about 3 μm-4.5 μm provided machine readable marks without internal damage (e.g.: cracking) of the wafer. Marking speeds of about 150 mm/sec were achieved, and it is expected that about 350 mm will be achievable with preferred laser parameters. The marking speed represents a relative improvement for marking in view of the large number of articles to be marked at high resolution. An exemplary range of operation includes pulse width of about 10-15 ns, repetition rate of about 15-30 KHz, with focused spot size of about 30-35 μm for marks on Silicon wafers. Another range may include a pulsewidth of up to about 50 ns, and a minimum repetition rate of about 10 KHz. Micro-cracking is also prevented by limiting penetration of the beam to a depth of less than about 10 μm. It is expected that a wavelength of 1.064 μm will be suitable for marking metal workpieces, with frequency doubled operation for Silicon wafer marking. Further details on a preferred laser and associated characteristics are disclosed in SECTION 4 which follows entitled “Laser Parameters and Mark Quality.”
The inspection system will preferably provide feedback regarding mark quality as rapidly as possible to maximize yield. For instance, a wafer may have 30,000 chip scale packages as articles. A marking field may have at least a thousand die. A separate inspection system with “standard” lighting for viewing marks may be an advantage to establish correlation between various stages of the wafer and device assembly steps wherein the marks may also be viewed. In an embodiment where the inspection system optical axis is separated the inspection may occur in a sequence where a first field is marked and then inspected. The inspection of the first field will occur while a second adjacent field is being marked when a large number of articles are to be inspected.
In an embodiment using a pair of galvanometer mirrors, data representing at least a sample of die (or other article) over the field may be acquired with a “through the lens” vision system (e.g.; a second simpler vision system for the case of wafer mark inspection). The data processing operation may overlap with positioning (indexing) to an adjacent field. It should be noted that the coaxial vision system might not require a vision system with complete inspection capability. For instance, the intensity or radiation pattern of the reflected scanned beam may be analyzed for early detection of gross mark defects or other process problems. For instance, a single photodetector may be used to analyze the reflected marking beam. Telecentric viewing (e.g.: received through lens 351) reduces variations with angle, which can provide for improved classification of signals.
Some Further Discussion of Various Alternatives:
In a preferred embodiment the workpiece 11 is translated when indexing to marking fields. However, the relative motion of the workpiece 11 and beam positioning sub-system 19 may include translation of at least a portion of the beam positioner (or a component). When marking wafers, a single X-Y stage moving the wafer allows for positioning of the alignment system 14, marking lens 351, inspection system 20, and possibly an optional mark verification reader. In an embodiment wherein the wafer is translated, alignment and beam scanning may be simplified. In an embodiment where the positioning sub-system or portion of the sub-system is translated fiber beam delivery from a remote laser source to marking head 19 may be used to an advantage.
In one embodiment for wafer marking a Z-axis stage 28 may be used. A range of at least + or −2 mm is preferred. The beam positioner 19 and lens 351 may move, but movement of the wafer is preferred. The Z motion may be determined by the focus of the alignment camera system components 13, 15. The sag and warp of the wafer is preferably compensated by movement (translation, roll, pitch) of the wafer with the positioning system 18, 17 or by movement of the beam positioner 19 as described above.
A total Z range of travel of about 12 mm, implemented with one or more translators, may be used to allow a robotic end effector to load a wafer while allowing for compensation of wafer sag by relative movement of the wafer and marking beam focus location.
A method for controlling contamination may be an advantage. For example, a tilted window, placed between lens 351 and the workpiece, with a slight amount of vibration may remove particles from the marking lens. Air pressure may be used to clean the lens during idle periods. A tilted window will displace the beam and aberrations may be introduced. Certain errors (e.g.: beam displacement) may be corrected during calibration. Alternatively, an “air knife” may be used to produce fast moving air across the lens.
An exemplary exclusion zone of about 2-3 mm is typically used.
The wafer nest may have vacuum applied on the 2 mm exclusion zone. The nest may be held with a kinematic mount.
The focal position of the alignment system lens 15 and camera 13 may be used for determining a Z-axis location and for fine positioning of the beam. In one embodiment the wafer is translated. Alternatively, the camera system may be focused and the position recorded. The position may then be related to the beam positioner coordinates (e.g.: the lens position) and the lens and positioner translated accordingly.
