A Thorough Review of Cooling Concepts and Thermal Management Techniques for Automotive WBG Inverters: Topology, Technology and Integration Level

: The development of electric vehicles (EVs) is an important step towards clean and green cities. An electric powertrain provides power to the vehicle and consists of a charger, a battery, an inverter, and a motor as the main components. Supplied by a battery pack, the automotive inverter manages the power of the motor. EVs require a highly efﬁcient inverter, which satisﬁes low cost, size, and weight requirements. One approach to meeting these requirements is to use the new wide-bandgap (WBG) semiconductors, which are being widely investigated in the industry as an alternative to silicon switches. WBG devices have superior intrinsic properties, such as high thermal ﬂux, of up to 120 W/cm 2 (on average); junction temperature of 175–200 ◦ C; blocking voltage limit of about 6.5 kV; switching frequency about 20-fold higher than that of Si; and up to 73% lower switching losses with a lower conduction voltage drop. This study presents a review of WBG-based inverter cooling systems to investigate trends in cooling techniques and changes associated with the use of WBG devices. The aim is to consider suitable cooling techniques for WBG inverters at different power levels. distribution in the IMD with a current-source inverter (CSI) for steady-state operation at a rated load (3 kW) with 25 ◦ C ambient air and an assumed air ﬂow of 1.03 × 10 − 2 m 3 /s. The maximum efﬁciency of the inverter was 97.7% with a peak temperature of 107 ◦ C for the WBG devices, and 96% for the machine with an inductor’s peak temperature of 101 ◦ C. These results show the feasibility of using air cooling for low-power IMD WBG-based solutions [13]. Another example of IMD cooling was presented in Lillo et al. in 2018. This is a liquid-cooled IMD in which each motor winding is connected to one power module through an innovative multilayer PCB structure, which is designed to minimize the space required for the end windings. The coolant ﬁrst passes through the power section, and the stator windings are then immersed in the same coolant. This solution is based on a direct substrate cooling method with targeted impingement cooling, which can provide an extremely low speciﬁc thermal resistance. It can be concluded that some techniques, which are not widely used and rarely applied alone, may improve the efﬁciency of matured liquid cooling. In Lillo’s study, the combination of spray cooling and immersion cooling provides effective liquid cooling in an IMD [14].


Introduction
The electric vehicle (EV) market has grown rapidly for several decades as an alternative to conventional internal-combustion engine (ICE) vehicles, resulting in reduced air pollution [1]. The trend of reducing emissions in the atmosphere is demonstrated by the results of the voluntary agreement between the European Commission and the European Automobile Manufacturers Association (ACEA), the Japanese Automobile Manufacturers Association, and the Korean Automobile Manufacturers Association. This agreement has a target to reduce CO 2 emissions by at least 37.5% compared to a 2021 baseline, which equates to the reduction of CO 2 emissions to 59.4 g/km by 2030 (fleet-average emissions are tested through the New European Driving Cycle) [2,3]. Moreover, the European Union (EU) has set targets to reduce greenhouse emissions in the 2030 climate and energy framework to 55% compared to 1990 [3]. According to the ACEA, the percentage of electrically chargeable and hybrid cars compared to the total of all newly registered cars in Europe increased from 4.3% to 22.4% between 2017 and 2020 ( Figure 1) [4]. The propulsion system plays a central role in providing tractive effort and energy management of an electric vehicle. The propulsion system includes the energy storage, power converters, electric motors, and associated controllers [5]. EV propulsion systems have been steadily improved, with the main development trends including higher power density, and reduction in size, weight, and cost [6]. The increase in power density and decrease in size can lead to poor heat dissipation and thermal stress, which in turn leads to additional failures of power converters [7]. Thus, to boost reliability, the main challenge of the propulsion's thermal management is to obtain a compromise between power density and size. This compromise allows thermal stress to be avoided and provides a better operating condition through a proper cooling system design [8].
Initially, cooling systems in EVs used forced air cooling, before evolving to liquid cooling as a more effective solution [9]. Because air cooling is less thermally efficient than liquid cooling, in current automotive applications the inverter air cooling is only possible for low power levels. Liquid cooling of EVs is usually provided using a low-temperature cooling loop (65 °C) for the power inverter [10,11], a high-temperature loop for the electric machine [12], and a separate battery-cooling loop (20-40 °C) [13][14][15]. This research focuses on the cooling design issues of automotive inverters with new wide-bandgap (WBG) semiconductors. Recent contributions regarding the advanced cooling of automotive inverters show two main approaches. The first approach is to eliminate the low temperature loop through high thermal performance cooling circuits, such as direct liquid cooling and dualsided cooling. [13,[16][17][18][19]. The second approach is to use air cooling for the inverter and machine, instead of separate cooling loops for each. An example is the integrated motor drive (IMD) [20]. The air-cooling loop for the IMDs requires the inverter to operate at high temperatures. Because the inverter operating temperatures are limited with silicon (Si) semiconductors, using new semiconductor materials may lead to wider applications of air cooling. WBG semiconductors are a possible substitution for Si semiconductors in automotive applications.
The thermal performance of a device can be evaluated using different parameters. The thermal resistance, Rth, is a parameter of power modules that indicates the heat transfer capability and determines the temperature difference between a heat source and a case or a heat sink, if specified. Therefore, the junction to heat sink thermal resistance value, Rthj-h, is a crucial parameter for heat dissipation efficiency. In addition, the design of power modules or cooling systems favors devices with a low Rth to maintain the junction temperature, Tj, within the acceptable range. The thermal resistance value is directly proportional to the material thickness (t) and inversely proportional to the thermal conductivity (σ) and dissipation area (A) [21][22][23][24][25][26][27][28].
High thermal performance of a power converter can be achieved in different ways, for example, using new materials with interesting thermal properties for power module manufacturing. Some of these materials can only be applied at manufacture, for example, The propulsion system plays a central role in providing tractive effort and energy management of an electric vehicle. The propulsion system includes the energy storage, power converters, electric motors, and associated controllers [5]. EV propulsion systems have been steadily improved, with the main development trends including higher power density, and reduction in size, weight, and cost [6]. The increase in power density and decrease in size can lead to poor heat dissipation and thermal stress, which in turn leads to additional failures of power converters [7]. Thus, to boost reliability, the main challenge of the propulsion's thermal management is to obtain a compromise between power density and size. This compromise allows thermal stress to be avoided and provides a better operating condition through a proper cooling system design [8].
Initially, cooling systems in EVs used forced air cooling, before evolving to liquid cooling as a more effective solution [9]. Because air cooling is less thermally efficient than liquid cooling, in current automotive applications the inverter air cooling is only possible for low power levels. Liquid cooling of EVs is usually provided using a lowtemperature cooling loop (65 • C) for the power inverter [10,11], a high-temperature loop for the electric machine [12], and a separate battery-cooling loop (20-40 • C) [13][14][15]. This research focuses on the cooling design issues of automotive inverters with new widebandgap (WBG) semiconductors. Recent contributions regarding the advanced cooling of automotive inverters show two main approaches. The first approach is to eliminate the low temperature loop through high thermal performance cooling circuits, such as direct liquid cooling and dual-sided cooling. [13,[16][17][18][19]. The second approach is to use air cooling for the inverter and machine, instead of separate cooling loops for each. An example is the integrated motor drive (IMD) [20]. The air-cooling loop for the IMDs requires the inverter to operate at high temperatures. Because the inverter operating temperatures are limited with silicon (Si) semiconductors, using new semiconductor materials may lead to wider applications of air cooling. WBG semiconductors are a possible substitution for Si semiconductors in automotive applications.
The thermal performance of a device can be evaluated using different parameters. The thermal resistance, R th , is a parameter of power modules that indicates the heat transfer capability and determines the temperature difference between a heat source and a case or a heat sink, if specified. Therefore, the junction to heat sink thermal resistance value, R thj-h , is a crucial parameter for heat dissipation efficiency. In addition, the design of power modules or cooling systems favors devices with a low R th to maintain the junction temperature, T j , within the acceptable range. The thermal resistance value is directly proportional to the material thickness (t) and inversely proportional to the thermal conductivity (σ) and dissipation area (A) [21][22][23][24][25][26][27][28].
High thermal performance of a power converter can be achieved in different ways, for example, using new materials with interesting thermal properties for power module manufacturing. Some of these materials can only be applied at manufacture, for example, new materials for substrates and semiconductors. Other thermal interface materials (TIMs) can be chosen at the time of device development [29]. Due to their commercial availability, power modules can be used as inverter switches [8]. Another approach is to change the basic concepts of cooling. Examples of this approach are the design of new power module packages with reduced thermal impedance, and new hybrid (air-liquid) or integrated (at the power module or inverter level) cooling techniques [30]. Thus, cooling circuit design tends to apply new materials with the lowest thermal resistance and, at the same time, implement an optimal cooling architecture. Combining these approaches provides the highest power density and the most efficient cooling of the device.
WBG semiconductors have striking intrinsic properties and are currently widely available commercially [31,32]. The power electronic industry has adopted this type of semiconductor during the past decade. As a result, numerous studies about WBG inverter cooling have been conducted. This paper presents a review of different WBG inverter cooling approaches to examine the changes in cooling associated with WBG devices. The paper describes cooling techniques for automotive inverters as a part of EVs, and explores how the use of WBG semiconductors may result in more compact and effective cooling systems.
The paper is divided into eight sections. Section 2 describes new WBG semiconductors with a focus on their thermal properties. Section 3 details the classification of cooling approaches for automotive inverters. Sections 4 and 5 include examples of recent inverter air-and liquid-cooling implementations, respectively. Section 6 focuses on promising cooling techniques, which may soon attract more attention, related to WBG semiconductors. Section 7 is dedicated to summarizing the main research findings and highlighting trends in automotive inverter cooling systems. Section 8 presents the main conclusions.

