Method and apparatus for simultaneously cutting multiple slices from a workpiece

Abstract

The invention relates to a method and a wire saw for simultaneously cutting a plurality of slices from a workpiece during a slicing process. The method is characterized in that a non-linear pitch function dTAR(WP) is selected based on a target thickness characteristic value function TTAR(WP), a pitch function dINI(WP) and a thickness characteristic value function TINI(WP), wherein dTAR(WP) associates adjacent grooves in the outer shell portion of the wire guide roll at a position WP with a pitch during the slicing process; TINI(WP) associates slices obtained at the position WP during a plurality of previous slicing processes with thickness characteristic values measured on the slices using the wire saw; dINI(WP) associates adjacent grooves in the outer shell portion of the wire guide roll at the position WP with the pitch during the previous slicing processes; TTAR(WP) associates slices that have been cut at the position WP during the slicing operation with target thickness values; and WP denotes an axial position of the adjacent grooves with respect to an axis of the wire guide roll.

Description

Technical Field

[0001] The present invention relates to a method and apparatus for simultaneously cutting multiple slices from a workpiece by means of a wire saw, and a wire saw suitable for carrying out the method. Background Technology

[0002] Many applications require thin, and particularly uniform, slices of specific materials. One example of a slice subject to particularly stringent requirements regarding uniformity and planar parallelism on both the front and back sides is a slice made of semiconductor material; these slices are called “wafers” and are used as substrates for manufacturing microelectronic components. Wire sawing, a technique used to cut multiple slices simultaneously from a workpiece, is particularly important for producing such slices due to its economic efficiency.

[0003] For example, the method and apparatus required for this purpose are known from DE 10 2016 211 883 A1 or DE 10 2013 219 468 A1. In this method, a wire is helically guided around at least two wire guide rollers, such that the two wire guide rollers stretch a wire mesh consisting of wire segments extending parallel to each other, the wire mesh facing the workpiece. The wire guide rollers are in the form of straight cylinders, rotatable about their axes, and have a plurality of annular closed grooves on their side surfaces, the plurality of annular closed grooves extending in a plane perpendicular to the axis and guiding the wire. The rotation of the wire guide rollers creates relative motion between the wire segments and the workpiece. The wire saw also has a feeding device (apparatus) to which the workpiece is secured via a saw beam (sacrificial beam) and which feeds the workpiece to the wire mesh. The relative motion, in the presence of abrasive, results in the removal of material from the workpiece through contact with the wire mesh. With continued feeding using relative motion and in the presence of abrasive, wire segments form slicing gaps in the workpiece by material removal, and the wire mesh slowly works through the entire workpiece until it is completely within the saw beam, whereby the workpiece is attached to the saw beam by means of adhesive joints. The workpiece is then completely separated into slices, which hang on the saw beam like the teeth of a comb, held only by the adhesive joints.

[0004] Wire sawing can be categorized into wire abrasive slivers and wire grinding slivers. Wire abrasive slivers are characterized by a three-body interaction between the wire and the workpiece, where the wire initially does not carry the abrasive, which is supplied loosely in slurry form. Wire grinding slivers, on the other hand, are characterized by a two-body interaction between the abrasive and the workpiece, where the abrasive is firmly bonded to the surface of the wire.

[0005] Wire is typically made of hypereutectoid pearlitic steel (piano wire). The steel wire is usually coated with a very thin layer of brass or zinc, whose ductility acts as a drawing aid (solid lubricant) during wire production and also provides corrosion protection for the finished wire. In in-line grinding, the abrasive is typically silicon carbide (SiC) suspended in a viscous carrier fluid composed of oil or ethylene glycol. In in-line grinding, the abrasive is typically diamond, which is fixed to the wire surface by means of electroplating or synthetic resin bonding, or by form fitting (press-in, roll-in). Wire is supplied from wire stock, usually in spool form, to a wire mesh. This spool is called the new wire spool. After passing through the wire mesh, the wire is supplied to wire stock, also usually in spool form. This spool is called the wear wire spool. In in-line sawing, smooth wire (plain or ordinary wire) is used, while in in-line grinding, so-called structured wire is also used. The smooth line takes the form of a cylinder with a very large height (i.e., the length of the line), and the line diameter corresponds to the diameter of this cylinder. The structured line comprises a smooth line with multiple protrusions and depressions arranged along its entire length in a direction perpendicular to the longitudinal direction of the line. The surface of the structured line thus has protrusions and depressions in which slurry can accumulate without being wiped away when the line enters the slicing gap or during further processing as the line passes through the slicing gap. The structured line has the diameter of the base smooth line and an effective diameter corresponding to the diameter of the cylinder with the smallest diameter that completely encompasses all the depressions and protrusions (the so-called envelope). WO 2006 / 067062 A1 describes an example of a structured line.

[0006] The slice is typically in the form of a cylinder with a small height, and therefore has a bottom top surface (back side), a top top surface (front side), and side surfaces (edges of the slice). The front and back sides of the slice form its primary surfaces, and the edges of the slice form secondary surfaces.

[0007] The groove on the wire guide roller that first contacts the new wire supplied to the wire mesh can be called the first groove. The half of the wire mesh containing the first groove is called the new wire (supply) side of the wire mesh. The slice formed on its main surface by the wire segment of the first groove can be called the first slice, and the half of the workpiece containing the first slice can be called the starting end of the workpiece. The groove that last contacts the wear line that has detached from the wire mesh can be called the last groove. The side of the wire mesh containing the last groove is called the wear line (detachment) side of the wire mesh. The slice formed on its main surface by the wire segment of the last groove can be called the last slice, and the half of the workpiece containing the last slice can be called the ending end (end) of the workpiece.

