A method for forming forced convection to enhance heat transfer in a local high power heat generating area

Abstract

The application discloses a method for forming forced convection to local high-power heating area to strengthen heat exchange, which comprises the following steps: step 1, providing a high-power electronic device, at least one heat dissipation surface of the high-power electronic device has a local high-power heating area; step 2, at least partially immersing the high-power electronic device in an insulating cooling working medium; step 3, directionally and forcibly conveying the low-temperature insulating cooling working medium to the local high-power heating area; and step 4, discharging the insulating cooling working medium, which is heated after flowing through the local high-power heating area, from the local high-power heating area to the immersion environment around the high-power electronic device. The method meets the high heat dissipation requirement of the local high-power area and the conventional heat dissipation requirement of the remaining area, can be directly applied to an existing unmodified chip or power module, has low upgrading cost and is easy to integrate into an existing immersion cooling system.

Description

Technical Field

[0001] This invention belongs to the field of heat dissipation technology for electronic devices, and specifically relates to a method for enhancing heat transfer by forming forced convection in local high-power heat-generating areas. Background Technology

[0002] Currently, liquid cooling is the mainstream solution for some high-power electronic devices. However, different parts of a high-power electronic device have different heat generation powers and different heat dissipation requirements. Existing natural immersion methods are insufficient to meet the higher heat dissipation requirements of local high-power areas. If the entire high-power electronic device is immersed in a high-speed flowing working fluid to improve the heat dissipation efficiency of local parts with high heat dissipation requirements, it will face unnecessary flow resistance and overcooling of the entire area, resulting in low energy efficiency.

[0003] In particular, with the rapid development of AI chips and HPC processors towards high power and high-density integration, chip heat flux density has exceeded 200W / cm², and next-generation chips are expected to reach 300-500W / cm², placing even higher demands on heat dissipation efficiency. Existing mainstream liquid cooling solutions all have some shortcomings: 1. Microchannel cold plate liquid cooling.

[0004] Microchannel cold plate liquid cooling is currently the mainstream forced convection liquid cooling solution. It sets up a micron-level channel structure inside the cold plate and uses forced convection to improve the convective heat transfer coefficient. However, this solution requires the cold plate to be attached to the chip surface through a thermally conductive interface material (TIM). Heat must be conducted through a complex path: chip → TIM1 → metal cap (IHS) → TIM2 → cold plate substrate → coolant inside the microchannel. When the heat flux density is ≥200W / cm², the thermal resistance of the TIM accounts for 30%-50% of the total thermal resistance, becoming a heat dissipation bottleneck. Moreover, chip warpage and assembly tolerances will further aggravate poor contact, significantly reducing heat transfer efficiency.

[0005] 2. Traditional single-phase immersion liquid cooling.

[0006] The chip surface has a flat structure, resulting in a small effective contact area between the liquid and the chip. Relying solely on natural convection for heat transfer, the boundary layer thermal resistance is high, and the heat transfer efficiency is low, making it unsuitable for high heat flux density scenarios (>150W / cm²).

[0007] 3. Passive phase change immersion liquid cooling.

[0008] The chip surface is flat and the number of boiling nuclei is small; relying solely on natural convection and pool boiling for heat exchange, a stable vapor film is easily formed under high heat flux density, triggering boiling crisis (CHF) and directly causing thermal runaway of the chip.

[0009] 4. Silicon-based microchannel liquid cooling solution.

[0010] Microchannels need to be etched at the wafer level, which has extremely high technical barriers, high costs, and requires deep integration with chip manufacturers; and once the microchannels are blocked or damaged, the entire chip is scrapped, resulting in extremely poor maintainability.

[0011] Therefore, a new technical solution needs to be researched to address the above problems. Summary of the Invention

[0012] In view of this, the present invention addresses the shortcomings of existing technologies. Its main objective is to provide a method for enhancing heat transfer by forming forced convection in localized high-power heating areas. This method only forces the high-power areas to be flushed, while the remaining areas remain in a natural immersion environment. This satisfies the high heat dissipation requirements of the localized high-power areas and the conventional heat dissipation requirements of the remaining areas. Moreover, through localized forced transport and natural return, it does not have special requirements on the morphology of the heating surface. It is applicable to smooth surfaces, rough surfaces, and reinforced surfaces with micropillars or microgrooves. This allows the method to be directly applied to existing unmodified chips or power modules without increasing microstructure processing costs or requiring significant modifications to existing immersion tanks. The upgrade cost is low, and it is easy to integrate into existing immersion cooling systems.

[0013] To achieve the above objectives, the present invention adopts the following technical solution: A method for enhancing heat transfer by creating forced convection in a localized high-power heating region includes the following steps: Step 1: Provide a high-power electronic device, wherein at least one heat dissipation surface of the high-power electronic device has a local high-power heat generation area; Step 2: Immerse the high-power electronic device at least partially in an insulating cooling medium; Step 3: The low-temperature insulating cooling medium is forcibly and directionally transported to the local high-power heating area, forcing the insulating cooling medium to flow at high speed through the local high-power heating area at the heat dissipation surface, thereby generating forced convection heat transfer in the local high-power heating area; Step 4: Discharge the insulating cooling medium that has heated up after flowing through the local high-power heating area from that area into the immersion environment surrounding the high-power electronic device; Step 5: Collect the heated insulating cooling medium in the immersion environment, cool it externally, and then forcibly transport it to the local high-power heating area to form a cycle.

