JP7217409B2 - Crack growth device and crack growth method - Google Patents

Description

TECHNICAL FIELD The present invention relates to a technique for cleaving a wafer in which a modified region is formed by laser light.

In Patent Document 1, a semiconductor substrate placed with its back surface facing upward is irradiated with a laser to form a modified region inside the substrate, an expanding tape is attached to the back surface of the semiconductor substrate, and a knife is cut from above the expanding tape. It describes cutting a semiconductor substrate into chips by applying an edge and splitting the substrate with the modified region as a base point.

Further, in Patent Document 1, a semiconductor substrate placed with its back surface facing upward is irradiated with a laser to form a modified region inside the substrate, then the substrate is ground to be thinned, and expanded onto the back surface of the semiconductor substrate. It is described that the tape is attached and the expanded tape is stretched to split the substrate with the modified region as a base point.

In Patent Document 2, a semiconductor substrate placed with its back surface facing upward is irradiated with a laser beam to form a modified region inside the substrate, thereby generating cracks in the thickness direction of the semiconductor substrate, thereby reducing the back surface of the substrate. It describes cutting a semiconductor substrate into chips by exposing cracks on the back surface by grinding and chemical etching. Patent Document 2 discloses that cracks occur in the thickness direction from the modified region by generating thermal stress by applying a natural or relatively small force, such as an artificial force or a temperature difference, to the substrate. described to occur.

Japanese Patent No. 3624909 Japanese Patent No. 3762409

In the invention described in Patent Document 1, the substrate is cracked by applying an external force locally with a knife edge, and bending stress and shear stress are applied to the substrate in order to apply the external force locally. However, it is difficult to uniformly distribute bending stress and shear stress over the entire surface of the substrate. For example, when bending stress or shearing stress is applied to the substrate, the stress is concentrated on some weak point, and the minimum required stress cannot be uniformly applied to the desired portion efficiently.

Therefore, there is a problem that the cracking of the substrate varies, and if the cracking does not progress slowly, the substrate will break even in the chip. In addition, in the case where stress is applied locally and sequentially to the portion to be cut of the substrate, for example, in the case of collecting a large number of chips from a single substrate, there are many cutting lines. , there is a problem that the productivity is very low.

In addition, there is a problem that when the substrate is split by applying an external force, a very large stress is required to split the wafer if the substrate is not processed to be thin.

In particular, when the thickness of the substrate is sufficiently thick relative to the width of the modified depth of laser processing, even if an external force is applied, the substrate may not be cut cleanly and perpendicularly due to the influence of the sudden external force. Therefore, it may be necessary to irradiate several laser pulses in multiple stages in the thickness direction of the substrate.

Further, in the inventions described in Patent Documents 1 and 2, the modified region formed inside the substrate by laser irradiation eventually remains on the chip cross section. Therefore, dust may be generated from the modified region of the chip cross section. In addition, as a result of local fracture of the chip cross-section, the fractured cross-section may trigger the chip to break. As a result, there is a problem that the bending strength of the chip is reduced.

The invention described in Patent Document 2 describes that cracks naturally occur in the thickness direction from the modified region. In order to inevitably obtain the stable effect of breaking, it is sometimes necessary to take arbitrary means, and if it breaks naturally, it does not correspond to arbitrary means.

It is also conceivable to generate thermal stress by applying a temperature difference as a relatively small force to cause cracking in the thickness direction from the modified region. In this case, there is a problem of how to give a uniform thermal gradient in the plane of the substrate. That is, even if a thermal gradient is artificially provided, heat is dispersed in the substrate by heat conduction so as to partially alleviate the thermal gradient. Therefore, there is a very difficult problem in how to continuously form a stable thermal gradient (stable temperature difference) sufficient to cut a given substrate.

In addition, in the invention described in Patent Document 2, after the semiconductor substrate is ground, the back surface is chemically etched. is formed, and a work-affected layer remains. When the surface is chemically etched, portions with large lattice distortion such as microcracks are selectively etched. As a result, microcracks are rather encouraged to become large cracks. For this reason, not only the cutting starting point region but also the microcracks formed by grinding and etching may occasionally cause fracture, which poses a problem that stable cutting is difficult.

In addition, the substrate surface is not mirror-finished because the etching promotes unevenness on the substrate surface. As a result, since unevenness remains even in the divided chips, it is highly possible that the chip will be broken from a portion with large unevenness, that is, a microcrack, and the die strength of the chip will be low.

Furthermore, when an etchant is applied to a portion modified by laser processing, the modified portion melts once, recrystallizes, and solidifies, so that large grain boundaries are formed.

When the etchant acts on such grain boundaries, the etching progresses so that the silicon grains are peeled off from the grain boundaries.

As for the polishing process, specifically, p.12_1 line of the cited document describes that the grinding process and the back surface are subjected to chemical etching. In the case of chemical etching, even if the crack is left as it is, the chemical etchant penetrates into the crack and has the effect of dissolving the crack.

In particular, if the cracks extend to the surface of the substrate, the etchant penetrates between the chip and the film bonding the chips, causing a problem of peeling the chip from the film.

Etching proceeds along the crack even if the crack does not reach the surface of the substrate. In particular, when etching is performed in a thin state after grinding, the etchant permeates into the cracks by capillary action, melting the crack tips and deepening the microcracks further, and furthermore, new etchant enters there. Become. In this case, if the etchant acts on the vicinity of the device surface, the vicinity of the device surface may be dissolved, and etching may proceed to the inside of the element. Moreover, even if there is no problem at that time, the etchant left on the wall surface of the wafer may enter the interior of the device and cause defects later. Therefore, while the processing strain on the wafer surface is removed, the cracks are further promoted, and secondary problems associated therewith may be induced.

Moreover, in such a case, cracks may eventually extend to the surface, and the etchant may partially enter between the chip and the film, causing the chip to peel off during processing. In particular, when the substrate becomes thinner after grinding, cracks tend to progress even with a small amount of external force, and when the crack reaches the surface, chip separation occurs. You have to do something to prevent it from progressing any further.

Patent Document 2 describes that cracks are generated from the modified region by grinding the back surface of the substrate, but Patent Document 2 does not describe a method for fixing the substrate when the substrate is ground. . When the outer periphery of the substrate is supported by fitting the substrate into a retainer or the like as shown in FIG. 20, or when only a part of the substrate is sucked as shown in FIG. In such a case, even if the rear surface of the substrate is ground, cracks do not occur from the modified region because they are not restrained.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a method for cutting a semiconductor substrate that can efficiently obtain chips of stable quality.

In order to achieve the above object, one aspect of a semiconductor substrate cleaving method according to the present invention is a wafer cleaving method for cleaving a wafer having a modified region formed therein by a laser beam, wherein the wafer surface is uniformly placed on a table. and grinding and removing the back surface of the wafer while the wafer is vacuum-sucked, and while allowing the microcracks extending from the modified region to grow in the depth direction of the wafer, leaving the advanced microcracks. a grinding step of grinding and removing the modified region; a mirror-finishing step of performing chemical mechanical polishing on the back surface of the wafer after the grinding step; and a step of cleaving the wafer after the mirror-finishing step. Characterized by

In one aspect of the method for cleaving a semiconductor substrate according to the present invention, the grinding step includes grinding away from the back surface of the wafer to a portion in front of the modified region, and removing microcracks extending from the modified region to the depth of the wafer. and a second grinding step of grinding and removing the modified region formed inside the wafer while leaving the extended microcrack.

Further, in one aspect of the method for cleaving a semiconductor substrate according to the present invention, it is preferable that the step of mirror-finishing is performed by removing the work-affected layer introduced in the grinding step while leaving the microcracks. .

According to the method for cutting a semiconductor substrate of the present invention, cutting can be performed reliably and efficiently, and chips of stable quality can be efficiently obtained.

1 is a diagram showing an overview configuration of a laser dicing apparatus 1; FIG. FIG. 2 is a conceptual configuration diagram showing the configuration of driving means of the laser dicing apparatus 1; FIG. 2 is a plan view showing the configuration of driving means of the laser dicing apparatus 1; FIG. 2 is a diagram showing an overview configuration of a grinding device 2; FIG. 2 is a partially enlarged view of the grinding device 2; FIG. 2 is a partially enlarged view of the grinding device 2; 4 is a schematic diagram of a polishing stage of the grinding device 2; FIG. It is a figure which shows the detail of the chuck|zipper of the grinding device 2, (a) is a top view, (b) is sectional drawing, (c) is the elements on larger scale. FIG. 2 is a side view of the tape peeling device 3; The side sectional view which showed the mode of cutting. FIG. 4 is a side cross-sectional view showing a state of a pressed elastic body; The perspective view which showed the mode of cutting. The perspective view which showed the sticking method of a tape. FIG. 2 is a flowchart showing a method for cutting a semiconductor substrate according to the present invention; The figure explaining a grinding removal process. It is a diagram for explaining crack growth in the grinding removal process, (a) is a schematic diagram during grinding, (b) is the state of the back surface of the wafer W, (c) is the state of the front surface of the wafer W, and (d) is the state of the wafer W. Cross section. FIG. 4 is a view for explaining the surface state of the back surface of the wafer W after the grinding removal process; The figure shown about the conditions of crack growth evaluation. The figure which showed the evaluation result of crack progress evaluation. The figure which shows the fixed state at the time of the conventional board|substrate grinding. The figure which shows the fixed state at the time of the conventional board|substrate grinding.

