Substrate dividing method

Technical Field

0001The present invention relates to a substrate dividing method used for dividing a substrate such as a semiconductor substrate in a step of making a semiconductor device or the like.

Background Art

0002As semiconductor devices have been becoming smaller in recent years, there are cases where a semiconductor substrate is thinned to a thickness of several tens of micrometers in a step of making a semiconductor device. When thus thinned semiconductor substrate is cut and divided by a blade, chipping and cracking occur more than in the case where a semiconductor substrate is thicker, thereby causing a problem that the yield of semiconductor chips obtained by dividing the semiconductor substrate decreases.

0003Known as semiconductor substrate dividing methods which can solve such a problem are those described in Japanese Patent Application Laid-Open Nos. <patcit id="pcit0001" dnum="JPSHO6438209B"><text>SHO 64-38209</text></patcit> and <patcit id="pcit0002" dnum="JPSHO624341A"><text>SHO 62-4341</text></patcit>.

0004In the methods described in these publications, a semiconductor substrate having a front face formed with a functional device is inscribed with a groove by a blade on the front face side, then an adhesive sheet is attached to the front face, so as to hold the semiconductor substrate, and the rear face of the semiconductor substrate is ground until the groove formed beforehand is exposed, thereby thinning the semiconductor substrate and dividing the semiconductor substrate.

0005<patcit id="pcit0003" dnum="EP1022778A1"><text>EP1022778A1</text></patcit> describes a method of dividing a wafer and method of manufacturing a semiconductor device wherein grooves are formed in a surface of a wafer, on which surface semiconductor elements are formed, along dicing lines or chip parting lines on the wafer. The grooves are deeper than a thickness of a finished chip. A holding sheet is attached on that surface of the wafer on which the semiconductor elements are formed. The bottom surface of the wafer is lapped and polished to the thickness of the finished chip, thereby dividing the wafer into chips. When the wafer is divided into the chips, the lapping and polishing is continued until the thickness of the wafer becomes equal to the thickness of the finished chip, even after the wafer has been divided into the chips by the lapping and polishing.

Disclosure of the Invention

0006If the grinding of the rear face of the semiconductor substrate is performed by surface grinding in the methods described in the above-mentioned publications, however, chipping and cracking may occur at side faces of the groove formed beforehand in the semiconductor substrate when the surface-ground face reaches the groove.

0007In view of such a circumstance, it is an object of the present invention to provide a substrate dividing method which can prevent chipping and cracking from occurring, and thin and divide a substrate.

0008For achieving the above-mentioned object, the substrate dividing method in accordance with the present invention is defined in claim 1.

0009In the step of grinding the substrate, the substrate is ground such that the substrate attains a predetermined thickness after the starting point region for cutting is formed within the substrate. Here, even when the ground surface reaches the fracture generated from the starting point region for cutting acting as a start point, cut surfaces of the substrate cut by the fracture remain in close contact with each other, whereby the substrate can be prevented from chipping and cracking upon grinding.

0010This can prevent chipping and cracking from occurring, and can thin and divide the substrate.

0011Here, the light-converging point refers to a location at which laser light is converged. The grinding encompasses shaving, polishing, chemical etching, and the like. The starting point region for cutting refers to a region to become a start point for cutting when the substrate is cut. Therefore, the starting point region for cutting is a part to cut where cutting is to be performed in the substrate. The starting point region for cutting may be produced by continuously forming a modified region or intermittently forming a modified region.

0012The substrate encompasses semiconductor substrates such as silicon substrates and GaAs substrates. Since the substrate can be ground after forming the functional device, a chip thinned so as to conform to a smaller size of a semiconductor device, for example, is obtained. Here, the functional device refers to light-receiving devices such as photodiodes, light-emitting devices such as laser diodes, circuit devices formed as circuits, etc.

0013Preferably, the step of grinding the substrate includes a step of subjecting the rear face of the substrate to chemical etching. When the rear face of the substrate is subjected to chemical etching, the rear face of the substrate becomes smoother as a matter of course. Also, since the cut surfaces of the substrate cut by the fracture generated from the starting point region for cutting acting as a start point remain in close contact with each other, only edge parts on the rear face of the cut surfaces are selectively etched, so as to be chamfered. This can improve the transverse rupture strength of chips obtained by dividing the substrate, and prevent chipping and cracking from occurring in the chips.

