Review of Power Electronics Components at Cryogenic Temperatures

Abstract

In order to apply power electronics systems to applications such as superconducting systems under cryogenic temperatures, it is necessary to investigate the characteristics of different parts in the power electronics systems. This paper reviews the influence of cryogenic temperature on power semiconductor devices including Si and wide bandgap switches, integrated circuits, passive components, interconnection and dielectric materials, and some typical cryogenic converter systems. Also, the basic theories and principles are given to explain the trends for different aspects of cryogenically cooled converters. Based on the review, Si active power devices, bulk CMOS based integrated circuits, nanocrystalline and amorphous magnetic cores, NP0 ceramic and film capacitors, thin/metal film and wirewound resistors are the components suitable for cryogenic operation. Pb-rich PbSn solder or In solder, classic PCB material, most insulation papers and epoxy encapsulant are good interconnection and dielectric parts for cryogenic temperatures.

I. Introduction

In general, there are two main motivations to use cryogenically-cooled power electronics systems [1], [2]. First, some special applications like spacecraft based electronic systems or superconducting machines require cryogenic temperatures for their operation. In such applications, power electronics systems supply power to electric equipment or machines. Conventionally, the power electronics systems are placed inside some thermal insulation and the temperature is maintained at room temperature (~300 K) [1]-[4]. However, the extra thermal insulation and temperature regulation increase the complexity, weight, volume and cost. Thus, it would be beneficial if the power electronics systems can also operate at cryogenic temperatures.

Second, research has already shown that some semiconductor devices have improved performance at low temperatures like lower on-resistance and faster switching speed [1]-[3], [5]-[7], which means that making power electronics systems work at cryogenic temperatures can contribute to lower power dissipation, smaller volume and weight.

Generally speaking, the temperature can be called cryogenic when it is lower than 123 K. The low temperature environment can be created with the help of liquefied natural gas (111 K), liquid nitrogen (77 K) or liquid helium (4 K). Compared to power electronics systems at room temperature (~300 K), the requirements at such cryogenic temperatures are similar, which mainly includes high efficiency, high reliability and high power density. However, due to the significant temperature change, both the static and dynamic characteristics of the parts in power electronics systems can change significantly.

A typical power converter mainly consists of the following:

1. Power semiconductor devices, which include diodes and active switches such as power MOSFETs.

2. Integrated circuits, which include analog parts and circuits in the control or gate drives.

3. Passive components, which include transformers, inductors, capacitors and resistors in power or control circuits.

4. Interconnection and dielectric materials, which include solder, PCB and insulation materials providing electrical or mechanical support for the converter.

It is essential to understand the properties of all the aspects mentioned above at cryogenic temperatures so that one can select proper parts for the design.

Some converters have been tested at cryogenic temperatures, and some of the characteristics of the aforementioned parts are reported [1]-[4], [8]-[16]. Performance of power semiconductor devices are shown in [1]-[4], [10], [13], but they mainly focus on Si diodes and MOSFETs. Properties of integrated circuits are presented in [2]-[4], [8], [9], while the characteristics of some passive components are analyzed in [1]-[4], [11]. However, most of these papers are more like testing reports without much theoretical analysis, and none of the existing literature covers all of the four different types of components mentioned above.

This paper makes a comprehensive investigation of the existing literature on different parts in power electronics systems operating at cryogenic temperatures. Section II analyzes the characteristics of power semiconductor devices. Section III presents the behavior of integrated circuits. Section IV shows the features of passive components. Section V demonstrates the properties of interconnection and dielectric materials. Section VI provides a brief conclusion.

II. Power Semiconductor Devices

Generally, the characterization of semiconductor devices at cryogenic temperature follows the rules and procedures at room temperature except that the device under test is located in a cryogenic chamber or vacuum station. The applied gate drives and conditioning circuits are identical to the testing at room temperature, so that the result can be fairly compared with the room temperature testing.

A. Si P-N Diode

The Si P-N diode is one of the very first semiconductor devices developed for power circuits, and its performance at cryogenic temperatures has been investigated by several research groups [2], [3], [5]-[7], [17].

