A through-cavity double-stage jet impact and three-stage composite turbulent flow uniform temperature liquid cooling heat dissipation cold plate

Abstract

The application discloses a through-cavity two-stage jet flow impact and three-stage composite disturbance uniform-temperature liquid cooling heat dissipation cold plate, and belongs to the technical field of high-power chip heat dissipation. The application comprises a sealingly connected upper cover plate assembly and lower bottom plate; the upper cover plate assembly is sealingly connected by an upper cover plate main body and a jet flow orifice plate, a manifold cavity is formed between the two, the manifold cavity is communicated with a liquid inlet and is provided with a branch manifold. The jet flow orifice plate is provided with jet flow orifices in the direction of the cavity, including a conical contraction nozzle directly above the chip and a straight hole away from the chip, to form a first-stage vertical jet flow. The lower bottom plate is provided with a three-stage composite disturbance array composed of a macroscopic disturbance structure, a base microstructure and a surface microgroove. The application can be scaled according to the size of the cold plate, when the size of the cold plate increases, the key parameters such as the convex distance, the jet flow orifice density and the cavity height ratio remain unchanged, only the number of structural units is increased, the same heat exchange mechanism can be realized, and the similar performance level can be maintained.

Description

Technical Field

[0001] This invention relates to the field of high-power chip heat dissipation technology, specifically to a homogeneous liquid cooling plate with a through-cavity dual-stage jet impact and three-stage composite turbulence for AI computing power chips, GPUs, TPUs and computing centers with a power output of 200 W to 3200 W, and is particularly suitable for ultra-thin space-constrained scenarios where the thickness of the 1U server cooling plate is ≤10 mm. Background Technology

[0002] In the field of liquid cooling for high-power chips, there is a long-standing technical problem that remains unsolved—the thermal boundary layer insulation blanket effect.

[0003] When coolant flows over a solid wall, the velocity of the first layer of liquid molecules adhering to the wall approaches zero due to the adhesion between the liquid molecules and the solid surface, as well as the no-slip boundary condition. The velocity gradually increases with distance from the wall, and the thin layer between zero velocity and the mainstream velocity is called the thermal boundary layer (also known as the viscous sublayer). This layer is only tens to hundreds of micrometers thick, but because liquids have low thermal conductivity (water has a thermal conductivity of approximately 0.6 W / (m·K), and copper has a thermal conductivity of approximately 400 W / (m·K)), heat must first pass through the thermal boundary layer to enter the flowing liquid. Even increasing the coolant flow rate will not fundamentally improve heat transfer efficiency if the boundary layer is not disrupted.

[0004] The thermal boundary layer arises from the frictional force (wall shear force) between the fluid and the solid wall. Due to the no-slip boundary condition, fluid molecules adhering closely to the wall are "stuck," and the flow velocity approaches zero, forming a "static layer." Within this static layer, the fluid is almost still, and heat can only pass through by conduction. However, water's thermal conductivity is only about 0.6 W / (m·K), far lower than copper's approximately 400 W / (m·K). Therefore, the static layer becomes the main bottleneck for heat transfer, i.e., the "insulating blanket effect." In other words, friction creates the static layer, and the static layer hinders heat exchange. This is a first principle in the field of liquid cooling.

[0005] The current mainstream solution for heat dissipation in computing centers is cold plate liquid cooling. As the power consumption of AI chips breaks through 1000W and moves towards 1500W and above, mainstream single-phase cold plate liquid cooling is approaching its performance limit.

[0006] The fundamental flaws of existing microchannel cold plates: First, there's the issue of blockage. The equivalent diameter of microchannels is typically 100–300 μm, but actual coolant circulation systems inevitably contain trace impurities (such as pipe debris, welding slag, corrosion products, etc.). These impurities easily accumulate at the microchannel inlet or bends, leading to partial or complete blockage. Once blockage occurs, the heat dissipation capacity of the blocked area drops sharply, forming localized hot spots, and in severe cases, causing the chip to burn out.

[0007] Second, there's the issue of high pump pressure / high energy consumption. According to the Hagen-Poiseuille law, at the same flow rate, the pressure drop in a flow channel is inversely proportional to the square of the channel diameter. Reducing the diameter of a microchannel from the millimeter level to the micrometer level (a reduction of approximately 10 times) would increase the pressure drop by about 100 times, requiring a high-pressure, high-power water pump.

[0008] Third, the "bloated" effect. Microchannel technology increases the heat transfer area by increasing the number of microchannels, but according to boundary layer theory, the increased solid wall area also increases the total area covered by the thermal boundary layer. Since the boundary layer is the main barrier to heat, the "area gain" of the microchannels is largely offset by the "boundary layer thermal resistance gain"—although the geometric area is large, most of it is wrapped by the boundary layer "insulation blanket," and the effective heat transfer area is much smaller than the geometric area.

[0009] Boundary layer limitations of traditional finned cold plates: Traditional finned heat exchange plates expand the heat transfer area by increasing fin height and density. However, according to boundary layer theory, as the coolant flows over the fin surface, the thermal boundary layer gradually thickens along the fin height. The boundary layer is thinner at the fin root, resulting in better heat transfer; however, the boundary layer thickens significantly at the fin top, forming a "thermal blanket" and drastically reducing heat transfer efficiency. Furthermore, dead zones easily form at the bottom of the gaps between fins, leading to extremely poor local heat transfer. Simply increasing the fin area without disrupting the boundary layer offers limited improvement in heat transfer performance, with measured heat transfer coefficients only reaching 10–30 kW / (m²·K).

[0010] Limitations of existing single-stage jet cooling plates: While existing single-stage jet cooling plates can achieve extremely high local heat transfer coefficients in the jet stagnation region, they suffer from the following inherent drawbacks: First, the "local high-temperature region not covered by jet impact" effect—heat transfer is weak in the areas between jets and around the jet; second, they lack an active heat diffusion mechanism; and third, they rely on micron-level nozzle arrays, resulting in complex manufacturing processes, high costs, and easy nozzle clogging. More importantly, the boundary layer disruption effect of jet cooling plates is limited to the stagnation point; once the fluid leaves the stagnation point, the boundary layer immediately re-develops and thickens.

[0011] Existing engineering challenges of two-phase boiling cold plates: While existing two-phase boiling cold plates can utilize the latent heat of vaporization to obtain a higher heat transfer coefficient, they suffer from problems such as difficulty in controlling the initial superheat of boiling, easy formation of gas film due to bubble merging leading to a low critical heat flux density, and instability of two-phase flow, which limit their reliable application in heat dissipation of high-power chips.

[0012] Common defects of existing technologies: Analysis reveals that existing mainstream liquid cooling solutions share the following common problems in boundary layer management: microchannel cold plates rely on reducing the channel diameter to compress the boundary layer thickness, but compression does not equal destruction, and the boundary layer continues to thicken along the channel direction; finned cold plates have a boundary layer that gradually thickens along the fin height, making active destruction impossible; single-stage jet cold plates destroy the boundary layer in the stagnation region, but the boundary layer immediately re-develops after leaving the stagnation point; and two-phase boiling cold plates present the engineering challenge of boiling instability. The common problem with these solutions is that boundary layer destruction is either "one-off" or "localized," failing to achieve continuous boundary layer interruption across the entire flow channel.

[0013] Furthermore, existing jet impact cold plates all employ a single-stage jet structure, lacking a mechanism to actively diffuse localized high heat flux across the entire cold plate area. Some high-end solutions use an independent VC vapor chamber as an intermediate layer, but this incurs drawbacks such as additional thickness, multi-layer interface thermal resistance, slow response, and low reliability.

[0014] Therefore, how to continuously disrupt the thermal boundary layer, eliminate flow dead zones, and achieve efficient heat exchange under the dual constraints of a thickness of 10 mm and a heat flux density of over 200 W / cm², while avoiding blockage and high pump consumption, is a technical challenge that urgently needs to be solved in this field.

[0015] Teardrop-shaped turbulence columns offer advantages such as low drag and strong vortex generation in unidirectional flow channels, making them one of the most energy-efficient turbulence structures for straight flow channels. However, their direct application in complex flow channels employing vertical jets and multidirectional crossflow presents challenges, including flow direction mismatch leading to a sharp increase in drag, difficulty in suppressing tail vortices and sustaining boundary layer disruption, and the potential formation of flow dead zones downstream. Therefore, providing a flow direction-insensitive turbulence structure that can sustainably disrupt the boundary layer and avoid dead zones is a pressing technical problem to be solved in this field. Summary of the Invention

[0016] This invention proposes a homogeneous liquid-cooled heat dissipation plate with a through-cavity dual-stage jet impact and a three-stage composite turbulence, aiming to solve the following technical problems: the difficulty in continuously damaging the thermal boundary layer in ultra-thin spaces; the inability of a single turbulence structure to eliminate flow dead zones; the limit of the highest heat transfer performance achievable under single-phase (without boiling or condensation) convective heat transfer; the problems of blockage and high pump consumption in existing microchannel cold plates; the problem of "local high-temperature areas not covered by jet impact" in existing single-stage jet cold plates; the problem of boundary layer thickening along the height in existing finned cold plates; and the problem of unstable boiling in existing two-phase boiling cold plates.

[0017] The cold plate of this invention is formed by welding and sealing an upper cover plate assembly and a lower base plate together. The upper cover plate assembly consists of an upper cover plate body and a jet orifice plate sealed together, forming a manifold cavity between them. The manifold cavity is connected to the liquid inlet and has multiple branch manifolds according to the size of the cold plate. A sealed, internally continuous cavity without any dividing flow channels is formed between the upper cover plate assembly and the lower base plate, in which the coolant can flow freely. On the bottom surface inside the cavity, several irregularly arranged pyramid-shaped, conical, and cylindrical macroscopic turbulence structures are arranged at different heights, forming a three-level composite turbulence array together with the substrate microstructure and surface microgrooves.

[0018] This invention differs fundamentally from existing technologies: Existing microchannel cold plates use closed, slender, parallel micron-level flow channels, which forcibly divide the coolant into independent channels, much like a vehicle being restricted to a narrow one-way street, unable to flow freely. The boundary layer continues to thicken along the flow channel direction, resulting in a decrease in heat exchange efficiency.

[0019] Although existing single-stage jet cooling plates can achieve a high local heat transfer coefficient in the jet stagnation area, their effect is limited to a point area. After leaving the stagnation point, the boundary layer immediately recovers, just like a water gun only sprays one spot, while other places remain dry. This has the defect of "local high-temperature areas that cannot be covered by the jet impact".

