An embedded thermal management structure for a spaceborne computing chip and its fabrication method
By employing a four-layer functional partitioning architecture consisting of a near-junction heat extraction layer, a uniform distribution layer, a phase change heat expansion layer, and a vapor conduction layer in the satellite computing chip, combined with high thermal conductivity composite materials and microchannel arrays, the heat dissipation problem of the satellite computing chip in the space environment was solved, achieving stable and efficient two-phase flow and heat conduction.
Patent Information
- Application Number
- CN202611290930.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-25
- Publication Date
- 2026-11-17
AI Technical Summary
Existing heat dissipation technologies for satellite computing chips are unable to achieve ultra-high heat flux density in space environments and the two-phase flow is unstable, failing to meet the heat dissipation requirements under microgravity and thermal vacuum conditions.
It adopts a four-layer functional partition architecture consisting of a near-junction heat extraction layer, a uniform distribution layer, a phase change heat expansion layer, and a steam conduction layer. Combined with high thermal conductivity composite materials and microchannel arrays, it achieves stable heat dissipation through two-phase boiling phase change heat extraction.
Under ultra-high heat flux density conditions, stable two-phase flow and efficient heat conduction were achieved, reducing thermal resistance and eliminating structural stress problems caused by thermal expansion coefficient mismatch.
Smart Images

Figure CN123124812A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the fields of aerospace electronic thermal management and micro / nano manufacturing technology, specifically to an embedded thermal management structure for a spaceborne computing chip and its fabrication method. Background Technology
[0002] With the continuous improvement of onboard artificial intelligence inference processing and high-speed communication baseband processing capabilities, the power consumption density of satellite computing chips is increasing exponentially.
[0003] In high-power spaceborne applications, wide-bandgap semiconductor devices, such as gallium nitride and silicon carbide, have achieved local heat flux densities exceeding 500 watts per square centimeter, and in some extreme conditions, even surpassing 1500 watts per square centimeter. In the vacuum environment of orbit, heat cannot be dissipated through air convection; therefore, the chip's heat must be dissipated via conduction and radiation.
[0004] Existing heat dissipation technologies have the following main shortcomings: heat pipe solutions have long heat diffusion paths and slow thermal response, making it difficult to achieve efficient heat extraction at the millimeter scale; pump-driven single-phase liquid cooling only utilizes the sensible heat of the working fluid, which has an upper limit on heat dissipation capacity and depends on moving parts; while conventional single-layer microchannels have two-phase flow instability and are not specifically designed for space environments such as microgravity and thermal vacuum, making it difficult to balance ultra-high heat flux density heat extraction with working fluid flow stability.
[0005] Therefore, there is an urgent need for an embedded thermal management structure that can achieve ultra-high heat flux density heat extraction in a space environment while maintaining the stability of two-phase flow. Summary of the Invention
[0006] To address the technical bottlenecks in existing satellite computing chip heat dissipation technologies, such as low heat extraction efficiency of heat pipe solutions, limited single-phase liquid cooling capacity, and unstable two-phase flow and lack of adaptability to the space environment in conventional microchannels, this application provides an embedded thermal management structure for spaceborne computing chips and its fabrication method. This structure comprises a four-layer functional partition architecture consisting of a near-junction heat extraction layer, a uniform distribution layer, a phase change heat expansion layer, and a vapor conduction layer arranged sequentially. The synergistic cooperation between near-junction heat extraction from ultra-high thermal conductivity composite materials and two-phase boiling phase change heat extraction from micropillar arrays achieves stable heat dissipation under ultra-high heat flux density conditions.
[0007] Specifically, this application provides the following technical solutions: This application provides an embedded thermal management structure for a spaceborne computing chip, comprising a near-junction heat extraction layer, a uniform distribution layer, a phase change heat expansion layer and a vapor conduction layer arranged in sequence. Each layer is embedded in the chip substrate to form a functional partitioned thermal management structure. The near-junction heat extraction layer is attached to the active region of the chip, and a microchannel array is provided in the near-junction heat extraction layer to capture the heat generated near the junction. The phase change heat expansion layer includes a boiling chamber, which is formed based on a first micropillar structure; The vapor conduction layer includes a first conduction network, which is used to orderly guide the vapor generated by the phase change to the outside of the chip.
[0008] Optionally, the microchannel arrays are parallel to each other.
[0009] Optionally, the cross-section of the microchannel array is rectangular or trapezoidal.
[0010] Optionally, the cross-sectional area of the flow channel in the uniformly distributed layer gradually changes along the flow direction.
[0011] Optionally, in the first micropillar structure, the spacing between the micropillars at the working fluid inflow end is smaller than the spacing between the micropillars at the working fluid outflow end.
[0012] Optionally, the first micropillar structure has at least two steps along the direction of fluid flow.
[0013] Optionally, a first transition structure is provided at the junction of the phase change heat expansion layer and the steam conduction layer.
[0014] This application also provides a method for fabricating an embedded thermal management structure for a spaceborne computing chip, including: A first etching process and a second etching process are sequentially performed on the back side of a semiconductor substrate to form a vapor conduction layer and a phase change heat expansion layer, respectively. A first substrate is prepared using a first composite process, and a microchannel array is processed on the substrate to obtain a near-junction heat extraction layer; The near-junction heat extraction layer is encapsulated and combined with one side of the phase change heat expansion layer of the semiconductor substrate, and then sealed after injecting a two-phase cooling working fluid to form an embedded two-phase thermal management structure.
[0015] Optionally, a second surface treatment is applied to the surface of the micropillar array chamber of the phase change heat expansion layer to enhance capillary force and bubble nucleation density.
[0016] Optionally, the steam guiding layer undergoes a first surface treatment to make the contact angle of the inner wall of the steam guiding channel greater than 120 degrees.
[0017] The beneficial effects of this application are as follows: By using a high thermal conductivity composite material as the near-junction heat extraction layer substrate, the heat flow path from the active region of the chip to the cooling medium is shortened, the near-junction thermal resistance is reduced, and the structural stress problem caused by the mismatch of thermal expansion coefficients under packaging thermal cycling is eliminated. Attached Figure Description
[0018] Figure 1 This is a schematic diagram of the overall structure of the four-layer thermal management structure provided in the embodiments of this application.
[0019] Figure 2 This is a schematic diagram of the near-junction heat extraction layer microchannel array structure provided in an embodiment of this application.
[0020] Figure 3 This is a schematic diagram of a uniformly distributed layer wedge-shaped manifold structure provided in an embodiment of this application.
[0021] Figure 4 This is a top view of the first micropillar structure of the phase change heat expansion layer provided in the embodiments of this application.
[0022] Figure 5 A schematic diagram of the first steam conduction network and the first transition structure of the steam conduction layer provided in the embodiments of this application.
[0023] Figure 6 This is a schematic diagram of the two-phase heat dissipation cycle principle provided in an embodiment of this application.
[0024] Figure 7 A schematic diagram illustrating the principle of bubble nucleation and steam generation in the first micropillar structure provided in this application embodiment.
[0025] Figure 8 This is a schematic diagram of the preparation method steps provided in the embodiments of this application.
[0026] In the diagram: 1. Active region of the chip; 2. Near junction heat extraction layer; 3. Uniform distribution layer; 4. Phase change heat expansion layer; 5. Vapor conduction layer. Detailed Implementation
[0027] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0028] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.
[0029] With the continuous improvement of onboard artificial intelligence inference processing and high-speed communication baseband processing capabilities, the power consumption density of satellite computing chips is increasing exponentially. In the vacuum environment on orbit, heat dissipation cannot rely on air convection. The heat of the chip must be discharged through conduction and radiation. The heat dissipation structure must simultaneously meet the requirements of ultra-high heat flux density heat extraction, stable gas-liquid two-phase flow under microgravity conditions, and reliable sealing in the thermal vacuum environment.
[0030] The thermal management structure provided in this application can be applied to spaceborne high-power-density chips such as spaceborne artificial intelligence inference chips, spaceborne high-speed communication baseband chips, and phased array radar T / R component power amplifier chips, and can also be applied to other aerospace electronic devices with ultra-high local heat flux density.
[0031] like Figure 1 As shown in the figure, this application provides an embedded thermal management structure for a spaceborne computing chip, including a near-junction heat extraction layer, a uniform distribution layer, a phase change heat expansion layer and a vapor conduction layer arranged in sequence. Each layer is embedded in the chip substrate to form a functional partitioned thermal management structure.
