EP1644895A2 - Procede d'apprentissage applique a une base de connaissances pour la classification automatique de defauts - Google Patents

Procede d'apprentissage applique a une base de connaissances pour la classification automatique de defauts

Info

Publication number
EP1644895A2
EP1644895A2 EP04735896A EP04735896A EP1644895A2 EP 1644895 A2 EP1644895 A2 EP 1644895A2 EP 04735896 A EP04735896 A EP 04735896A EP 04735896 A EP04735896 A EP 04735896A EP 1644895 A2 EP1644895 A2 EP 1644895A2
Authority
EP
European Patent Office
Prior art keywords
defects
adc
data
user
button
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
EP04735896A
Other languages
German (de)
English (en)
Inventor
Dirk SÖNKSEN
Ralf Friedrich
Andreas Draeger
Detlef Schupp
Thin Van Luu
Wolfgang Langer
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
KLA Tencor MIE GmbH
Original Assignee
Leica Microsystems CMS GmbH
Vistec Semiconductor Systems GmbH
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from DE102004022717A external-priority patent/DE102004022717B4/de
Application filed by Leica Microsystems CMS GmbH, Vistec Semiconductor Systems GmbH filed Critical Leica Microsystems CMS GmbH
Publication of EP1644895A2 publication Critical patent/EP1644895A2/fr
Withdrawn legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
    • G01N21/9501—Semiconductor wafers
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84—Systems specially adapted for particular applications
    • G01N21/88—Investigating the presence of flaws or contamination
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G01N2021/8854—Grading and classifying of flaws
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30148—Semiconductor; IC; Wafer