[0014] As a preferred embodiment, the high-power electronic device is a server, a high-performance computing processor, or a data center computing chip.

[0015] As a preferred embodiment, the high-power electronic devices are completely immersed in an insulating cooling medium within an immersion tank.

[0016] As a preferred embodiment, the local high-power heating area may or may not have microstructures forming microchannels.

[0017] As a preferred embodiment, the forced delivery is achieved by a pump, which is installed on the circulation pipeline and connected to the inlet of the local high-power heating area.

[0018] As a preferred embodiment, a low-temperature insulating coolant is output from the coolant distribution unit and guided by a flow channel to be preferentially, forcibly, and directionally delivered to the local high-power heating area, forming forced convection scouring. The coolant absorbs heat from the chip in the local high-power heating area through single-phase convection or phase change boiling. The heated single-phase fluid or gas-liquid two-phase mixture naturally flows back to the immersion tank due to density difference and system pressure difference. After the heated insulating coolant enters the coolant distribution unit to complete cooling, it re-enters the circulation. In the cycle, the forced delivery is driven only by the original pump pressure of the coolant distribution unit, without the addition of any additional power components.

[0019] As a preferred embodiment, the same liquid cooling system dissipates heat for multiple high-power electronic devices, which are located in multiple different immersion tanks or in the same immersion tank. For the flow channel of each local high-power heat generation area, a low-temperature cooling working fluid input pipe is connected. All low-temperature cooling working fluid input pipes adopt a pipe diameter differentiation design or a throttling orifice matching design to realize the flow rate distribution according to the power consumption of the local high-power heat generation area.

[0020] As a preferred embodiment, a forced flushing channel is integrated on the local high-power heating area. The forced flushing channel has a working fluid inlet and an outlet. The inlet and outlet are located on the same heat dissipation surface of the high-power electronic device, or on different heat dissipation surfaces.

[0021] As a preferred embodiment, a heat dissipation cover is installed on the local high-power heating area to form the forced flushing channel located between the heat dissipation cover and the surface of the local high-power heating area.

[0022] As a preferred embodiment, the microstructure is integrated on the surface of the local high-power heating region, and the microchannel uses the surface of the local high-power heating region as the inner surface of the microchannel, allowing the cooling working fluid to directly contact it.

[0023] Compared with existing technologies, this invention has significant advantages and beneficial effects. Specifically, as can be seen from the above technical solution, it mainly involves forced flushing only in local high-power areas, while the remaining areas remain in a natural immersion environment. This satisfies the high heat dissipation requirements of the local high-power areas and the conventional heat dissipation requirements of the remaining areas. Compared with some existing methods where natural immersion is insufficient to meet the higher heat dissipation requirements of local high-power areas, this invention effectively solves the problem of low heat dissipation efficiency in these local high-power areas. Furthermore, compared with immersing the entire high-power electronic device in a high-speed flowing working fluid to improve heat dissipation efficiency, this method achieves precise local cooling, avoiding unnecessary flow resistance and over-cooling across the entire area, resulting in higher energy efficiency.

[0024] Moreover, this method, through localized forced delivery and natural recirculation, has no special requirements on the morphology of the heating surface. It is applicable to smooth, rough, or reinforced surfaces with micropillars or microgrooves, allowing direct application to existing, unmodified chips or power modules without increasing microstructure processing costs. In implementation, this method only requires adding a guide channel (such as a nozzle or spout) to the coolant distribution unit outlet to guide the low-temperature working fluid to the localized heating area. No major modifications to the existing immersion tank are needed, resulting in low upgrade costs and easy integration into existing immersion cooling systems. Furthermore, natural recirculation reduces pumping power consumption and noise. Because the working fluid density decreases after heating in the localized high-power area, it naturally recirculates back to the cooling unit under gravity or pressure difference, requiring no additional electrical energy consumption. Compared to a fully forced circulation scheme, this method can reduce pump power consumption by 30% to 50% while also reducing fluid flow noise.