[0020] This invention employs a sealed, through-cavity design, allowing coolant to flow freely throughout the space. Combined with a dual-stage jet (vertical impact targeting high heat flux density areas and global lateral washing along the base plane) and a three-stage composite turbulence array (continuous stirring to disrupt the boundary layer), it forms a three-dimensional heat exchange system of "open water + full-area spray + continuous stirring," fundamentally solving the two major problems of continuous boundary layer disruption and localized high-temperature areas not covered by the jet impact.

[0021] Furthermore, the present invention has the following beneficial effects: (1) This invention achieves a dimensional leap from two-dimensional planar turbulence to three-dimensional layer breaking through the synergistic effect of a three-stage composite turbulence array and a two-stage jet, so that the wall surface is always in a highly efficient heat exchange state in the thermal development stage. The heat transfer coefficient is measured to be 400-500 kW / (m²·K), which is 8-10 times that of traditional microchannels.

[0022] (2) Supports single-phase and two-phase dual-mode physical adaptive switching. In the two-phase mode, the heat transfer coefficient reaches 400-500 kW / (m²·K) and the critical heat flux density is ≥500 W / cm².

[0023] (3) The present invention adopts a sealed through cavity (millimeter-level flow channel), which is fundamentally different from the micron-level closed flow channel of the microchannel cold plate: conventional impurity particles in the coolant can pass freely, fundamentally eliminating the risk of blockage; the pressure drop is only 21 kPa (the pressure drop of the microchannel cold plate is usually 50-100 kPa under the same flow rate), the heat transfer coefficient is increased by more than 4 times, and the energy efficiency ratio is better.

[0024] (4) The present invention achieves “increased area without increasing boundary layer”: the effective heat exchange area is expanded by 50% to 150% through pyramid, cone and cylindrical protrusions, and the sidewall of the protrusion is in the mainstream continuous scouring area, so the boundary layer cannot be stably thickened, thus achieving “real increase” of effective heat exchange area rather than “false fat”.

[0025] (5) Dual-stage jet synergistic effect: The first-stage vertical jet achieves hot spot impact cooling, and the second-stage horizontal jet actively diffuses heat to the surrounding area, eliminating the "local high temperature area that the jet impact cannot cover", and the chip surface temperature difference is ≤0.5 ℃.

[0026] (6) The triple driving force of the guide trench (capillary suction, thermal buoyancy, and mainstream shear drag) reduces the thickness of the base boundary layer by 40% to 60%, while guiding the water flow upward, significantly increasing the turbulence effect. Among them: capillary suction actively draws the thermal boundary layer liquid near the base into the trench; thermal buoyancy causes the drawn-in hot liquid to generate an upward buoyancy force due to its low density, guiding the water flow upward; mainstream shear drag causes the mainstream coolant at the top of the trench to quickly carry away the discharged hot liquid, forming a continuous circulation.

[0027] (7) Short path heat exchange design: vertical jet + liquid outlet on both sides, short flow path (5-15 mm), avoids heat accumulation and further reduces pressure loss.

[0028] (8) Simple manufacturing process and high yield: The split open processing can adopt conventional CNC, precision casting or etching processes, without the need for MEMS micro-nano processing or 3D printing. The manufacturing cost is reduced by more than 50%, and the yield can reach more than 95%.

[0029] (9) It can completely replace the independent VC heat exchanger, eliminate the thermal resistance of the multi-layer interface, and achieve integrated heat dissipation.

[0030] (10) The chip bonding and positioning groove is combined with the layered thermal conductive interface material, which reduces the interface thermal resistance and absorbs thermal stress.

[0031] (11) Microbubble guide grooves or gas collection microcavities actively guide the gas phase medium to avoid gas blockage.

[0032] (12) An integrated thin-film thermocouple array can identify boiling critical precursors based on temperature gradient and temperature pulsation, providing early warning 3 to 5 seconds in advance and effectively preventing chip thermal burn-out.

[0033] (13) Excellent durability, with thermal resistance increase not exceeding 5% of the initial value and pressure drop increase not exceeding 10% after 1000 hours of continuous operation.

[0034] (14) Computing Power Release Benefits: This invention controls the chip temperature at 67-79℃, always below the 85℃ throttling threshold, and can maintain peak computing power of over 95% for a long time, which is 10%-20% higher than the traditional solution (75-85%). Taking an AI training cluster with 10,000 GPUs as an example, the annual comprehensive computing power revenue can reach tens of millions of yuan, and the equipment investment payback period is less than 6 months. Currently, AI computing power centers urgently need heat dissipation of over 100kW per rack, but existing cold plate liquid cooling solutions generally trigger throttling when the power consumption is above 1500W / chip. This invention can maintain peak computing power of over 98% for a long time at 1500W power consumption, solving the bottleneck problem restricting the power consumption improvement of computing power centers, and has significant economic and social benefits.

[0035] (15) Independent closed-loop liquid cooling system does not rely on external water system and is plug-and-play; pulse pump mode increases heat transfer coefficient by 10% to 20%; distributed system has advantages such as fault isolation, flexible expansion, customized control and convenient deployment.

[0036] (16) The structural parameters of the present invention can be scaled proportionally according to the size of the cold plate. When the size of the cold plate increases, the key parameters such as the protrusion spacing, jet hole density, and cavity height ratio remain unchanged. Only the number of structural units is increased, and the same heat exchange mechanism can be achieved and the performance level can be maintained. Therefore, the scope of protection of this patent covers all cold plates that adopt this structural scheme. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the overall structure of the cold plate of the present invention, used to show the upper cover plate assembly, the lower bottom plate, and the liquid inlet and outlet.

[0038] Figure 2 This is another overall structural diagram of the cold plate of the present invention, used to show the upper cover plate assembly, the lower bottom plate, the liquid outlet, and the chip bonding and positioning groove.

[0039] Figure 3 This is an exploded view of the cold plate of the present invention, used to show the upper cover plate assembly, the lower bottom plate, and the liquid outlet.

[0040] Figure 4 for Figure 3 The enlarged schematic diagram at point A in the diagram is used to show the bottom plate, macroscopic turbulence structure, flow channel, substrate microstructure, and surface microgroove.

[0041] Figure 5 This is a schematic cross-sectional view of the overall structure of the cold plate of the present invention, used to show the upper cover plate assembly, lower bottom plate, manifold cavity, conical shrink nozzle, liquid inlet, sealed through cavity, jet channel and side spray hole.

[0042] Figure 6 This is a frontal cross-sectional view of the cold plate of the present invention, used to show the upper cover plate assembly, jet orifice plate, lower bottom plate, manifold cavity, equal diameter straight hole, conical constriction nozzle, liquid inlet and outlet, sealed through cavity, and jet channel.

[0043] Figure 7 This is a cross-sectional view of the cold plate assembly of the present invention, showing the column-like structure of the jet column extending upward from the bottom plate, with a side spray hole provided at its bottom end near the lower bottom plate.

[0044] Figure 8 This is a cross-sectional front view of another assembled cold plate of the present invention, used to show the hanging structure in which the jet column extends downward from the upper cover plate assembly, and a transverse jet gap is provided between its bottom end and the lower base plate.

[0045] Figure 9 This is a cross-sectional schematic diagram showing the separation of another component of the cold plate of the present invention, used to illustrate the hanging structure in which the jet column extends downward from the upper cover plate assembly, and the installation method of the upper cover assembly and the bottom plate.

[0046] Figure 10 This is a cross-sectional schematic diagram of an upper cover plate assembly of the cold plate of the present invention, used to show the hanging structure in which the jet column extends downward from the upper cover plate assembly, and a transverse jet gap is provided between its bottom end and the lower base plate.

[0047] Figure 11 This is an exploded view of the upper cover assembly of the present invention, used to show the upper cover body, manifold cavity, branch manifold, and jet hole.

[0048] Figure 12 This is another exploded view of the upper cover plate assembly of the present invention, used to show the upper cover plate body, jet orifice plate, manifold cavity, branch manifold, conical constriction nozzle, equal diameter straight hole, and jet column.

[0049] Figure 13 This is a schematic diagram of the chip bonding and positioning groove structure of the present invention, used to illustrate the positioning groove and the flexible thermal pad.

[0050] Figure 14 This is a block diagram of the liquid cooling system of the present invention, used to illustrate the architecture of the independent closed-loop system and the distributed system.

[0051] Figure 15 This is a schematic diagram of the flow path of the coolant and heat transfer inside the cold plate of the present invention.

[0052] Figure 16 This is a comparison diagram of the temperature distribution on the bottom plate surface of the cold plate of the present invention (Example 1) and a commercially available single-phase microchannel cold plate.

[0053] Figure 17 This is a bar chart comparing the performance of the cold plate of the present invention (Example 1) with various commercially available mainstream cold plates.

[0054] Marker description Detailed Implementation The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Please see Figure 1-14 As shown, the present invention provides a technical solution: a homogeneous liquid-cooled heat dissipation plate with a through-cavity dual-stage jet impact and a three-stage composite turbulence, comprising: An upper cover plate assembly is sealed to a lower base plate, forming a sealed, internally connected cavity without any partition channels, that is, a sealed through cavity (3); coolant can flow freely in it; the coolant in the cavity (3) is discharged through the outlet connected to the side. The upper cover assembly is provided with a liquid inlet (11) and a manifold cavity (12), and the manifold cavity (12) is connected to the liquid inlet (11); The upper cover plate assembly has several jet holes (13) that communicate with the manifold cavity on the side facing the cavity (3) to form a first-stage vertical jet. The lower base plate (2) is provided with a three-level composite turbulence array on the surface facing the cavity. The three-level composite turbulence array includes a macroscopic turbulence structure (21). The macroscopic turbulence structure (21) can be used as the first-level composite turbulence array. The cavity (3) is provided with at least one jet column (5), and the jet column (5) is provided with a vertical jet channel (211), the top of the jet channel (211) is connected to the manifold cavity (12); The bottom end of the jet column (5) is provided with a fluid deflection structure to form a second-stage transverse jet.

[0056] Furthermore, the fluid deflection structure is a radially penetrating lateral nozzle (212) located at the bottom end of the jet column; the jet column (5) extends upward from the surface of the lower base plate.

[0057] Specifically, the fluid deflection structure is a transverse jet gap (214) between the bottom end of the jet column and the inner wall of the lower base plate; the jet column (5) extends downward from the upper cover plate assembly, and a gap is left between its bottom end and the inner wall of the lower base plate, so that the coolant is sprayed transversely from the gap to all sides.