[0008] One problem with wire sawing is that the thickness of the slices obtained through the slicing operation has a finite width distribution. Typically, further (machined) processing of wire-sawed slices, such as by grinding, polishing, etching, or a series of these steps, is subject to stringent requirements regarding the thickness distribution of the wire-sawed slices. For example, no slices or only a maximum proportion of slices determined by economic considerations can be below the minimum thickness to maintain a minimum material thickness above the target slice thickness required for subsequent material removal in each case, so as to completely remove surface layers damaged by wire sawing. Meanwhile, it is generally true that wire-sawed slices must not exceed a given maximum thickness for subsequent processing, for example, because excessive subsequent material removal is uneconomical. For instance, slices in grinding operations following wire sawing, as part of a so-called batch processing of multiple slices simultaneously, must have an initial thickness distribution that is very narrow, because otherwise the load on the worktable would be unevenly distributed across the individual slices, potentially overloading them and causing them to break, or causing worktable misalignment, resulting in slices with a wedge-shaped thickness profile. For example, during subsequent grinding processes, the grinding wheels may become clogged and dull, or, in the case of multiple grinding wheels (double-sided machining), if the amount of material removed during the grinding operation is too large, there may be uneven material removal from the front and back sides of the slice. Besides the minimum and maximum thicknesses mentioned as examples, other thickness parameters can be envisioned, for which the wire saw slices must have a narrow distribution to allow for economical post-wire sawing processing operations and to obtain slices with the desired high uniformity and planar parallelism on both the front and back sides, such as the average slice thickness.

[0009] The grooves of a line guide roller typically have a V-shaped or U-shaped cross-sectional profile, comprising a groove side, a groove bottom, and a recess (dent) depth. The groove side opens to the ungrooved cylindrical side surface of a cylinder, the shape of which the line guide roller has, and has an opening angle that "grips" the line and prevents (counters) the line from accidentally "jumping out" during rotation of the line guide roller. The groove bottom is the portion of the profile closest to the axis of the line guide roller, and the line supplied to the line guide roller with a given tension will abut against this portion due to that tension. The recess depth (groove depth) represents the distance between the point of the groove closest to the axis of the line guide roller and the side surface of the line guide roller. For the groove bottom, straight / linear (can-shaped), pointed (V-shaped), and curved (U-shaped) forms are known.

[0010] In in-line grinding, the rolling of loose abrasive between the wire and the workpiece, which causes the removal of brittle, corrosive material from the workpiece, also leads to undesirable material removal from the wire: as a result of wear caused by abrasion, the diameter of smooth wires or the effective diameter of structured wires decreases. In contrast, in in-line grinding, no movement occurs between the abrasive and the wire: particularly, the wire diameter changes only slightly, primarily as a result of very low abrasion of the diamond or as a result of occasional detachment of individual diamonds from their bond with the wire surface.

[0011] It is further known that, as a result of abrasion, the thinning of the line, especially in the case of line grinding slices, results in an increase in slice thickness from the new line side to the worn line side due to the constant pitch of the grooves in the line guide rollers.

[0012] In the prior art, measures are known for counteracting uneven slice thickness caused in this way.

[0013] DE 10 2010 005 718 A1 describes a line guide roller in which the radius of curvature of the bottom of the U-shaped groove of each groove, the opening angle of the groove side, or the groove depth varies along the axis of the line guide roller. Variations in groove pitch are also provided.

[0014] Many patents describe cylindrical line guide rollers with a pitch that decreases continuously from the first groove to the last groove: CN 2015 166 49U, KR 100 445 192B, CN 2012 25584Y and CN 101 879 759A.

[0015] However, none of the known wire guide rollers in the prior art can solve the problem of the wide (wide range) distribution of slice thickness after wire sawing. Summary of the Invention

[0016] Therefore, the object of the present invention is to provide a method for providing slices that have a sufficiently narrow distribution in terms of uniformity and planar parallelism on the front and back sides, with regard to thickness parameters that are critical for subsequent processing.

[0017] The object of the present invention is achieved by a method for simultaneously cutting multiple slices from a workpiece having a workpiece axis by means of a wire saw during a slicing operation. The method includes moving the wire of the wire saw relative to the workpiece by rotating the wire guide roller of the wire saw about its axis. The wire guide roller has a housing portion provided with wire guide grooves and stretches a wire mesh composed of wire segments of the wire.

[0018] In the presence of abrasive, the workpiece is fed to the wire mesh perpendicular to the wire segment, the wire segment operating completely through the workpiece, wherein...

[0019] The nonlinear pitch function dTAR(WP) is selected based on the target thickness eigenvalue function TTAR(WP), the pitch function dINI(WP), and the thickness eigenvalue function TINI(WP). dTAR(WP) is the pitch assigned (allocated) to the adjacent grooves in the outer shell of the wire guide roller at position WP during the slicing operation. TINI(WP) is the thickness eigenvalue measured on the slice obtained by means of the wire saw at position WP during multiple previous slicing operations.

[0020] dINI(WP) is the pitch of the adjacent groove at position WP in the outer shell of the line guide roller during the previous slicing operation, TTAR(WP) is the target thickness characteristic value assigned to the slice cut at position WP during the slicing operation, and WP represents the axial position of the adjacent groove about the axis of the line guide roller.

[0021] The pitch dTAR(WP) is preferably selected such that the equation dTAR(WP) = dINI(WP) + TTAR(WP) - TINI(WP) is satisfied.