BEST MODE FOR CARRYING OUT THE INVENTION Preferred embodiments of the method for cutting a semiconductor substrate according to the present invention will be described below in detail with reference to the accompanying drawings.

The present invention comprises a laser dicing device 1, a grinding device 2, a tape peeling device 3, and a transport device (not shown) for transporting a wafer processed by the laser dicing device 1 to the grinding device 2 or the tape peeling device 3. , and an expanding device for separating the wafer split by the tape peeling device 3 .

<About device configuration>
(1) Laser dicing machine 1 FIG. As shown in the figure, the laser dicing apparatus 1 of this embodiment mainly includes a wafer moving section 11, a laser head 40 comprising a laser optical section 20 and an observation optical section 30, a control section 50, and the like. .

The wafer moving unit 11 includes a suction stage 13 that holds the wafer W by suction, and an XYZθ table 12 that is provided on the body base 16 of the laser dicing apparatus 1 and precisely moves the suction stage 13 in the XYZθ directions. The wafer moving unit 11 precisely moves the wafer W in the XYZθ directions in the drawing.

A back grind tape (hereinafter referred to as BG tape) B having an adhesive material is applied to the surface of the wafer W formed with devices by a tape applying device (not shown), and the wafer W is placed on the suction stage 13 so that the back surface faces upward. be.

A dicing sheet having an adhesive material may be attached to one surface of the wafer W, and the wafer W may be placed on the suction stage 13 in a state integrated with the frame via the dicing sheet. In this case, it is placed on the suction stage 13 so that the surface faces upward.

The laser optical unit 20 is composed of a laser oscillator 21, a collimating lens 22, a half mirror 23, a condensing lens (collecting lens) 24, driving means 25 for finely moving the laser light parallel to the wafer W, and the like. A laser beam oscillated from a laser oscillator 21 is condensed inside the wafer W through an optical system such as a collimating lens 22, a half mirror 23, a condensing lens 24, and the like. The position of the focal point in the Z direction is adjusted by finely moving the condenser lens 24 in the Z direction by means of Z fine movement means 27, which will be described later.

The conditions of the laser beam are as follows: the light source is a semiconductor laser pumped Nd:YAG laser, the wavelength is 1064 nm, the laser beam spot cross-sectional area is 3.14×10 −8 cm 2 , the oscillation mode is a Q-switch pulse, the repetition frequency is 100 kHz, and the pulse width. is 30 ns, the output is 20 μJ/pulse, the laser beam quality is TEM00, and the polarization characteristic is linear polarization. Further, the conditions for the condensed lens 24 are a magnification of 50 times, N.O. A. is 0.55, and the transmittance for the laser light wavelength is 60%.

The observation optical unit 30 includes an observation light source 31, a collimating lens 32, a half mirror 33, a condensing lens 34, a CCD camera 35 as observation means, an image processing unit 38, a television monitor 36, and the like.

In the observation optical unit 30 , the illumination light emitted from the observation light source 31 irradiates the surface of the wafer W through an optical system such as a collimate lens 32 , a half mirror 33 and a condensed lens 24 . Reflected light from the surface of the wafer W passes through the condensing lens 24, the half mirrors 23 and 33, and the condensing lens 34 and enters a CCD camera 35 as an observation means, and the surface image of the wafer W is captured.

This imaging data is input to the image processing unit 38 and used for alignment of the wafer W, and displayed on the television monitor 36 via the control unit 50 .

The control unit 50 includes a CPU, a memory, an input/output circuit unit, etc., and controls the operation of each unit of the laser dicing apparatus 1 .

The laser dicing apparatus 1 is composed of a wafer cassette elevator, wafer transfer means, an operation panel, indicator lamps, and the like (not shown).

The wafer cassette elevator vertically moves the cassette in which the wafers are stored and positions it at the transfer position. The transport means transports wafers between the cassette and the suction stage 13 .

Switches and a display device for operating each part of the dicing apparatus 10 are attached to the operation plate. The indicator lamps indicate the operating status of the dicing apparatus 10, such as during processing, end of processing, emergency stop, and the like.

FIG. 2 is a conceptual diagram illustrating the details of the driving means 25. As shown in FIG. The drive means 25 includes a lens frame 26 that holds the condensed lens 24, a Z fine movement means 27 that is attached to the upper surface of the lens frame 26 and finely moves the lens frame 26 in the Z direction of the drawing, and a holding frame 28 that holds the Z fine movement means 27. , PZ1, PZ2, etc., which are linear fine movement means for finely moving the holding frame 28 in parallel with the wafer W. As shown in FIG.

A piezoelectric element that expands and contracts when a voltage is applied is used for the Z-fine movement means 27 . The expansion and contraction of the piezoelectric element slightly moves the condenser lens 24 in the Z direction so that the focal point of the laser beam is precisely positioned in the Z direction.

The holding frame 28 is supported by two pairs of parallel springs composed of four piano wires (not shown), is movable in the XY directions, and is restricted in movement in the Z direction. The method of supporting the holding frame 28 is not limited to this. For example, the holding frame 28 may be vertically sandwiched between a plurality of balls to restrict movement in the Z direction and be supported so as to be movable in the XY directions.

The linear fine movement means PZ1 and PZ2 use a piezoelectric element like the Z fine movement means 27, one end of which is fixed to the case body of the laser head 40, and the other end of which is in contact with the side surface of the holding frame 28.

3 is a plan view of the driving means 25. FIG. As shown in FIG. 3, two linear fine movement means PZ1 and PZ2 are arranged in the X direction, one end of each of which is fixed to the case body of the laser head 40, and the other end is in contact with the side surface of the holding frame 28. . Therefore, by controlling the applied voltage, the condensing lens 24 can be reciprocally finely moved in the X direction, and the laser beam can be reciprocally finely moved or oscillated in the X direction.

A piezoelectric element may be used for one of the linear fine movement means PZ1 and PZ2, and an elastic member such as a spring may be used for the other. Also, three linear fine movement means may be arranged on the circumference.

A laser beam L is emitted from a laser oscillator 21, and the inside of the wafer W is irradiated with the laser beam L via an optical system such as a collimating lens 22, a half mirror 23, a condensing lens 24, and the like. The Z-direction position of the focal point of the laser beam L to be irradiated is accurately set at a predetermined position inside the wafer by the Z-direction position adjustment of the wafer W by the XYZθ table 12 and the position control of the condensed lens 24 by the Z fine movement means 27. set.

In this state, the XYZθ table 12 is processed and fed in the X direction, which is the dicing direction. It is vibrated in parallel in the X direction or arbitrary XY directions, and the focal point of the laser beam L forms a modified region K while vibrating minutely inside the wafer. As a result, along the cutting line of the wafer W, one line of modified region K is formed inside the wafer W by multiphoton absorption.

Vibration in the Z direction may be applied by the Z fine movement means 27 as required. In addition, by feeding the wafer W in the X direction while slowly reciprocating the laser light L in the X direction, which is the processing direction, the laser light L is repeatedly radiated back and forth like perforations. good too.

When one line of the modified region is formed along the cutting line, the XYZθ table 12 is indexed and fed by one pitch in the Y direction, and the next line is similarly formed with the modified region.

When the modified regions are formed along all the cutting lines parallel to the X direction, the XYZθ table 12 is rotated by 90°, and all the lines orthogonal to the previous lines are similarly formed.

(2) Grinding Device 2 FIG. 4 is a perspective view showing the general configuration of the grinding device 2 . The main body 112 of the grinding device 2 is provided with an alignment stage 116 , a rough grinding stage 118 , a fine grinding stage 120 , a polishing stage 122 , a polishing cloth cleaning stage 123 , a polishing cloth dressing stage 127 and a wafer cleaning stage 124 .

Rough grinding stage 118, fine grinding stage 120, and polishing stage 122 are partitioned by a partition plate 125 (not shown in FIG. 4) as shown in FIG. It is prevented from scattering on the stage.

As shown in FIG. 5, the partition plate 125 is fixed to the index table 134 and formed in a cross shape so as to partition the four chucks 132, 136, 138, 140 installed on the index table 134. .

The rough grinding stage 118 is a stage that performs rough grinding, and as shown in FIG. The fine grinding stage 120 is a stage that performs fine grinding, and is surrounded by the side surface of the main body 112 , the top plate 129 and the partition plate 125 like the rough grinding stage 118 . Brushes (not shown) are provided on the top and side surfaces of the partition plate 125 to isolate the rough grinding stage 118 and fine grinding stage 120 from the outside. Through holes 128A and 129A through which the heads of the respective stages are inserted are formed in the top plates 128 and 129, respectively.