In one embodiment slight relative movement of the Z-axis position may be used to compensate for sag and warp. For instance, a change in the z-axis position may be effected at a plurality of marking locations over a 100 mm marking field. For instance, Z-translation may occur at nine locations (e.g.: to compensate from center to edge).
The X-Y table may have a range of travel of about 12-18 inches, with linear encoders for position feedback.
An inspection module may have optical resolution of about 4 microns.
A telecentric lens may be used with the fine alignment system.
The inspection module 20 may also be used for certain alignment operations (e.g. locating a fiducial on backside) and may be calibrated using a transparent alignment target to establish correspondence with the coordinate system of the fine alignment camera 13.
The recommended marking depth for optimum reading, while avoiding substrate damage, may be about 3.5 microns. The laser system may be configured for a maximum mark depth of about 10 microns.
Embodiments of the present invention may be used to mark wafers with programmable field sizes and number of fields (e.g.: 9-16 fields of view on a wafer having a diameter in a range of 150-300 mm), focusing options (e.g.: 3 focus positions for wafers 775 microns thick with increasing density for thinner wafers) and various marking speeds (e.g.: 150-250 mm/sec).
Various exemplary and non-limiting system parameters and associated tolerances may include:
Numerous alternatives may be used to practice the invention. Variations of the positioner type, number of positioners, vision systems, focusing hardware, laser types including q-switched and fiber lasers, may be used. Furthermore, the choice of serial/parallel operation of multiple markers and inspectors for efficient production time management and yield improvement, including cluster tools and statistical process control may be incorporated for use with a precision marking system of the invention. Further, it is contemplated that the pattern recognition and marking techniques of present invention may be used alone or in combination with other production processes, for instance the “dicing” operation described in the aforementioned '943 patent.
SECTION 1—2D/3D Calibration
Various commercially available marking and workpiece processing systems calibrate the laser marking field by marking a grid on test mirror and measuring the grid on a separate coordinate measuring or metrology machine. It is an iterative process and very time consuming. Other laser systems use the on-line through-scan-lens vision system to calibrate the laser-marking field on the same side. Alternatively, a substrate or disposable workpiece may be marked.
In accordance with the present invention, “two-dimensional calibration” utilizes an x-y stage a pair of stages translating the workpiece and/or marking head, or other arrangement winch allows the on-line machine vision sub-system 14 of
Calibration may be system dependent and manual, automatic, or semi-automatic. By way of example, four steps for calibration are shown below to illustrate aspects of overall system calibration:
Yet another calibration step may be applied to compensate for X-Y stage tolerances.
Yet another calibration may be applied to calibrate the scan lens 351 of marking head 19 in
These basic steps above may be sufficient and preferred in a system wherein marking performance is substantially invariant with depth (e.g.: large depth of focus, relatively large laser spots, relatively small wafers having exemplary thickness of about 775 microns and minimal sag).
In one embodiment the alignment vision subsystem 14 of
Compensation for workpiece sag and warpage may require maintaining the same spot size with different working distances. Besides, there is an increasing need to change laser beam spot sizes during operation to meet different application parameters, such as line width, character size, mark contrast, hard-mark, soft-mark, throughput, etc. Three-dimensional calibration provides calibration at a plurality of marking positions along the Z-axis. As a result, the laser marking field capability is provided for changing the laser beam working distance and/or spot size automatically while maintaining the laser beam position accuracy.
There is also an increasing need to change the size of the field of view (FOV) of the machine vision system during operation to meet different application requirements. Three-dimensional calibration on machine vision allows the system to change the size of FOV automatically and maintain the vision dimension accuracy at the same time.
The calibration techniques described herein are not restricted to “topside” imaging and “bottomside” marking. For example, the process may be applied to wafer marking in a system where a chuck holds the wafer in a vertical position, and the marking and illumination beams are substantially horizontal. Likewise, the workpiece may be marked from the topside based on calibration and reference data from the bottom-side. Similarly, the process may be adapted for calibrating separated alignment and marking fields, both covering regions of a single side of a workpiece.
