Wafer dicing process

Abstract

The present invention discloses a wafer scribing process, comprising: expanding the width of the virtual copper wires in the saw street and continuously filling them to form a shear buffer copper strip; performing double-pulse laser annealing on the shear buffer copper strip to generate a twinned softening layer; performing back infrared gradient preheating based on the thermoplastic transition temperature of the copper strip and correcting the tool feed rate and rotation speed in combination with vibration data coupling to obtain a plastic initial saw kerf; performing layered progressive deep cutting according to multiple cutting depth data, so that the copper strip slides under the pulse load to absorb energy and elastically resets through the low dielectric layer to form a full-depth saw kerf; performing constant temperature crystal locking and gradient cooling in a nitrogen environment to stabilize the twin structure of the copper strip and interlock it with the dielectric layer shrinkage hoop, thereby generating a flexible bridging unit on the sidewall of the saw kerf. The technical solution of the present invention can reduce the peak shear stress between the shunt layers at the moment of cutting, suppress the selective delamination of the multi-layer interconnection interface, and improve the yield rate and long-term reliability of ultra-thin wafers.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor device manufacturing, and in particular to a wafer dicing process. Background Art

[0002] Wafer scribing is a critical back-end process in semiconductor device manufacturing. Its purpose is to separate integrated circuit chips, manufactured in batches on a single wafer, from the wafer substrate to form independent device units. This process typically uses a diamond blade to mechanically cut along pre-set scribing lines, or a laser beam to thermally cut, dividing the wafer into hundreds to thousands of precisely dimensioned chips with intact edges. With the advancement of semiconductor technology, modern wafer scribing requires not only ensuring the geometric accuracy and surface quality of the chips, but also ensuring the structural integrity and electrical stability of complex multi-layer structures during the separation process.

[0003] Modern semiconductor devices typically contain multiple metal wiring layers separated by low-dielectric-constant materials, forming a complex, heterogeneous multilayer material system. During the dicing process, the mechanical tool's entry generates complex stress wave propagation within the wafer. Due to significant differences in the elastic modulus, thickness, and wiring density of different metal layers, the stress transmission and response at the interfaces between layers are highly non-uniform. Furthermore, when stress waves propagate between multiple metal layer interfaces, a superposition effect occurs, causing shear stresses at certain specific interfaces to far exceed the stress levels of a single layer of material. This stress accumulation process is selective, meaning that stress concentration tends to occur at interfaces with the greatest material property differences, while other interfaces may remain within safe stress ranges. Due to this interlayer shear stress accumulation effect, even if the surface dicing quality appears good, some metal layer interfaces within the device may have already undergone microscopic separation or damage. These hidden defects can lead to premature device failure during subsequent reliability testing, seriously affecting the long-term reliability of the product. Summary of the Invention

[0004] The main purpose of the present invention is to solve the technical problem that the existing wafer dicing process has interlayer shear stress accumulation on the multi-layer metal wiring wafer and induces interface delamination failure.

[0005] A first aspect of the present invention provides a wafer scribing process, the wafer scribing process comprising:

[0006] The dummy copper wires in the saw street area are subjected to line width expansion processing, and the expanded dummy copper wires are continuously filled to form a shear buffer copper strip located between the low-k dielectric layer and the passivation layer and running through the multi-layer interconnect structure along a preset scribing direction;

[0007] Performing a double-pulse laser annealing treatment on the surface of the shear buffer copper strip, inducing the formation of a twin structure inside the shear buffer copper strip by controlling the pulse duration and energy density, so that the yield strength of the shear buffer copper strip is lower than the yield strength of adjacent copper interconnects;

[0008] performing infrared preheating on the back side of the wafer according to the thermoplastic transition temperature of the shear buffer copper strip to cause the shear buffer copper strip to enter a plastic deformation state, and adjusting the tool feed rate according to the obtained saw street area vibration data to form an initial cutting saw kerf;

[0009] The wafer is subjected to a layered progressive deep cutting process according to the initial cutting saw kerf, the shear buffer copper tape is subjected to slip deformation under the action of the tool pulse load and reset by the elastic action of the low dielectric constant dielectric layer, thereby forming a full-depth cutting saw kerf;

[0010] The full-depth cutting saw seam is subjected to a nitrogen environment cooling treatment, so that the twin structure inside the shear buffer copper strip is stabilized and forms a mechanical constraint structure with the adjacent dielectric layer, forming a bridging unit on the side wall of the saw seam.

[0011] Preferably, the method of performing line width expansion processing on the dummy copper wires in the saw street area and continuously filling the expanded dummy copper wires to form a shear buffer copper tape located between the low-k dielectric layer and the passivation layer and penetrating the multi-layer interconnect structure along a preset scribing direction includes:

[0012] Performing line width redefinition processing on the interconnection layout according to the saw street dicing coordinates to generate mask data of an amplified area that is continuous along the dicing direction and has a width greater than that of the original dummy copper wire;

[0013] Performing trench etching on a dummy copper wire in a low-k dielectric layer according to the amplification region mask data to obtain an amplification groove having parallel sidewalls and a depth consistent with the thickness of the dielectric layer;

[0014] Depositing a conductive copper seed layer on the amplification tank to obtain a continuous seed layer that covers the tank inner wall and is conductive with the tank bottom;

[0015] The amplification tank body is electroplated and filled according to the continuous seed layer, and then the filled copper body is chemically mechanically planarized to obtain a shear buffer copper strip whose top surface is flush with the upper surface of the low-k dielectric layer.

[0016] Preferably, the double-pulse laser annealing treatment is performed on the surface of the shear buffer copper strip, and a twin structure is induced inside the shear buffer copper strip by controlling the pulse duration and energy density so that the yield strength of the shear buffer copper strip is lower than the yield strength of adjacent copper interconnects, including:

[0017] The surface of the shear buffer copper strip is subjected to a first pulse thermal shock treatment. The first pulse duration and energy density are adjusted according to the thickness of the shear buffer copper strip and the thermal diffusivity of the low-k dielectric layer, thereby forming a semi-molten layer along the surface of the shear buffer copper strip.

[0018] Performing a sub-pulse laser shock treatment on the surface of the shear buffer copper strip according to the thermal platform of the semi-molten layer to obtain a twin nucleation network located inside the semi-molten layer and continuously distributed laterally;

[0019] Performing a heat-maintaining expansion treatment on the interior of the shear buffer copper strip according to the twin nucleation network to obtain a twin expansion band that penetrates along the thickness direction of the shear buffer copper strip;

[0020] The shear buffer copper strip is subjected to a stepwise cooling and solidification treatment according to the twin extension band to obtain a softened shear buffer copper strip having a yield strength lower than that of adjacent copper interconnects.

[0021] Preferably, the infrared preheating treatment is performed on the back side of the wafer according to the thermoplastic transition temperature of the shear buffer copper strip to make the shear buffer copper strip enter a plastic deformation state, and the tool feed rate is adjusted according to the obtained saw street area vibration data to form an initial cutting saw kerf, including:

[0022] The infrared irradiation power is gradient-set according to the thermoplastic transition temperature of the shear buffer copper strip and the lateral width of the saw street, and the back power distribution data is obtained. The power corresponding to the back power distribution data at the center of the saw street is higher than the power corresponding to the edge of the saw street.

[0023] Performing a zoned infrared radiation heating process on the back side of the wafer according to the back side power distribution data to obtain a plastic preheating zone located at the center of the saw street and having a temperature within a range of ±5 degrees Celsius from the thermoplastic transition temperature;

[0024] performing coupling correction processing on the tool feed rate and the tool rotation speed according to the vibration amplitude data and the vibration main frequency data obtained in the plastic preheating zone to obtain a first feed rate and a first rotation speed;

[0025] A shallow scratching process is performed on the tool according to the first feed rate and the first rotation speed to obtain an initial cutting saw kerf that deeply covers the outer layer copper interconnection.

[0026] Preferably, the infrared irradiation power is gradient-set according to the thermoplastic transition temperature of the shear buffer copper strip and the lateral width of the saw street to obtain backside power distribution data, including:

[0027] The thermoplastic transition temperature of the shear buffer copper strip is divided into temperature intervals to obtain the center target temperature and the edge target temperature;

[0028] Performing interpolation calculation on the center target temperature and the edge target temperature according to the transverse width of the saw street to obtain a continuous temperature gradient curve;

[0029] Power mapping processing is performed on the infrared radiation power according to the temperature gradient curve to obtain back surface power distribution data.

[0030] Preferably, the coupling correction processing of the tool feed rate and the tool rotation speed is performed based on the vibration amplitude data and the vibration main frequency data obtained in the plastic preheating zone to obtain the first feed rate and the first rotation speed, including:

[0031] performing amplitude-threshold comparison processing on the vibration amplitude data obtained in the plastic preheating zone to obtain a feed rate adjustment coefficient;

[0032] Performing main frequency-harmonic deviation calculation processing on the vibration main frequency data obtained in the plastic preheating zone to obtain a speed adjustment coefficient;

[0033] A joint correction process is performed on the original feed rate and the original speed according to the feed rate adjustment coefficient and the speed adjustment coefficient to obtain a first feed rate and a first speed.

[0034] Preferably, the wafer is subjected to a layered progressive deep cutting process according to the initial cutting saw kerf, the shear buffer copper tape is subjected to slip deformation under the action of the tool pulse load and is reset by the elastic action of the low dielectric constant dielectric layer to form a full-depth cutting saw kerf, comprising:

[0035] Performing layered cutting depth planning processing on the initial cutting saw kerf depth data to obtain multiple cutting depth data that increase in sequence along the saw street direction and correspond to the combination of the thickness of the metal layer and the dielectric layer;

[0036] Performing a first-level deep cutting process on the wafer according to the multiple-level cutting depth data to obtain first-level saw kerf displacement data that generates a sliding displacement in the transverse direction of the shear buffer copper strip;

[0037] A rebound buffer process is performed on the wafer based on the first-level saw kerf displacement data, and the shear buffer copper tape is reset by utilizing the elastic effect of the low-k dielectric layer. Meanwhile, a deviation correction is performed on the next-level cutting depth data to obtain the corrected cutting depth data.

[0038] The subsequent deep cutting process and rebound buffer process are cyclically executed according to the corrected cutting depth data until the silicon base surface is cut deep enough to obtain a full-depth cutting saw kerf.