Fig. 1 plots the forward voltage of Si P-N diodes at different temperatures. The plotted value is normalized to that at room temperature. The forward voltage of a diode consists of the knee voltage and the voltage drop across the on-resistance. The knee voltage increases when the temperature decreases because of the drop in intrinsic carrier concentration at low temperature. Meanwhile, the on-resistance decreases when temperature drops from room temperature to around 100 K due to the increase of carrier mobility. But when temperature further decreases, the resistance increases because of the impact from carrier freeze-out [18]. Thus, the net effect can cause P-N diodes to exhibit different temperature coefficients depending on the current amplitude. Generally, within the current rating, the forward voltage shows negative temperature coefficient because the knee voltage is dominant as shown in Fig. 1. However, when the current is large enough, the on-resistance becomes dominant, and the forward voltage drop has positive temperature coefficient until the temperature is below 100 K.

Fig. 1.

Normalized forward voltage of Si P-N diodes.

Fig. 2 shows the normalized breakdown voltage of Si P-N diodes. The breakdown voltage declines because the mean free path of the carrier increases at low temperature, which contributes to higher impact ionization efficiency. So, more electron-hole pairs with high energy are created to launch impact ionization, and the avalanche is enhanced.

Fig. 2.

Normalized breakdown voltage of Si P-N diodes.

In terms of the switching performance of Si P-N diode at cryogenic temperatures, it is found that both the peak reverse current and the recovery time reduce due to the reduced carrier lifetime [6], [7].

B. Si and SiC Schottky Diode

Unlike the P-N diode, Si and SiC Schottky diodes do not have a reverse recovery issue, which makes them more suitable for high switching frequency circuits. The temperature dependent forward voltage drop of Si and SiC Schottky diodes is similar to that of a Si P-N diode [2], [5], [6]. However, the on-resistance of SiC Schottky diodes increases more rapidly than Si Schottky diodes when the temperature is lower than 100 K due to larger influence by carrier freeze-out in SiC devices [2].

The breakdown voltage of Si and SiC Schottky diodes is stable or increases slightly at lower temperatures because of the reduced space charge generation caused by lower intrinsic carrier concentration [2], [6].

C. Si MOSFET

Extensive testing and analysis have been conducted for Si power MOSFETs at cryogenic temperatures [6], [7], [19]-[27]. Fig. 3 demonstrates the normalized on-resistance of Si MOSFETs versus temperature. The on-resistance decreases significantly from room temperature to about 100 K due to the increased carrier mobility. However, the on-resistance increases as temperature drops below 100 K because of carrier freeze-out. For the same reason, the transconductance increases with decreasing temperature until around 100 K. The threshold voltage increases at low temperature because of the reduction of intrinsic carrier concentration. The body diode of the Si MOSFET is a Si P-N diode, so its behavior follows that of a Si P-N diode as analyzed above.

Fig. 3.

Normalized on-resistance of Si MOSFETs.

Fig. 4 gives the drain-source breakdown voltage of Si MOSFETs, and it reduces as the temperature drops. This is mainly due to the increase in the mean free path of carriers and higher impact ionization.

Fig. 4.

Normalized breakdown voltage of Si MOSFETs.

Due to the increased inversion layer mobility and its resultant higher transconductance, the switching speed can be faster and the switching loss can be reduced with lower temperature [26].

D. SiC MOSFET

Wide bandgap devices are getting more and more popular because of their high thermal conductivity, high carrier mobility and saturated electron velocity compared to conventional Si devices. Among them, the silicon carbide (SiC) MOSFET is a strong competitor with the Si IGBT in similar voltage and power applications. At room temperature, SiC MOSFETs provide faster switching speed, lower switching loss and higher operating temperature. However, the existing research shows that SiC MOSFETs have relatively poor performance under cryogenic temperatures [28]-[35].

Fig. 5 illustrates the change of threshold voltage of SiC MOSFETs at different temperatures. Compared to the result by the conventional temperature dependent equation to calculate threshold voltage, the tested result appears to increase much more dramatically with the decrease of temperature. This phenomenon is related to the number of occupied interface traps. When the temperature decreases, the number of occupied interface traps increases rapidly, which results in the higher threshold voltage [31].

Fig. 5.

Normalized threshold voltage of SiC MOSFETs.

