Abstract

Power densification and rising module heat losses cannot be managed by traditional “external-to-case” cooling solutions. This is especially pronounced in high voltage systems, where intervening layers of insulating material between the power devices and cooling solution need to be sufficiently thick to provide adequate voltage isolation. As operating voltages increase, the required thicknesses for these insulating layers become so large that they limit the ability to extract the heat. A direct cooling approach that addresses voltage separation issues represents a unique opportunity to deliver coolant to the hottest regions, while opening up the opportunity for increased scaling of power electronics modules. However technical concerns about long-term performance of coolants and their voltage isolation characteristics coupled with integration challenges impede adoption. Here, the reliability and performance of a dielectric fluid of the hydrofluoroether type, HFE7500, are examined to advance the feasibility of a direct cooling approach for improved thermal management of high-voltage, high-power module. The breakdown voltage of the dielectric fluid is characterized through relevant temperatures, flow rates, and electric fields with the ultimate goal of developing design rules for direct integrated cooling schemes.

1 Introduction

High-power electronic systems will be needed to drive the increasing electrification of mobility as well as the shift from hydraulic and pneumatic actuation to electric power. One way to enable more power is through the development of modules that operate at higher voltages, which has been made possible by the emergence of wide band gap (WBG) semiconductor devices with voltage ratings exceeding 1.2 kV. Additionally, these WBG devices exhibit high-temperature tolerance and low drain-source on resistance (RDSon) capabilities, enabling high amperage power modules that further increase power density; however, packaging and cooling techniques are struggling to keep up. To meet the demands of increased power density, new advancements in packaging materials [1], packaging architectures [2], and thermal management have emerged.

Novel cooling solutions that have been explored include those aimed at increasing the heat transfer area through microchannel flow passages [3], single phase immersion cooling in dielectric fluids to unlock higher densities, particularly in computer servers [4,5], novel fluid delivery methods via sprays [6] or jets [7] for enhanced heat transfer rates and cooling uniformity [8] as well as two-phase flow [9] to exploit the latent heat of boiling. A common theme for many of these efforts is the minimization of the convective thermal resistance (Rconv), which still leaves a large component of the overall thermal resistance, specifically junction-case resistance (Rjc), largely untouched. Minimizing Rjc, which is made up of the thermal resistances of all layers between the power device and heat sink, has been the subject of numerous research activities including those seeking to replace traditional ceramic-copper substrates with more conductive alternatives [10]. While these approaches deliver some improvement in thermal conductivity, the most effective way to minimize Rjc is to bring the coolant as close to the power devices as possible.

Integrated or direct cooling of power dense electronics has attracted significant research activity including efforts by Kwon et al. [11], Birbarah et al. [12] and Jung et al. [13]. While these direct cooling efforts have demonstrated the ability to improve the thermal management of power dense electronics, the research efforts have largely focused on low voltage applications. In high voltage packages, there are implementation challenges pertaining to the risk of phenomena such as partial discharge and breakdown, which complicate implementation.

High voltage systems face a tradeoff between electric field (E-field) mitigation and thermal management. On the one hand, thick or stacked ceramic substrates [14,15] with high dielectric properties are required between the power devices and heat sink to reduce the risk of partial discharge and breakdown due to the high E-field across the system. On the other hand, this isolates the power devices from the sink, thereby limiting the ability to extract heat. Existing research has considered solutions such as double-sided cooling or high-performance cooling schemes including jet impingement [14] onto the copper metallization of the thick DBC substrates. However, such solutions are only effective up to a point, after which further increases in the heat transfer coefficient have little to no effect, due to increased (Rjc). Direct cooling with dielectric fluids represents an opportunity to meet the cooling demands in such high-V power electronics. In a recent effort, researchers at the Army Research Laboratory demonstrated direct cooled packages, including a stacked diode module [16] that was demonstrated to operate at >20 kV while being directly cooled by forced convection of a dielectric fluid (DF) of the hydrofluoroether type. However, dielectric breakdown (BD) tests on the package showed a lower breakdown voltage (VBD) for the cooled stack compared to the uncooled stacked module, which the authors attributed to failure through the fluid or the package. This result highlights the fact that while direct cooling with DF's is a viable approach for thermal management of high voltage packages, concerns over long-term performance of the coolants and their voltage isolation characteristics remain a barrier to adoption.

Direct forced convective cooling, in particular, exposes the fluid to changing conditions in terms of flow and temperature, which have been shown to significantly increase or decrease the voltage shielding capacity of DF's. Breakdown data from several reviewed studies [1720] are plotted in Fig. 1 as a function of flow velocity, indicating a clear impact of enforced flow. To crystallize the effects of flow and facilitate comparison, the data from each study is plotted as VBD/VBD,0 where VBD is the average breakdown voltage at a given flow velocity and VBD,0 is the respective average breakdown voltage under static flow conditions. In these works, the augmentation of breakdown strength with flow is attributed to the sweeping action of flow on charge carriers and impurities, whereas the reverse impact of flow is attributed to, among other factors, the onset of turbulence that provides low pressure regions that act as nucleation sites for breakdown and triboelectric charging of the fluid, which results in local E-field enhancement. The observed differences in the data from the reviewed studies is due to several factors including varying flow conditions, flow geometries, electrode configurations and other factors all of which can influence breakdown strength. However, while this information is available for hydrocarbons of paraffinic and isoparaffinic type, of which mineral oils for transformer cooling applications are notable examples, there remains a lack of information on fluorinated compounds, including perfluorocarbons and hydrofluoroethers such as HFE7500. These fluids, which possess low viscosity over a broad temperature range, are well suited to electronics cooling applications when it comes to pumping power requirements.

Fig. 1
Literature review of breakdown voltage (VBD) as a function of flow. For each study, the average VBD at each flow velocity is normalized with respect to the corresponding breakdown voltage under static flow conditions (VBD,0).
Fig. 1
Literature review of breakdown voltage (VBD) as a function of flow. For each study, the average VBD at each flow velocity is normalized with respect to the corresponding breakdown voltage under static flow conditions (VBD,0).
Close modal

In direct cooling applications where the fluid is subject to forced convection, a better understanding of the effects of flow conditions needs to be built to ensure safe implementation of these types of DF's in high voltage, high power electronics cooling. This effort seeks to advance the feasibility of direct cooling by assessing the voltage blocking characteristics of fluorinated dielectric fluid through relevant temperatures, flow rates, and electric fields.