The polishing stage 122 performs chemical mechanical polishing, and is covered by a casing 126 having a top plate 126A, as shown in FIG. 5, in order to isolate it from the other stages. The top plate 126A is formed with a through hole 126C through which the head of each stage is inserted.

As shown in FIG. 6, a brush 126B is attached to the side surface of the casing 126 through which the partition plate 125 passes. The side 125B is contacted. As a result, when the chuck 140 is positioned at the processing position, it is held in a substantially airtight state by the casing 126, the partition plate 125, and the brush 126B.

Since the polishing stage 122 performs chemical mechanical polishing, the polishing liquid contains a chemical polishing agent. If the grinding liquid is mixed with such a polishing liquid, the concentration of the chemical polishing agent is lowered, resulting in an increase in processing time. By keeping the polishing stage 122 in a substantially airtight state, it is possible to prevent the grinding fluid and processing debris used in the fine grinding stage 120 from entering the polishing stage 122, and also prevent the grinding fluid used in the polishing stage 122 from entering the polishing stage 122. can be prevented from scattering from the polishing stage 122 . Therefore, it is possible to prevent machining problems caused by mixing of both machining liquids.

FIG. 7 is a structural diagram of the polishing stage 122. As shown in FIG. In the polishing stage 122, the wafer W is polished with a polishing cloth 156 and slurry supplied from the polishing cloth 156, and a process-affected layer formed on the rear surface of the wafer W is removed by rough polishing and fine polishing. The work-affected layer is a general term for streaks and work strain (crystal is altered) caused by grinding.

A polishing cloth 156 of the polishing stage 122 is attached to a polishing head 161 connected to an output shaft 160 of a motor 158 . Guide blocks 162, 162 constituting linear motion guides are provided on the side surface of the motor 158, and the guide blocks 162, 162 are engaged with guide rails 166 provided on the side surface of the support plate 164 so as to be vertically movable. It is Therefore, the polishing cloth 156 is attached to the support plate 164 together with the motor 158 so as to be vertically movable.

A support plate 164 is provided at the tip of an arm 168 arranged horizontally. A base end of the arm 168 is connected to an output shaft 174 of a motor 172 arranged within the casing 170 . Therefore, when the motor 172 is driven, the arm 168 can rotate about the output shaft 174 . As a result, the range of the polishing position of the polishing cloth 56 (see the solid line in FIG. 3), the polishing cloth cleaning position by the polishing cloth cleaning stage 123 (see the two-dot chain line in FIG. 3), and the dressing position by the polishing cloth dressing stage 127 is determined. can be moved inside. When the polishing cloth 156 is moved to the polishing cloth washing position, the surface of the polishing cloth 156 is washed by the polishing cloth washing stage 123 to remove polishing dust and the like adhering to the surface. Examples of the polishing cloth 156 include foamed polyurethane, polishing cloth, etc. The polishing cloth cleaning stage 123 is provided with a removing member such as a brush for removing polishing debris. This removal member is rotationally driven when cleaning the polishing cloth 156 , and the polishing cloth 156 is similarly rotationally driven by the motor 158 . The polishing cloth dressing stage 127 employs the same material as the polishing cloth 156, such as polyurethane foam.

Guide blocks 176, 176 constituting linear motion guides are provided on the side surface of the casing 170. These guide blocks 176, 176 are vertically movable on a guide rail 180 provided on the side surface of a housing 178 for the screw feeder. is engaged in Also, a nut member 179 protrudes from the side surface of the casing 170 . The nut member 179 is screwed through an opening (not shown) formed in the housing 178 to a threaded rod 181 of a screw feeder provided in the housing 178 . An output shaft 184 of a motor 182 is connected to the upper end of the threaded rod 181 . Therefore, when the motor 82 is driven and the threaded rod 181 is rotated, the casing 70 is moved up and down by the feeding action of the screw feeder and the rectilinear action of the guide block 176 and the guide rail 180 . As a result, the polishing cloth 156 is largely moved in the vertical direction, and the gap between the polishing head 161 and the wafer W is set to a predetermined gap.

A piston 188 of an air cylinder device 186 is connected to the upper surface of the motor 158 through a through hole 169 of the arm 168 . A regulator 190 for controlling the internal pressure P of the cylinder is connected to the air cylinder device 186 . Therefore, when the internal pressure P is controlled by the regulator 190, the pressing force (pressure contact force) of the polishing pad 156 against the wafer W can be controlled.

In the present embodiment, the polishing cloth 156 is used as the polishing body, but the present invention is not limited to this, and as long as the work-affected layer can be removed, for example, a polishing grindstone or electrophoresis of abrasive grains can be used. may apply. When a polishing grindstone, electrophoresis of abrasive grains, or the like is applied, it is preferable to perform quantitative polishing.

Returning to the description of FIG. The alignment stage 116 is a stage that aligns the wafer W transferred from the laser dicing apparatus 1 by a transfer device (not shown) to a predetermined position. The wafer W aligned on the alignment stage 116 is sucked and held by a transfer robot (not shown), then conveyed toward an empty chuck 132 and held by the suction surface of the chuck 132 .

The chuck 132 is installed on the index table 134, and chucks 136, 138, and 140 having the same function are installed on the circumference of the index table 134 around the rotating shaft 135 at intervals of 90 degrees. . A spindle (not shown) of a motor (not shown) is connected to the rotating shaft 135 .

The chuck 136 is positioned on the rough grinding stage 118 in FIG. 4, and the sucked wafer W is rough ground here. The chuck 138 is positioned on the fine grinding stage 120 in FIG. 4, and the chucked wafer W is finish ground (fine grinding, spark-out) here. The chuck 140 is positioned on the polishing stage 122 in FIG. 4, where the chucked wafer W is polished to remove the work-affected layer and variations in the thickness of the wafer W caused by grinding.

The chucks 132, 136, 138, 140 will now be described. Since the chucks 136, 138, and 140 have the same configuration as the chuck 132, the chuck 132 will be described, and the description of the chucks 136, 138, and 140 will be omitted.

8A and 8B are diagrams showing the details of the chuck 132, in which (a) is a plan view of the chuck 132, (b) is a cross-sectional view along line AA' in (a), and (c) is an enlarged view of part B in (b). It is a diagram.

The chuck 132 is configured by fitting a mounting table 132a made of a porous material (for example, porous ceramics) into a chuck body 132b made of a dense body. A suction hole 132c for vacuum suction is formed on the lower side of the chuck body 132b into which the mounting table 132a is fitted. The chuck 132 is desirably made of a material with low thermal conductivity.

The wafer W is mounted on the mounting table 132a with the BG tape B interposed therebetween, as shown in FIG. 8(c). As shown in FIG. 8C, the mounting table 132a is formed such that when the wafer W is mounted on the mounting table 132a, part of the outer periphery of the wafer W protrudes from the mounting table 132a. is about 1.5 mm. The wafer W used in this embodiment has a diameter of approximately 12 inches and a thickness t of approximately 775 μm.

As shown in FIGS. 8A and 8B, the suction holes 132c are arranged so as to cover substantially the entire mounting table 132a. A fluid coupling (not shown) is connected to the suction hole 132c, and a suction pump (not shown) connected to this fluid coupling sucks air. Therefore, substantially the entire surface of the wafer W is firmly vacuum-sucked to the surface of the mounting table 132a. As a result, the wafer W and the mounting table 132a can be brought into close contact with each other without any positional deviation.

As shown in FIG. 7, the chucks 132, 136, 138 and 140 are rotated by the driving force of a spindle 194 and a motor 192 respectively connected to the lower surface thereof. The motor 192 is supported by the index table 134 via a support member 193 . Thus, each time the chucks 132, 136, 138, 140 are moved by the motor 137, the spindle 194 is separated from the chucks 132, 136, 138, 140, or the chucks 132, 136 are moved to the spindle 194 installed at the next movement position. , 138 and 140 can be saved.

A piston 119 of the cylinder device 117 is connected to the lower portion of the motor 192 . When the piston 119 is extended, it is inserted into recesses (not shown) formed in the lower portions of the chucks 132, 136, 138, 140 and connected. Then, the chucks 132 , 136 , 138 , 140 are lifted from the index table 134 by the continued extension of the piston 119 and positioned at grinding positions by the grindstones 146 , 154 .

The control unit 100 includes a CPU, a memory, an input/output circuit unit, etc., and controls the operation of each unit of the grinding apparatus 2 .

The thickness of the wafer W sucked and held by the chuck 132 is measured by a pair of measurement gauges (not shown) connected to the controller 100 . Each of these measurement gauges has a contactor, one of which is in contact with the upper surface (back surface) of the wafer W, and the other contactor is in contact with the upper surface of the chuck 132 . These measurement gauges can detect the thickness of the wafer W as a difference between the in-process gauge readings with the upper surface of the chuck 132 as a reference point. In addition, the thickness measurement by the measuring gauge may be performed in-line. Also, the method of measuring the thickness of the wafer W is not limited to this.