[0022] This method considers the thickness characteristics of slices from i previous slicing operations performed by wire sawing, using a wire guide roller with an initial distribution of groove pitches corresponding to a pitch function dINI(WP). A second pitch function dINI(WP) is predefined, and for each previous slicing operation, a groove pitch dependent on the position WP of the adjacent groove is assigned to adjacent grooves in the outer shell of the wire guide roller. The groove pitch also corresponds to the distance between two wire segments in the wire mesh traveling through adjacent grooves. The position WP is preferably a discrete length specification indicating the axial position of the adjacent grooves about the axis of the wire guide roller, and thus also indicating the position of the slice in the wire mesh.

[0023] For example, the initial distribution of the pitch function dINI(WP) can be selected such that the pitch between adjacent grooves decreases by the same amount continuously as the position WP changes. The method according to the invention also includes the designation of a target thickness feature function TTAR(WP), which assigns a target thickness feature value to the slice cut at position WP during the slicing operation. Minimum thickness, maximum thickness, and average thickness are particularly suitable as target thickness features. Target thickness features are particularly suitable for material removal requirements in subsequent processing steps. For example, the target thickness feature value could be the minimum thickness that each slice must have, such that sufficient material removal is available for subsequent grinding processes, for example, to completely remove layers near the surface that have been damaged by wire sawing.

[0024] Thickness characteristic values ​​can be based on full-area measurements. However, measuring the surface of the slice along its diameter, particularly along its diameter in the supply direction, is sufficient and preferred for determining the thickness characteristic values. Measurements along the diameter in the supply direction are also used as the basis for the thickness characteristic values ​​of the examples presented subsequently and the thickness characteristic values ​​of the comparative examples.

[0025] Starting from i previous slicing operations, i corresponding slices with a thickness feature value TINI(WP) are generated for each position WP. The thickness feature value function TINI(WP) assigns a common thickness feature value to the slices obtained at position WP by means of a wire saw during previous slicing operations. This common thickness feature value is measured and formed by averaging (arithmetic mean) or calculated as a quantile. The minimum thickness of the slice is, for example, such a thickness feature value and corresponds to the value of the minimum distance between corresponding measurement points on the front and back sides of the slice. The number of measurement points on the slice side (face) is preferably at least 25. The measurement points are preferably arranged (presented) along the diameter in the supply direction. Instead of the average value of the minimum thickness, for example, the average value of the maximum thickness or the average value of the average thickness of the slices at position WP can also be used as the thickness feature value, or the thickness of the quantile of the minimum thickness, the quantile of the maximum thickness, or the quantile of the average thickness can be used.

[0026] Regarding the thickness characteristic value, all slices are evaluated from the same location WP in the mesh. For example, the thickness characteristic value could be the (-3σ) quantile (=0.135 percentile, i.e., 0.135%). The (-3σ) quantile of the minimum thickness of slices from previous slicing operations means that 0.135% of all slices at location WP are below that minimum thickness.

[0027] As an example, the target thickness eigenvalue function TTAR(WP) can be specified such that it assigns the same constant to each location WP, TTAR(WP) = constant, for example, because the (-3σ) quantile of the minimum thickness of the slice to be cut means (pre-defined) that it has the same value for all locations WP.

[0028] Based on the specified target thickness eigenvalue function TTAR(WP) and the thickness eigenvalue function TINI(WP) derived from measurement, a first pitch function dTAR(WP) is determined, which assigns the groove pitch between two adjacent grooves associated with each position WP. The first pitch function dTAR(WP) is preferably obtained by means of the following equation: dTAR(WP) = dINI(WP) + TTAR(WP) - TINI(WP).

[0029] This invention can be implemented using a smooth line, which is in the form of a straight cylinder with a very large height (i.e., the length of the line). The diameter of the cylinder represents the diameter of the smooth line, and the axis of the cylinder represents the longitudinal axis of the smooth line.

[0030] This invention is particularly preferably implemented by means of a wire grinding and slicing method using structured lines, which are smooth lines with a plurality of recesses and protrusions in a direction perpendicular to the line axis along their entire length. In the case of structured lines, the longitudinal axis of the line is the axis of a straight cylinder with a minimum diameter that completely encloses the structured line (envelope), the diameter of which is called the effective diameter of the structured line.

[0031] This invention can be implemented using a unidirectional line travel configuration. In the case of unidirectional line sawing, the line moves in exactly one direction along its longitudinal axis throughout the entire time period of the slicing operation. In this case, the rate of line supply can be variable.

[0032] This invention is particularly preferably implemented using a pilgrim step method. This pilgrim step method describes the movement of a line in the form of a continuous series of pairs of reversed directions throughout the slicing operation. Each pair of reversed directions includes the line moving a first length along its longitudinal direction in a first direction and then moving a second length in a second direction exactly opposite to the first direction, wherein the first length is selected to be greater than the second length. The pairs of reversed directions can differ in terms of the length and speed of the line movement.

[0033] As an example, the first length could be 320m, and the second length could be 240m. During a pair of directional reversals, 320m + 240m = 560m of wire thus moves through the slicing gap; however, in one complete pilgrimage cycle, the wire moves only a further total of 320m – 240m = 80m from the new wire stock to the worn wire stock. In the example described, as a result of wire sawing in the pilgrimage step method, the wire is utilized with a coefficient of (320 + 240) / (320 – 240) = 560 / 80 = 5. This is advantageous because very high wire loss would lead to a situation where the wire is used only once (unidirectional slicing method), and the worn wire would only suffer slight wear, which is very uneconomical. As a result of the selection of the first length L1 and the second length L2, a utilization rate r = (L1 + L2) / (L1 - L2) is preferably selected such that the wire is subjected to wear to such an extent that, under a selected tensile stress (whereby the selected tensile stress is used to supply the wire to or remove it from the wire mesh), the wire does not break and has not become non-circular (elliptical) as a result of uneven wear, resulting in a variability in the width of the resulting slicing gap. This variability in width will lead to inhomogeneity and a lack of planar parallelism in terms of the flatness of the front and back sides of the adjacent slices defining the slicing gap, making the slices unsuitable for the proposed (claimed) application.