2 Procedure and Testing

2.1 Dielectric Fluid Selection.

The breakdown characterization efforts focused on HFE7500, a hydrofluoroether type DF developed by 3 M. HFE7500 is nontoxic and nonflammable, colorless and exhibits good thermal and chemical stability while being compatible with most common metals and hard plastics [21]. Additionally, HFE7500 is environmentally friendly with no-ozone depleting potential as well as little global warming potential. In Table 1, the thermophysical properties of HFE7500, as obtained from the manufacturer's data sheet, are presented [22]. Efforts were made to verify some of these properties using Differential Scanning Calorimetry for thermal properties and Rheometry for viscosity. As shown in Table 1 and corroborated by rheometry, HFE7500 exhibits low viscosity even at low temperatures, reducing pumping requirements and making it well suited to power electronics cooling.

Table 1

Thermophysical and electrical properties of HFE7500 at 25 °C and 1 atm per the manufacturer's datasheet

Chemical name2-Trifluoromethyl-3-ethoxydodeca-fluorohexane
Chemical formulaC9H5F15O
Electrical resistivity2.2 × 108 Ω cm
Relative permittivity, εr5.8
Boiling point128 °C
Freezing point−100 °C
Density, ρ1614 kg/m3
Kinematic viscosity, ν7.7 × 10−7 m2/s
Dynamic viscosity, μ0.0012 (0.0021) Pa·s
Specific heat, Cp1128 (1138) J/KgK
Thermal conductivity, k0.065 W/mK
Chemical name2-Trifluoromethyl-3-ethoxydodeca-fluorohexane
Chemical formulaC9H5F15O
Electrical resistivity2.2 × 108 Ω cm
Relative permittivity, εr5.8
Boiling point128 °C
Freezing point−100 °C
Density, ρ1614 kg/m3
Kinematic viscosity, ν7.7 × 10−7 m2/s
Dynamic viscosity, μ0.0012 (0.0021) Pa·s
Specific heat, Cp1128 (1138) J/KgK
Thermal conductivity, k0.065 W/mK

Values in parentheses represent measured data.

2.2 Dielectric Fluid Breakdown Characterization.

A setup that enables DF testing through relevant temperatures, flow, and E-field was assembled. As shown in Figs. 2 and 3, the setup consists of a flow loop that features a pump ① (INTG7-060) for flowrate control and an inline heater ② for temperature regulation. The AC heater is rated for 120 V 400 W and is used in tandem with a PID controller to maintain fluid temperature at a desired value. Feedback to the PID controller is provided by a thermocouple ③ placed upstream of the test section. To apply a field to the fluid inside the test section ④, a Matsasuda Rb60 power supply   Ⓐ rated for 30 kV, 2 mA is used. In the current permutation of the apparatus, the voltage from the power supply is applied to a pair of point-point electrodes  Ⓑ , whose spacing is precisely controlled to vary E-field intensity. The point-point electrodes were chosen to minimize flow perturbation and capture the effects of E-field concentrations analogous to those at sharp corners, which represents worst-case E-field concentrations within a high voltage package [23]. Electrode spacing is varied relative to a reference point, where the electrodes are brought into electrical contact as verified with a multimeter's continuity test. The desired electrode spacing (d) can then be achieved by offsetting each electrode by a distance equal to d/2 relative to the reference point. For precise positioning, each electrode is attached to a micrometer actuator  Ⓔ , which is used to specify the offset to within a resolution of 10 μm.

Fig. 2
Schematic of the experimental setup for dielectric fluid breakdown characterization
Fig. 2
Schematic of the experimental setup for dielectric fluid breakdown characterization
Close modal
Fig. 3
Illustration of the test section and the circuitry used to control the high voltage power supply and monitor leakage current between the electrodes
Fig. 3
Illustration of the test section and the circuitry used to control the high voltage power supply and monitor leakage current between the electrodes
Close modal

Voltage application to the electrodes   Ⓑ   is controlled via a combination of a labview program Ⓒ and an NI USB-6001 DAQ system   Ⓓ  . The two are used in tandem to generate an analog signal (Vcon-in) to the RB60 supply  Ⓐ   . This external control voltage is varied from 0 to +5 Vdc, with the upper bound corresponding to a 30 kV output. During testing, Vcon-in is ramped such that the voltage applied to the electrodes increases at a rate of 100 Vdc/s. Throughout operation, the RB60 supply   Ⓐ outputs analog signals Vmoni and Imoni, which make it possible to monitor the supplied voltage as well as the leakage current flowing between the electrodes, respectively. The Imoni signal ranges from 0 to 5 Vdc, with the upper limit corresponding to a leakage current of 2 mA. To stop each test, a 1 mA leakage current limit is prescribed in the labview Ⓒ code.

Additional equipment such as a thermistor ⑤ (SEN-AP008B0), pressure sensor ⑥, flow sensor ⑦ (FTB2002-C) as well as an inline coolant filter ⑧ (INS-FLTR03) for keeping impurities from the test cell are also incorporated into the flow loop. The data from the sensors ⑤⑥⑦ is recorded separately using an Arduino Mega 2560 board. The test cell where the fluids are subjected to high E-field is additively manufactured using Peopoly Moai Hi-Temp Nex Resin that handles up to 180 °C. This makes it possible to drive the fluid to high temperatures to account for the growing push for elevated coolant inlet temperatures that reduce power and size demands on the heat exchanger/condenser side.

3 Results

During testing, the breakdown voltage of the dielectric fluid was determined by monitoring the increase in the leakage current between the electrodes with increasing voltage. As indicated in Fig. 4, the breakdown voltage (VBD) of the fluid is then identified as the voltage at which a sharp rise is observed in the measured leakage current due the development of a conductive path through the fluid. Based on this principle, the breakdown voltage of HFE7500 was tested under varying flow velocities and fluid temperature.