Definitions

  • the invention relates to a method for teaching in a knowledge-based database for automatic error classification.
  • wafers or masks are processed sequentially in a multiplicity of process steps during the production process.
  • the requirements for the quality of the structures formed on the wafers increase
  • the quality, accuracy and reproducibility of the components and process steps that handle the wafer are correspondingly required.
  • the invention has for its object to provide a method with which a simple and fast possibility is offered and all data and files required for an ADC run ADC Run (Knowledge Base, AutoAlignment Focus Setup) This object is achieved by a method with the features of claim 1
  • Leica ADC HP provides a simple and quick way to generate all the data and files required for an ADC run (knowledge base, auto alignment, focus setup). In some cases, specified data and files are used no longer, as in previous ADC versions, a manual classification of the defects on a wafer is necessary, the time required to teach in a knowledge base for creating a new ADC recipe can be reduced by up to 50%.
  • the pre-grouping function also improves Pregrouping Function 'in many cases the quality of the knowledge base, which in turn has a direct influence on the accuracy of the ADC run.
  • ADC HP is presented as an independent learning mode LearnMode in the Leica ADC.
  • the user has to specify, confirm and change the necessary data in individual steps individual steps are presented as separate pages in the Leica ADC HP Dialo g displayed
  • the user guidance to the individual page is done in the so-called wizard style ie ubei ⁇ Back> and ⁇ Next> buttons.
  • the new learning mode has the advantage over the previous learning mode that the new learning mode is uncomplicated and requires a reduced number of steps are to be carried out by the operator in the correct order. Pre-classified defects were required for the previous learning mode. All that the new learning mode required is one or more wafers with as many unclassified defects as possible.
  • an interaction with the Viscon surface is necessary, the Leica ADC HP dialog not displayed modal but top most The dialog can be automatically made invisible or the user can make it invisible or visible again
  • the process of learning a knowledge-based database for automatic mistake classification includes the following steps: Selecting a review data file Input of parameters and data by the user on a page (50) of the learning mode, the parameters and the data being known to the user, starting an alignment procedure and a procedure for adjusting the light intensity, automatically setting the optimal intensity of the lighting by starting some defects and, if necessary, adjusting for the optimal illumination, checking the detection using a few examples, the optimization of the detection parameters being carried out using images, - automatic approach to all defects of the wafer or the wafers, the respective defect being detected and the respective defect being described by a descriptor is assigned, and analysis and automatic grouping of the descriptors of the defects
  • the input of parameters and data comprises the selection of the elements present on the semiconductor substrate, wherein the elements can be memory circuits, logic circuits, a bare wafer without resist, or with resist.
  • the parameters or data of the layers on the wafer include the specification of a polymer layer, an oxide layer, a contact or a metal layer
  • the user selects the type of illumination, at least one lens used and one type of focus.
  • the type of illumination bright field, UV or DUV can be selected.
  • the default setting is bright field and the lens is 100 times Veigroßei uny
  • a manual two point alignment is performed, with a first point being manually aligned by moving a table during the Teaching in the first point automatically saves data for the auto-alignment file.
  • Each alignment point is taught in with three different magnifications of the lenses.
  • the optimal intensity of the lighting is set by randomly selecting a certain number of defects.
  • To set the optimal intensity of the lighting only defects that are not larger than 25% of the video picture width and -
  • the analysis and automatic grouping of the descriptors of the defects divides the recorded thumbnails of the defects into groups.
  • the display shows the first nine examples of a selected group of defects in a thumbnail representation
  • FIG. 1 shows an overview of a schematic structure of a Wafei inspection device in which the method according to the invention is implemented
  • FIG. 4 shows a “Leica ADC HP Control Desk” window, which clearly summarizes the ADC tasks, some of which are already available in previous ADC versions
  • FIG. 5 shows a page of the learning mode, which the user calls up and thereby opens an input file, ie a review -Data files indicates
  • FIG. 7 shows a page of the learning mode, via which the user specifies the ADC KnowledgeBase data
  • FIG. 9 shows a page of the learning mode, via which the user carries out an automatic light adjustment
  • FIG. 10 shows a page of the learning mode, via which the user achieves an optimization of the setting of the detection parameters
  • Figure 13 is a representation of a Change Sensitivity dialogue
  • Figure 1 4 is an illustration of a window that issues a warning to the user
  • 16 shows a page of the learning mode, via which the user automatically generates a knowledge base
  • 1 shows a representation of the Defect Code Mapping dialog
  • FIG. 19 is an illustration of a dialog for starting an ADC run
  • FIG. 20 is an illustration of a finish dialog
  • a scanning table 4 is integrated on a base frame 2 as a support table for the wafer 8.
  • the scanning table 4 can be moved in an X coordinate direction and a Y coordinate direction
  • the wafer 8 to be examined is placed or hooked on the scanning table 4.