Brief Description of the Drawings

0014<ul id="ul0001" list-style="none" compact="compact"><li><figref idref="f0001">Fig. 1</figref> is a plan view of an object to be processed during laser processing in the laser processing method in accordance with an embodiment of the present invention;</li><li><figref idref="f0001">Fig. 2</figref> is a sectional view of the object to be processed taken along the line II-II of <figref idref="f0001">Fig. 1</figref>;</li><li><figref idref="f0002">Fig. 3</figref> is a plan view of the object to be processed after laser processing by the laser processing method in accordance with the embodiment;</li><li><figref idref="f0002">Fig. 4</figref> is a sectional view of the object to be processed taken along the line IV-IV of <figref idref="f0002">Fig. 3</figref>;</li><li><figref idref="f0003">Fig. 5</figref> is a sectional view of the object to be processed taken along the line V-V of <figref idref="f0002">Fig. 3</figref>;</li><li><figref idref="f0003">Fig. 6</figref> is a plan view of the object to be processed cut by the laser processing method in accordance with the embodiment;</li><li><figref idref="f0004">Fig. 7</figref> is a graph showing relationships between the electric field intensity and crack spot size in the laser processing method in accordance with the embodiment;</li><li><figref idref="f0005">Fig. 8</figref> is a sectional view of the object to be processed in a first step of the laser processing method in accordance with the embodiment;</li><li><figref idref="f0005">Fig. 9</figref> is a sectional view of the object to be processed in a second step of the laser processing method in accordance with the embodiment;</li><li><figref idref="f0006">Fig. 10</figref> is a sectional view of the object to be processed in a third step of the laser processing method in accordance with the embodiment;</li><li><figref idref="f0006">Fig. 11</figref> is a sectional view of the object to be processed in a fourth step of the laser processing method in accordance with the embodiment;</li><li><figref idref="f0007">Fig. 12</figref> is a view showing a photograph of a cut section in a part of a silicon wafer cut by the laser processing method in accordance with the embodiment;</li><li><figref idref="f0008">Fig. 13</figref> is a graph showing relationships between the laser light wavelength and the internal transmittance of a silicon substrate in the laser processing method in accordance with the embodiment;</li><li><figref idref="f0009">Fig. 14</figref> is a schematic diagram of the laser processing apparatus in accordance with Example 1;</li><li><figref idref="f0010">Fig. 15</figref> is a flowchart for explaining the laser processing method in accordance with Example 1;</li><li><figref idref="f0011">Fig. 16</figref> is a view showing the semiconductor substrate after a step of forming a starting point region for cutting in accordance with Example 1;</li><li><figref idref="f0012">Fig. 17</figref> is a view for explaining a step of attaching a protective film in accordance with Example 1;</li><li><figref idref="f0012">Fig. 18</figref> is a view for explaining a step of grinding the semiconductor substrate in accordance with Example 1;</li><li><figref idref="f0012">Fig. 19</figref> is a view for explaining a step of attaching an expansion film in accordance with Example 1;</li><li><figref idref="f0013">Fig. 20</figref> is a view for explaining a step of peeling the protective film in accordance with Example 1;</li><li><figref idref="f0013">Fig. 21</figref> is a view for explaining a step of expanding the expansion film and picking up semiconductor chips in accordance with Example 1;</li><li><figref idref="f0014">Fig. 22</figref> is a view showing chamfers formed at edge parts on the rear face side of cut surfaces of semiconductor chips after the step of grinding the semiconductor substrate in accordance with Example 1;</li><li><figref idref="f0015">Fig. 23A</figref> is a view for explaining a case where a molten processed region remains in a cut surface of a semiconductor chip after the step of grinding the semiconductor substrate in accordance with Example 1, while a fracture reaches the front face before the step of grinding the semiconductor substrate;</li><li><figref idref="f0015">Fig. 23B</figref> is a view for explaining a case where a molten processed region remains in a cut surface of a semiconductor chip after the step of grinding the semiconductor substrate in accordance with Example 1, while a fracture does not reach the front face before the step of grinding the semiconductor substrate;</li><li><figref idref="f0015">Fig. 24A</figref> is a view for explaining a case where a molten processed region does not remain in a cut surface of a semiconductor chip after the step of grinding the semiconductor substrate in accordance with Example 1, while a fracture reaches the front face before the step of grinding the semiconductor substrate;</li><li><figref idref="f0015">Fig. 24B</figref> is a view for explaining a case where a molten processed region does not remain in a cut surface of a semiconductor chip after the step of grinding the semiconductor substrate in accordance with Example 1, while a fracture does not reach the front face before the step of grinding the semiconductor substrate;</li><li><figref idref="f0016">Fig. 25A</figref> is a view for explaining a case where a molten processed region remains in an edge part on the rear face side of a cut surface of a semiconductor chip after the step of grinding the semiconductor substrate in accordance with Example 1, while a fracture reaches the front face before the step of grinding the semiconductor substrate;</li><li><figref idref="f0016">Fig. 25B</figref> is a view for explaining a case where a molten processed region remains in an edge part on the rear face side of a cut surface of a semiconductor chip after the step of grinding the semiconductor substrate in accordance with Example 1, while a fracture does not reach the front face before the step of grinding the semiconductor substrate;</li><li><figref idref="f0016">Fig. 26A</figref> is a sectional view of a marginal part of the semiconductor substrate before the step of grinding the semiconductor substrate in accordance with Example 1;</li><li><figref idref="f0016">Fig. 26B</figref> is a sectional view of the marginal part of the semiconductor substrate after the step of grinding the semiconductor substrate in accordance with Example 1;</li></ul>

Best Modes for Carrying Out the Invention

0015In the following, a preferred embodiment of the present invention will be explained in detail with reference to drawings. The substrate dividing method in accordance with this method comprises the steps of irradiating a substrate with laser light while positioning a light-converging point within the substrate, so as to form a modified region due to multiphoton absorption within the substrate, thereby forming a starting point region for cutting; and then grinding the substrate such that the substrate attains a predetermined thickness.

0016First, a laser processing method carried out in the step of forming the starting point region for cutting, multiphoton absorption in particular, will be explained.

0017A material becomes optically transparent if its absorption bandgap EG is greater than a photon energy hν. Hence, the condition under which absorption occurs in the material is hν > EG. However, even when optically transparent, the material yields absorption under the condition of nhv > EG (n = 2, 3, 4, ...) if the intensity of laser light is very high. This phenomenon is known as multiphoton absorption. In the case of pulse waves, the intensity of laser light is determined by the peak power density (W/cm2) of laser light at a light-converging point thereof. The multiphoton absorption occurs, for example, at a peak power density (W/cm2) of 1 × 108 (W/cm2) or higher. The peak power density is determined by (energy per pulse of laser light at the light-converging point)/(laser light beam spot cross-sectional area × pulse width). In the case of a continuous wave, the intensity of laser light is determined by the electric field strength (W/cm2) of laser light at the light-converging point.

0018The principle of laser processing in accordance with the embodiment utilizing such multiphoton absorption will now be explained with reference to <figref idref="f0001 f0002 f0003">Figs. 1 to 6</figref>. <figref idref="f0001">Fig. 1</figref> is a plan view of a substrate 1 during laser processing; <figref idref="f0001">Fig. 2</figref> is a sectional view of the substrate 1 taken along the line II-II of <figref idref="f0001">Fig. 1</figref>; <figref idref="f0002">Fig. 3</figref> is a plan view of the substrate 1 after laser processing; <figref idref="f0002">Fig. 4</figref> is a sectional view of the substrate 1 taken along the line IV-IV of <figref idref="f0002">Fig. 3</figref>; <figref idref="f0003">Fig. 5</figref> is a sectional view of the substrate 1 taken along the line V-V of <figref idref="f0002">Fig. 3</figref>; and <figref idref="f0003">Fig. 6</figref> is a plan view of the cut substrate 1.