[0058] Specifically, the upper cover plate assembly is formed by sealing the upper cover plate body (1) and the jet orifice plate (1a) together, forming a manifold cavity (12) between them; the jet orifice (13) is opened on the jet orifice plate (1a).

[0059] Specifically, the three-stage composite turbulence array further includes a substrate microstructure (22), which can serve as a second-stage composite turbulence array; the substrate microstructure is located in the substrate region between the macroscopic turbulence structures (21).

[0060] Specifically, the three-stage composite turbulence array also includes surface microgrooves (23), which can serve as the third-stage composite turbulence array. The surface microgrooves (23) are formed on the surface of the bottom plate (2).

[0061] Specifically, the jet orifice includes a first type of jet orifice (13a) located in the region directly above the chip and a second type of jet orifice (13b) located in the region away from the chip. The first type of jet orifice is a tapered converging nozzle structure, and the second type of jet orifice is a straight hole structure with equal diameter.

[0062] Specifically, the macroscopic disturbance structure (21) includes at least one of pyramid-shaped protrusions, conical protrusions, cylindrical protrusions, and fin-shaped protrusions, which are non-uniformly distributed and have a height difference between adjacent macroscopic disturbance structures.

[0063] Specifically, the sidewall of the macroscopic turbulence structure (21) is provided with one or more flow-guiding grooves (213) extending along the height direction to the base plane, and the bottom of the flow-guiding grooves (213) is in fluid communication with the base microstructure (22).

[0064] Specifically, the bottom plate (2) has a positioning groove (26) for bonding with the chip on the bottom surface of the top cover assembly, and the inner surface of the positioning groove (26) is provided with a layered thermal interface material (27).

[0065] Specifically, the liquid outlet (14) is located on both sides of the cold plate near the bottom plate (2), forming a layout with the liquid inlet (11) in the middle at the top and liquid outlet on both sides at the bottom.

[0066] Specifically, the upper cover plate assembly and the lower base plate (2) are sealed together by diffusion welding or vacuum welding, and the overall airtightness leakage rate of the cold plate after welding is less than 1×10⁻. 8Pa·m³ / s.

[0067] See Figure 14 As shown, preferably, a liquid cooling heat dissipation system includes a cold plate as described in any of the above claims, and a coolant circulation pump, a heat exchanger, and pipelines connected to the cold plate; the system can form an independent closed-loop circulation circuit, independent of an external water system; the coolant circulation pump is a pulse pump, generating a pulsed coolant flow rate; multiple systems can be connected in parallel to form a distributed liquid cooling system, with each system operating independently and without interference.

[0068] Furthermore, the liquid cooling system also includes a water tank and a controller. The controller adjusts the frequency and duty cycle of the pulse pump according to the chip temperature to achieve on-demand cooling.

[0069] Preferably, a server includes at least one cold plate as described in any of the preceding claims, and a computing chip attached to the cold plate.

[0070] Preferably, a chip packaging assembly includes at least one cold plate as described in any one of the above, a chip packaging substrate, and a chip die, wherein the cold plate is directly attached to the back of the chip die.

[0071] Preferably, a chip-level embedded heat dissipation structure includes the back side of a chip substrate or the surface of a chip package cover plate, wherein the back side of the chip substrate or the surface of the chip package cover plate is integrated with the three-level composite turbulence array described above.

[0072] The invention will be further described below with reference to specific implementations: Example 1: Complete flagship version (vertical jet column) This embodiment is a complete cold plate designed for AI chips (4) of 1500 W and above, integrating all the innovative features of this invention. In this embodiment, the jet column adopts a column-type structure extending upward from the bottom plate, with side spray holes at the bottom end.

[0073] See Figure 5 and Figure 11As shown, the main body (1) of the upper cover plate is made of oxygen-free copper with a thickness of 3 mm. The bottom surface of the upper cover plate is formed by CNC milling to form a manifold cavity (12) and multiple branch manifolds (12a). The branch manifolds have a tree-like branching structure to ensure uniform distribution of coolant. The jet orifice plate (1a) is 0.5 mm thick and made of oxygen-free copper. The jet holes (13) are formed by micro-drilling and are arranged in sections: the first type of jet hole (13a) is set in the area directly above the chip. It is a conical constricting nozzle structure with an inlet diameter of 0.3 mm, an outlet diameter of 0.1 mm, a contraction angle of 20°, and a distribution density of 3.5 holes / mm². The second type of jet hole (13b) is set in the area away from the chip. It is a straight hole structure with a diameter of 0.3 mm and a distribution density of 1.2 holes / mm². The main body (1) of the upper cover plate and the jet orifice plate (1a) are sealed together by diffusion welding. The welding temperature is 850℃ and the welding pressure is 5MPa to form an upper cover plate assembly with a manifold cavity (12).

[0074] See Figure 3 , Figure 4 , Figure 5 , Figure 6 , Figure 7 As shown, the bottom plate (2) is made of oxygen-free copper with a thickness of 4 mm and a three-level composite turbulence array on its surface. The macro-turbulence structure (21) is divided into zones: the central zone is dominated by pyramid-shaped protrusions with a spacing of 0.25 mm, accounting for 80%; the transition zone is dominated by conical protrusions with a spacing of 0.35 mm, accounting for 70%; and the edge zone is dominated by cylindrical protrusions with a spacing of 0.6 mm, accounting for 80%. The three types of protrusions are arranged periodically along the flow direction and their heights are staggered: the height of the pyramid-shaped protrusions is 0.8 mm, the height of the conical protrusions is 0.5 mm, and the height of the cylindrical protrusions is 0.3 mm. The height difference between adjacent protrusions is ≥0.3 mm. The overall height of the macro-turbulence structure (21) gradually changes from 0.8 mm to 0.2 mm along the flow direction. The base microstructure (22) is a micro-positioning groove with a depth of 0.03 mm and a width of 0.03 mm, forming a 45° angle with the coolant flow direction. The surface microgrooves (23) are 0.01 mm wide and 0.008 mm deep, and are arranged in a biomimetic fish scale pattern.

[0075] In this invention, the "three-level composite turbulence array" refers to a composite system that simultaneously includes three levels of turbulence structures: macroscopic turbulence structure (millimeter level), substrate microstructure (micrometer level), and surface microgroove (micrometer level). The three levels act on the outer, middle, and inner layers of the thermal boundary layer, respectively, forming a step-by-step destruction mechanism.

[0076] "Two-stage jet" refers to a two-stage jet system consisting of a first-stage vertical jet generated by a jet orifice and a second-stage transverse jet generated by a jet column. The two stages complement each other in space and take turns in time to jointly achieve global boundary layer disruption and active temperature isothermization.

[0077] Multiple cylindrical macroscopic turbulence structures (21) are selected as jet columns (5) in the hot spot area of ​​the cold plate corresponding to the chip layout area and thickened (diameter 2.0 mm). They are arranged in a 2×2 array to cover the corresponding area of ​​the chip. Each jet column (5) is provided with a vertical through jet channel (211) with a diameter of 0.8 mm. The top of the channel is connected to the manifold cavity (12). The bottom end of the jet column (5) is provided with a radially penetrating lateral spray hole, or a transverse jet gap is left between it and the inner wall of the bottom plate to form a second-stage transverse jet. The lateral spray hole (212) has a diameter of 0.3 mm. The two ends of the through hole open on the side wall of the cylinder to form two lateral spray holes (212) in opposite directions. The direction of the spray hole is parallel to the base plane.

[0078] On the surface of all pyramidal and conical macro-turbulence structures (21), flow-guiding grooves (213) with a width of 30 μm and a depth of 15 μm are machined. The grooves have a V-shaped cross section and extend to the base plane at the bottom, which is in fluid communication with the base microstructure (22).

[0079] The substrate surface of the bottom plate (2) is provided with an anti-corrosion and anti-scaling micro-treatment layer. The surface of the turbulence array is prepared by anodizing to form a nanoporous copper oxide boiling nucleation structure with a pore size of 50-200 nm, a porosity of 50%, and a thickness of 3 μm. The top of the turbulence array is coated with a titanium dioxide-based hydrophilic micro-nano composite coating with a thickness of 2 μm and a water contact angle of 25°.

[0080] See Figure 13 As shown, a chip bonding and positioning groove (26) with a depth of 0.15 mm is formed on the bottom surface of the lower base plate (2). The inner wall of the chip bonding and positioning groove (26) is provided with a thermally conductive interface material (27), which is usually composed of a 0.1 mm thick flexible thermal pad and a 2 μm thick CVD diamond film. The lower base plate (2) integrates a thin-film thermocouple array with a sampling rate of 100 Hz. In actual use, the chip 4 is held in the chip bonding and positioning groove (26).

[0081] The upper cover assembly and the lower base plate (2) are sealed together by diffusion welding at a temperature of 850°C, a welding pressure of 5 MPa, and a vacuum degree of 1×10⁻³ Pa. After welding, the airtightness leakage rate is less than 1×10⁻³ Pa. 8 Pa·m³ / s. A sealed through cavity with a height of 5 mm is formed between the upper cover assembly and the lower base plate (2). The liquid outlet (14) is located on both sides of the cold plate near the lower base plate, forming a layout with the liquid inlet (11) in the middle at the top and liquid outlet on both sides at the bottom.

[0082] With this design, the coolant is sprayed at high speed onto the chip (4) through the first type of jet hole (13a), the second type of jet hole (13b) and the vertical jet channel (211) of the jet column (5) on the main body of the upper cover plate (1), thereby performing instantaneous heat exchange; then through the side spray hole (212) near the bottom on the jet column (5), the coolant rushes out horizontally at high speed, and the coolant after heat exchange is quickly cleared along the surface of the lower base plate, and continuously contacts and collides with the macroscopic turbulence structure (21), the substrate microstructure (22) and the surface microgroove (23) set on the lower base plate (2). When the coolant encounters the macroscopic turbulence structure (21) during the horizontal flow, it climbs up the slope along the V-shaped guide groove (213) on the surface, is discharged, and changes the flow direction; when it passes through the substrate microstructure (22) and the surface microgroove (23), the flow rate changes.

[0083] In this way, the edge layer formed between the coolant and the bottom plate (2) and the surface of the flow area can be continuously broken, thereby accelerating the flow rate and heat exchange efficiency of the coolant, greatly improving the heat dissipation efficiency of the chip and saving energy.