[0039] Preferably, the layered cutting depth planning processing is performed on the initial cutting saw kerf depth data to obtain multiple cutting depth data that increase in sequence along the saw street direction and correspond to the combination of the thickness of the metal layer and the dielectric layer, including:

[0040] Performing layer thickness accumulation processing based on the metal layer thickness data and the dielectric layer thickness data to obtain a multi-layer stack cumulative depth sequence;

[0041] Performing energy-depth matching processing on tool single pulse energy constraint data according to the multi-layer stack cumulative depth sequence to obtain a cutting depth increment sequence;

[0042] The initial cutting kerf depth data is subjected to depth segmentation processing according to the cutting depth increment sequence to obtain multiple levels of cutting depth data.

[0043] Preferably, the full-depth cutting saw kerf is subjected to a nitrogen environment cooling treatment, the twin structure inside the shear buffer copper strip is stabilized and forms a mechanical constraint structure with the adjacent dielectric layer, and a bridging unit is formed on the sidewall of the saw kerf, comprising:

[0044] According to the difference between the shear buffer copper strip twin stable temperature zone and the linear thermal expansion coefficient of the low dielectric constant dielectric layer, a double window setting process is performed on the nitrogen cooling stage to obtain the first constant temperature window parameter and the second temperature reduction gradient parameter;

[0045] performing a constant temperature dwell process on the full-depth cutting saw kerf according to the first constant temperature window parameters, so that the twin interfaces inside the shear buffer copper strip are rearranged during the constant temperature dwell period and axial residual compressive stress data of the shear buffer copper strip is obtained;

[0046] performing radial gradient cooling on the full-depth saw kerf according to the second temperature-lowering gradient parameter, so as to cause the shear buffer copper strip to undergo reversible lateral contraction and drive the adjacent low-k dielectric layer to produce nested rebound, thereby obtaining a coupled shrinkage hoop interface;

[0047] A stress redistribution process is performed on the full-depth cutting saw seam based on the coupled shrinkage interface and the axial residual compressive stress data. During the stress redistribution process, the residual compressive stress of the shear buffer copper strip and the rebound elasticity of the low dielectric constant dielectric layer are used to achieve interlocking, forming a bridging unit that continuously extends along the side wall of the saw seam.

[0048] Preferably, the step of performing radial gradient cooling on the full-depth saw kerf according to the second temperature-lowering gradient parameter to cause the shear buffer copper strip to undergo reversible lateral contraction and drive the adjacent low-k dielectric layer to produce nested rebound, thereby obtaining a coupled shrinkage interface, comprises:

[0049] Performing center-edge partitioning and setting processing on the nitrogen flow rate according to the second cooling gradient parameter to obtain center flow rate data and edge flow rate data;

[0050] performing zoned nitrogen supply processing according to the central flow data and the edge flow data to obtain radial temperature gradient data;

[0051] According to the radial temperature gradient data, a synchronous matching process is performed on the lateral shrinkage of the shear buffer copper strip and the rebound of the low dielectric constant dielectric layer to obtain a coupled shrinkage hoop interface.

[0052] This approach begins with the structural transformation of the dummy copper conductors within the saw street area. Originally serving only as filler, these dummy copper conductors are transformed into shear buffer copper strips with specific geometric characteristics through line width expansion and continuous filling processes. This copper strip is located between the low-k dielectric layer and the passivation layer, extending throughout the multilayer interconnect structure along the predetermined dicing direction. This spatial arrangement ensures that the buffer copper strip can intercept stress waves propagating from the surface to the deeper layers, while its continuous geometry ensures complete and consistent stress interception. The key function of the shear buffer copper strip lies in its controllable yield properties. Through a double-pulse laser annealing treatment, a specific twin structure is induced within the copper strip. Twin boundaries, acting as crystal defects, provide preferential slip paths for dislocation motion, significantly reducing the material's yield strength. When the yield strength of the shear buffer copper strip is lower than that of the adjacent copper interconnects, the strip will preferentially undergo plastic deformation under applied stress, while the surrounding functional copper interconnects remain elastic. This differentiated mechanical response is the physical basis for achieving selective stress absorption. The preheat treatment utilizes the thermoplastic transition properties of the laser-modified shear buffer copper strip, forcing it into a plastic deformation state through infrared heating. The thermoplastic transition temperature is the critical point at which a material transitions from elastic to plastic behavior. At this temperature, the material's resistance to dislocation movement is significantly reduced, significantly reducing its deformation resistance. Simultaneously, real-time acquisition of vibration data provides feedback for dynamic adjustment of cutting parameters. When abnormal vibration is detected, reducing the tool feed rate prolongs the stress action time, providing a sufficient time window for the plastic response of the buffer copper strip, ensuring that stress can be dissipated promptly rather than propagating deeper into the metal layer. Each time the tool reaches a new metal layer interface, the pre-placed shear buffer copper strip undergoes slip deformation under the pulsed load. During this slip deformation process, the elastic strain energy of the shear stress is converted into plastic deformation work and heat energy, achieving irreversible energy dissipation. The elastic action of the low-k dielectric layer causes the buffer copper strip to partially reset after the load is removed. This reset mechanism maintains the geometric integrity of the buffer structure, preparing it for stress absorption by the next layer. Through this layer-by-layer processing method, the shear stress that would have accumulated and superimposed between the multi-layer interfaces is dissipated segment by segment, and the actual stress borne by each interface is far below the material's damage threshold. The cooling treatment stage completes the transition from temporary buffering to permanent protection. During the cooling process in the nitrogen environment, the twin structure inside the shear buffer copper tape undergoes a stable rearrangement, and the dislocation density and distribution tend to a balanced state. At the same time, the thermal shrinkage of the material causes the buffer copper tape to form a close mechanical constraint relationship with the adjacent dielectric layer. This constraint structure forms a continuously extended bridging unit on the sidewall of the saw seam, providing continuous buffering protection for the thermomechanical stress that may occur during subsequent packaging processes and device use. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying any creative work.

[0054] Figure 1 A schematic diagram of a wafer dicing process according to an embodiment of the present invention;

[0055] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0057] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, movement status, etc. between the various components under a certain specific posture. If the specific posture changes, the directional indications will also change accordingly.

[0058] In addition, the descriptions of "first", "second", etc. in the present invention are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, "and / or" in the full text includes three solutions. Taking A and / or B as an example, it includes technical solution A, technical solution B, and technical solution that satisfies both A and B. In addition, the technical solutions between the various embodiments can be combined with each other, and must be based on the ability of ordinary technicians in this field to implement. When the combination of technical solutions is mutually contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0059] An embodiment of the present application provides a wafer scribing process. Figure 1 A flow chart of a wafer scribing process according to an embodiment of the present application is provided. In this embodiment, the method includes:

[0060] See also Figure 1 , performing line width expansion processing on the dummy copper wires in the saw street area, and continuously filling the expanded dummy copper wires to form a shear buffer copper tape located between the low-k dielectric layer and the passivation layer and running through the multi-layer interconnect structure along a preset scribing direction;

[0061] In one embodiment of the present invention, the method of performing line width expansion processing on the dummy copper wires in the saw street area and continuously filling the expanded dummy copper wires to form a shear buffer copper tape located between the low-k dielectric layer and the passivation layer and penetrating the multi-layer interconnect structure along a preset scribing direction includes:

[0062] Performing line width redefinition processing on the interconnection layout according to the saw street dicing coordinates to generate mask data of an amplified area that is continuous along the dicing direction and has a width greater than that of the original dummy copper wire;

[0063] Performing trench etching on a dummy copper wire in a low-k dielectric layer according to the amplification region mask data to obtain an amplification groove having parallel sidewalls and a depth consistent with the thickness of the dielectric layer;

[0064] Depositing a conductive copper seed layer on the amplification tank to obtain a continuous seed layer that covers the tank inner wall and is conductive with the tank bottom;

[0065] The amplification tank body is electroplated and filled according to the continuous seed layer, and then the filled copper body is chemically mechanically planarized to obtain a shear buffer copper strip whose top surface is flush with the upper surface of the low-k dielectric layer.

[0066] The following is a detailed description of the steps involved in the above embodiment:

[0067] Saw street coordinates and layout redefinition are implemented using electronic design automation software to achieve geometric adjustments. Saw street coordinates refer to the predefined cutting path location information in the wafer layout design, including X- and Y-coordinates, as well as cutting direction information, stored as digital coordinates in the layout database. The interconnect layout is a design file that describes the geometric relationships of multiple layers of metal wiring, typically organizing the wiring information for different metal layers in a hierarchical manner. Line width redefinition involves digitally modifying the geometric dimensions of the original dummy copper conductors. In practice, dummy copper conductor objects within the saw street area are located in the layout editing software. The original width of these dummy copper conductors is typically the standard minimum line width. Using the software's geometry editing function, the target dummy copper conductor is selected and a width expansion operation is performed to increase the conductor width from its original size to the desired target size. The expansion process is symmetrical about the conductor's central axis. The augmentation area mask data is the photolithography mask production data generated after the width adjustment. This data records the precise boundary coordinates and geometric outline of the augmented copper conductor in a vector graphics format. The layout software outputs the modified geometric information in the standard data format required for mask making. For example, when expanding a dummy copper wire originally 2 microns wide to 6 microns, the software automatically calculates the new boundary coordinates and generates the corresponding mask pattern data. This line width expansion creates the necessary space for the subsequent formation of an effective stress buffer structure. The increased cross-sectional area of ​​the expanded wire provides greater deformation space and greater load-bearing capacity under stress.

[0068] The trench etching process uses reactive ion etching (RIE) technology to precisely remove material from the low-k dielectric layer. Low-k dielectric layers refer to insulating materials with a dielectric constant significantly lower than that of conventional silicon dioxide. These primarily include porous SiOCH and SiOF materials. These materials have a porous structure or contain organic groups, which are used to reduce parasitic capacitance between metal interconnects. The amplification area mask data is first transferred to the photoresist layer on the wafer surface through a photolithography process, forming an etch mask corresponding to the designed pattern. Trench etching is performed in a plasma etcher. Fluorine-containing chemical gases such as CF4 and CHF3 are introduced into the etching chamber, where they are excited by an RF electric field to form a plasma. The active fluorine atoms in the plasma chemically react with the dielectric material, while high-energy ions physically bombard the material surface. These two effects synergistically achieve selective material removal. The etching process utilizes an anisotropic mode, ensuring that the etching direction is perpendicular to the wafer surface by controlling the gas flow ratio and RF power. The amplified slot refers to the three-dimensional groove structure formed in the dielectric layer after etching is completed. Its cross-sectional shape is determined by the mask pattern and has steep and parallel sidewall profiles. The depth is consistent with the thickness of the dielectric layer, which means that the etching process accurately penetrates the current dielectric layer and stops when it contacts the surface of the underlying dummy copper conductor. The etching depth is monitored in real time using endpoint detection technologies such as plasma emission spectroscopy or laser interferometry. When the detection system recognizes the characteristic signal of the underlying copper material, the equipment automatically stops the etching process. Precise depth control ensures the consistency of the slot geometry, provides a standardized space container for subsequent copper material filling, and avoids damage to the underlying metal structure.

