The present invention concerns a method for preventing crack propagation in brittle materials.

PRIOR AND RELATED ART

Silicon based photovoltaic devices are typically fabricated on relatively large wafers that are very thin. The wafers are necessarily thin because of the high cost of solar grade silicon. Despite the high strength of silicon, thin silicon wafers are easily broken because silicon is a brittle material and device manufacturing introduces defects. Defects can act as stress concentration points. When stress is concentrated to the extent that a crack can propagate, the result is a broken wafer.

A large number, hundreds of millions, of wafers are used annually. Hence, a few percent of broken wafers is also a large number, which have a significant cost. Because of the high cost of yield loss due to broken wafers there are efforts to eliminate the root cause of the cracks and to minimize the stress to which wafers are subjected in order to limit crack propagation. There have been no known efforts to modify the existing internal stresses in wafers directly.

There is no known prior technology to stop crack propagation by modifying the internal stress in silicon wafers.

SUMMARY OF THE INVENTION

The invention concerns a method for preventing crack propagation in an object made of brittle material, comprising the steps of - heating a selected region of the object to induce plastic deformation, - cooling the selected region sufficiently fast to retain the deformation.

The heating and cooling of selected regions in an object made of a brittle material so that it deforms and does not return to its original shape upon cooling changes the internal stress in the material. It must be understood that cooling does not have to be active. Cooling in ambient conditions can be sufficiently fast to retain the deformation. If the new internal stress field is such that there is compressive stress perpendicular to the direction of crack propagation there is less probability that a crack will propagate and less probability that the object will break.

One such region of compressive stress can be put in front of a known propagating crack. Alternatively, several regions of compressive stress may be located in an arced pattern in front of a propagating crack. Repeating the heating and cooling of selected regions in a predetermined pattern is a statistical approach that may increase the probability for a propagating crack to encounter such a region of compressive stress, and hence reduces the probability of a break. Applying this concept to silicon wafers, it is possible to form a cylinder, which is in tension, through the wafer. In reaction to the cylindrical volume in tension there will be a cylindrically symmetric zone, around the volume or column in tension, which has tangential compression. Creating pre-stressed regions in a silicon wafer can prevent or reduce crack propagation and hence reduce the probability that the wafer will break.

One method of heating selected cylindrical volumes in a silicon wafer is with a laser. The wave length and power density of the laser beam must be carefully selected so that the silicon is heated to the point that the silicon undergoes plastic deformation without significant melting or ablating.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention will be better understood from the following detailed description with reference to the accompanying drawings, in which like numerals refer to like parts, and in which:

Figure 1 shows a volume of heated silicon in a preferred embodiment of the invention.

Figure 2 is a cross section of a wafer with a cylindrical zone after heating.

Figure 3 illustrates the stress relationship - the heated column which plastically deforms when hot is under tension after cooling. The surrounding zone is in compression. Figures 4a and 4b illustrates crack termination at single spot.

Figure 5 illustrates crack termination by a group of deformed spots

Figure 6 shows a multiple spot pattern to prevent crack propagation over a larger area.

DETAILED DESCRIPTION OF THE INVENTION

Laser processing of silicon wafers has nearly exclusively used wavelengths in the visible spectrum, which is strongly absorbed by silicon. This is because to scribe or cut silicon, ablation is desirable. However, it is possible to heat silicon without ablating material using near infrared lasers. Silicon has an indirect band gap that corresponds to about 1100nm wavelength light. Because the band gap is indirect, the absorption of light at wavelengths slightly shorter than 1100nm is relatively weak. The majority of light passes through a significant depth of silicon. Hence a laser beam of infrared light can pass through a silicon wafer that is on the order of 200 microns thick.

By using a photon energy slightly above the indirect bandgap energy, that is for silicon in the example above using light having a wavelength slightly shorter than 1100nm, it is thus possible to achieve light absorption that is great enough to heat the material, but not so high that it cannot penetrate through the material. Light should pass through the material and heat it all the way through.