The controller 100 rotates the index table 134 by 90 degrees in the direction of arrow R in FIG. The back surface of W is rough ground. As shown in FIG. 4, the cup-shaped grindstone 146 is connected to an output shaft (not shown) of a motor 148 and attached to a grindstone feeder 152 through a support casing 150 of the motor 148 . The grindstone feeder 152 vertically moves the cup-shaped grindstone 146 together with the motor 148, and the cup-shaped grindstone 146 is pressed against the back surface of the wafer W by this downward movement. Thereby, the back surface of the wafer W is roughly ground. The control unit 100 sets the amount of downward movement of the cup-shaped grindstone 146 and controls the motor 148 . The amount of downward movement of the cup-shaped grindstone 46, that is, the amount of grinding by the cup-shaped grindstone 146 is set based on the pre-registered reference position of the cup-shaped grindstone 146 and the thickness of the wafer W detected by the measurement gauge. be done. The control unit 100 also controls the rotation speed of the cup-shaped grindstone 146 by controlling the rotation speed of the motor 148 .

After the cup-shaped grindstone 146 is retracted from the wafer W, the thickness of the wafer W whose back surface has been roughly ground on the rough grinding stage 118 is measured by a measurement gauge (not shown) connected to the controller 100 . The index table 134 is rotated 90 degrees in the direction of arrow R in FIG. Grinded and sparked out. Since the structure of the fine grinding stage 120 is the same as that of the rough grinding stage 118, the description thereof is omitted here. The amount of grinding by the cup-shaped grindstone 154 is set by the controller 100 , and the processing movement amount and the number of revolutions of the cup-shaped grindstone 154 are controlled by the controller 100 .

After the cup-shaped grindstone 154 is retracted from the wafer W, the thickness of the wafer W whose back surface has been finely ground by the fine grinding stage 120 is measured by a measurement gauge (not shown) connected to the control unit 100 . When the index table 134 is rotated 90 degrees in the direction of arrow R in FIG. Then, the back surface of the wafer W is mirror-finished. The vertical movement distance of the abrasive cloth 156 is set by the controller 100, and the position of the abrasive cloth 156 is controlled by controlling the motor 182 by the controller 100. FIG. Further, the control unit 100 controls the rotation speed of the motor 158 , that is, the rotation speed of the polishing pad 156 .

After the wafer W is polished by the polishing stage 122, the arm 168 is rotated by the control unit 100, and the polishing cloth 156 is retracted from the position above the wafer W. , and transferred to the wafer cleaning stage 124 . A stage having a rinse cleaning function and a spin drying function is applied as the wafer cleaning stage 124 . The polished wafer W is free of the work-affected layer, so that it is not easily damaged during transportation by the robot and during cleaning on the wafer cleaning stage 124 .

The wafer W that has been cleaned and dried on the wafer cleaning stage 124 is sucked and held by a hand (not shown) of a robot (not shown) and stored on a predetermined shelf of a cassette (not shown).

(3) Tape peeling device 3 FIG. 9 is a side view showing the structure of the tape peeling device 3. As shown in FIG. The BG tape B or expanded tape E is adhered to the wafer W by the tape peeling device 3 . Further, the wafer W ground and polished by the grinding device 2 is cleaved by the tape separating device 3 .

The tape peeling device 3 includes a table 211 rotatable by a driving device (not shown). An elastic body 212 on which the wafer W is placed is attached to the upper surface of the table 211 .

A supply reel 213 and a peeling roller 214 are provided above the table 211 . , and is wound up on a take-up reel 218 .

A pressing member 219 for breaking the wafer W is provided in the vicinity of the peeling roller 214 (for example, above the outer peripheral portion of the table 211).

The elastic body 212 attached to the upper surface of the table 211 is an elastic body having a small compression set and no voids, and is arranged in a grid pattern or at intervals equal to or less than the size of the chips C formed on the wafer W as shown in FIG. Grooves 212a, 212a, . . . are formed in parallel. The elastic body 212 is connected to a vacuum source (not shown), has a BG tape B adhered to its front surface by means of grooves 212a, 212a . It is formed so as to be able to suck and fix the wafer W that has been formed.

As a material for the elastic body 212, for example, SBR (styrene butadiene rubber), NBR (nitrile butadiene rubber), or the like can be suitably used. Note that the groove 212a may have a hole shape or an uneven shape.

For example, when a sponge-like elastic body 212B in which a large number of small voids are formed as shown in FIG. Distortion occurs. Therefore, if the elastic body 212B is repeatedly used, the elastic modulus gradually changes due to deformation, and stable cutting cannot be performed.

Further, as shown in FIG. 11(b), in an elastic body 212C having no grooves, holes, unevenness, etc., when a part of the elastic body is pushed, the volume around it is constant, so the pushed-in surrounding portion rises. It will happen. In this case, even if an attempt is made to split the wafer by pressing locally to sink the wafer, the stress is dispersed and efficient splitting becomes difficult.

On the other hand, as shown in FIG. 11(c), the elastic body 212 has room for deformation in the lateral direction due to the grooves 212a against pressure from above. Stress concentrates on S, and cutting can be performed reliably and efficiently.

As shown in FIG. 10, the pressing member 219 has a radius R smaller than the interval L between the cutting lines S on which the modified layer P is formed. Since the radius R is smaller than the interval L, when the pressing member 219 presses one cutting line S, it does not press another cutting line S. As a result, the stress is concentrated on the cutting line S where the cutting is desired, and the cutting can be reliably and efficiently performed.

As shown in FIG. 12, the wafer W is cleaved by attaching a BG tape B to the front surface of the wafer W on which the modified layer P is formed along the cutting line S and an expanding tape E to the back surface. The tape peeling device 3 adjusts the position of the wafer W by rotating the table 211 so that the pressing member 219 is parallel to one of the cutting lines S formed in a grid pattern as shown in FIG.

In this state, the pressing member 219 is pressed against the wafer W and rolled to cut the wafer W along one of the cutting lines S. As shown in FIG. After cutting one cutting line S, the table 211 on which the wafer W is placed is rotated 90 degrees by a drive means (not shown) so that the other cutting line S and the pressing member 219 are arranged in parallel. In this state, the pressing member 219 is pressed against the wafer W and rolled to cut the wafer W along the other cutting line S. As shown in FIG.

Since the BG tape B and the expanding tape E are attached to the front and back surfaces of the wafer W, after cutting one cutting line S, when cutting the next cutting line S, the BG tape B and the expanding tape E are applied. The position is restrained, and the gap between chips on the cutting line S that has already been cleaved is prevented from widening. As a result, not only the vertical deformation of the elastic body 212 but also the horizontal deformation due to the widening of the gap between the chips C can be prevented from dispersing the local stress of the pressing member 219, thereby concentrating the stress. The cutting can be performed reliably and efficiently.

In addition, since the BG tape B and the expand tape E are attached to the front and back surfaces of the wafer W, fragments generated when the wafer W is cut do not adhere to the elastic body 212, and the next wafer W can be attached to the elastic body 212. 212 is not adversely affected by debris.

The peeling roller 214 is made of an elastic material that is softer than the elastic material 212 . The BG tape B has, for example, an ultraviolet curable adhesive on the adhesive surface, and is irradiated with ultraviolet rays by an ultraviolet irradiation device (not shown) provided inside or outside the tape peeling device 3 to reduce the adhesive force. . As shown in FIG. 9, a peeling roller 214 is pressed downward and rolled laterally (in the direction of arrow A in FIG. 9) to roll the peeling film 215 onto the BG tape B whose adhesive strength has decreased. affixed.

After the peeling film 215 is adhered, the take-up unit 218A including the take-up reel 218 and the guide roller 217 rolls the peeling film 215 along with the peeling roller 214 in the horizontal direction (arrow B direction shown in FIG. 9). ), the BG tape B is separated from the wafer W. As shown in FIG.

When the peeling roller 214 adheres the peeling film 215 or peels off the BG tape B, the cutting line S for dividing the chip C is aligned with the peeling roller 214 as shown in FIG. By rotating the table 211 by 45 degrees, the wafers W arranged in parallel are arranged so that the cutting line S obliquely intersects the peeling roller 214 as shown in FIG. 13(b).

As a result, the positions of the cut chips C do not shift when the peeling film 215 is attached or when the BG tape B is peeled off, and there is no adverse effect on subsequent processes such as picking up the chips C after expansion. .

(4) Expanding Device Next, an expanding device (not shown) will be described. A conventional normal expanding device can be used as the expanding device. For example, an expanding device disclosed in Japanese Unexamined Patent Application Publication No. 2007-173587 and configured as follows can be used.

That is, the peripheral portion of the expanding tape E is fixed to a frame-like frame F. As shown in FIG. A ring member is in contact with the lower surface of the inner portion of the peripheral portion of the expanded tape E. As shown in FIG. The outer periphery of the upper surface of this ring member is smoothly rounded. A downward force is applied to the frame F to push it downward. As a result, the dicing tape is expanded to increase the distance between chips. At this time, since the outer periphery of the upper surface of the ring member is smoothly R-chamfered, the expanding tape E is smoothly expanded.