[0034] The axis of the workpiece is preferably oriented parallel to the axis of the wire guide roller. However, for example, if the workpiece is an ingot (bulk) made of single-crystal semiconductor material and it is desirable to obtain a slice (wafer) with defined crystal dislocations, it may be advantageous to pivot the axis of the workpiece relative to the axis of the wire guide roller. For example, if the wafer is additionally provided with a layer of further single-crystal semiconductor material applied by epitaxy after wire sawing and subsequent processing (grinding, grinding, etching, polishing), dislocation (misalignment) is advantageous because dislocation is accompanied by an increased density of atomic steps at the growth boundary surface, which makes particularly defect-free growth of the epitaxial layer possible.

[0035] This method can also be implemented using a line guide roller whose groove has a depth corresponding to a depth function t(WP) dependent on the position WP. As a result of line wear, the depth is preferably selected to be proportional to a decrease in the line diameter (smooth line) or a decrease in the effective diameter (structured line). The depth of the groove is therefore understood to refer to the size (degree) of the groove in the direction toward the axis of the line guide roller, that is, the difference between the radius of the straight cylinder (the line guide roller in the form of said cylinder) and the distance of a point on the groove profile (which has the minimum distance relative to the axis of the line guide roller) relative to the axis of the line guide roller.

[0036] This objective is also achieved by a wire saw for simultaneously cutting multiple slices from a workpiece, the wire saw comprising a wire guide roller that stretches a wire mesh and has an axis and a housing portion, the housing portion having multiple grooves for guiding the wire, the pitch between adjacent grooves at position WP following a nonlinear pitch function dTAR(WP), where WP represents the axial position of adjacent grooves about the axis of the wire guide roller.

[0037] The nonlinear pitch function dTAR(WP) preferably satisfies the equation dTAR(WP)=dINI(WP)+TTAR(WP)-TINI(WP), where TTAR(WP) is the target thickness feature value function, which assigns the target thickness feature value to the slice to be cut at position WP, TINI(WP) is the thickness feature value function, which assigns the thickness feature value measured on the slice to the slice obtained at position WP by means of a wire saw during multiple previous slicing operations, and dINI(WP) is the pitch function, which assigns the pitch of the adjacent groove in the outer shell of the wire guide roller at position WP to the pitch presented during previous slicing operations.

[0038] Exemplary embodiments of the present invention will now be described with reference to the accompanying drawings. Attached Figure Description

[0039] Figure 1 The components of a wire saw are shown.

[0040] Figure 2 The line guide roller is shown in cross-section.

[0041] Figure 3 The (-3σ) quantile of minimum thickness TMIN for the comparative example and the first example, depending on the location WP, is shown.

[0042] Figure 4 The groove pitch WGRP depending on the position WP is shown in the comparative example, the first example, and the second example.

[0043] Figure 5 The cumulative frequency CF of minimum thickness TMIN and maximum thickness TMAX of slices from the comparative example, the first example and the second example are shown.

[0044] Figure 6 The maximum thickness TMAX of the slices, which depends on the location WP, is shown in relation to the comparative example and the first example.

[0045] Figure 7 The maximum thickness TMAX of slices from the comparative example and the second example is shown, depending on the position WP.

[0046] Figure 8The minimum thickness TMIN of the slices associated with the comparative example and the second example is shown, depending on the position WP.

[0047] Figure 9 The (+3σ) quantile of the maximum width KMAX of the slicing gap is shown from the grinding slicing operation performed using smoothed lines and structured lines.

[0048] Figure 10 The typical (characteristic) thickness characteristics of semiconductor slices obtained by grinding and slicing are shown.

[0049] List of reference numerals and abbreviations used

[0050] Line 1

[0051] 2-line section

[0052] 3 Left side guide rollers

[0053] 4 right-side guide rollers

[0054] 5. Axis of the left side guide roller

[0055] 6. The axis of the right-side guide roller

[0056] 7. Rotation of the left side guide roller

[0057] 8. Rotation of the right-side guide roller

[0058] Supply of 9 lines (new lines)

[0059] Line 10 is detached (wear line).

[0060] 11-line network

[0061] 12 workpieces

[0062] Direction of movement of section 13

[0063] 14. Workpiece axis

[0064] 15-beam saw

[0065] 16 Adhesive Joint

[0066] 17 Supply Direction

[0067] 18 grooves

[0068] 19 Left nozzle

[0069] 20 Right side nozzle

[0070] 21 Exit Opening

[0071] 22 Left-side jet

[0072] 23 Right-side jet

[0073] 24 cutting depth

[0074] 25 slice gap

[0075] 26 Marking notch

[0076] 27-line guide roller

[0077] Core section of 28-line guide roller

[0078] Coating of 29-line guide rollers

[0079] 30 is the (-3σ) quantile of the minimum thickness when the groove pitch is linearly reduced with respect to position WP.

[0080] 31 Regression curve with respect to 30

[0081] 32. (-3σ) quantile of minimum thickness when the groove pitch is changed non-linearly with respect to position WP.

[0082] Regression curve of 33 on 32

[0083] 34. Linearly decreasing groove pitch

[0084] 35 groove pitch for a constant (-3σ) quantile of minimum thickness

[0085] 36 Groove pitch for constant (+3σ) quantiles at maximum thickness

[0086] 37. Cumulative frequency CF of minimum thickness under linearly decreasing groove pitch

[0087] 38. Cumulative frequency CF of the maximum thickness under linearly decreasing groove pitch

[0088] 39. The cumulative frequency CF of the minimum thickness under the condition of optimizing the nonlinearly varying groove pitch for a constant (-3σ) quantile of the minimum thickness.