0019As shown in <figref idref="f0001">Figs. 1 and 2</figref>, the front face 3 of the substrate 1 has a desirable line along which the substrate should be cut 5 for cutting the substrate 1. The line along which the substrate should be cut 5 is a linearly extending virtual line (the substrate 1 may also be formed with an actual line acting as the line along which the substrate should be cut 5). In the laser processing in accordance with this embodiment, the substrate 1 is irradiated with laser light L such that a light-converging point P is positioned within the semiconductor substrate 1 under a condition causing multiphoton absorption, so as to form a modified region 7. Here, the light-converging point is a location where the laser light L is converged.

0020The laser light L is relatively moved along the line along which the substrate should be cut 5 (in the direction of arrow A), so as to move the light-converging point P along the line along which the substrate should be cut 5. This forms the modified region 7 along the line along which the substrate should be cut 5 only within the substrate 1 as shown in <figref idref="f0002 f0003">Figs. 3 to 5</figref>, and the modified region 7 forms a starting point region for cutting (part to cut) 8. In the laser processing method in accordance with this embodiment, no modified region 7 is formed upon heating the substrate 1 by causing the substrate 1 to absorb the laser light L. Instead, the laser light L is transmitted through the semiconductor substrate 1, so as to generate multiphoton absorption within the semiconductor substrate 1, thereby forming the modified region 7. Hence, the laser light L is hardly absorbed by the front face 3 of the semiconductor substrate 1, whereby the front face 3 of the semiconductor substrate 1 does not melt.

0021If a start point exists at a location to cut when cutting the substrate 1, the substrate 1 fractures from this start point and thus can be cut with a relatively small force as shown in <figref idref="f0003">Fig. 6</figref>. This makes it possible to cut the substrate 1 without generating unnecessary fractures in the front face 3 of the substrate 1.

0022There seem to be the following two ways of cutting the substrate from the starting point region for cutting acting as a start point. The first case is where, after forming the starting point region for cutting, an artificial force is applied to the substrate, so that the substrate fractures from the starting point region for cutting acting as a start point, whereby the substrate is cut. This is the cutting in the case where the substrate has a large thickness, for example. The application of an artificial force encompasses application of bending stress and shearing stress along the starting point region for cutting of the substrate, and exertion of a temperature difference upon the substrate to generate thermal stress, for example. The other case is where a starting point region for cutting is formed, so that the substrate is naturally fractured in a cross-sectional direction (thickness direction) of the substrate from the starting point region for cutting acting as a start point, whereby the substrate is cut. This is enabled, for example, by forming the starting point region for cutting by a single row of modified regions when the substrate has a small thickness, and by a plurality of rows of modified regions aligned in the thickness direction when the substrate has a large thickness. Even in the case of natural fracturing, fractures do not extend to the front face at a location not formed with the starting point region for cutting in the part to cut, whereby only the part corresponding to the location formed with the starting point region for cutting can be fractured. Thus, fracturing can be regulated well. Such a fracturing method with favorable controllability is quite effective, since semiconductor substrates such as silicon wafers have recently been apt to become thinner.

0023The modified region formed by multiphoton absorption in this example includes the following cases (1) to (3). Cases (1) and (3) are not concerned by the present invention:

(1) Case where the modified region is a crack region including one or a plurality of cracks

0024A substrate (e.g., glass or a piezoelectric material made of LiTa03) is irradiated with laser light while a light-converging point is positioned therewithin under a condition with an electric field intensity of at least 1 × 108 (W/cm2) at the light-converging point and a pulse width of 1 µs or less. This pulse width is a condition under which a crack region can be formed only within the substrate while generating multiphoton absorption without causing unnecessary damages to the substrate. This generates a phenomenon of optical damage due to multiphoton absorption within the substrate. This optical damage induces thermal distortion within the substrate, thereby forming a crack region therewithin. The upper limit of electric field intensity is 1 × 1012 (W/cm2), for example. The pulse width is preferably 1 ns to 200 ns, for example. The forming of a crack region due to multiphoton absorption is described, for example, in "<nplcit id="ncit0001" npl-type="s"><text>Internal Marking of Glass Substrate by Solid-state Laser Harmonics", Proceedings of 45th Laser Materials Processing Conference (December 1998), pp. 23-28</text></nplcit>.

0025The inventors determined relationships between the electric field intensity and the magnitude of crack by an experiment. Conditions for the experiment are as follows: <ol id="ol0001" compact="compact"><li>(A) Substrate: Pyrex (registered trademark) glass (having a thickness of 700 µm)</li><li>(B) Laser <ul id="ul0002" list-style="none" compact="compact"><li>Light source: semiconductor laser pumping Nd:YAG laser Wavelength: 1064 nm</li><li>Laser light spot cross-sectional area: 3.14 × 10-8 cm2</li><li>Oscillation mode: Q-switch pulse</li><li>Repetition frequency: 100 kHz</li><li>Pulse width: 30 ns</li><li>Output: output &lt; 1 mJ/pulse</li><li>Laser light quality: TEM00</li><li>Polarization characteristic: linear polarization</li></ul></li><li>(C) Light-converging lens Transmittance with respect to laser light wavelength: 60% </li><li>(D) Moving speed of a mounting table mounting the substrate: 100 mm/sec</li></ol>

0026Here, the laser light quality being TEM00 indicates that the light convergence is so high that light can be converged up to about the wavelength of laser light.