In order to optimize the system for different application parameters, sometimes one or more machine settings might require adjustment during the operation. When the change in setting affects the system accuracy, a new calibration will be required. The three-dimensional calibration process is used to create multiple layers of calibration files with respect to different system settings. A three-dimensional calibrated system can switch between different settings automatically and achieve the required performance and accuracy by using the corresponding calibration files. Exemplary methods to achieve three-dimensional calibration for different settings on the system include:
1. Laser beam spot size versus laser working distance: Use z-stage 28, and/or a combination of relative motion of chuck 17, and/or motion of an optical sub-system within marking head 19 to relatively position the test mirror to different working distances with respect to the laser source. Varying the working distance de-focuses the laser beam and provides different spot size at the work surface. It has been determined that a defocused spot provides acceptable mark quality for certain workpieces, and hence is considered. The two-dimensional calibration described above is repeated for each working distance. As the result, a group of calibrated algorithms or look up tables for different spot sizes with corresponding working distances is generated.
2. Laser beam spot size versus laser beam expander setting: Use an expander for focus control, zoom expansion control, or the combination. For instance, a computer controlled embodiment of the expander 49 shown in
3. Laser beam working distance versus laser beam expander setting: Use an expander for focus control, zoom expansion control, or the combination. For instance, a computer controlled embodiment of the expander 49 shown in
4. Machine vision field of view versus vision lens/camera setting: Adjust the zoom and focus on vision lens/camera 13, 15 of sub-system 14 to achieve different sizes of field of view on a work surface. Repeat and generate a calibration algorithm or look up table for each vision lens/camera setting. As the result, a group of calibrated algorithms or look up tables for different fields of view with corresponding lens/camera settings is generated. On an alternative arrangement, “software zoom” capability provides for a useable range of operation without requiring moving parts. In yet another arrangement the digital and optical techniques may be combined.
In a preferred arrangement capability will be provided for adjustment of system parameters (e.g. laser beam working distance and spot size) while maintaining calibration in the presence of “sag” or workpiece warpage. The warpage may be significant relative to the depth of focus for smaller spot sizes, particularly for thinner wafers or workpieces (e.g. 300 μm thick, 300 mm diameter). In one embodiment the alignment vision system 14 (e.g. positioned relative to the first side) and marker coordinates may be calibrated with at least the following steps:
Use a precisely made grid template 91 (shown in
X-Y Table Calibration:
Use the fine alignment camera sub-system and x-y stage 18 to measure a precisely made full field size grid, which approximates or matches the workpiece dimension (e.g. largest workpiece to be processed with the system). This step will compensate for static errors (e.g. tolerance stackup), including non-linearity and non-orthogonality of the stages.
Marker Field Calibration:
Laser mark a full field size 24 grid on a mirror 92, as shown in FIG. 11C. Use the calibrated fine alignment camera (from step 1) and the calibrated x-y table (from step 2) to measure each mark 95 of the grid on the mirror over a marking field 24. This step will compensate for geometric distortion of the laser scanning lens and Galvanometer system and other static errors.
Three Dimensional Marker Field Calibration:
In order to compensate for wafer sagging and warpage, the wafer is marked a plurality of levels along the Z-axis 26. Multiple marker field calibrations may be required. In this case, relative motion of one or more of the (1) stage 18, (2) marking head 19 or internal optical components, for instance expander components 49 of
Three Dimensional Fine Alignment Camera Calibration:
In order to compensate for different wafer thickness, focusing of the fine alignment camera is set at some slightly different surface levels. The focusing operation may include translation of the fine alignment sub-system 14 along the Z-axis, or by adjustment of lens system 15, or in combination. Similarly, a Z-axis stage may be used to translate the workpiece. Multiple vision field calibrations may be required. In this case, fine alignment camera will focus at several pre-determined surface levels along the Z-axis. Step 1 is then repeated for each surface level.
The technique in Step 4 will also allow setting different spot sizes (by de-focusing) on the fly for different applications. Various curve fitting methods known in the art may be applied at each of the calibration steps to improve precision. The technique in Step 5 can also be applied to register the mark inspection camera 20 and fine alignment sub-system. For instance, the optical centerline 29 may be approximately aligned at setup and the calibration procedure used to precisely register the sub-systems. This is desirable so that the inspected marks may be displayed with a mark overlaying the corresponding die (for visual inspection), for instance. Software will be programmed to select correct calibration files for proper application.