Fig. 6 shows the on-resistance of SiC MOSFETs at different temperatures. It is explained in [31] that the total on-resistance R_total consists of two parts: the channel resistance R_ch and the residual resistance R_s. R_s includes the drift region, JFET, substrate and contact resistances, and is much larger than R_ch at high temperature. R_s drops with decreasing temperature due to higher carrier mobility similar to a Si MOSFET. However, R_ch has a negative coefficient and becomes dominant at low temperature, which makes the total on-resistance increase rapidly. This is also caused by the increase of the interface state density since large amounts of electrons are trapped and few free electrons are available for conduction of the inversion layer. In addition, it is suggested by [32] that carrier freeze-out is also a contributor for the increase of on-resistance.

Fig. 6.

Normalized on-resistance of SiC MOSFETs.

The drain-source breakdown voltage of a SiC MOSFET keeps relatively constant for a wide temperature range because the impact ionization efficiency does not increase [28]. It is reported that the switching speed of SiC MOSFETs does not improve or even gets worse at cryogenic temperatures [28], [30].

E. GaN HEMT

In addition to SiC MOSFETs, gallium nitride (GaN) high electron mobility transistor (HEMT) is another wide bandgap power device with appealing features at room temperature. Due to the different device structure compared to a MOSFET, GaN HEMT shows different characteristics under cryogenic temperatures [35]-[53].

Fig. 7 shows the on-resistance of GaN HEMTs at different temperatures. It is observed that the on-resistance keeps decreasing with temperature. The on-resistance of the GaN HEMT R_total mainly includes contact resistance of drain and source electrodes R_c, source to gate and gate to drain resistance R_SG and R_GD as well as the two-dimensional electron gas (2DEG) channel resistance R_ch. Among them, R_ch is the dominant part and is determined by the electron mobility in 2DEG [41]. Due to the weak influence of Coulomb scattering, the electron mobility keeps increasing until the temperature reduces to around 40 K. Moreover, there is no sign of carrier freeze-out in GaN HEMTs due to the characteristics of 2DEG at cryogenic temperatures. Therefore, the transconductance also increases with the decrease of temperature due to the increase of electron mobility.

Fig. 7.

Normalized on-resistance of GaN HEMTs.

The breakdown mechanism of GaN HEMTs is different from that of MOSFETs and is more complicated because of its lateral structure [45], [46], [52]. Several mechanisms can contribute to the breakdown of GaN HEMT: source to drain punch-through, leakage through gate Schottky junction, vertical leakage of substrate, and impact ionization between source and drain [46].

Not much research has been conducted in analyzing the breakdown of GaN HEMTs at cryogenic temperatures. However, the temperature dependent trend shows that vertical leakage of the substrate is the dominant factor that impacts the breakdown behavior. Fig. 8 in [52] shows the forward and reverse biased breakdown behavior of a GaN HEMT. It can be seen that although the leakage current during the increase of voltage changes with temperature, the breakdown voltage keeps almost constant from 423 K to room temperature, which indicates that it is not caused by impact ionization. The potential reason for the vertical breakdown is that, electrons are injected from the substrate and trapped in the buffer. With the increase of voltage bias, the trapped carriers are ionized and more free electrons are generated. The relatively constant breakdown voltage of GaN HEMTs at cryogenic temperatures is verified in [38].

Fig. 8.

Vertical breakdown behavior of GaN HEMT [48]. (a) Forwardbiased, (b) Reverse-biased.

The threshold voltage of normally off p-GaN HEMTs should have positive temperature coefficient at low temperatures because the increase in intrinsic carrier concentration and ionization in the p-GaN layer and GaN buffer. This matches with the trend of devices from GaN Systems and Panasonic [38], [50]. However, V_th of GaN FET from EPC shows negative coefficient as reported in [37], whose reason is not reported.

The current collapse phenomenon (kink effect) due to the trapped electrons in the surface states is a potential issue for GaN HEMTs. Several papers have analyzed the kink effect at cryogenic temperatures [39], [42], [44], [47], [49]. It is found that the kink effect is more severe when the temperature decreases since lower temperature enhances the surface traps.