[0032] The near-junction heat extraction layer is closely attached to the active region of the chip and uses a high thermal conductivity composite material as the matrix. It contains a microchannel array to effectively capture the heat generated near the junction. The phase change heat expansion layer includes a boiling chamber formed by a first micropillar structure. The cooling medium undergoes two-phase boiling heat exchange in the boiling chamber. The first micropillar structure maintains a continuous supply of liquid film by capillary force. The uniform distribution layer is located between the near-junction heat extraction layer and the phase change heat expansion layer and is used for the uniform distribution of the cooling medium. The vapor conduction layer includes a first conduction network, namely a micromanifold vapor channel network, which orderly guides the vapor generated by the phase change to the outside of the chip.
[0033] like Figure 2 As shown, the near-junction heat extraction layer uses a diamond-copper composite material as its matrix and contains a microchannel array to capture heat generated near the junction. The linear expansion coefficient of the diamond-copper composite material is close to that of gallium nitride, which can eliminate structural stress problems caused by thermal expansion coefficient mismatch during encapsulation thermal cycling.
[0034] In terms of integration, the active region of the chip and the near-junction heat extraction layer can be directly bonded or connected through a metallization transition layer. The metallization transition layer adopts a multi-layer metal system, which has good bonding force with the back metal of the chip and the composite material matrix, thereby effectively reducing contact thermal resistance and compensating for the thermal expansion difference between the chip and the composite material. There is no macroscopic gap between the heat-receiving surface of the near-junction heat extraction layer and the back of the active region of the chip, so that heat can be conducted to the cooling working fluid in the microchannel through the shortest path, shortening the near-junction heat flow path.
[0035] like Figure 3As shown, the uniform distribution layer is located between the near-junction heat extraction layer and the phase change heat expansion layer. It adopts a wedge-shaped manifold structure, and the cross-sectional area of the flow channel gradually changes along the flow direction, so that the cooling working fluid is injected into the phase change heat expansion layer in the manner of jet impact, thereby destroying the thermal boundary layer and enhancing heat transfer.
[0036] When the high-speed jet is ejected from the narrow slit of the uniform distribution layer, it first impacts the bottom plate of the phase change heat expansion layer. After impacting the bottom plate, the jet spreads outwards and then hits the micro-pillars on the top surface, forming a secondary impact. At the secondary impact position, the fluid undergoes another drastic change in velocity and pressure, which further damages the thermal boundary layer, increases the heat transfer coefficient, and forms a rebound thermal enhancement process.
[0037] It should be noted that the intensity of the secondary impact mainly depends on the distance from the narrow slit outlet of the uniform distribution layer to the bottom plate of the phase change heat expansion layer, i.e., the height of the jet chamber. This height is determined in conjunction with the depth of the phase change heat expansion layer chamber and the height of the micropillar, so that the jet maintains a high impact velocity when it reaches the top micropillar, thereby maintaining an effective thermal boundary layer destruction effect on the surface of the micropillar.
[0038] Generally, jet impingement heat transfer has the highest heat transfer coefficient in the stagnation region. At the stagnation point, the fluid velocity is zero, the static pressure is maximum, the boundary layer thickness is thinnest, and the heat transfer capacity is strongest. From the stagnation point outward along the radial direction, the fluid accelerates its flow, the boundary layer gradually thickens, and the heat transfer coefficient decreases with distance.
[0039] In addition, the narrow-slit jet array of the uniformly distributed layer creates multiple stagnation zones on the surface of the phase change heat expansion layer. Each stagnation zone and the jet impact zone overlap, forming an overall uniform high heat transfer coefficient distribution on the surface of the micropillar array.
[0040] like Figure 4 As shown, the phase change heat expansion layer includes a boiling chamber formed by the first micropillar structure, in which the cooling working fluid undergoes two-phase boiling heat exchange.
[0041] The first micropillar structure relies on capillary force to maintain a continuous supply of liquid film, preventing localized drying. Simultaneously, the micropillar array increases the heat exchange area, thereby promoting bubble nucleation and detachment. It should be noted that in the microgravity environment of orbit, gravity-driven natural convection and gas-liquid stratification disappear, and surface tension becomes the dominant force governing the movement of the gas and liquid phases. The micropillar array forms periodic contact line pinning at the gas-liquid interface, causing the gas-liquid interface to expand and stabilize within the micropillar gaps. This suppresses the gas-liquid interface instability caused by the aggregation of vapor into large bubbles within the chamber, thus inhibiting two-phase flow instability.
[0042] The first capillary structure adopts a gradient spacing design. The working fluid inflow end is close to the chip heat source, where the heat flux density is the highest. The micropillar spacing is reduced to 50μm to 80μm to increase the evaporation area and capillary driving force, ensuring liquid film replenishment under high evaporation rate. The working fluid outflow end is far away from the heat source, where the evaporation intensity decreases. The micropillar spacing is increased to 100μm to 200μm to increase porosity, reduce flow resistance, and prevent local drying.
[0043] like Figure 7 As shown, the surface of the first micropillar structure is provided with a micro-nano composite structure. The microcavities and pits in the micro-nano composite structure constitute the vaporization nucleus, increasing the bubble nucleation density and reducing the superheat required for bubble nucleation. After the bubbles generated by phase change nucleate on the surface of the micropillar, they absorb the latent heat of vaporization and grow gradually. When the bubble diameter increases to an order of magnitude comparable to the micropillar spacing, they detach from the micropillar surface under the shearing action of steam flow and the disturbance of adjacent nucleation sites. The steam is discharged through the first guiding network. The micropillar spacing constrains the bubble detachment size, preventing the bubble from agglomerating with adjacent bubbles over a large area before detachment, thus maintaining a stable nucleated boiling state.
[0044] During boiling heat transfer, the gas-liquid two-phase flow pattern in the phase change chamber successively exhibits bubbly flow, slug flow, and annular flow as the heat flux density increases: When the heat flux density is low, bubbles nucleate, grow, and detach on the surface of the micropillars in the form of discrete bubbles, and vapor flows with the liquid phase in the form of dispersed bubbles, i.e., bubble flow. When the heat flux density increases, the number of bubbles increases, and adjacent bubbles are confined and aggregated in the gaps between the micropillars to form slug-shaped bubbles with a size comparable to the channel, i.e. slug flow. When the heat flux density increases further, the liquid phase is squeezed towards the wall to form an annular liquid film, and vapor occupies the core of the channel, i.e., annular flow.
[0045] The micro-pillar gaps in the first micro-pillar structure create spatial constraints on the bubble size, causing the bubbles to be divided by the micro-pillars during the slug flow stage. This makes it difficult to form a large-area vapor film covering the entire heat exchange surface, thus maintaining the heat exchange in a condition dominated by nucleation boiling and avoiding a transition to film boiling that would lead to a sharp deterioration in heat exchange performance.
[0046] like Figure 5 As shown, the vapor conduction layer includes a first conduction network that orderly guides the vapor generated by the phase change to the outside of the chip. The first conduction network includes vapor collection channels and conduction channels. The collection channels are distributed in a tree-like or parallel manner to orderly collect the vapor and residual fluid generated in each region of the phase change heat expansion layer. The conduction channels guide the vapor to the outlet on the side of the chip.
[0047] Among them, the tree-like distribution has each level of converging channels flowing from downstream to upstream, which is suitable for scenarios with a large phase change heat expansion layer area. In the parallel distribution, the length of each converging channel is consistent with the cross-section and the flow path is of equal length, which is suitable for scenarios where the heat source area is distributed in a regular rectangle, and can ensure that the steam output resistance of each area is balanced.
[0048] The tree-like distribution can adopt a two- or three-stage confluence structure: the first-stage confluence channel is arranged along the edge of each heat exchange unit of the phase change heat expansion layer to collect the steam generated by each unit; the second-stage confluence channel merges the steam of adjacent first-stage channels; the third-stage confluence channel collects all the steam and connects it to the steam outlet through hole.
[0049] It should be noted that the cross-sectional dimensions of each channel increase gradually from downstream to upstream, so that the steam velocity remains relatively stable during the convergence process, avoiding pressure drop concentration caused by excessively high local velocity. The proportion of the gradual increase in the channel cross-section is determined according to the gas volume of each convergence, ensuring that the Mach number of the steam in each channel is at a low level, suppressing noise and vibration caused by high-speed steam flow.