Fig. 4
Leakage current as a function of the applied voltage for HFE7500 under static flow conditions and with an electrode gap of 1 mm
Fig. 4
Leakage current as a function of the applied voltage for HFE7500 under static flow conditions and with an electrode gap of 1 mm
Close modal

As the fluid passes through the test section, the flow regimes are not only controlled by the internal channel geometry (cross-field flow) but also by flow over the cylindrical electrodes. Since, as illustrated in Fig. 5, the flow field has both geometries potentially participating in the local fluidic phenomena, the Reynolds numbers associated with each flow type are computed to crystallize any impact on breakdown mechanisms. In the computation of these Reynolds numbers, ρ is the density of HFE7500, V is the flow velocity and μ is the dynamic viscosity. For flow over the electrodes, the characteristic length is the needle diameter (D) whereas for channel flow, this corresponds to the hydraulic diameter Dh, which is determined using the cross-sectional area A and the perimeter P of the channel.

Fig. 5
The flow regimes governing flow through the test section along with equations for calculating the corresponding Reynolds numbers
Fig. 5
The flow regimes governing flow through the test section along with equations for calculating the corresponding Reynolds numbers
Close modal

To illustrate any dependency of breakdown strength on the flow regimes, the dielectric strength of HFE7500 is plotted as a function of Rechannel and Recyl in Fig. 6. For these tests, temperature and electrode gap were fixed at room temperature and 1 mm, respectively. Furthermore, each data point in the figure represents 10 breakdown measurements of HFE7500. Results indicated that cross-field flow can deteriorate the breakdown voltage, while higher flowrates in the range of Rechannel > 15,000 can result in an enhancement of the breakdown voltage. Figure 6 also shows that turbulent conditions overlap for cross-field flow and for flow over the electrodes. For channel flow, the onset of turbulence is typically taken to occur at 2300 < Rechannel< 4000. At the same time, for flow over the electrodes, vortex shedding, which is brought about by the detachment of the boundary layer as the fluid passes over the electrodes, starts at Recyl 90 and transitions to turbulence at 150 < Recyl< 300 [24]. In this Recyl range, flow in the wake of the electrodes is characterized by random vortex formation and, similarly to channel flow, low pressure regions that can act as nucleation sites for breakdown. The fluctuation of breakdown strength observed within these transitional ranges for Recyl and Rechannel can be attributed to turbulent conditions, which influence local pressure and in turn bubble nucleation and breakdown. Temperature-dependent measurements (see Fig. 7) indicated a slight decrease in breakdown strength with increasing temperature for Rechannel< 17,000, while no discernible trend can be observed with Rechannel> 17,000.

Fig. 6
Breakdown (BD) field as a function of enforced flow velocity for HFE7500, showing that turbulent conditions overlap for flow over a cylinder and for cross-field flow leading to large fluctuations in BD strength
Fig. 6
Breakdown (BD) field as a function of enforced flow velocity for HFE7500, showing that turbulent conditions overlap for flow over a cylinder and for cross-field flow leading to large fluctuations in BD strength
Close modal
Fig. 7
Breakdown (BD) field as a function of HFE7500 temperature at 1 mm gap spacing, indicating minimal impact of the variable
Fig. 7
Breakdown (BD) field as a function of HFE7500 temperature at 1 mm gap spacing, indicating minimal impact of the variable
Close modal
The E-field experienced by the DF is dependent on both electrode geometry and electrode gap. While initial testing used a 1 mm electrode gap, further testing also considered varied E-fields strength to establish a relationship between electrode gap and VBD. This would inform the layout of direct cooled packages including the layout of devices, traces, interconnects, and terminals. Results, which are presented in Fig. 8, were fitted to a power law model that is typically used to describe the relationship between breakdown voltage of liquid dielectrics and electrode gap spacing
VBD=Adn
(1)

where d denotes the gap length in mm and A and n are constants. Table 2 summarizes the values of A and n associated with the experimental data. Predictions made using these parameters are plotted in Fig. 8 along with 95% confidence level (CL) intervals for the fitted line. The model, which describes the experimental data well particularly at high Rechannel values, can be used to make decisions on the layout of direct cooled packages to minimize the risk of a discharge event.

Fig. 8
Model for describing the breakdown voltage dependency on gap spacing for different flow regimes and at room temperature, with results showing higher divergence in the laminar and transitional regimes
Fig. 8
Model for describing the breakdown voltage dependency on gap spacing for different flow regimes and at room temperature, with results showing higher divergence in the laminar and transitional regimes
Close modal
Table 2

Constants for the power law model for describing breakdown voltage as a function of gap length

An
RechannelEstimateStd ErrorEstimateStd Error
017.21.020.470.031
59014.71.030.540.045
295714.61.040.60.059
17,74015.61.020.660.034
An
RechannelEstimateStd ErrorEstimateStd Error
017.21.020.470.031
59014.71.030.540.045
295714.61.040.60.059
17,74015.61.020.660.034

4 Discussion

The voltage blocking performance of dielectric liquids has drawn significant research attention, but their breakdown mechanisms remain to be fully understood, nonetheless. Broadly speaking, two predominant theories of pure liquid breakdown have emerged in literature: electronic breakdown theory and bubble/cavitation initiated breakdown [25]. The former involves the generation of free charge carriers at the liquid-electrode interface via field ionization or field emission. The continuous prebreakdown conduction currents observed in Fig. 4 are partly attributed to these mechanisms [26]. In the field ionization process, neutral molecules adjacent to a positive electrode (anode) are ionized by valence electrons tunneling from the liquid into the metal surface, and thus injecting positive ions in the fluid. This process is governed by the difference between the ionization energy of the liquid (I) and the work function of the electrode surface (φ), i.e., the energy required to free an electron from the surface of the electrode. Field emission on the other hand takes the reverse path with electrons being emitted from the tip or surface irregularities of a negative electrode (cathode), where the work functions (φ) are much lower compared to the rest of the electrode and local field enhancement by a factor of up to 300 is possible [27]. Consequently, at moderate applied voltages (2–5 kV), electrons at these preferred injection sites on the cathode surface gain enough energy to tunnel through the potential energy barrier and into the liquid in a phenomenon known as field emission [25,28]. Once injected, these electrons acquire energy from the field (E) and are accelerated toward the anode. The electronic breakdown theory postulates that some of the electrons gain sufficient energy to ionize liquid molecules on collision, creating positive ions and more electrons that initiate avalanche and eventually breakdown. The onset of avalanche is met when the energy acquired by an electron (eEλ) along its mean free path (λ) is equivalent to the ionization energy (chv) of the liquid molecule (where hv is the quantum of energy lost in the ionization process and c is a constant) [25].