SECTION 2—Feature Detection and Fine Alignment
In the GSI Lumonics WH4100 wafer marker, offered by the assignee of the present invention, a fine alignment vision sub-system corrects rotational or offset errors (X, Y, Angle) which are introduced when a wafer is placed in the marking station. A manual “teach tool” allows the user to train the system to recognize three non-collinear points on the wafer that is to be used for the correction. The operator selects three regions of the wafer (e.g. three corners of the overall pattern 115 of FIG. 12. During 4100 operation a positioner then positions the camera over the wafer and a die corner is visually selected. A “vision model” of the region is generated using an iterative trial and error process with various adjustments. For instance, lighting adjustments are used to enhance contrast so that an acceptable match (“model score”) is obtained at each of the measurement locations. Manual evaluation of the results is required with the system. The model information is then used to determine mark locations on the bottom side of the wafer.
The model 4100 is used to process wafers up to 200 mm in diameter using a “full-field” backside laser marker (e.g.: marker field covers the entire wafer). However, future generation marking systems will require marking of wafers up to 300 mm, for example, with miniature die or packages of finer dimensions (e.g. 0.5 mm). Also, smaller wafers may also be produced in the future with die sizes a fraction of a millimeter.
By way of example,
Those skilled in the art of machine vision measurement and pattern recognition will recognize that a number of tools may be used to obtain the information be used for the automatic teaching method. For example, the AcuWin vision software provided by Cognex is suitable for performing various internal “matching” operations. WO0161275, earlier cited herein, also teaches various automatic learning algorithms for use in a 3D system for inspection.
In one embodiment, during the training operation, a wafer is loaded into the system after the pre-alignment step. The algorithm then determines at least one of three regions for training based on wafer map information. The region information will often be replicated over the wafer, so a single pattern may apply to the entire wafer. Preferably, the system is calibrated with the 2D/3D calibration process prior to teaching, but a complete calibration may not always be required. Referring to
The purpose of the FineAlignment training procedure is to generate this information for a particular part type. The table below contains preferred information about a part type that is to be entered into the system before training can begin.
Three pieces of information are sought:
With this information the location of the upper-left corner 606 of the die pattern bounding box in the primary coordinate system may be determined. The origin of the die pattern coordinate system is then a die_pitch_y up and a die_pitch_x to the left of that as shown. With the position and orientation of the die pattern coordinate system known, the stage may be moved relative to any die location on the wafer.
Upon determining the locations of two actual die corners along the left and top edges of the die pattern, and with capability for positioning any die location in the field of view, a search is performed (e.g.: search up/down and left/right) from these two corners looking for the last die in each direction. The target dies for this algorithm are 602, 601,603 in
A vision model is to be generated in the area around each corner (including at least a portion of all four neighboring die locations). The model may include various features corresponding to the model of
Various alternatives may be used to practice a semi-automatic or automatic training algorithm. For instance, additional die may be selected throughout the wafer and least squares fitting done to improve estimates.
An overall fine alignment process may be semi-automatic, but with an algorithm for automatic measurement of the die pitch with enhanced accuracy. By way of example, the process may begin with a wafer transport tool moving a wafer to the nest. A user interface and display allows an operator to move a wafer stage 18 of
In at least one embodiment the die pitch is measured prior to setting up the at least second and third alignment locations or the at least three locations 601,602,603 used to transform coordinates. The operator may position the stage and view the wafer to identify a suitable row of die and further identify die corners, for instance the lower left and upper left corner of a die. The stage may then be moved (e.g.: interactively) to the next die and a corner location identified from which the die pitch in a first direction is estimated. The process is then repeated in the orthogonal direction.
Preferably the estimate is improved using a program to obtain additional data by traversing the wafer along rows and columns, identifying useable die, and locating features (e.g.: corners) of the die with a pattern recognition algorithm. The data may be obtained at each row or column where useful data is available, or in larger increments. The averaging spacing may be estimated and related to a wafer map.
In a present system of the invention “ease of use” and minimal operator intervention are considered beneficial improvements. Operator inputs may be valuable to verify a column of die are useable, for instance. In one embodiment the operator may verify that a selected die corner is useable and in a “topmost” column.
The additional locations for pattern matching are the selected, the stage positioned, and a test to verify the correct pattern recognition software operation.
SECTION 3-Workpiece Chuck/Positioner
It is desirable to grip and hold workpieces of varying shapes for the application of second side marking based on first-side data. Similarly, a preferred arrangement can also be adapted for general “double sided” laser processing and/or inspection operations.