As a mechanism for pushing down the frame F, various known linear motion devices can be employed. For example, a cylinder member (using hydraulic pressure, pneumatic pressure, etc.), a linear motion device comprising a motor and a screw (a combination of a male screw as a shaft and a female screw as a bearing) can be employed.

<Method for cutting semiconductor substrate>
Next, a method for cutting the semiconductor substrate will be described. FIG. 14 is a flow chart showing the processing flow of the semiconductor substrate cutting method.

(1) BG tape attaching step (step S1)
In the method of cutting a semiconductor substrate according to the present invention, first, a BG tape B is adhered to the front side of the wafer W on which circuit patterns and the like are formed (step S1).

The sticking of the BG tape B to the wafer W is performed by a tape sticking device (not shown). When the BG tape B is applied, the table 211 is rotated by 45 degrees so that the cutting line S dividing the chip C obliquely intersects the application roller for applying the tape. This prevents the wafer W from breaking along the cutting line S due to the pressing force of the bonding roller while the BG tape B is being bonded to the wafer W.

(2) Laser modification step (step S2)
A wafer W having a BG tape B attached to its front surface is placed on the suction stage 13 of the laser dicing apparatus 1 so that the rear surface faces upward. The following processing is performed by the laser dicing apparatus 1 and controlled by the controller 50 .

When the laser beam L is emitted from the laser oscillator 21, the laser beam L is irradiated inside the wafer W via an optical system such as a collimating lens 22, a half mirror 23, a condensing lens 24, etc., and is reflected inside the wafer W. A quality region K is formed.

In this embodiment, since the thickness of the finally produced chip is approximately 50 μm, the laser beam is irradiated to a depth of approximately 60 μm to approximately 80 μm from the surface of the wafer W. FIG. In order to efficiently break the front surface (device surface) of the wafer W, a laser-modified region is formed at a relatively deep position close to the reference surface, which is a surface located on the back surface side by the thickness of the chip C from the front surface of the wafer W. Because it is necessary to form.

The controller 50 scans the surface of the wafer W with a pulsed processing laser beam L in parallel to form a plurality of discontinuous modified regions K inside the wafer W side by side. Inside the modified region K, minute holes (hereinafter referred to as cracks) are formed. Hereinafter, a region formed by arranging a plurality of discontinuous modified regions K, . . . is called a modified layer.

When the modified layer is formed along all of the cutting lines S shown in FIG. 10, the process of step S2 ends.

(3) Grinding removal step (step S3)
After the modified region K is formed along the cutting line S by the laser modification step (step S2), the wafer W is transferred from the laser dicing device 1 to the grinding device 2 by a transfer device (not shown). The following processing is performed by the grinding device 2 and controlled by the control section 100 .

The transported wafer W is placed on a chuck 132 (eg, chucks 136, 148, and 140 are also acceptable) with the rear surface facing upward, that is, with the BG tape B attached to the front surface of the wafer W facing downward. The entire surface is vacuum-sucked to the chuck 132 .

The index table 134 is rotated around the rotary shaft 135 to load the chuck 132 onto the rough grinding stage 118, and the wafer W is roughly ground.

Rough polishing is performed by rotating the chuck 132 and the cup-shaped grindstone 146 . In this embodiment, Vitrified #325 manufactured by Tokyo Seimitsu Co., Ltd. is used as the cup-shaped grindstone 146, and the rotation speed of the cup-shaped grindstone 146 is approximately 3000 rpm.

After rough polishing, the index table 134 is rotated about the rotary shaft 135 to carry the chuck 132 onto the fine grinding stage 120, and the wafer W is finely ground by rotating the chuck 132 and the cup-shaped grindstone 154. In the present embodiment, coarse resin #2000 manufactured by Tokyo Seimitsu Co., Ltd. is used as the cup-shaped grindstone 154, and the rotational speed of the cup-shaped grindstone 154 is approximately 2400 rpm.

In this embodiment, as shown in FIG. 15, the wafer W is ground to a target surface, that is, to a depth of approximately 50 μm from the surface of the wafer W by combining rough polishing and fine polishing. In this embodiment, roughly 700 μm of grinding is performed in rough polishing and approximately 30 to 40 μm of approximately 30 to 40 μm in fine polishing. You may determine the amount of grinding so that it may become.

Therefore, as shown in FIG. 15, the modified layer is removed in the grinding process, and the modified region K by the laser beam does not remain in the cross section of the chip C which is the final product. Therefore, it is possible to prevent the modified layer from being crushed from the cross section of the chip, crack the chip C from the crushed portion, or generate dust from the crushed portion.

Further, in the present embodiment, the grinding removal step includes a crack propagation step for propagating cracks in the modified layer in the wafer W thickness direction. 16A and 16B are diagrams for explaining the mechanism by which cracks propagate, in which (a) is a schematic diagram during grinding, (b) is the back surface of the wafer W, (c) is the front surface of the wafer W, and (d) is 4 is a cross-sectional view of the wafer W during grinding; FIG.

Due to the grinding, the ground surface shown in FIG. 16(a), that is, the back surface of the wafer W expands due to grinding heat as shown in FIG. 16(b). On the other hand, the surface on the opposite side of the ground surface, that is, the surface of the wafer W, is vacuum chucked over substantially the entire surface by a vacuum chuck as shown in FIG. 16(c). is physically constrained against displacement to

That is, as shown in FIG. 16(d), the back surface (ground surface) of the wafer W tends to expand in the outer peripheral direction due to thermal expansion (displacement due to thermal expansion) in the case of a disk-like shape, whereas the front surface of the wafer W ( The attraction surface) is restrained so that each point on the wafer surface to be spread is not physically displaced. As a result, strain occurs inside the wafer, and cracks develop in the thickness direction of the wafer W due to this internal strain. This internal strain acts evenly between the portion that expands due to thermal expansion and each point on the wafer surface that is physically restrained. Crack growth due to internal strain is most efficient at the initial stage of grinding, that is, during rough polishing, when the amount of grinding is greatest and frictional force is large, ie, frictional heat can be large.

The laser-modified region is formed at a relatively deep position close to the thickness of the chip. Therefore, in the early stages of grinding, the distance from the ground surface to the laser modified layer is relatively long, but the target surface is relatively close to the modified layer to the extent that cracks are propagated. Therefore, in order to promote crack propagation, it is preferable to accelerate the thermal expansion of the wafer W by the heat of grinding at the initial stage of rough polishing.

Even if grinding is performed under conditions that cause thermal expansion of the wafer W, that is, conditions that generate a lot of frictional heat (for example, reducing the amount of grinding fluid), the shear stress of grinding is immediately applied to the modified layer. It's nothing. In the present embodiment, cracks are not propagated by shear stress due to grinding, but thermal expansion due to grinding heat is the dominant factor in crack propagation.

When a crack is propagated by internal strain, the crack can be uniformly propagated at each point in the wafer W surface regardless of the wafer W without being caused by variations in rigidity in the wafer W surface. . Therefore, it is possible to prevent the stress from concentrating on a portion with weak rigidity due to the presence of defects in the surface of the wafer W, as in the case of applying an artificial stress.

Moreover, when an external force is applied artificially, the stress is concentrated on the weak portion of the material, so it is difficult to control the cracks to grow uniformly and gently, and the wafer is completely cleaved. On the other hand, in the case of propagation of cracks due to internal strain in the present embodiment, the cracks can be subtly propagated because the internal strain is caused by the degree of thermal expansion. That is, the crack can grow between the target surface and the wafer W surface. Therefore, it becomes possible to divide efficiently in the wafer dividing step (step S6) described later.

Internal strain due to thermal expansion of the wafer W is distinguished from so-called thermal stress caused by temperature difference. Thermal stress is generated in proportion to the temperature gradient, but in this embodiment the generated heat escapes to the chucks 132, 136, 138 and 140, so no thermal stress is generated.

(4) Chemical mechanical polishing step (step S4)
This process is performed by the grinding device 2 and controlled by the controller 100 .

After the fine polishing, the index table 134 is rotated around the rotary shaft 135 to load the chuck 132 onto the polishing stage 122, and the polishing cloth 156 of the polishing stage 122 performs chemical mechanical polishing. The damaged layer formed on the back surface of the wafer W is removed, and the back surface of the wafer W is mirror-finished.

In this embodiment, a non-woven cloth impregnated with polyurethane (for example, TS200L manufactured by Tokyo Seimitsu Co., Ltd.) is used as the polishing cloth 156, colloidal silica is used as the slurry, and the rotational speed of the polishing cloth 156 is approximately 300 rpm.

Due to the grinding removal step (step S3), the rear surface of the wafer W is formed with a large number of irregularities as shown in FIG. When polishing is performed by chemical etching, since the surface shape is maintained as it is, there is a risk that cracks may occur from the concave portions, and the surface cannot be mirror-finished. On the other hand, in the present embodiment, since chemical mechanical polishing is performed, processing distortion caused by processing is removed, and irregularities on the surface are removed to make the surface mirror-finished.