Given the dense nature of the liquid phase, it is a point of contest that injected electrons can gain enough energy to trigger Townsend-type impact ionization and electron multiplication in the liquid phase [29,30]. This has led to the rise of an alternative breakdown theory centered around bubble-initiated discharge. The bubble or cavitation theory proposes that a gaseous or vapor phase of lower dielectric strength triggers total breakdown of the liquid medium. Bubble formation in pure liquid dielectrics may be a consequence of several processes [1] including gas pockets on electrode surfaces, cavitation due to turbulence, and gaseous products released through the dissociation of liquid molecules upon impact with electrons. Electronic processes can also initiate bubble formation. For example, the injection of electrons by field emission at a point cathode or surface asperities can induce local heating resulting in the formation of gaseous cavities [27,31]. Once formed, these bubbles represent low density regions where the electron mean free path is significantly greater than in the surrounding liquid medium. This allows electrons to acquire enough energy from the local field to trigger impact ionization and avalanche discharge within the bubble and release more energy. The vapor phase discharges result in conduction current fluctuations, similar to those observed before breakdown in Fig. 4, while also facilitating the propagation of the low-density front across the gap [27]. For cases where the point electrode is negative (cathode), these low-density regions develop into streamers [29] that propagate toward the anode, eventually causing total breakdown of the liquid.

Streamer initiation for a point anode is less understood and less reproducible; however, it propagates at higher velocities and leads to lower breakdown [32,33]. Given the positive DC polarity applied to the point electrodes in this study (see Fig. 3), point anode-initiated streamers are expected to play a role in breakdown. According to Ingebrigsten et al. [34], there exists a voltage-dependent critical volume around a point electrode where the field is strong enough to initiate avalanche within a fluid. When a seed electron is emitted from the cathode, it is injected at the location most favorable to initiate avalanche required to vaporize the fluid. For a point positive electrode, on the other hand, the likelihood that field ionization supplies a favorably placed seed electron is a stochastic event that leads to a distribution of avalanches in the vicinity of the point anode with varying capacity to vaporize the fluid. Consequently, unlike at the cathode, streamer inception voltages at a point anode are less reproducible and more affected by external factors such as hydrostatic pressure [35] and mechanical cavitation due to fluid motion [36]. This is observable in the flow-dependent data in Fig. 6 where there is a large spread in the breakdown of HFE7500 as function of enforced motion, especially for Rechannel ranging from 2000 to 6000 due to the onset of internal flow turbulence. Even at internal flow Rechannel below this range, vortices due to flow over the cylindrical electrodes can induce low pressure regions for streamer nucleation and reduced breakdown strength. In the absence of enforced fluid motion, mechanical cavitation can still ensue due to electrohydrodynamic (EHD) motion [31,3739]. With EHD motion, the free charge generated by field emission or field ionization at the electrode interface is driven toward the oppositely charged electrode by electrophoretic forces (also called Coulomb forces). As the ions move, they collide and transfer some of their momentum to neutral liquid molecules, inducing bulk flow in the direction of the ion. The ratio of the induced liquid motion to ion drift velocity is always greater than one for liquids, highlighting the prominent role played by the induced motion in the transport of charge carriers (electroconvection) [25]. Furthermore, at strong enough fields, the induced flow becomes turbulent [40], which influences local pressure and in turn streamer inception and breakdown of the liquid medium. This phenomenon is a key driver of BD for cases in this study where the fluid is not subjected to forced convection. Imposing external forced motion on the fluid can however introduce a sweeping action on ionic charge carriers, which can disrupt the turbulence induced by EHD motion from fully developing in the interelectrode gap. Increasing temperature should also increase the propensity for the formation of the low-density regions for BD nucleation due to changes in fluid properties such as viscosity, density, and surface tension; however, within the tested range, increasing temperature only resulted in marginal reduction of BD strength for the laminar and transitional flow regimes.

In commercial dielectric liquids, the breakdown is also heavily influenced by the presence of impurities. These impurities can take the form of gaseous inclusions of lower breakdown strength and liquid contaminants such as water globules [41]. Commercial liquids also contain solid impurities which can be in the form of suspended solid particles or fibers. The suspended particle theory [25] of breakdown suggests that when a field is applied to a dielectric liquid of permittivity εDF containing a particle of permittivity εp, the said particle gets polarized and is subjected to a force exerted by the applied field. For εp > εDF, which is generally the case for solid particles and water globules, this force directs the particle toward the region of highest stress, i.e., the uniform field region in plate-plate or sphere-sphere electrode configurations subject to DC voltage. Once inside this uniform field region (dE/dx = 0), no force acts on the particle allowing it to reach an equilibrium state. An electric flux concentration then forms at the particle surface due its higher permittivity relative to the surrounding medium. The higher flux concentration attracts more particles, creating a particle chain that may eventually bridge the gap and initiate breakdown. In cases with too few particles to bridge the gap, local field enhancement due to their presence may induce microdischarges that can lead to early breakdown [25].

Liquid contaminants such as water globules can also significantly lower breakdown strength by becoming unstable under the imposition of a field [41]. Such impurities elongate under the influence of electrostatic forces leading to the formation of a low resistance bridge across the electrode gap and reduced breakdown strength. For conductive liquid impurities such as water where the ratio of εwaterDF is large, there exists a critical field given by Eq. (2) at which the globule elongation destabilizes and expands rapidly to trigger breakdown channels [41,42]. Based on empirical data, this field is given by
Ecrit=0.46(σRεDF)1/2
(2)

where R is the initial radius of the droplet of water, εDF is the permittivity of HFE7500 (i.e., εDF= ε0εr where εr is the relative permittivity of HFE7500 and ε0 is the permittivity of a vacuum), and σ is the surface tension of the water globule (0.043 N/m). Assuming a 1 μm initial water globule in HFE7500, this critical field inside the globule would be about 13 kV/mm, which is well within the range of values reached in this study, highlighting the significant effect that even microscopic quantities of water can have on breakdown.