[0069] The conductive copper seed layer is deposited using magnetron sputtering technology to form a continuous metal film on the surface of the tank. This extremely thin copper film, with a thickness of tens of nanometers, provides a continuous conductive path for the subsequent electroplating process, ensuring that the plating current is evenly distributed across the entire surface of the tank. Deposition occurs in a magnetron sputtering system, where a high-purity copper target is mounted within a vacuum chamber, with the wafer placed on a substrate opposite the target. During deposition, argon gas is introduced into the vacuum chamber and an RF electric field is applied. The argon ions, accelerated by the electric field, bombard the copper target surface, sputtering copper atoms from the target and depositing them on the wafer. The wafer is slowly rotated during deposition to ensure that copper atoms reach the tank surface from different angles, achieving uniform coverage of the bottom and sidewalls. A continuous seed layer is defined as a continuous, uninterrupted, and hole-free copper film formed across the entire surface of the tank. Its continuity is verified by scanning electron microscopy. Conductivity with the bottom of the tank indicates good metallic contact between the deposited seed layer and the existing dummy copper conductors below, with contact resistance low enough to support current conduction during the subsequent electroplating process. During sputtering deposition, copper atoms have excellent surface migration capabilities, bypassing the edges of the tank and depositing onto the sidewalls and bottom surfaces, forming conformal coverage. A uniform and continuous seed layer is a prerequisite for high-quality electroplating fill. It ensures that the current density remains relatively uniform across the tank during the electroplating process, avoiding uneven fill or defect formation caused by localized current overshoot or undershoot.

[0070] Electroplating filling and chemical mechanical planarization process completes the precise formation of copper structure through the combination of electrochemical deposition and mechanical polishing. Electroplating filling process is carried out in electrochemical deposition equipment, where the wafer is immersed in an electrolyte containing copper sulfate as the cathode, and the copper anode provides the source of copper ions. The copper sulfate in the electrolyte provides Cu 2+ ions, Cu under the action of DC electric field 2+The ions are reduced to metallic copper on the wafer surface and deposited on the seed layer. The electroplating process uses pulsed or direct current (DC) mode, precisely controlling the current density to achieve bottom-up filling of the trench. This ensures that the copper material first nucleates and grows at the bottom of the trench and then gradually fills the entire trench space. A filled copper structure completely fills the expanded trench and slightly protrudes above the dielectric layer. Overfilling ensures that there are no voids or gaps within the trench. Chemical mechanical planarization (CMP) is performed in a CMP system, which combines chemical etching and mechanical polishing to remove excess copper. The CMP process uses a polishing slurry containing fine abrasive particles and an oxidizing agent. The wafer is placed in contact with a rotating polishing pad under controlled pressure. The oxidizing agent in the polishing slurry forms an easily removable oxide layer on the copper surface, while the abrasive particles mechanically remove the oxide layer and some copper material. The polishing process continues until all copper material protruding above the dielectric layer is completely removed. The shear buffer copper tape is the final copper structure with precise geometric dimensions. Its top surface forms a completely flush composite surface with the upper surface of the low-k dielectric layer, with surface flatness reaching nanometer-level precision. This flat surface provides an ideal working platform for subsequent laser processing steps, ensuring uniform distribution and absorption of laser energy. The expanded copper tape also has a larger cross-sectional area and volume, providing an ample material foundation for withstanding shear stress and undergoing plastic deformation.

[0071] Please continue reading Figure 1 , performing a double-pulse laser annealing treatment on the surface of the shear buffer copper strip, inducing the formation of a twin structure inside the shear buffer copper strip by controlling the pulse duration and energy density, so that the yield strength of the shear buffer copper strip is lower than the yield strength of the adjacent copper interconnects;

[0072] In one embodiment of the present invention, the double-pulse laser annealing treatment is performed on the surface of the shear buffer copper strip, and a twin structure is induced inside the shear buffer copper strip by controlling the pulse duration and energy density, so that the yield strength of the shear buffer copper strip is lower than the yield strength of adjacent copper interconnects, including:

[0073] The surface of the shear buffer copper strip is subjected to a first pulse thermal shock treatment. The first pulse duration and energy density are adjusted according to the thickness of the shear buffer copper strip and the thermal diffusivity of the low-k dielectric layer, thereby forming a semi-molten layer along the surface of the shear buffer copper strip.

[0074] Performing a sub-pulse laser shock treatment on the surface of the shear buffer copper strip according to the thermal platform of the semi-molten layer to obtain a twin nucleation network located inside the semi-molten layer and continuously distributed laterally;

[0075] Performing a heat-maintaining expansion treatment on the interior of the shear buffer copper strip according to the twin nucleation network to obtain a twin expansion band that penetrates along the thickness direction of the shear buffer copper strip;

[0076] The shear buffer copper strip is subjected to a stepwise cooling and solidification treatment according to the twin extension band to obtain a softened shear buffer copper strip having a yield strength lower than that of adjacent copper interconnects.

[0077] The following is a detailed description of the steps involved in the above embodiment:

[0078] First-pulse thermal shock treatment uses a pulsed laser to precisely input thermal energy into the surface of the shear buffer copper strip. This process involves instantaneously heating the copper strip's surface with a single laser pulse. Its purpose is to create a localized high-temperature region on the strip's surface without affecting the underlying structure. In practice, a solid-state laser with a wavelength of 532 nanometers is used as the heat source, and the laser beam is precisely aligned with the shear buffer copper strip's surface area via an optical focusing system. The pulse duration is adjusted based on the thickness of the shear buffer copper strip. Thinner strips use shorter pulse durations to prevent heat conduction to the bottom, while thicker strips require longer pulse durations to ensure sufficient surface heating. The thermal diffusivity of low-k dielectric layers refers to the dielectric material's ability to conduct heat, which affects the speed and range of laser heat diffusion around the copper strip. Energy density, the energy carried by the laser pulse per unit area, is precisely controlled by adjusting the laser power and spot size. For example, for a 500-nanometer-thick shear-buffered copper ribbon, the pulse duration was set to 50 nanoseconds, and the energy density was controlled at 0.2 joules per square centimeter, ensuring that the surface temperature of the copper ribbon instantly rose to near its melting point while the interior temperature remained relatively low. The semi-molten layer refers to a thin region on the surface of the copper ribbon that has partially melted due to laser heating. The copper atoms in this region have high mobility, but the overall structure has not yet fully liquefied. Precise regulation of pulse duration and energy density enables precise control of the heating depth and temperature gradient, avoiding structural damage caused by complete melting while creating the temperature and structural conditions ideal for twin formation.

[0079] Sub-pulse laser shock treatment (SLSP) utilizes the thermal plateau effect of the semi-molten layer to induce the formation of twin structures. The thermal plateau refers to a region of the semi-molten layer with uniform temperature distribution and high atomic mobility formed after laser heating. This region provides an ideal temperature and energy environment for subsequent structural modification. Sub-pulse laser shock treatment involves applying a second laser pulse immediately after the initial pulse, with a time interval controlled in the microsecond range to exploit the residual heat of the semi-molten layer. The energy density of the secondary pulse is set to 60-80% of that of the initial pulse, and the pulse duration is shortened to 20-30 nanoseconds, creating a strong thermal shock effect. Under the action of the secondary pulse, copper atoms within the semi-molten layer undergo rapid displacement and rearrangement. The high temperature gradient and rapid cooling conditions promote the formation of twin structures within the copper crystals. A twin nucleation network refers to the distribution of interconnected twin initiation sites within the semi-molten layer. These nucleation sites form a network across the lateral extent of the copper strip. This lateral continuity means that the twin nucleation sites are continuous across the width of the copper strip, forming a network structure that runs across the entire cross-section of the strip. For example, in a shear buffer copper ribbon with a width of 7 microns, the twin nucleation network is densely distributed with a spacing of approximately 500 nanometers, ensuring a foundation for twin modification across the entire cross-section. Double-pulse laser treatment establishes physical conditions favorable for twin formation on the copper ribbon's surface through precise control of the temperature field. The continuous distribution of the twin nucleation network provides a uniform germination foundation for subsequent structural expansion.

[0080] The thermal expansion treatment promotes the expansion of the twin structure deeper into the copper strip through temperature control. This process involves maintaining appropriate temperature conditions after the twin nucleation network is formed, allowing the twin structure to expand from the surface into the interior of the copper strip. In practice, an infrared heater gently heats the treatment area, maintaining the temperature within the range of 200-300 degrees Celsius—a temperature sufficient to activate the diffusion of copper atoms without destroying the already formed twin structure. The heating time is determined by the thickness of the shear buffer copper strip; a 500-nanometer-thick copper strip requires 10-15 minutes of heating to allow sufficient heat transfer to the bottom of the strip. During the thermal expansion process, the twin nucleation points on the surface serve as the starting points for structural expansion. Under thermal activation, the copper atoms undergo orderly arrangement and displacement along specific crystallographic directions. The twin expansion zone refers to a region of twin structure formed by the twin nucleation network extending deep into the copper strip. This zone is continuously distributed along the thickness of the shear buffer copper strip and ultimately extends through the entire thickness of the strip. Through-through means that the twin structure extends from the top surface of the copper strip to the interface with the underlying structure, forming a complete thickness-wise modification. For example, in a 500-nanometer-thick shear buffer copper strip, the twin extension zone begins at the surface and gradually expands downward, ultimately forming a modified region that extends throughout the entire 500-nanometer thickness. The temperature-controlled expansion process ensures the deep distribution and continuity of the twin structure, resulting in uniform modification of the mechanical properties throughout the shear buffer copper strip, avoiding the potential uneven performance issues that can arise from surface-only modification.