In other words, in order to improve the quality of the chip C, which is the final product, two processes are indispensable: a grinding removal process using a grindstone and a chemical mechanical polishing process using free abrasive grains containing a chemical liquid using a polishing cloth. .

(5) Expanding tape application step (step S5)
An expanding tape E is adhered to the back surface of the wafer W on which the chemical mechanical polishing step (step S4) has been performed. An expandable tape is a type of elastic tape and is stretchable.

The expansion tape E is adhered to the wafer W by a tape adhering device (not shown) in the same manner as the BG tape B is adhered in step S1. When the expanding tape E is applied, the table 211 is rotated by 45 degrees so that the cutting line S for dividing the chip C, like the BG tape B, intersects the application roller for applying the tape at an angle. Let

As a result, the wafer W is not broken along the cutting line S by the pressing force of the sticking roller while the expand tape E is being stuck to the wafer W.

(6) Wafer cutting step (step S6)
The wafer W to which the expanding tape E is adhered is transported to the tape peeling device 3, and the pressing member 218 is pressed against the wafer W to split the wafer W along the cutting lines S, whereby the wafer W is divided into individual chips C. split.

In cleaving the wafer W, an elastic member 212 is provided on the upper surface of the wafer W so that the pressing member 219 is parallel to one of the cutting lines S, which are cutting lines formed in a grid pattern, as shown in FIG. placed on the table 211 . In this state, the pressing member 219 is pressed against the wafer W and rolled to cut the wafer W along one of the cutting lines S. As shown in FIG. After cleaving one cutting line S, the table 211 on which the wafer W is mounted is rotated 90 degrees by a driving means (not shown) so that the other cutting line S and the pressing member 219 are arranged in parallel. The wafer W is cut along the cutting line S.

The elastic body 212 attached to the upper surface of the table 211 is an elastic body having a small compression set and no voids, and is arranged in a grid pattern or at intervals equal to or less than the size of the chips C formed on the wafer W as shown in FIG. Grooves 212a, 212a, . . . are formed in parallel. As a result, as shown in FIG. 11(c), the elastic body 212 has room for deformation in the horizontal direction due to the grooves 212a against pressure from above. The stress is concentrated on the surface, and the cutting can be performed reliably and efficiently.

As shown in FIG. 10, the pressing member 219 has a radius r smaller than the interval l between the cutting lines S on which the modified regions K are formed. When pressing, other cutting lines S are not pressed. As a result, the stress is concentrated on the cutting line S where the cutting is desired, and the cutting can be reliably and efficiently performed.

When cutting the wafer W, the BG tape B and the expanding tape E are attached to the front and back surfaces. The position is restrained by E, and the gap between chips on the cutting line S that has already been cleaved does not widen greatly. As a result, the local stress of the pressing member 218 is not dispersed due to not only vertical deformation by the elastic body 212 but also horizontal deformation due to the widening of the gap between the chips C, and the stress is concentrated. The cutting can be performed reliably and efficiently.

In addition, since the BG tape B and the expand tape E are attached to the front and back surfaces of the wafer W, fragments generated when the wafer W is cut do not adhere to the elastic body 212, and the next wafer W can be attached to the elastic body 212. 212 is not adversely affected by debris.

(7) BG tape peeling process (step S7)
A peeling film 215 is adhered to the BG tape B of the cut wafer W by a peeling roller 214, and the BG tape B is peeled off.

The peeling roller 214 is made of an elastic material that is softer than the elastic material 212 . The BG tape B has, for example, an ultraviolet curable adhesive on the adhesive surface, and is irradiated with ultraviolet rays by an ultraviolet irradiation device (not shown) provided inside or outside the tape peeling device 3 to reduce the adhesive force. . As shown in FIG. 9, a peeling roller 214 is pressed downward and rolled laterally (in the direction of arrow A in FIG. 9) to roll the peeling film 215 onto the BG tape B whose adhesive strength has decreased. affixed.

After the release film 215 is adhered, the take-up unit 218A including the two take-up reels 18 and the guide roller 217 moves the release film 215 along with the release roller 214 in the horizontal direction (arrow B in FIG. 9) while winding up the release film 215. direction), the BG tape B is peeled off from the wafer W.

When the peeling roller 214 adheres the peeling film 215 or peels off the BG tape B, the cutting line S for dividing the chip C is aligned with the peeling roller 214 as shown in FIG. By rotating the table 211 by 45 degrees, the wafers W arranged in parallel are arranged so that the streets S intersect with the peeling roller 214 at an angle as shown in FIG. 13(b).

As a result, the positions of the cut chips C do not shift when the peeling film 215 is attached or when the BG tape B is peeled off, and there is no adverse effect on subsequent processes such as picking up the chips C after expansion. .

(8) Wafer Separation Step (Step S8)
After peeling off the BG tape B, the wafer W is placed on an expansion table (not shown) with the expansion tape E facing down, and the expansion tape E is removed by pushing down the frame F or by raising the table with the frame F fixed. is stretched, and the interval between chips C is widened.

The wafer W after expansion is picked up from the expansion tape E for each chip C and transported to various subsequent processes.

<<Crack growth evaluation by polishing>>
Next, evaluation of crack growth by polishing in the grinding removal step (step S3) will be described with reference to FIGS. 18 and 19. FIG. The polishing method, the dividing/separating method, their conditions, etc. are basically as described in steps S1 to S8. FIG. 18 is a diagram showing conditions for crack growth evaluation, and FIG. 19 is a diagram showing evaluation results of crack growth evaluation.

In FIGS. 18A, 18B, and 18C, the horizontal axis is common, each position corresponds to each other, and indicates the polishing time (s). The vertical axis of FIG. 18(A) indicates the cutting speed (polishing speed) (μm/s), the vertical axis of (B) indicates ON/OFF of water supply to the grindstone during polishing, and (C). indicates the temperature (° C.) of the back surface of the wafer W during polishing.

As shown in FIG. 18A, while varying the polishing rate, rough polishing was performed to a total thickness of 710 μm, followed by fine polishing of 13 μm (not shown), and chemical mechanical polishing (not shown) of 2 μm. During wafer polishing, as shown in FIG. 18B, water supply to the grindstone or wafer was interrupted during rough polishing. The water supply was stopped after t1 seconds had elapsed after the start of polishing, and resumed after t2 seconds had elapsed after the start of polishing. Water was supplied at a flow rate of 10 L/min. After that, the steps S5, S6, and S7 were performed to cut the wafer, and the state of the cutting was observed and evaluated. The results are summarized in FIG. 19, with ◯ indicating good cracking without chip cracking or generating dust, and x indicating chip cracking or generating dust.

As shown in FIG. 18C, the back surface temperature of the wafer W suddenly starts rising t1 seconds after the start of polishing when water supply is stopped, and falls immediately after t2 seconds after the start of polishing when water supply is resumed. began to

As shown in FIG. 19, when the rear surface temperature of the wafer W reached 70° C. or higher due to polishing, the wafer was cleaved favorably. This is probably because the cracks formed by the laser propagated due to the heat of polishing.

Therefore, the grinding apparatus according to the present invention includes means for measuring the temperature of the wafer, means for turning on and off the water supply to the grindstone or the wafer during grinding, and the wafer temperature reaching a predetermined value after a predetermined time has elapsed from the start of grinding. means for controlling the water supply to the wafer to be turned off until the As a result, the crack formed by the laser can be propagated, and the wafer can be cleaved satisfactorily.

As described above, according to the present embodiment, cracks in the modified region formed by laser light by grinding can be propagated. You can make sure that no space is left over. Therefore, problems such as cracking of the chip C and generation of dust from the cross section of the chip C can be prevented. Therefore, chips with stable quality can be efficiently obtained. In addition, the stress of the pressing member can be concentrated on the cutting line of the wafer, so that the cutting can be reliably and efficiently performed. Further, no contamination is left on the elastic body on which the wafer is placed at the time of cutting, and the wafer is not adversely affected even if the cutting is continuously performed.

(Appendix)
As can be seen from the above detailed description of the embodiments, this specification includes disclosure of various technical ideas including the inventions described below.

(Appendix 1) a step of attaching a back grind tape to the front surface of a wafer; a modified region forming step of forming microvoids in the modified region by forming the modified region; a vacuum suction step for independently and uniformly vacuum-sucking the wafer to the table; in the thickness direction of the wafer; after the grinding step, chemically and mechanically polishing the wafer in which the micropores are developed in the thickness direction of the wafer in the grinding step; a step of attaching an expanding tape to the back surface of the wafer polished to a high degree; and pressing the wafer, to which the expanding tape is attached to the back surface, with a pressing member via the back grind tape to cut the wafer. A splitting step, a peeling step of attaching a peeling tape for peeling the back grind tape, peeling the back grind tape with a peeling roller, and stretching the split wafer by the expanding tape. and a dividing/separating step of dividing the semiconductor substrate into a plurality of chips.

