Small electronic devices generate a lot of heat, which must be dissipated in order to maintain the normal performance of the device. This is an indispensable part of the microchip: microfluidic system.
An energy-saving way to improve the performance of electronic systems is to integrate microfluidic cooling channels into the chip to prevent overheating.However, state-of-the-art microfluidic cooling systems have previously been designed and constructed separately from electronic chips to prevent channels from being integrated into circuits to provide direct cooling at hot spots.Since this integration greatly increases the complexity of chip manufacturing, it may increase costs.Van Erp et al. wrote and reported in the journal 《Nature》 an electronic device designed with an integrated microfluidic cooling system that is closely aligned with the electronic components and manufactured using a single low-cost process.
Power electronic devices are solid-state electronic devices that can convert electrical energy into different forms and are widely used in various daily applications.From computers to battery chargers, from air conditioners to hybrid electric vehicles, and even satellites.The growing demand for more and more efficient and smaller power electronic devices means that the amount of conversion power per unit volume of these devices has greatly increased.In turn, it increases the heat flux of the equipment, that is, the heat generated per unit area.The heat generated in this way is becoming a big problem.The energy and water consumed by data centers in the United States are the same as the energy and water consumed by Philadelphia, and they must meet their residential needs.
Microfluidic cooling systems have great potential to reduce the temperature of electronic devices because heat can be efficiently transferred to these systems.Three microfluidic cooling designs have been developed.The first type is used to cool the chips covered by the protective cover.Heat is transferred from the chip through the cover to the cold plate, which contains microfluidic channels through which the liquid coolant flows.Two layers of thermal interface material (TIM) are used to help transfer heat from the lid to the cold plate: one layer is between the lid and the board, and the other layer is between the lid and the die(Semiconductor wafers have been made from chips).
In the second design, the chip has no cover, so the heat is transferred directly from the back of the chip to the microfluidic cooling plate through a single TIM layer.The main disadvantage of these two methods is the need for a TIM layer-even if the TIM is designed to effectively transfer heat, there will still be thermal resistance at the interface between the TIM layer and the mold, lid, and cold plate.
The effective way to solve this problem is to make the coolant directly contact the chips-this is the third conventional designFor example, die direct jet cooling is a valuable technique in which liquid coolant is jetted from a microchannel nozzle directly on the back of the chip. Because there is no TIM layer, the cooling efficiency of this method is very high, and there is no need to change in the process of manufacturing the chip.However, manufacturing microfluidic devices is generally expensive. Low-cost, polymer-based technology has been developed, but is not compatible with existing production and assembly processes of electronic devices.
Figure 1 Photoelectric integration. Atabaki et al. reported a technique for integrating electronic and photonic devices on a single silicon microchip.The author adds spacers (islands) of the insulator material silicon dioxide to the bulk silicon substrate. For simplicity, only one island is shown here.Then, they deposited a polysilicon film on top. Photonic devices and electronic devices called transistors are made from this film. The former is located in the silicon-on-insulator area, and the latter is located in bulk silicon.
Another method of bringing the coolant in direct contact with the backside of the chip is embedded liquid cooling, where the cold liquid is pumped through straight parallel microchannels (SPMC) etched directly in the semiconductor device.This can effectively turn the back of the chip into a heat sink and provide excellent cooling performance.However, compared with other methods, the mold requires additional processing.The main disadvantage of SPMC is that with the passage of fluid, the pressure in the channel will greatly increase, which means that a high-power pump is required, which increases energy consumption and cost, and generates potentially destructive mechanical stress on the semiconductor device.Another big disadvantage is that there will be a high temperature gradient on the chip, which will cause thermomechanical stress and cause local warpage of the thin chip.
Three-dimensional cooling systems called embedded microchannel manifolds (EMMCs) have great potential for reducing pumping power requirements and temperature gradients compared to SPMCs.In these systems, a 3D layered manifold (a channel component with multiple ports for distributing coolant) provides multiple inlets and outlets for embedded microchannels, thereby dividing the coolant flow into multiple parallel sections.However, the integration of EMMC into the chip of the power electronic device increases the complexity and cost of constructing the device.Therefore, the previously reported EMMC is designed and manufactured as a separate module, which is then combined with a heat source or commercial chip to evaluate its cooling performance.
Van Erp et al. have achieved a breakthrough in this technology by developing what they call a monolithic integrated manifold microchannel (mMMC) system.The system integrates EMMC with chips in a single chip and co-manufactures them.Therefore, the buried channel is embedded directly below the effective area of the chip, allowing the coolant to pass directly under the heat source (Figure 2).
Figure 2 Integrated cooling system of the microchip.Van Erp et al. have developed the overall design of chips for electronic devices.The microchannel system is manufactured together with the chip and used as a cooling system.The cold water passes through the manifold and feeds the water into the microchannels made of silicon.The water passes directly under the gallium nitride layer of the semiconductor, which contains the components of the electronic device (not shown).Therefore, cold water can effectively dissipate the heat generated by the equipment, thereby ensuring good performance. The metal contacts on the top seal the channels.
The construction process of mMMC includes three steps.First, the slit is etched into a silicon substrate coated with a layer of semiconductor gallium nitride (GaN); the depth of the slit defines the depth of the channel to be created.Next, a process called isotropic gas etching is used to widen the gap in the silicon to the final width of the channel; this etching process also causes short portions of the channel to be connected to create a longer channel system.Finally, the opening in the GaN layer at the top of the channel is sealed with copper, and then electronic devices can be fabricated in the GaN layer.Unlike previously reported methods of making manifold microchannels, van Erp and colleagues' process does not require a bond or interface between the manifold and the device.
The authors also implemented their design and construction strategy to create power electronic modules that convert alternating current (ac) to direct current (dc).Experiments on this device have shown that only 0.57 W cm^-2 pump power can cool a heat flux exceeding 1.7 kW/cm².In addition, since the deterioration caused by self-heating is eliminated, the conversion efficiency of the liquid-cooled device is much higher than that of the uncooled device.
The results of Van Erp and his colleagues are impressive, but as with any technological advancement, there is still a lot of work to be done.For example, it is necessary to study the structural integrity of a thin GaN layer over time to understand its stabilization time.In addition, the author used an adhesive with a maximum operating temperature of 120°C to connect the microchannel in the device to the fluid delivery channel on the support circuit board.This means that the assembled system will not be able to withstand higher temperatures,For example, the typical temperature (250°C) involved in the reflow soldering process is a commonly used process in the manufacture of electronic equipment.Therefore, there will be a need to develop fluid connections that are compatible with the temperatures used in manufacturing.
Another future research direction is to apply the mMMC concept to the latest AC/DC converter design-the design reported by van Erp and colleagues is a simple test case.In addition, the author used only liquid water for single-phase cooling in the experiment (that is, the water did not become too hot to become a gas).It will be very useful to characterize the cooling and electrical performance of its equipment in a two-phase flow cooling system.In this system, heat is dissipated due to the evaporation of the fluid, and the water may freeze or directly contact the chip.Therefore, water may not be an ideal coolant for real-world applications.Future work should study the use of different liquid coolants.
Although there are still challenges to be solved, the work of van Erp and colleagues is a big step towards a low-cost, ultra-compact and energy-efficient power electronics cooling system.Their method is superior to state-of-the-art cooling technology and may make devices that generate high heat fluxes part of our daily lives.
Nature 585, 188-189 (2020)