Wide-Bandgap Semiconductors
WBG semiconductors are a promising alternative to Si semiconductors [32][33][34][35][36][37]. Because the WBG switches are commercially available in wide ranges of current and voltage ratings, this technology is mature enough to be used in industry [26,27,33,38,39]. Compared to Si devices, WBG devices have a number of superior physical properties, such as high thermal flux of up to 120 W/cm 2 (Si-10 W/cm 2 ); high junction temperature, T jmax , of 175 to 200 • C (Si-T jmax = 150 • C); high limit of blocking voltage (Si-maximum 6.5 kV); higher switching frequency (about 20-fold compared to Si); and lower switching losses (up to 73% lower) with a lower conduction voltage drop [40][41][42]. Spider diagrams comparing the parameters of Si with those of SiC and GaN are shown in Figure 2 [43]. Despite the advantages of WBG devices and their decreasing cost, it should be noted that they are still more expensive than conventional semiconductors [44]. new materials for substrates and semiconductors. Other thermal interface materials (TIMs) can be chosen at the time of device development [29]. Due to their commercial availability, power modules can be used as inverter switches [8]. Another approach is to change the basic concepts of cooling. Examples of this approach are the design of new power module packages with reduced thermal impedance, and new hybrid (air-liquid) or integrated (at the power module or inverter level) cooling techniques [30]. Thus, cooling circuit design tends to apply new materials with the lowest thermal resistance and, at the same time, implement an optimal cooling architecture. Combining these approaches provides the highest power density and the most efficient cooling of the device.
WBG semiconductors have striking intrinsic properties and are currently widely available commercially [31,32]. The power electronic industry has adopted this type of semiconductor during the past decade. As a result, numerous studies about WBG inverter cooling have been conducted. This paper presents a review of different WBG inverter cooling approaches to examine the changes in cooling associated with WBG devices. The paper describes cooling techniques for automotive inverters as a part of EVs, and explores how the use of WBG semiconductors may result in more compact and effective cooling systems.
The paper is divided into eight sections. Section 2 describes new WBG semiconductors with a focus on their thermal properties. Section 3 details the classification of cooling approaches for automotive inverters. Sections 4 and 5 include examples of recent inverter air-and liquid-cooling implementations, respectively. Section 6 focuses on promising cooling techniques, which may soon attract more attention, related to WBG semiconductors. Section 7 is dedicated to summarizing the main research findings and highlighting trends in automotive inverter cooling systems. Section 8 presents the main conclusions.

Wide-Bandgap Semiconductors
WBG semiconductors are a promising alternative to Si semiconductors [32][33][34][35][36][37]. Because the WBG switches are commercially available in wide ranges of current and voltage ratings, this technology is mature enough to be used in industry [26,27,33,38,39]. Compared to Si devices, WBG devices have a number of superior physical properties, such as high thermal flux of up to 120 W/cm 2 (Si-10 W/cm 2 ); high junction temperature, Tjmax, of 175 to 200 °C (Si-Tjmax = 150 °C); high limit of blocking voltage (Si-maximum 6.5 kV); higher switching frequency (about 20-fold compared to Si); and lower switching losses (up to 73% lower) with a lower conduction voltage drop [40][41][42]. Spider diagrams comparing the parameters of Si with those of SiC and GaN are shown in Figure 2 [43]. Despite the advantages of WBG devices and their decreasing cost, it should be noted that they are still more expensive than conventional semiconductors [44].   The most developed and widespread WBG devices are those using gallium nitride (GaN) [45][46][47][48] and silicon carbide (SiC) [32,44,[49][50][51][52] semiconductors. These two types of WBG devices are becoming increasingly popular in the field of EVs due to their higher blocking voltage, frequency, and operation temperature [53,54]. The growth in the market revenue of WBG devices can be seen from the Yole reports' road maps (see Figure 3) [55,56]. Despite the fact that the GaN semiconductors are considered to be suitable for use in automotive inverters, GaN technology does not currently have many commercially available high voltage modules compared to SiC [57][58][59][60]. Therefore, most of the examples in this paper are based on SiC inverters. The most developed and widespread WBG devices are those using gallium nitrid (GaN) [45][46][47][48] and silicon carbide (SiC) [32,44,[49][50][51][52] semiconductors. These two types o WBG devices are becoming increasingly popular in the field of EVs due to their highe blocking voltage, frequency, and operation temperature [53,54]. The growth in the marke revenue of WBG devices can be seen from the Yole reports' road maps (see Figure 3 [55,56]. Despite the fact that the GaN semiconductors are considered to be suitable for us in automotive inverters, GaN technology does not currently have many commercially available high voltage modules compared to SiC [57][58][59][60]. Therefore, most of the example in this paper are based on SiC inverters. A previous study [61] using a WBG switch in a special high-temperature packag with integrated microchannel liquid cooling showed results for peak power dissipation capability up to 5 kW/cm 2 , i.e., significantly higher than that of a Si die. Theoretically these advantages could lead to positive changes in the cooling of a power converter, fo instance, by reducing the volume and production costs of the cooling system, combined with an increase in power density [44]. However, WBG technology is still constrained b limitations in its thermal packaging technology because of low thermal conductivit [62,63]. To increase operation temperatures up to the device's physical limitations (175 200 °C), high-temperature packaging should be designed. Examples can be found in th literature; for instance, Lu et al. presented a high-temperature packaged GaN High Elec tron Mobility Transistor (HEMT), which showed sufficient performance operating at 25 °C [64]. In addition, in 2020 Suganuma considered new highly conductive materials fo interconnections inside a package to provide optimal thermal resistance at operating tem peratures greater than 200 °C [65].
Due to their advantages, WBG devices have good thermal performance propertie that can justify optimization or downgrading of an application's existing cooling system for instance, using cheaper air cooling instead of liquid cooling. Overall, this can lowe the volume and cost of EV powertrains, which is the main trend in the EV industry. As an example, in 2011 Zhang H. et al. [53] performed an extensive comparison of Si and SiC inverters. As a result, the fuel-to-wheel efficiency of hybrid electric vehicles (HEVs) wa increased by 4.8% due to better regenerative braking. Fuel usage decreased from 3.94 to 3.36 L/100 km, which is consistent with an equivalent fuel economy improvement of abou 13.8%. The efficiency of plug-in hybrid electric vehicles (PHEVs) increased by 13.8%, and the average electricity consumption during the Urban Dynamometer Driving Schedul (UDDS) drive cycle was reduced from 425.1 to 308.1 J/m, representing a decrease of 27.5% due to the use of SiC switches. In addition, trends in EV inverters, as suggested by Zen et al. in 2020, show that WBG inverters have had a higher peak power and power density compared to Si devices during the past decade [18]. A previous study [61] using a WBG switch in a special high-temperature package with integrated microchannel liquid cooling showed results for peak power dissipation capability up to 5 kW/cm 2 , i.e., significantly higher than that of a Si die. Theoretically, these advantages could lead to positive changes in the cooling of a power converter, for instance, by reducing the volume and production costs of the cooling system, combined with an increase in power density [44]. However, WBG technology is still constrained by limitations in its thermal packaging technology because of low thermal conductivity [62,63]. To increase operation temperatures up to the device's physical limitations (175-200 • C), high-temperature packaging should be designed. Examples can be found in the literature; for instance, Lu et al. presented a high-temperature packaged GaN High Electron Mobility Transistor (HEMT), which showed sufficient performance operating at 250 • C [64]. In addition, in 2020 Suganuma considered new highly conductive materials for interconnections inside a package to provide optimal thermal resistance at operating temperatures greater than 200 • C [65].
Due to their advantages, WBG devices have good thermal performance properties that can justify optimization or downgrading of an application's existing cooling system, for instance, using cheaper air cooling instead of liquid cooling. Overall, this can lower the volume and cost of EV powertrains, which is the main trend in the EV industry. As an example, in 2011 Zhang H. et al. [53] performed an extensive comparison of Si and SiC inverters. As a result, the fuel-to-wheel efficiency of hybrid electric vehicles (HEVs) was increased by 4.8% due to better regenerative braking. Fuel usage decreased from 3.94 to 3.36 L/100 km, which is consistent with an equivalent fuel economy improvement of about 13.8%. The efficiency of plug-in hybrid electric vehicles (PHEVs) increased by 13.8%, and the average electricity consumption during the Urban Dynamometer Driving Schedule (UDDS) drive cycle was reduced from 425.1 to 308.1 J/m, representing a decrease of 27.5%, due to the use of SiC switches. In addition, trends in EV inverters, as suggested by Zeng et al. in 2020, show that WBG inverters have had a higher peak power and power density compared to Si devices during the past decade [18].