  • An observation device which is preferably equipped with a microscope objective 7, is connected to the base frame 2 via a carrier unit 9.
  • the microscope objective 7 enables the wafer 8 to be observed on an enlarged scale.
  • Several microscope objectives 7 can be attached to one Revolvers (not shown) may be provided, so that observation at different magnifications is possible.
  • the structures of the wafer 8 which have been observed in an enlarged manner can be viewed directly via an eyepiece 5 or via a display 11 which is connected to a CCD camera 13
  • an electronics unit 15 is provided, with the aid of which a system can be automated.
  • the electronics unit 15 is used to control the scanning table 14, to read out the camera 13 and to control the display 11.
  • the wafer holder 16 is usually designed in such a way that it is connected to the Examining wafer 8 so that f is fi during the examination period.
  • the scanning table 14 is designed to be movable in an X coordinate position and a Y coordinate direction that are perpendicular to each other observation point on the wafer 8 are brought under the optical axis 7a of the microscope objective 7 (FIG. 1)
  • FIG. 2 shows the ADC HP Toolbar button 20, with which the user calls up the function for automatic error detection.
  • the ADC HP dialog is opened via an ADC HP Toolbar button 20 or via a main toolbar 19 of the Viscon application 21 in the “DC” menu or in the context menu of the 'ADC dialog (see Fig. 3) Every user (from user level operator ") has access to this menu entry. Since ADC HP is a separate option, the menu entry will only be visible, if ADC HP is also installed installed This option is protected, similarly as before, via a registry entry which is generated by the installation program when the option is selected
  • Fig. 4 shows the so-called 'Leica ADC HP Control Desk' window 25. It clearly summarizes the ADC tasks that are already available in previous ADC versions in a window and serves as the starting point for starting individual modules 26, 27, 28 and 29 In detail, this is the ⁇ Learn Recipe, learning and creating a new ADC recipe, and a knowledge base with subsequent ADC run (Run Recipe), the edit recipe to edit an existing knowledge base, the expand recipe to expand an existing knowledge base and the run recipe to start an ADC run
  • a button is provided for each module.
  • this is a Learn Recipe button 26, a "Edit Recipe” button 27, an "Expand Recipe” button 28 and "Run Recipe” button 29
  • "Edit Recipe” After pressing this button 27, the user must select an existing knowledge base file. This is started by the external application KB Wizard and the content of the file is displayed. The data can edited there and the knowledge base tested as a whole ⁇ "Expand Recipe” With this button 28, the user selects an existing knowledge base file and a review data file.
  • the ADC HP - Learn mode is displayed as a non-modal dialog. The user has to enter the necessary data or select files in eight consecutive steps, ie on eight pages. The final page only shows the result of the ADC HP - Learn run
  • buttons 30, 31 buttons 30, 31 (wizard style), as far as the current state allows (see Fig. 5)
  • the display of the individual pages does not depend on user level - exception are additional user interface elements that are only visible for development - user level.These are only visible during the development phase and are removed in the release version or generally for everyone User level to be invisible
  • FIG. 5 shows a page 33 of the learning mode, which the user calls up and thereby opens an input file 34, ie indicates a review data files.
  • the page 33 is referred to as an open input file.
  • the page 33 shows the data file (without path).
  • a FileOpen button 35 the directories for the user to enter the file are displayed. If an input file has been determined, it is opened temporarily, but the Viscon sequencer is not started.
  • the file is hard-coded as a script file
  • FIG. 6 shows a further page 38 of the learning mode, which the user calls up and thus assigns a name for a recipe file Recipe File.
  • Page 33 is labeled Recipe File.
  • the actuation of the back button 30 is not permitted in this page 38.
  • the actuation of the Next - Button 31 is allowed if a valid input file 37 is selected.
  • the Cancel button 39 the user can cancel the ADC HP Learn mode.
  • the Leica ADC HP recipe file is displayed in an EditEo 4U.
  • the komponentenu read name components are put together as specified and the resulting file name (ending with vsl) is displayed.
  • the name components are separated by a _ character (unteistnch) Invalid letters in the resulting file name are removed and hyphens are replaced by an underscore.
  • the user also has the option of changing the specified name (in whole or in part) according to his ideas.
  • Result Recipe File sequence control file for an ADC run
  • the file "EasyADCRun vsl" is used (hard-coded) in copy.
  • Page 38 contains several checkboxes 41, 42, 43 and 44. Checkboxes 41, 42, 43 and 44 are used to determine the name components. Default Lotld, Stepld and Setupld are used The resulting file name (without ending vsl ') is also used as a default for other files (AutoAlignment, focus setup file, etc.).
  • the data file of the results "Result data file” is always given the same name as the input file and same format type and written in the standard result directory
  • the back button 30 is allowed and the next button 31 is allowed if at least s a name component has been selected
  • the Cancel button 39 is permitted
  • FIG. 7 is a page 50 of the learning mode, via which the user specifies the ADC KnowledgeBase data.
  • This page 50 is labeled ADC Basic Data.
  • ADC Basic Data In a selection column 51 with the designation “Structure Type, the user can choose between memory .memory and logic circuit logic Select An additional selection for a bare, unstructured wafer, 'bare wafer' is also possible.
  • ADC run mode repetitive or random mode
  • auto alignment mode normal auto alignment or bare wafer alignment