Generally, at least one workpiece positioner is provided to relatively position the workpieces, and configured so as to support and position workpieces of varying specified dimensions. The arrangement allows radiation beams to directly irradiate the first and second sides of the workpiece over a large working area. Further, damage to the workpiece is avoided which might result from mounting on a fixture. Still further, a desirable arrangement allows for a robot driven end effector to load a workpiece without movement of chuck.
In at least one embodiment a method and system for edge chucking and focusing populated and blank silicon wafers of variable diameters and thickness is used. The method and system may also be used for other applications, for example in a micromachining process where a radiation beam is to irradiate both sides of the workpiece.
The “chuck” system includes one or more positioners for supporting workpieces of varying sizes, and for fine positioning of the workpiece with one or more degrees of freedom. The chuck system is mechanically coupled to the X-Y translation stage 18 of
The workpiece positioner (e.g. positioning sub-system) may be constructed as shown in
A chuck configuration utilizing, but not limited to, nor requiring, three positioners, driven by closed loop linear steppers or servos, is the preferred method for holding most workpieces.
In operation, under control of a computer program, the workpiece is loaded by adjusting the distance between support 53 with the first axis drive(s) to match the width of the workpiece. At least the height, and preferably the attitude is controlled with the additional axis. This generally provides, when used in combination with other system components, at least five axes of adjustment (e.g.: X,Y,Z, roll, pitch). Further, the adjustment may be dynamic and occur during the laser processing operation or during idle periods.
SECTION 4 Laser Parameters and Mark Quality
It is desirable to produce high contrast, machine-readable marks, at high speed in a designated region (e.g. specified by length, width, and depth). Further, conformance to industry specifications prohibits damaging or otherwise adversely affecting the function or operation of the articles (e.g. a semiconductor device).
Preferably the laser output will be generated from an Neodymium Vanadate laser with a wavelength of 1064 nm for processing metal based substrates. The output will be frequency doubled using the second harmonic generator 72 to be 532 nm for non-metal substrates (e.g. silicon or gallium arsenide).
When practicing the present invention various alternatives may be used for pulse energy control, for instance, controlling (pulsing) the pump diode power for “marking on demand” with a series of pulses. U.S. Pat. Nos. 5,854,805, 5,812,569 describe such methods as applied to workpiece processing. A method of pulse control in laser systems is also described in U.S. Pat. No. 6,339,604. Various combinations of pump, q-switch, and optical switch controls may also be of benefit for controlling the energy output, improving reliability, etc.
In a preferred embodiment for marking, a telecentric lens 76 and optical sub-system 74 are used to control the spot size and distribution, which preferably will include optics for varying the spot-size and focus position under computer control.
In an application to laser marking output pulses are produced having a set of pre-determined pulse characteristics including a repetition rate (and corresponding temporal pulse spacing), pulse width, and output energy.
Selected pulses gated by the switch 73 or otherwise controlled (which may be a “burst” or “string” of pulses) irradiate the wafer 77 surface at a first predetermined marking location within the marking field of the mirrors 75. The stage 79 may be a step and repeat stage used when the workpiece is larger than the marking field (e.g. as also illustrated for the “second side” case of FIG. 2A). Referring to
Preferably, the pulse width will be within a range of about 10 to 15 nsec to produce a mark with sufficient contrast.
The energy per pulse incident on the surface is preferably in the range of 0.00023 to 0.00025 Joules (eg: 230-250 microjoules) produces high quality marks on Silicon wafers.
Preferably, the marking speed is improved to a higher linear speed on the wafer surface 77, with a relatively high pre-determined pulse frequency of the laser 71. By way of example, a repetition rate of about 15-30 KHz, for instance 25 KHz, provides significant improvement over earlier wafer marking systems used at both near Infrared and Green wavelengths. With a preferred spot size of about 30-35 μm, linear marking speed greater than 150 mm/sec is a relative improvement over previous wafer mark systems. A speed of about of 350 mm/sec is expected for use in a system having the preferred laser pulse characteristics. Reduced solid state laser power at high repetition rates constrained earlier performance, and separation of spots on the surface were observed which limited mark quality.