0027<figref idref="f0004">Fig. 7</figref> is a graph showing the results of the above-mentioned experiment. The abscissa indicates peak power density. Since laser light is pulse laser light, its electric field intensity is represented by the peak power density. The ordinate indicates the size of a crack part (crack spot) formed within the substrate processed by one pulse of laser light. Crack spots gather, so as to form a crack region. The size of a crack spot refers to that of the part of dimensions of the crack spot yielding the maximum length. The data indicated by black circles in the graph refers to a case where the light-converging lens (C) has a magnification of x100 and a numerical aperture (NA) of 0.80. On the other hand, the data indicated by white circles in the graph refers to a case where the light-converging lens (C) has a magnification of x50 and a numerical aperture (NA) of 0.55. It is seen that crack spots begin to occur within the substrate when the peak power density reaches about 1011 (W/cm2), and become greater as the peak power density increases.

0028A mechanism by which the substrate is cut upon formation of a crack region in the laser processing in accordance with this embodiment will now be explained with reference to <figref idref="f0005 f0006">Figs. 8 to 11</figref>. As shown in <figref idref="f0005">Fig. 8</figref>, the substrate 1 is irradiated with laser light L while positioning the light-converging point P within the substrate 1 under a condition where multiphoton absorption occurs, so as to form a crack region 9 therewithin along a line along which the substrate should be cut. The crack region 9 is a region including one or a plurality of crack spots. The crack region 9 forms a starting point region for cutting. As shown in <figref idref="f0005">Fig. 9</figref>, the crack further grows while using the crack region 9 as a start point (i.e., using the starting point region for cutting as a start point). As shown in <figref idref="f0006">Fig. 10</figref>, the crack reaches the front face 3 and rear face 21 of the substrate 1. As shown in <figref idref="f0006">Fig. 11</figref>, the substrate 1 breaks, so as to be cut. The crack reaching the front face and rear face of the substrate may grow naturally or grow as a force is applied to the substrate.

(2) Case where the modified region is a molten processed region

0029A substrate (e.g., a semiconductor material such as silicon) is irradiated with laser light while a light-converging point is positioned therewithin under a condition with an electric field intensity of at least 1 × 108 (W/cm2) at the light-converging point and a pulse width of 1 µs or less. As a consequence, the inside of the substrate is locally heated by multiphoton absorption. This heating forms a molten processed region within the substrate. The molten processed region refers to a region once melted and then re-solidified, a region just in a melted state, or a region in the process of re-solidifying from its melted state, and may also be defined as a phase-changed region or a region having changed its crystal structure. The molten processed region may also be regarded as a region in which a certain structure has changed into another structure in monocrystal, amorphous, and polycrystal structures. Namely, it refers to a region in which a monocrystal structure has changed into an amorphous structure, a region in which a monocrystal structure has changed into a polycrystal structure, and a region in which a monocrystal structure has changed into a structure including an amorphous structure and a polycrystal structure, for example. When the substrate is a silicon monocrystal structure, the molten processed region is an amorphous silicon structure, for example. The upper limit of electric field intensity is 1 × 1012 (W/cm2), for example. The pulse width is preferably 1 ns to 200 ns, for example.

0030By an experiment, the inventors have verified that a molten processed region is formed within a silicon wafer.

Conditions for the experiment are as follows:

0031<ol id="ol0002" compact="compact"><li>(A) Substrate: silicon wafer (having a thickness of 350 µm and an outer diameter of 10.2 cm (4 inches))</li><li>(B) Laser <ul id="ul0003" list-style="none" compact="compact"><li>Light source: semiconductor laser pumping Nd:YAG laser</li><li>Wavelength: 1064 nm</li><li>Laser light spot cross-sectional area: 3.14 × 10-8 cm2</li><li>Oscillation mode: Q-switch pulse</li><li>Repetition frequency: 100 kHz</li><li>Pulse width: 30 ns</li><li>Output: 20 µJ/pulse</li><li>Laser light quality: TEM00</li><li>Polarization characteristic: linear polarization</li></ul></li><li>(C) Light-converging lens Magnification: x50 N. A.: 0.55</li></ol>

0032Transmittance with respect to laser light wavelength: 60%

(D) Moving speed of a mounting table mounting the substrate: 100 mm/sec

0033<figref idref="f0007">Fig. 12</figref> is a view showing a photograph of a cut section in a part of a silicon wafer cut by laser processing under the above-mentioned conditions. A molten processed region 13 is formed within a silicon wafer 11. The size of the molten processed region 13 formed under the above-mentioned conditions is about 100 µm in the thickness direction.

0034The fact that the molten processed region 13 is formed by multiphoton absorption will now be explained. <figref idref="f0008">Fig. 13</figref> is a graph showing relationships between the wavelength of laser light and the transmittance within the silicon substrate. Here, respective reflecting components on the front face side and rear face side of the silicon substrate are eliminated, whereby only the transmittance therewithin is represented. The above-mentioned relationships are shown in the cases where the thickness t of the silicon substrate is 50 µm, 100 µm, 200 µm, 500 µm, and 1000 µm, respectively.

0035For example, it is seen that laser light transmits through the silicon substrate by at least 80% at 1064 nm, where the wavelength of Nd:YAG laser is located, when the silicon substrate has a thickness of 500 µm or less. Since the silicon wafer 11 shown in <figref idref="f0007">Fig. 12</figref> has a thickness of 350 µm, the molten processed region 13 due to multiphoton absorption is formed near the center of the silicon wafer, i.e., at a part separated from the front face by 175 µm. The transmittance in this case is 90% or greater with reference to a silicon wafer having a thickness of 200 µm, whereby the laser light is absorbed within the silicon wafer 11 only slightly and is substantially transmitted therethrough. This means that the molten processed region 13 is not formed by laser light absorption within the silicon wafer 11 (i.e., not formed upon usual heating with laser light), but by multiphoton absorption. The forming of a molten processed region by multiphoton absorption is described, for example, in "<nplcit id="ncit0002" npl-type="s"><text>Processing Characteristic Evaluation of Silicon by Picosecond Pulse Laser", Preprints of the National Meeting of Japan Welding Society, No. 66 (April 2000), pp. 72-73</text></nplcit>.