Efforts were made to minimize the influence of particulates including the incorporation of a mesh-based filter upstream of the test chamber as well as the tracking of the sequence in which tests were performed, which provided a way of capturing the effects of impurity generation. These impurities could include dissolved water, gaseous products of liquid molecule dissociation and even particulates due to electrode erosion and other components of the flow loop such as the pump. The significance of particulate generation on the breakdown data from Fig. 6 was examined through a screen test in jmp software. The standard least squares was used to construct a model with breakdown as the response and model effects being Rechannel, same day test sequence and the interaction between the two. The Prob > F statics, which assess the significance level of the variables, are reported in Table 3. The Prob>F or p-value is the probability that the null hypothesis is true (i.e., the predictor does not contribute to the response). P-values closer to zero reject the hypothesis and thereby confirming the significance of the predictor. P-values < 0.01 are considered statically significant in this study. As shown in Table 3, Rechannel is significant in explaining the variance in breakdown strength and while test sequence might have an effect, it is not significant. Based on the Prob > F value for the interaction between the two variables, it can also be concluded that no interaction is detected.

Table 3

Effect test summary and report showing that same day test sequence is not a significant predictor of breakdown strength

PredictorF RatioProb > F or P-value (Probability that the predictor does not impact BD strength)Statically significant
Rechannel7.24220.00763Yes
Same day test sequence0.46620.49542No
Rechannel*Same day test sequence0.01610.89918No
PredictorF RatioProb > F or P-value (Probability that the predictor does not impact BD strength)Statically significant
Rechannel7.24220.00763Yes
Same day test sequence0.46620.49542No
Rechannel*Same day test sequence0.01610.89918No

To supplement the time-dependent analysis of the breakdown data, Ultraviolet/visible (UV/Vis) spectroscopy was used to assess the extent of particulate generation following repeated discharge events. This was done by analyzing the difference in the light absorbance of various fluid conditions/treatments including a pristine HFE7500 sample and a sample that was drawn from the flow loop reservoir after 95 breakdown events. A contaminated sample that showed significantly reduced breakdown strength was also tested to crystallize the change in absorption due to contaminants. The tests were performed using air as the background and covered a broad range of wavelengths to detect any changes. The results in Fig. 9 show negligible water content as indicated by the absence of a water peak, which would appear in the 1000 nm wavelength range. A closer examination of the peaks in the 280–320 wavelength range also reveals that the peak at 287 nm wavelength shifted for the fluid that has undergone repeated discharge events, which indicates degradation of the fluid itself. Furthermore, a peak that is not present in the pristine nor the 95BD sample is present at 292 nm for the contaminated sample indicating particulate uptake that is not apparent for the other two samples.

The imposition of cross-field flow on the dielectric fluid can have a sweeping effect on these water globules and other impurities leading to the rise in breakdown observed in Fig. 6 at Rechannel > 17,000. This sweeping effect can also act on the charge carriers resulting in the delay of breakdown. As discussed earlier, for the point electrodes used in this study, there is a role played by streamers formed at the anode. In this case, formation of such streamers is driven by field ionization wherein electrons tunneling from the liquid leave behind positive ionic charges of low mobility close to the anode. These space charge concentrations at the anode aid in local field enhancement, avalanche initiation for streamer formation and result in a virtual extension of the electrode into the gap. This enhanced field imparts higher velocities to streamers initiated at the point-positive electrode, resulting in lower breakdowns [32,33]. The introduction of forced motion can inhibit the formation of the intense local field by sweeping the ionic charges and delaying breakdown nucleation. As the ions drift through the interelectrode gap, their velocity vion can be expressed as
vion=μionE
(3)
where μion is the ionic mobility and E is the applied field. Although ion mobility is difficult to measure, it can be estimated for nonpolar fluids using Walden's rule [43]. In this model, the ion is considered to be a sphere of radius R drifting through a medium of dynamic viscosity μ. By assuming that the electric force (F = eE, where e is the elementary charge) acting on the ion is balanced by the viscous force which is given by Stoke's law (F=6πμRvion), Walden's rule gives the mobility as:
μion=e6πRμ
(4)

Considering R to be 10−9 m, the ionic drift velocity of HFE7500 is estimated to be 6.85 × 10−9 m2/Vs. For static flow conditions wherein average E-fields reached values as high as 16.8 kV/mm before breakdown, this corresponds to an ionic drift velocity of 0.12 m/s. On the other hand, at the highest Rechannel, the average flow velocity through the test section was as high as 1.9 m/s, lending support to the idea of the sweeping action of flow on space charges and impurities.

Fig. 9
UV spectrum of HFE samples comparing various fluid conditions/treatments along with the UV spectrum for pure water, showing negligible water contamination, but also shifting peaks for fluid that has undergone breakdown
Fig. 9
UV spectrum of HFE samples comparing various fluid conditions/treatments along with the UV spectrum for pure water, showing negligible water contamination, but also shifting peaks for fluid that has undergone breakdown
Close modal

In Fig. 10, it can be observed that higher flowrates exhibit a slow build up in leakage current compared to static flow conditions. This suggests that due to the sweeping effect, there is a slow buildup of space charge hindering the formation of an intense local field at the point electrodes and resulting in a breakdown that is more systematic. This contrasts with the more arbitrary breakdown at static and low flow (Rechannel = 2957) conditions where impurities and cavity-initiated breakdown dominate.