[0025] Secondly, this method utilizes only the existing pump pressure in the coolant distribution unit (CDU) to drive the cryogenic working fluid to forcibly flush the locally heated areas. The heated working fluid then naturally returns due to density and system pressure differences, eliminating the need for additional circulation or return pumps in the entire circulation loop. This reduces the number of components, lowers system cost and leakage risk, and improves long-term reliability. Furthermore, by selecting insulating cooling fluids with different boiling points, this method can flexibly operate in single-phase flow mode (using high-boiling-point working fluid) or phase-change boiling mode (using low-boiling-point working fluid). Single-phase mode is suitable for medium heat flux density scenarios, offering a simple and stable system. Phase-change mode utilizes latent heat of vaporization to maintain stable chip temperature under extremely high heat flux densities, and the bubbles generated during boiling help enhance turbulence. Users can switch working fluids according to actual power requirements without altering the hardware design. To more clearly illustrate the structural features and effects of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0026] Figure 1This is a step diagram illustrating the method of the present invention for forming forced convection in a local high-power heating area to enhance heat transfer; Figure 2 This is a schematic diagram of the overall structure of a forced convection immersion liquid cooling system based on chip surface microstructures, according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the overall structure of a forced convection immersion liquid cooling system based on chip surface microstructures, according to another embodiment of the present invention. Figure 4 This is a schematic diagram of the forced convection scouring of the microstructure of the heat dissipation surface of the heat-generating chip of the present invention. Figure 1 ; Figure 5 yes Figure 4 The diagram shows the exploded state of the structure shown. Figure 6 yes Figure 4 The diagram shows the forced convection scouring flow direction of the structure shown. Figure 7 This is a schematic diagram of the forced convection scouring of the microstructure of the heat dissipation surface of the heat-generating chip of the present invention. Figure 2 ; Figure 8 yes Figure 7 The diagram shows the forced convection scouring flow direction of the structure shown. Figure 9 This is a schematic diagram of the microstructure of the heat dissipation surface of the heat-generating chip of the present invention. Figure 1 ; Figure 10 yes Figure 9 A cross-sectional view of the heating chip shown; Figure 11 This is a schematic diagram of the microstructure of the heat dissipation surface of the heat-generating chip of the present invention. Figure 1 ; Figure 12 yes Figure 11 A cross-sectional view of the heating chip shown; Figure 13 This is a structural diagram showing the microstructure of the present invention fabricated separately and then mounted on the heat dissipation surface of the finished heat-generating chip. Figure 14 This is a structural diagram of the microstructure of the present invention integrated onto a heat sink cover and then mounted on the heat sink surface of the finished heat-generating chip. Detailed Implementation

[0027] Please refer to Figures 1 to 14 As shown, it illustrates a specific embodiment of the present invention.

[0028] A method for enhancing heat transfer by creating forced convection in a localized high-power heating region includes the following steps: Step 1: Provide a high-power electronic device, wherein at least one heat dissipation surface of the high-power electronic device has a local high-power heat generation area; the high-power electronic device is an AI server, a high-performance computing processor, or a data center computing chip, etc.

[0029] Step 2: Immerse the high-power electronic device at least partially in an insulating cooling medium; for example, a server is usually fully immersed in an insulating cooling medium in an immersion tank.

[0030] Step 3: The low-temperature insulating cooling medium is forcibly and directionally transported to the local high-power heating area, forcing the insulating cooling medium to flow at high speed through the local high-power heating area at the heat dissipation surface, thereby generating forced convection heat transfer in the local high-power heating area; the local convection heat transfer coefficient of the local high-power heating area is greater than the natural convection heat transfer coefficient in the immersion environment.

[0031] Step 4: The insulation that heats up after flowing through the local high-power heating area The cooling medium is discharged from this area into the immersion environment surrounding the high-power electronic devices; Step 5: Collect the heated insulating cooling medium from the immersion environment, cool it externally, and then forcibly transport it to the local high-power heating area to form a cycle. The external cooling is carried out by air cooling, liquid cooling, or phase change condensation. In this way, only the local high-power area is forcibly flushed, while the rest of the area remains in a natural immersion environment, meeting the high heat dissipation requirements of the local high-power area and the conventional heat dissipation requirements of the rest of the area. Compared with some existing methods where natural immersion is difficult to meet the higher heat dissipation requirements of local high-power areas, this method effectively solves the problem of low heat dissipation efficiency in local high-power areas. Compared with the method of immersing the entire high-power electronic device in a high-speed flowing medium to improve heat dissipation efficiency, this method achieves precise local cooling, avoids unnecessary flow resistance and over-cooling of the entire area, and has a higher energy efficiency ratio.

[0032] In implementation, microstructures can be provided to form microchannels in the localized high-power heating area, or no microstructures may be provided. The microstructures consist of micropillars, microfins, microchannels, or micropore arrays, with their gaps or internal spaces forming microchannels for the directional flow of the cooling medium. The microstructures are integrated onto the surface of the localized high-power heating area, and the microchannels utilize the surface of the localized high-power heating area as their inner surface, allowing the cooling medium to directly contact them.

[0033] In implementation, a forced flushing channel is integrated into the localized high-power heating area. The forced flushing channel has a working fluid inlet and an outlet, which are located on the same heat dissipation surface of the high-power electronic device, or on different heat dissipation surfaces. A heat sink is installed on the localized high-power heating area to form the forced flushing channel located between the heat sink and the surface of the localized high-power heating area.

[0034] In one embodiment, the forced delivery is achieved by a pump, which is installed on the circulation pipeline and connected to the inlet of the local high-power heating zone. When using a pump alone, the flow velocity of the working fluid through the local high-power heating zone is controlled by adjusting the driving force of the forced delivery, so that the local convective heat transfer coefficient of that zone is greater than the natural convective heat transfer coefficient of that zone without forced delivery.