[0089] 40. The cumulative frequency CF of the maximum thickness under the condition of nonlinearly varying groove pitch optimized for a constant (-3σ) quantile of the minimum thickness.

[0090] 41. The cumulative frequency CF of the minimum thickness under the condition of optimizing the nonlinearly varying groove pitch for a constant (+3σ) quantile of the maximum thickness.

[0091] 42. The cumulative frequency CF of the maximum thickness under the condition of nonlinearly varying groove pitch optimized for a constant (-3σ) quantile of the maximum thickness.

[0092] 43. The (+3σ) quantile of maximum thickness TMAX with respect to position WP when the groove pitch is linearly reduced.

[0093] 44 Regression curve on 43

[0094] 45. Maximum thickness TMAX with respect to position WP under the condition of nonlinear groove pitch optimization for a constant (-3σ) quantile of minimum thickness.

[0095] 46 Regression curve on 45

[0096] 47. Minimum thickness TMIN with respect to position WP, optimized for a constant (+3σ) quantile of maximum thickness with respect to a nonlinear groove pitch.

[0097] Regression curve of 48 on 47

[0098] 49. The (+3σ) quantile of the maximum slice gap width KMAX when the groove pitch is linearly reduced for the structured line.

[0099] Regression curve of 50 on 49

[0100] 51. The (+3σ) quantile of the maximum slice gap width KMAX when the groove pitch is linearly reduced for a smooth line.

[0101] 52 Regression curve on 51

[0102] 53 Monotonically decreasing section

[0103] 54 monotonically increasing segments

[0104] 55 Monotonically decreasing section

[0105] 56-cut wedge

[0106] 57 Saddle

[0107] 58. Scanning direction along the supply direction

[0108] 59 slices center plane

[0109] 60 slices

[0110] 61. (+3σ) quantile of maximum thickness TMAX with respect to position WP under the condition of nonlinear groove pitch optimization for a constant (+3σ) quantile of maximum thickness.

[0111] Regression curve of 62 on 61

[0112] α opening angle

[0113] CF cumulative frequency

[0114] d groove pitch

[0115] The pitch between the second groove and the first groove d1

[0116] dINI(WP) pitch function

[0117] dTAR(WP) nonlinear pitch function

[0118] Diameter of D-line guide roller

[0119] DIFF difference

[0120] i-cutting counter

[0121] Maximum width of KMAX slice gap

[0122] Length of L-line guide roller

[0123] r is the radius of curvature at the bottom of the groove.

[0124] s is the width of the cylindrical portion between adjacent grooves.

[0125] s1 Width of the cylindrical portion between the first and second grooves

[0126] s(WP) width function

[0127] The depth of the groove

[0128] t(WP) depth function

[0129] Depth of the first groove t1

[0130] Minimum thickness of TMIN slices

[0131] Maximum thickness of TMAX slices

[0132] TINI(WP) thickness eigenvalue function

[0133] TTAR(WP) target thickness eigenvalue function

[0134] VAR variance

[0135] Groove pitch of WGRP line guide roller

[0136] The position of the groove pitch or slice position of two adjacent grooves on the WP line guide roller. Detailed Implementation

[0137] Figure 1The components of a wire saw are shown, which help to understand the invention: wire 1 is supplied from wire stock / prepared stock (new wire spool, not shown) in direction 9 and is spirally guided in grooves 18 surrounding at least two wire guide rollers 27, and a wire mesh 11 composed of wire segments 2 extending parallel to each other is produced. Figure 1 The example of the wire saw shows two wire guide rollers 27, namely the left wire guide roller 3 and the right wire guide roller 4. As a result of the rotation of the wire guide rollers 3 and 4 about their axes 5 and 6 in the rotational directions 7 and 8, the wire segment 2 and the wire mesh 11 move relative to the workpiece 12 in a relative manner in direction 13. The axes 5 and 6 of the wire guide rollers 3 and 4 are oriented parallel to each other. The workpiece 12 has an axis 14 and is connected to the saw beam 15 by a glue joint (glue joint) 16. The saw beam is fastened to a supply device (not shown) that supplies the workpiece 12 in a direction 17 perpendicular to the wire mesh 11. Preferably, the workpiece 12 is made of a single-crystal semiconductor material, such as an ingot made of single-crystal silicon. In this case, the workpiece is in the shape of a straight cylinder and has marking notches 26 for crystal orientation provided parallel to its axis 14. On both sides of the workpiece 12, jets 22 and 23 of liquid cutting aid are supplied to the wire mesh 11 through nozzles 19 and 20 via openings 21 in the nozzles. If the wire saw is operated using the grinding slicing method, the cutting aid is a cooling lubricant that does not contain abrasive substances, and wire 1 is a piano wire whose surface contains firmly bonded diamond particles with abrasive properties. If the wire saw is operated using the grinding slicing method, the cutting aid is a slurry of silicon carbide (SiC) in a liquid carrier consisting of oil or ethylene glycol.

[0138] As a result of supply along direction 17, wire mesh 11 comes into contact with workpiece 12. As a result of relative motion 13 between wire mesh 11 and workpiece 12, in the presence of abrasive material, wire segments 2 cause material removal from workpiece 12. With further supply 9 with relative motion 13, the material removal caused by wire segments 2 forms slicing gaps 25 in workpiece 12. The depth reached by the wire segments in workpiece 12 that have passed through the workpiece is called the cutting depth 24. When the cutting depth 24 exceeds the diameter of the workpiece, all wire segments 2 of wire mesh 11 have passed through the workpiece slice and reached the retaining beam 15: the workpiece has been completely separated into multiple slices. After passing through the wire mesh, wire 1 leaves wire mesh 11 along direction 10 and goes to the used wire stock (wearing wire spool, not shown).