Fig. 10
I–V plots for different Rechannel values showing slow build up in leakage current at higher flow due to the strengthening sweeping action of flow
Fig. 10
I–V plots for different Rechannel values showing slow build up in leakage current at higher flow due to the strengthening sweeping action of flow
Close modal

5 Conclusions

As part of the objective to enable direct cooling for improved thermal management of high-voltage, high-power modules, efforts have been undertaken to assess the voltage blocking characteristics of hydrofluoroether type dielectric fluid. The characterization tests are performed over a range of relevant temperature and flow conditions. Results indicated that cross-field flow can deteriorate the breakdown voltage at low velocities, while higher flowrates in the range of Rechannel > 15,000 can result in an enhancement of the breakdown voltage. The observed drop in dielectric strength can be attributed to various mechanisms, primarily turbulence-induced cavitation that facilitates breakdown nucleation. While impurities can cause reduced breakdown strength, Rechannel > 15,000 can have a sweeping effect on these impurities as well as on ionic charge carriers, hindering the formation of an intense local field at the point electrodes and resulting in delayed inception of breakdown. The sweeping action of charge carriers and impurities is supported by IV plots which show a slow buildup of prebreakdown conduction currents at Rechannel = 22175. This contrasts with the I-V curves for static and Rechannel = 2957, wherein breakdown occurs arbitrarily with no sustained rise in current, pointing to the dominant role played by impurities, flow-induced cavitation as well as rapid ionization and avalanche that generate fast-moving low-density streamers that trigger breakdown. As temperature increases, the ability for cavity-initiated breakdown also increases due to the changes in fluid properties such as viscosity, density, and surface tension. This is, to a certain extent, supported by the slight drop observed in breakdown strength with increasing temperature for Rechannel in the laminar and transitional regime, while, conversely, no effect is seen for the fully turbulent Rechannel.

Fluid motion and impurity content are seen to be significant drivers of the voltage blocking characteristics of HFE7500, and any implementation of direct cooling of high voltage power electronics would need to adopt careful filtering and impurity monitoring measures. Additionally, while enforced motion is shown to lower breakdown strength of HFE7500, using flow velocities of Rechannel > 15,000 can take advantage of the sweeping action of flow to restore breakdown strength.

Acknowledgment

The authors would like to thank the Center for Power Optimization of Electro-Thermal Systems (POETS) for their ongoing support. Any opinions, finding, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. The authors would also like to thank Michael Lynn for his help on the additive manufacturing needs of this study and Yalin Wang for his input in the assembly of the high voltage test setup.

Funding Data

  • National Science Foundation (Award No. 2014-00555-04; Funder ID: 10.13039/100000001).

Data Availability Statement

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.