  • Layer Type the user can choose whether one or more layers on the wafer
  • Layers are also applied. It is also of interest which layers are applied to the wafer. Without resist, w / o resist, with resist is called with resist "(see FIG. 7). The resist or other layers are on the wafer 8 or Semiconductor substrate applied The preselection or the default setting is w / o Resist In others The user can select the type of layer.
  • a polymer layer is labeled "Poly”
  • an oxide layer is labeled with oxide
  • a contact is labeled "Contact”
  • Metal layer is labeled "Metal”.
  • the process of hanging up the various layers can also be selected
  • B an oxide layer (oxide) applied in front of the polymer layer is called 'before poly'.
  • the choice of the layer type metal gives the user the possibility between a single metal layer (Metal 1), a double metal layer (Metal 2) and an n-fold Select metal layer (n-metal) Whether there is a main layer and a subordinate layer is used to determine the random mode and the focus type.
  • the default settings for the layers are poly, for, Ox ⁇ d Before Poly and for, Metal Metal 1 Oxid- und Metal -Underlayer radio boxes are only activated if "Oxid" or "Metal have been selected beforehand. Otherwise they become inactive In a further selection column 53, the user can select the illumination mode .Illumination Mode
  • the radio boxes with the designation BF for bright field, UV for ultraviolet and DUV for deep UV are available to the user.
  • the available lenses are displayed to the user in a list box 54 , only the lenses that match the selected ADC type are displayed.
  • Bright field BF ' is selected for the default setting and a lens with 100x or lower magnification is suggested
  • Table I shows the resulting focus setting based on the selected data
  • the default values of the "TV Focus Flexible 2" mode are used.
  • the back button 30, the next: button 31 and the cancel button 39 are permitted in this window. If the ⁇ Next> button 31 is pressed, the ADC HP dialog becomes invisible.
  • the same changes are made for the named copy of the "EasyADCRun" file (the later ADC Run Recipe).
  • the input file is loaded with the adapted script file and the VisconNT sequencer is started.
  • the file is automatically loaded until the wafer Selection processed
  • the standard wafer selection dialog is used and displayed.
  • the learning mode is dependent on the setting of the layers on the wafer of the corresponding AutoAlignment (SemiAuto or later Bare Wafer Alignment) started
  • the user can carry out a manual two-point - Al ⁇ gnme.nt, whereby he only aligns the very first point manually (moving the table with a joystick or by double-clicking the mouse in the live video image) and must confirm.
  • data for the auto-alignment file are automatically saved.
  • Each alignment point is taught-in with three different objective magnifications, whereby the maximum magnifying lens is specified by the selection on page 50 (ADC Basic Data).
  • the second point is already automatically learned and aligned based on the stored data of the first point.
  • the selected ADC lens is always specified by the software. This lens must be used because it is required for later light balancing (the alignment point method used) he If the point entered is not found, the second point is shifted by a to the center of the wafer and the structure is searched there again. The second point is shifted by six. This before the alignment ends with an error. In this case, the user is informed Displayed window that says that the alignment is canceled and the wafer is unloaded
  • the Viscon sequencer is paused (built-in pause action (without MessageBox display) in the Easy ADC Script File), the ADC HP dialog is visible again and shows the next page.
  • the Next button 31 is not allowed , if alignment is carried out or was aborted by an error.
  • the Next button 31 is allowed if the alignment was successful.
  • the Cancel button 39 is allowed and aborts the entire ADC HP learning mode
  • FIG. 9 is a page 70 of the learning mode via which the user carries out an automatic light adjustment.
  • This page is labeled Light Adjustment.
  • a Perform Automatic Lightadjust ent Buttons 71 select a certain number of points (defects from the data file) at random. If there is size information, only defects are selected that are not larger than 25% of the video picture width and height. These defects are approached and pictures are taken on Lamp brightness "start value is determined on the basis of a histogram evaluation and set on the microscope. This means that the brightness is reduced in such a way that no defect image is" overdriven ". For this, all available color channels are examined and set accordingly
  • a status box 72 "ProgressControlBox 'and an info box 73" Read-Only EditBox' are shown.
  • the progress during the automatic light adjustment "LightAdjustments” is shown in status box 72.
  • a status text is displayed in the info box 73 if successful or unsuccessful.
  • the back button 30 is not permitted if the light adjustment is carried out.
  • the back button 30 is permitted if the light adjustment is rejected.
  • the wafer is unloaded and the page 50 ADC Basic Data is displayed.
  • the next button 31 is permitted if the light adjustment was successful
  • the Cancel button is allowed when light adjustment is carried out and all open files are closed and deleted
  • FIG. 10 is a page 80 of the learning mode, via which the user achieves an optimization of the setting of the detection parameters.
  • This page 80 is labeled Optimize ADC Detection '.
  • the process is started with a button 81, which is labeled Start Optimization " Ensure that the standard values for focus setting and detection ⁇ araineter work on the selected wafer. If this is not the case, the user has the option to change the default values once more.
  • the Viscon sequencer is started, defects selected, approached and pictures taken
  • the text on the button 81 changes to 'Stop Optimization'