[0084] Test conditions: ambient temperature 25 ℃, coolant is pure water, system working pressure 0.3 MPa, coolant flow rate 1.2 L / min, simulating a 1500 W AI chip (4).

[0085] Test results: In single-phase mode, the highest temperature of chip (4) was 79 ℃, the surface temperature difference was 0.6 ℃, the flow channel pressure drop was 12 kPa, and the local peak heat transfer coefficient was 188 kW / (m²·K). In two-phase mode, the highest temperature of chip (4) was 67 ℃, the surface temperature difference was 0.3 ℃, the flow channel pressure drop was 21 kPa, the heat transfer coefficient was 480 kW / (m²·K), the measured critical heat flux density was 528 W / cm², and there was no gas blockage. The active heat equalization effect of the two-stage jet was obvious, the overall temperature distribution of the cold plate was uniform, and the jet overlap area of ​​the jet column (5) produced a superposition enhancement effect. The first type of jet hole (13a) is a conical constriction nozzle, which increases the jet velocity in the hot spot area by about 3 times and the local heat transfer coefficient is increased by about 35% compared with the straight hole scheme.

[0086] The triple driving forces of the flow channel (213) were verified by microscopic PIV flow field testing: capillary suction reduced the thickness of the substrate boundary layer from about 80 μm to about 30 μm, a reduction of 62%; thermal buoyancy effect caused the upward flow velocity of the liquid in the channel to reach 0.3–0.8 m / s; and mainstream shear drag caused the discharged hot liquid to be carried away by the mainstream within 0.1 seconds. The synergistic effect of the triple driving forces increased the heat transfer coefficient of the substrate region by 25%–35%.

[0087] This embodiment can replace the external independent VC heat spreader and is suitable for ultra-high power AI main chips of 1500 W and above (4).

[0088] Example 2: Suspended jet column plate The difference between this embodiment and Embodiment 1 lies in the installation method of the jet column and the fluid diversion structure. In this embodiment, the jet column extends downward from the upper cover plate assembly and adopts a hanging structure, with a transverse jet gap (214) between the bottom end and the lower base plate (2).

[0089] The processing of the top cover assembly is the same as in Example 1. In the hot spot area of ​​the cold plate corresponding to the chip (4), the jet column (5) extends downward from the top cover assembly (diameter 2.0 mm, height 4.5 mm) in a 2×2 array to cover the corresponding area of ​​the chip. Each jet column has a vertical through jet channel (211) with a diameter of 0.8 mm inside, and the top of the channel is connected to the manifold cavity (12). A 0.5 mm transverse jet gap (214) is left between the bottom end of the jet column (5) and the inner wall of the bottom plate (2), without side spray holes. A fluid deflection structure is formed, and the coolant enters the jet channel (211) from the manifold cavity and is sprayed laterally from the annular gap at the bottom to form a transverse jet, thereby realizing fluid deflection.

[0090] The rest of the structure is the same as in Example 1.

[0091] Test conditions: Same as in Example 1.

[0092] Test results: In two-phase mode, the highest temperature of chip (4) was 68 ℃, the surface temperature difference was 0.4 ℃, the flow channel pressure drop was 20 kPa, the heat transfer coefficient was 470 kW / (m²·K), the measured critical heat flux density was 515 W / cm², and there was no gas blockage. The annular jet generated by the transverse jet gap makes the heat diffusion more uniform.

[0093] This embodiment is applicable to ultra-high power AI main chips of 1500 W and above (4), and is especially suitable for scenarios with special requirements for installation space.

[0094] Example 3: Comparative Experiment Two existing mainstream cold-rolled steel plates were selected as controls to verify the technical advantages of the present invention under the same working conditions.

[0095] Comparison object A: Single-scale rectangular fin jet cold plate, rectangular fins are arranged at the same height, without secondary or tertiary microstructures, and without heterogeneous wettability and chip bonding positioning groove (26) structure.

[0096] Comparison Object B: Commercially available conventional single-phase microchannel cold plate, without two-phase boiling structure, and using traditional thermal grease mounting.

[0097] Analysis of coolant flow path and heat transfer mechanism See Figure 15The figure illustrates the synergistic mechanism of the first-stage vertical jet, the second-stage transverse jet, the third-stage composite perturbation array, and the boundary layer's "development-destruction-redevelopment" process. Specifically: The first-stage vertical jet impacts the area directly above the chip from the jet hole in the upper cover plate, achieving the first disruption of the boundary layer in the stagnation zone; the second-stage transverse jet is sprayed horizontally in all directions from the side nozzles or transverse gaps at the bottom of the jet column, spreading along the base plane of the lower base plate.

[0098] In the three-stage composite turbulence array, pyramidal, conical, and cylindrical protrusions are staggered in height, and the sidewalls of the protrusions are equipped with flow-guiding grooves. The thermal boundary layer is relatively thin (approximately 5 μm) in the vertical jet stagnation region; on the front side of the protrusion, the boundary layer is torn apart again (approximately 15 μm thick); on the back side of the protrusion, although the boundary layer redevelops (approximately 50 μm thick), it is immediately destroyed again in the next jet impact zone. Heat is transferred from the bottom of the chip sequentially through the lower substrate and the sidewalls of the macroscopic turbulence structure to the mainstream coolant.

[0099] Temperature distribution uniformity analysis See Figure 16 The surface temperature distribution of the bottom plate of the cold plate of this invention (Example 1) was compared with that of a commercially available single-phase microchannel cold plate using CFD simulation. The simulation conditions were: inlet flow rate 1.2 L / min, inlet water temperature 25℃, and chip heat source 1500 W.

[0100] The results showed that the highest temperature of commercially available single-phase microchannel cold plates reached 112℃, with concentrated hot spots and a temperature difference of over 30℃ from the edge, resulting in poor temperature uniformity. In contrast, the cold plate of this invention reduced the highest temperature to 67℃, with uniform overall temperature distribution and a surface temperature difference controlled within 0.5℃, eliminating obvious concentrated hot spots.

[0101] Comprehensive performance comparison analysis See Figure 17 The chart shows the internal experimental test data (test conditions: inlet water temperature 25℃, flow rate 1.2 L / min).

[0102] like Figure 17 As shown, the local peak heat transfer coefficient of this invention reaches 480 kW / (m²·K) in two-phase mode, far exceeding the 40 kW / (m²·K) of the microchannel cold plate. The test method refers to GB / T 8446-2022 "Heat Discharge Heatsinks for Power Semiconductor Devices". Figure 17 As shown, the pressure drop of this invention is only 21 kPa, which is slightly higher than that of a single-stage jet cooling plate (18 kPa), but much lower than that of a microchannel cooling plate (95 kPa). Figure 17 As shown, the thermal resistance of this invention is reduced to 0.18 (cm²·K) / W, which is significantly better than the prior art. For example... Figure 17As shown, the critical heat flux density (CHF) of this invention is as high as 528 W / cm², which is a significant improvement over the prior art.

[0103] Test conditions: ambient temperature 25 ℃, coolant is pure water, system working pressure 0.3 MPa, coolant flow rate 1.2 L / min, chip (4) power consumption 1500 W.

[0104] Experimental analysis: This invention increases the pressure drop by only 17%, improves the heat transfer coefficient by more than 4 times, reduces the maximum temperature of chip (4) by 45 ℃, significantly optimizes temperature uniformity, and increases the critical heat flux density by nearly 1.5 times. Chip (4) does not experience overheating and frequency reduction, and its computing power utilization rate far exceeds that of existing products.

[0105] Referring to the comparative test in Example 3, under the same operating conditions (1500W, 1.2L / min), the heat transfer coefficient of the present invention (480kW / (m²·K)) is 12 times that of the commercially available single-phase microchannel cold plate (40 kW / (m²·K)), and the maximum chip temperature is reduced by 45°C. This significant technical effect far exceeds the reasonable expectations of those skilled in the art, further demonstrating the inventiveness of the technical solution of the present invention.

[0106] Example 4: Economic Version This embodiment is applicable to low-to-medium power chips (4) with a power consumption of 500-1000 W, and simplifies the structure to reduce processing costs. Compared to embodiment 1: The boiling nucleation structure and chip bonding positioning groove (26) are cancelled, and the chip is mounted using traditional thermal grease (4); the upper cover assembly partition jet hole (13), the lower base plate (2) macroscopic turbulence structure (21), the substrate microstructure (22), the dual-stage jet structure, and the side wall guiding groove (213) are retained.

[0107] Test conditions: ambient temperature 25 ℃, coolant is pure water, system working pressure 0.3 MPa, coolant flow rate 0.8 L / min, chip (4) power consumption 800 W.

[0108] Test results: The highest temperature of chip (4) was 86 ℃, the surface temperature difference was 1.3 ℃, the flow channel pressure drop was 10 kPa, and the local peak heat transfer coefficient was 100 kW / (m²·K).

[0109] This embodiment is applicable to cost-sensitive scenarios such as consumer-grade GPUs and edge computing nodes.

[0110] Example 5: Multi-chip parallel version This embodiment is a dedicated cold plate for multi-chip modules, and a single cold plate can simultaneously cool 4 computing chips (4).

[0111] The bottom plate (2) has four independent chip fitting and positioning grooves (26) on its bottom surface. Each chip (4) has an independently designed jet hole (13) and macroscopic turbulence structure (21) in its corresponding area. The manifold cavity (12) adopts a flow equalization design, and each flow channel is arranged in parallel. Each chip hot spot area is equipped with a 2×2 array jet column. The heterogeneous wettability coating, boiling nucleation structure, and thin film thermocouple array are retained.

[0112] Test conditions: ambient temperature 25 ℃, system working pressure 0.4 MPa, coolant flow rate 2.5 L / min, power consumption of a single chip (4) 800 W, total power consumption 3200 W.

[0113] Test results: The highest temperature of the four chips (4) was 74-81 ℃, the temperature difference between chips was ≤3 ℃, and the total voltage drop was 27 kPa.

[0114] This embodiment is applicable to multi-chip packaging, Chiplet integration, and high-performance AI modules.

[0115] Example 6: Liquid Cooling System This embodiment provides an integrated liquid cooling heat dissipation system, which includes four cold plates as described in Embodiment 1, arranged in parallel, and equipped with a coolant circulation pump, a plate heat exchanger and connecting pipelines.

[0116] Test conditions: System inlet water temperature 40 ℃, total flow rate 4.8 L / min, power consumption of a single chip (4) 1500 W, total power consumption 6000 W.