References

1.
Gersh
,
J.
,
DiMarino
,
C.
,
DeVoto
,
D.
,
Paret
,
P.
,
Major
,
J.
, and
Gage
,
S.
,
2021
, “
Evaluation of Low-Pressure-Sintered Multi-Layer Substrates for Medium-Voltage SiC Power Modules
,” 2021 IEEE Applied Power Electronics Conference and Exposition (
APEC
), Phoenix, AZ, June 14–17, pp.
20
26
.10.1109/APEC42165.2021.9487244
2.
Boteler
,
L. M.
,
Niemann
,
V. A.
,
Urciuoli
,
D. P.
, and
Miner
,
S. M.
,
2017
, “
Stacked Power Module With Integrated Thermal Management
,” 2017 IEEE International Workshop on Integrated Power Packaging (
IWIPP
), Delft, The Netherlands, Apr. 5–7, pp.
1
5
.10.1109/IWIPP.2017.7936764
3.
Zhu
,
N.
,
Mantooth
,
H. A.
,
Xu
,
D.
,
Chen
,
M.
, and
Glover
,
M. D.
,
2017
, “
A Solution to Press-Pack Packaging of SiC MOSFETS
,”
IEEE Trans. Ind. Electron.
,
64
(
10
), pp.
8224
8234
.10.1109/TIE.2017.2686365
4.
Shinde
,
P. A.
,
Bansode
,
P. V.
,
Saini
,
S.
,
Kasukurthy
,
R.
,
Chauhan
,
T.
,
Shah
,
J. M.
, and
Agonafer
,
D.
,
2019
, “
Experimental Analysis for Optimization of Thermal Performance of a Server in Single Phase Immersion Cooling
,”
ASME
Paper No. IPACK2019-6590. 10.1115/IPACK2019-6590
5.
Bansode
,
P. V.
,
Shah
,
J. M.
,
Gupta
,
G.
,
Agonafer
,
D.
,
Patel
,
H.
,
Roe
,
D.
, and
Tufty
,
R.
,
2019
, “
Measurement of the Thermal Performance of a Custom-Build Single-Phase Immersion Cooled Server at Various High and Low Temperatures for Prolonged Time
,”
ASME J. Electron. Packag.
,
142
(
1
), p.
011010
.10.1115/1.4045156
6.
Bostanci
,
H.
,
Van Ee
,
D.
,
Saarloos
,
B. A.
,
Rini
,
D. P.
, and
Chow
,
L. C.
,
2009
, “
Spray Cooling of Power Electronics Using High Temperature Coolant and Enhanced Surface
,”
2009 IEEE Vehicle Power and Propulsion Conference
, Dearborn, MI, Sept. 7–10, pp.
609
613
.10.1109/VPPC.2009.5289793
7.
Wei
,
T.-W.
,
Oprins
,
H.
,
Cherman
,
V.
,
De Wolf
,
I.
,
Beyne
,
E.
,
Yang
,
S.
, and
Baelmans
,
M.
, May
2018
, “
3D Printed Liquid Jet Impingement Cooler: Demonstration, Opportunities and Challenges
,” 2018 IEEE 68th Electronic Components and Technology Conference (
ECTC
), San Diego, CA, May 29–June 1, pp.
2389
2396
.10.1109/ECTC.2018.00360
8.
Whitt
,
R.
,
Huitink
,
D.
,
Emon
,
A.
,
Deshpande
,
A.
, and
Luo
,
F.
,
2021
, “
Thermal and Electrical Performance in High-Voltage Power Modules With Nonmetallic Additively Manufactured Impingement Coolers
,”
IEEE Trans. Power Electron.
,
36
(
3
), pp.
3192
3199
.10.1109/TPEL.2020.3015226
9.
Fu
,
B.-R.
,
Lee
,
C.-Y.
, and
Pan
,
C.
,
2013
, “
The Effect of Aspect Ratio on Flow Boiling Heat Transfer of HFE-7100 in a Microchannel Heat Sink
,”
Int. J. Heat Mass Transfer
,
58
(
1–2
), pp.
53
61
.10.1016/j.ijheatmasstransfer.2012.11.050
10.
Schulz-Harder
,
J.
,
Dezord
,
J. B.
,
Schaeffer
,
C.
,
Avenas
,
Y.
,
Puig
,
O.
,
Rogg
,
A.
,
Exel
,
K.
, and
Utz-Kistner
,
A.
,
2006
, “
DBC (Direct Bond Copper) Substrate With Integrated Flat Heat Pipe
,”
Twenty-Second Annual IEEE Semiconductor Thermal Measurement and Management Symposium
, Dallas, TX, Mar. 14–16, pp.
152
156
.10.1109/STHERM.2006.1625221
11.
Kwon
,
B.
,
Foulkes
,
T.
,
Yang
,
T.
,
Miljkovic
,
N.
, and
King
,
W. P.
,
2020
, “
Air Jet Impingement Cooling of Electronic Devices Using Additively Manufactured Nozzles
,”
IEEE Trans. Compon., Packag., Manuf. Technol.
,
10
(
2
), pp.
220
229
.10.1109/TCPMT.2019.2936852
12.
Birbarah
,
P.
,
Gebrael
,
T.
,
Foulkes
,
T.
,
Stillwell
,
A.
,
Moore
,
A.
,
Pilawa-Podgurski
,
R.
, and
Miljkovic
,
N.
,
2020
, “
Water Immersion Cooling of High Power Density Electronics
,”
Int. J. Heat Mass Transfer
,
147
, p.
118918
.10.1016/j.ijheatmasstransfer.2019.118918
13.
Jung
,
K. W.
,
Kharangate
,
C. R.
,
Lee
,
H.
,
Palko
,
J.
,
Zhou
,
F.
,
Asheghi
,
M.
,
Dede
,
E. M.
, and
Goodson
,
K. E.
, May
2017
, “
Microchannel Cooling Strategies for High Heat Flux (1 kW/cm2) Power Electronic Applications
,” 2017 16th IEEE Intersociety Conference on Thermal and Thermomechanical Phenomena in Electronic Systems (
ITherm
), Orlando, FL, May 30–June 2, pp.
98
104
.10.1109/ITHERM.2017.7992457
14.
DiMarino
,
C. M.
,
Mouawad
,
B.
,
Johnson
,
C. M.
,
Boroyevich
,
D.
, and
Burgos
,
R.
,
2020
, “
10-kV SiC MOSFET Power Module With Reduced Common-Mode Noise and Electric Field
,”
IEEE Trans. Power Electron.
,
35
(
6
), pp.
6050
6060
.10.1109/TPEL.2019.2952633
15.
Deshpande
,
A.
,
Luo
,
F.
,
Iradukunda
,
A.
,
Huitink
,
D.
, and
Boteler
,
L.
,
2019
, “
Stacked DBC Cavitied Substrate for a 15-kV Half-Bridge Power Module
,” 2019 IEEE International Workshop on Integrated Power Packaging (
IWIPP
), Toulouse, France, Apr. 24–26, pp.
12
17
.10.1109/IWIPP.2019.8799077
16.
Boteler
,
L. M.
,
Hinojosa
,
M.
,
Niemann
,
V. A.
,
Miner
,
S. M.
, and
Gonzalez-Nino
,
D.
, May
2017
, “
High Voltage Stacked Diode Package With Integrated Thermal Management
,” 2017 16th IEEE Intersociety Conference on Thermal and Thermomechanical Phenomena in Electronic Systems (
ITherm)
, Orlando, FL, May 30–June 2, pp.
913
920
.10.1109/ITHERM.2017.7992583
17.
Ikeda
,
M.
,
Teranishi
,
T.
,
Honda
,
M.
, and
Yanari
,
T.
,
1981
, “
Breakdown Characteristics of Moving Transformer Oil
,”
IEEE Trans. Power Appar. Syst
,
PAS-100
(
2
), pp.
921
928
.10.1109/TPAS.1981.316952
18.
Schmidt
,
W. F.
,
1989
, “
Electrical Breakdown of Liquid Hydrocarbons in a Test Cell With One Rotating Electrode
,”
IEEE Trans. Electr. Insul.
,
24
(
2
), pp.
179
183
.10.1109/14.90268
19.
Zaky
,
A. A.
,
Megahed
,
I. Y.
, and
Abdallah
,
M. A.
,
1996
, “
Effect of Cross-Field Flow on Conduction Current and Breakdown in Transformer Oil Using Point-to-Plane Electrodes and Direct Voltage
,”
ICDL'96. 12th International Conference on Conduction and Breakdown in Dielectric Liquids