[0050] Furthermore, the cross-sectional dimension of the first channeling network is larger than the micropillar spacing of the phase change heat expansion layer to reduce steam flow resistance. Simultaneously, a first transition structure, a tapered gradually expanding transition structure, is provided at the interface between the steam channeling layer and the phase change heat expansion layer. The cross-section of the flow channel gradually expands from the micropillar gaps towards the steam channeling layer. Within the expanding region, the bubbles are continuously guided towards the steam channeling layer by the pressure gradient caused by the increased cross-section, rather than remaining trapped in the micropillar gaps, thus preventing localized drying. In this embodiment, rounded corners are used at the flow channel corners, with a radius of not less than 50 μm, to eliminate dead zones where gas accumulates.
[0051] It should be noted that the first transition structure is formed through a gradual etching depth or step-by-step etching method, which smoothly expands the flow channel cross-section from the micropillar gap towards the vapor channeling layer, avoiding local pressure drop concentration caused by abrupt changes in the cross-section. The rounded corner structure is formed by the rounded corner design of the etching mask pattern and the corner rounding effect of isotropic etching, requiring no additional processing steps. The inner wall of the first channeling network undergoes a first surface treatment (i.e., plasma fluorination treatment) to form a hydrophobic surface with a contact angle greater than 120°. When vapor condenses upon encountering cold on the inner wall of the channeling channel, the hydrophobic surface inhibits the spread of the condensate film, causing the condensate to exist in the form of droplets and roll off under the action of vapor flow, preventing the accumulation of liquid film from blocking the channel cross-section and reducing flow resistance.
[0052] like Figure 6As shown, when the chip heats up, the heat is efficiently captured by the diamond and copper microchannels in the near-junction heat extraction layer. The cooling working fluid absorbs heat and heats up in the microchannels. After being heated, the working fluid is evenly distributed by the uniform distribution layer and enters the phase change heat expansion layer. A thin liquid film is formed on the surface of the micropillar array and nucleation boiling occurs. After absorbing the latent heat of vaporization, it is converted into steam. The steam is orderly discharged from the chip through the first drainage network of the steam drainage layer and enters the external condensation radiation unit. The steam is condensed into liquid at the radiation heat sink. The condensed liquid working fluid flows back to the steam drainage layer. In the closed loop, the capillary pressure difference formed by the meniscus inside the first capillary structure provides the driving force for the return flow, which enables the liquid working fluid to overcome the flow resistance of the loop and flow back to the inlet of the near-junction heat extraction layer, completing the two-phase heat transfer cycle.
[0053] In this embodiment, the external condensation radiation unit includes a radiant heat dissipation plate and a condensation channel connected to the steam outlet. After the steam enters the condensation channel, it condenses on the low-temperature wall surface of the radiant heat dissipation plate. The released latent heat of condensation is radiated out into space through the radiant heat dissipation plate. The liquid film formed by condensation flows into the return pipe under the action of wall tension and flow, and returns to the near-junction heat extraction layer through the liquid working fluid inlet.
[0054] In the condensate return path, the condensate channels are arranged in the gravity-assisted direction. After the condensate forms a liquid film on the channel wall, it converges towards the channel outlet under the action of gravity and steam flow shear force, and enters the liquid working fluid inlet near the junction heat exchange layer through the return pipeline. The liquid working fluid in the return pipeline flows under the drive of capillary pressure difference. The inner diameter of the pipeline is matched with the flow rate of the liquid working fluid to avoid excessive resistance of the liquid column in the pipeline.
[0055] A liquid storage chamber is installed at the inlet of the return pipeline. The volume of the storage chamber is designed to match the condensate flow rate and the return rate, buffering fluctuations in the condensate return flow and preventing flow oscillations caused by the mixing of gas and liquid phases in the return pipeline. It should be understood that the cross-sectional dimensions of the condensate flow channel are designed according to the condensate flow rate to ensure smooth condensate discharge and prevent condensate accumulation that could block the steam passage. Furthermore, this self-circulating process requires no external pump power; under steady-state conditions, the capillary pressure difference and the loop pressure drop reach dynamic equilibrium, ensuring continuous and stable operation of the two-phase circulation in a microgravity environment.
[0056] It should be noted that the loop pressure drop consists of frictional pressure drop, local resistance pressure drop, acceleration pressure drop, and gravity pressure drop. In the microgravity environment on orbit, the gravity pressure drop is approximately zero. The pressure drop mainly comes from frictional and local resistance as the working fluid flows through microchannels, narrow slits, micro-column gaps, and drainage channels.
[0057] The capillary pressure difference formed at the gas-liquid interface by the first capillary structure must be greater than or equal to the total pressure drop of the loop for the liquid working fluid to overcome the flow resistance and continue to flow back. The magnitude of the capillary pressure difference depends on the radius of curvature of the meniscus in the micro-column gap. The smaller the radius of curvature of the meniscus, the greater the capillary pressure difference. When the heat load increases and the evaporation rate increases, the meniscus shrinks, reducing the radius of curvature and increasing the capillary pressure difference. This automatically compensates for the increased loop pressure drop, forming a self-regulating negative feedback mechanism that enables the two-phase cycle to maintain stable operation under variable heat load conditions.
[0058] The steam outlet is connected to an external condensation and radiation unit via a pipeline, which together with the liquid working fluid inlet near the junction heat exchange layer forms a closed two-phase circulation loop.
[0059] The startup process of a two-phase cycle includes two stages: cold start-up and steady-state establishment. During the cold start-up phase, after the chip is powered on, the working fluid in the near junction heat extraction layer is heated and the temperature rises. The heated working fluid enters the phase change heat expansion layer and begins to nucleate and boil after reaching the saturation temperature on the surface of the micropillar. The initial amount of steam gradually accumulates, and the steam is discharged along the first guiding network and enters the external condensation radiation unit. During the steady-state establishment phase, the condensate reflux rate gradually matches the evaporation rate, the pressure and temperature in the loop tend to stabilize, the capillary pressure difference and the loop pressure drop reach dynamic equilibrium, and the two-phase cycle enters a stable operating state.
[0060] It should be noted that the start-up transition time depends on the heat load, the liquid filling rate, and the heat dissipation of the condenser unit. During the start-up phase, a low-power preheating method is used to gradually bring the working fluid to the boiling point, avoiding the pressure shock caused by instantaneous boiling. During the steady-state phase, the jet flow rate is finely adjusted based on the chip surface temperature gradient signal to suppress temperature fluctuations and ensure that the two-phase cycle maintains a stable matching relationship between condensation pressure and boiling pressure under different orbital heat load changes. Insufficient liquid filling rate or insufficient condensation capacity will prolong the start-up transition time.
[0061] In addition, there is a startup power threshold for two-phase circulation. When the chip heat load is below this threshold, the working fluid temperature rise is insufficient to reach the saturation temperature, boiling cannot start, and the heat is dissipated only by single-phase convective heat dissipation, which has limited heat dissipation capacity. When the heat load exceeds the startup threshold, boiling starts, the heat transfer coefficient increases stepwise, the chip surface temperature drops, and startup characteristics are formed.
[0062] The startup power threshold is related to the working fluid saturation temperature, filling rate, and initial temperature. The startup threshold can be lowered by selecting a low-boiling-point working fluid or increasing the filling rate. When selecting the working fluid, low-boiling-point working fluids are preferred under high power density conditions to ensure sufficient boiling driving force, while high specific heat capacity working fluids are preferred under low power density conditions to reduce circulation flow requirements. Both types of working fluids must have good chemical compatibility with chip packaging materials, piping materials, and sealing materials to avoid corrosion or material swelling during long-term operation. After the working fluid selection is determined, the filling volume is calculated based on a predetermined filling rate of the radiator's internal volume and accurately measured at the filling port using a weighing method to shorten the startup time under low heat load conditions.
[0063] In addition, under steady-state conditions, the pressure in the circuit is determined by the saturation pressure of the working fluid corresponding to the condensation temperature of the condensation unit, and the working fluid temperature in the boiling chamber reaches the saturation temperature under that pressure. The lower the condensation temperature, the lower the saturation pressure of the circuit, the lower the saturation temperature in the boiling chamber, the greater the heat exchange temperature difference between the chip and the working fluid, and the stronger the heat extraction capacity.
[0064] The external condensing radiant unit radiates heat from the low-temperature background of deep space. The condensation temperature depends on the thermal equilibrium temperature of the radiant heat sink. The area of the radiant heat sink is determined based on the rated heat load and the space radiation background temperature, ensuring that the steady-state condensation temperature is within the design range. Simultaneously, the surface of the radiant heat sink is coated with a high-emissivity thermal control coating with an emissivity of not less than 0.9, enabling efficient heat dissipation via radiation under the low-temperature background of deep space. The area of the radiant heat sink is directly proportional to the rated heat load and inversely proportional to the temperature difference from the radiation background temperature. Under a rated heat load of 200W, the area of the radiant heat sink is determined according to the radiation heat balance equation, ensuring that the steady-state condensation temperature is within the design range of -10℃ to 30℃. This guarantees that the condensation pressure and boiling pressure of the two-phase cycle are within the range of working fluid property matching, ensuring stable operation of the two-phase cycle under rated conditions.