[0139] Figure 2A line guide roller 27 with a diameter D and a length L in a cross-section including axis 6 is shown. The line guide roller 27 includes a core portion 28 and a shell portion 29. The core portion 28 may, for example, include a chamber for cooling the line guide roller. The shell portion 29 includes a coating made of a wear-resistant material, such as thermosetting polyurethane. The coating has an annular closed groove 18 arranged in a plane perpendicular to axis 6 and extending (traveling) around the line guide roller. According to the invention, depending on its position WP, the pitch of adjacent grooves measured along axis 6 from a first groove with depth t1 to the last groove is selected such that the pitch follows a nonlinear pitch function dTAR(WP). The nonlinear pitch function dTAR(WP) depends on the thickness characteristic value of the slice from the previous slicing operation, the pitch function of the groove pitch during the previous slicing operation, and a specified target thickness characteristic value function TTAR(WP). As shown, for example, by different pitches of the first groove pitch d1 and any groove pitch d, dTAR(WP) varies in a nonlinear manner. The groove has an opening angle α on its side and a radius of curvature r at its bottom. Furthermore, according to one embodiment of the invention, the groove depth t can vary with position WP according to a depth function t(WP), as shown by the depth t1 of the first groove and the depth t of any arbitrary groove, as illustrated in the example. Depending on the groove pitch dTAR(WP), which varies with position WP, and optionally depending on the variable groove depth t(WP), when the radius of curvature r at the bottom of the groove is the same and the opening angle α on the side of the groove is the same, the residual cylindrical region on the side surface between the grooves has different widths according to the width function s(WP). Width s1 is illustrated by an example of the cylindrical portion between the first and second grooves.

[0140] Figure 10 A schematic diagram of a semiconductor slice 60 obtained by wire grinding is shown in the thickness illustration, i.e., in a perspective view, a planar central region 59 is present between the front and rear sides of the semiconductor slice. In this thickness illustration, any undulations in the slice that would be revealed in the curvature of the central region 59 have been eliminated. This central region contains precisely the midpoint of the paths connecting corresponding points on the front and rear sides, respectively. In each case, the corresponding point is the point where the connecting path is perpendicular to the central region. In the thickness illustration, the semiconductor slice 60 obtained by wire grinding typically has a cutting wedge 56, i.e., a slice thickness reduction zone in the initial contact area between the workpiece and the wire mesh. Cutting (i.e., the feeding of the workpiece during the slicing operation) is performed along the scan line 58 from... Figure 10 The line moves from left to right. It is perpendicular to the supply direction, i.e., perpendicular to scan line 58. Preferably, the thickness characteristic value is measured parallel to scan line 58.

[0141] The cutting wedge 56 is generated because new wire is present in the wire mesh at the start of the slicing operation, resulting in a wide slicing gap and thus a reduced slice thickness. The wire present in the wire mesh at the start of the slicing operation is the wire that was last cut across the saw beam in the previous slicing operation. The saw beam is made of a material that is generally softer and easier to work with than semiconductor materials. The saw beam is made of, for example, sintered carbon, composite plastics, or glass. The cut through the saw beam therefore results in only a small amount of wear on the wire present in the wire mesh at the start of the slicing operation. Once a cutting depth of several centimeters has been reached in the workpiece, with further wire feeding and initial wear due to contact with the workpiece, the wire in the wire mesh is replaced by wire segments whose diameter is determined by the ratio of chip volume to wire feed rate, resulting in a thinner slicing gap and a greater slice thickness.

[0142] In the region of maximum engagement length of the workpiece, that is, at the center of scan line 58, the thickness of the semiconductor slice 60 obtained by wire grinding is saddle-shaped 57. This is due to the fact that during entry into the workpiece and during further penetration in the direction of the workpiece center, as a result of the cutting aid being wiped off, used up, and dripped off the workpiece edge, the wire segment loses most of the cutting aid (slurry) surrounding it, which also determines the width of the resulting slice gap. Figure 10 Two such thickness-reduced regions 57 are shown at the edges of the semiconductor slice where the travel direction along the line is opposite to each other. During the sawing process (performed) according to the pilgrimage-style method, the line segment has thus entered the workpiece in two directions.

[0143] This behavior and the resulting slice shape explained as follows: Figure 9 The graph (profile) shows the observed maximum width of the slicing gap depending on the position WP. The set of points 51 and the associated regression curve 52 show the (+3σ) quantile of the maximum width of the slicing gap from approximately 2000 slicing operations, where a silicon slice (wafer) with a diameter of 300 mm was cut using a smooth line. For comparison, the set of points 49 and the associated regression curve show the (+3σ) quantile of the maximum width of the slicing gap depending on the position WP, where approximately 1000 slicing operations were performed using a structured line to obtain a semiconductor slice with a diameter of 300 mm.

[0144] The graph of regression curve 52 shows that, when using a smoothing line, the maximum width of the slice gap depends non-linearly on the position WP, and therefore cannot be compensated for by a linear decrease in the groove pitch to obtain slices with, for example, the most similar thickness possible from each position WP. The set of points 49 and the graph of the associated regression curve 50 show that, when using a structured line, the maximum width of the slice gap dependent on the position WP can even increase at some positions. Regression curve 50 has, for example, a segment 53 with a non-linearly decreasing maximum slice gap width; further, a segment 54 in which the maximum slice gap width increases with increasing position WP, despite the fact that the effective diameter of the structured line decreases with increasing cumulative engagement length into the workpiece; and a segment 55 with a non-linearly decreasing maximum slice gap width.