, Roma, Italy, July 15–19, pp.
283
286
.10.1109/ICDL.1996.565473
20.
Theodossiou
,
G. G.
,
Lee
,
M. J.
, and
Nelson
,
J. K.
,
1984
, “
Influence of Enforced Liquid Motion on Breakdown Properties
,”
Conference on Electrical Insulation Dielectric Phenomena - Annual Report 1984
, Claymont, DE, Oct. 21–25, pp.
279
284
.10.1109/EIDP.1984.7684000
21.
Tuma
,
P. E.
,
2001
, “
Using Segregated HFEs as Heat Transfer Fluids as Heat Transfer Fluids Avoiding Problems in System Design
,”
Chem. Process.
, 64(2), pp.
47
50
.https://www.semanticscholar.org/paper/Using-Segregated-HFEs-as-Heat-Transfer-Fluids-in-Tuma/3ca3e62184b0aa7319ccf680c4eff16325e849f7
22.
3M Electronics Markets Materials Division,
2022
, “
3M-Novec-7500-Engineered Fluid
,” 3M Company, St. Paul, MN, accessed Mar. 20, 2022, https://multimedia.3m.com/mws/media/65496O/3m-novec-7500-engineered-fluid.pdf
23.
Frey
,
D.
,
Schanen
,
J. L.
,
Auge
,
J. L.
, and
Lesaint
,
O.
,
2003
, “
Electric Field Investigation in High Voltage Power Modules Using Finite Element Simulations and Partial Discharge Measurements
,”
38th IAS Annual Meeting on Conference Record of the Industry Applications Conference
, Salt Lake City, UT, Oct. 12–16, pp.
1000
1005
.10.1109/IAS.2003.1257662
24.
Lienhard
,
J. H.
,
1966
, “
Synopsis of Lift, Drag, and Vortex Frequency Data for Rigid Circular Cylinders
,” Technical Extension Service,
300
,
Washington State University, Pullman, WA
.https://www.uh.edu/engines/vortexcylinders.pdf
25.
Kuffel
,
E.
,
Zaengl
,
W. S.
, and
Kuffel
,
J.
,
2000
, “
Chapter 6—Breakdown in Solid and Liquid Dielectrics
,”
High Voltage Engineering Fundamentals
, 2nd ed.,
E.
Kuffel
,
W. S.
Zaengl
, and
J.
Kuffel
, eds.,
Newnes
,
Oxford, UK
, pp.
367
394
.
26.
André
,
D.
,
2011
, “
Conduction and Breakdown Initiation in Dielectric Liquids
,”
2011 IEEE International Conference on Dielectric Liquids
, Trondheim, Norway, June 26–30, pp.
1
11
.
27.
Forster
,
E. O.
, and
FitzPatrick
,
G. J.
,
1985
, “
Electric Breakdown in Dielectric Liquids
,”
Phys. Technol.
,
16
(
6
), pp.
282
287
.10.1088/0305-4624/16/6/I02
28.
Aljure
,
M.
,
Becerra
,
M.
, and
Karlsson
,
M. E.
,
2019
, “
On the Injection and Generation of Charge Carriers in Mineral Oil Under High Electric Fields
,”
J. Phys. Commun.
,
3
(
3
), p.
035019
.10.1088/2399-6528/ab0d59
29.
Sharbaugh
,
A. H.
,
Devins
,
J. C.
, and
Rzad
,
S. J.
,
1978
, “
Progress in the Field of Electric Breakdown in Dielectric Liquids
,”
IEEE Trans. Electr. Insul
,
EI-13
(
4
), pp.
249
276
.10.1109/TEI.1978.298076
30.
Forster
,
E. O.
,
1985
, “
A Critical Assessment of Electrical Breakdown in Liquid Dielectrics
,”
Conference on Electrical Insulation Dielectric Phenomena-Annual Report 1985
, Amherst, NY, Oct. 20–24, pp.
15
26
.10.1109/CEIDP.1985.7728126
31.
Watson
,
P. K.
,
1985
, “
Electrostatic and Hydrodynamic Effects in the Electrical Breakdown of Liquid Dielectrics
,”
IEEE Trans. Electr. Insul.
,
EI-20
(
2
), pp.
395
399
.10.1109/TEI.1985.348860
32.
Qian
,
J.
,
Joshi
,
R. P.
,
Schamiloglu
,
E.
,
Gaudet
,
J.
,
Woodworth
,
J. R.
, and
Lehr
,
J.
,
2006
, “
Analysis of Polarity Effects in the Electrical Breakdown of Liquids
,”
J. Phys. D: Appl. Phys
,
39
(
2
), pp.
359
369
.10.1088/0022-3727/39/2/018
33.
Bhatt
,
M.
, and
Bhatt
,
P.
,
2022
, “
Effect of Voltage Type and Polarity on Streamer Dynamics in Transformer Oil
,”
Mater. Today: Proc.
, 62(13), pp.
7131
7136
.10.1016/j.matpr.2022.02.184
34.
Ingebrigtsen
,
S.
,
Lundgaard
,
L. E.
, and
Åstrand
,
P.-O.
,
2007
, “
Effects of Additives on Prebreakdown Phenomena in Liquid Cyclohexane: I. Streamer Initiation
,”
J. Phys. D: Appl. Phys.
,
40
(
17
), pp.
5161
5169
.10.1088/0022-3727/40/17/022
35.
Lesaint
,
O.
,
2016
, “
Prebreakdown Phenomena in Liquids: Propagation ‘Modes’ and Basic Physical Properties
,”
J. Phys. D: Appl. Phys.
,
49
(
14
), p.
144001
.10.1088/0022-3727/49/14/144001
36.
Lesaint
,
O.
, and
Costeanu
,
L.
,
2017
, “
Positive Streamer Inception in Cyclohexane: Evidence of Formative Time and Cavitation Process
,” 2017 IEEE 19th International Conference on Dielectric Liquids (
ICDL
), Manchester, UK, June 25–29, pp.
1
4
.10.1109/ICDL.2017.8124600
37.
McCluskey
,
F. M. J.
, and
Denat
,
A.
,
1996
, “
The Behavior of Small Bubbles Generated by Electrical Current Impulses Over a Wide Range of Applied Pressures
,”
J. Appl. Phys.
,
80
(
4
), pp.
2049
2059
.10.1063/1.363098
38.
Grassi
,
W.
,
Testi
,
D.
, and
Vista
,
D. D.
,
2007
, “
Optimal Working Fluid and Electrode Configuration for EHD-Enhanced Single-Phase Heat Transfer
,”
J. Enhanced Heat Transfer
,
14
(
2
), pp.
161
173
.10.1615/JEnhHeatTransf.v14.i2.60
39.
Darabi
,
J.
, and
Wang
,
H.
,
2005
, “
Development of an Electrohydrodynamic Injection Micropump and Its Potential Application in Pumping Fluids in Cryogenic Cooling Systems
,”
J. Microelectromech. Syst.
,
14
(
4
), pp.
747
755
.10.1109/JMEMS.2005.845413
40.
Butcher
,
M.
,
Neuber
,
A. A.
,
Cevallos
,
M. D.
,
Dickens
,
J. C.
, and
Krompholz
,
H.
,
2006
, “
Conduction and Breakdown Mechanismsin Transformer Oil
,”
IEEE Trans. Plasma Sci.
,
34
(
2
), pp.
467
475
.10.1109/TPS.2006.872487
41.
Lucas
,
J. R.
,
2001
,
High Voltage Engineering
,
University of Moratuwa
,
Moratuwa, Sri Lanka
.
42.
Garton
,
C. G.
,
Krasucki
,
Z.
, and
Bowden
,
F. P.
,
1964
, “
Bubbles in Insulating Liquids: Stability in an Electric Field
,”
Proc. R. Soc. London, Ser. A
,
280
(
1381
), pp.
211
226
.10.1098/rspa.1964.0141
43.
Schmidt
,
W. F.
,
1999
, “
Electrons, Holes, and Ions in Non-Polar Dielectric Liquids
,” Proceedings of 1999 IEEE 13th International Conference on Dielectric Liquids (
ICDL'99
) (Catalogue No. 99CH36213), Nara, Japan, July 25, pp.
1
8
.10.1109/ICDL.1999.798854