Classification of Existing Cooling Techniques
The cooling of the inverter aims to provide an acceptable junction temperature of the power semiconductors during operation. Operating at acceptable temperatures increases the reliability and efficiency of the device. The cooling systems for automotive inverters have adopted numerous different architectures and solutions ( Figure 4). Research on gas-cooled inverters, for example, can be found in Chintamani's paper [66]. Nevertheless, two fundamentally different approaches can be considered in inverter cooling: air and liquid cooling.

Classification of Existing Cooling Techniques
The cooling of the inverter aims to provide an acceptable junction temperature of the power semiconductors during operation. Operating at acceptable temperatures increases the reliability and efficiency of the device. The cooling systems for automotive inverters have adopted numerous different architectures and solutions ( Figure 4). Research on gascooled inverters, for example, can be found in Chintamani's paper [66]. Nevertheless, two fundamentally different approaches can be considered in inverter cooling: air and liquid cooling. Air cooling can be divided into forced and non-forced. The main advantages of air cooling are the implementation simplicity and low cost. Until recently, there has been little interest in air cooling for automotive inverters because the overall efficacy of this form of cooling is lower than that of liquid cooling. Thus, liquid cooling is a more popular and widespread solution. However, the emergence of WBG semiconductors with the ability to operate at higher temperatures has increased the interest in air cooling. Forced air cooling provides the principal operation efficiency with cheaper and easier cooling (see Section 4). Non-forced air cooling is not used in automotive applications because of the low heat dissipation capability.
Liquid cooling can be classified on basis of water-based or non-water-based coolant liquids. Water cooling is the simplest and most popular approach for automotive inverters, and non-water cooling usually implies that the driving motor in EVs uses oil cooling [67][68][69][70]. With regard to automotive inverter cooling, non-water cooling involves the use of water composites and has a special application in liquid cooling. The most popular water composite coolant is 50% water and 50% ethylene glycol [71] (see Section 5).
Regarding the type of contact with the coolant, liquid cooling can be categorized into direct and indirect cooling. Indirect cooling is an approach in which the power module has no direct contact with the liquid. As can be seen from Figure 5a, a conventional packaging scheme for indirect cooling includes a power device mounted through the solder over a ceramic substrate. A commonly used substrate is direct bonded copper (DBC). The substrate, in turn, is placed through DBC solder on the base plate (Cu, AlSiC, etc.). Electrical interconnections are provided via wire bonds. The entire construction is located on the heat sink or cold plate through TIMs to ensure heat dissipation [8]. All of these material and layer configurations affect the final thermal resistance. The difference between the direct and indirect liquid cooling is show in Figure 5. In direct liquid cooling, the base plate, TIM, and heat sink are replaced with only a base plate, thus reducing the number of layers inside the module (Figure 5b); as a result, the thermal resistance decreases. For instance, Mcpherson et al. (2017) compared the thermal resistance of three module configurations: (a) indirect cooling with a conventional flat cold plate; (b) direct cooling with a soldered pin-fin cold plate; and (c) direct cooling with a silver sintered pin-fin cold plate. The latter configuration almost halved the thermal resistance compared with indirect cooling [72]. In the case of indirect cooling, the thermal conductivity of the base plate, TIM, and heat sink play a key role in the overall heat dissipation capability [8,21,35,73]. The Air cooling can be divided into forced and non-forced. The main advantages of air cooling are the implementation simplicity and low cost. Until recently, there has been little interest in air cooling for automotive inverters because the overall efficacy of this form of cooling is lower than that of liquid cooling. Thus, liquid cooling is a more popular and widespread solution. However, the emergence of WBG semiconductors with the ability to operate at higher temperatures has increased the interest in air cooling. Forced air cooling provides the principal operation efficiency with cheaper and easier cooling (see Section 4). Non-forced air cooling is not used in automotive applications because of the low heat dissipation capability.
Liquid cooling can be classified on basis of water-based or non-water-based coolant liquids. Water cooling is the simplest and most popular approach for automotive inverters, and non-water cooling usually implies that the driving motor in EVs uses oil cooling [67][68][69][70]. With regard to automotive inverter cooling, non-water cooling involves the use of water composites and has a special application in liquid cooling. The most popular water composite coolant is 50% water and 50% ethylene glycol [71] (see Section 5).
Regarding the type of contact with the coolant, liquid cooling can be categorized into direct and indirect cooling. Indirect cooling is an approach in which the power module has no direct contact with the liquid. As can be seen from Figure 5a, a conventional packaging scheme for indirect cooling includes a power device mounted through the solder over a ceramic substrate. A commonly used substrate is direct bonded copper (DBC). The substrate, in turn, is placed through DBC solder on the base plate (Cu, AlSiC, etc.). Electrical interconnections are provided via wire bonds. The entire construction is located on the heat sink or cold plate through TIMs to ensure heat dissipation [8]. All of these material and layer configurations affect the final thermal resistance. The difference between the direct and indirect liquid cooling is show in Figure 5. In direct liquid cooling, the base plate, TIM, and heat sink are replaced with only a base plate, thus reducing the number of layers inside the module (Figure 5b); as a result, the thermal resistance decreases. For instance, Mcpherson et al. (2017) compared the thermal resistance of three module configurations: (a) indirect cooling with a conventional flat cold plate; (b) direct cooling with a soldered pin-fin cold plate; and (c) direct cooling with a silver sintered pin-fin cold plate. The latter configuration almost halved the thermal resistance compared with indirect cooling [72]. In the case of indirect cooling, the thermal conductivity of the base plate, TIM, and heat sink play a key role in the overall heat dissipation capability [8,21,35,73]. The direct cooling approach involves eliminating the cold plate or heat sink, in addition to the TIMs, and placing the power module in direct contact with water through the base plate only [72,74].
nergies 2021, 14, x FOR PEER REVIEW 6 of 21 direct cooling approach involves eliminating the cold plate or heat sink, in addition to the TIMs, and placing the power module in direct contact with water through the base plate only [72,74].
(a) (b) From the construction perspective, there are single-sided ( Figure 6a) and doublesided ( Figure 6d) liquid cooling systems [54]. In single-sided cooling, the power module is cooled only from one side. The standard single-sided cooling method uses a power module with a cold plate attached to the base plate trough the thermal interface materials [75,76]. Double-sided cooling involves changing the module construction in such a manner that two substrates are assembled one above the other with the power devices embedded between them [77]. The double-sided design allows the cooling efficiency to be increased because twice as much heat can be dissipated [28,78]. In addition, the bonding wires inside the module can be eliminated, which improves reliability [76]. The doublesided technique can be an efficient solution because the dissipation of power from both of the semiconductor's surfaces leads to a reduction in thermal impedance of up to 30% [79] (a) The cold plate configuration, when applicable, can affect the heat dissipation capability of a power module. The dissipation capacity is related to the heat dissipation surface, which is usually determined by the cold plate surface area. The more common structures of cold plates are flat, multiple channel, microchannel, and pin fin ( Figure 6). Regarding direct liquid cooling, when the cold plate is eliminated, the base plate can fulfill the functions of the cold plate and can also be implemented in different configurations.
In addition to standard air and liquid cooling, several other cooling techniques can be used for WBG inverters (see Section 6). These inverters have a smaller chip size but operate at higher temperatures. From the cooling system design perspective, this implies  [54]. In single-sided cooling, the power module is cooled only from one side. The standard single-sided cooling method uses a power module with a cold plate attached to the base plate trough the thermal interface materials [75,76]. Double-sided cooling involves changing the module construction in such a manner that two substrates are assembled one above the other with the power devices embedded between them [77]. The double-sided design allows the cooling efficiency to be increased because twice as much heat can be dissipated [28,78]. In addition, the bonding wires inside the module can be eliminated, which improves reliability [76]. The double-sided technique can be an efficient solution because the dissipation of power from both of the semiconductor's surfaces leads to a reduction in thermal impedance of up to 30% [79].
nergies 2021, 14, x FOR PEER REVIEW 6 of 21 direct cooling approach involves eliminating the cold plate or heat sink, in addition to the TIMs, and placing the power module in direct contact with water through the base plate only [72,74].
(a) (b) From the construction perspective, there are single-sided ( Figure 6a) and double sided ( Figure 6d) liquid cooling systems [54]. In single-sided cooling, the power module is cooled only from one side. The standard single-sided cooling method uses a power module with a cold plate attached to the base plate trough the thermal interface materials [75,76]. Double-sided cooling involves changing the module construction in such a man ner that two substrates are assembled one above the other with the power devices embed ded between them [77]. The double-sided design allows the cooling efficiency to be in creased because twice as much heat can be dissipated [28,78]. In addition, the bonding wires inside the module can be eliminated, which improves reliability [76]. The double sided technique can be an efficient solution because the dissipation of power from both o the semiconductor's surfaces leads to a reduction in thermal impedance of up to 30% [79] (a) (b) (c) (d) The cold plate configuration, when applicable, can affect the heat dissipation capa bility of a power module. The dissipation capacity is related to the heat dissipation sur face, which is usually determined by the cold plate surface area. The more common struc tures of cold plates are flat, multiple channel, microchannel, and pin fin ( Figure 6). Re garding direct liquid cooling, when the cold plate is eliminated, the base plate can fulfil the functions of the cold plate and can also be implemented in different configurations.
In addition to standard air and liquid cooling, several other cooling techniques can be used for WBG inverters (see Section 6). These inverters have a smaller chip size bu operate at higher temperatures. From the cooling system design perspective, this implies the use of hot and small areas to dissipate heat. Hoque et al. (2020) presented a modular The cold plate configuration, when applicable, can affect the heat dissipation capability of a power module. The dissipation capacity is related to the heat dissipation surface, which is usually determined by the cold plate surface area. The more common structures of cold plates are flat, multiple channel, microchannel, and pin fin ( Figure 6). Regarding direct liquid cooling, when the cold plate is eliminated, the base plate can fulfill the functions of the cold plate and can also be implemented in different configurations. In addition to standard air and liquid cooling, several other cooling techniques can be used for WBG inverters (see Section 6). These inverters have a smaller chip size but operate at higher temperatures. From the cooling system design perspective, this implies the use of hot and small areas to dissipate heat. Hoque et al. (2020) presented a modular heat sink for enhanced thermal performance of GaN switches [80]. Another approach to increasing the thermal performance of a small chip is to use an electric field technology; an example for GaN transistors can be found in [81]. Heat pipes may be a solution when some area require a higher heat dissipation capability. A jet impingement technique is also considered to be an effective cooling approach for automotive inverter applications [82]. Moreover, immersion cooling allows high heat transfer coefficients (h > 2 kW/(m 2 ·K)) to be achieved for the entire cooling system [83]. The basic principle of an immersion cooling technique is the submergence of the power modules in a thermally and not electrically conductive liquid [84]. In addition, integrated motor drive (IMD) topologies are currently gaining attention. Because of its proximity to the motor, the integrated inverter should work at higher temperatures. Therefore, WBG inverters, which have higher operating temperatures, are suitable for integration with the electrical machine. However, the proper cooling design of the integrated system remains important, as shown in recent studies [85].