For semiconductors with direct bandgaps the photon energy may correspond to the direct bandgap energy. However, for some direct bandgap semiconductors the absorption may rise so sharply that it will not, in practice, be possible to achieve the desired effect. The laser can be either continuous or pulsed. For silicon, the best choice is a continuous infrared laser. A continuous IR laser can be directed to the selected spot for a short time. The time required depends on the spot size and energy density. For a 200 micron diameter spot, the following approximate time and energy are required:

The total energy required increases with time because of heat loss due to conduction and radiation. It must be understood that the spots may have another width than 200 micrometers, and does not need to be circular. Thus, the columnar volumes are not necessarily cylindrical. Further, the main direction of the columns may be inclined at some angle other than 90° with respect to the surface 3, i.e. the columns does not necessarily have to be perpendicular to the surface 3.

Figure 1 shows a silicon wafer 1 with a volume 2 of heated silicon, i.e. a cylinder or column through the wafer which is heated nearly uniformly. Because the absorption coefficient for the selected wavelength of light is relatively small, the intensity of the light is significant through the entire thickness of the wafer.

When a laser beam of cylindrical symmetry impinges on a silicon wafer 1 at normal incident, i.e. substantially perpendicular to the illuminated surface 3, it heats a column 2 through the wafer. When the column 2 is heated to the point that the material plastically deforms in response to the pressure caused by local heating and thermal expansion, then the laser is turned off and the silicon wafer 1 allowed too cool. It must be understood that cooling does not have to be active. Cooling in ambient conditions can be sufficiently fast to retain the deformation. The heated volume in the wafer would contract on cooling but it is constrained by the surrounding silicon. The heated column 2 of silicon has plastically deformed when it was hot to relieve internal stress. As it cools, the stress builds up again, but because of the lower temperature there is less plastic deformation and the stress remains in the silicon. The silicon wafer 1 is left with a columnar volume 2 which is in tension. See Fig 3. In response to the tension the silicon wafer has a larger column 4, surrounding the column 2 in tension, which has a component of compression. This larger columnar volume 4 has compressive stress in the tangential direction. This stress field can be cylindrically symmetric, as shown in fig 3.

Figure 2 is a cross section of a wafer 1 having a cylindrical zone 2 after heating. Because of thermal expansion and plastic deformation the top and bottom of the heated cylinder extends above and below the original surfaces.

Figure 3 illustrates the predominant stress relationships. The heated column

2, which was plastically deformed when hot, is now under tension after cooling. More precisely, the cylindrical volume 2 forms a region having a predominantly radial tensile stress after cooling. The substantially radial tensile stress field 3 is indicated entirely within region 2, but in reality there will be no clearly defined radius, such as the one indicated by the dash-dot line 2, inside of which there will be only radial tensile stress. Rather, the predominantly radial stress field induced by heating and cooling will extend some distance outside the columnar volume indicated by circles 2 in the figures. The surrounding zone 4 is substantially in compression, as illustrated by arrows in the tangential direction. That is, numeral 4 refers to a surrounding zone with predominantly tangential compressive stress. Due to inhomogenities in the material, non-uniform heating and/or cooling, there may be smaller zones with a predominant tensile stress in the tangential direction. This will be the exception rather than the rule.

Figure 4a illustrates a primary mechanism for crack termination at a single spot. Cracks propagate when there is a predominant component of tensile stress perpendicular to the crack. Similarly, cracks tend to terminate when the predominant tensile component is parallel to the crack. In figure 4a, an incoming crack 5 approaches the columnar volume or spot 2 from a substantially radial direction. Once it reaches the region 7 where is runs parallel to the radial stress field 3 indicated by a single-headed arrow, there will not be sufficient tensile stress perpendicular to the crack to keep tearing it open or promote crack propagation. A crack 5 entering a region of predominantly radial tensile stress from a substantially radial direction will thus tend to terminate at a mean radius indicated by the dotted line 6, which in general will differ from the deformed zone 2. Such a mean radius 6 is defined by the mean value theorem even if the spot 2 has a non-circular form. We emphasize that the figures are drawn to illustrate crack termination in a clear manner, and thus that the relationship between spot 2 and mean radius 6 is not necessarily to scale.