[0145] As a comparative example Figure 3 A set of points 30 representing the (-3σ) quantile of the minimum thickness TMIN of semiconductor slices obtained from approximately 1000 slicing operations using a structured wire with a core wire diameter of 175 μm, plotted relative to the position WP in the wire mesh, is shown. Straight cylindrical silicon (100) single crystals are cut using wire guide rollers with a pitch that decreases linearly from the first groove to the last groove, where up to three ingots with a total length of up to 400 mm (variable) are sequentially mounted on a saw beam during the slicing operations. Measurements are taken only for every 25th semiconductor slice, resulting in a set of points 30 exhibiting considerable diffusion with respect to a third-order polynomial regression curve 31 despite the high number of slicing operations and the (-3σ) quantile, which is typically perturbed for statistical reasons. For each slicing operation, a 60 km structured wire is used, resulting in a 12 μm difference between the effective diameter of the new wire supplied to the wire mesh and the effective diameter of the used wire removed from the wire mesh. The groove pitch between the first and second grooves is 1134 μm, and the groove pitch between the penultimate and last grooves is 1122 μm. Therefore, the total linear decrease is 1134 μm – 1122 μm = 12 μm, making the minimum thickness TMIN of the first and last semiconductor slices preferably the same. However, it has been shown that despite the linear decrease in groove pitch, according to the set of points 30, semiconductor slices from the middle position deviate from each other by up to 5.7 μm in terms of regression curve 31, and by up to 12.3 μm in terms of quantiles. Therefore, despite the linear decrease in groove pitch, valuable semiconductor material is wasted as a result of producing unnecessarily thick semiconductor slices.

[0146] Figure 4The initial distribution of the groove pitch 34 as a pitch function for the comparative example, the first example, and the second example is shown. A linear decrease in the groove pitch was chosen in each case. For the first and second examples, this graph of the groove pitch 34 corresponds to the function dINI(WP), with the slicing operation of the comparative example forming the previous slicing operation. Figure 4 A graph of the groove pitch 35 is also shown, corresponding to the nonlinear pitch function dTAR(WP) calculated for the first example using the equation dTAR(WP) = dINI(WP) + TTAR(WP) - TINI(WP). The function TINI(WP) represents the thickness at the (-3σ) quantile of the minimum thickness of the slice from the previous slicing operation, and the target thickness eigenvalue function represents the constant (-3σ) quantile of the minimum thickness: TTAR(WP) = 878.6 μm. The graph of the groove pitch 35 does not merely decrease monotonically, but also temporarily exhibits a region of gradual increase. This region of the graph has a similar characteristic to... Figure 9 The regression curve for the (+3σ) quantile of the maximum slice gap width shown in the figure has a similar structure to regions 53, 54, and 55.

[0147] As a first example, Figure 3 A set of points 32 is shown representing the (-3σ) quantile of the minimum thickness TMIN from approximately 1000 slicing operations performed by grinding the wafers, with the thickness characteristic value measured only on every 25 semiconductor wafers. Except for the groove pitch, the conditions for grinding the wafers are the same as in the comparative example. The groove pitch 35 follows a nonlinear pitch function dTAR(WP) calculated by means of an equation. The fitting polynomial 33 for the set of points 32 is almost (substantially) constant for all positions WP in the mesh. The method implemented according to the invention thus provides semiconductor wafers with a (-3σ) quantile of the minimum thickness distributed narrowly (within a narrow range) around its target value of 878.6 μm.

[0148] Figure 5 The cumulative frequencies of minimum thickness TMIN and maximum thickness TMAX for all measured slices are shown, specifically the cumulative frequency 37 for the minimum thickness of the slice in the comparative example and the cumulative frequency 39 for the minimum thickness of the slice in the first example. Similarly, the cumulative frequency 38 for the maximum thickness of the slice in the comparative example and the cumulative frequency 40 for the maximum thickness of the slice in the first example are plotted, as well as the cumulative frequency 41 for the minimum thickness TMIN and the cumulative frequency 42 for the maximum thickness of the slice in the second example.

[0149] Figure 6The set of points 43 of the (+3σ) quantile of the maximum thickness TMAX of the slice in the comparative example and the associated regression curve 44 are shown, as well as the set of points 45 of the (+3σ) quantile of the maximum thickness TMAX of the slice in the first example and the associated regression curve 46.

[0150] The results of the second example show that, for example, the groove pitch can also be optimized towards the maximum thickness of the slice, making it as uniform as possible. Figure 7 The set of points 61 and the associated regression curve 62 for the (+3σ) quantile of the maximum thickness of the slice under the condition that, during the slicing operation, the quantile is related to the maximum thickness of the slice. Figure 4 The groove pitch 36 shown is similar, and the groove pitch varies with the position WP(in the wire mesh). This graph corresponds to the graph of the nonlinear pitch function dTAR(WP), where, during the solution of the equation, the (+3σ) quantile of the maximum thickness is estimated as the thickness eigenvalue of the thickness eigenvalue function TINI(WP), and the constant (+3σ) quantile of the maximum thickness is estimated for the target thickness eigenvalue function. For comparison, Figure 7 Again, a set of points 43 for the (+3σ) quantile of the maximum thickness of the slice in the comparative example and the associated regression curve 44 are shown.