A laser pulse is focused into a spot diameter to produce energy density within a predetermined range. The minimum distance between a pair of machine readable marks may be further reduced by controlling the spot diameter with optics 74. Such an arrangement may include a “zoom” beam expander in 74 or other optical components which are removable/insertable, preferably under computer control (e.g.: as shown in FIGS. 4 and 6). The spot size adjustment is generally desirable to control the mark linewidth and contrast. A spot diameter in a range of about 30 to 35 μm and a working distance to the workpiece of about 220 mm to 250 mm represent exemplary ranges of operation. The smaller spot size provides improved capability for producing higher mark density compared to earlier marking systems, and higher speed is provided with the pulse characteristics.
Results for backside marking of Silicon wafers have shown the depth range of a mark is to be about 3 to about 4.5 μm so to produce a machine readable mark 781 with enough contrast to the background. The result was contrary to an expectation that larger penetration depth was required. The results also provided additional margin for avoiding damage.
Further analysis of the marks indicated sufficient contrast for machine readability over a range of about 3-4.5 μm. Increasing the mark depth to the larger numbers, for instance 9.91 μm, produced cracking.
An interferometric scan was obtained of a wafer marked at a repetition rate of 10 KHz and a marking speed of about 120 mm/sec. Severe cracking as exemplified by the “spiky” data which results from structural variations at a depth of about 9 μm or more. Scans also showed good results with maximum depth of about 4 μm.
The shifted wavelength may be below the absorption edge of the workpiece material, but need not be restricted to 532 nm. For instance, the workpiece may be Silicon wafers or metal. The wavelength will preferably be substantially less than the absorption edge of Silicon (1.12 μm) for marking in accordance with the present invention.
Suitable lasers may include commercially available diode pumped (DPL) Nd:YAG lasers with about 6 Watts IR output, and output 3 Watts in the Green. An alternative though more expensive, is a 10 Watt (W) DPL laser with about 10 W IR and 5 W green output power. Preferably, the optical system will contain high efficiency optical components to minimize losses.
The Vanadate laser is preferred for marking Silicon wafers, but is not essential for practicing the invention. The desired pulse characteristics may be implemented with other designs, provided all specifications (e.g. beam quality, stability) are met. For instance, a fiber optic amplified system (e.g. Master Oscillator Power Amplifier) may be used to produce short pulses at relatively high rates. A solid state laser, including a fiber laser, with a slower repetition rate but sufficient power may be “pulse stretched” with a delay line and beam combiner(s) to increase the output repetition rate of the laser system.
SECTION 5—Precision Teletronics Lens
In precision laser marking and other similar material processing applications, for instance embedded resistor trimming, there is a need to produce fine spot sizes so as to control the width and contrast of a mark (or kerf) while maintaining precise spot placement over a relatively large 3-dimensional field. For instance, a 300 mm wafer may have die sizes ranging from about one millimeter or less with a tightly constrained marking region defined within the die. A spot size of about 30 μm will produce high contrast marks, but the depth of focus is about four-times less than that of earlier marking systems. For thin wafers the warpage may be a significant fraction of the depth of focus, so the three-dimensional spot size/spot placement considerations are valuable.
In non-telecentric scanned laser systems, spot placement errors at a workpiece plane will vary with depth, and may significantly degrade the system accuracy. Such z-axis error may be the result of workpiece tilt, defocus, sag, warp or any deviation from an ideal target plane. For non-normal incident angles of the scanned beam, the z coupling is approximately the deviation angle from normal incidence times the local z error. Preferably, a telecentric scan lens is used for focusing the marking laser onto the field. The telecentric scan lens, well known in the field of laser scanning, is used to maintain a near normal incidence angle of the beam to the workpiece thereby minimizing z coupling and the resulting x and y position errors. The approximate invariance of angle over the field may also have other advantages, such as providing for coaxial detection of reflected radiation. Coaxial detection can be used with many know methods to determine focus position, for example astigmatic spot detection.
Considering the first order scan lens properties, placing the scan origin at the front focal plane of the lens will produce a telecentric scan. In practice, there are non-linearity errors in the lens design that deviate from perfect telecentricity. Those skilled in the art of scan lens design recognize that correction of these errors is possible by modifying the individual lens elements and/or adding additional elements to the lens design.