[0035] In another embodiment, a low-temperature insulating cooling medium is output from the CDU device, i.e., the coolant distribution unit. Guided by a flow channel, the low-temperature cooling medium is preferentially, forcibly, and directionally transported to localized high-power heating areas (e.g., the microstructure area on the surface of the heating chip), forming forced convection scouring. The cooling medium absorbs heat from the chip in the microstructure area of ​​the localized high-power heating area through single-phase convection or phase-change boiling. The heated single-phase fluid or gas-liquid two-phase mixture naturally flows back into the immersion tank due to density difference and system pressure difference (the density of the cooling medium decreases after absorbing heat and heating in the microstructure area, forming a density difference with the unheated cooling medium in the immersion environment, thus generating a natural convection pressure difference). The returning liquid, i.e., the heated insulating cooling medium, enters the coolant distribution unit to complete cooling and gas-liquid separation, and then re-enters the circulation. Taking the heating chip as an example, the insulating cooling medium is a dielectric liquid, and its boiling point is selected according to the target operating mode: in single-phase mode, a medium with a boiling point higher than the chip's operating temperature is selected; in phase-change mode, a medium with a boiling point lower than or close to the chip's operating temperature is selected. The coolant distribution unit is a cold energy distribution unit, which integrates a circulating pump, a heat exchanger, and an optional gas-liquid separator. In the circulation process, the forced delivery is driven solely by the existing pump pressure of the coolant distribution unit, without the addition of any extra power components. Furthermore, by selecting insulating coolants with different boiling points, the system can operate in single-phase or phase-change mode. The forced delivery is achieved through the circulating pump built into the coolant distribution unit. The output head of the circulating pump must overcome the local resistance of the flow channel and the flow resistance in the microstructure region, but does not need to overcome the gravity head of the return flow section in the loop.

[0036] When dissipating heat for multiple high-power electronic devices, the method is based on the same liquid cooling system and dissipates heat for multiple high-power electronic devices located in multiple different immersion pools or in the same immersion pool; for the flow channel of each local high-power heat generation area, a low-temperature cooling working fluid input pipe is connected, and all low-temperature cooling working fluid input pipes adopt a pipe diameter differential design or a throttling orifice matching design to realize the flow rate distribution according to the power consumption of the local high-power heat generation area.

[0037] To better meet the high heat dissipation efficiency requirements of AI servers, high-performance computing processors, or data center computing chips, the above-described method for enhancing heat transfer by forming forced convection in local high-power heat-generating areas is implemented based on a forced convection immersion liquid cooling system with microstructures on the chip surface.

[0038] like Figures 2 to 14 As shown, the system and method are applied to high-power electronic devices (e.g., AI servers with heat flux density > 200W / cm², high-performance computing processors, data center computing chips, etc.). The high-power electronic device includes an electronic device body and at least one heat-generating chip 10 disposed on the electronic device body.

[0039] The heat dissipation surface of the heat-generating chip 10 is provided with a microstructure heat exchange region, which includes a microstructure array. The gaps between the microstructures 11-1 form a forced scouring channel 11-2. The liquid cooling system includes a closed loop in which an insulating cooling medium circulates. The closed loop includes an immersion tank 100, a low-temperature cooling medium inlet pipe 200, a high-temperature cooling medium outlet pipe 300, a circulation pump, and a cooling unit.

[0040] In this embodiment, a CDU device 400 is connected between the outlet of the heated cooling medium output pipe 300 and the inlet of the low-temperature cooling medium input pipe 200, and the circulating pump and cooling unit are formed on the CDU device 400. Typically, the CDU device 400 is a cooling capacity distribution unit in a liquid cooling system, short for Coolant Distribution Unit. Its core function in the liquid cooling system is to drive the circulation of the cooling medium and efficiently remove and exchange heat generated by servers, etc. Taking a server as an example, its built-in pump provides circulation power for the cooling medium, and through a plate heat exchanger, it transfers the heat from the server side to an external cooling water system (such as a dry cooler or cooling tower). Furthermore, by controlling the pump speed and valves, it precisely regulates the temperature and flow rate of the cooling medium to prevent the server from overheating or condensation. For liquid-liquid CDUs, they connect to external chiller units, have high heat exchange efficiency, and are suitable for large data centers. For liquid-gas CDUs, they have built-in fans and cooling coils, do not require an external cold source, and offer flexible installation but are relatively noisy.

[0041] The high-power electronic devices are housed within the immersion tank. Taking server 1000 as an example, the entire server 1000 is immersed in immersion tank 100. For instance, the electronic device body has a motherboard 1001, and a heating chip 10 is mounted on the motherboard 1001. The low-temperature cooling medium input pipe 200 extends into the immersion tank 100, and the outlet of the low-temperature cooling medium input pipe 200 is connected to the inlet of the forced flushing channel 11-2. The outlet of the forced flushing channel 11-2 is connected to the immersion tank 100, and the outlet of the immersion tank 100 is connected to the inlet of the heating cooling medium output pipe 300. The cooling unit is located between the outlet of the heating cooling medium output pipe 300 and the inlet of the low-temperature cooling medium input pipe 200.