0036Here, a fracture is generated in the cross-sectional direction while using a molten processed region as a start point, whereby the silicon wafer is cut when the fracture reaches the front face and rear face of the silicon wafer. The fracture reaching the front face and rear face of the silicon wafer may grow naturally or grow as a force is applied to the silicon wafer. The fracture naturally grows from the starting point region for cutting to the front face and rear face of the silicon wafer in any of the cases where the fracture grows from the molten processed region in a melted state and where the fracture grows from the molten processed region in the process of re-solidifying from the melted state. In any of these cases, the molten processed region is formed only within the silicon wafer. In the cut section after cutting, the molten processed region is formed only therewithin as shown in <figref idref="f0007">Fig. 12</figref>. When a molten processed region is formed within the substrate, unnecessary fractures deviating from a line along which the substrate should be cut are hard to occur at the time of fracturing, which makes it easier to control the fracturing.

(3) Case where the modified region is a refractive index change region (not concerned by the invention).

0037A substrate (e.g., glass) is irradiated with laser light while a light-converging point is positioned therewithin under a condition with an electric field intensity of at least 1 × 108 (W/cm2) at the light-converging point and a pulse width of 1 ns or less. When multiphoton absorption is generated within the substrate with a very short pulse width, the energy caused by multiphoton absorption is not transformed into thermal energy, so that a permanent structural change such as ionic valence change, crystallization, or polarization orientation is induced within the substrate, whereby a refractive index change region is formed. The upper limit of electric field intensity is 1 × 1012 (W/cm2), for example. The pulse width is preferably 1 ns or less, more preferably 1 ps or less, for example. The forming of a refractive index change region by multiphoton absorption is described, for example, in "<nplcit id="ncit0003" npl-type="s"><text>Formation of Photoinduced Structure within Glass by Femtosecond Laser Irradiation", Proceedings of 42th Laser Materials Processing Conference (November 1997), pp. 105-111</text></nplcit>.

0038The cases of (1) to (3) are explained as modified regions formed by multiphoton absorption in the foregoing. When a starting point region for cutting is formed as follows in view of the crystal structure of the substrate, the cleavage property thereof, and the like, the substrate can be cut with a smaller force and a higher accuracy while using the starting point region for cutting as a start point.

0039Namely, in the case of a substrate made of a monocrystal semiconductor having a diamond structure such as silicon, the starting point region for cutting is preferably formed in a direction along the (111) plane (first cleavage plane) or (110) plane (second cleavage plane). In the case of a substrate made of a III-V family compound semiconductor having a zinc ore type structure such as GaAs, the starting point region for cutting is preferably formed in a direction along the (110) plane. In the case of a substrate having a hexagonal crystal structure such as sapphire (Al2O3), a starting point region for cutting is preferably formed in a direction along the (1120) plane (A plane) or (1100) plane (M plane) while using the (0001) plane (C plane) as a principal plane.

0040When the substrate is formed with an orientation flat along a direction to be formed with the starting point region for cutting (e.g., in a direction along the (111) plane in the monocrystal silicon substrate) or a direction orthogonal to the direction to be formed with the starting point region for cutting, the starting point region for cutting extending along the direction to be formed with the starting point region for cutting can be formed in the substrate in an easy and accurate manner with reference to the orientation flat.

0041In the following, the present invention will be explained more specifically with reference to Examples.

Example 1

0042Example 1 of the substrate dividing method in accordance with the present invention will now be explained. Example 1 is directed to a case where the substrate 1 is a silicon wafer (having a thickness of 350 µm and an outer diameter of (10.2 cm (4 inches)) ("substrate 1" will hereinafter be referred to as "semiconductor substrate 1" in Example 1), whereas the front face 3 of the semiconductor substrate 1 is formed with a plurality of functional devices in a device manufacturing process.

0043First, before explaining a step of forming a starting point region for cutting within the semiconductor substrate 1, a laser processing apparatus employed in the step of forming a starting point region for cutting will be explained with reference to <figref idref="f0009">Fig. 14. Fig. 14</figref> is a schematic diagram of the laser processing apparatus 100.

0044The laser processing apparatus 100 comprises a laser light source 101 for generating laser light L; a laser light source controller 102 for controlling the laser light source 101 so as to regulate the output, pulse width, etc. of laser light L and the like; a dichroic mirror 103, arranged so as to change the orientation of the optical axis of laser light L by 90°, having a function of reflecting the laser light L; a light-converging lens 105 for converging the laser light L reflected by the dichroic mirror 103; a mounting table 107 for mounting a semiconductor substrate 1 irradiated with the laser light L converged by the light-converging lens 105; an X-axis stage 109 for moving the mounting table 107 in the X-axis direction; a Y-axis stage 111 for moving the mounting table 107 in the Y-axis direction orthogonal to the X-axis direction; a Z-axis stage 113 for moving the mounting table 107 in the Z-axis direction orthogonal to the X- and Y-axis directions; and a stage controller 115 for controlling the movement of these three stages 109, 111, 113.

0045The Z-axis direction is a direction orthogonal to the front face 3 of the semiconductor substrate 1, and thus becomes the direction of focal depth of laser light L incident on the semiconductor substrate 1. Therefore, moving the Z-axis stage 113 in the Z-axis direction can position the light-converging point P of laser light L within the semiconductor substrate 1. This movement of light-converging point P in X(Y)-axis direction is effected by moving the semiconductor substrate 1 in the X(Y)-axis direction by the X(Y)-axis stage 109 (111).