  • the progress of the picture acquisition is displayed in a storage box 82. The user can cancel the process by printing again.
  • the Viscon sequencer is started again, the button text changes to Stop Optimization and a predetermined number of defects in the current wafer is selected
  • the defects are approached and a special ADC action is triggered.
  • This action records the images, detects the defects via an already developed ADC routine, and saves the images temporarily until the image acquisition of all defects is completed.
  • the progress of this process is displayed using the status box The user can abort the process by pressing the button again
  • thumbnail dialog 90 full screen display on the screen
  • the thumbnail dialog 90 is essentially divided into a first area 91, a second area 02, a third area 93 and a fourth area 94.
  • the first area 91 comprises a horizontal list in which the thumbnails 91, 91 2 , 91 3 , , 91 n with detection marking and defect ID (defect identification).
  • the currently selected image is displayed maximally with a resolution of 640x480 pixels in the second area 92. If available, the reference images are also shown reduced in size in the third area 93.
  • the current image selection can be changed by means of a mouse click, cursor keys and / or the browser buttons 95 under the defect image
  • the defect marking can be switched off and on again via a ⁇ H ⁇ de Defect Detect ⁇ on> button 96.
  • the browser buttons 95 are used to select and display the next or previous defect image.
  • the ⁇ H ⁇ de Defect Detect ⁇ on> button 96 is designed as a toggle button, and thus the detection marker can be switched visible or invisible
  • a Focus Difference - Defect Reference Button 97 enables the display of a messagebox 86 (see Fig. 12). Upon confirmation, the selected defect image (and existing reference images) is discarded, ie deleted from the display. An internal threshold value (default 30%) of the discarded insufficient images is discarded If bad focus images are exceeded, the focus values are changed (ie changing from laser to TV focus or changing the TV focus offset in 500 nm steps). The defects are then approached again and data is recorded. To do this, the Thumbnail Dialog 90 is closed and the ADC HP Dialog is displayed again during the scanning process
  • a Wrong Defect Detection button 86 enables the detection threshold for the selected image to be redetermined. For this purpose, a new dialog 80 is displayed (see FIG. 10).
  • a default button 88 enables all changes to the detection parameters of all images to be undone.
  • the list is rebuilt with the original values
  • the Optimize dialog is closed, the ADC HP dialog is restored Visible and the new overall detection threshold is entered in the knowledge base.
  • the Cancel button 39 closes the Optimize dialog and the ADC HP dialog becomes visible again. All changes are discarded
  • FIG. 13 is a representation of a dialog 100 for changing the sensitivity.
  • Change sensitivity The dialog 100 is used to determine the optimum setting for the detection sensitivity detection threshold of the selected defect image.
  • the defect image 101 is displayed in the center with the associated detection threshold in the dialog 100. on page 50 ADC Basic Data) indirect detection threshold has been used, a value of 50% is assumed. With two buttons 102, the sensitivity of the detection can be reduced or increased.
  • the defect image 101 shown in the middle shows the defect detection with the currently selected sensitivity The value is shown below the defect image 101.
  • a reduced image 103 is shown on the left next to the defect image 101 and illustrates the change in detection with reduced sensitivity.
  • a reduced image 103 is shown on the right next to the defect image 101 and ve illustrates the change in detection with increased sensitivity.
  • the current sensitivity is changed to this value and the image is now shown in the middle.
  • the changes to the left and right are then new calculated
  • a hide defect detection button 105 is designed as a toggle button. The detection marking is hereby switched visible or invisible
  • An S of the 106 with the name Sensitivity Step Size is used to change the strength of the changes in sensitivity when the button 102 is pressed.
  • a delete image button 107 is used to reject a defect for further evaluation. The defect is removed from the list in the optimization dialog This dialog is closed and the user goes to the previous dialog
  • FIG. 14 is a representation of a window 110 which issues a warning to the user. If an internal threshold value (default 30%) of the discarded wrong detection images is exceeded, new defects can be selected (automatically) and data can be recorded.
  • An Apply button 107 starts The application FIG. 15 shows a notification window 110 for the acceptance of the new detection threshold.
  • the notification window notifies the user that by accepting the new detection threshold, the detections of all other images also change when you press the ⁇ Refresh> button 87 in the Optimize dialog 90 the new value is applied to all other images
  • the detection threshold of the middle image display is adopted and the user returns to the previous dialog 100
  • FIG. 16 is a page 120 of the learning mode, via which the user automatically generates knowledge -Base carries out With a Start Collecting Data - Button 121 all necessary data of all defects of all selected Wafei are recorded and saved
  • the status is displayed to the user in a status box 122 and an info box 123.
  • info box 123 those still to be processed are displayed Defects from the total number (eg, 267 of 750)
  • the actuation of the back button 30 is not permitted when the data acquisition procedure is in progress.
  • the actuation of the next button 31 is also not permitted when the data acquisition procedure is in progress.
  • the actuation of the cancel - Button 39 is not allowed if the data acquisition procedure is running. If the operation of the Cancel button 39 is allowed, then all open files are closed and deleted