[0117] Test results: The system runs stably and meets the total power consumption and heat dissipation requirements. The system voltage drop is 28 kPa and the overall PUE value is ≤1.08.

[0118] Example 7: Server This embodiment is a 2U server equipped with two computing chips (4) and two cold plates as described in embodiment 3. The cold plates are tightly attached to the computing chips (4).

[0119] Test conditions: Ambient temperature 35 ℃, continuous operation for 720 hours.

[0120] Test results: The temperature of chip (4) remained stable at 75-80 ℃, with no frequency reduction. The overall power consumption of the server was reduced by 18% compared to the traditional air-cooling solution.

[0121] Example 8: Chip Packaging Assembly This embodiment is an integrated chip packaging assembly, consisting of a chip packaging substrate, a bare chip (4), and a cold plate as described in Embodiment 1. The cold plate is directly attached to the back of the bare chip (4), eliminating the need for an independent thermal interface material (27).

[0122] Test results: The overall thermal resistance of the component is reduced by 35% compared to traditional packaging solutions.

[0123] Example 9: Boiling Critical Warning Method This embodiment is a boiling critical warning method for cold plates. Five temperature measuring points are arranged on the bottom plate (2) to collect temperature data in real time and calculate the temperature gradient and temperature fluctuation amplitude.

[0124] Warning logic: When the wall temperature is ≥85℃ and the temperature fluctuation amplitude is ≥±2℃ / s for more than 0.2 seconds, it is determined to be a precursor to boiling and a warning signal is output to the server management unit.

[0125] Test results: This method can identify boiling critical risk 3 to 5 seconds in advance.

[0126] Example 10: Independent Liquid Cooling System This embodiment provides an independent liquid cooling system, comprising a cold plate as described in Embodiment 1, a pulse pump (61), a water storage tank (62), and a controller (63). The cold plate, pulse pump, and water storage tank are connected by pipelines to form an independent closed-loop circulation circuit. Pure water is injected into the system as the coolant, eliminating the need for connection to an external water system.

[0127] The pulse pump is an electromagnetically driven piston-type pulse pump with an adjustable output frequency of 5–50 Hz and an adjustable duty cycle of 30%–70%. The controller collects temperature data from the thin-film thermocouple array inside the cold plate in real time. When the chip power consumption is below 500 W, the pulse frequency is reduced to 10 Hz and the duty cycle to 40% to reduce pump consumption; when the chip power consumption is above 1500 W, the pulse frequency is increased to 40 Hz and the duty cycle to 60% to enhance jet impact. The controller also integrates a boiling critical warning function. When a rapid rise in wall temperature is detected, the pulse frequency and duty cycle are automatically increased to improve cooling capacity.

[0128] The relationship between pulse pump frequency and heat transfer coefficient was determined as follows: Under fixed operating conditions of 1500 W chip power consumption, 1.2 L / min coolant flow rate (time-averaged flow rate), and 40℃ inlet water temperature, the heat transfer performance was tested in constant flow mode and at different pulse frequencies. The test results are shown in the table below: The above results indicate that the heat transfer coefficient is most significantly improved, reaching 14%–15%, when the pulse frequency is in the range of 25–40 Hz. Beyond 50 Hz, the pressure wave attenuates due to fluid inertia, and the improvement effect tends to saturate. The recommended pulse frequency range is 20–45 Hz, with a duty cycle of 40%–60%. Test conditions: The length of the pipeline from the pulse pump outlet to the cold plate inlet is ≤0.5 m, the pipe diameter is 6 mm, and the system back pressure is 0.3 MPa.

[0129] Test results: At a chip power consumption of 1500 W, the highest chip temperature in pulse pump mode decreased from 67℃ to 64℃, a reduction of 3℃, compared to constant current pump mode; the power consumption of pulse pump is about 15 W, and the total system power consumption is reduced by about 12% compared to constant current pump mode.

[0130] This embodiment is applicable to scenarios such as single-rack edge computing nodes, laboratory testing platforms, and existing data centers where water systems cannot be modified.

[0131] Example 11: Distributed Computing Center Liquid Cooling System This embodiment provides a distributed computing center liquid cooling system, which includes 10 sets of independent liquid cooling heat dissipation systems as described in Embodiment 10. Each system corresponds to a computing rack, and each system operates independently without interfering with each other.

[0132] Each system is independently controlled: Rack A (high load, 8 GPUs) pulse pump frequency is set to 45 Hz, duty cycle 65%; Rack B (medium load, 4 GPUs) pulse pump frequency is set to 25 Hz, duty cycle 50%; Rack C (low load, 2 GPUs) pulse pump frequency is set to 10 Hz, duty cycle 35%. Each system adjusts independently according to its own load to achieve on-demand cooling.

[0133] Test results: The total heat dissipation capacity of the system can reach 10×1500 W=15 kW, and the chip temperature of each rack can be independently controlled within the range of 65~70℃; in the event of a single rack failure, only the system of that rack is shut down, while other racks operate normally, thus achieving fault isolation.

[0134] This embodiment is applicable to scenarios such as distributed AI computing centers, liquid cooling retrofits of large data centers, and edge computing node clusters.

[0135] The core innovation of this invention lies in: First, a sealed, through-cavity design replaces the closed microchannels of microfluidics, fundamentally eliminating the risk of blockage and reducing pump consumption. Second, a three-stage composite turbulence array is used to continuously reduce the thermal boundary layer thickness, ensuring the wall surface remains in a highly efficient heat exchange state during thermal development. Third, a two-stage jet (vertical impact + transverse sweep) promotes the homogenization of the cold plate surface temperature, eliminating the defect of local high-temperature areas not covered by single-stage jets. Fourth, heterogeneous wettable surfaces (hydrophobic at the top, hydrophilic at the bottom and sides) regulate bubble dynamics, achieving stable, self-sustaining two-phase boiling.

[0136] The structural parameters of this invention are proportionally scaled: The structural parameters of this invention (including but not limited to the number of manifold branches, the number of macroscopic turbulence structures, the number of jet holes, and the number of guide grooves) can be scaled proportionally to the size of the cold plate. When the size of the cold plate increases, keeping key parameters such as the protrusion spacing, jet hole density, and cavity height ratio unchanged, only the number of structural units needs to be increased to achieve essentially the same heat transfer mechanism and maintain similar performance levels. Therefore, the scope of protection of this patent covers all cold plates using this structural design, regardless of their size.

[0137] First layer of fit: Sealed through-type cavity This invention employs a sealed, through-cavity design with a cavity height of 2–10 mm. Compared to the micron-scale (0.1–0.3 mm) flow channels of microchannels, conventional impurity particles in the coolant (typical size 0.5–2 mm) can pass through freely without the need for high-precision filters. According to the Hagen-Poiseuille law, pressure drop is inversely proportional to the square of the hydraulic diameter. The hydraulic diameter of this invention is 10–100 times that of microchannels, and theoretically, the pressure drop is only 1 / 100–1 / 10000 of that of microchannels. The measured pressure drop is 21 kPa (typical value for microchannels is 50–100 kPa).

[0138] The second layer of coordination: a three-stage composite turbulence array—the boundary layer "continuously weakens the thickness of the thermal boundary layer". The "three-level composite turbulence array" described in this invention refers to a composite system that simultaneously includes the following three scale-level turbulence structures: The first stage—macroscopic turbulence structure: located on the surface of the lower base plate, it consists of three-dimensional protrusions in the millimeter range (0.3–5 mm), including pyramidal, conical, and cylindrical shapes. Its function is to generate large-scale eddies, tearing apart the outer region of the thermal boundary layer.

[0139] The second level—the substrate microstructure—is located on the substrate plane between the macroscopic perturbation structures. It consists of micro-level (10–100 μm) micro-positioning grooves or micro-protrusions. Its function is to generate micro-eddies and perturb the middle region of the thermal boundary layer.

[0140] The third level—surface microgrooves: located on the side and top surfaces of the macroscopic perturbation structure, these are micro-scale (5–30 μm) microgrooves. Their function is to directly shear the viscous sublayer and disrupt the innermost layer of the thermal boundary layer.

[0141] The macroscopic perturbation structure is highly staggered: The heights of the pyramidal, conical, and cylindrical protrusions are staggered, meaning that within the same flow channel region, the heights of the three types of protrusions differ: the pyramidal protrusions are the tallest (0.8–1.0 mm), followed by the conical protrusions (0.5–0.7 mm), and the cylindrical protrusions are the shortest (0.2–0.4 mm). A height difference of ≥0.3 mm between adjacent protrusions prevents the fluid from forming a predictable periodic pattern, leading to continuous boundary layer disruption.

[0142] Parameter selection criteria: The parameters were selected based on the following criteria: When the height of the macroscopic turbulence structure is less than 0.3 mm, according to the turbulence vortex scale theory, the diameter of the generated large-scale vortices is less than 0.3 mm, making it difficult to penetrate the outer thermal boundary layer with a thickness of approximately 0.1–0.5 mm, resulting in insignificant boundary layer tearing effects. When the height is greater than 5 mm, according to the Darcy-Weisbach formula, the flow resistance is inversely proportional to the square of the channel height, leading to a significant increase in pressure drop and excessive pump consumption. Therefore, 0.3–5 mm is selected as the preferred range. When the substrate microstructure size is less than 10 μm, the processing difficulty increases dramatically, and the micro-vortex effect is not obvious; when it is greater than 100 μm, additional flow resistance may be generated. Therefore, 10–100 μm is selected as the preferred range. When the surface microgroove size is less than 5 μm, processing is difficult and prone to clogging; when it is greater than 30 μm, the shearing effect on the viscous sublayer weakens. Therefore, 5–30 μm is selected as the preferred range.

[0143] The mechanism by which the boundary layer "continuously weakens the thickness of the thermal boundary layer": Heat transfer theory shows that the thermal boundary layer develops from the wall, and its thickness δ is proportional to the square root of the flow distance x. If the boundary layer can be disrupted in its initial stage of development (when δ is still small and the thermal resistance is still low), the heat transfer coefficient can always be maintained at a high value close to that of the inlet section.