[0065] like Figure 8 As shown in the embodiments of this application, a method for fabricating an embedded thermal management structure for a spaceborne computing chip is also provided, comprising the following steps: S100, the first etching process and the second etching process are sequentially performed on the back side of the semiconductor substrate to form a vapor conduction layer channel network and a phase change heat expansion layer micropillar array chamber, respectively.
[0066] Specifically, it includes the following sub-steps: S110 provides a semiconductor substrate.
[0067] A 4-inch double-sided polished single-crystal silicon wafer was selected as the processing substrate, with a wafer thickness of 500μm to 700μm. The wafer thickness must simultaneously meet the requirements for structural strength and etching depth. If the thickness is too small, it is difficult to support the etching depth of the three-layer structure and ensure the structural strength after bonding. If the thickness is too large, it will increase thermal resistance and processing costs. A thickness range of 500μm to 700μm can achieve a balance between etching depth requirements and structural strength.
[0068] Semiconductor substrates are used to support the fabrication of uniform distribution layers, phase change heat expansion layers, and vapor conduction layers.
[0069] Before processing, the silicon wafer is cleaned according to standard procedures to remove surface particulate contaminants, metal ions, and organic residues, so as to avoid contaminants forming masking defects during subsequent etching and affecting the processing accuracy of channels and micropillar structures.
[0070] S120, the first etching process is used to process the steam conduction layer channel network.
[0071] The first etching process is carried out on the back of the silicon wafer using deep reactive ion etching, with an etching depth of 200μm to 400μm, to form a vapor conduction layer channel network.
[0072] Among them, deep reactive ion etching adopts an alternating cycle of etching and passivation. Etching gas and passivation gas are periodically introduced. The etching gas isotropically etches silicon, and the passivation gas deposits a polymer passivation layer on the etched sidewall. The alternating cycle of etching and passivation suppresses lateral etching, achieving high aspect ratio vertical sidewall etching. The channel sidewall is steep, meeting the processing requirements of deep trench channels.
[0073] In the deep reactive ion etching process, a photoresist mask and a silicon oxide hard mask are sequentially prepared on the silicon wafer surface. The photoresist is patterned by photolithography to form planar patterns of channels and micropillars. The silicon oxide hard mask protects the non-etched areas during the etching process, improves the mask selectivity, and prevents significant loss at the mask edges during deep trench etching, thus ensuring the perpendicularity and dimensional accuracy of the sidewalls of the channel and micropillar patterns.
[0074] In this embodiment, the etching rate of deep reactive ion etching is related to the process parameters, with a typical etching rate of 5 μm / min to 15 μm / min, a mask selection ratio of not less than 50:1, and a sidewall verticality deviation controlled within ±1°. During the etching process, the silicon wafer temperature is controlled by cooling with an electrostatic chuck to prevent photoresist charring and sidewall passivation layer failure caused by etching heat, thus ensuring the uniformity of deep trench etching.
[0075] The vapor conduit channel network comprises vapor collection channels and conduit channels, with channel widths ranging from 200 μm to 500 μm, arranged in a tree-like pattern and converging towards the vapor outlet vias at the chip edge. The inner walls of the channels are made of... and The mixed gas plasma fluorination treatment, i.e. the first surface treatment, forms a hydrophobic surface with a contact angle greater than 120° to reduce the adhesion of vapor condensate film and reduce flow resistance.
[0076] Plasma fluorination is performed in a reactive ion etching chamber, whereby... and The gas is introduced into the chamber at a set flow rate. Under the influence of the radio frequency electric field, it dissociates into fluorine-containing active groups. These fluorine-containing groups react with the silicon surface of the inner wall of the channel to form a low-surface-energy fluorocarbon film, thereby reducing the surface energy of the inner wall and achieving hydrophobic properties against the aqueous working fluid. By controlling the gas flow rate ratio, radio frequency power, and processing time, the contact angle can be adjusted to be greater than 120°.
[0077] S130, the second etching process is used to process the phase change heat expansion layer micropillar array chamber.
[0078] Above the vapor channeling layer, close to the chip, a second etching process is carried out using deep reactive ion etching, with an etching depth of 100μm to 300μm, to form the first micropillar structure, which is a boiling chamber composed of a micropillar array.
[0079] The diameter of the micropillar is typically in the range of 30μm to 100μm, and the height of the micropillar is typically in the range of 50μm to 500μm; in this embodiment, the diameter of the micropillar is 50μm and the height is 200μm.
[0080] The cross-sectional shape of the micropillars can be exemplarily circular, square, or conical. Circular micropillars have smooth sidewalls during processing, resulting in low etching stress concentration and high mechanical strength. Square micropillars have a larger sidewall area than circular micropillars of the same size, slightly increasing the heat exchange area, and the gaps between adjacent micropillars are uniform, leading to better capillary force distribution. Conical micropillars have a smaller top cross-sectional area than the root, reducing steam rise resistance while maintaining capillary channels. Micropillars of different cross-sectional shapes can all constitute the first micropillar structure, relying on capillary force to maintain a continuous supply of liquid film.
[0081] For micropillars with a height of 50μm to 500μm and a diameter of 30μm to 100μm, the height-to-diameter ratio is in the range of 0.5 to 10. When the height-to-diameter ratio is large, the micropillars are at risk of bending or collapsing due to surface tension during the cleaning, drying and working fluid filling process after etching.
[0082] To ensure the structural integrity of the micropillar array, critical point drying or gradual displacement drying processes are used after etching to reduce the lateral force of liquid bridge surface tension on the micropillars during the drying process. At the same time, by controlling the spacing between micropillars to be no less than 0.5 times the diameter of the micropillars, adjacent micropillars are prevented from sticking together due to liquid bridge aggregation during the drying process, thus ensuring that the micropillar array remains upright and uniformly arranged.
[0083] During the etching process, a rounded corner naturally forms at the root of the micropillar. This rounded corner reduces stress concentration at the root of the micropillar, improving its fatigue resistance under vibration and thermal cycling conditions. The radius of the rounded corner is related to the etching process parameters, typically ranging from 0.1 to 0.3 times the micropillar diameter. The rounded corner also increases the connection area between the micropillar and the substrate, facilitating heat transfer from the root to the substrate and reducing the risk of localized overheating. Simultaneously, the micropillar array relies on capillary action between the micropillars to maintain a continuous supply of liquid film, preventing localized drying.
[0084] The surface of the micropillar array chamber is processed by femtosecond laser to form a micro-nano composite structure, i.e., a second surface treatment, to enhance capillary force and bubble nucleation density. Specifically, the femtosecond laser pulse width is on the order of femtoseconds, and the laser energy density is set to a range higher than the material ablation threshold but lower than the thermal damage threshold. By controlling the laser scanning path and the number of scans, a micro-nano composite structure combining microcavities and pits is formed on the micropillar surface. The ultrashort pulse of the femtosecond laser causes the material to vaporize and be removed in an extremely short time, resulting in a minimal heat-affected zone that does not damage the integrity of the micropillar structure itself.
[0085] Near the chip side, the micropillar spacing is 50μm to 80μm (60μm in this embodiment), and the micropillar spacing gradually increases to 100μm to 200μm (150μm in this embodiment) along the working fluid flow direction, thereby forming the first capillary structure, namely the gradient capillary core structure.
[0086] In this embodiment, the height of the micro-columns is set in three steps along the working fluid flow direction, with the height increasing step by step to control a relatively uniform flow rate. The step-by-step increase in micro-column height causes the cross-sectional area of the flow channel within the micro-column gaps to gradually change along the working fluid flow direction. The flow velocity of the working fluid adjusts accordingly as it flows through each step: the lower micro-column region near the inlet has a relatively large flow channel cross-section and a lower working fluid velocity, which is beneficial for the working fluid to spread fully in the inlet region; as the downstream micro-column height increases, the flow channel cross-section contracts, and the flow velocity increases, compensating for the flow rate decrease caused by evaporation downstream. This results in a relatively uniform working fluid flow rate in each heat exchange zone, avoiding excess flow in some areas and insufficient liquid supply in others.