Air Cooling
Air cooling is the simplest and cheapest cooling approach. The main problem with air cooling is a low heat flux. As a result, this technique is only suitable for low power applications. However, the high rated temperature and thermal flux dissipation capability of WBG semiconductors can compensate for the low flux of air cooling. Although the power density of the liquid-cooled SiC inverters is higher than that of the Si inverters, a liquid-cooling system implies additional components, which increase the volume and reduce the overall power density. Thus, the air-cooling system, which is the simplest implementation, can increase the power density of the inverter [18]. Hensler et al. (2017) proposed an unusual air-cooled inverter configuration, in which the heat sink has the shape of a hexagon, and the inner volume of the heat sink is separated by a plastic tube as an air path, resulting in a very compact solution. The principle of the hexagon heat sink is depicted in Figure 7a [86]. Another example was developed by Liu et al. using 3D aircooled SiC packaging for a 20 kW inverter (see the concept in Figure 7b) [87]. The research of  showed an air-cooled 500 kW modular inverter with a conventional heat sink, in which the efficiency of the inverter was about 98.74% [88]. increasing the thermal performance of a small chip is to use an electric field technology; an example for GaN transistors can be found in [81]. Heat pipes may be a solution when some area require a higher heat dissipation capability. A jet impingement technique is also considered to be an effective cooling approach for automotive inverter applications [82]. Moreover, immersion cooling allows high heat transfer coefficients (h > 2 kW/(m 2 ·K)) to be achieved for the entire cooling system [83]. The basic principle of an immersion cooling technique is the submergence of the power modules in a thermally and not electrically conductive liquid [84]. In addition, integrated motor drive (IMD) topologies are currently gaining attention. Because of its proximity to the motor, the integrated inverter should work at higher temperatures. Therefore, WBG inverters, which have higher operating temperatures, are suitable for integration with the electrical machine. However, the proper cooling design of the integrated system remains important, as shown in recent studies [85].