[0081] The step-down cooling and solidification treatment stabilizes the twin structure and ultimately determines the modification effect by controlling the cooling rate. The step-down cooling and solidification treatment refers to a process in which the twin structure inside the copper strip is gradually stabilized and solidified by lowering the temperature in stages. In specific implementation, the temperature is first slowly lowered from the holding temperature to 150 degrees Celsius and maintained for 5 minutes to initially release the stress near the twin boundary. The temperature is then further lowered to 100 degrees Celsius and maintained for 10 minutes to promote further stabilization of the twin structure. Finally, the temperature is lowered to room temperature at a rate of 10 degrees Celsius per minute, and the entire cooling process lasts about 30 minutes. The step-down cooling method avoids the thermal stress and structural defects that may be caused by rapid cooling, ensuring that the twin structure completes the final morphological fixation under stable conditions. The softened shear buffer copper strip refers to a copper structure that has undergone a complete twin modification treatment, which contains a large number of twin boundaries. These twin boundaries, as the preferred path for dislocation movement, significantly reduce the yield strength of the material. A lower yield strength than adjacent copper interconnects means that under the same applied stress, the softened shear buffer copper tape will preferentially undergo plastic deformation, while the surrounding standard copper interconnects remain elastic. For example, the yield strength of a standard copper interconnect is approximately 180 MPa, while the yield strength of the twinned shear buffer copper tape is reduced to 80-120 MPa, achieving a 40-60% strength reduction. The controlled cooling process ensures the stability of the twin structure and the durability of the modification effect, allowing the shear buffer copper tape to reliably perform its stress absorption function during the subsequent dicing process while avoiding structural integrity issues that could be caused by over-softening.

[0082] Please continue reading Figure 1 , performing infrared preheating on the back side of the wafer according to the thermoplastic transition temperature of the shear buffer copper strip, so that the shear buffer copper strip enters a plastic deformation state, and adjusting the tool feed rate according to the obtained saw street area vibration data to form an initial cutting saw kerf;

[0083] In one embodiment of the present invention, the back side of the wafer is subjected to infrared preheating treatment according to the thermoplastic transition temperature of the shear buffer copper strip to cause the shear buffer copper strip to enter a plastic deformation state, and the tool feed rate is adjusted according to the obtained saw street area vibration data to form an initial cutting saw kerf, including:

[0084] The infrared irradiation power is gradient-set according to the thermoplastic transition temperature of the shear buffer copper strip and the lateral width of the saw street, and the back power distribution data is obtained. The power corresponding to the back power distribution data at the center of the saw street is higher than the power corresponding to the edge of the saw street.

[0085] Performing a zoned infrared radiation heating process on the back side of the wafer according to the back side power distribution data to obtain a plastic preheating zone located at the center of the saw street and having a temperature within a range of ±5 degrees Celsius from the thermoplastic transition temperature;

[0086] performing coupling correction processing on the tool feed rate and the tool rotation speed according to the vibration amplitude data and the vibration main frequency data obtained in the plastic preheating zone to obtain a first feed rate and a first rotation speed;

[0087] A shallow scratching process is performed on the tool according to the first feed rate and the first rotation speed to obtain an initial cutting saw kerf that deeply covers the outer layer copper interconnection.

[0088] The following is a detailed description of the steps involved in the above embodiment:

[0089] The gradient setting process for infrared irradiation power is implemented through a computer control system, enabling precise power distribution design. The thermoplastic transition temperature (TPT) refers to the critical temperature at which the shear buffer copper strip transitions from elastic to plastic behavior. This temperature is reduced to 85-95°C for twin-modified copper strips. The lateral width of the saw street refers to the spatial dimension of the saw street perpendicular to the scribing direction. The lateral width of the saw street for standard semiconductor devices ranges from 50-100 microns. The gradient setting process involves calculating the required infrared power at different locations based on the heating target and geometric constraints. In practice, the control system first reads the TPT parameter of the shear buffer copper strip and then acquires the lateral width data for the saw street in the wafer layout. The system divides the saw street area into multiple sub-areas, creating a grid with a 2-micron spacing from the center of the saw street toward the edge. For each grid point, the system calculates the required heating power based on the distance from the center of the saw street. The highest power value is set at the center of the saw street, with the power decreasing linearly toward the edge. Backside power distribution data refers to a data set that records the infrared heating power values ​​corresponding to various locations on the backside of the wafer. This data is stored as coordinate-power pairs. For example, for a saw street with a width of 80 microns, the infrared power is set to 2 watts per square centimeter at the center, decreasing to 0.5 watts per square centimeter at the edge, and a linear distribution is applied in the middle. Setting the power at the center of the saw street higher than the power at the edge creates a temperature gradient from the center to the edge, ensuring that the shear buffer copper strip area reaches the plastic transition temperature while the surrounding area remains relatively cool, thus avoiding the thermal impact of an excessively large heating range on surrounding device structures.

[0090] Zoned infrared heating utilizes a multi-point, independently controlled infrared heating array to achieve precise temperature distribution control. Backside power distribution data is transmitted to an infrared heating controller, which drives multiple independent infrared emitting units based on this data. Each infrared emitting unit targets a specific area on the backside of the wafer and uses a halogen lamp or ceramic heater as a heat source. Zoned infrared heating involves applying infrared radiation of varying intensities to different areas of the wafer backside according to a preset power distribution. During the heating process, the radiant energy from the infrared emitting units is transmitted through the wafer's silicon substrate to the shear buffer copper strip area on the front side. Due to the excellent infrared transmittance of silicon, the infrared energy is effectively transferred to the front side structure. The heating time is controlled within 30-60 seconds to ensure a stable temperature distribution. The plastic preheat zone is a region where the temperature reaches the thermoplastic transition temperature of the shear buffer copper strip. This region is located at the center of the saw street and is continuously distributed along the dicing direction. A temperature range of ±5°C (100-100°C) around the thermoplastic transition temperature ensures that the shear buffer copper strip remains in a plastic state without causing structural damage. For example, when the thermoplastic transition temperature of the shear buffer copper strip is 90°C, the temperature of the plastic preheating zone is controlled within the range of 85-95°C. Temperature monitoring is performed in real time using an infrared thermal imager, and the heating process is completed when the temperature in the center of the saw path reaches the target range. The five-degree temperature control accuracy ensures that the shear buffer copper strip enters the ideal plastic state. If the temperature is too low, the copper strip maintains high rigidity, while if the temperature is too high, there is a risk of over-softening or structural damage.

[0091] Vibration data acquisition and coupled parameter correction processing enable dynamic optimization of scribing parameters through real-time monitoring and feedback control. After the plastic preheating zone reaches the target temperature, vibration sensors installed on the scribing equipment acquire vibration signals from the sawing area. Vibration amplitude data refers to the maximum displacement value of the vibration signal in the time domain, reflecting the intensity of mechanical vibration during the scribing process. Vibration dominant frequency data refers to the frequency component with the highest energy concentration in the vibration signal spectrum, reflecting the primary characteristic frequency of vibration. During the acquisition process, the vibration sensor continuously records the vibration signal at a sampling rate of 1000 times per second, with a data acquisition time of 5-10 seconds. Coupled correction processing is a calculation process that jointly adjusts scribing process parameters by comprehensively considering both vibration amplitude and dominant frequency. In specific implementation, the control system performs frequency domain analysis on the acquired vibration data, extracting the primary frequency components and corresponding amplitude information. When the vibration amplitude exceeds a preset threshold, the system reduces the tool feed rate to reduce mechanical shock. When the dominant vibration frequency approaches the natural frequency of the equipment, the system adjusts the tool speed to avoid resonance. The primary feed rate refers to the tool movement speed after vibration feedback correction, and the primary speed refers to the corrected tool rotation speed. For example, when a vibration amplitude of 5 microns and a main frequency of 200 Hz are detected, the system adjusts the feed rate from the original 10 mm / min to 7 mm / min and the rotation speed from 30,000 rpm to 25,000 rpm. Dynamic parameter adjustment enables real-time matching of the scribing process with the material state, ensuring that appropriate cutting parameters are used when the shear buffer copper strip is in a plastic state, maximizing the stress buffering effect while avoiding excessive mechanical shock.

[0092] Shallow scribing utilizes parameter-optimized tool operation to achieve precise initial kerf formation. A first feed rate and a first rotational speed are input into the dicing equipment's motion control system as tool control parameters. Shallow scribing involves the process of making an initial deep cut into the wafer surface using optimized tool parameters, with the cut depth controlled to cover the outermost copper interconnect structures. In specific implementation, the diamond blade, driven by the control system, rotates at a first rotational speed while simultaneously moving along the saw street at a first feed rate. The tool's descent depth is controlled by the Z-axis control system, with the cut depth set to penetrate the outermost copper interconnect layer and dielectric layer without touching underlying structures. Deep coverage of the outer copper interconnects means the cut depth is sufficient to completely separate the topmost metal wiring structures, a depth determined by the specific device layer thickness design. The initial kerf is the trench structure formed during the first cut, which is continuous along the saw street direction and has a regular geometric profile. For example, for a device with an outer copper interconnect thickness of 800 nanometers, the shallow scribing cut depth is set to 1 micron to ensure complete separation of the outer layers. During the cutting process, the shear buffer copper strip, in its plastic state, deforms in a controlled manner under the action of the cutter, absorbing some of the cutting impact energy and reducing stress transfer to deeper structures. The optimized cutting parameters, matched to the material's plastic state, achieve a high-quality initial cut, resulting in a smooth kerf with precise geometry, creating ideal starting conditions for subsequent deeper cutting.

[0093] In one embodiment of the present invention, the infrared irradiation power is gradient-set according to the thermoplastic transition temperature of the shear buffer copper strip and the lateral width of the saw street to obtain the back power distribution data, including:

[0094] The thermoplastic transition temperature of the shear buffer copper strip is divided into temperature intervals to obtain the center target temperature and the edge target temperature;

[0095] Performing interpolation calculation on the center target temperature and the edge target temperature according to the transverse width of the saw street to obtain a continuous temperature gradient curve;

[0096] Power mapping processing is performed on the infrared radiation power according to the temperature gradient curve to obtain back surface power distribution data.

[0097] The following is a detailed description of the steps involved in the above embodiment:

[0098] Temperature interval division processing refers to the data processing process of dividing the continuous temperature range into discrete control points according to the spatial position requirements. During specific implementation, the control system reads the thermoplastic transition temperature parameter of the shear buffer copper strip, which is 90 degrees Celsius. The system divides the saw street area into two control zones: the central area and the edge area according to the heating requirements. The center target temperature is set to the thermoplastic transition temperature value, that is, 90 degrees Celsius, to ensure that the shear buffer copper strip in the center of the saw street fully enters the plastic state. The edge target temperature is set to a value lower than the thermoplastic transition temperature, specifically 60 degrees Celsius, to avoid overheating of the edge area of ​​the saw street. The dual-point temperature control strategy ensures the spatial selectivity of the heating effect. The shear buffer copper strip area reaches the plastic state while the surrounding structure remains at a relatively low temperature, avoiding adverse effects of heat diffusion on adjacent device structures.