However, it should be noted that the existing literature focuses on depletion mode GaN HEMTs with relatively small voltage and current rating instead of enhancement mode power GaN HEMTs. So it is still not clear what the current collapse issue of power GaN HEMTs is at low temperature.

F. Si IGBT

Though SiC MOSFETs are attracting more and more attention, Si insulated gate bipolar transistor (IGBT) is still widely used in medium to high power applications due to its low cost and high reliability and availability. IGBTs combine the advantages of MOSFETs and BJTs so they have lower voltage drop compared to MOSFETs and lower power for the gate drive compared to BJTs. Since the aforementioned review has shown that SiC MOSFETs are not good candidates for cryogenic application, it is of great importance if Si IGBTs can show superior performance at low temperature.

The main drawback of IGBTs at room temperature is the current tailing caused by the removal of excess carriers stored in the N- drift region. Consequently, the turn-off time of IGBTs is significantly increased, and the switching frequency of the converter is limited. The switching performance of IGBTs at cryogenic temperatures has been analyzed [5]-[7], [34], [54], [55]. As shown in Fig. 9, the normalized turn-off time of IGBTs is plotted for different temperatures. The tailing time decreases from 2.8 μs to 0.4 μs as temperature drops from 287 K to 50 K, which means that the switching speed and loss of IGBT at cryogenic temperatures are much improved compared to room temperature. The decrease of the tailing time with lower temperature is due to the reduction of IGBT’s inherent PNP transistor gain. Because the lifetime of minority carriers in IGBTs decreases at low temperature, the gain of inherent PNP transistor β reduces and leads to faster decay of collector current.

Fig. 9.

Normalized turn-off time of Si IGBTs.

The trend of static characteristics of IGBTs is similar to that of a Si MOSFET. When the temperature drops, the forward breakdown voltage decreases due to higher impact ionization efficiency; the threshold voltage increases due to the reduction of intrinsic carrier concentration; the transconductance increases and the forward voltage drop decreases due to the increase of carrier mobility.

G. SiGe Diode and HBT

The use of germanium (Ge) has been considered since the freeze-out temperature of dopants in Ge is much lower than in Si, and also the carrier mobility in Ge is higher than that in Si at low temperature [7].

Power semiconductor devices based on SiGe are developed by GPD Optoelectronics Corp in cooperation with Auburn University, motivated by the NASA deep space exploration program [56], [57]. A 50 V/ 5 A SiGe heterojunction bipolar transistor was designed and fabricated, which was examined along with a SiGe diode in a 100 W, 24 V/48 V, 100 kHz boost power converter at low temperature. The DC current gain does not drop and the switching performance improves at cryogenic temperatures, which makes the SiGe HBT a promising candidate for cryogenic power converters.

However, some contradictory results are reported by the University of Akron and NASA with the same SiGe HBT in [58]. The testing shows that the DC current gain drops significantly when the temperature is lower than 200 K. Because there is no detailed description of the testing setup, applied device structure, and measurement methods, it is not clear why this result occurs. Before SiGe power devices can be utilized in real converters, much more device development and testing are needed.

H. Summary

Fig. 10 and Fig. 11 compare the performance of diodes and active switches at cryogenic temperatures with their individual performance at room temperature. The specific performance at room temperature of each device is normalized to 1. A higher value at cryogenic temperatures means higher voltage, higher resistance, or longer switching time. Note that the value of the point indicates the relative increase or decrease in the characteristic of one device at cryogenic temperatures compared to its own characteristic at room temperature. For instance, the Si IGBT has larger switching time improvement at cryogenic temperatures than Si MOSFET, so the IGBT has a lower point. But it does not mean that at cryogenic temperatures, Si IGBT can switch faster than Si MOSFET. With the plotted comparison, it can be concluded that Si diodes, Si MOSFETs, GaN HEMTs and Si IGBTs can be adopted in applications requiring lower loss, while SiC diodes and MOSFETs have poor conduction performance, but are more suitable for the cases requiring stable drain-source breakdown voltage.

Fig. 10.

Comparison of diodes’ performance at cryogenic temperatures with their individual performance at room temperature.

Fig. 11.

Comparison of switches’ performance at cryogenic temperatures with their individual performance at room temperature.