[0087] In an alternative implementation, the micropillar array can be arranged in an alternating pattern with an angle of 30°, which increases the heat transfer area and has a limited increase in flow resistance compared to a parallel arrangement.
[0088] In another embodiment, the first and second etching processes can employ inductively coupled plasma etching (ICP-E). ICP-E generates high-density plasma within the etching chamber using an RF coil, and modulates the ion bombardment energy using an independently controlled bias power supply. This achieves anisotropic etching of high aspect ratio microstructures at a high etching rate, resulting in good sidewall perpendicularity and low surface roughness.
[0089] The first and second etching processes are performed sequentially on the same side of the silicon wafer. The two etching processes share the same photolithographic alignment reference. The relative position of the two patterns is controlled by overlay alignment marks, so that the positional deviation between the vapor conduction layer channel network and the phase change heat expansion layer micropillar array chamber is controlled within the design tolerance range.
[0090] After two etching processes, wet cleaning was used to remove residual polymer and passivation layer from the channels and micropillars. The cleaning solutions were deionized water, diluted hydrofluoric acid, and deionized water in sequence. The diluted hydrofluoric acid was used to remove residual fluorocarbon polymer deposited during the passivation process. After cleaning, nitrogen blowing or critical point drying was used to remove moisture to prevent the micropillars from collapsing due to liquid surface tension under high aspect ratio conditions. After drying, the micropillar array was microscopically inspected to confirm that there were no residues, no collapses, and no pattern defects.
[0091] The etching depth is monitored and controlled in real time by etching time and etching rate, or the depth is measured after etching using a profilometer and white light interferometer to ensure that the etching depth of the vapor conduction layer is 200μm to 400μm and the etching depth of the phase change heat expansion layer is 100μm to 300μm. The two layers form a preset height difference in the thickness direction, providing a definite chamber volume for subsequent bonding and working fluid flow.
[0092] S140, processed uniformly distributed layer wedge-shaped manifold structure.
[0093] A uniform distribution layer is fabricated on the silicon substrate near the chip side of the phase change heat expansion layer. The uniform distribution layer adopts a wedge-shaped manifold structure, and the flow channel has a geometric shape with gradually changing width in the flow direction. The cross-sectional area of the inlet flow channel gradually decreases along the flow direction, and the cross-sectional area of the outlet flow channel gradually increases. The width of the narrow slit is smaller than the width at the inlet.
[0094] For example, the inlet channel width of the wedge manifold is 800 μm to 1200 μm, the outlet slit width is 100 μm to 300 μm, and the channel depth is 150 μm to 300 μm. The number of manifold channels corresponds to the number of heat exchange units in the phase change heat expansion layer. Each slit is uniformly arranged along the surface of the phase change heat expansion layer, and the slit spacing is 500 μm to 1000 μm. The ratio of the inlet width to the slit width determines the acceleration factor of the jet. The larger the ratio, the higher the jet outlet velocity and the greater the impact heat transfer coefficient. However, the loop pressure drop increases accordingly. The jet outlet velocity and the jet impact heat transfer coefficient have a power function relationship. The increase in jet velocity further thins the boundary layer of the stagnation zone, and the heat transfer coefficient increases approximately to the power of 0.5 to 0.8 of the velocity.
[0095] It should be understood that, within the allowable voltage drop range of the circuit, increasing the jet velocity by increasing the ratio of the inlet width to the slit width can improve the heat transfer coefficient without increasing the working fluid flow rate, thereby reducing the chip surface temperature under the same heat flux density. The ratio of the inlet width to the slit width is determined in conjunction with the allowable voltage drop and the target heat transfer coefficient.
[0096] The flow distribution principle of the wedge manifold is that the cross-section of the flow channel gradually shrinks along the flow direction, the working fluid velocity gradually increases, the static pressure gradually decreases, and the static pressure difference at each narrow slit outlet is controlled within a small range, thereby making the flow rate at each outlet tend to be consistent.
[0097] If a constant cross-section flow channel is used, the downstream static pressure will be lower than the upstream static pressure due to frictional pressure drop during the flow process, and the downstream outlet flow rate will increase accordingly, resulting in uneven flow distribution. The wedge-shaped cross-section compensates for the frictional pressure drop by reducing the cross-sectional area, so that the static pressure at each outlet is balanced, and the flow rate is evenly distributed.
[0098] The size of the wedge angle determines the compensation capability. The larger the wedge angle, the faster the cross-section shrinks, and the more pressure drop along the flow path is compensated. However, an excessively large wedge angle will lead to an increase in the outlet flow velocity difference. The wedge angle is determined in conjunction with the flow channel length and the target flow deviation to keep the flow deviation of each outlet within ±10%.
[0099] Furthermore, the flow distribution of the wedge-shaped manifold is achieved according to a gradient law that decreases from the inlet width to the end width. The inlet pressure of each jet slit is determined by the friction resistance along the manifold and the local resistance of the jet, so that the convective heat transfer coefficient of each jet impact zone on the chip surface presents a gentle gradient distribution along the flow direction, avoiding the concentration of local hot spots. The flow cross section of the manifold inlet section contracts faster and the jet pressure increases accordingly, while the contraction of the end section slows down and the jet pressure decreases. This further matches the heat transfer and heat flux density distribution in the central and edge regions, ensuring that the high heat flux density region receives stronger jet cooling and avoiding energy waste caused by overcooling in the low heat flux density region.
[0100] The uniform distribution layer is used to collect all the cooling working fluid flowing out of the upstream near-junction heat extraction layer and distribute it evenly to each heat exchange unit of the downstream phase change heat expansion layer. The cooling working fluid is injected into the phase change heat expansion layer from the narrow slit in the form of a high-speed jet, forming a jet impact.
[0101] The wedge-shaped manifold channel of the uniform distribution layer is formed on the silicon substrate using the same deep reactive ion etching process as the second etching process. The channel depth is determined in conjunction with the boiling chamber depth of the phase change heat expansion layer, and the narrow slit outlet is aligned with the boiling chamber inlet of the phase change heat expansion layer.
[0102] In the deep reactive ion etching process, the width of the flow channel is gradually changed along the flow direction by controlling the photolithographic pattern. The inlet flow channel is wide and the outlet is narrow. A wedge-shaped cross section can be formed in one etching, without the need for secondary alignment.
[0103] As the cooling medium flows along the flow direction within the wedge-shaped manifold, the cross-section of the flow channel gradually contracts, the flow velocity gradually increases, and the static pressure gradually decreases, forming a high-speed jet at the narrow slit outlet. The jet impacts the surface of the micro-pillar array of the phase change heat expansion layer in a vertical direction, and the high-pressure fluid near the stagnation zone diffuses rapidly in all directions, effectively destroying the near-wall thermal boundary layer and increasing the heat transfer coefficient. The gradually changing cross-section of the wedge-shaped manifold makes the flow distribution at each outlet more uniform, avoiding insufficient liquid supply to some heat exchange units while excessive flow to other units, thus ensuring uniform boiling in all regions of the phase change heat expansion layer.
[0104] In another embodiment, the uniform distribution layer can adopt a rectangular manifold structure with a constant cross-section. The cross-section of each branch channel remains consistent along the flow direction. The channel processing is simple and suitable for application scenarios where the heat source area is small or the requirements for flow distribution uniformity are relatively relaxed. Compared with the wedge manifold, the rectangular manifold with a constant cross-section has relatively limited effect on improving flow distribution uniformity, reducing pressure drop and improving boiling heat transfer stability. However, it can still achieve the function of uniformly distributing the cooling working fluid from the near-junction heat extraction layer to each heat exchange unit of the phase change heat expansion layer.
[0105] S200, a high thermal conductivity composite material substrate (i.e., the first substrate) is prepared using a first composite process, and a microchannel array is processed on the substrate to obtain a near-junction heat extraction layer.
[0106] Specifically, it includes the following sub-steps: S210, the first substrate is prepared using the first composite process.
[0107] The first composite process is carried out by gas pressure infiltration or spark plasma sintering, using diamond particles coated with carbide-forming elements as raw materials to prepare diamond-coated composite material blocks.
[0108] The carbide-forming element is Ti, W, or Cr, which is deposited on the surface of the diamond particles to ensure that the composite material has high thermal conductivity. During the high-temperature process, the carbide-forming element reacts chemically with the surface of the diamond particles to form a carbide interface layer. This interface layer improves the wettability between diamond and molten copper, allowing the molten copper to fully fill the gaps between the diamond particles during impregnation. On the other hand, it reduces the interfacial thermal resistance between diamond and copper, reducing phonon scattering loss at the interface, thus enabling the composite material to achieve an overall thermal conductivity close to the theoretical value.