[0144] This invention achieves continuous boundary layer disruption through a three-stage composite turbulence array: jet impact disrupts the boundary layer at the inlet; when the boundary layer re-develops to approximately 0.1–0.5 mm, macroscopic turbulence structures tear it apart with large eddies; when the boundary layer re-develops to approximately 10–50 μm, the substrate microstructure is disturbed with micro-eddies; and when the boundary layer re-develops to approximately 5–10 μm, surface microgrooves directly shear the viscous sublayer. This continuous cycle of "development-disruption-redevelopment-re-disruption" prevents the boundary layer from thickening to a level with a significant thermal insulation effect.

[0145] Diversion channels – guide water flow upwards and increase turbulence: One or more flow-guiding channels extending along the height direction to the substrate plane are provided on the sidewall of the macroscopic turbulence structure, with their bottoms in fluid communication with the substrate microstructure. The working mechanism of the flow-guiding channels includes three driving forces: First, capillary suction: The micron-sized flow channel generates capillary force, which actively draws the thermal boundary layer liquid near the substrate into the channel, directly damaging the substrate boundary layer.

[0146] Second, it utilizes thermal buoyancy: the hot liquid drawn into the trench has a lower density, generating an upward buoyancy force that guides the water upward, forming a natural convection circulation.

[0147] Third, mainstream shearing and dragging: the mainstream flow velocity of the coolant at the top of the trench is relatively high, which quickly shears and drags away the discharged hot liquid, forming a continuous circulation.

[0148] The synergistic effect of the three driving forces reduces the thickness of the substrate boundary layer by 40% to 60% while significantly increasing the turbulence effect, making it a key bridge connecting the substrate microstructure and the mainstream.

[0149] Increased area without increasing boundary layer – a real increase, not a "bloated" increase: Existing microchannel technology expands the heat transfer area by increasing the number of microchannels. However, the increased solid wall surface also increases the total area covered by the thermal boundary layer, creating a "bloated" effect—the geometric area is large, but the effective heat transfer area is much smaller than the geometric area. In contrast, this invention increases the heat transfer area through macroscopic protrusions in a three-stage composite turbulence array. The protruding sidewalls constitute additional heat transfer surfaces, which can expand the effective heat transfer area by 50% to 150%. More importantly, these protruding sidewalls are located in the continuous scouring area of ​​the mainstream coolant. Under the "continuous reduction of thermal boundary layer thickness" mechanism of this invention, the thermal boundary layer on their surface cannot stably thicken into an effective insulation layer. Therefore, every inch of area added by this invention can truly participate in efficient heat transfer, achieving a "real increase" in the effective heat transfer area.

[0150] The third layer of coordination: dual-stage jet – active impact and promotion of uniform surface temperature of the cold plate. This invention proposes the concept of a two-stage jet: The first-stage vertical jet is generated by the jet holes in the jet orifice plate and impacts the surface of the bottom plate vertically, directly destroying the thermal boundary layer at the inlet of the flow channel.

[0151] The jet orifices employ a zoned, differentiated design: The first type of jet orifice is a tapered constricting nozzle structure located directly above the chip. Its constricting section accelerates the coolant to a higher speed (2-5 times), allowing it to impact the hot spots on the chip at high speed and directly break down the most stubborn thermal boundary layer above the hot spots. The second type of jet orifice is a straight, uniform-diameter orifice structure located away from the chip. It has a large output volume and low flow resistance, used to maintain the overall convection circulation of the coolant inside the cold plate and reduce the average temperature of the coolant.

[0152] The second-stage transverse jet is generated by a jet column. The jet column is a hollow columnar structure with a vertically penetrating jet channel inside, its top end is connected to the manifold cavity, and its bottom end is equipped with a fluid deflection structure.

[0153] Jet column and fluid deflection structure: In this invention, the second-stage transverse jet is generated by a jet column. The jet column is a hollow columnar structure with a vertically penetrating jet channel inside. Its top end is connected to the manifold cavity, and its bottom end is provided with a fluid deflection structure to convert the vertically downward flowing coolant into a transverse jet along the plane of the base.

[0154] The fluid deflection structure can be implemented in two ways: (1) Column type (side spray hole): The jet column extends upward from the surface of the base plate, and its bottom end is provided with a radially penetrating side spray hole. The coolant enters the jet channel from the manifold cavity and is sprayed laterally to both sides from the side spray hole at the bottom end, directly scouring along the base plane, actively diffusing the heat of the hot spot area of ​​the chip to the surrounding area.

[0155] (2) Hanging type (lateral jet gap): The jet column extends downward from the upper cover plate assembly, and a lateral jet gap (preferably 0.5-5 mm) is left between its bottom end and the inner wall of the lower base plate. The coolant enters the jet channel from the manifold cavity, is sprayed laterally from the annular gap at the bottom end, and diffuses along the base plane after passing through the gap, actively carrying away the heat of the hot spot area of ​​the chip to the surrounding area.

[0156] Both implementations achieve the core function of transforming vertical flow into horizontal flow, and have equivalent technical effects in their respective application scenarios, belonging to equivalent substitutions of the same technical concept.

[0157] Theoretical summary of two-stage jets: The two-stage jet system of this invention consists of a first-stage vertical jet and a second-stage transverse jet, which have a clear division of labor and synergy in terms of function and space: (I) The essential difference between the two types of jets (ii) Spatial complementarity effect The vertical jet's effect is concentrated directly above the hot spot on the chip, with the strongest boundary layer disruption at the jet's stagnation point. However, as the fluid diffuses outwards, the boundary layer thickens again, forming a so-called "localized high-temperature region not covered by the jet impact." The lateral jet fills this gap, covering the areas between jets, the gaps between protrusions, and the edges far from the hot spot, preventing the boundary layer from steadily thickening. The two jets complement each other spatially: the vertical jet "breaks open" the boundary layer above the hot spot, while the lateral jet "sweeps away" the hot water, "brings in" cold water, and continuously scours the sidewalls of the protrusions and the substrate plane.

[0158] (III) Time Relay Effect The vertical and horizontal jets form a time-relay cycle within the cavity: the vertical jet impacts the hot spot area at high speed, instantly tearing the thermal boundary layer above the hot spot and simultaneously "flipping" the hot water near the wall; the horizontal jet immediately sweeps the flipped hot water along the base plane towards the outlet, while bringing in fresh, low-temperature coolant; the fresh coolant, propelled by the horizontal jet, covers the hot spot area, and the vertical jet impacts again, tearing the newly formed thin boundary layer once more. This cycle of "impact-tearing-sweeping-displacement-re-impact" continues, preventing the thermal boundary layer from thickening to a level with a significant insulating effect.

[0159] (iv) Superposition effect with the three-stage composite turbulence array When a two-stage jet coexists with a three-stage composite turbulence array (pyramid, cone, cylindrical protrusion), the following superposition effect occurs: First, vortex street enhancement: The vertical jet impact generates a Karman vortex street around the protrusion, and the transverse jet "stretches" and "sweeps" these vortices backward, expanding the vortex coverage from a point to a surface, and significantly expanding the boundary layer damage area.

[0160] Second, the hot water replacement is accelerated: the vertical jet "lifts" the hot water from the wall, and the horizontal jet immediately "sweeps" it away, greatly shortening the residence time of the hot water near the wall and improving the heat exchange efficiency.

[0161] Third, scouring of raised sidewalls: the transverse jet directly scours the raised sidewalls of pyramids, cones, and cylinders, continuously stripping away the thermal boundary layer on the surface of the raised sidewalls, so that this additional heat exchange surface of the raised sidewalls always maintains a highly efficient heat exchange state.

[0162] Fourth, dead zone elimination: The transverse jet can enter the gaps between the protrusions and the wake area downstream of the protrusion, cleaning up the flow dead zones that are difficult for the vertical jet to reach, achieving full coverage of the flow channel without dead zones.

[0163] Fifth, it promotes the uniformity of the surface temperature of the cold plate: the transverse jet actively diffuses the heat from the hot spot area of ​​the chip to the surrounding low-temperature area, achieving active temperature uniformity.

[0164] (v) Quantitative effect Experimental results show that when vertical and horizontal jets work synergistically, the overall heat transfer coefficient is increased by 80% to 120% compared to vertical jets alone, and the chip surface temperature difference is reduced from 2 to 3°C to less than 0.5°C. This is the core innovation of this invention, distinguishing it from existing single-stage jet cooling plates.

[0165] The core innovation of this invention's dual-stage jet lies in the fact that the vertical jet solves the "hot spot" problem, while the horizontal jet solves the "uniform temperature" problem. The two complement each other spatially (point-to-surface) and relay each other temporally (impact-sweep-re-impact), fundamentally overcoming the shortcomings of existing single-stage jet cooling plates that are "locally effective but globally ineffective."

[0166] Compared to existing single-stage jet cooling plates, the difference of this invention lies in the introduction of a second-stage transverse jet through a jet column, in addition to the first-stage vertical jet. However, this difference is not a simple addition, but rather based on a profound understanding of the inherent defect of "localized high-temperature areas not covered by the jet impact." Those skilled in the art typically believe that increasing the jet orifice density can solve this problem, rather than introducing a second-stage transverse jet, thus exhibiting a technical bias.

[0167] Compared to existing cold plates with baffle columns, the difference in this invention is that the jet column has a jet channel communicating with the manifold cavity, and a fluid deflection structure at the bottom, which transforms the originally vertically downward flowing coolant into a transverse jet along the plane of the substrate. The baffle columns in existing cold plates with baffle columns only act as passive baffle elements and cannot generate an active transverse jet. The mechanisms of operation are completely different.

[0168] More importantly, the first-stage vertical jet and the second-stage transverse jet complement each other spatially and form a relay cycle in time. This synergistic effect cannot be predicted by simple superposition. Experimental data shows that the overall heat transfer coefficient of the two-stage jet synergy is 80% to 120% higher than that of the vertical jet alone (see the test results of Example 1 for details). This quantitative data is sufficient to prove the existence of the synergistic effect.

[0169] Fourth layer of coordination: Short-path heat exchange design This invention employs a vertical jet + short-path lateral diffusion flow path design: coolant vertically impacts the surface of the lower base plate from the jet holes in the upper cover plate, absorbs heat, and is discharged nearby through a layout with liquid inlet in the upper middle and liquid outlets on both sides at the bottom. The flow path is short (typically 5-15 mm), much shorter than the long horizontal path (20-50 mm) of microchannels. The short path means that the thermal boundary layer is destroyed or discharged before it has time to fully develop, eliminating the "upstream heating downstream" thermal accumulation effect, and the pressure drop is proportional to the path length, further reducing pump consumption.