Air Cooling
Air cooling is the simplest and cheapest cooling approach. The main problem with air cooling is a low heat flux. As a result, this technique is only suitable for low power applications. However, the high rated temperature and thermal flux dissipation capability of WBG semiconductors can compensate for the low flux of air cooling. Although the power density of the liquid-cooled SiC inverters is higher than that of the Si inverters, a liquid-cooling system implies additional components, which increase the volume and reduce the overall power density. Thus, the air-cooling system, which is the simplest implementation, can increase the power density of the inverter [18]. Hensler et al. (2017) proposed an unusual air-cooled inverter configuration, in which the heat sink has the shape of a hexagon, and the inner volume of the heat sink is separated by a plastic tube as an air path, resulting in a very compact solution. The principle of the hexagon heat sink is depicted in Figure 7a [86]. Another example was developed by Liu et al. using 3D air-cooled SiC packaging for a 20 kW inverter (see the concept in Figure 7b) [87]. The research of  showed an air-cooled 500 kW modular inverter with a conventional heat sink, in which the efficiency of the inverter was about 98.74% [88]. Zeng et al. proposed a design of a light and compact EV inverter without a liquid cooling system. The highest temperature reached by the SiC MOSFET was 110 °C, which is the rated temperature of WBG semiconductors. It was also proved that the forced aircooling technique provides a thermal flux of 200 W/cm 2 , and the SiC MOSFET module has its own heat flux of 120 W/cm 2 , which is 10-fold higher than that of the Si alternative. The high heat flux of the WBG module means that heat can be more easily dissipated from the chip surface. Thus, the smaller heat flux of air cooling, compared to liquid cooling, is cooling technique provides a thermal flux of 200 W/cm 2 , and the SiC MOSFET module has its own heat flux of 120 W/cm 2 , which is 10-fold higher than that of the Si alternative. The high heat flux of the WBG module means that heat can be more easily dissipated from the chip surface. Thus, the smaller heat flux of air cooling, compared to liquid cooling, is sufficient for WBG power converters. Moreover, the optimized design of the Al heat sink reduces the thermal resistance of the heatsink from 0.44 to 0.35 K/W [18].
Another air-cooled SiC inverter was introduced by Wang et al. (2019), in which the power module was cooled with a thermally optimized heat sink (the principle is shown Figure 8). The optimized heat sink was generated using a MATLAB Genetic Algorithm (GA) and manufactured using 3D printing technology. This comprehensive heat sink allows sufficient heat to be dissipated from the semiconductors to enable the use of air cooling rather than liquid cooling. Moreover, two modules in one phase were split into two submodules, i.e., a top-side module and a bottom-side module, attached to two separate heat sinks, which were placed one above the other for better heat dissipation. Thermal optimization resulted in a reduction in the volume of the heat sink of 27% and reduced the junction temperature from 108.3 to 102.1 • C [89].  2019), in which the power module was cooled with a thermally optimized heat sink (the principle is shown Figure 8). The optimized heat sink was generated using a MATLAB Genetic Algorithm (GA) and manufactured using 3D printing technology. This comprehensive heat sink allows sufficient heat to be dissipated from the semiconductors to enable the use of air cooling rather than liquid cooling. Moreover, two modules in one phase were split into two submodules, i.e., a top-side module and a bottom-side module, attached to two separate heat sinks, which were placed one above the other for better heat dissipation. Thermal optimization resulted in a reduction in the volume of the heat sink of 27% and reduced the junction temperature from 108.3 to 102.1 °C [89].

Liquid Cooling
Liquid cooling is currently the best option for EVs because of higher thermal transfer capability compared to air. Moreover, for power levels applied to automotive inverters, liquid cooling has proven effectiveness, which justifies the cost and challenges in implementation [74]. Indirect liquid cooling with flat cold plate is a mature technology and can now be found in the literature as a "conventional" or "standard" structure (with Rth ≥ 0.8 K/W). Direct and indirect liquid cooling structures with complex cold plates are more advanced topologies [73]. Nevertheless, the use of the direct liquid cooling approach has increased because it shows the best thermal performance. By eliminating some layers of the power module solution, such as the TIM layer and the cold plate, the direct cooling approach allows the thermal resistance of the power module to be reduced by up to 30% [90]. Among the wide range of direct liquid cooling approaches, jet impingement (0.44-0.48 K/W), turbulator (0.2-0.55 K/W), and microchannel (0.13-0.24 K/W) technologies can be distinguished as major groups [91]. The direct liquid approach shows better thermal properties, such as a halving in the thermal resistance compared to indirect liquid cooling [92]. However, direct liquid cooling has a more complex implementation process, which results in higher costs. Thus, the application of direct liquid cooling is justified only to achieve higher thermal performance. It is worth noting that there can be a power module scale for power electronic devices, called integrated power module cooling, and a power converter scale to evaluate the efficiency and thermal performance. In the following subsections, single-sided and double-sided cooling approaches are discussed for direct and indirect cooling implementation.

Liquid Cooling
Liquid cooling is currently the best option for EVs because of higher thermal transfer capability compared to air. Moreover, for power levels applied to automotive inverters, liquid cooling has proven effectiveness, which justifies the cost and challenges in implementation [74]. Indirect liquid cooling with flat cold plate is a mature technology and can now be found in the literature as a "conventional" or "standard" structure (with R th ≥ 0.8 K/W). Direct and indirect liquid cooling structures with complex cold plates are more advanced topologies [73]. Nevertheless, the use of the direct liquid cooling approach has increased because it shows the best thermal performance. By eliminating some layers of the power module solution, such as the TIM layer and the cold plate, the direct cooling approach allows the thermal resistance of the power module to be reduced by up to 30% [90]. Among the wide range of direct liquid cooling approaches, jet impingement (0.44-0.48 K/W), turbulator (0.2-0.55 K/W), and microchannel (0.13-0.24 K/W) technologies can be distinguished as major groups [91]. The direct liquid approach shows better thermal properties, such as a halving in the thermal resistance compared to indirect liquid cooling [92]. However, direct liquid cooling has a more complex implementation process, which results in higher costs. Thus, the application of direct liquid cooling is justified only to achieve higher thermal performance. It is worth noting that there can be a power module scale for power electronic devices, called integrated power module cooling, and a power converter scale to evaluate the efficiency and thermal performance. In the following subsections, single-sided and double-sided cooling approaches are discussed for direct and indirect cooling implementation.

Single-Sided Cooling
Single-sided liquid cooling is still the most widespread and cost-effective solution for automotive inverters. At present, indirect liquid cooling development is represented by new structures of cold plates and heat sinks, where applicable. For instance, Mademlis et al.
(2021) presented a heat sink with rectangular pin fins and optimized cooling channels [93]. The most common cold plate type for WBG inverters in recent contributions has a pin fin structure [94].
Qi et al. (2019) investigated a 30 kVA three-phase SiC inverter with a liquid-cooled cold plate for high-temperature operation. The inverter continuously operated for several hours in a high-temperature thermal chamber, where an ambient temperature of 180 • C was achieved. The maximum temperature of the commercial SiC modules, in the special high-temperature packaging, reached 150 • C, whereas the temperature of the cold plate and coolant were 85 and 50 • C, respectively [95]. In [96], a 100 kW SiC MOSFET three-phase inverter (34 kW/L) with a flat cold plate was tested at 105 • C ambient and 65 • C coolant temperatures. The thermal resistance from the junction to the coolant was 0.16 K/W, and the maximum die temperature was 113 • C. A two-level optimized cold plate was presented in Gurpinar et al. (2018) for a liquid-cooled 125 kW inverter, and the results showed a high thermal performance at a maximum operating capability of SiC-MOSFETs [97]. The twolevel cold plate cools a power module and a DC link capacitor bank from different sides; this concept is depicted in Figure 9. Another example of cold plate thermal optimization can be found in Huber et al. (2018), in which six different SiC power modules were presented with a minimal thermal resistance of 0.331 K/W [98].

Single-Sided Cooling
Single-sided liquid cooling is still the most widespread and cost-effective solution for automotive inverters. At present, indirect liquid cooling development is represented by new structures of cold plates and heat sinks, where applicable. For instance, Mademlis et al. (2021) presented a heat sink with rectangular pin fins and optimized cooling channels [93]. The most common cold plate type for WBG inverters in recent contributions has a pin fin structure [94].
Qi et al. (2019) investigated a 30 kVA three-phase SiC inverter with a liquid-cooled cold plate for high-temperature operation. The inverter continuously operated for several hours in a high-temperature thermal chamber, where an ambient temperature of 180 °C was achieved. The maximum temperature of the commercial SiC modules, in the special high-temperature packaging, reached 150 °C, whereas the temperature of the cold plate and coolant were 85 and 50 °C, respectively [95]. In [96], a 100 kW SiC MOSFET threephase inverter (34 kW/L) with a flat cold plate was tested at 105 °C ambient and 65 °C coolant temperatures. The thermal resistance from the junction to the coolant was 0.16 K/W, and the maximum die temperature was 113 °C. A two-level optimized cold plate was presented in Gurpinar et al. (2018) for a liquid-cooled 125 kW inverter, and the results showed a high thermal performance at a maximum operating capability of SiC-MOSFETs [97]. The two-level cold plate cools a power module and a DC link capacitor bank from different sides; this concept is depicted in Figure 9. Another example of cold plate thermal optimization can be found in Huber et al. (2018), in which six different SiC power modules were presented with a minimal thermal resistance of 0.331 K/W [98]. Becker et al. carried out thermal optimization of a three-phase SiC MOSFET 1200 V 150 A automotive inverter with indirect liquid 50/50% water-glycol cooling and a copper heat sink with round pins. A cooling liquid circulates inside an aluminum case with a liquid pressure drop of 25 kPa at the pin area. The SiC power module with high conductivity Si3N4 (90 W/(m·K)) substrate was pressure contacted with a pin fin heat sink. The experiment included changing the number of parallel modules and the distance between them. It was found that if the distance between the chips was reduced from 8 to 1 mm, only three parallel SiC devices could be used rather than 4. The simulation of the complete assembled inverter showed that the junction temperature of the hottest chip was reduced from 226 to 154 °C with small temperature differences of ∆T = 5.1 °C between the chips. Although 154 °C is still high, the maximum Tj of the SiC semiconductor used in this research was around 170 °C. Moreover, the hot spot temperature can be optimized by changing the inlet liquid temperature and flow. As a result, the design allowed the thermal resistance to be reduced by 32.1% [99].