[0042] In the case where a CDU device is not used, it is only necessary to ensure that the circulating pump is set in the closed circulation loop to preferentially, forcibly, and directionally deliver the low-temperature cooling medium in the low-temperature cooling medium input pipe 200 into the inlet of the forced flushing channel 11-2, so that it flows through the forced flushing channel 11-2 in the microstructure heat exchange area, and then enters the immersion tank 100 through the outlet of the forced flushing channel 11-2, so that the low-temperature cooling medium forms a forced convection flush in the forced flushing channel 11-2 to quickly absorb the heat of the heating chip 10; and the heated cooling medium in the immersion tank 100, after absorbing the heat of the high-power electronic device, flows back to the cooling unit through the heated cooling medium output pipe 300, is cooled by the cooling unit, and then sent back to the low-temperature cooling medium input pipe 200 for recirculation. Since the cooling medium is in direct contact with the microstructure heat exchange area, there is no thermal resistance from the thermally conductive interface material (TIM).

[0043] like Figures 4 to 6 As shown, the inlet of the forced flushing channel 11-2 is located at or near the top center. The cryogenic cooling medium inlet pipe 200 is connected to the inlet of the forced flushing channel 11-2 via connector 14, and the outlet of the forced flushing channel 11-2 is located on the periphery. After the cryogenic cooling medium enters, it is ejected from the periphery after passing through the forced flushing channel 11-2. The flow path is short, and the outlet of the channel is aligned with the microstructure area, forming a jet impact effect, which further enhances heat transfer. like Figures 7 to 8 As shown, the inlet of the forced flushing channel 11-2 is located at one end of the periphery, and the outlet of the forced flushing channel 11-2 is located at the other opposite end of the periphery. After the low-temperature cooling working fluid enters, the channel outlet is aligned with the microstructure region, forming a transverse flushing effect, further enhancing heat transfer.

[0044] The microstructure 11-1 is one or more combinations of a needle-fin array (e.g., a micron-scale needle-fin array), a trench structure, a porous sintered layer, or a toothed structure. The microstructure on the chip surface significantly increases the heat transfer area, effectively increasing it by 5-20 times; combined with forced convection scouring to enhance boundary layer heat transfer, the heat transfer efficiency is 3-5 times higher than that of traditional single-phase immersion liquid cooling. The heat transfer region of the microstructure is located on the back of the packaging cover plate or the back of the silicon substrate of the heating chip. The microstructure is formed in the post-packaging process of the heating chip through processes such as attachment, welding, electroplating, sintering, etching, or scraping, without involving the processing of the chip's active area.

[0045] like Figure 9 and Figure 10 As shown, it displays a heat-generating chip 10' with a toothed structure package. That is, a toothed structure is integrally formed on the back of the package cover plate 11 or the back of the silicon substrate of the heat-generating chip as a microstructure 11-1, and a gap is formed between adjacent teeth as a forced flushing channel 11-2. The toothed structure and the active region 12 of the chip are located on two opposite sides.

[0046] like Figure 11 and Figure 12 As shown, it displays a heat-generating chip 10 with a pin-wing structure package. That is, a pin-wing structure is integrally formed on the back of the package cover plate 11 or the back of the silicon substrate of the heat-generating chip as a microstructure 11-1, and a gap is formed between adjacent pins as a forced flushing channel 11-2. The pin-wing structure and the active area 12 of the chip are located on two opposite sides.

[0047] In other embodiments, the microstructure can also be fabricated separately and then mounted on the heat dissipation surface of the finished heat-generating chip, such as by attachment or soldering. Figure 13 and Figure 14 As shown, the microstructure 11-1 can be fabricated separately and then mounted on the back of the packaging cover plate or the back of the silicon substrate of the finished heat-generating chip. In actual fabrication, the microstructure 11-1 can be fabricated separately and then mounted on the back of the packaging cover plate 11 or the back of the silicon substrate of the finished heat-generating chip. This is independent of the chip fabrication process. Based on this heat-generating chip with a microstructure, a heat dissipation cover plate 13 can then be installed. Alternatively, the microstructure can be fabricated separately and then integrated onto the heat dissipation cover plate 13, and then mounted onto the heat-generating chip together with the heat dissipation cover plate 13.

[0048] The insulating cooling medium can operate in a single-phase liquid state or a phase change state (by selecting insulating cooling media with different boiling points, the system can operate in single-phase mode or phase change mode). After absorbing heat from the chip in the microstructure heat exchange region through single-phase convection or phase change boiling, the cooling medium forms a heated single-phase fluid or a gas-liquid two-phase mixture and flows back to the coolant distribution unit. After completing cooling and gas-liquid separation, it is circulated again. In single-phase mode, the heated liquid phase fluid naturally flows into the immersion pool due to the system pressure difference and returns to the CDU device for cooling. Single-phase mode strengthens boundary layer disturbance and destroys the thermal boundary layer. In phase change mode, the gas-liquid two-phase mixture naturally flows into the immersion pool due to the density difference and system pressure difference. After preliminary gas-liquid separation in the immersion pool, the liquid phase returns to the CDU device for cooling, and the gas phase is condensed by the condenser and then flows back. Phase change mode provides a large number of stable vaporization nuclei, destroys the vapor film, and significantly increases the critical heat flux density (CHF). Specifically, when the system operates in phase change mode, the microstructured heat exchange region provides a vaporization nucleus. Combined with forced convection scouring to disrupt the vapor film, the boiling process is confined to the nucleation boiling range. The insulating cooling medium includes, but is not limited to, any one of: electronic fluorinated liquid, insulating mineral oil, synthetic hydrocarbon oil, and silicone oil. When a low-boiling-point medium (such as fluorinated liquid) is used, it can operate in phase change mode, utilizing latent heat transfer to achieve higher heat dissipation capacity; when a high-boiling-point medium (such as mineral oil or synthetic oil) is used, it can operate in single-phase mode, significantly reducing system costs.