Typically, an x y galvanometer scan system has two scan mirrors, as shown in
With field dependent error, a portion of the field may be selected to reduce errors at the workpiece. For instance, a small central portion of the field is used and material is processed with improved telecentricity. With one scan mirror located near the front focal plane of the lens, a first axis of the field addressed with this mirror will have better telecentricity than a second axis addressed by a second mirror more remote from the front focal plane. In this case, a portion of the field having improved telecentricity may be selected with a larger dimension along the first axis and a smaller dimension in the second axis, for example a rectangular field. It is also recognized that by using a rectangular field, the first axis may be larger than the edge a square field. Selecting a portion of the field may reduce other field dependent errors such as thermal drift of X-Y glavanometer deflectors. For example, a quadrant of the field where gain drift is mitigated in part by offset drift in each galvanometer may be selected to reduce beam-positioning errors.
For embodiments using through the lens viewing, the scan lens is typically required to image a target at wavelengths other than the processing wavelength. Color correction elements can be used in a design to improve viewing performance. Telecentric scan lenses with color correction for through the lens viewing are know, for instance the scan lens used in the commercially available GSI Lumonics Model W672 laser trimmer.
A preferred embodiment for precision laser marking of large wafers and similar applications includes a three-element telecentric scan lens 990 as shown in FIG. 13A. This lens has an effective focal length of 155 mm at 532 nm and is capable of forming 30 micron spots over a scan field of 80 mm square. The total path length is about 360 mm. With uncorrected, spaced mirrors the telecentricity error is approximately 2 degrees.
In the precision marking system, wherein three dimensional tolerances determine system performance, the spot placement accuracy of the lens system is to be maintained by including a method for three-dimensional calibration. In one embodiment the wafer is positioned with a workpiece positioner so that a best fit plane (over the wafer) is aligned normal to the marking head. A location is then determined relative to best focus position of the telecentric system of FIG. 13A. The beam positioner is directed based upon the location of features and stored calibration data.
At least one embodiment of the present invention may include a precision scan lens with improved telecentricity when compared with a conventional non-telecentric scan lens. In one example, the maximum angle incident at the workpiece may be less than about half of the maximum angle of the beam incident on the scan lens entrance pupil. In another example, the maximum deviation angle to the workpiece may be limited to less than other 10 degrees. This type of scan lens can be smaller, and may be less complex than a larger telecentric scan lens. Thus, a precision scan lens with improved telecentricity may be used to provide a design compromise with both a level of improved marking accuracy with changes in the workpiece height and reduced lens size, complexity and cost.
Backside Mark Visual Inspection with Frontside Die Registration
In early versions of certain backside wafer marking systems an infrared source was used to “backlight” a wafer so as to view backside features. With high density circuitry increasing at a rapid rate, the “backlight” approach will not always be possible in the future.
In one embodiment of a wafer marking system used to form marks on the backside of a wafer, an inspection feature includes a registered display of the mark and die. In a preferred embodiment inspection feature uses two cameras, one above and one below the wafer.
In at least one embodiment of the system, the equipment calibrates the bottom camera system 20 to the top camera system. Preferably, the cameras are in fixed positions. One or more cameras may have a zoom lens which is manually adjustable. In one arrangement, a calibration target of a transparent surface is placed between the two cameras. The image is acquired with both cameras. The images are superimposed and, using pattern-matching software, for instance commercially available tools from Cognex Inc, a correction offset, angle, and scale is calculated to align the bottom camera's image to the top camera.
During the inspection operation the top camera is used to acquire an image of the die on the topside of the wafer. The bottom camera is used to acquire an image of the mark on the backside of the wafer. By superimposing the coordinate systems of the two images, analysis determines the accuracy of the mark with respect to the die.
During inspection, this calibration data is applied to the mark image. Using pattern matching or OCR software the location of the mark relative to the location of the die is known.
It is to be understood that this feature is not restricted to top and backside wafer marking, but may be applied to any two sides or separated fields.
Inspection of marks may be done on-line or off-line. The inspection may include a random sample of die or up to 100% inspection. In at least one embodiment an operator may setup a region of interest 900 within a backside image corresponding to at least a portion of a die as shown in FIG. 17B. Preferably, the operator will be able to adjust 901 the area of interest, as shown in
While the best mode for carrying out the invention has been described in detail, those familiar with the art to which this invention relates will recognize various alternative designs and embodiments for practicing the invention as defined by the following claims.