[0109] The gas pressure impregnation method involves pre-forming the plated diamond particles and placing them in a crucible. Under vacuum or a protective atmosphere, molten copper is pressurized and impregnated into the gaps between the diamond particles. The impregnation temperature and pressure are determined in conjunction with the fluidity of the molten copper and the size of the diamond particles. After impregnation, the temperature is slowly lowered to release thermal stress.
[0110] The spark plasma sintering method involves mixing diamond particles with copper powder and loading them into a graphite mold. The mixture is then rapidly heated by a pulsed current and subjected to axial pressure. The surface of the particles is activated by the spark plasma, achieving low-temperature rapid densification sintering. This method has a short sintering time and can suppress harmful interfacial reactions between the diamond particles and the copper matrix.
[0111] Both of the above processes can produce diamond-copper composite material bulks with good interfacial bonding and low porosity. The composite material bulk thickness is greater than 1 mm, with a thermal conductivity typically ranging from 2200 W / (m·K) to 2600 W / (m·K), not lower than 600 W / (m·K), and a coefficient of linear expansion of [missing value]. / K to / K, approximately the same as gallium nitride chip materials The linear expansion coefficient of / K is matched, eliminating the structural stress problem caused by the mismatch of thermal expansion coefficients under packaging thermal cycling.
[0112] In another embodiment, the high thermal conductivity composite matrix can be a graphene-copper composite material. The graphene-copper composite material uses copper as the matrix and graphene as the reinforcing phase. Graphene forms a continuous thermal conductivity network within the copper matrix, utilizing the high phonon thermal conductivity of graphene to improve the overall thermal conductivity of the composite material, achieving a thermal conductivity of not less than 600 W / (m·K). By adjusting the amount and distribution of graphene, the coefficient of linear expansion of the composite material can be matched with that of the gallium nitride chip material, eliminating structural stress problems caused by the mismatch in thermal expansion coefficients during packaging thermal cycling. The density of the graphene-copper composite material is lower than that of the diamond-copper composite material, which helps reduce the overall weight of the thermal management structure in spaceborne applications.
[0113] In an alternative implementation, a split design can be adopted: microchannel grooves are formed on the back side of a silicon substrate using micro-nano fabrication, and a pre-prepared diamond-copper composite sheet with a thickness of 200 μm to 500 μm is bonded to the microchannel using a bonding process. This method avoids the difficulty of directly processing diamond-copper composite materials and reduces process complexity.
[0114] S220, laser-processed microchannel array.
[0115] Microchannel arrays are directly formed on the surface of a diamond-copper composite substrate by laser processing.
[0116] The laser processing uses a pulsed laser with a wavelength selected that has a high absorption rate for diamond and copper composite materials. The pulse width is on the order of nanoseconds to picoseconds. The material is removed layer by layer along a preset path by scanning with a high repetition frequency of pulses. The scanning strategy adopts multiple shallow cuts and deep cycles, and the thickness removed in a single operation is controlled on the order of micrometers to avoid thermal damage and microcracks to the material caused by a single high-energy pulse. After processing, the bottom surface of the channel is flat and there is no obvious recast layer on the sidewall.
[0117] After laser processing, the microchannels undergo post-processing to remove the recast layer and debris: First, ultrasonic cleaning is used to remove the processing debris attached to the channel. Then, chemical polishing or electrochemical polishing is used to remove the recast layer, making the bottom and sidewall surfaces of the channel smooth, reducing flow resistance and eliminating the source of cracks in the recast layer. After the treatment is completed, the channel surface is subjected to microscopic inspection to confirm that the channel dimensions and surface quality meet the design tolerance requirements.
[0118] Typically, microchannel arrays have channel widths ranging from 50 μm to 200 μm, depths ranging from 100 μm to 300 μm, and channel spacing ranging from 100 μm to 400 μm. The microchannel cross-sections are rectangular or trapezoidal. Among them, rectangular cross-section microchannels have vertical sidewalls, uniform channel cross-sections, high processing precision, and the ratio of heat exchange area to flow channel volume is easy to control. Trapezoidal cross-section microchannels have a bottom width smaller than the top width and inclined sidewalls, which is beneficial for slag removal and liquid filling during laser processing, and can reduce the bottom retention area for the same opening width.
[0119] In practical applications, one of the cross-sectional shapes can be selected based on the processing technology and heat exchange requirements. The microchannel walls undergo surface modification treatment to acquire hydrophilicity or controllable wettability, promoting the formation of a thin liquid film of the cooling medium within the channel and reducing thermal resistance. Surface modification treatment, exemplarily, employs plasma treatment or chemical vapor deposition to introduce polar functional groups or deposit a hydrophilic thin film on the microchannel walls. This reduces the contact angle between the wall and the cooling medium, allowing the cooling medium to spread and form a continuous thin liquid film on the wall, increasing the effective contact area between the liquid film and the wall, and reducing the thermal resistance of single-phase convective heat transfer. The total thickness of the near-junction heat extraction layer can be set to 1 mm, 1.5 mm, or 2 mm.
[0120] Based on the above-mentioned layer structure, the typical size system of this embodiment is as follows: the total thickness of the near-junction heat extraction layer is 1mm to 2mm, wherein the thickness of the composite material substrate is greater than 1mm and the microchannel depth is 100μm to 300μm; the uniform distribution layer has a channel depth of 150μm to 300μm; the phase change heat expansion layer has a boiling chamber depth of 100μm to 300μm, a micropillar diameter of 50μm, and a micropillar height of 200μm; the vapor conduction layer has a channel depth of 200μm to 400μm; the total thickness of the four-layer structure is approximately 1.6mm to 3.2mm, which is compatible with the thickness of the chip substrate, so that the thermal management structure is embedded in the chip package without increasing the package thickness.
[0121] In another embodiment, when the area of the chip's heat-generating region is large, the microchannel array of the near-junction heat-generating layer adopts a manifold-type inlet distribution structure. The cooling working fluid is evenly distributed to each microchannel region through the inlet manifold network, avoiding the problem of downstream temperature rise and uneven temperature distribution between microchannels caused by the working fluid being heated step by step along the flow direction under a single inlet structure. This allows the near-junction heat-generating layer to maintain uniform heat-generating capability in large-area heat-generating scenarios.
[0122] In the manifold-type inlet distribution structure, the inlet manifold is arranged perpendicular to the microchannel. The manifold is connected to the inlet of each microchannel, and the cooling working fluid enters each microchannel simultaneously through the inlet manifold, resulting in balanced inlet pressure in each microchannel. Compared with the single-inlet series structure, the manifold-type inlet makes the flow distribution in each microchannel no longer depend on the natural distribution of pressure drop along the flow path, but is determined by the manifold geometry and the throttling at the microchannel inlet. The flow rate of each channel can be made more consistent by adjusting the local resistance at the microchannel inlet, thus suppressing the temperature unevenness caused by the flow rate attenuation of the downstream channel in large-area heat generation scenarios.
[0123] The near-junction heat extraction layer employs a high-efficiency, high-flow-rate single-phase forced convection design, with its contribution to temperature rise controlled within 30% of the total allowable temperature rise. This leaves over 70% of the temperature rise margin for the phase change heat spreader layer to trigger boiling. Before reaching the phase change heat spreader layer, the heat generated by the chip must pass through the solid thermal conduction path of the near-junction heat extraction layer substrate and the convective heat transfer path of the single-phase liquid within the microchannel, both of which generate temperature rises. If the temperature rise of the near-junction heat extraction layer is too large, the temperature of the working fluid reaching the phase change heat spreader layer will be close to the saturation temperature, resulting in insufficient superheat margin for boiling and making it difficult to maintain stable nucleated boiling.
[0124] In this embodiment, the near-junction heat extraction layer shortens the solid heat conduction path and reduces the fin temperature gradient by improving the microchannel fin efficiency. It also enhances the single-phase convective heat transfer coefficient and reduces the liquid temperature rise by increasing the flow velocity within the microchannel, ensuring that the working fluid retains sufficient superheat margin when it flows out of the near-junction heat extraction layer. The remaining temperature rise margin of more than 70% is used for boiling phase change heat transfer within the phase change heat expansion layer, enabling the working fluid to stably nucleate on the surface of the micropillar and absorb the latent heat of vaporization, thus achieving a functional connection between near-junction heat extraction and phase change heat expansion.