According to the invention described in Supplementary Note 1, the back grind tape for protection is attached to the surface of the wafer by the mechanism for attaching the tape. After the backgrinding tape is adhered, microvoids are formed in the modified region by entering a laser beam from the rear surface of the wafer along the cutting line to form a modified region inside the wafer. Almost the entire surface of the wafer on which the modified regions are formed is independently and evenly attracted to each position of the substrate, and the wafer is ground from the rear surface to remove the modified regions.

At this time, due to the grinding heat generated by the grinding, the wafer surface undergoes thermal expansion in the radial direction and tends to spread. On the other hand, however, the wafer vacuum chuck with a large heat capacity tries to prevent the wafer from spreading due to its expansion.

As a result, while the front surface of the wafer expands due to thermal expansion, the rear surface of the wafer that is chucked does not expand due to the vacuum chuck and tries to maintain the same state. As a result, the modified region formed inside the wafer acts to further expand the modified region according to the difference in expansion between the front surface and the rear surface of the wafer, and the crack progresses further. The modified region has large crystal grains and is fragile because there is also a part that has been melted once and recrystallized by being irradiated with a laser beam. When such a modified region appears on the side surface of a chip in the future, dust may be generated from the side surface of the chip, and large crystal grains may be chipped from the side surface of the chip. However, since the microcracks that develop from the modified region are pure crystal planes, even if these planes appear on the chip side in the future, they may generate dust from the chip side or become large crystal grains and chip. There is no such thing as.

In this way, the wafer is scraped and removed by the grinding process, and the modified region is removed by extending the microvoids in the thickness direction of the wafer. Next, in order to make the process-affected layer formed by grinding and the microvoids that have developed stand out, the entire surface of the wafer is subjected to chemical mechanical polishing (details will be described later) to completely remove the process-affected layer. do. As a result, only microvoids remain on the surface, and the remaining area becomes a perfect mirror surface without machining strain. Thereafter, an expanded film is applied by a tape application mechanism to the back surface of the wafer that has not yet cracked.

The wafer to which the expand film is adhered is turned over and placed on a table provided in a mechanism for peeling off the back grind tape. Subsequently, one of the cutting lines is cleaved by a pressing member provided in the peeling mechanism positioned parallel to the cutting line. After one cutting line is completely broken, the wafer is rotated 90 degrees and the other cutting line is broken.

After breaking, the adhesive strength of the back grind tape is reduced by ultraviolet irradiation, heating, or the like. A peeling roller is used to adhere a peeling tape to the back grind tape whose adhesive force has been reduced, and the tape is peeled off. The wafer from which the backgrinding tape has been peeled off is separated from the chips by expanding the expand film. As a result, the modified region formed by the laser beam can be prevented from remaining on the cross section of the chip. Therefore, it is possible to prevent defects such as cracking of the chip and generation of dust from the cross section of the chip, and to efficiently obtain chips of stable quality.

In addition, since the back surface of the ground wafer is chemically and mechanically polished before the wafer is divided, the die strength of the chip can be increased. Here, the chemical mechanical polishing is largely distinguished from the etching process in the polishing process described in Cited Document 2.

First, the chemical solution in the present application is different from the etching solution in Cited Document 2. The etchant in Cited Document 2 has the effect of naturally dissolving the substrate, that is, etching the substrate by the action of the liquid on the surface of the substrate. As a result, the etchant permeates even the cracked portion and naturally etches the surroundings, which has the effect of further enlarging the modified layer formed by the laser. Further, in the case of the conventional etchant, as described above, the etchant penetrates too much into the cracks, propagating the cracks, and finally penetrates into the portion between the chip and the film. As a result, the etchant flows to the chip surface, and the chip surface is corroded by the etchant, and each chip device may not function.

On the other hand, the polishing agent (slurry) used in the chemical mechanical polishing of the present application has no etching action under static conditions. That is, even if only the abrasive is supplied to the wafer, the etching does not proceed at all, and the wafer surface is merely modified. Therefore, even if the abrasive gets into the cracks on the planned cutting line of the wafer, it does not melt the surrounding silicon, so the cracks do not progress any further.

As a result, cracks may grow in the polishing process, that is, the etching process in Cited Document 2, but in the polishing process of the present application, the wafer is not etched at all even if the abrasive is added and left as it is, so cracks hardly grow. . As a result, the polishing agent permeates and the chip is peeled off from the film without permission, and the polishing agent flows into the chip surface and erodes the chip, causing the chip device to stop functioning. There is no

Here, the mechanism of chemical mechanical polishing is as follows. That is, in polishing, an abrasive is first applied to the wafer, which only chemically modifies the wafer surface. Next, since the modified surface is softer, the abrasive grains contained in the slurry act on the wafer surface in this state, so even with a very small stress, the wafer surface can be removed efficiently. be.

For example, for processes that remove silicon, silica-based slurries are used. When using a silica-based slurry, the following reactions occur on the wafer surface.

[Number 1]
Si(-OH)+SiOH=SiO2+H2O
That is, there are Si atoms on the wafer surface immediately after normal polishing. This Si atom is hydrated on the surface in water, and -OH exists on the Si atom surface. The OH groups attached to the surface of the Si substrate are linked to the SiOH of the silica sol present in the liquid and the SiOH present on the silica particle surface.

However, this alone does not progress the etching. As a result, in this state, the polishing pad containing a large amount of silica particles is moved relative to the substrate, and the silica particles act on the substrate surface under a certain amount of momentum, thereby maintaining a certain temperature. It is removed mechanically as the chemical reaction proceeds under the environment and pressure environment.

At this time, the silica sol and silica particles that permeate the cracks do not have the effect of further propagating the cracks. This is because the polishing pad does not enter into the cracks, so that no mechanical action works and, as a result, no chemical removal action occurs.

As a result, the work-affected layer on the surface of the substrate generated during grinding is supplied with a slurry having chemical properties containing silica sol and silica particles, and the polishing pad comes into contact with the substrate, causing mechanical friction. Together with this, only the surface of the substrate is chemically and mechanically polished and removed (Tshiro Doi, ed.: Detailed explanation of semiconductor CMP technology (Kogyo Chosakai) (2001), pp. 40-42).

As a result, most of the work-affected layer on the wafer surface is removed. Also, the formed wafer surface becomes a mirror surface. The principle of this mirror surface is as follows when compared with the chemical etching described in Cited Document 2.

In the case of chemical etching, as described above, lattice strain and crystal defect portions in crystals such as silicon are selectively etched. As a result, etch pits may be formed and large hollows may occur along the grain boundaries, which cannot be avoided in principle.

On the other hand, in the case of chemical-mechanical polishing, as mentioned above, the chemical removing action does not work unless the mechanical action works. The mechanical action, which here is the rubbing of the wafer surface by the polishing pad, has nothing to do with the crystalline state of the substrate surface and occurs with equal probability on all substrate surfaces. . Therefore, all surfaces are uniformly removed regardless of the crystal state, and chemical action works, so that the surface becomes a uniform surface with no crystal defects and a mirror surface. Here, the term "mirror surface" is ambiguous from a relative point of view, so in the present application, it is defined as a mirror surface obtained by chemical mechanical polishing.

By performing such chemical mechanical polishing, it is possible to obtain an ideal mirror surface state, which is greatly different from that in Cited Document 2, and to prevent erosion due to penetration of the etchant between the chip and the film.

It should be noted that the chips are not yet completely cracked after chemical mechanical polishing. Only microvoids developed from the modified layer remain on the surface part after chemical mechanical polishing. Note that the modified layer has already been removed by the previous grinding process.

If the state in which the micropores developed from the modified layer are left, for example, in the surface state after grinding and etching as shown in Cited Document 2, the micropores developed from the modified layer and It is no longer possible to clearly distinguish from the unevenness promoted by the etching treatment after grinding. Therefore, when breaking the chip, the chip is not necessarily broken from the microvoids, but depending on the case, it is conceivable that the chip may be broken from the irregularities formed by further enhancing the grinding streaks by etching.

When chemical mechanical polishing is applied to the surface as in the present application, there is almost no irregularity on the surface of the wafer after the treatment, and only microvoids developed from the modified layer remain. Therefore, when the cutting treatment is performed in this state, the cracks are further propagated from the microvoids, resulting in the cutting.

As described above, according to the method of the present invention, the substrate is not completely divided even after grinding and chemical mechanical polishing, although the microvoids are enlarged and the cracks are propagated. If a crack develops during grinding or chemical mechanical polishing and the substrate is completely split, the chips on the periphery of the wafer, in particular, cannot withstand the shear stress during grinding and polishing, and are separated from the suction table. There is a problem that causes chipping.

However, in the present invention, even after grinding and polishing, although cracks grow, they are not completely divided. In order to allow the crack to grow further from the state in which the material is not completely divided and to achieve complete cleaving, a further cleaving process is required. In the cleaving step, after an expand tape is attached to the back surface of the wafer, a pressing member is pressed through the tape to locally apply a bending stress to the wafer. As a result, it is possible to efficiently split from cracks developed by grinding.

Here, when the expanding tape is not attached, the pressing member directly presses the wafer, and when the wafer is cleaved, the vibration of the crack is propagated in the wafer plane. The propagated vibration may concomitantly crack other portions of the cutting line or crack portions other than the cutting line.