[0099] The interpolation calculation process uses a linear interpolation algorithm to calculate the temperature distribution in the middle position according to the geometric dimensions of the saw street. Interpolation calculation processing refers to the process of calculating the corresponding values ​​of each point in the middle through a mathematical algorithm under the condition that the values ​​of the two endpoints are known. During the specific implementation, the system obtains the horizontal width data of the saw street, and the standard value is 80 microns. The system divides the saw street from the center to the edge into 40 calculation nodes at a spacing of 2 microns, and each node corresponds to a temperature control point. For any node X microns away from the center of the saw street, its target temperature value is calculated by a linear interpolation formula. The continuous temperature gradient curve refers to a smooth temperature distribution function formed by connecting all temperature control points. The curve linearly decreases from 90 degrees Celsius at the center of the saw street to 60 degrees Celsius at the edge. The linear gradient distribution ensures the continuity and controllability of temperature changes, avoids the concentration of thermal stress that may be caused by sudden temperature changes, and realizes a smooth transition from the high temperature zone in the center to the low temperature zone at the edge.

[0100] Power mapping processing refers to the data conversion process of calculating the heating power distribution required to achieve the target temperature distribution. During specific implementation, the system calls the material thermal property database to obtain parameters such as the thermal conductivity and specific heat capacity of the silicon substrate. For each temperature control point on the temperature gradient curve, the system calculates the required heating power density based on the difference between the target temperature and the ambient temperature at that point. The infrared irradiation power corresponding to the target temperature of 90 degrees Celsius at the center position is set to 2.5 watts per square centimeter, and the power corresponding to 60 degrees Celsius at the edge position is 1.0 watt per square centimeter. The back power distribution data refers to a complete data set that records the infrared heating power values ​​corresponding to each spatial position on the back of the wafer. The data is stored in the format of coordinate-power value pairs and transmitted to the heating control system. The precise mapping relationship between power and temperature ensures that the heating system can generate the required power distribution on the back of the wafer according to the design requirements, realize precise temperature control of the front shear buffer copper strip area, and meet the spatial selectivity requirements of plastic activation.

[0101] In one embodiment of the present invention, the coupling correction processing of the tool feed rate and the tool rotation speed is performed based on the vibration amplitude data and the vibration main frequency data obtained in the plastic preheating zone to obtain the first feed rate and the first rotation speed, including:

[0102] performing amplitude-threshold comparison processing on the vibration amplitude data obtained in the plastic preheating zone to obtain a feed rate adjustment coefficient;

[0103] Performing main frequency-harmonic deviation calculation processing on the vibration main frequency data obtained in the plastic preheating zone to obtain a speed adjustment coefficient;

[0104] A joint correction process is performed on the original feed rate and the original speed according to the feed rate adjustment coefficient and the speed adjustment coefficient to obtain a first feed rate and a first speed.

[0105] The following is a detailed description of the steps involved in the above embodiment:

[0106] The amplitude-threshold comparison process converts vibration intensity into a feed rate adjustment parameter through a numerical comparison algorithm. The amplitude-threshold comparison process refers to the calculation process of numerically comparing the measured vibration amplitude data with a preset threshold value. In specific implementation, the control system obtains vibration amplitude data from the vibration sensor in the plastic preheating zone, which represents the maximum value of the vibration displacement. The system presets a vibration amplitude threshold value of 3 microns. When the measured vibration amplitude exceeds this threshold, it indicates that the cutting process has generated excessive mechanical impact. The feed rate adjustment coefficient is a numerical parameter used to correct the original feed rate and is calculated from the ratio of the vibration amplitude to the threshold value. When the vibration amplitude is 2 microns, the adjustment coefficient is 1.0, maintaining the original rate. When the vibration amplitude is 4 microns, the adjustment coefficient drops to 0.7, indicating a 30% reduction in speed. The threshold comparison process achieves a quantitative correlation between vibration intensity and cutting parameters, and real-time monitoring of the vibration amplitude provides reliable feedback for feed rate adjustment.

[0107] The main frequency-harmonic deviation calculation process uses frequency domain analysis to identify resonance risks and generate speed adjustment parameters. This process involves calculating the deviation between the measured main vibration frequency and the equipment's harmonic frequencies. In practice, the system performs spectral analysis on the main vibration frequency data, extracting the frequency component with the highest energy concentration as the main frequency value. The system pre-stores the equipment's natural frequency and its harmonic frequencies, including key frequency points such as 200 Hz, 400 Hz, and 600 Hz. The deviation calculation determines the degree of resonance risk by comparing the main frequency value with the nearest harmonic frequency. The speed adjustment coefficient is a numerical parameter used to modify the original speed, with the adjustment amplitude determined by the deviation. For example, if the main vibration frequency is 195 Hz and close to its 200 Hz harmonic, the deviation is 5 Hz. The system generates a speed adjustment coefficient of 0.85, indicating a 15% speed reduction to avoid resonance. Frequency deviation analysis effectively prevents the adverse effects of resonance on cutting quality and ensures that the tool's operating frequency remains within the equipment's sensitive frequency range.

[0108] The joint correction process uses a multi-parameter coupling algorithm to simultaneously adjust the feed rate and speed parameters. The joint correction process refers to the calculation process of collaboratively adjusting the original cutting parameters by comprehensively considering the feed rate adjustment coefficient and the speed adjustment coefficient. In specific implementation, the system reads the original feed rate and original speed as reference parameters. The original feed rate is set to 10 mm per minute, and the original speed is set to 30,000 rpm. The system multiplies the original parameters with the corresponding adjustment coefficients to obtain the corrected parameter values. The first feed rate is calculated by multiplying the original feed rate by the feed rate adjustment coefficient, and the first speed is calculated by multiplying the original speed by the speed adjustment coefficient. For example, when the feed rate adjustment coefficient is 0.7 and the speed adjustment coefficient is 0.85, the first feed rate is 7 mm per minute and the first speed is 25,500 rpm. The dual-parameter joint adjustment achieves a precise match between the cutting process and the material state, ensuring that the optimal cutting parameter combination is used when the shear buffer copper strip is in a plastic state, maximizing the stress absorption effect and ensuring the stability of the cutting quality.

[0109] Please continue reading Figure 1 , the wafer is subjected to a layered progressive deep cutting process according to the initial cutting saw kerf, the shear buffer copper tape is subjected to slip deformation under the action of the tool pulse load and reset by the elastic action of the low dielectric constant dielectric layer, thereby forming a full-depth cutting saw kerf;

[0110] In one embodiment of the present invention, the wafer is subjected to a layered progressive deep cutting process according to the initial cutting saw kerf, the shear buffer copper tape is subjected to slip deformation under the action of the tool pulse load and is reset by the elastic action of the low dielectric constant dielectric layer to form a full-depth cutting saw kerf, including:

[0111] Performing layered cutting depth planning processing on the initial cutting saw kerf depth data to obtain multiple cutting depth data that increase in sequence along the saw street direction and correspond to the combination of the thickness of the metal layer and the dielectric layer;

[0112] Performing a first-level deep cutting process on the wafer according to the multiple-level cutting depth data to obtain first-level saw kerf displacement data that generates a sliding displacement in the transverse direction of the shear buffer copper strip;

[0113] A rebound buffer process is performed on the wafer based on the first-level saw kerf displacement data, and the shear buffer copper tape is reset by utilizing the elastic effect of the low-k dielectric layer. Meanwhile, a deviation correction is performed on the next-level cutting depth data to obtain the corrected cutting depth data.

[0114] The subsequent deep cutting process and rebound buffer process are cyclically executed according to the corrected cutting depth data until the silicon base surface is cut deep enough to obtain a full-depth cutting saw kerf.

[0115] The following is a detailed description of the steps involved in the above embodiment:

[0116] The layered depth of cut planning process uses a data analysis system to convert the wafer's complex multi-layer structure into a stepped cutting depth sequence. Layered depth of cut planning refers to the data processing process that develops a staged cutting depth plan based on the thickness information of the wafer's multi-layer structure. In practice, the control system reads the initial kerf depth data, which records the current cutting depth after the shallow scribe process is completed, with a standard value of 1 micron. The system then accesses the wafer structure database to obtain detailed thickness information for each layer, including the specific thickness values ​​for each metal and dielectric layer. A metal and dielectric layer thickness combination refers to a composite structural unit consisting of adjacent metal and dielectric layers. The total thickness of each combination determines the incremental depth of cut for each step. The system analyzes each layer combination from the surface to the depth. The first step consists of an 800-nanometer metal layer and a 500-nanometer underlying dielectric layer, for a total thickness of 1.3 microns. The second step consists of a 600-nanometer metal layer and a 400-nanometer dielectric layer, for a total thickness of 1.0 micron. Multi-step depth of cut data refers to a data sequence containing the target depths for each cutting stage, with each step increasing in depth along the saw street. For example, based on an initial depth of 1 micron, the first-level cutting depth is 2.3 microns, the second level is 3.3 microns, and the third level is 4.5 microns, increasing in sequence until the silicon base depth is reached. This step-by-step planning avoids the excessive stress impact that may be caused by a single deep cut. The cutting depth at each level is controlled within the material's tolerance range, ensuring that the shear buffer copper strip can effectively absorb stress at each stage without structural failure.

[0117] First-stage deep cutting utilizes precise depth control technology to selectively cut specific layers. First-stage deep cutting refers to the process of precisely cutting the wafer according to the first-stage depth value in the multi-stage depth data. During implementation, the Z-axis control system of the dicing equipment lowers the cutter to the depth specified by the first-stage depth data, i.e., 2.3 microns. The cutter then makes a lateral cut at this depth, penetrating the combined structure of the first metal layer and the underlying dielectric layer. During the cutting process, the shear buffer copper strip undergoes plastic deformation under the tool load, absorbing some of the cutting energy through slip. Slip displacement refers to the lateral dislocation movement of the shear buffer copper strip under shear stress, and the amount of displacement indicates the degree of stress absorption. First-stage kerf displacement data is measured by recording the lateral slip distance of the shear buffer copper strip, acquired in real time using a high-precision displacement sensor. For example, during a first-stage deep cut, the shear buffer copper strip experiences a lateral slip displacement of 0.8 microns, which is recorded as first-stage kerf displacement data. The occurrence of slip deformation indicates that the shear buffer copper strip successfully assumed the stress absorption function, converting the shear stress originally transmitted to the brittle interface into its own plastic deformation energy, effectively protecting the metal-dielectric layer interface from stress concentration damage.