[0151] Figure 8 The set of points 47 of the (-3σ) quantile 47 of the minimum thickness of the slice of the second example and the associated regression curve 48 are shown, and again for comparison, the set of points 30 of the (-3σ) quantile of the minimum thickness of the slice of the comparative example and the associated regression curve 31 are shown.

[0152] Table 1 summarizes the results:

[0153]

[0154] Table 1 indicates the variance (VAR) of the fitted polynomial on which the regression curves are based and the difference (DIFF) based on the position (WP) in the mesh, where the thickness distribution of the two examples has been altered relative to the comparison example by the difference (DIFF). In the first example (where the target thickness eigenvalue function specifies a constant (-3σ) quantile of the minimum thickness TMIN of the slice), the variance in the comparison example (linear decrease in groove pitch) has decreased from 5.2 μm to 0 μm. In contrast, the variance of the (+3σ) quantile of the relevant maximum thickness TMAX of the slice has increased from 38.3 μm to 58.6 μm. In the second example (where the target thickness eigenvalue function specifies a constant (+3σ) quantile of the maximum thickness TMAX of the slice), the variance of the (+3σ) quantile of the maximum width TMAX of the slice has decreased from 38.3 μm to 0 μm, while the variance of the (-3σ) quantile of the minimum thickness TMIN has increased from 5.2 μm to 63.5 μm.

[0155] The above description of the illustrative embodiments should be understood as exemplary. This disclosure enables those skilled in the art to understand the invention and its related advantages, and also allows for changes and modifications to the described structures and methods, which are equally apparent to those skilled in the art. All such changes and modifications, and their equivalents, should therefore be covered by the scope of the claims.

Claims

1. A method for simultaneously cutting multiple slices from a workpiece having a workpiece axis using a wire saw during a slicing operation, comprising: The wire of the wire saw is moved relative to the workpiece by the rotation of the wire guide roller of the wire saw about its axis. The wire guide roller has a housing portion with a wire guide groove, and the wire guide roller stretches a wire mesh formed by the wire segments of the wire. The workpiece is fed to the wire mesh perpendicular to the wire segment in the presence of abrasive, the wire segment working completely through the workpiece, characterized by selecting a nonlinear pitch function dTAR(WP) based on a target thickness feature function TTAR(WP), a pitch function dINI(WP), and a thickness feature function TINI(WP), where dTAR(WP) is the pitch of the adjacent groove at position WP in the outer shell of the wire guide roller assigned during the slicing operation, TINI(WP) is the thickness feature value of the slice obtained at position WP by means of the wire saw during multiple previous slicing operations assigned to the slice measured on the slice, dINI(WP) is the pitch of the adjacent groove at position WP in the outer shell of the wire guide roller assigned during previous slicing operations, TTAR(WP) is the target thickness feature value assigned to the slice cut at position WP during the slicing operation, and WP represents the axial position of the adjacent groove about the axis of the wire guide roller.

2. The method as described in claim 1, characterized in that, Perform multiple previous slice operations of number i, and select the pitch dTAR(WP) such that the equation dTAR(WP) = dINI(WP) + TTAR(WP) - TINI(WP) is satisfied.

3. The method as described in claim 1 or 2, characterized in that, The measured thickness characteristic value is the average of the minimum thickness, the average of the maximum thickness, the average of the average thickness, or the thickness at the quantile of the minimum thickness, the quantile of the maximum thickness, or the quantile of the average thickness of the measured slice.

4. In the method of any one of claims 1 to 3, TTAR(WP) is a constant.

5. The method according to any one of claims 1 to 4, wherein the line is a smooth piano line composed of hypereutectoid pearlite, and the abrasive is provided in the form of a slurry composed of silicon carbide (SiC) in a carrier liquid composed of ethylene glycol or oil.

6. The method of claim 4, wherein the line is further provided with a plurality of protrusions and recesses perpendicular to the longitudinal direction of the line.

7. The method of any one of claims 1 to 6, wherein the movement of the line involves a series of consecutive pairs of direction reversals, and in each case a pair of direction reversals includes the line first moving a first length in a first direction along the longitudinal direction of the line and the line subsequently moving a second length in a second direction exactly opposite to the first direction, the first length being selected to be greater than the second length.

8. The method according to any one of claims 1 to 7, wherein the workpiece axis is oriented parallel to the axis of the line guide roller.

9. The method of any one of claims 1 to 8, wherein the groove has a depth following a depth function t(WP), the depth function t(WP) being the depth of the groove at the location WP that is proportional to the envelope of the line.

10. A wire saw for simultaneously cutting multiple slices from a workpiece, comprising a wire guide roller having a stretched wire mesh and an axis and a housing portion, the housing portion being provided with a plurality of grooves for guiding the wire, the pitch between adjacent grooves at a position WP following a nonlinear pitch function dTAR(WP), WP representing the axial position of the adjacent grooves about the axis of the wire guide roller, and the nonlinear pitch function dTAR(WP) being selected such that the equation dTAR(WP) = dINI(WP) + TTAR(WP) - TINI(WP) is satisfied, TTAR(WP) is a target thickness feature function, which assigns a target thickness feature value to the slice to be cut at the position WP, TINI(WP) is a thickness feature function, which assigns a thickness feature value measured on the slice to the slice obtained by means of the wire saw at the position WP during a plurality of previous slicing operations, and dINI(WP) is a pitch function, which is the pitch of the adjacent grooves in the housing portion of the wire guide roller at the position WP assigned during the previous slicing operations.

11. The wire saw as claimed in claim 10, characterized in that, The measured thickness characteristic value is the average of the minimum thickness, the average of the maximum thickness, the average of the average thickness, or the thickness at the quantile of the minimum thickness, the quantile of the maximum thickness, or the quantile of the average thickness of the measured slice.

Images