0046The laser light source 101 is an Nd:YAG laser generating pulse laser light. Known as other kinds of laser usable as the laser light source 101 include Nd:YVO4 laser, Nd:YLF laser, and titanium sapphire laser. For forming a molten processed region, Nd:YAG laser, Nd:YVO4 laser, and Nd:YLF laser are preferably employed. Though pulse laser light is used for processing the semiconductor substrate 1 in Example 1, continuous wave laser light may be used as long as it can cause multiphoton absorption.

0047The laser processing apparatus 100 further comprises an observation light source 117 for generating a visible light beam for irradiating the semiconductor substrate 1 mounted on the mounting table 107, and a visible light beam splitter 119 disposed on the same optical axis as that of the dichroic mirror 103 and light-converging lens 105. The dichroic mirror 103 is disposed between the beam splitter 119 and light-converging lens 105. The beam splitter 119 has a function of reflecting about a half of a visual light beam and transmitting the remaining half therethrough, and is arranged so as to change the orientation of the optical axis of the visual light beam by 90°. About a half of the visible light beam generated from the observation light source 117 is reflected by the beam splitter 119, and thus reflected visible light beam is transmitted through the dichroic mirror 103 and light-converging lens 105, so as to illuminate the front face 3 of the semiconductor substrate 1 including the line along which the substrate should be cut 5 and the like.

0048The laser processing apparatus 100 further comprises an image pickup device 121 and an imaging lens 123 which are disposed on the same optical axis as that of the beam splitter 119, dichroic mirror 103, and light-converging lens 105. An example of the image pickup device 121 is a CCD camera. The reflected light of the visual light beam having illuminated the front face 3 including the line along which the substrate should be cut 5 and the like is transmitted through the light-converging lens 105, dichroic mirror 103, and beam splitter 119 and forms an image by way of the imaging lens 123, whereas thus formed image is captured by the image pickup device 121, so as to yield imaging data.

0049The laser processing apparatus 100 further comprises an imaging data processor 125 for inputting the imaging data outputted from the image pickup device 121, an overall controller 127 for controlling the laser processing apparatus 100 as a whole, and a monitor 129. According to the imaging data, the imaging data processor 125 calculates focal point data for positioning the focal point of the visible light generated from the observation light source 117 onto the front face 3. According to the focal point data, the stage controller 115 controls the movement of the Z-axis stage 113, so that the focal point of visible light is positioned on the front face 3. Hence, the imaging data processor 125 functions as an autofocus unit. Also, according to the imaging data, the imaging data processor 125 calculates image data such as an enlarged image of the front face 3. The image data is sent to the overall controller 127, subjected to various kinds of processing therein, and then sent to the monitor 129. As a consequence, an enlarged image or the like is displayed on the monitor 129.

0050Data from the stage controller 115, image data from the imaging data processor 125, and the like are fed into the overall controller 127. According to these data as well, the overall controller 127 regulates the laser light source controller 102, observation light source 117, and stage controller 115, thereby controlling the laser processing apparatus 100 as a whole. Thus, the overall controller 127 functions as a computer unit.

0051With reference to <figref idref="f0009">Figs. 14</figref> and <figref idref="f0010">15</figref>, a step of forming a starting point region for cutting in the case using the above-mentioned laser processing apparatus 100 will be explained. <figref idref="f0010">Fig. 15</figref> is a flowchart for explaining the step of forming a starting point region for cutting.

0052Light absorption characteristics of the semiconductor substrate 1 are determined by a spectrophotometer or the like which is not depicted. According to the results of measurement, a laser light source 101 generating laser light L having a wavelength to which the semiconductor substrate 1 is transparent or exhibits a low absorption is chosen (S101). Subsequently, the thickness of the semiconductor substrate 1 is measured. According to the result of measurement of thickness and the refractive index of the semiconductor substrate 1, the amount of movement of the semiconductor substrate 1 in the Z-axis direction is determined (S103). This is an amount of movement of the semiconductor substrate 1 in the Z-axis direction with reference to the light-converging point P of laser light L positioned at the front face 3 of the semiconductor substrate 1 in order for the light-converging point P of laser light L to be positioned within the semiconductor substrate 1. This amount of movement is fed into the overall controller 127.

0053The semiconductor substrate 1 is mounted on the mounting table 107 of the laser processing apparatus 100. Subsequently, visible light is generated from the observation light source 117, so as to illuminate the semiconductor substrate 1 (S105). The illuminated front face 3 of the semiconductor substrate 1 including the line along which the substrate should be cut 5 is captured by the image pickup device 121. The line along which the substrate should be cut 5 is a desirable virtual line for cutting the semiconductor substrate 1. Here, in order to obtain semiconductor chips by dividing the semiconductor substrate 1 into the functional devices formed on its front face 3, the line along which the substrate should be cut 5 is set like a grid running between the functional devices adjacent each other. The imaging data captured by the imaging device 121 is sent to the imaging data processor 125. According to the imaging data, the imaging data processor 125 calculates such focal point data that the focal point of visible light from the observation light source 117 is positioned at the front face 3 (S107).

0054The focal point data is sent to the stage controller 115. According to the focal point data, the stage controller 115 moves the Z-axis stage 113 in the Z-axis direction (S109). As a consequence, the focal point of visible light from the observation light source 117 is positioned at the front face 3 of the semiconductor substrate 1. According to the imaging data, the imaging data processor 125 calculates enlarged image data of the front face 3 of the semiconductor substrate 1 including the line along which the substrate should be cut 5. The enlarged image data is sent to the monitor 129 by way of the overall controller 127, whereby an enlarged image of the line along which the substrate should be cut 5 and its vicinity is displayed on the monitor 129.

0055Movement amount data determined in step S103 has been fed into the overall controller 127 beforehand, and is sent to the stage controller 115. According to the movement amount data, the stage controller 115 causes the Z-axis stage 113 to move the substrate 1 in the Z-axis direction to a position where the light-converging point P of laser light L is positioned within the semiconductor substrate 1 (S111).