[0118] The rebound buffer process utilizes the elastic properties of the dielectric material to partially reset the shear buffer copper strip and optimize subsequent cutting parameters. Rebound buffering refers to the physical process of using the elastic restoring force of the low-k dielectric layer to return the slipped shear buffer copper strip to a relatively stable position. In practice, when the tool moves away from the cutting position, the low-k dielectric layer, which has been subjected to compressive deformation, elastically rebounds, generating a restoring force on the surface of the shear buffer copper strip. Because the dielectric layer material has a certain elastic modulus, its elastic action can drive the shear buffer copper strip to reverse displacement, partially returning it from its maximum slip position to a position close to its initial position. During this reset, the shear buffer copper strip moves approximately 60-70% of its slip displacement, from a 0.8 micron slip position to a 0.3 micron position. The system simultaneously applies deviation correction to the next-level depth of cut data. Deviation correction is a calculation process that adjusts the preset depth of cut value based on actual slip and rebound conditions. The corrected depth of cut data is calculated by subtracting the residual slip displacement of 0.3 microns from the original second-level depth of cut value of 3.3 microns, resulting in a corrected value of 3.0 microns. For example, if the first-gear saw kerf displacement data shows a slip of 0.8 microns and a remaining 0.3 microns after rebound, the second-gear cutting depth is corrected from the preset 3.3 microns to 3.0 microns. Deviation correction ensures accurate depth control for each cutting step, preventing cumulative errors from affecting subsequent cutting accuracy, while fully utilizing the elastic recovery capacity of the shear buffer copper strip to prepare for stress absorption in the next step.

[0119] Cyclic deep cuts and rebound buffering achieve full-depth through-cuts by repeatedly executing an optimized cutting and recovery sequence. Cyclic execution refers to the repeated execution of subsequent deep cuts and rebound buffering operations according to the same process flow. In practice, the system sequentially executes deep cuts through the second, third, and final stages, each using the corrected depth of cut data as the target depth. After each stage, rebound buffering is immediately performed to measure the new kerf displacement data and calculate the deviation correction for the next stage. The silicon basal plane, the bottommost silicon substrate surface of the wafer, serves as the final target depth for the cut. Cutting deep to the basal plane means completely penetrating all metal and dielectric layers to the silicon substrate surface. A full-depth kerf is a continuous, geometrically defined trench extending from the wafer surface to the basal plane. For example, for a multilayer structure with a total thickness of 10 microns, a full-depth kerf of 10 microns is achieved after six stages of deep cut. The repeated slip-and-rebound cycles of the shear buffer tape during each processing step achieve segmented stress absorption and release, preventing excessive stress accumulation at any single interface. This cyclic processing mechanism fully leverages the reusable nature of the shear buffer tape, enabling it to continuously perform its stress buffering function throughout the cutting process, ensuring that each metal-dielectric interface is separated at a safe stress level. This fundamentally addresses the problem of selective interface failure caused by shear stress accumulation between layers in multilayer structures.

[0120] In one embodiment of the present invention, the layered cutting depth planning processing is performed on the initial cutting saw kerf depth data to obtain multiple cutting depth data that increase in sequence along the saw street direction and correspond to the combination of the thickness of the metal layer and the dielectric layer, including:

[0121] Performing layer thickness accumulation processing based on the metal layer thickness data and the dielectric layer thickness data to obtain a multi-layer stack cumulative depth sequence;

[0122] Performing energy-depth matching processing on tool single pulse energy constraint data according to the multi-layer stack cumulative depth sequence to obtain a cutting depth increment sequence;

[0123] The initial cutting kerf depth data is subjected to depth segmentation processing according to the cutting depth increment sequence to obtain multiple levels of cutting depth data.

[0124] The following is a detailed description of the steps involved in the above embodiment:

[0125] Layer thickness accumulation processing converts the thickness information of the multi-layer structure into continuous depth position data through a numerical calculation system. Layer thickness accumulation processing refers to the data processing process of adding the thickness data of each metal layer and dielectric layer layer by layer in order from the surface layer to the deep layer. In specific implementation, the control system reads the metal layer thickness data and dielectric layer thickness data from the wafer structure database. These data record the precise thickness value of each structural layer. The system performs cumulative calculations according to the actual arrangement order of the layer structure. The thickness of the first metal layer is 800 nanometers as the first cumulative value. The thickness of the first metal layer is 800 nanometers plus the thickness of the first dielectric layer is 500 nanometers to obtain the second cumulative value of 1300 nanometers. The thickness of the second metal layer is 600 nanometers and continues to accumulate to obtain the third cumulative value of 1900 nanometers. The cumulative depth sequence of a multi-layer stack refers to an ordered data set containing the depth position of the interface of each layer. The sequence records the cumulative distance from the wafer surface to each key interface. For example, for a structure containing three metal and three dielectric layers, the cumulative depth sequence is 800 nm, 1300 nm, 1900 nm, 2400 nm, 2800 nm, and 3300 nm, with each value corresponding to a critical layer interface. This cumulative processing enables a digital description of complex multi-layer structures, providing a precise spatial reference for subsequent cutting depth planning, ensuring that each cut accurately reaches the predetermined layer interface.

[0126] Energy-depth matching uses engineering calculations to optimize the matching of tool energy limits with cutting depth requirements. Energy-depth matching involves determining appropriate depth increments for each cutting stage based on the tool's single-pulse energy constraints. The tool's single-pulse energy constraint represents the maximum energy the tool can output in a single cutting motion. This value is determined by the tool's material properties and the equipment's power limitations. In implementation, the system reads the tool's single-pulse energy constraint data, with a standard value of 0.5 joules per pulse. The system analyzes the differences between adjacent depth values ​​in the accumulated depth sequence of a multi-layer stack and calculates the material removal and energy required for each depth increment. For depth increments with large thickness differences, the system decomposes them into multiple smaller depth increments to meet the energy constraints. The depth increment sequence is the sequence of depth increments for each step, resulting from energy matching optimization. Each increment falls within the tool's energy constraints. For example, if a layer spacing of 1000 nanometers exceeds the single-pulse energy limit, the system decomposes it into two 500-nanometer depth increments. Energy constraint matching ensures that each cutting step is performed under the optimal working condition of the tool, avoiding tool overload or reduction in cutting quality caused by excessive cutting depth, while ensuring the stability and repeatability of the cutting process.

[0127] Depth segmentation uses a data organization algorithm to convert incremental depths of cut into a complete, step-by-step cutting plan. Depth segmentation refers to the calculation process that generates target depths for each step based on the incremental depth of cut sequence and the initial kerf depth data. In practice, the system uses the initial kerf depth data as the starting point for the segmentation calculation. This data records the current cutting depth after the shallow entry process is complete. The system then adds the initial depth value to the first increment in the incremental depth of cut sequence to generate the first step depth of cut data. The system then adds the first step depth to the second increment to generate the second step depth of cut data, and so on until all depth increments are covered. Multi-step depth of cut data is a complete data set containing the target depths for each cutting stage. This data set determines the specific execution plan for the layered progressive cutting process. For example, based on an initial depth of 1000 nm and a sequence of incremental depths of cut of 500 nm, 600 nm, and 400 nm, the generated multi-step depth of cut data is 1500 nm for the first step, 2100 nm for the second step, and 2500 nm for the third step. Segmented processing achieves a systematic decomposition of complex cutting tasks. Each cutting depth is within the tool's capabilities and matches the material's structural characteristics, ensuring the operability of the layered progressive cutting strategy and the consistency of the cutting quality.

[0128] Please continue reading Figure 1 The full-depth cutting saw seam is cooled in a nitrogen environment, the twin structure inside the shear buffer copper strip is stabilized and forms a mechanical constraint structure with the adjacent dielectric layer, and a bridging unit is formed on the side wall of the saw seam.

[0129] In one embodiment of the present invention, the full-depth cutting saw kerf is subjected to a nitrogen environment cooling treatment, the twin structure inside the shear buffer copper tape is stabilized and forms a mechanical constraint structure with the adjacent dielectric layer, and a bridging unit is formed on the sidewall of the saw kerf, comprising:

[0130] According to the difference between the shear buffer copper strip twin stable temperature zone and the linear thermal expansion coefficient of the low dielectric constant dielectric layer, a double window setting process is performed on the nitrogen cooling stage to obtain the first constant temperature window parameter and the second temperature reduction gradient parameter;

[0131] performing a constant temperature dwell process on the full-depth cutting saw kerf according to the first constant temperature window parameters, so that the twin interfaces inside the shear buffer copper strip are rearranged during the constant temperature dwell period and axial residual compressive stress data of the shear buffer copper strip is obtained;

[0132] performing radial gradient cooling on the full-depth saw kerf according to the second temperature-lowering gradient parameter, so as to cause the shear buffer copper strip to undergo reversible lateral contraction and drive the adjacent low-k dielectric layer to produce nested rebound, thereby obtaining a coupled shrinkage hoop interface;

[0133] A stress redistribution process is performed on the full-depth cutting saw seam based on the coupled shrinkage interface and the axial residual compressive stress data. During the stress redistribution process, the residual compressive stress of the shear buffer copper strip and the rebound elasticity of the low dielectric constant dielectric layer are used to achieve interlocking, forming a bridging unit that continuously extends along the side wall of the saw seam.

[0134] The following is a detailed description of the steps involved in the above embodiment:

[0135] The dual-window setting process determines the phased parameter control scheme for the nitrogen cooling process through material thermophysical property analysis and calculation. The twinning stability temperature range refers to the temperature range in which the twin structure inside the shear buffer copper strip remains stable. The twinning stability temperature range of the laser-modified copper strip is 120-180 degrees Celsius. The linear thermal expansion coefficient difference refers to the difference in volume change rate between the shear buffer copper strip and the low-k dielectric layer material when the temperature changes. The linear thermal expansion coefficient of copper material is approximately 17×10 -6 / degrees Celsius, SiOCH dielectric material is about 25×10 -6 / degrees Celsius, the difference is 8×10 -6 / degrees Celsius. The dual-window setting process refers to the parameter design process that divides the cooling process into two independent control stages according to the difference in material properties. During the specific implementation, the control system analyzes the temperature range of the twin stable temperature zone and the influence of the difference in thermal expansion coefficient, and determines that the first stage requires constant temperature treatment in the twin stable temperature zone to complete structural stabilization. The first constant temperature window parameter is set to a temperature of 150 degrees Celsius and a duration of 15 minutes. This parameter ensures that the twin structure is fully rearranged in the stable temperature zone. The second cooling gradient parameter is set to a cooling rate of 5 degrees Celsius per minute and a gradient range from 150 degrees Celsius to room temperature of 25 degrees Celsius. This parameter controls the coordinated shrinkage of the copper strip and the dielectric layer. For example, when the shear buffer copper strip twin stable temperature zone is 120-180 degrees Celsius and the difference in thermal expansion coefficient with the dielectric layer is 8×10 -6 When the temperature rises to 150 degrees Celsius, the first constant temperature window parameter is set at 150 degrees Celsius for 15 minutes, and the second cooling gradient parameter is set to a slow cooling rate of 5 degrees Celsius per minute. The staged cooling control avoids thermal stress mismatch caused by rapid cooling, ensuring the stabilization of the twin structure and the orderly thermal contraction process.