Double-Sided Cooling
Double-sided liquid cooling is considered for motor driving applications. For instance, in 2020 Liu et al. [100] presented double-sided cooling of the SiC power module, and Möller et al. [92] suggested a double-sided cooling concept for an automotive Becker et al. carried out thermal optimization of a three-phase SiC MOSFET 1200 V 150 A automotive inverter with indirect liquid 50/50% water-glycol cooling and a copper heat sink with round pins. A cooling liquid circulates inside an aluminum case with a liquid pressure drop of 25 kPa at the pin area. The SiC power module with high conductivity Si 3 N 4 (90 W/(m·K)) substrate was pressure contacted with a pin fin heat sink. The experiment included changing the number of parallel modules and the distance between them. It was found that if the distance between the chips was reduced from 8 to 1 mm, only three parallel SiC devices could be used rather than 4. The simulation of the complete assembled inverter showed that the junction temperature of the hottest chip was reduced from 226 to 154 • C with small temperature differences of ∆T = 5.1 • C between the chips. Although 154 • C is still high, the maximum T j of the SiC semiconductor used in this research was around 170 • C. Moreover, the hot spot temperature can be optimized by changing the inlet liquid temperature and flow. As a result, the design allowed the thermal resistance to be reduced by 32.1% [99].

Double-Sided Cooling
Double-sided liquid cooling is considered for motor driving applications. For instance, in 2020 Liu et al. [100] presented double-sided cooling of the SiC power module, and Möller et al. [92] suggested a double-sided cooling concept for an automotive application. Moreover, a double-sided liquid cooling module with jet impingement was reported in Tang et al. [101].
Catalano et al. [28] carried out an investigation of SiC-based power modules arranged in a half-bridge to prove the efficiency of double-sided cooling (the concept of the doublesided module is shown in Figure 10). To compare the thermal performance of single-and double-sided cooling, the thermal resistance was calculated for a wide range of boundary conditions. The heat transfer coefficient was used as a figure of merit for the boundary conditions; thus, h = 101 W/(m 2 ·K) means natural convection and h = 109 W/(m 2 ·K) indicates contact with an ideal heat sink. Moreover, two direct-bonded copper substrates based on Si 3 N 4 and AlN were compared. According to the research results, double-sided cooling shows lower thermal resistance only when a high heat transfer coefficient is achieved (h > 106 W/(m 2 ·K)). In the case of the high heat transfer coefficient and the AlN type of substrate, the reduction of thermal resistance was up to 70%. It should be noted that the cost of the AlN substrate is high; thus, using this substrate is justified only if highly efficient cooling is needed.
Energies 2021, 14, x FOR PEER REVIEW 10 of 2 application. Moreover, a double-sided liquid cooling module with jet impingement wa reported in Tang et al. [101]. Catalano et al. [28] carried out an investigation of SiC-based power modules arrange in a half-bridge to prove the efficiency of double-sided cooling (the concept of the double sided module is shown in Figure 10). To compare the thermal performance of single-an double-sided cooling, the thermal resistance was calculated for a wide range of boundar conditions. The heat transfer coefficient was used as a figure of merit for the boundar conditions; thus, h = 101 W/(m 2 ·K) means natural convection and h = 109 W/(m 2 ·K) ind cates contact with an ideal heat sink. Moreover, two direct-bonded copper substrate based on Si3N4 and AlN were compared. According to the research results, double-side cooling shows lower thermal resistance only when a high heat transfer coefficient i achieved (h > 106 W/(m 2 ·K)). In the case of the high heat transfer coefficient and the AlN type of substrate, the reduction of thermal resistance was up to 70%. It should be note that the cost of the AlN substrate is high; thus, using this substrate is justified only if highl efficient cooling is needed. In 2018, Hitachi's Research & Development Group developed a full-bridge SiC auto motive inverter with an improved thermal design. According to the research result, th double-sided liquid cooling application ensured a 35% reduction in thermal resistanc with a final value of about 0.175 K/W. Moreover, the double-sided structure enabled solution to be derived for the issue that occurs when the parallel connections of semicon ductors are required to share a large output current. In common structures, in which a interconnections are implemented via Al wires, the imbalance in the current can be con siderable, especially when the switching frequency is high. A chip-connecting structur was built to provide an isometric current pattern, leading to a reduction in the curren deviation between each chip from 10% to 2%. As a result, the overall design was able t reduce power losses by 60% in comparison to the conventional single-side cooled silico modules [76].

Other Promising Cooling Approaches
The tendency toward higher powers and smaller sizes requires more efficient coolin with high heat flux. Although WBG devices differ from Si devices in terms of a large num ber of parameters, the most significant in terms of cooling are the smaller chip size an high heat flux due to the intrinsic material properties. When combined, these two proper ties are challenging for a cooling system, because it should dissipate a greater amount o heat from a smaller size. To ensure uniform temperature distribution, a cooling system should have high dissipation capability. WBG semiconductors become "hot points" dur ing unbalanced temperature distribution, which lowers the reliability of the entire device Thus, the WBG inverters' cooling requires a high cooling capacity and homogenous tem perature maintenance of the power modules' chips. This section presents cooling tech niques and approaches that provide a high thermal performance and may be successfull In 2018, Hitachi's Research & Development Group developed a full-bridge SiC automotive inverter with an improved thermal design. According to the research result, the double-sided liquid cooling application ensured a 35% reduction in thermal resistance with a final value of about 0.175 K/W. Moreover, the double-sided structure enabled a solution to be derived for the issue that occurs when the parallel connections of semiconductors are required to share a large output current. In common structures, in which all interconnections are implemented via Al wires, the imbalance in the current can be considerable, especially when the switching frequency is high. A chip-connecting structure was built to provide an isometric current pattern, leading to a reduction in the current deviation between each chip from 10% to 2%. As a result, the overall design was able to reduce power losses by 60% in comparison to the conventional single-side cooled silicon modules [76].