  • the process is as follows: The Viscon sequencer is started again and all defects in the input file are selected. In a first step, defects are approached on the wafer or wafers, pictures are taken, descriptors are generated and in the ADC result data the defect is saved The images of the defect are saved with the following settings W ⁇ te to Archive File '.' All Images' - 'Image Compression' Yes Leica-ImageStore (Leica image memory ) No.
  • the Viscon sequencer pauses at the basket level (before saving the output file).
  • the groups are generated from the collection of the descriptors (pre-grouping).
  • pre-grouping attempts to create a maximum of only 20 groups. Groups with fewer than 2 examples are discarded. The resulting groups are temporarily copied to the knowledge base, whereby the defect code and defect description of each group are numbered first (1, 2, 3 or EasyClassl , EasyClass2, EasyClass3)
  • a dialog 130 for classifying the defect code mapping is displayed (see FIG. 17).
  • the defect code mapping dialog 130 is essentially represented by a first window 131, a second window 132, a third window 133 and a fourth window 134.
  • a folder icon is shown for each group generated in the fourth step
  • the first nine examples of the selected group are shown in a thumbnail display.
  • the current defect code table is displayed in the third window 133.
  • the icon of this class is changed by getting a green check 136 and the corresponding defect code text is displayed.
  • This division mapping can also be carried out by double-clicking in the defect code table.
  • the toggle button 139 enables a section around the defect marking to be shown in the original size of the example images. If the defect marking in a sample image is too large, the display does not change. Pressing the toggle button 139 again leads to reduced full screen display back The actuation of an Apply button 129 is permitted if all
  • Defect groups were treated, i.e. mapped or marked as to be deleted
  • a sixth step an attempt is made to confirm the number of individual examples per mapped group (groups marked as to be deleted are not used and discarded). This is necessary so that certain groups with a large number of defects do not dominate the knowledge base and defects preferably assigned to this class.
  • the result is transferred to the knowledge base and the user goes to the ADC learning mode.
  • the Cancel button 39 is pressed on the display 11, the information dialog 140 shown in FIG. 18 after confirmation of the ⁇ Yes> - Buttons 141 the mapping and the entire ADC learning mode is canceled 19 shows a representation of a dialog 150 for starting an ADC run.
  • ADC Runs "With a start ADC Run button 151, after printing the button 151, all defects of the selected wafers are classified offline" (ie without restarting).
  • the classification done with the current ADC knowledge base Dialog 150 includes a status box 152 and an info box 153 With status box 153, the total number of defects to be classified is displayed (e.g. 123 out of 750)
  • the actuation of the back button 30 is not permitted when the offline ADC is running.
  • the actuation of the next button 31 is not permitted when the offline ADC is running.
  • the actuation of the cancel button 39 is not permitted when the offline ADC is running Cancel button 39 is allowed, all open files are closed and deleted. If the Next button 31 was printed, the sequencer is started again. The result data file is written and the sequencer ends automatically, whereby all open files are closed
  • FIG. 20 shows an illustration of a dialog 160 for the termination.
  • An info box 161 is provided for an output file.
  • the info box 161 is used to display the stored data file. Only the file name is displayed
  • the Easy ADC VSL file is also displayed in a Read Only EditBox 162.
  • info box 162 the generated ADC run file is displayed. Only the file name is displayed. The complete path is displayed in a tooltip
  • the number of total defects Total defects are displayed in an info box 163 In the info box 163 you can see the total number of all defects
  • the found defects are detected in an info box 164.
  • the redetection of the defects in percent is also shown in an info box I b6.
  • the defects detected with ADC are displayed absolutely and in percent.
  • An LED 149 shows in color whether the percentage value is above is a predefined value If the value is above the predefined value, the LED 149 is green, otherwise the LED 149 is red.
  • the number of classified defects Defects Classified "is displayed in a read info box 165.
  • the percentage of the classified defects, Class ⁇ f ⁇ ab ⁇ l ⁇ ty is also displayed in an info box 167.
  • the display of defects classified with ADC is absolute and in percent.
  • An LED 148 shows in color whether the percentage is above a predefined value
  • Green means that the percentage is above the predefined value. If the value is below, the display is red
  • a report button 147 displays a report dialog 170 (FIG. 21).
  • This report dialog 170 is user-dependent. An extended report is only displayed from user level engineer. The actuation of the finish button 146 is permitted. This report is ended -Dialog 170 is user dependent
  • FIG. 21 shows the report dialog 170 with the expanded display of the data in an info box 171.
  • the data, file information (P) output file name (+ path), recipe information (P), ADC HP recipe file name (+ path),
  • the report dialog 170 is provided with a print button 171.
  • a print preview of the ADC HP report is shown on the display 11.
  • the printout can be printed out using the standard printer.
  • the printout is in landscape format, since the paths are usually not fully displayed or printed out in portrait format become
  • FIG. 22 is a representation of a printed Easy ADC report 180.
  • the ADC HP report can be saved as a text file (extension TXT) when it is actuated