[0136] The constant temperature dwell treatment promotes the rearrangement of the twin structure and optimizes the stress state within the shear buffer copper strip through precise temperature control. This heat treatment involves maintaining a constant temperature at the full-depth saw kerf within the first constant temperature window. A nitrogen cooling system precisely controls the ambient temperature at 150°C for 15 minutes, with a nitrogen flow rate of 10 liters per minute to ensure uniform temperature distribution. During the constant temperature dwell period, thermally activated dislocation rearrangement and interface migration occur within the shear buffer copper strip, leading to a uniform twin density and distribution. Complete rearrangement of the twin interfaces occurs when the spatial position and orientation of the twin boundaries reach a stable configuration at their lowest energy state. Axial residual compressive stress data refers to the retained compressive stress along the length of the shear buffer copper strip, acquired in real time using strain gauge sensors. The system continuously measures the stress state of the shear buffer copper strip during the constant temperature dwell process, recording the gradual decrease in axial compressive stress from an initial 150 MPa to a stable 80 MPa. For example, during a 15-minute dwell at 150°C, the shear buffer copper strip's axial residual compressive stress releases from 150 MPa and stabilizes at 80 MPa. This data is recorded as the axial residual compressive stress. This dwell fully stabilizes the twin structure and orderly releases internal stress, creating ideal initial conditions for the subsequent shrinkage and solidification process, ensuring the final bridge structure possesses stable mechanical properties.

[0137] Radial gradient cooling utilizes zoned temperature control technology to achieve coordinated shrinkage of the shear buffer copper strip and the dielectric layer. Radial gradient cooling involves differentially cooling the full-depth saw kerf from the center to the edge according to a second cooling gradient parameter. In practice, the nitrogen cooling system utilizes a multi-point, independently controlled array of cooling nozzles. The nitrogen flow rate in the center zone is set at 15 liters per minute, with a cooling rate of 5 degrees Celsius per minute, while the nitrogen flow rate in the edge zone is set at 8 liters per minute, with a cooling rate of 3 degrees Celsius per minute. Reversible lateral shrinkage refers to the reversible lateral dimensional reduction of the shear buffer copper strip during cooling. The amount of shrinkage is directly related to temperature changes and the thermal expansion coefficient of the material. Nested rebound refers to the cooperative deformation recovery of the low-k dielectric layer driven by the shear buffer copper strip's contraction. The porous structure of the dielectric layer undergoes elastic recovery after the external force is removed. A coupled shrinkage interface is the intimate contact formed between the shear buffer copper strip and the low-k dielectric layer during the coordinated shrinkage process. This interface exhibits the mechanical characteristics of mutual restraint and support. During cooling, the shear buffer copper strip shrinks lateraly by approximately 0.2 microns, driving corresponding deformation recovery in the surrounding dielectric layer, forming a stable mechanical constraint. For example, when the temperature drops from 150°C to 25°C, the shear buffer copper strip shrinks lateraly by 0.2 microns, driving a nested rebound of 0.15 microns in the adjacent SiOCH dielectric layer, which together form a coupled shrinkage interface. Gradient cooling ensures matching shrinkage rates between the different materials, avoiding interface separation or excessive stress concentration caused by uncoordinated shrinkage.

[0138] Stress redistribution processing achieves the final formation of a stable bridging structure through the synergistic action of multiple materials. Stress redistribution processing refers to a mechanical process that utilizes coupled shrinkage interface and axial residual compressive stress data to adjust the multi-directional stress state. In specific implementation, the system calculates the three-dimensional stress distribution within the shear buffer copper strip based on the axial residual compressive stress data of 80 MPa and the constraints of the coupled shrinkage interface. The interlocking of residual compressive stress and rebound elasticity refers to the mechanical coupling state in which the axial compressive stress of the shear buffer copper strip and the radial elastic constraint of the low-k dielectric layer form a mutually balanced state. This interlocking mechanism achieves a stable mechanical equilibrium through the orthogonal configuration of stress directions: the axial compressive stress prevents tensile failure of the copper strip, while the radial constraint prevents lateral displacement of the copper strip. A bridging unit is a composite structural unit continuously distributed along the sidewall of the saw kerf. This unit is composed of the shrinkage-cured shear buffer copper strip and the tightly fitted dielectric layer. The continuous ductility characteristic indicates that the bridging unit maintains structural integrity and mechanical continuity along the entire length of the saw kerf. For example, under the combined action of 80 MPa axial residual compressive stress and the 0.2 micron coupling hoop constraint, a continuous bridge unit is formed along the sidewall of a 7 micron wide saw kerf. This stress redistribution transforms the shear buffer copper tape from a temporary buffering function to a permanent protective structure. The bridge unit is capable of withstanding long-term thermomechanical cyclic loads, providing continuous stress buffering protection for the device during subsequent packaging and use.

[0139] In one embodiment of the present invention, the radial gradient cooling process is performed on the full-depth cutting saw kerf according to the second temperature reduction gradient parameter, so that the shear buffer copper strip undergoes reversible lateral contraction and drives the adjacent low-k dielectric layer to produce nested rebound, thereby obtaining a coupled shrinkage interface, including:

[0140] Performing center-edge partitioning and setting processing on the nitrogen flow rate according to the second cooling gradient parameter to obtain center flow rate data and edge flow rate data;

[0141] performing zoned nitrogen supply processing according to the central flow data and the edge flow data to obtain radial temperature gradient data;

[0142] According to the radial temperature gradient data, a synchronous matching process is performed on the lateral shrinkage of the shear buffer copper strip and the rebound of the low dielectric constant dielectric layer to obtain a coupled shrinkage hoop interface.

[0143] The following is a detailed description of the steps involved in the above embodiment:

[0144] The center-edge zoning setting process divides the nitrogen cooling system into different supply areas through a flow control algorithm. The center-edge zoning setting process refers to the data processing process of dividing the saw street area into a center area and an edge area according to the radial position based on the second cooling gradient parameter, and setting the corresponding nitrogen flow parameters for each area. During specific implementation, the control system reads the second cooling gradient parameter, which stipulates a cooling requirement of a cooling rate of 5 degrees Celsius per minute. The system divides the saw street area into sections according to the radial distance from the center line of the shear buffer copper strip. The area within 2 microns from the center line is defined as the center area, and the area within 2-10 microns from the center line is defined as the edge area. The center flow data refers to the volume flow value of nitrogen supplied to the center area. According to the requirement of a faster cooling rate for the center area, the center flow data is set to 15 liters per minute. The edge flow data refers to the volume flow value of nitrogen supplied to the edge area. Considering that the edge area needs a slower cooling to avoid sudden changes in thermal stress, the edge flow data is set to 8 liters per minute. For example, when the second cooling gradient parameter requires an overall cooling rate of 5 degrees Celsius per minute, the nitrogen flow rate in the central area is set to 15 liters per minute to achieve faster local cooling, and the nitrogen flow rate in the edge area is set to 8 liters per minute to achieve gradual cooling. This zoned flow setting enables spatially selective control of the cooling process. High-flow cooling in the central area ensures that the shear buffer copper strip quickly reaches the target temperature, while low-flow cooling in the edge area avoids the adverse effects of excessive cooling on surrounding structures.

[0145] The zoned nitrogen supply process utilizes a multi-channel, independently controlled gas distribution system to achieve precise, zoned cooling. This refers to the cooling process in which nitrogen is supplied to corresponding zones based on center and edge flow data. Specifically, the nitrogen cooling system utilizes a multi-nozzle array configuration. Three high-flow nozzles are deployed in the center zone, each delivering 5 liters per minute (15 liters per minute), for a total flow rate of 15 liters per minute. Four low-flow nozzles are deployed in the edge zones, each delivering 2 liters per minute (8 liters per minute). Each nozzle is precisely regulated using independent flow control valves, ensuring that the actual supply flow rate deviates from the set flow rate by less than ±5%. Radial temperature gradient data, which measures the temperature distribution from the center of the saw blade to the edge, is acquired in real-time using an infrared thermal imager. During the cooling process, the temperature in the center zone decreases at a rate of 6 degrees Celsius per minute, while the temperature in the edge zone decreases at a rate of 3 degrees Celsius per minute, resulting in a significant radial temperature difference. For example, after 5 minutes of cooling, the temperature in the center area drops from 150 degrees Celsius to 120 degrees Celsius, and the temperature in the edge area drops from 150 degrees Celsius to 135 degrees Celsius, with a radial temperature gradient of 15 degrees Celsius. Precise zoned supply control achieves the preset radial temperature distribution, and the temperature difference between the center and the edge creates ideal driving conditions for the differentiated shrinkage behavior of different materials.

[0146] Synchronous proportioning processing achieves coordinated shrinkage of the shear buffer copper strip and the dielectric layer through a material deformation matching algorithm. Synchronous proportioning processing refers to the adjustment process of calculating and controlling the lateral shrinkage of the shear buffer copper strip and the rebound of the low dielectric constant dielectric layer based on the radial temperature gradient data. During specific implementation, the control system reads the radial temperature gradient data and analyzes the temperature difference and its changing trend between the central area and the edge area. The lateral shrinkage of the shear buffer copper strip refers to the lateral size reduction value of the copper strip under the action of the radial temperature gradient. This shrinkage is directly related to the temperature change and the thermal expansion coefficient of the copper material. When the temperature of the central area drops by 30 degrees Celsius, the lateral shrinkage of the shear buffer copper strip is calculated to be 0.24 microns. The rebound of the low dielectric constant dielectric layer refers to the elastic recovery deformation value of the SiOCH material driven by the contraction of the shear buffer copper strip. This rebound depends on the elastic modulus and porosity characteristics of the dielectric layer. The matching treatment adjusts the cooling rate of the edge region to maintain a proportional relationship between the dielectric layer's rebound and the copper strip's contraction. When the shear buffer copper strip shrinks by 0.24 microns, the dielectric layer's rebound is controlled to 0.18 microns. A coupled shrinkage interface is a tightly fitting interface formed between the shear buffer copper strip and the low-k dielectric layer during the synchronous shrinkage process. This interface exhibits the mechanical characteristics of mutual restraint and support. During interface formation, the contraction of the copper strip generates inward tensile stress, while the rebound of the dielectric layer generates outward compressive stress. These two stresses reach equilibrium at the interface. For example, when the shear buffer copper strip shrinks by 0.24 microns and the dielectric layer rebounds by 0.18 microns, a stable mechanical constraint relationship is formed at the interface, with an interfacial compressive stress of approximately 50 MPa. Synchronous matching ensures the coordinated deformation behavior of the different materials, avoiding interface separation or excessive stress concentration caused by shrinkage mismatch. The formation of the coupled shrinkage interface provides a reliable mechanical foundation for subsequent long-term stability, enabling the bridge unit to withstand repeated thermomechanical cycling loads while maintaining structural integrity.