Other Promising Cooling Approaches
The tendency toward higher powers and smaller sizes requires more efficient cooling with high heat flux. Although WBG devices differ from Si devices in terms of a large number of parameters, the most significant in terms of cooling are the smaller chip size and high heat flux due to the intrinsic material properties. When combined, these two properties are challenging for a cooling system, because it should dissipate a greater amount of heat from a smaller size. To ensure uniform temperature distribution, a cooling system should have high dissipation capability. WBG semiconductors become "hot points" during unbalanced temperature distribution, which lowers the reliability of the entire device. Thus, the WBG inverters' cooling requires a high cooling capacity and homogenous temperature maintenance of the power modules' chips. This section presents cooling techniques and approaches that provide a high thermal performance and may be successfully applied to emerging WBG power electronics systems.
New materials could be used to enhance the heat dissipating ability and provide a balanced temperature distribution. The use of metal foam for a heat sink, for instance, was presented by Lee et al. with Samsung Electronics Co in 2020. The authors investigated two heat sink structures: a metal foam (Type 1), and a metal foam with a pin fin baseplate (Type 2) for high power IGBTs. The concept of a metal foam layout in a power module is shown for Type 1 and Type 2 in Figure 11a and 11b, respectively. Porous heat sinks provide better heat transfer with a low-pressure drop in comparison to a microchannel technology, which usually requires pressure drops of 30-100 kPa [102]. The presented liquid cooling design provides a low R th = 0.185 K/W and a low inlet pressure P in = 5-15 kPa. Although the results were presented for IGBTs, the high thermal flux capability of porous heat sinks makes this technology suitable for WBGs. For instance, Takai et al. (2017) suggested the use of a porous heat sink for a GaN-based inverter [103]. heat sink structures: a metal foam (Type 1), and a metal foam with a pin fin baseplate (Type 2) for high power IGBTs. The concept of a metal foam layout in a power module is shown for Type 1 and Type 2 in Figure 11a and 11b, respectively. Porous heat sinks provide better heat transfer with a low-pressure drop in comparison to a microchannel technology, which usually requires pressure drops of 30-100 kPa [102]. The presented liquid cooling design provides a low Rth = 0.185 K/W and a low inlet pressure Pin = 5-15 kPa. Although the results were presented for IGBTs, the high thermal flux capability of porous heat sinks makes this technology suitable for WBGs. For instance, Takai  The optimal structure of a cold plate can also increase the heat transfer capability of the device. Cold plates with direct single-sided liquid cooling have a higher thermal performance and simpler implementation process compared, for example, to double-sided modules. Thus, to design highly effective cooling for a WBG inverter, complex cold plate structures can be used. The simplest approach to enhance the effectiveness of a cold plate or base plate is to develop a structure with a larger surface for heat removal. In 2018, for instance, Infineon Technologies presented the liquid direct cooled Hybrid-PACKTM with different base plate structures. The thermal experiment included the measurement of three cold plate types: ribbon-bonded (this concept is depicted in Figure 12), pin fin, and flat. At the flow rate of 10 l/min, the result for the flat plate was Rth = 0.153 K/W, for the ribbon bonded plate Rth = 0.132 K/W, and for the pin fin structure Rth = 0.125 K/W [73]. Despite the fact that the thermal resistance of the pin fin cold plate was lower, the difference was not large (about 5.6%), and the manufacture of a ribbon-bonded cold plate is simpler and cheaper [90]. Heat pipes are passive two-phase heat transfer devices. This cooling approach includes an evaporator, a condenser, a vapor line, a liquid line, and a vapor-removal channel [61]. It should be noted, however, that two-phase cooling systems, in general, are more effective than single phase systems [104]. The operational principle of a heat pipe, as The optimal structure of a cold plate can also increase the heat transfer capability of the device. Cold plates with direct single-sided liquid cooling have a higher thermal performance and simpler implementation process compared, for example, to double-sided modules. Thus, to design highly effective cooling for a WBG inverter, complex cold plate structures can be used. The simplest approach to enhance the effectiveness of a cold plate or base plate is to develop a structure with a larger surface for heat removal. In 2018, for instance, Infineon Technologies presented the liquid direct cooled Hybrid-PACKTM with different base plate structures. The thermal experiment included the measurement of three cold plate types: ribbon-bonded (this concept is depicted in Figure 12), pin fin, and flat. At the flow rate of 10 l/min, the result for the flat plate was R th = 0.153 K/W, for the ribbon bonded plate R th = 0.132 K/W, and for the pin fin structure R th = 0.125 K/W [73]. Despite the fact that the thermal resistance of the pin fin cold plate was lower, the difference was not large (about 5.6%), and the manufacture of a ribbon-bonded cold plate is simpler and cheaper [90].
Energies 2021, 14, x FOR PEER REVIEW heat sink structures: a metal foam (Type 1), and a metal foam with a pin fin (Type 2) for high power IGBTs. The concept of a metal foam layout in a power m shown for Type 1 and Type 2 in Figure 11a and 11b, respectively. Porous heat s vide better heat transfer with a low-pressure drop in comparison to a microchan nology, which usually requires pressure drops of 30-100 kPa [102]. The present cooling design provides a low Rth = 0.185 K/W and a low inlet pressure Pin = 5 Although the results were presented for IGBTs, the high thermal flux capability o heat sinks makes this technology suitable for WBGs. For instance, Takai et al. (2 gested the use of a porous heat sink for a GaN-based inverter [103]. The optimal structure of a cold plate can also increase the heat transfer cap the device. Cold plates with direct single-sided liquid cooling have a higher the formance and simpler implementation process compared, for example, to dou modules. Thus, to design highly effective cooling for a WBG inverter, complex c structures can be used. The simplest approach to enhance the effectiveness of a c or base plate is to develop a structure with a larger surface for heat removal. In instance, Infineon Technologies presented the liquid direct cooled Hybrid-PACK different base plate structures. The thermal experiment included the measur three cold plate types: ribbon-bonded (this concept is depicted in Figure 12), pin flat. At the flow rate of 10 l/min, the result for the flat plate was Rth = 0.153 K/W ribbon bonded plate Rth = 0.132 K/W, and for the pin fin structure Rth = 0.125 K Despite the fact that the thermal resistance of the pin fin cold plate was lower, t ence was not large (about 5.6%), and the manufacture of a ribbon-bonded cold simpler and cheaper [90]. Heat pipes are passive two-phase heat transfer devices. This cooling app cludes an evaporator, a condenser, a vapor line, a liquid line, and a vapor-remova [61]. It should be noted, however, that two-phase cooling systems, in general, effective than single phase systems [104]. The operational principle of a heat Heat pipes are passive two-phase heat transfer devices. This cooling approach includes an evaporator, a condenser, a vapor line, a liquid line, and a vapor-removal channel [61]. It should be noted, however, that two-phase cooling systems, in general, are more effective than single phase systems [104]. The operational principle of a heat pipe, as shown in Figure 13, is to dissipate the heat from small hot sources to larger areas through evaporation and condensation [105,106]. This circulation of liquid leads to an increased heat transfer and, most importantly, provides higher thermal uniformity [107]. The heat transfer capability of heat pipes is determined only by the cooling capacity of the refrigerant [79]. Moreover, adding heat pipes allows the same heat transfer level to be obtained from smaller devices [108]. Therefore, heat pipes in cooling systems are particularly suitable for WBG devices because they provide higher heat dissipation from small areas, leading to a better temperature distribution among power devices. For instance, Gupta et al. (2018) presented cold plates with embedded heat pipes that reduced the maximum temperature of the device by 37 • C and improved temperature distribution [108]. In addition, Chang et al. proposed a heat pipe integrated heat sink, which decreased the temperature of the heatsink by 6 • C and the thermal resistance between the coolant and chip junction 10-fold [109].
ies 2021, 14, x FOR PEER REVIEW transfer capability of heat pipes is determined only by the cooli erant [79]. Moreover, adding heat pipes allows the same heat tran from smaller devices [108]. Therefore, heat pipes in cooling syste able for WBG devices because they provide higher heat dissipatio ing to a better temperature distribution among power devices. F (2018) presented cold plates with embedded heat pipes that red perature of the device by 37 °C and improved temperature distri Chang et al. proposed a heat pipe integrated heat sink, which de of the heatsink by 6 °C and the thermal resistance between the c 10-fold [109]. Immersion cooling is one of the promising alternatives to c ing. This method consists of placing heat sources with direct con to dissipate the generated heat (the principle is shown in Figure  requirement for immersion cooling systems is to use an electrical ing fluid, which allows homogenous cooling. Equal thermal distr of increasing the reliability for high power density power device fluids suitable for immersion cooling are de-ionized water, m based fluids, and synthetic oil [83]. Immersion cooling, in gener or two-phase. The former includes using coolants only in one s the latter implies physical state changes. This cooling approach h coefficient, which is important for increasing electrical power r implemented as one cooling loop for the inverter and motor in E main motor 105 °C coolant loop [68]. Yuki et al. (2020) presented immersion-cooled SiC inverter with a high heat flux of 500 W/cm a cooling technique can meet all of the requirements for automo simple, compact, and highly effective [110]. Immersion cooling is one of the promising alternatives to conventional liquid cooling. This method consists of placing heat sources with direct contact in circulating liquid to dissipate the generated heat (the principle is shown in Figure 14). The most important requirement for immersion cooling systems is to use an electrically non-conductive working fluid, which allows homogenous cooling. Equal thermal distribution is crucial in terms of increasing the reliability for high power density power devices [104]. Potential types of fluids suitable for immersion cooling are de-ionized water, mineral oil, fluorocarbonbased fluids, and synthetic oil [83]. Immersion cooling, in general, is either single-phase or two-phase. The former includes using coolants only in one state of aggregation, and the latter implies physical state changes. This cooling approach has a higher heat transfer coefficient, which is important for increasing electrical power requirements, and can be implemented as one cooling loop for the inverter and motor in EVs, thus eliminating the main motor 105 • C coolant loop [68]. Yuki et al. (2020) presented an on-board two-phase immersion-cooled SiC inverter with a high heat flux of 500 W/cm 2 , and showed that such a cooling technique can meet all of the requirements for automotive cooling because it is simple, compact, and highly effective [110].
The integrated motor drive (IMD) has emerged on the automotive powertrain market, and is rapidly evolving to become a key technology for compact and efficient motor drives [111]. Cooling technologies are being developed with integrated motor drives [85]. The recent literature has shown interest in WBG inverters for integrated solutions [112][113][114][115], in which air cooling can be applied for low-power motor drives. For instance, Dai et al. shared an air-cooled IMD concept in which the operation procedure of the air-cooling system is as follows: (1) inlet air is first drawn radially across the heat sink, which conducts heat out of the power modules; (2) the air is then pulled axially through the motor's airgap to cool the machine. A thermal finite element analysis showed the predicted temperature distribution in the IMD with a current-source inverter (CSI) for steady-state operation at a rated load (3 kW) with 25 • C ambient air and an assumed air flow of 1.03×10 −2 m 3 /s. The maximum efficiency of the inverter was 97.7% with a peak temperature of 107 • C for the WBG devices, and 96% for the machine with an inductor's peak temperature of 101 • C. These results show the feasibility of using air cooling for low-power IMD WBG-based solutions [13]. Another example of IMD cooling was presented in Lillo et al. in 2018. This is a liquid-cooled IMD in which each motor winding is connected to one power module through an innovative multilayer PCB structure, which is designed to minimize the space required for the end windings. The coolant first passes through the power section, and the stator windings are then immersed in the same coolant. This solution is based on a direct substrate cooling method with targeted impingement cooling, which can provide an extremely low specific thermal resistance. It can be concluded that some techniques, which are not widely used and rarely applied alone, may improve the efficiency of matured liquid cooling. In Lillo's study, the combination of spray cooling and immersion cooling provides effective liquid cooling in an IMD [14]. Immersion cooling is one of the promising alternatives to conventional liqu ing. This method consists of placing heat sources with direct contact in circulatin to dissipate the generated heat (the principle is shown in Figure 14). The most im requirement for immersion cooling systems is to use an electrically non-conductiv ing fluid, which allows homogenous cooling. Equal thermal distribution is crucial of increasing the reliability for high power density power devices [104]. Potential fluids suitable for immersion cooling are de-ionized water, mineral oil, fluoro based fluids, and synthetic oil [83]. Immersion cooling, in general, is either singl or two-phase. The former includes using coolants only in one state of aggregati the latter implies physical state changes. This cooling approach has a higher heat coefficient, which is important for increasing electrical power requirements, and implemented as one cooling loop for the inverter and motor in EVs, thus elimina main motor 105 °C coolant loop [68]. Yuki et al. (2020) presented an on-board tw immersion-cooled SiC inverter with a high heat flux of 500 W/cm 2 , and showed th a cooling technique can meet all of the requirements for automotive cooling beca simple, compact, and highly effective [110]. The integrated motor drive (IMD) has emerged on the automotive powertra ket, and is rapidly evolving to become a key technology for compact and efficien