Landscapes

  • Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Immunology (AREA)
  • Health & Medical Sciences (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Pathology (AREA)
  • Analytical Chemistry (AREA)
  • Quality & Reliability (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
  • Image Analysis (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

Procédé d'apprentissage appliqué à une base de connaissances pour la classification automatique de défauts. L'utilisateur est déchargé de toute une série d'entrées, étant donné que le système exécute un mode d'apprentissage automatique qui nécessite un nombre réduit d'entrées effectuées par l'utilisateur.
EP04735896A 2003-07-12 2004-06-03 Procede d'apprentissage applique a une base de connaissances pour la classification automatique de defauts Withdrawn EP1644895A2 (fr)

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
DE10331646 2003-07-12
DE102004022717A DE102004022717B4 (de) 2003-07-12 2004-05-07 Verfahren zum Einlernen einer wissensbasierten Datenbasis für die automatische Fehlerklassifikation
PCT/EP2004/051008 WO2005006002A2 (fr) 2003-07-12 2004-06-03 Procede d'apprentissage applique a une base de connaissances pour la classification automatique de defauts

Publications (1)

Publication Number Publication Date
EP1644895A2 true EP1644895A2 (fr) 2006-04-12

Family

ID=34066319

Family Applications (1)

Application Number Title Priority Date Filing Date
EP04735896A Withdrawn EP1644895A2 (fr) 2003-07-12 2004-06-03 Procede d'apprentissage applique a une base de connaissances pour la classification automatique de defauts

Country Status (3)

Country Link
US (1) US7623698B2 (fr)
EP (1) EP1644895A2 (fr)
WO (1) WO2005006002A2 (fr)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9613411B2 (en) * 2014-03-17 2017-04-04 Kla-Tencor Corp. Creating defect classifiers and nuisance filters
US9798954B1 (en) * 2016-12-15 2017-10-24 Federal Home Loan Mortgage Corporation System, device, and method for image anomaly detection
WO2019194064A1 (fr) * 2018-04-02 2019-10-10 日本電産株式会社 Dispositif de traitement d'image, procédé de traitement d'image, système d'inspection d'apparence et procédé d'inspection d'apparence
US12086519B2 (en) * 2020-11-03 2024-09-10 Changxin Memory Technologies, Inc. Method and apparatus for setting wafer script, device and storage medium
US11756186B2 (en) * 2021-09-15 2023-09-12 Mitutoyo Corporation Workpiece inspection and defect detection system utilizing color channels

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6292582B1 (en) * 1996-05-31 2001-09-18 Lin Youling Method and system for identifying defects in a semiconductor
US6246787B1 (en) * 1996-05-31 2001-06-12 Texas Instruments Incorporated System and method for knowledgebase generation and management
US6104835A (en) * 1997-11-14 2000-08-15 Kla-Tencor Corporation Automatic knowledge database generation for classifying objects and systems therefor
US6408219B2 (en) * 1998-05-11 2002-06-18 Applied Materials, Inc. FAB yield enhancement system
WO2001041068A1 (fr) * 1999-11-29 2001-06-07 Olympus Optical Co., Ltd. Systeme de detection de defaut
US6456899B1 (en) * 1999-12-07 2002-09-24 Ut-Battelle, Llc Context-based automated defect classification system using multiple morphological masks
US20020184172A1 (en) * 2001-04-16 2002-12-05 Vladimir Shlain Object class definition for automatic defect classification
US6792366B2 (en) * 2001-12-11 2004-09-14 Hitachi, Ltd. Method and apparatus for inspecting defects in a semiconductor wafer