III. Integrated Circuits

A. Si BJT Based

Several papers have shown that Si BJTs perform poorly under low temperature [59]-[63], which means that integrated circuits based on BJT are not suitable for cryogenic temperature operation.

Fig. 12 in [60] illustrates the DC current gain β of a typical NPN BJT. It is clear that β drops significantly with the decrease of temperature. This is mainly because the bandgap in emitter region drops at low temperature, which results in severe decrease of emitter injection efficiency. Another contributor is the reduction of base transport factor caused by the reduction of carrier lifetime.

Fig. 12.

DC current gain of BJT [60].

For the other characteristics, the base-collector breakdown voltage decreases with decreasing temperature due to the typical P-N diode behavior. The collector-emitter breakdown voltage increases due to the reduced current gain. The switching speed increases because of the increase of diffusion coefficient and decrease of carrier lifetime. However, the advantage is not enough to compensate the decreased current gain.

B. Bulk CMOS Based

Previous discussion indicates that power MOSFETs have improved performance at low temperature. For bulk CMOS based integrated circuit, the main improvement is the switching speed [64]-[67]. Due to the increase of carrier mobility and saturation velocity at low temperature, the transconductance of both n-channel and p-channel MOSFETs increases. The unity gain frequency of bulk CMOS increases as temperature drops [68]. Therefore, bulk CMOS based integrated circuits are more suitable for high frequency operation at cryogenic temperatures.

The absolute value of the threshold voltage of both n-channel and p-channel MOSFETs increase as temperature decreases because of the increased surface potential caused by decreased intrinsic carrier concentration.

Despite the benefit of faster switching speed brought by low temperature, reliability issues such as gate degradation caused by hot carriers should be paid special attention to [66]-[70]. With increased carrier mobility and mean free path, the probability for carriers to gain enough energy to enter and get trapped in the gate oxide increases. Therefore, interface states are easier to be formed at low temperature, which is the source of transconductance degradation and threshold voltage shift. However, this phenomenon does not damage the device directly, but it limits the long-term reliability and lifetime. Fig. 13 from [69] shows the transconductance degradation of a CMOS analog circuit at different temperatures after the same operating time. The degradation increases when the temperature drops. The potential solution to mitigate the impact of hot carrier effect is to reduce the gate voltage at cryogenic temperatures so that the electric field is decreased, and the possibility for carriers to gain enough energy to flow into the gate oxide is reduced.

Fig. 13.

Transconductance degradation of CMOS analog circuit [69].

C. SOI CMOS Based

Silicon on insulator (SOI) technology can provide faster speed, lower power dissipation, and smaller package compared to conventional CMOS-based integrated circuits. Similar to bulk CMOS technology, the speed of SOI device improves at low temperature. However, SOI devices suffer from kink effect at low temperature especially for partially depleted SOI (PDSOI) [68], [71]-[73]. This phenomenon is one of the main floating body effects. Due to the higher impact ionization efficiency at low temperature, carriers are generated but they are not able to flow through the silicon substrate and part of them are trapped to form a forward bias of the body region, which reduces the threshold voltage and consequently increases the channel current. The kink effect decreases the voltage gain, increases loss, induces low frequency noise, and reduces lifetime of the device. This issue can be suppressed by using fully depleted SOI (FD-SOI).

D. SiGe HBT Based

As mentioned above, SiGe HBT devices are expected to have better performance at cryogenic temperatures. Compared to power devices, there are more reports of SiGe technology in analog circuits [59], [74]-[78]. It has been shown that SiGe HBTs show significant increase in current gain with the decrease of temperature.

E. Summary

It is concluded from the above review that bulk CMOS technology is promising for low temperature operation due to the higher switching speed capability, while the Si BJT is not suitable at cryogenic temperatures because of the significant degradation of current gain. Though SOI CMOS based and SiGe HBT based technology also show some benefit at low temperature, they are not likely to be used in power converters due to the limited availability and cost issues since bulk CMOS operates acceptably.