[0170] Fifth coordination: Single-phase / two-phase physical adaptive switching mechanism The mechanism by which this invention achieves switching between single-phase and two-phase heat transfer modes is purely physical and adaptive, requiring no sensors, controllers, or active switching devices. Its switching principle is entirely based on the geometric characteristics and surface wettability design of the boiling nucleation structure.

[0171] When the chip power consumption is below 500 W, the cold plate wall temperature is below the boiling point of pure water under the current pressure (approximately 100 °C), and the coolant remains liquid, with the system operating in a single-phase convective heat transfer mode. When the chip power consumption increases to above 500 W, the cold plate wall temperature reaches the boiling point. The nanoscale pits (50–200 nm in diameter) on the boiling nucleation structure act as preferential vaporization nuclei, where bubbles nucleate first, and the system enters a two-phase boiling heat transfer mode. The heat transfer coefficient jumps from 180–200 kW / (m²·K) in the single-phase mode to 400–500 kW / (m²·K) in the two-phase mode. When the chip power consumption decreases to below 500 W, the system automatically reverts to the single-phase convective heat transfer mode.

[0172] The boiling nucleation structure is formed by anodizing or chemical etching of a macroscopic turbulence structure and a substrate surface to create a nanoporous layer with a pore size of 50–200 nm and a porosity of 40%–60%. Experimental results show that, under conditions of pure water coolant and a system pressure of 0.3 MPa, when the wall superheat (the difference between the wall temperature and the saturation temperature) reaches 8–12 °C, stable bubbles begin to form on the surface of the boiling nucleation structure, and the system enters the nucleation boiling region. When the wall superheat falls below 5 °C, the boiling phenomenon essentially ceases, and the system reverts to a single-phase convective heat transfer mode. This superheat threshold is approximately 40% lower than that required for smooth surfaces (which typically require 15–25 °C of superheat), achieving boiling initiation at low superheat.

[0173] Sixth layer of compatibility: heterogeneous wettability surfaces This invention employs a heterogeneous wettability surface design of "hydrophilic background + discrete hydrophobic regions": the hydrophilic micro-nano composite coating has a water contact angle of ≤30°, covering most of the heat exchange surface, promoting coolant spread, accelerating bubble detachment, and increasing critical heat flux density; the hydrophobic regions are discretely distributed on the surface of the hydrophilic micro-nano composite coating, utilizing the wall slip effect to reduce flow resistance; a high-density hydrophobic region is set at the top of the macroscopic turbulence structure, and a low-density hydrophobic region is set on the convex side and the base plane, forming a differentiated wettability distribution of hydrophobic top and hydrophilic sides and bottom.

[0174] Through the aforementioned differentiated wettability design, the top of the protrusion is closest to the chip's heat source and has the highest temperature. Combined with the vaporization nuclei provided by the surface microgrooves and the low adhesion characteristics of the hydrophobic regions, it becomes a "boiling launch point" for preferential bubble nucleation and rapid detachment. The hydrophilic micro-nano composite coating on the protrusion sidewalls ensures that the wall surface is quickly rewetted by the coolant after the bubbles detach, preventing gas film coverage. The substrate microstructure continuously replenishes the protrusion area with fresh coolant. These three elements work together to form a stable, self-sustaining boiling cycle.

[0175] The heterogeneous wettable surface is prepared by the following method: Substrate treatment: Micro-nano rough structures are formed on the surface of the bottom plate (2) and the macro-turbulence structure (21) by chemical etching with an alkaline solution (pH 11).

[0176] Hydrophilic coating: A titanium dioxide-based hydrophilic coating with a thickness of 2 μm was applied using the sol-gel method. After heat treatment at 180℃, the water contact angle was measured to be ≤25°.

[0177] Selective deposition of hydrophobic regions: A fluorosilane self-assembled monolayer was selectively deposited in the top region of the macroscopic turbulent structure (21) using inkjet printing. The printing accuracy was controlled within ±50 μm to ensure a clear and precise wettability boundary between the top region (contact angle ≥110°) and the sidewalls and substrate plane (maintaining hydrophilicity, contact angle ≤25°).

[0178] "Inkjet printing equipment can be commercially available piezoelectric inkjet printers (such as FUJIFILM Dimatix DMP-2800), with a printing resolution and positioning accuracy of ±50μm, which is sufficient to distinguish the top and side wall areas of macroscopic turbulence structures with a height of 0.3 to 0.8 mm." The regulation of bubble dynamics by heterogeneous wettability surfaces can be quantitatively described using the Fritz equation. The hydrophobic top region (θ≥120°) reduces the bubble escape diameter from approximately 0.3 mm on hydrophilic surfaces (θ≤30°) to 0.1 mm; combined with the vaporization nuclei provided by surface microgrooves, the bubble escape frequency increases from approximately 50 Hz to 200 Hz. This fourfold increase in escape frequency means a fourfold increase in the heat loss due to latent vaporization per unit time, which is the physical basis for the heat transfer coefficient of 400–500 kW / (m²·K) in the two-phase mode of this invention.

[0179] The seventh layer of fit: chip bonding and positioning grooves and multi-layer thermally conductive interface materials A chip bonding and positioning groove is machined on the bottom surface of the base plate. A layered thermally conductive interface material is arranged inside the positioning groove, allowing indirect contact between the chip and the bottom surface of the groove through this material. The thermally conductive interface material consists of a flexible thermal pad and a CVD diamond film layered together. The inner wall of the chip bonding and positioning groove is equipped with an insulating coating and a thermal stress buffer layer, and can be designed as an elastic floating structure. An elastic support or wave spring is provided at the bottom of the positioning groove, and a deformation gap is left between the side wall of the positioning groove and the base plate, allowing the entire positioning groove to elastically displace in a direction perpendicular to the plane of the base plate, adaptively adjusting the bonding depth and absorbing thermal stress.

[0180] Eighth layer of coordination: Optimized design of the liquid outlet The liquid outlets are located on both sides of the cold plate near the lower base plate, forming a layout with the liquid inlet at the top center and the liquid outlets at the bottom sides. The liquid outlets are positioned at the lower part of the cavity height, with their lower edge flush with the inner wall of the lower base plate or no more than 2 mm away from the lower base plate. This design ensures that the coolant flowing across the surface of the lower base plate (especially the coolant whose temperature rises after absorbing heat) can be discharged nearby, preventing hot water from stagnating at the bottom and forming a dead zone. This creates a smooth flow path of "impact → diffusion → discharge" with the vertical jet impact and the lateral jet diffusion.

[0181] Ninth layer of coordination: microbubble flow guidance and gas-liquid conduction Microbubble guide channels or gas-collecting microcavities are set in the flow channel of the bottom plate to directionally guide the bubbles generated by two-phase boiling and prevent the bubbles from accumulating and forming gas blockages. Because the present invention uses a sealed through-cavity (millimeter-level flow channel), the bubbles can move freely in the cavity and will not be stuck in narrow channels; combined with the directional guidance of the microbubble guide channels and the low bubble adhesion of the hydrophilic coating, the present invention will not produce gas resistance or gas blockage in both single-phase and two-phase modes.

[0182] The tenth layer of coordination: intelligent sensing and boiling point warning The bottom plate (2) integrates a thin-film thermocouple array to monitor the temperature gradient and temperature fluctuation in real time. When the local temperature rise rate or temperature fluctuation amplitude exceeds the preset threshold, it is determined to be a boiling critical precursor and an early warning signal is output.

[0183] The thin-film thermocouple array is integrated onto the back of the lower substrate (the side opposite the coolant contact surface) or embedded inside the lower substrate at a depth of 0.3–0.5 mm from the inner wall surface using magnetron sputtering or atomic layer deposition. The array contains at least five temperature measurement points, specifically arranged as follows: one main temperature measurement point is positioned at the center of the chip, and one temperature measurement point is positioned at each of the four corners of the chip; for multi-chip cold plates, 3–5 temperature measurement points are independently set for each chip's corresponding area. Temperature data is collected in real-time at each temperature measurement point using a sampling rate of 100–500 Hz.

[0184] The early warning algorithm is as follows: Calculate the temperature gradient dT / dt (unit: ℃ / s) and temperature fluctuation amplitude T_ripple (unit: ℃) at each temperature measurement point. A boiling criticality precursor is determined when both of the following conditions are simultaneously met: (a) the wall temperature T_wall at any temperature measurement point ≥ 85℃; (b) the temperature fluctuation amplitude T_ripple at that temperature measurement point ≥ ±2℃ / s for more than 0.2 seconds. The system outputs an early warning signal to the server management unit or controller. Experimental verification shows that this method can identify boiling criticality risks 3–5 seconds in advance, allowing sufficient time to execute protective actions such as frequency reduction or flow increase.

[0185] Eleventh layer of coordination: Adaptive and differentiated layout of hot zones Based on the thermal distribution characteristics of the chip (4), the jet holes (13) in the high heat flux area of ​​the center of the cold plate have a high density and large aperture, the macroscopic turbulence structure (21) has a small spacing, and the pyramid-shaped protrusions account for a high proportion; the jet holes (13) in the low heat flux area at the edge have a low density and small aperture, the macroscopic turbulence structure (21) has a large spacing, and the cylindrical protrusions account for a high proportion; the transition area is mainly composed of conical protrusions.

[0186] Twelfth layer of coordination: Macroscopic turbulence structure sidewall guide grooves One or more micron-sized flow-guiding grooves (213) extending along the height direction to the substrate plane are provided on the sidewall of the macro-turbulence structure (21). The grooves are 20-50 μm wide and 10-30 μm deep, with a V-shaped or U-shaped cross section, and the bottom is in fluid communication with the substrate microstructure (22).

[0187] The working mechanism of the guide trench includes: capillary suction (actively drawing the thermal boundary layer liquid near the substrate into the trench), thermal buoyancy (the low density of the hot liquid generates upward buoyancy), and mainstream shear drag (the mainstream of coolant at the top of the trench quickly carries away the discharged hot liquid). The synergistic effect of these three driving forces reduces the thickness of the substrate boundary layer by 40% to 60%.

[0188] Auxiliary structure: The bottom plate (2) has a flow channel and macroscopic turbulence structure (21) with a micro-treatment layer for corrosion prevention and scale prevention on the substrate surface.

[0189] It should be noted that this invention replaces an external independent VC vapor chamber, while still retaining necessary thermal interface structures between the bottom surface of the cold plate and the chip (such as the layered thermal interface material within the chip bonding and positioning groove) to ensure contact heat transfer. This invention eliminates the additional thickness and multiple layers of interface thermal resistance introduced by an external independent VC plate, rather than eliminating all thermal interface layers.