Due to the inhomogenities and non-uniform heating or cooling, the tensile stress field 3 will be more or less inhomogeneous. That is, it may comprise tangential components in addition to the radial components as shown by the single-headed arrows in figure 4b. Figure 4b shows a crack 5 propagating near a spot 2 in a substantially tangential direction. The substantially radial tensile stress on the left hand side of figure 4b tends to "tear" the crack open. Hence the crack will propagate in a substantially tangential direction around the spot 2 as illustrated. Once the crack 5 reaches the region 7, where it runs parallel to the predominant tensile stress, there is no substantial stress component tearing the crack open, and the crack 5 will terminate. This is also illustrated by a compression field indicated by a double-headed arrow perpendicular to the crack 5 in region 7. As the inhomogenities of the material will be randomly distributed, zones of predominantly tangential tensile stress will be located at various radii from the center of a deformed spot. By the mean value theorem, however, there will be a mean expected radius around a heated spot of any shape that will tend to inhibit crack propagation. This mean radius is illustrated by the dotted line 6 in figures 4a and 4b. It is determined by the radius at which the radial tensile field of figure 3 starts to dominate, by the inhomogenities in the material, and/or the temperature changes during heating and cooling of he columnar volume illustrated by spots 2.

From the discussion in connection with figure 4a above, it follows that heating, e.g. by laser, and creating a compression zone directly in line with a crack, as shown in figure 4a, inhibits further propagation. If the location of a crack is known, one laser heating process can be applied to create a deformed region 2 with radial tensile stress surrounded by a compression zone 4 directly in line with the crack. Due to inhomogeneous stress fields, such spots may also, at least to some extent, prevent propagation of cracks entering a region near a heated column from a substantially tangential direction as illustrated in figure 4b.

In both cases, crack propagation will tend to stop at or near a mean radius 6 around a deformed spot 2. In other words, a crack 5 terminates when it runs into the compression zone 4 around a spot 2.

In practice, wafers break during handling or processing when there is some form of applied stress. The combination of applied stress and internal stress determine the magnitude and direction of stress at a particular point at a particular time. When the internal compressive stress in the tangential direction is larger than the tangential component of an applied stress, then a crack will not propagate in the radial (perpendicular) direction.

Figure 5 shows an example of crack termination by a group of deformed spots 2. When the location or direction of a crack 5 is not known precisely, or if the crack direction is likely to change because of non-uniformities in the silicon wafer, several compression zones 4 can be located in an arc pattern centred on the assumed location of the crack tip. The compression zones 4 are provided by heating and cooling cylindrical or non-cylindrical columns 2 as discussed above.

As is clear from figure 6, identification of cracks is not required. The two previously described applications require identification of crack location and orientation, as the inhomogenities required to obtain the effect illustrated in figure 4b cannot always be relied upon in practice. It is possible to forgo the crack identification process. Creation of compression zones can also be used for general crack-propagation prevention. This is a statistical approach. The more compression zones created, the higher the probability that one of them will stop a random crack from propagating. Any pattern can be selected for the location of compression zones. Patterns can range from closest packing to random.

The radius of the tension cylinder can vary from a few 10's of microns to several hundred microns depending on the laser source and optics. The radius is typically on the order of 200 microns. The radius of the cylindrical compression zone, where the compressive stress is significant, is not more than two times the radius of the tension cylinder. Hence, choosing an average spot size larger than 200 microns may be advantageous for the purpose of crack inhibition.

A full closest pack pattern requires a prohibitively large number of compression zones. However, multi-crystalline silicon wafers rarely break in an exactly straight line. It is possible to make a pattern of compression zones that have space between the zone, where cracks will propagate, but still have a high probability that the crack will run into a compression zone within a short distance.

Both mono-crystalline silicon wafers and multi-crystalline silicon wafers can be treated with compression zones. The larger number of stress concentrating defects makes multi-crystalline silicon wafers the prime candidate for treatment with compression zones. However, saw damage on non-polished mono-crystalline silicon wafers provide enough micro-cracks to make compression zone treatment valuable for mono-crystalline silicon wafers also.

In another aspect, the invention concerns an apparatus for carrying out the method above. The apparatus comprises a heating device (100) capable of heating a selected region of the object to induce plastic deformation. In a preferred embodiment the heating device is a laser emitting light with a wavelength slightly shorter than a bandgap of the material. For the indirect bandgap of silicon described above, the laser may be e.g. an Nd:YAG laser emitting light with 1064nm, although other lasers, pulsed or continuous, are contemplated.

A motor proves a relative movement between the heating device and the object. This motor may move a table with the silicon wafer underneath a fixed laser, or the laser may be moved around illuminating points on the surface of the wafer.

A conveyor moves the object to and/or from the table.