IV. Passive Components

A. Inductors and Transformers

An inductor or transformer in a power converter mainly includes two parts: core and winding. Research has been conducted for different kinds of high frequency core material including powder, ferrite, nanocrystalline and amorphous [79][91]. The comparison is given in Table I. Among them, ferrite core shows the worst performance, which has significant decrease in permeability and increase in loss, making it not suitable for cryogenic applications. Different materials in powder cores show different characteristics. The permeability of Molypermalloy (MP) and High Flux core keeps constant while Kool Mμ loses 40% of its permeability at 77 K. The loss of High Flux and Kool Mμ cores is relatively stable at low temperature while the loss of Molypermalloy core increases by 40%. The amorphous and nanocrystalline cores have similar or even higher permeability and saturation flux density at cryogenic temperatures though the magnetic loss increases.

TABLE I.

Comparison of different core materials at cryogenic temperatures with room temperature

Material	Powder			Ferrite	Nano-crystalline	Amorphous
	Molypermalloy (MP)	High Flux	Kool Mμ
Permeability	--	--	↓	↓↓	↑	--
Loss	↑	--	--	↑↑	↑	↑
Saturation flux density	N/A	N/A	N/A	Not clear	↑	↑

↑↑: increase significantly ↑: increase slightly --: keep constant ↓: decrease slightly ↓↓: decrease significantly

For the winding loss calculation, the anomalous skin effect is worth paying attention to [92]-[97]. According to the classical skin effect theory, the skin depth δ and the resistance at skin depth R_s is expressed as

$$\begin{gathered} \delta =\sqrt{\frac{2}{{\mu }_0\omega \sigma }} \\ {R}_s=\sqrt{\frac{{\mu }_0\omega }{2\sigma }} \end{gathered}$$ (1)

where μ₀ is the permeability of free space, ω is the angular frequency of current, and σ is the DC conductivity.

It can be seen that the surface resistance is inversely proportional to the square root of the DC conductivity σ. Since σ increases significantly at cryogenic temperatures, the resistance is expected to decrease. The classical theory is based on the current density equation

$$\overrightarrow{J}=\sigma \overrightarrow{E}$$ (2)

where J is the current density and E is the electric field. However, this equation is only valid when the skin depth is much longer than the mean free path of the electrons, which means that an electron does not experience electric field change before it collides. Meanwhile, at low temperature and high frequency applications, the mean free path of the electrons becomes even higher than the skin depth. In such case, the classical skin effect theory is not valid and the anomalous skin effect occurs.

With anomalous skin effect, the relationship between surface resistance and angular frequency of current is derived from [95]

$$R_s={\left(\frac{l}{\sigma }\right)}^{\frac{2}{3}}{\left(\frac{\sqrt{3}{{\mu }_0}^2}{16\pi }\right)}^{\frac{1}{3}}{\omega }^{\frac{2}{3}}$$ (3)

where l/σ is a constant depending on material. For copper, it equals to 6.8×10⁻¹⁶ Ω·m². Thus, the surface resistance is independent of DC conductivity with the impact of anomalous skin effect. Fig. 14 shows the calculated surface resistance of the copper with residual resistivity ratio (RRR) equal to 2000 based on classical and anomalous skin effect at 100 kHz, 1 MHz and 10 MHz respectively. It is observed that at low temperature where the mean free path of the electron is larger than skin depth, the anomalous skin effect theory provides higher resistance than classical theory. With higher frequency, the distance between the result of anomalous theory and classical theory gets larger. Therefore, designers cannot depend only on classical skin effect theory to calculate winding loss since the anomalous skin effect can increase the AC winding resistance at low temperature and high frequency.

Fig. 14.

Surface resistance with classical and anomalous skin effect of RRR= 2000 copper at different frequency.

B. Capacitors

Capacitors made from different materials operating at cryogenic temperatures have been investigated by several papers [2]-[4], [98]-[103]. Due to the difference in dielectric constant [98], capacitors can show different temperature dependent characteristics in capacitance and dissipation factor. Table II shows the change of capacitance and dissipation factors with temperature. It is concluded that NP0, polypropylene, polyphenylene sulfide (PPS), and mica perform well at cryogenic temperatures. For applications requiring high capacitance value, tantalum is preferred although its dissipation factor significantly increases while an electrolytic capacitor loses most of its capacitance at low temperature.

TABLE II.