0056Subsequently, laser light L is generated from the laser light source 101, so as to irradiate the line along which the substrate should be cut 5 in the front face 3 of the semiconductor substrate 1. Then, the X-axis stage 109 and Y-axis stage 111 are moved along the line along which the substrate should be cut 5, so as to form a molten processed region along the line along which the substrate should be cut 5, thereby forming a starting point region for cutting within the semiconductor substrate 1 along the line along which the substrate should be cut 5 (S113).

0057The foregoing completes the step of forming a starting point region for cutting, thereby forming the starting point region for cutting within the semiconductor substrate 1. When the starting point region for cutting is formed within the semiconductor substrate 1, a fracture is generated in the thickness direction of the semiconductor substrate 1 from the starting point region for cutting acting as a start point naturally or with a relatively small force exerted thereon.

0058In Example 1, the starting point region for cutting is formed at a position near the front face 3 side within the semiconductor substrate 1 in the above-mentioned step of forming a starting point region for cutting, and a fracture is generated in the thickness direction of the semiconductor substrate 1 from the starting point region for cutting acting as a start point. <figref idref="f0011">Fig. 16</figref> is a view showing the semiconductor substrate 1 after the starting point region for cutting is formed. As shown in <figref idref="f0011">Fig. 16</figref>, fractures 15 generated from the starting point region for cutting acting as a start point are formed like a grid along lines to cut, and reach only the front face 3 of the semiconductor substrate 1 but not the rear face 21 thereof. Namely, the fractures 15 generated in the semiconductor substrate 1 separate a plurality of functional devices 19 formed like a matrix on the front face of the semiconductor substrate 1 from each other. The cut surfaces of the semiconductor substrate 1 cut by the fractures 15 are in close contact with each other.

0059Here, "the starting point region for cutting is formed at a position near the front face 3 side within the semiconductor substrate 1" means that a modified region such as a molten processed region constituting a starting point region for cutting is formed so as to shift from the center position in the thickness direction of the semiconductor substrate 1 (i.e., half thickness position) toward the front face 3. Namely, it refers to a case where the center position of the width of the modified region in the thickness direction of the semiconductor substrate 1 is shifted toward the front face 3 from the center position in the thickness direction of the semiconductor substrate 1, and is not limited to the case where the whole modified region is located on the front face 3 side from the center position in the thickness direction of the semiconductor substrate 1.

0060The step of grinding the semiconductor substrate 1 will now be explained with reference to <figref idref="f0012 f0013">Figs. 17 to 21. Figs. 17 to 21</figref> are views for explaining respective steps including the step of grinding the semiconductor substrate. In Example 1, the semiconductor substrate 1 is thinned from the thickness of 350 µm to a thickness of 50 µm.

0061As shown in <figref idref="f0012">Fig. 17</figref>, a protective film 20 is attached to the front face 3 of the semiconductor substrate after the starting point region for cutting is formed. The protective film 20 is used for protecting the functional devices 19 formed on the front face 3 of the semiconductor substrate 1 and holding the semiconductor substrate 1. Subsequently, as shown in <figref idref="f0012">Fig. 18</figref>, the rear face 21 of the semiconductor substrate 1 is subjected to surface grinding and then chemical etching, whereby the semiconductor substrate 1 is thinned to the thickness of 50 µm. As a consequence, i.e., because of the grinding of the rear face 21 of the semiconductor substrate 1, the rear face 21 reaches the fractures 15 generated from the starting point region for cutting acting as a start point, whereby the semiconductor substrate 1 is divided into semiconductor chips 25 having the respective functional devices 19. Examples of the chemical etching include wet etching (HF·HNO3) and plasma etching (HBr·Cl2).

0062Then, as shown in <figref idref="f0012">Fig. 19</figref>, an expansion film 23 is attached so as to cover the rear faces of all the semiconductor chips 25. Thereafter, as shown in <figref idref="f0013">Fig. 20</figref>, the protective film 20 attached so as to cover the functional devices of all the semiconductor chips 25 are peeled off. Subsequently, as shown in <figref idref="f0013">Fig. 21</figref>, the expansion film 23 is expanded, so that the semiconductor chips 25 are separated from each other, and a suction collet 27 picks up the semiconductor chips 25.

0063As explained in the foregoing, the substrate dividing method in accordance with Example 1 can grind the rear face 21 of the semiconductor substrate 1 after forming the functional devices 19 on the front face 3 of the semiconductor substrate 1 in the device manufacturing process. Also, because of the following effects respectively exhibited by the step of forming a starting point region for cutting and the step of grinding the semiconductor substrate, the semiconductor chips 25 thinned so as to respond to the smaller size of semiconductor devices can be obtained with a favorable yield.

0064Namely, the step of forming a starting point region for cutting can prevent unnecessary fractures and melting deviated from a desirable line along which the substrate should be cut for cutting the semiconductor substrate 1 from occurring, and thus can keep unnecessary fractures and melting from occurring in the semiconductor chips 25 obtained by dividing the semiconductor substrate 1.

0065The step of forming a starting point region for cutting does not melt the front face 3 of the semiconductor substrate 1 along the line along which the substrate should be cut, and thus can narrow the gap between the functional devices 19 adjacent each other, thereby making it possible to increase the number of semiconductor chips 25 separated from one semiconductor substrate 1.

0066On the other hand, the step of grinding the semiconductor substrate subjects the rear face 21 of the semiconductor substrate 1 to surface grinding such that the semiconductor substrate 1 attains a predetermined thickness after the starting point region for cutting is formed within the semiconductor substrate 1. Here, even if the rear face 21 reaches the fractures 15 generated from the starting point region for cutting acting as a start point, the cut surfaces of the semiconductor substrate 1 cut by the fractures 15 are in close contact with each other, whereby the semiconductor substrate 1 can be prevented from chipping and cracking because of the surface grinding. Therefore, the semiconductor substrate 1 can be made thinner and divided while preventing the chipping and cracking from occurring.