[0049] The same liquid cooling system dissipates heat from multiple high-power electronic devices. Since each high-power electronic device has at least one heat-generating chip on its body, at least multiple low-temperature cooling medium input pipes are correspondingly provided. Since multiple high-power electronic devices are located in multiple different immersion pools, multiple heating cooling medium output pipes are also correspondingly provided. Alternatively, if multiple high-power electronic devices are located in the same immersion pool, at least one heating cooling medium output pipe is provided. Or, the number of immersion pools is less than the total number of high-power electronic devices, allowing the high-power electronic devices to be distributed across these pools. For example, five high-power electronic devices, one in one immersion pool and the other four in another. Multiple low-temperature cooling medium input pipes employ a differentiated pipe diameter design or a matching throttling orifice design to achieve precise flow distribution according to the power consumption of the heat-generating chip. The entire cycle is driven solely by the existing pump pressure of the coolant distribution unit, without the need for additional power components or independent micro-pumps for individual heat-generating chips.

[0050] If the same liquid cooling system is used to dissipate heat from multiple heat-generating chips of a high-power electronic device, multiple low-temperature cooling working fluid input pipes are provided accordingly; the multiple low-temperature cooling working fluid input pipes adopt a pipe diameter differential design or a throttling orifice matching design to achieve precise flow distribution according to the power consumption of the heat-generating chip, without the need to add an independent micro pump for each heat-generating chip.

[0051] Therefore, if the same liquid cooling system is used to dissipate heat from multiple high-power electronic devices, and if some high-power electronic devices have a heat-generating chip and other high-power electronic devices have multiple heat-generating chips, a corresponding low-temperature cooling working fluid input tube is set for each heat-generating chip, and the number of heating cooling working fluid output tubes is usually matched with the number of immersion tanks.

[0052] Of course, in addition to the forced convection flushing design mentioned above, the cooling medium input pipe and cooling medium output pipe connected to the immersion tank can be flexibly configured as needed. Furthermore, for the low-temperature cooling medium input pipe, its inlet to the forced flushing channel can be in a sealed connection, or one or more drain holes can be pre-reserved on the portion of the low-temperature cooling medium input pipe extending into the immersion tank. These drain holes directly connect to the inside of the immersion tank. Part of the cooling medium in the low-temperature cooling medium input pipe enters the immersion tank directly through the drain holes, while the other part enters the immersion tank after passing through the forced flushing channel.

[0053] The aforementioned forced convection immersion liquid cooling system based on chip surface microstructures includes the following steps during operation: (1) The CDU (coolant distribution unit) outputs a low-temperature insulating coolant; (2) The low-temperature cooling medium is preferentially, forcibly, and directionally delivered to the microstructure area on the surface of the heat-generating chip through the low-temperature cooling medium input pipe to form forced convection scouring; (3) The cooling medium absorbs heat from the chip in the microstructure region through single-phase convection or phase change boiling; (4) After heating, the single-phase fluid or gas-liquid two-phase mixture will naturally flow back to the immersion tank due to the density difference and the system pressure difference; (5) The return fluid enters the coolant distribution unit through the heated cooling working fluid output pipe to complete cooling and gas-liquid separation, and then re-enters the circulation.

[0054] In addition, several specific implementation methods for the cooling working fluid are provided: 1. Fluorinated liquid single-phase immersion forced circulation cooling system Microstructure fabrication: Copper microfin arrays are directly grown on the back of the chip packaging cover using a suitable process; tooth height: 0.5-2.5mm; tooth gap / thickness: 0.1-1mm.

[0055] Cooling medium: Electronic fluorinated liquid (boiling point ≥60℃) is used, and it operates in a single-phase liquid state.

[0056] Flow channel design: CDU liquid outlet → main distribution pipe → branch pipe aligned with chip tooth area → lateral flushing heat exchange.

[0057] Operating parameters: Coolant flow rate: 0.5-2m / s; Chip surface temperature rise: ≤25℃.

[0058] Applicable scenarios: Chips with heat flux density of 100-250W / cm², and scenarios with low cost sensitivity and high reliability requirements.

[0059] 2. Mineral oil single-phase immersion forced circulation cooling system Microstructure fabrication: A pre-fabricated aluminum microfin array is attached to the back of the chip packaging cover using a welding process; fin height: 0.5-2mm; fin thickness: 0.1-0.3mm; fin spacing: 0.4-0.8mm.

[0060] Cooling medium: Insulating mineral oil or synthetic hydrocarbon oil is used, and the cost is only 1 / 10 to 1 / 5 of that of fluorinated liquid.