[0125] In terms of quantitative relationships, the temperature rise of the near-junction heat extraction layer consists of two parts: the solid thermal conduction temperature rise and the single-phase liquid convection temperature rise. Solid thermal conductivity temperature rise is related to the thickness and thermal conductivity of the composite material. The higher the thermal conductivity and the thinner the composite material, the smaller the solid thermal conductivity temperature rise. Single-phase liquid convection temperature rise is related to the heat transfer coefficient and liquid temperature rise within the microchannel. The narrower the channel and the higher the flow velocity, the larger the heat transfer coefficient and the smaller the liquid temperature rise. By improving the microchannel fin efficiency, i.e., increasing the ratio of fin height to spacing and maintaining sufficient heat exchange between the fin base and the fluid, the temperature gradient inside the fins can be reduced, making the fin base temperature close to the fluid temperature, thereby reducing the solid thermal conductivity temperature rise under the same heat flux density. Under the constraint of a total allowable temperature rise of 65K, the near-junction heat extraction layer temperature rise is controlled within 19.5K, and the phase change heat expansion layer temperature rise margin is not less than 45.5K, reserving sufficient superheat drive for boiling heat transfer.
[0126] S300, the near-junction heat extraction layer is packaged and combined with one side of the phase change heat expansion layer of the semiconductor substrate, and then sealed after injecting a two-phase cooling working fluid.
[0127] Specifically, it includes the following sub-steps: S310, packaged together.
[0128] The near-junction heat extraction layer, after processing, is aligned with one side of the phase change heat expansion layer of the semiconductor substrate and then packaged and bonded using a bonding process.
[0129] In this embodiment, the bonding process employs one of plasma-activated bonding, eutectic bonding, or thermocompression bonding. Plasma-activated bonding activates the bonding surface with oxygen or nitrogen plasma, achieving high-strength direct bonding at a relatively low temperature, resulting in low bonding temperature and low thermal stress. Eutectic bonding forms a metallurgical bond through a metal eutectic layer above the eutectic temperature, providing high sealing strength and good airtightness. Thermocompression bonding achieves bonding by applying pressure and temperature to induce plastic deformation at the contact surface, and is a simple process.
[0130] In this embodiment, considering the sealing reliability requirements of the thermal vacuum environment, a eutectic bonding method can be adopted to ensure that the bonding strength meets the requirements of the thermal vacuum test, and to align the microchannel outlet of the near-junction heat extraction layer with the inlet of the uniform distribution layer.
[0131] To improve the sealing reliability of the bonding interface, an annular sealing boss is set on the bonding surface between the near-junction heat extraction layer and the phase change heat expansion layer. The sealing boss is arranged around the narrow slit area of the uniform distribution layer and the edge of the boiling chamber of the phase change heat expansion layer. After bonding, the boss area forms a continuous sealing line, isolating the working fluid flow area from the external environment. The width of the sealing boss is 100μm to 300μm and the height is 5μm to 20μm. During bonding, the boss undergoes plastic deformation under pressure and temperature, filling the microscopic unevenness between the bonding surfaces and forming a dense sealing interface.
[0132] It should be noted that for the scheme using eutectic bonding, a eutectic solder layer is pre-prepared on the surface of the sealing boss. The thickness of the eutectic solder layer is 2μm to 5μm. It melts and flows above the eutectic temperature and wets the bonding surface. After cooling, it forms a defect-free metallurgical seal. The narrow slit of the uniformly distributed layer is aligned with the boiling chamber of the phase change heat expansion layer to form a closed embedded two-phase thermal management structure.
[0133] Furthermore, the narrow slots in the uniformly distributed layer have the same shape as the narrow slots on the top surface of the phase change heat expansion layer. After bonding, the two are precisely aligned, ensuring that the cooling medium flowing through the near-junction heat extraction layer can smoothly enter the boiling chamber of the phase change heat expansion layer after carrying away some heat. The bonding process is carried out in a vacuum or protective atmosphere to avoid oxidation of the bonding interface at high temperatures. The bonding temperature of eutectic bonding is determined according to the eutectic layer material, and the heating and cooling rates are controlled to reduce bonding thermal stress. The bonding pressure ensures full contact between the two bonding surfaces, guaranteeing that the bonding interface is free of voids. After bonding, a preliminary leak test is performed on the package structure to confirm the hermeticity of the bonding interface before subsequent working fluid filling.
[0134] S320, working fluid filling and sealing.
[0135] Two-phase cooling medium is injected into the phase change chamber through the filling port, with a filling rate of 40% to 60% of the phase change chamber volume, and then the filling port is sealed.
[0136] The phase change chamber volume is composed of the boiling chamber, the narrow slit jet region, and the volume of the working medium space in the guide channel. This volume is determined by three-dimensional dimensional measurement after processing. The filling amount is calculated by multiplying the measured volume by the target filling rate. The filling amount measurement accuracy is controlled within ±1% to ensure that the filling rate of each batch of products is consistent and to avoid the drift of the two-phase cycle working point caused by the fluctuation of the filling rate.
[0137] The two-phase cooling medium is at least one of water, ammonia, heptafluoropentylmethyl ether, or ethanol. Heptafluoropentylmethyl ether has good dielectric properties and a boiling point of 61°C, which has been verified in aerospace microchannel heat dissipation research. Water has the highest latent heat of phase change, about 2257 kJ / kg, and is suitable for high-power heat dissipation scenarios. Ammonia has a boiling point as low as -33.3°C, making it suitable for low-temperature environments.
[0138] When selecting the working fluid, the on-orbit ambient temperature and heat dissipation temperature range should also be considered: water is suitable for scenarios with high chip operating temperatures, as its boiling point matches the heat dissipation temperature range well, and it has a large latent heat of vaporization and stable chemical properties; ammonia is suitable for low-temperature heat dissipation scenarios, but it is corrosive and requires the selection of compatible piping and sealing materials; HFE-7100 has a low boiling point and low dielectric constant, making it suitable for chips with electrical insulation requirements, but its latent heat of vaporization is lower than that of water, and its heat extraction capacity per unit volume of working fluid is relatively limited. The filling volume is determined based on the filling rate and chamber volume, and the filling pressure of different working fluids is determined based on their saturated vapor pressure to ensure that sufficient vapor space remains in the chamber after filling.
[0139] The working fluid is charged under vacuum. First, the sealed chamber is evacuated to a vacuum level not exceeding [value missing]. The non-condensable gases remaining in the chamber and pipeline are discharged, and then a metered amount of two-phase cooling working fluid is injected through the filling port. After the filling is completed, the filling port is sealed.
[0140] Before charging, the two-phase cooling working fluid is purified and degassed to remove dissolved air and other non-condensable gases, reduce the impurity content in the working fluid, and prevent impurities from depositing on the heat exchange surface during boiling to form fouling thermal resistance, and prevent non-condensable gases from accumulating on the condensation surface to form a gas film that hinders steam condensation. The degassed treatment is achieved through heating-vacuum circulation or membrane degassing. The purified working fluid is transferred to the charging device in a vacuum environment to prevent the purified working fluid from re-contacting the atmosphere.
[0141] Vacuum filling can prevent non-condensable gases from remaining in the chamber, occupying the phase change space, and adhering to the heat exchange surface to form a gas film, thus degrading the heat exchange performance and ensuring that steam can condense normally on the condensing surface after the two-phase cycle starts.
[0142] The liquid filling rate is controlled at 40% to 60% of the phase change chamber volume because if the liquid filling rate is too low, the liquid film supply will be insufficient, and there is a risk of local drying in the high heat flux area; if the liquid filling rate is too high, the steam space will be insufficient, the condensation pressure will increase, and the pressure drop balance of the two-phase cycle will be disrupted.
[0143] Furthermore, the selection of the liquid filling rate must also consider the impact of the on-orbit microgravity environment and attitude changes: if the liquid filling rate is too high, the liquid working fluid may occupy too much steam conduction space under microgravity conditions, reducing the effective cross-sectional area of the steam channel and increasing the steam discharge resistance. In severe cases, the liquid working fluid may be entrained by the steam into the condensation unit, affecting condensation reflux. If the liquid filling rate is too low, the liquid working fluid may not be able to establish a liquid film on the surface of all heat exchange units after the two-phase cycle starts, and some heat exchange units will dry out under low heat load. The liquid filling rate range of 40% to 60% achieves a balance between liquid working fluid supply and steam space, enabling the thermal management structure to maintain stable nucleation boiling and condensation reflux within the rated heat load range.