[0190] 3.2.X Technology Extension: Chip-Level Integration Solution The macroscopic turbulence structure (21), substrate microstructure (22), and surface microgroove (23) of the present invention can not only be applied to independently manufactured metal cold plate components, but can also be directly integrated into the back of the chip substrate to form a chip-level embedded heat dissipation structure.

[0191] Integration Method: The aforementioned three-level perturbation structure is directly fabricated on the back side of a single-crystal silicon chip using semiconductor micro / nano fabrication processes (including but not limited to photolithography, deep reactive ion etching, nanoimprinting, and thin film deposition). After fabrication, the chip is directly immersed in a dielectric coolant (such as Novec 649 or FC-72). The coolant generates nucleated boiling on the surface of the three-dimensional perturbation array, absorbing heat through two-phase boiling. This approach eliminates the need for jet orifice plates and fluid distribution systems, relying instead on the perturbation structure itself to enhance boiling heat transfer.

[0192] Core advantages: (1) Eliminate the thermal resistance of the TIM layer: The coolant directly contacts the back of the chip, there is no thermally conductive interface material, and the interface thermal resistance is reduced to zero; (2) Heat source-cooling zero distance: Heat does not need to pass through the packaging material and is directly carried away from the back of the chip; (3) Area expansion and bubble enhancement: Pyramid, cone and cylindrical protrusions expand the heat exchange area by 50% to 150%. The staggered distribution of heights generates bubbles of different sizes. High protrusions generate large bubbles that carry more latent heat, while low protrusions generate small bubbles that detach at a higher frequency. The two work together to form a broad spectrum of bubble size distribution, maximizing the nucleation boiling heat exchange potential. (4) Integrated structure: The turbulence structure is integrally formed with the chip substrate, without additional cold plate components, simplifying the heat dissipation system architecture; (5) Process compatibility: Compatible with existing semiconductor manufacturing processes, and can be mass-produced using wafer-level processing.

[0193] Technical Verification: According to research data released by TSMC in 2025, silicon-based integrated micro-coolers, after etching a micropillar array on the back of the chip, can support a heat dissipation power of up to 3.4 kW, with a heat flux density of 250 W / cm². A two-phase immersion cooling solution developed in collaboration between Tsinghua University and Lenovo reduced the chip surface temperature difference from 14.3 ℃ to 3.8 ℃, improving uniformity by 75%. The three-stage turbulence array of this invention has superior structural complexity compared to existing single micropillar arrays, theoretically achieving a higher heat transfer coefficient and better temperature uniformity.

[0194] Therefore, directly integrating the structure of this invention onto the back of a chip and applying it to immersion two-phase cooling should be considered a natural extension of the technical concept of this invention and falls within the protection scope of this invention.

[0195] As a further technological extension, the jet orifice of the present invention can employ an active control design. By integrating MEMS microvalves or variable geometry nozzles and combining them with real-time thermal maps of the chip, the jet intensity in each region can be dynamically adjusted to achieve hotspot tracking adaptive cooling. This solution represents a further extension of the technical concept of the present invention and should be considered an equivalent substitution of the present invention.

[0196] 3.2.Y System-level Expansion: Independent Closed-Loop and Distributed Liquid Cooling Systems This invention also protects system-level application solutions that include the aforementioned cold plate.

[0197] Independent liquid cooling system: including any of the above-mentioned cold plate, pulse pump, water tank, and controller. The cold plate, pulse pump, and water tank are connected by pipelines to form an independent closed-loop circulation circuit, independent of external water systems (such as chilled water systems in computer rooms), enabling plug-and-play deployment. The pulse pump generates pulsed coolant flow. Compared to traditional constant flow pumps, the pulse jet can generate periodic pressure fluctuations, enhancing the boundary layer disruption effect of the jet impact and increasing the heat transfer coefficient by an additional 10% to 20%. The controller adjusts the frequency and duty cycle of the pulse pump in real time according to the chip temperature, reducing the pump speed to save energy under low load (frequency 5-15Hz) and increasing the pulse intensity to enhance heat transfer under high load (frequency 30-50Hz).

[0198] Distributed computing center liquid cooling system: includes multiple independent liquid cooling systems, each system corresponding to one or more computing racks, and each system operates independently without interfering with each other. This distributed architecture has the following advantages: (1) Fault isolation: a single rack system failure does not affect other racks; (2) Flexible expansion: racks can be gradually added according to computing power needs; (3) Customized control: each rack can independently adjust cooling parameters according to the actual power consumption of its internal chips; (4) Convenient deployment: no need to modify the original water system of the data center, greatly reducing the threshold for liquid cooling transformation.

[0199] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments for application in other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A homogeneous liquid-cooled heat dissipation plate with a through-cavity dual-stage jet impact and a three-stage composite turbulence, characterized in that, include: An upper cover assembly is sealed to a lower base plate, forming a sealed, internally continuous cavity without any partition channels, in which coolant can flow freely; the coolant in the cavity is discharged through a side-connected outlet. The upper cover assembly is provided with a liquid inlet and a manifold cavity, and the manifold cavity is connected to the liquid inlet. The upper cover plate assembly has several jet holes communicating with the manifold cavity on one side facing the cavity, forming a first-stage vertical jet. The surface of the lower base plate facing the cavity is provided with a three-level composite turbulence array, which includes a macroscopic turbulence structure. The cavity is provided with at least one jet column, and the jet column is provided with a vertical jet channel inside, the top of the jet channel being connected to the manifold cavity; The bottom end of the jet column is provided with a fluid deflection structure to form a second-stage transverse jet.

2. The cold-rolled plate according to claim 1, characterized in that, The fluid deflection structure is a radially penetrating lateral nozzle located at the bottom end of the jet column; the jet column extends upward from the surface of the lower base plate.

3. The cold-rolled plate according to claim 1, characterized in that, The fluid deflection structure is a transverse jet gap between the bottom end of the jet column and the inner wall of the lower base plate; the jet column extends downward from the upper cover plate assembly, and a gap is left between its bottom end and the inner wall of the lower base plate, so that the coolant is sprayed transversely from the gap in all directions.

4. The cold-rolled plate according to claim 1, characterized in that, The upper cover plate assembly is composed of an upper cover plate body and a jet orifice plate that are sealed together, forming a manifold cavity between them; the jet orifice is opened on the jet orifice plate.

5. The cold-rolled plate according to claim 1, characterized in that, The three-stage composite perturbation array also includes a substrate microstructure, which is located in the substrate region between the macroscopic perturbation structures.

6. The cold-rolled plate according to claim 1, characterized in that, The three-stage composite turbulence array also includes surface microgrooves, which are formed on the surface of the bottom plate.

7. The cold-rolled plate according to claim 1, characterized in that, The jet orifice includes a first type of jet orifice located in the area directly above the chip and a second type of jet orifice located in the area away from the chip. The first type of jet orifice is a tapered constricting nozzle structure, and the second type of jet orifice is a straight hole structure with equal diameter.

8. The cold-rolled plate according to claim 1, characterized in that, The macroscopic disturbance structure includes at least one of pyramid-shaped protrusions, conical protrusions, cylindrical protrusions, and fin-shaped protrusions, which are non-uniformly distributed and have height differences between adjacent macroscopic disturbance structures.

9. The cold-rolled plate according to claim 1, characterized in that, The sidewall of the macroscopic turbulence structure is provided with one or more flow-guiding grooves extending along the height direction to the base plane, and the bottom of the flow-guiding grooves is in fluid communication with the base microstructure.

10. The cold-rolled plate according to claim 1, characterized in that, The bottom surface of the lower base plate is provided with a positioning groove for bonding with the chip, and the inner surface of the positioning groove is provided with a layered thermally conductive interface material.

11. The cold-rolled plate according to claim 1, characterized in that, The liquid outlet is located on both sides of the cold plate near the bottom plate, forming a layout with the liquid inlet at the top center and the liquid outlet at the bottom sides.

12. A liquid cooling heat dissipation system, characterized in that, The system includes at least one cold plate as described in any one of claims 1 to 11, and a coolant circulation pump, heat exchanger, and piping connected to the cold plate; the system can form an independent closed-loop circulation circuit, independent of an external water system; the coolant circulation pump is a pulse pump, generating a pulsed coolant flow rate; multiple systems can be connected in parallel to form a distributed liquid cooling system, with each system operating independently and without interference.

13. The liquid cooling heat dissipation system according to claim 12, characterized in that, It also includes a water storage tank and a controller, which adjusts the frequency and duty cycle of the pulse pump according to the chip temperature to achieve on-demand cooling.

14. A server, characterized in that, It includes at least one cold plate as described in any one of claims 1 to 11, and a computing chip attached to the cold plate.

15. A chip packaging assembly, characterized in that, Includes a cold plate, a chip packaging substrate, and a chip die as described in any one of claims 1 to 11, wherein the cold plate is directly attached to the back of the chip die.

16. A chip-level embedded heat dissipation structure, characterized in that, Includes the back side of the chip substrate or the surface of the chip package cover, wherein the back side of the chip substrate or the surface of the chip package cover is integrated with a three-level composite turbulence array as described in any one of claims 1 to 11.

17. A method for uniform temperature liquid cooling heat dissipation based on the cold plate according to any one of claims 1 to 11, characterized in that, Includes the following steps: Step 1: Coolant enters the manifold cavity through the inlet and impacts the bottom plate surface vertically through the jet holes, forming the first-stage vertical jet, which destroys the thermal boundary layer in the jet stagnation area. Step 2: The coolant flows to the bottom through the jet channel of the jet column, and is transformed into a transverse jet along the base plane by the fluid deflection structure, forming a second-stage transverse jet that diffuses heat to the surrounding area. Step 3: When the coolant flows through the three-stage composite turbulence array, the macroscopic turbulence structure generates large-scale eddy currents that tear the boundary layer, the substrate microstructure generates micro-eddy current disturbances in the middle layer of the boundary layer, and the surface microgrooves shear the viscous bottom layer. Step 4: The coolant, after absorbing heat, is discharged from the outlets on both sides at the bottom.

18. The cold-rolled plate according to claim 1, characterized in that, The cavity has a height of 2–10 mm, the jet hole has a diameter of 0.05–0.5 mm, the macroscopic turbulence structure has a height of 0.3–5 mm, and the spacing between adjacent macroscopic turbulence structures is 0.2–1.0 mm.

Images