[0061] Flow channel design: CDU liquid outlet → main distribution pipe → branch pipe aligned with chip tooth area → lateral flushing heat exchange.

[0062] Operating parameters: Coolant flow rate: 0.3-1.5m / s; Chip surface temperature rise: ≤30℃.

[0063] Applicable scenarios: Chips with heat flux density of 80-200W / cm², and cost-sensitive large-scale data center scenarios. Theoretical calculations show that under forced convection conditions, this solution improves heat exchange capacity by 3-5 times compared to natural convection immersion, meeting the heat dissipation requirements of medium power density chips.

[0064] 3. Fluorinated liquid phase change immersion forced circulation cooling system Microstructure fabrication: A porous copper sintered layer is fabricated on the back side of the chip silicon substrate using a sintering process; sintered layer thickness: 0.2-1 mm; porosity: 30%-60%.

[0065] Cooling medium: Low-boiling-point electronic fluorinated liquid (boiling point 30-50℃) is used, and it operates in the nucleation boiling range.

[0066] Flow channel design: The bottom-inlet and top-outlet layout is adopted. The low-temperature fluorinated liquid enters the microstructure area from the bottom of the chip, and the vapor flows into the vapor phase space of the immersion tank from the dedicated exhaust hole at the top of the chip.

[0067] Operating parameters: Coolant subcooling: ≥3℃; Chip surface temperature rise: ≤15℃; Maximum suitable heat flux density: >300W / cm².

[0068] Applicable scenarios: Next-generation high-power AI chips and HPC processors with heat flux density of 200-350W / cm².

[0069] Feasibility Study of the Technical Solution: Currently, microchannel cold plate liquid cooling is the mainstream technology for heat dissipation of AI servers and high-performance computing chips, and its feasibility has been fully verified by large-scale commercial applications. The core heat dissipation mechanism of this solution can be summarized as follows: setting up a micron-level channel structure inside the cold plate, and using forced convection to significantly improve the convective heat transfer coefficient. Publicly available industry data shows that this solution has been widely used in scenarios with heat flux densities of 150-200 W / cm².

[0070] This invention relocates the microstructure from inside the cold plate to the chip surface and changes the relative position of the coolant and microstructure from indirect contact (through the TIM and the cold plate substrate) to direct contact. The thermal resistance of a traditional cold plate solution can be decomposed as: R_total = R_TIM + R_base + R_convection, where: R_TIM: Thermal resistance of the thermal interface material, publicly available industry data is 0.1-0.2 K / W (depending on the material and process). R_base: Thermal resistance of the cold plate substrate (usually copper, about 1mm thick), with a theoretically calculated value of about 0.03-0.08 K / W; R_convection: The convective heat transfer resistance of the coolant within the microchannel.

[0071] This invention eliminates the TIM and cold plate substrate, therefore the thermal resistance of this solution is: R_total' = R_convection' + R_contact, where R_contact is the contact thermal resistance between the microstructure and the chip surface (which can be controlled to an extremely low level through processes such as welding and electroplating). Under the same "microstructure + forced convection" conditions, R_convection' and R_convection are comparable in magnitude. Since this solution eliminates the two positive thermal resistances R_TIM and R_base, for any given heat flux density, the chip temperature rise of this solution will necessarily be lower than that of the cold plate solution, with a temperature rise reduction of approximately (R_TIM + R_base) × Q. Taking a heat flux density of 200 W / cm² and a chip area of ​​1 cm² as an example, the temperature rise reduction is approximately 15-30℃.

[0072] In summary, the method of this invention for enhancing heat transfer by forming forced convection in local high-power heating areas only forces the high-power areas to be flushed, while the remaining areas remain in a natural immersion environment. This satisfies the high heat dissipation requirements of the local high-power areas and the conventional heat dissipation requirements of the remaining areas. Moreover, through local forced transport and natural return, it has no special requirements on the morphology of the heating surface. Whether it is a smooth surface, a rough surface, or a reinforced surface with micropillars or microgrooves, it is applicable. This allows the method to be directly applied to existing unmodified chips or power modules without increasing the cost of microstructure processing or making significant modifications to the existing immersion pool. The upgrade cost is low, and it is easy to integrate into existing immersion cooling systems.