[0144] After filling and sealing, helium mass spectrometry is used to detect leaks in the packaged structure. Once the leakage rate is confirmed to meet the sealing requirements, external piping connections and system integration are then performed. The filling port is located in a pre-reserved position on the packaged structure and uses a microporous structure to connect the phase change chamber to the outside. After filling, the micropores are sealed by laser welding or cold pressing, forming a seal with the same strength as the packaged structure. Laser welding results in a small heat-affected zone, preventing damage to adjacent bonding interfaces. Cold pressing achieves a heat-free seal through metal plastic deformation, making it suitable for temperature-sensitive packaged structures.
[0145] S330, External piping connection and system integration.
[0146] Stainless steel or titanium alloy capillary tubes are connected to the steam outlet through-hole and the liquid working fluid reflux inlet. The outer diameter of the capillary tube is 1 / 16 inch and the inner diameter is 0.75 mm. The steam outlet is connected to the external condensation radiation unit through the capillary tube and the flexible metal hose, which together with the liquid working fluid inlet near the junction heat exchange layer form a closed two-phase circulation loop.
[0147] The connections between the capillary tube and the steam outlet orifice, as well as the liquid working fluid reflux inlet, are sealed using brazing or laser welding to ensure the sealing reliability of the two-phase circulation loop under thermal vacuum conditions. The flexible metal hose absorbs thermal expansion and contraction deformation during temperature alternation through its corrugated structure, preventing thermal stress from being transferred to the bonding interface and weld joints. Pipe connections are secured using compression fittings or welding, and the pipe routing includes pre-reserved bends to compensate for assembly tolerances and prevent the pipe from applying additional stress to the encapsulation structure.
[0148] S340, Space Environment Adaptability Assessment.
[0149] The encapsulated thermal management structure undergoes space environment adaptability assessment, which includes thermal vacuum testing and thermal performance testing.
[0150] In the thermal vacuum test, the vacuum level is no greater than The temperature cycling range is -65℃ to +125℃, with no fewer than 100 cycles, to assess sealing reliability and material compatibility. Simultaneously, sealing reliability is assessed by monitoring leakage rate changes at the packaging interface using helium mass spectrometry, while material compatibility is assessed by comparing dimensional changes in each layer and the integrity of the bonding interface before and after the test to determine whether differences in thermal expansion of different materials under alternating temperature conditions lead to interface damage. During the test, the chip surface temperature and the activation characteristics of the thermal management structure are monitored concurrently to confirm that heat dissipation performance has not degraded after temperature cycling.
[0151] The temperature cycling rate is controlled between 5K / min and 10K / min, and the holding time for each extreme temperature is not less than 30 minutes to ensure that the packaging structure reaches a sufficient temperature equilibrium. During the cycling process, the temperature response of each measuring point of the packaging structure is recorded to determine whether there is local temperature lag or temperature abrupt change. After the temperature cycling is completed, the packaging structure is visually inspected and its dimensions are remeasured to confirm that there are no cracks, no delamination, and no permanent deformation.
[0152] In thermal performance testing, heat flux densities ranging from 500 W / cm² to 2000 W / cm² were applied through an integrated heating resistor on the back of the chip in a simulated vacuum environment, and the chip surface temperature distribution and steady-state operating temperature were measured. When conditions permit, the start-up characteristics of the heat dissipation structure and the stability of two-phase flow were tested in a microgravity environment such as a drop tower or parabolic flight to meet the engineering goal of ensuring that the chip surface temperature rise does not exceed 65 K.
[0153] The specific process for thermal performance testing is as follows: a thermocouple array is arranged on the chip surface or an infrared thermal imager is used to collect temperature distribution. The temperature collection points cover the center, edge and four corners of the active area of the chip. At the same time, auxiliary measurement points are arranged on the outer surface of the package corresponding to the near junction heat extraction layer and the phase change heat expansion layer to evaluate the temperature gradient between the layers.
[0154] The thermocouples employ a thin filament structure and are adhered to the measuring point to minimize disturbance to the temperature field. Emissivity calibration is performed before infrared thermal imager acquisition to ensure measurement accuracy. Heat flux density is applied in stages via integrated heating resistors on the back of the chip, with a heat flux density of 500 W / cm². 2 1000W / cm 2 1500W / cm 2 2000W / cm 2 The heat flux is applied in stages, and each stage is maintained until the temperature change rate of the chip surface is less than 0.5 K / min. The steady state is then determined, and the steady-state temperature distribution and the location of the highest temperature point under this heat flux density are recorded.
[0155] During loading, the transient response at startup is recorded synchronously to examine the transition process of the heat dissipation structure from cold startup to the establishment of a stable two-phase cycle. After the test, the heat flux density is gradually unloaded to examine the reversibility of the cycle. The judgment criterion is: at 2000 W / cm². 2 Under heat flux density, the maximum temperature rise on the chip surface is no more than 65K, the temperature difference between each temperature acquisition point is within the design allowable range, and no temperature sudden change or continuous temperature rise caused by local drying occurred during the test.
[0156] Long-term life assessment is conducted through accelerated life testing: power cycling and temperature cycling stress are superimposed on the rated thermal load, with a cumulative operating time of no less than 1000 hours. During the test, the changes in chip surface temperature, loop pressure, and working fluid charge are monitored periodically to assess the working fluid leakage rate and the degradation trend of heat transfer performance. After the test, destructive analysis is performed on the packaging structure to check the microstructure of the bonding interface, micropillar structure, and channel surface to confirm that there is no working fluid corrosion, no micropillar collapse, and no interface degradation. Based on this, the on-orbit working life of the thermal management structure is estimated to meet the mission requirements.
[0157] In the microgravity adaptability verification, the start-up characteristic test focuses on the start-up transition time of the heat dissipation structure under microgravity conditions and the temperature overshoot during the start-up process; the two-phase flow stability test focuses on the morphological changes of the gas-liquid interface, the frequency of bubble detachment, and the presence of unstable phenomena such as local drying or flow oscillation under microgravity conditions.
[0158] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. An embedded thermal management structure for a spaceborne computing chip, characterized in that, It includes a near-junction heat extraction layer, a uniform distribution layer, a phase change heat expansion layer and a vapor conduction layer arranged in sequence. Each layer is embedded in the chip substrate to form a functional partitioned thermal management structure. The near-junction heat extraction layer is attached to the active region of the chip, and a microchannel array is provided in the near-junction heat extraction layer to capture the heat generated near the junction; The phase change heat expansion layer includes a boiling chamber, which is formed based on a first micropillar structure; The vapor conduction layer includes a first conduction network, which is used to orderly guide the vapor generated by the phase change to the outside of the chip.
2. The embedded thermal management structure of the spaceborne computing chip according to claim 1, characterized in that, The microchannel arrays are parallel to each other.
3. The embedded thermal management structure of the spaceborne computing chip according to claim 1 or 2, characterized in that, The cross-section of the microchannel array is rectangular or trapezoidal.
4. The embedded thermal management structure of the spaceborne computing chip according to claim 1, characterized in that, The cross-sectional area of the flow channel in the uniformly distributed layer gradually changes along the flow direction.
5. The embedded thermal management structure of the spaceborne computing chip according to claim 1, characterized in that, In the first micropillar structure, the spacing between the micropillars at the working fluid inflow end is smaller than the spacing between the micropillars at the working fluid outflow end.
6. The embedded thermal management structure of the spaceborne computing chip according to claim 1, characterized in that, In the first micro-column structure, at least two steps are provided along the direction of fluid flow.
7. The embedded thermal management structure of the spaceborne computing chip according to claim 1, characterized in that, A first transition structure is provided at the junction of the phase change heat expansion layer and the steam conduction layer.
8. A method for fabricating an embedded thermal management structure for a spaceborne computing chip, characterized in that, include: A first etching process and a second etching process are sequentially performed on the back side of a semiconductor substrate to form a vapor conduction layer and a phase change heat expansion layer, respectively. A first substrate is prepared using a first composite process, and a microchannel array is processed on the substrate to obtain a near-junction heat extraction layer; The near-junction heat extraction layer is encapsulated and combined with one side of the phase change heat expansion layer of the semiconductor substrate, and then sealed after injecting a two-phase cooling working fluid to form an embedded two-phase thermal management structure.
9. The preparation method according to claim 8, characterized in that, The surface of the micropillar array chamber of the phase change heat expansion layer is subjected to a second surface treatment to enhance capillary force and bubble nucleation density.
10. The preparation method according to claim 8, characterized in that, The steam guiding layer undergoes a first surface treatment to make the contact angle of the inner wall of the steam guiding channel greater than 120 degrees.