Comparison of different capacitors at cryogenic temperatures with room temperature

Material	Ceramic				Film				Mica	Electrolytic	Tantalum
	X7R	Y5V	Z5U	NP0	Polypropy lene	Polyphenylene sulfide	Polyest er	Polycarbo nate
Capacitance	↓↓	↓↓	↓↓	--	--	--	↓	↓	--	↓↓	↓
Dissipation factor	↑↑	↑↑	↑↑	--	↓	--	↓	↓	--	↓↓	↑↑

↑↑: increase significantly ↑: increase slightly --: keep constant ↓: decrease slightly ↓↓: decrease significantly

C. Resistors

Resistors with different materials are compared at cryogenic temperatures in [2], [3], and the results are shown in Table III. Thin film, metal film and wirewound are good candidates for resistors while carbon and ceramic composition are poor at cryogenic temperatures.

TABLE III.

Comparison of different resistors at cryogenic temperatures with room temperature

Material	Thin film	Thick film	Metal film	Metal Oxide	Power film	Wirewound	Carbon Composition	Ceramic Composition
Resistance	--	↑	--	↑	↑	--	↑↑	↑↑

↑↑: increase significantly ↑: increase slightly --: keep constant

V. Interconnection and Dielectric Materials

A. Solder

Solders provide both mechanical and electrical connection between dies, packages, and PCBs. Solder alloys are subject to cyclic stress resulting from mechanical or thermal cycling. The thermal expansion mismatch between the die or package and the substrate or PCB can cause the solder joints to deform and fatigue when cycled from room temperature to cryogenic temperatures.

Soft solder alloys are most widely used in power electronics applications since they have the advantage of lower melting temperature during assembly. However, they increase in strength and decrease in ductility with decreased temperature [104]-[107]. The most common soft solder is PbSn alloy, and it is able to operate at cryogenic temperatures and maintains its ductility if it has high Pb content. However, as Sn content increases up to more than 40%, the alloy becomes brittle at cryogenic temperatures. PbSn alloy containing Sb can mitigate this issue, but it is still not recommended to use Sn-rich solder for low temperature.

Compared to standard PbSn alloy, pure indium or indium alloys have shown to be much better at cryogenic temperatures due to the greater ductility and longer lifetime [105], [107][110]. So indium solder is the best candidate in soft solders if not considering the availability and cost.

In addition to soft solders, hard solders which are also called brazes, like AuSn, AuGe and AuSi alloys, have higher melting temperature and do not undergo stress but transmit more stress to die or package. Therefore, the device is more likely to crack during thermal cycling. However, it is found that AuSn can work well at cryogenic temperatures [111].

B. Printed Circuit Boards

PCBs provide mechanical support, electrical connection as well as dielectric material in the circuit. They are usually made from glass fiber reinforced epoxy and can fatigue during thermal cycling due to thermal expansion mismatch. Based on the existing research [112]-[114], the fatigue occurs after more than 10,000 cycles at the pressure of 266 MPa, which is far beyond the normal operation environment. So the mechanical property of PCB material is strong enough to work for cryogenic applications.

The other concern is about the dielectric performance. From [115], [116], the flashover breakdown voltage in liquid nitrogen is higher than that in transformer oil at room temperature. In conclusion, the mechanical and electrical performance of PCBs do not decline significantly at cryogenic temperatures.

C. Dielectric Paper

Table IV shows the breakdown electric field of some commonly used dielectric paper materials at cryogenic temperatures [116]-[121]. In general, most dielectric papers have improved dielectric performance at low temperatures, and they can be implemented in cryogenic applications.

TABLE IV.

Breakdown electric field of dielectric materials at cryogenic temperatures

Name	Material	Eb (kV/mm)	Temp. (K)
Kapton	Polyimide	230–360	12
Teflon	Polytetrafluoroethylene	230–310	12
Mylar	Polyester	190–300	12
Nomex	Polyamide	65–90	12
Copaco rag	--	45–60	90
Kraft paper	--	20–40	77

D. Encapsulants

Encapsulants are special dielectric materials that are widely used in power device packaging and busbar fabrication. The most common material for typical encapsulant is silicone gel based and epoxy based. However, silicone gel cannot survive at cryogenic temperatures [122]. It is found that partial discharge occurs when the temperature is lower than 215 K and can significantly reduce the breakdown voltage of the silicone gel. Moreover, the change of the breakdown voltage cannot be recovered even once the temperature returns to room temperature. Therefore, most commercially available power modules utilizing silicone gel based encapsulants cannot be used for cryogenic applications. On the other hand, many epoxies can work at cryogenic temperatures [123]. The only issue is that epoxies tend to be more brittle as temperature drops. Thus, epoxies with lower moduli should be selected for operation at cryogenic temperatures.