[0073] In particular, this method is applied to forced convection immersion liquid cooling systems based on chip surface microstructures, offering numerous advantages such as zero TIM thermal resistance, dual-mode compatibility, flexible working fluid selection, high heat exchange efficiency, easy process implementation, extremely simple and reliable system, and a revolutionary reduction in thermal resistance path. Specifically: the cooling working fluid directly contacts the heat exchange surface of the heat-generating chip microstructure, minimizing the heat exchange path and overcoming the heat flux density bottleneck of cold plate liquid cooling to achieve zero TIM thermal resistance; the same system architecture supports single-phase and phase-change heat exchange, allowing flexible selection of working fluid and operating mode based on heat dissipation requirements and cost budget, achieving dual-mode compatibility; and it is compatible with various insulating cooling working fluids such as fluorinated liquids, mineral oils, and synthetic oils, significantly reducing system operating costs while mitigating the environmental policy risks associated with using a single working fluid (such as fluorinated liquids). Flexible working fluid selection; microstructure significantly increases heat exchange area, combined with forced convection scouring, resulting in heat exchange efficiency 3-5 times higher than traditional single-phase immersion liquid cooling; high heat exchange efficiency; package-level implementation, no deep bonding with wafer fabs, even unrelated to chip manufacturing processes, low risk, rapid iteration, good maintainability, and easy process deployment and application; no additional pumps or complex actuators, low failure rate, supports rapid upgrades and retrofits of existing immersion cabinets, extremely simple and reliable system; the TIM thermal resistance of traditional cold plate solutions (0.1-0.2K / W) plus the conductive thermal resistance of the cold plate substrate totals approximately 0.13-0.28K / W, while this invention reduces the interface thermal resistance to near zero through direct contact, reducing the total thermal resistance by more than an order of magnitude compared to traditional cold plates, and shortening the thermal resistance path. Given industry preconceived notions, this invention breaks through the technical bias that "microchannel + forced convection can only be implemented in cold plates," directly integrating this mature heat dissipation mechanism onto the surface of the heat-generating chip, while retaining the original high-efficiency heat dissipation mechanism, completely solving the inherent problems of indirect cooling at the architectural level.

[0074] The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the technical scope of the present invention. Therefore, any minor modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention shall still fall within the scope of the technical solution of the present invention.

Claims

1. A method for enhancing heat transfer by creating forced convection in a localized high-power heating region, characterized in that, Includes the following steps: Step 1: Provide a high-power electronic device, wherein at least one heat dissipation surface of the high-power electronic device has a local high-power heat generation area; Step 2: Immerse the high-power electronic device at least partially in an insulating cooling medium; Step 3: The low-temperature insulating cooling medium is forcibly and directionally transported to the local high-power heating area, forcing the insulating cooling medium to flow at high speed through the local high-power heating area at the heat dissipation surface, thereby generating forced convection heat transfer in the local high-power heating area; Step 4: Discharge the insulating cooling medium that has heated up after flowing through the local high-power heating area from that area into the immersion environment surrounding the high-power electronic device; Step 5: Collect the heated insulating cooling medium in the immersion environment, cool it externally, and then forcibly transport it to the local high-power heating area to form a cycle.

2. The method for forming forced convection to enhance heat transfer in a localized high-power heating area according to claim 1, characterized in that, The high-power electronic device is a server, a high-performance computing processor, or a data center computing chip.

3. The method for forming forced convection to enhance heat transfer in a localized high-power heating area according to claim 1, characterized in that, The high-power electronic devices are all immersed in the insulating cooling medium in the immersion tank.

4. The method for forming forced convection to enhance heat transfer in a localized high-power heating area according to claim 1, characterized in that, The local high-power heating area may or may not have microstructures forming microchannels.

5. The method for forming forced convection to enhance heat transfer in a localized high-power heating area according to claim 1, characterized in that, The forced delivery is achieved by a pump, which is installed on the circulation pipeline and connected to the inlet of the local high-power heating area.

6. The method for forming forced convection to enhance heat transfer in a localized high-power heating area according to claim 3, characterized in that, The low-temperature insulating coolant is output from the coolant distribution unit and guided by the flow channel to be preferentially, forcibly, and directionally delivered to the local high-power heating area, forming forced convection scouring; the coolant absorbs the chip heat in the local high-power heating area through single-phase convection or phase change boiling; the heated single-phase fluid or gas-liquid two-phase mixture naturally flows back to the immersion tank due to the density difference and system pressure difference; after the heated insulating coolant enters the coolant distribution unit to complete the cooling, it re-enters the circulation. In the cycle, the forced delivery is driven only by the original pump pressure of the coolant distribution unit, without adding any additional power components.

7. The method for forming forced convection to enhance heat transfer in a localized high-power heating area according to claim 6, characterized in that, The same liquid cooling system dissipates heat for multiple high-power electronic devices, which are located in multiple different immersion tanks or in the same immersion tank. For each local high-power heat generation area, a low-temperature cooling working fluid input pipe is connected to the flow channel. All low-temperature cooling working fluid input pipes adopt a pipe diameter differentiation design or a throttling orifice matching design to realize the flow rate distribution according to the power consumption of the local high-power heat generation area.

8. The method for forming forced convection to enhance heat transfer in a localized high-power heating area according to claim 1, characterized in that, The local high-power heating area is integrated with a forced flushing channel. The forced flushing channel is provided with a working fluid inlet and an outlet. The inlet and outlet are located on the same heat dissipation surface of the high-power electronic device, or on different heat dissipation surfaces.

9. The method for forming forced convection to enhance heat transfer in a localized high-power heating area according to claim 8, characterized in that, A heat dissipation cover is installed on the local high-power heating area to form the forced flushing channel located between the heat dissipation cover and the surface of the local high-power heating area.

10. The method for forming forced convection to enhance heat transfer in a localized high-power heating area according to claim 4, characterized in that, The microstructure is integrated on the surface of the local high-power heating area, and the microchannel uses the surface of the local high-power heating area as the inner surface of the microchannel, allowing the cooling working fluid to directly contact it.

Images