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See references of WO2005006002A3 *

Also Published As

Publication number Publication date
US7623698B2 (en) 2009-11-24
US20060245634A1 (en) 2006-11-02
WO2005006002A3 (fr) 2005-02-10
WO2005006002A2 (fr) 2005-01-20

Similar Documents

Publication Publication Date Title
DE102019114117B3 (de) Automatische Workflows basierend auf einer Erkennung von Kalibrierproben
DE60034055T2 (de) Vorrichtung und verfahren zur überprüfung die ortung aussagefähiger bildbereiche innerhalb einer probe in einem abbildungssystem
DE69429145T2 (de) Klassifikation und Prüfvorrichtung für Teilchen in einer Flüssigkeit
DE10011200A1 (de) Verfahren zur Bewertung von Strukturfehlern auf einer Waferoberfläche
DE102019209088A1 (de) Verfahren zum Bewerten einer Laserschnittkante, mobiles Endgerät und System
DE10300545A1 (de) Vorrichtung, Verfahren, Speichermedium und Datenstruktur zur Kennzeichnung und Speicherung von Daten
DE102021100444A1 (de) Mikroskopiesystem und verfahren zum bewerten von bildverarbeitungsergebnissen
WO2022135787A1 (fr) Procédé et dispositif de contrôle de qualité optique lors de la fabrication de cartes de circuits imprimés
DE102020126554A1 (de) Mikroskopiesystem und verfahren zum überprüfen von eingabedaten
DE102019133685A1 (de) Informationsverarbeitungssystem und -verfahren
DE69225998T2 (de) Verfahren zum Erfassen und Analysieren defekter Halbleiterelemente
DE102010007730B4 (de) Verfahren und Vorrichtung zum Einstellen eines geeigneten Auswerteparameters für ein Fluoreszenzmikroskop
EP1797533A1 (fr) Procede et dispositif pour segmenter une representation numerique de cellules
DE102020206753A1 (de) Verfahren und Vorrichtung zum Bestimmen eines Kontaminationszustands einer Oberfläche eines Facetten-Elements
DE10116830A1 (de) Dateiverwaltungsvorrichtung und Datenträger zum Aufzeichnen eines Dateiverwaltungsprogramms
WO2005006002A2 (fr) Procede d'apprentissage applique a une base de connaissances pour la classification automatique de defauts
DE102004022717B4 (de) Verfahren zum Einlernen einer wissensbasierten Datenbasis für die automatische Fehlerklassifikation
EP3482348A1 (fr) Procédé et dispositif pour catégoriser une surface de rupture d'un élément
DE19637741B4 (de) Automatisches Zytologisches Probenanalysesystem unter Verwendung von individualisierten Patientendaten
EP4016082B1 (fr) Procédé et dispositif de détection d'une présence d'un type de motif fluorescent sur une section d'organe par microscopie immunofluorescente
DE102020127071B3 (de) Verfahren und Mikroskop mit einer Einrichtung zum Erfassen von Verlagerungen einer Probe gegenüber einem Objektiv
DE102008001174A1 (de) Inspektionssystem und -verfahren für die optische Untersuchung von Objektoberflächen, insbesondere von Waferoberflächen
WO2022238343A1 (fr) Procédé d'évaluation de modèles de simulation en microscopie
DE10231981A1 (de) Abbildungsvorrichtung für ein Mikroskop
DE102020130260A1 (de) Härteprüfer und Programm

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

17P Request for examination filed

Effective date: 20060207

AK Designated contracting states

Kind code of ref document: A2

Designated state(s): DE FR GB

DAX Request for extension of the european patent (deleted)
RBV Designated contracting states (corrected)

Designated state(s): DE FR GB

17Q First examination report despatched

Effective date: 20060602

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: VISTEC SEMICONDUCTOR SYSTEMS GMBH

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN

18D Application deemed to be withdrawn

Effective date: 20080103