However, by attaching the expanding tape and applying stress by the pressing member through the tape, the tape absorbs the vibration of the crack when the wafer is cleaved, so that unnecessary vibration is not generated. In addition, since the expand tape is adhered in a stretched state without wrinkles or deflection, a constant tension is constantly applied to the wafer, and the wafer is restrained. When the wafer is restrained when the wafer is cleaved, the bending deformation of various parts of the wafer is restrained, and the pressing force of the pressing member bends against the developed crack part where the rigidity is weak inside the wafer. Instead, it concentrates as shear stress. As a result, only the portions to be originally cut are efficiently cut, and portions other than the cutting line are protected and will not crack.

In particular, when tapes are attached to both surfaces of the wafer, the vibration of cracking is completely contained, and bending of the wafer is further constrained, so that efficient breaking becomes possible.

(Appendix 2) a step of attaching a back grind tape to the front surface of a wafer; By forming the modified region, substantially the entire surface of the wafer on which the modified region is formed in the modified region forming step of forming microvoids in the modified region, and the modified region forming step is unified. a step of sucking the wafer to the table independently in each region; and grinding and removing a portion before the reformed region formed inside the wafer by irradiating the laser beam while the wafer is sucked, and removing the reformed region. A first grinding step for growing microcracks extending from the surface of the substrate in the depth direction of the substrate, a second grinding step for grinding and removing the modified region formed inside the wafer, and a chemical slurry for modifying the wafer surface. While chemical mechanical polishing is performed using a polishing pad, the work-affected layer introduced in the first and second grinding steps is removed while leaving microcracks extending from the modified region, and the surface is mirror-finished. a step of attaching an expanding tape to the back surface of the wafer having a mirror-finished surface; and pressing the wafer having the expanding tape attached to the back surface with a pressing member through the back grinding tape. a cleaving step of cleaving the wafer; a peeling step of sticking a peeling tape for peeling off the backgrind tape and peeling the backgrind tape with a peeling roller; and a splitting/separating step of splitting the tape into a plurality of chips by stretching the tape.

According to the invention described in Supplementary Note 2, the modified region is formed at a deep position inside the wafer at the beginning, and the first grinding step initially generates grinding heat while performing removal processing, thereby improving the quality. A crack formed in the region can be propagated to a deeper position in the wafer.

However, in the first grinding step, if the laser-modified region is also ground and removed with the same capacity as the unmodified region, the modified region has large crystal grain boundaries, so large crystal grains may be chipped off, Along with such crystal grains, even more critical cracks may develop. Therefore, in the first grinding step, it is preferable to grind up to a portion before the modified region where the crystallinity is constant.

The second grinding step primarily grinds the laser-modified region. At this time, the grinding wheel also has a high count, that is, a grinding wheel with a finer grain size than in the first grinding process is used, and compared to the first grinding process, fatal cracks and defects derived from the modified region are used. Grind gently so as not to induce The second grinding process is only for the modified region, and the grinding rate is set to be smaller than that of the first grinding process, so that the area is finely ground.

Then, after the laser-modified region is removed, chemical mechanical polishing is finally performed. In chemical mechanical polishing, a polishing pad made of polymer or non-woven fabric is pressed against the wafer while a chemical solution that modifies the wafer is supplied to chemically and mechanically polish the wafer.

If chemical mechanical polishing is performed on the above-mentioned modified region, large crystal grains may chip off from the modified region. Since chemical mechanical polishing uses a polishing pad such as non-woven fabric or foamed polyurethane, if chipped crystal grains enter the surface of such a pad, the chipped crystal grains will constantly cause scratches during polishing. In such a case, the polished surface is full of scratches without fulfilling the purpose of mirror-finishing while removing the work-affected layer in the grinding process.

For this reason, the modified layer introduced by the laser in the previous second grinding step must be completely removed when introduced into the chemical mechanical polishing step.

(Appendix 3) A modified layer is formed by laser light along the cutting line, the back grind tape is attached to the front surface, and the expand tape is attached to the back surface, and an elastic body is attached to the upper surface of the wafer. and the table is rotated so that the pressing member is parallel to the cutting line, and the pressing member is pressed against the wafer and rolled to cut the cutting line. The method for cutting a semiconductor substrate according to Supplementary Note 1 or Supplementary Note 2.

According to the invention described in Supplementary Note 3, the back grind tape is attached to the front surface of the wafer when the wafer is cleaved. An expanding tape is attached to the back surface by a tape attaching mechanism.

The wafer with the backgrinding tape and the expanding tape attached to both sides thereof is placed on a table having an elastic body attached to the upper surface thereof, which is provided in a mechanism for peeling off the backgrinding tape. Subsequently, a pressing member provided in the peeling mechanism is positioned parallel to the cutting line, and is pressed and rolled to break one of the cutting lines.

After one cutting line is completely broken, the wafer is rotated 90 degrees and the other cutting line is broken. The cleaved wafer is subjected to ultraviolet irradiation, heating, or the like, so that the adhesion of the back grind tape is reduced. A peeling roller is used to adhere a peeling tape to the back grind tape whose adhesive force has been reduced, and the tape is peeled off. The wafer from which the backgrinding tape has been peeled off is separated from the chips by expanding the expand film. As a result, the cleaving is reliably and efficiently performed, and continuous cleaving can be performed without adversely affecting subsequent wafers.

(Appendix 4) The semiconductor substrate cutting method according to appendix 3, wherein the pressing member is formed to have a radius smaller than the interval between the cutting lines.

According to the invention described in Supplementary Note 4, the radius of the pressing member is formed to be smaller than the interval between the cutting lines formed on the wafer to be pressed. Since the radius is smaller than the interval of the cutting lines, the pressing member does not press a plurality of cutting lines at once and the stress is dispersed. As a result, stress can be concentrated on one cutting line, and the cutting can be performed reliably and efficiently.

(Appendix 5) Cutting a semiconductor substrate according to appendix 3 or appendix 4, wherein grooves, holes, or irregularities are formed in the elastic body at intervals equal to or less than the size of the chips formed on the wafer. Method.

According to the invention described in Supplementary Note 5, the surface of the elastic body on which the wafer is placed is grooved, perforated, or unevenly processed. The elastic body is preferably an elastic body having an initial elastic modulus and a void-free elastic body having a small compression set so that the elastic modulus changes little after several uses. By forming grooves, holes, or irregularities in the elastic body at intervals equal to or smaller than the size of the chips formed on the wafer, the wafer can be locally deformed, and the cutting can be reliably and efficiently performed.

(Appendix 6) When the back grind tape or the expand tape is applied, the application roller is rotated in a direction that obliquely intersects the cutting lines formed in a grid pattern to apply the tape. The method for cutting a semiconductor substrate according to any one of appendices 3 to 5, characterized in that:

According to the invention described in Supplementary Note 6, when attaching the back grind tape or the expand tape to the front surface and the rear surface of the wafer, the table on which the wafer is placed is rotated by about 45 degrees to form a grid pattern. After pressing the sticking roller in a direction that obliquely intersects the cut line, the sticking roller is rolled to stick the tape.

If the cutting line is broken by the pressing force of the tape sticking roller, it becomes difficult to stick the tape accurately due to air bubbles entering the sticking surface and deviation of the chip direction. By moving the tape application roller so as to intersect the cutting line so as to be inclined, the cutting can be reliably and efficiently performed without adversely affecting the subsequent processes.

DESCRIPTION OF SYMBOLS 1... Laser dicing apparatus, 2... Grinding apparatus, 3... Tape peeling apparatus, 11... Wafer moving part, 13... Adsorption stage, 20... Laser optical part, 30... Observation optical part, 40... Laser head, 50... Control part, DESCRIPTION OF SYMBOLS 118... Rough grinding stage, 120... Fine grinding stage, 122... Polishing stage, 132, 136, 138, 140... Chuck, 146, 154... Cup type grindstone, 156... Polishing cloth, 211... Table, 212... Elastic body, 212a Groove 213 Supply reel 214 Peeling roller 215 Peeling tape 216, 217 Guide roller 218 Winding reel 219 Pressing member B Chip BG tape C Chip E Expand Tape, F... Frame, K... Modified area, L... Laser beam, S... Cutting line, W... Work

Claims (4)

A crack propagation device for propagating a crack from a laser-modified region formed inside a wafer,
Crack propagating means for changing the degree of crack extension by controlling the temperature of the back surface of the wafer when grinding the back surface of the wafer,
Crack growth device. The crack propagating means controls the temperature of the back surface of the wafer by controlling water supply when the grinding is performed.
The crack propagation device according to claim 1. A crack propagation method for propagating a crack from a laser-modified region formed inside a wafer,
A crack growth step of changing the degree of extension of the crack by controlling the temperature of the back surface of the wafer when grinding the back surface of the wafer,
Crack growth method. The crack growth step controls the temperature of the back surface of the wafer by controlling water supply when the grinding is performed.
The crack propagation method according to claim 3.

Images