VI. CRYOGENIC CONVERTER SYSTEMS

Not many converters developed for cryogenic operation have been reported [2], [3], [11]-[16]. In general, the converters are tested inside a cryogenic chamber or box. The topologies, power rating, applied devices, and the switching frequencies are summarized in Table V.

TABLE V.

Converters in existing literature

Ref	Topology	Power (W)	Device	fs(kHz)
[2]	Full bridge	20	Si MOSFET	20
[3]	Full bridge	24	Si MOSFET	N/A
[11]	Switched capacitor	1 k	GaN HEMT	120
[12]	Half bridge	320	Si MOSFET	10
[13]	Full bridge	2.5k	Si MOSFET	20
[14]	Buck	175	Si MOSFET	50
[15]	Buck	55	Si MOSFET	200
[6]	Full bridge	8	Si MOSFET	50

The purpose of building these converters is mainly to verify the function and feasibility of the concept to run power converters at cryogenic temperatures, so the power ratings are normally low. Bridge type topologies are popular because most of them are for motor drives. Considering the device performance and availability, Si MOSFETs are mostly used, and the switching frequency is usually lower than 100 kHz. There is no reference about using Si IGBTs or SiC MOSFETs to build and run converters for high power applications at cryogenic temperatures.

It should be noted that, this review paper mainly focuses on the properties of individual parts in power converter systems. However, interactions such as the coefficient of thermal expansion (CTE) among the materials, components and systems are also essential and require special attention and calculation when developing cryogenic power converters.

VII. CONCLUSIONS

This paper reviews the behavior of different parts in power electronics systems operating at cryogenic temperatures, which include power semiconductor devices, integrated circuits, passive components, interconnection and dielectric materials. The characteristics, basic theory, and comparison are given for each part.

1. For power semiconductors, Si P-N diode shows improved on-resistance and switching performance at cryogenic temperatures, while the knee voltage and breakdown voltage decrease. Si Schottky diode has improved on-resistance but knee voltage increases, while a SiC Schottky diode has larger on-resistance at low temperature. Si MOSFET has smaller on-resistance and faster switching speed; however, the breakdown voltage decreases. SiC MOSFET shows poor performance at low temperature. GaN HEMT has improved on-resistance, switching speed and stable breakdown voltage. Si IGBT shows improved forward voltage drop and switching speed but lower breakdown voltage.

2. For integrated circuits, Si BJT has poor performance while bulk CMOS, SOI CMOS and SiGe HBT work well at low temperatures. Considering the requirement of normal power electronics systems, bulk CMOS is the best candidate considering availability and cost.

3. In terms of passive components, ferrite core has poor performance at cryogenic temperatures while powder, amorphous and nanocrystalline cores can work at low temperature although the loss increases to some extent. Anomalous skin effect should be taken into consideration when calculating the winding loss. NP0 is the most suitable material for ceramic capacitors, while most film capacitors can work well at cryogenic temperatures. Both electrolytic and tantalum capacitors have worse performance. Metal film, thin film, and wirewound are candidates for resistors.

4. For interconnection and dielectric materials, Pb-rich PbSn or In alloys are preferred solders. Classical PCB is good enough for both mechanical and dielectric support. Most dielectric paper materials can work at cryogenic temperatures. Epoxy based encapsulant is preferred. However, commercially available power modules with silicone gel encapsulant fail at cryogenic temperatures.

5. Existing cryogenic converter systems are mostly designed for low power and low voltage applications, and based on Si MOSFETs. The switching speed of the power devices is normally less than 100 kHz.

From the review, cryogenic temperature can cause significant change in properties of different materials, which results in the variation of the converter efficiency and reliability compared with room temperature case. Thus, designers need to carefully select materials for cryogenic applications, and this paper can serve as initial guidance for designers to follow in building power electronics systems.