0067The close contact of the cut surfaces in the semiconductor substrate 1 is also effective in preventing the grinding dust caused by the surface grinding from entering the fractures 15, and keeping the semiconductor chips 25 obtained by dividing the semiconductor substrate 1 from being contaminated with the grinding dust. Similarly, the close contact of the cut surfaces in the semiconductor substrate 1 is effective in reducing the chip-off of the semiconductor chips 25 caused by the surface grinding as compared with the case where the semiconductor chips 25 are separated from each other. Namely, as the protective film 20, one with a low holding power can be used.

0068Since the rear face 21 of the semiconductor substrate 1 is subjected to chemical etching, the rear faces of the semiconductor chips 25 obtained by dividing the semiconductor substrate 1 can be made smoother. Further, since the cut surfaces of the semiconductor substrate 1 caused by the fractures 15 generated from the starting point region for cutting acting as a start point are in close contact with each other, only edge parts of the cut surfaces on the rear face side are selectively etched as shown in <figref idref="f0014">Fig. 22</figref>, whereby chamfers 29 are formed. Therefore, the transverse rupture strength of the semiconductor chips 25 obtained by dividing the semiconductor substrate 1 can be improved, and the chipping and cracking in the semiconductor chips 25 can be prevented from occurring.

0069The relationship between the semiconductor chip 25 and the molten processed region 13 after the step of grinding the semiconductor substrate includes those shown in <figref idref="f0015 f0016">Figs. 23A to 25B</figref>. The semiconductor chips 25 shown in these drawings have their respective effects explained later, and thus can be used according to various purposes. <figref idref="f0015">Figs. 23A, 24A</figref>, and <figref idref="f0016">25A</figref> show the case where the fracture 15 reaches the front face 3 of the semiconductor substrate 1 before the step of grinding the semiconductor substrate, whereas <figref idref="f0015">Figs. 23B, 24B</figref>, and <figref idref="f0016">25B</figref> show the case where the fracture 15 does not reach the front face 3 of the semiconductor substrate 1 before the step of grinding the semiconductor substrate. Even in the case of <figref idref="f0015">Figs. 23B, 24B</figref>, and <figref idref="f0016">25B</figref>, the fracture 15 reaches the front face 3 of the semiconductor substrate 1 after the step of grinding the semiconductor substrate.

0070In the semiconductor chip 25 having the molten processed region 13 remaining within the cut surface as shown in <figref idref="f0015">Figs. 23A and 23B</figref>, the cut surface is protected by the molten processed region 13, whereby the transverse rupture strength of the semiconductor chip 25 improves.

0071The semiconductor chip 25 in which the molten processed region 13 does not remain within the cut surface as shown in <figref idref="f0015">Figs. 24A and 24B</figref> is effective in the case where the molten processed region 13 does not favorably influence the semiconductor device.

0072In the semiconductor chip 25 in which the molten processed region 13 remains in an edge part on the rear face side of the cut surface as shown in <figref idref="f0016">Figs. 25A and 25B</figref>, the edge part is protected by the molten processed region 13, whereby the chipping and cracking in the edge part can be prevented from occurring as in the case where the edge part of the semiconductor chip 25 is chamfered.

0073The rectilinearity of the cut surface obtained after the step of grinding the semiconductor substrate improves more in the case where the fracture 15 does not reach the front face 3 of the semiconductor substrate 1 before the step of grinding the semiconductor substrate as shown in <figref idref="f0015">Figs. 23B, 24B</figref>, and <figref idref="f0016">25B</figref> than in the case where the fracture 15 reaches the front face 3 of the semiconductor substrate 1 before the step of grinding the semiconductor substrate as shown in <figref idref="f0015">Figs. 23A, 24A</figref>, and <figref idref="f0016">25A</figref>.

0074Whether the fracture reaches the front face 3 of the semiconductor substrate 1 or not depends on not only the depth of the molten processed region 13 from the front face 3, but also the size of the molten processed region 13. Namely, when the molten processed region 13 is made smaller, the fracture 15 does not reach the front face 3 of the semiconductor substrate 1 even if the depth of the molten processed region 13 from the front face 3 is small. The size of the molten processed region 13 can be controlled by the output of the pulse laser light in the step of forming a starting point region for cutting, for example, and becomes greater and smaller as the output of the pulse laser light is higher and lower, respectively.

0075In view of a predetermined thickness of the semiconductor substrate 1 thinned in the step of grinding the semiconductor substrate, it is preferred that marginal parts (outer peripheral parts) of the semiconductor substrate 1 be rounded by at least the predetermined thickness by chamfering beforehand (e.g., before the step of forming a starting point region for cutting). <figref idref="f0016">Figs. 26A and 26B</figref> are respective sectional views of a marginal part of the semiconductor substrate 1 before and after the step of grinding the semiconductor substrate in accordance with Example 1. The thickness of the semiconductor 1 shown in <figref idref="f0016">Fig. 26A</figref> before the step of grinding the semiconductor substrate is 350 µm, whereas the thickness of the semiconductor 1 shown in <figref idref="f0016">Fig. 26B</figref> after the step of grinding the semiconductor substrate is 50 µm. As shown in <figref idref="f0016">Fig. 26A</figref>, a plurality of (seven here) rounded portions are formed at the marginal part of the semiconductor substrate 1 beforehand by chamfering with a thickness of 50 µm each, i.e., the marginal part of the semiconductor substrate 1 is caused to have a wavy form. As a consequence, the marginal part of the semiconductor substrate 1 after the step of grinding the semiconductor substrate 1 attains a state rounded by chamfering as shown in <figref idref="f0016">Fig. 26B</figref>, whereby the chipping and cracking can be prevented from occurring at the marginal part, and handling can be made easier because of an improvement in mechanical strength.

Industrial Applicability

0076As explained in the foregoing, the present invention can thin and divide the substrate while preventing the chipping and cracking from occurring.