[0147] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. All equivalent structural transformations made by using the contents of the present invention description and drawings under the inventive concept of the present invention, or direct / indirect application in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A wafer scribing process, characterized in that: include: The dummy copper wires in the saw street area are subjected to line width expansion processing, and the expanded dummy copper wires are continuously filled to form a shear buffer copper strip located between the low-k dielectric layer and the passivation layer and running through the multi-layer interconnect structure along a preset scribing direction; Performing a double-pulse laser annealing treatment on the surface of the shear buffer copper strip, inducing the formation of a twin structure inside the shear buffer copper strip by controlling the pulse duration and energy density, so that the yield strength of the shear buffer copper strip is lower than the yield strength of adjacent copper interconnects; performing infrared preheating on the back side of the wafer according to the thermoplastic transition temperature of the shear buffer copper strip to cause the shear buffer copper strip to enter a plastic deformation state, and adjusting the tool feed rate according to the obtained saw street area vibration data to form an initial cutting saw kerf; The wafer is subjected to a layered progressive deep cutting process according to the initial cutting saw kerf, the shear buffer copper tape is subjected to slip deformation under the action of the tool pulse load and reset by the elastic action of the low dielectric constant dielectric layer, thereby forming a full-depth cutting saw kerf; The full-depth cutting saw seam is subjected to a nitrogen environment cooling treatment, so that the twin structure inside the shear buffer copper strip is stabilized and forms a mechanical constraint structure with the adjacent dielectric layer, forming a bridging unit on the side wall of the saw seam.

2. The wafer scribing process according to claim 1, wherein: The method comprises: performing line width expansion processing on the dummy copper wires in the saw street area, and continuously filling the expanded dummy copper wires to form a shear buffer copper tape located between the low-k dielectric layer and the passivation layer and penetrating the multi-layer interconnect structure along a preset scribing direction, comprising: Performing line width redefinition processing on the interconnection layout according to the saw street dicing coordinates to generate mask data of an amplified area that is continuous along the dicing direction and has a width greater than that of the original dummy copper wire; Performing trench etching on a dummy copper wire in a low-k dielectric layer according to the amplification region mask data to obtain an amplification groove having parallel sidewalls and a depth consistent with the thickness of the dielectric layer; Depositing a conductive copper seed layer on the amplification tank to obtain a continuous seed layer that covers the tank inner wall and is conductive with the tank bottom; The amplification tank body is electroplated and filled according to the continuous seed layer, and then the filled copper body is chemically mechanically planarized to obtain a shear buffer copper strip whose top surface is flush with the upper surface of the low-k dielectric layer.

3. The wafer scribing process according to claim 1, wherein: The double-pulse laser annealing treatment is performed on the surface of the shear buffer copper strip, and a twin structure is induced inside the shear buffer copper strip by controlling the pulse duration and energy density, so that the yield strength of the shear buffer copper strip is lower than the yield strength of adjacent copper interconnects, including: The surface of the shear buffer copper strip is subjected to a first pulse thermal shock treatment. The first pulse duration and energy density are adjusted according to the thickness of the shear buffer copper strip and the thermal diffusivity of the low-k dielectric layer, thereby forming a semi-molten layer along the surface of the shear buffer copper strip. Performing a sub-pulse laser shock treatment on the surface of the shear buffer copper strip according to the thermal platform of the semi-molten layer to obtain a twin nucleation network located inside the semi-molten layer and continuously distributed laterally; Performing a heat-maintaining expansion treatment on the interior of the shear buffer copper strip according to the twin nucleation network to obtain a twin expansion band that penetrates along the thickness direction of the shear buffer copper strip; The shear buffer copper strip is subjected to a stepwise cooling and solidification treatment according to the twin extension band to obtain a softened shear buffer copper strip having a yield strength lower than that of adjacent copper interconnects.

4. The wafer scribing process according to claim 1, wherein: The method includes performing infrared preheating on the back side of the wafer according to the thermoplastic transition temperature of the shear buffer copper strip to cause the shear buffer copper strip to enter a plastic deformation state, and adjusting the tool feed rate according to the obtained saw street area vibration data to form an initial cutting saw kerf, including: The infrared irradiation power is gradient-set according to the thermoplastic transition temperature of the shear buffer copper strip and the lateral width of the saw street, and the back power distribution data is obtained. The power corresponding to the back power distribution data at the center of the saw street is higher than the power corresponding to the edge of the saw street. Performing a zoned infrared radiation heating process on the back side of the wafer according to the back side power distribution data to obtain a plastic preheating zone located at the center of the saw street and having a temperature within a range of ±5 degrees Celsius from the thermoplastic transition temperature; performing coupling correction processing on the tool feed rate and the tool rotation speed according to the vibration amplitude data and the vibration main frequency data obtained in the plastic preheating zone to obtain a first feed rate and a first rotation speed; A shallow scratching process is performed on the tool according to the first feed rate and the first rotation speed to obtain an initial cutting saw kerf that deeply covers the outer layer copper interconnection.

5. The wafer scribing process according to claim 4, wherein: The infrared irradiation power is gradient-set according to the thermoplastic transition temperature of the shear buffer copper strip and the lateral width of the saw street to obtain backside power distribution data, including: The thermoplastic transition temperature of the shear buffer copper strip is divided into temperature intervals to obtain the center target temperature and the edge target temperature; Performing interpolation calculation on the center target temperature and the edge target temperature according to the transverse width of the saw street to obtain a continuous temperature gradient curve; Power mapping processing is performed on the infrared radiation power according to the temperature gradient curve to obtain back surface power distribution data.

6. The wafer scribing process according to claim 4, wherein: The coupling correction processing of the tool feed rate and the tool rotation speed is performed based on the vibration amplitude data and the vibration main frequency data obtained in the plastic preheating zone to obtain the first feed rate and the first rotation speed, including: performing amplitude-threshold comparison processing on the vibration amplitude data obtained in the plastic preheating zone to obtain a feed rate adjustment coefficient; Performing main frequency-harmonic deviation calculation processing on the vibration main frequency data obtained in the plastic preheating zone to obtain a speed adjustment coefficient; A joint correction process is performed on the original feed rate and the original speed according to the feed rate adjustment coefficient and the speed adjustment coefficient to obtain a first feed rate and a first speed.

7. The wafer scribing process according to claim 1, wherein: The wafer is subjected to a layered progressive deep cutting process according to the initial cutting saw kerf, the shear buffer copper tape is subjected to slip deformation under the action of the tool pulse load and is reset by the elastic action of the low dielectric constant dielectric layer to form a full-depth cutting saw kerf, including: Performing layered cutting depth planning processing on the initial cutting saw kerf depth data to obtain multiple cutting depth data that increase in sequence along the saw street direction and correspond to the combination of the thickness of the metal layer and the dielectric layer; Performing a first-level deep cutting process on the wafer according to the multiple-level cutting depth data to obtain first-level saw kerf displacement data that generates a sliding displacement in the transverse direction of the shear buffer copper strip; A rebound buffer process is performed on the wafer based on the first-level saw kerf displacement data, and the shear buffer copper tape is reset by utilizing the elastic effect of the low-k dielectric layer. Meanwhile, a deviation correction is performed on the next-level cutting depth data to obtain the corrected cutting depth data. The subsequent deep cutting process and rebound buffer process are cyclically executed according to the corrected cutting depth data until the silicon base surface is cut deep enough to obtain a full-depth cutting saw kerf.

8. The wafer scribing process according to claim 7, wherein: The layered cutting depth planning process is performed on the initial cutting saw kerf depth data to obtain multiple cutting depth data that increase in sequence along the saw street direction and correspond to the combination of the thickness of the metal layer and the dielectric layer, including: Performing layer thickness accumulation processing based on the metal layer thickness data and the dielectric layer thickness data to obtain a multi-layer stack cumulative depth sequence; Performing energy-depth matching processing on tool single pulse energy constraint data according to the multi-layer stack cumulative depth sequence to obtain a cutting depth increment sequence; The initial cutting kerf depth data is subjected to depth segmentation processing according to the cutting depth increment sequence to obtain multiple levels of cutting depth data.

9. The wafer scribing process according to claim 1, wherein: The full-depth cutting saw kerf is subjected to a nitrogen environment cooling treatment, the twin structure inside the shear buffer copper strip is stabilized and forms a mechanical constraint structure with the adjacent dielectric layer, and a bridge unit is formed on the sidewall of the saw kerf, including: According to the difference between the shear buffer copper strip twin stable temperature zone and the linear thermal expansion coefficient of the low dielectric constant dielectric layer, a double window setting process is performed on the nitrogen cooling stage to obtain the first constant temperature window parameter and the second temperature reduction gradient parameter; performing a constant temperature dwell process on the full-depth cutting saw kerf according to the first constant temperature window parameters, so that the twin interfaces inside the shear buffer copper strip are rearranged during the constant temperature dwell period and axial residual compressive stress data of the shear buffer copper strip is obtained; performing radial gradient cooling on the full-depth saw kerf according to the second temperature-lowering gradient parameter, so as to cause the shear buffer copper strip to undergo reversible lateral contraction and drive the adjacent low-k dielectric layer to produce nested rebound, thereby obtaining a coupled shrinkage hoop interface; A stress redistribution process is performed on the full-depth cutting saw seam based on the coupled shrinkage interface and the axial residual compressive stress data. During the stress redistribution process, the residual compressive stress of the shear buffer copper strip and the rebound elasticity of the low dielectric constant dielectric layer are used to achieve interlocking, forming a bridging unit that continuously extends along the side wall of the saw seam.

10. The wafer scribing process according to claim 9, wherein: The step of performing radial gradient cooling on the full-depth saw kerf according to the second temperature reduction gradient parameter to cause the shear buffer copper strip to undergo reversible lateral contraction and drive the adjacent low-k dielectric layer to produce nested rebound, thereby obtaining a coupled shrinkage interface, includes: Performing center-edge partitioning and setting processing on the nitrogen flow rate according to the second cooling gradient parameter to obtain center flow rate data and edge flow rate data; performing zoned nitrogen supply processing according to the central flow data and the edge flow data to obtain radial temperature gradient data; According to the radial temperature gradient data, a synchronous matching process is performed on the lateral shrinkage of the shear buffer copper strip and the rebound of the low dielectric constant dielectric layer to obtain a coupled shrinkage hoop interface.

Images