Comparison of Cooling Techniques
WBG inverters operate at higher switching frequencies, power levels, and temperatures. The properties of WBG devices comply with increasing power levels and switching frequency requirements for power electronics devices. As shown by the literature review in this paper, the main advantages of WBG devices in cooling design are:

•
The higher operating temperature of WBG devices leads to the design and use of simpler and cheaper cooling techniques; • The higher switching frequencies and lower losses increase the overall efficiency of the device; • The smaller size and higher heat flux of the WBG die is suitable for compact devices, although this is challenging for cooling system design because more heat should be dissipated from a smaller area.
A summary of cooling techniques and their application to WBG converters is shown in Table 1. Overall, both air-and liquid-cooled WBG inverters are presented with different optimized configurations of cold plates or heat sinks, with the purpose of increasing the heat dissipation capability of the device. Air cooling is a cheaper and simpler solution compared to liquid cooling. Liquid cooling requires additional auxiliaries (pumps, radiator, compressor, and chiller), which increase the weight and volume of the system. Because of higher working temperatures and heat flux of WBG devices, air cooling can be applied to a wider power range of inverters compared to Si devices. As can be seen from recent studies, introducing WBG devices on the market transformed the trend from liquid to forced air cooling for low-power (up to 50 kW) [89] and low power density (up to 18 kW/L) [87] applications. In addition, some studies are devoted to volume hexagonal or cubic air-cooled inverters. Such construction increases the power density of the inverter; for example, the standard construction with a common flat and finned heat sink achieves 13 kW/L [18] and both of the presented examples of volume inverters achieve 17.2 and 18.6 kW/L [86,87]. Another approach of applying air cooling for the power inverter is a modular inverter structure. Modular air-cooled inverters from [87,88] are solutions with high power levels. Due to the use of separated modules, each module-inverter has a low power level and can be air cooled.
Nevertheless, liquid cooling is still the most widespread solution for power electronics applications in EVs, and indirect liquid cooling is still feasible. Single-sided indirect cooling is developing via improvement of the cold plate, as can be seen in [93,94]. Moreover, the single-sided direct and indirect cooling for WBG inverters has been optimized in two ways: by changing the distance between the chips and changing the number of power switches [96][97][98][99]. Because WBGs have a smaller chip size, additional power switches can be used without an increase in the volume of the device, to provide better current sharing in one inverter leg as more parallel chips are used. From the thermal optimization perspective, in contrast, the reduced number of chips enables a larger distance between them. When chips have a larger distance between them, applied cooling has better heat dissipation capability and temperature distribution. The main drawback of this optimization is the increase in the operating current for each power switch. However, in the case of WBG semiconductors, this problem can be compensated for by using higher rated currents of WBGs. The double-sided cooling approach shows good thermal properties; for example, reduced thermal resistance up to 35% [28]. Such cooling is suitable for high thermal performance and high temperature solutions. Although the heat dissipation capability of the cooling system is low, double-sided cooling does not justify the complexity of its implementation. In addition, power modules with double-sided cooling have higher reliability due to the replacement of the bonding wires by more reliable interconnectors. Furthermore, the double-sided power module can be designed for an optimized current path between parallel modules [76].

Conclusions
This paper provides a technical assessment and review analysis of the cooling technologies for automotive WBG inverters. The WBG semiconductors, classification of cooling techniques, and details of each cooling technique are reviewed in detail. These semiconductors have different intrinsic properties that can affect the thermal properties of the overall system. WBG technology is a highly promising solution for the next generation of power electronics and automotive powertrains due to their superior intrinsic characteristics, such as high thermal flux, available junction temperature, limit of the blocking voltage, and switching frequency, and low switching losses and conduction voltage drop.
The commercial availability of WBG devices enables the use of cheaper and simpler air-cooling approaches for low-power motor drives up to 50 kW, rather than liquid cooling. Two main approaches aimed at increasing the heat dissipation capability of air cooling can be highlighted: the volume and the modular design of power inverter. Whereas the former provides a compact inverter with a larger surface to dissipate heat, the latter modular design provides all of the required power of the device divided between separate modules, in which each module has lower power and can be cooled using the forced-air approach.
Liquid cooling remains the most cost-effective solution for a wide range of power levels. Numerous studies have been conducted on the different optimization approaches, which include new shapes of cold plates or heat sinks, the use of new materials, and the combination of two or more cooling techniques. Single-sided indirect liquid cooling remains highly popular and shows good thermal performance, particularly when new WBGs with better thermal properties are applied. Because WBGs have a smaller chip size, smaller losses, and a higher rated current and voltage, some studies suggest thermal optimization can be achieved by changing the number of semiconductors in one inverter leg and by changing the distance between the power switches. Direct liquid cooling has better thermal performance than indirect cooling, but adds complexity to the implementation. Such a cooling approach can be useful for high performance and high temperature solutions. Indirect and direct double-sided cooling are applicable for devices with a high heat transfer coefficient, because only in this case is their implementation complexity justified.
According to the trend of increasing rated power levels of power inverters, high heat flux systems, such as immersion cooling, may soon become widespread. Moreover, WBG power modules, in the case of high power levels of the device, need to be cooled with special cooling approaches, which are capable of effectively dissipating heat from a small hot area. A porous heat sink, heat pipes, and special thermal packaging may be a solution. In addition, high operating temperatures of WBGs also provide an opportunity for the use of IMDs with a shared motor and inverter cooling.
Due to the growing demand for increasing power levels in motor drives, cooling technology is an important part of the system. The results of this paper could be used for many future studies on the topic.