Pool boiling performances comparison of FC-72 and Novec 649 in the presence of a DC electric field

Electronics cooling by direct immersion of the components in a dielectric liquid to obtain pool boiling is a valuable solution to dissipate the high heat fluxes of latest and increasingly challenging requirements. In the last decades, perfluorocarbons such as FC-72 were intensively studied and used. However, such fluids will be replaced by other refrigerants with lower Global Warming Potential. The Novec 649 is a suitable alternative to FC-72 as most of the thermophysical properties are similar. Enhanced boiling of FC-72 due to electric fields has been a subject of studies in the past years, for ground and space applications. The aim of this work is to compare the boiling performances of the two fluids in the presence of a DC electric field. Boiling curves were obtained at ambient pressure with fully degassed fluids and at different subcooling levels. Results showed that the performances of the two fluids are comparable in the absence of the electric field, but there are some differences when the electric field is on: electric field enhances the heat transfer coefficient of Novec 649 at low heat fluxes and the Critical Heat Flux at high subcooling. Boiling patterns and bubbles dynamics appear quite different between the two fluids in the presence of the electric field. These observations are explained with the fact that while FC-72 behaves as an electric insulator, while Novec 649 is a leaky-dielectric in the examined conditions. Thus, Novec 649 is a valid replacement of FC-72 and further investigations are needed to better quantify its boiling performances under the action of an electric field.


Introduction
Boiling heat transfer is recognized as one of the most effective ways to remove heat.In the last decades, the cooling needs of microelectronics and power electronics increased considerably.An alternative to traditional air and indirect liquid cooling consists of the direct immersion cooling, where the electronic components dissipate substantial amounts of heat by boiling of a surface in direct contact of a dielectric liquid.The refrigerant fluid should have good dielectric and thermophysical properties, chemical compatibility with the materials in contact with, chemical stability, low flammability, low toxicity, and low environmental impact [1].
Concerning the last point, refrigerants must avoid ozone depletion and greenhouse effect.In the last four decades, the perfluorohexane (C6F14 that can be found on the market as FC-72 Fluorinert TM ) has been studied and used for cooling of power electronics.However, in atmosphere FC-72 is a greenhouse gas having GWP of 9000 over a one-hundred-year lifetime [2].For this reason, it will be replaced by other coolants, such as the Novec TM 649 Engineered Fluid, commercial name of the CF3CF2C(O)CF(CF3).Thermophysical properties of the two fluids are comparable (see Table 1), but the Novec 649 has global warming potential of one over a one-hundred-year lifetime.Thus, it is interesting to compare the boiling performances of the two fluids in view of a replacement.
The very first pool boiling experiments performed with Novec 649 are due to Forrest et al. [1] in 2009: they used a nickel wire and compared with FC-72, founding that the performances are comparable.Novec 649 showed slightly higher heat transfer coefficient but moderately lower Critical Heat Fux (CHF).More recently, other authors compared the two fluids using enhanced surfaces such as copper surfaces with deep minichannels [3], microporous copper surfaces and hydrophilic and hydrophobic nanostructured surfaces [4].The results are similar to the ones by [1], i.e., similar performances between the two fluids were observed.In addition to the enhanced boiling surfaces, the dielectric properties of dielectric refrigerants allow to impose an external electric field to enhance boiling performances.Indeed, the presence of two phases and the liquid-vapor interface provide an electric force on bubbles, which can improve the heat transfer coefficient and delay the Critical Heat Flux [5].These effects are particularly efficient in microgravity conditions, where buoyancy force is negligible.The use of an external electric field can significantly improve the boiling performances in space applications, restoring or exceeding the ground ones [6].When a fluid is in the presence of an electric field, it is subjected to a volumetric electric force fe [7]: The first term of the right side of equation ( 1) is named electrophoresis and depends on the electric field E and the free charge density ρ E ; it is present when charge build-up occurs, and it is usually neglected in dielectric fluids.The second term is the dielectrophoresis, that is proportional to the electric field squared and the gradient of electric permittivity ε R (ε 0 is the vacuum permittivity).The third term is called electrostriction (ρ is the mass density), and it is an irrotational term, thus it does not induce fluid motion.According to [8], the relaxation time τ = ε 0 ε R /σ is the ratio between the electric permittivity and the electric conductivity of a fluid σ, and represents the characteristic time that charges need to move from the bulk to the external boundaries -such as the liquid-vapor interfacesexcluding the electric field from the bulk.The value of τ must be compared with a characteristic time t c of the system, such as the period of electric field frequency or the interface oscillation.charges have no time to move towards the boundaries and thus the fluid behaves as a perfect insulator: electric field penetrates in the bulk and the dielectrophoretic and electrostriction terms of equation ( 1) dominate (electrophoretic term is negligible).If τ ≪ t c the charges reach the system boundaries within the characteristic time of the system, excluding the electric field from the bulk and making the fluid a conductor.In this case, the electrophoretic term can overcome the other terms.There is another condition, when τ ~ t c and the electric field is present in the bulk but there is also free charges buildup; the fluid in these conditions is named leaky-dielectric [9].In general, the electrophoretic term (due to the presence of free charges) is larger than the other two, when present.For insulating fluids, high electric permittivity gradients and electric field intensities influence the fluid dynamic, as dielectrophoretic and electrostriction terms are involved.Electric fields have been studied in the past to enhance boiling performances of FC-72.The aim of this work is to experimentally investigate the enhancement capability of the electric field with Novec 649 also compare it with the FC-72.

Experimental methods
Figure 1.Schematic of the boiling support, the test cell and the and the pressurization systems.Experiments were conducted using a dedicated apparatus, capable to obtain boiling curves controlling pressure, subcooling level and Non-Condensable Gases (NCG) concentration; in this framework, ambient pressure and complete degassed fluids were used, while different subcooling levels were set.The experimental facility is made of five subsystems: the boiling section (including the heater and the electrode to produce the electric field), the test cell, the pressure control, the temperature control and the power supplies and data acquisition systems.Part of the apparatus is shown in figure 1.
The boiling section is a 3D printed polycarbonate (PC) support fixed with screws to the test cell top; thermophysical properties of the PC allow to reach temperatures above 120 °C and ensure proper thermal insulation (k = 0.2 W/m/K).The square 10 × 10 × 0.5 mm 3 P-doped N-type silicon heaters that produce boiling are glued to the support by means of thermal and electrical insulating glue; the doped silicon used is electrically conductive allowing Joule heating by means of a DC power supply (ELC ALR3206D) connected with two copper wires soldered by ultrasonic welding technology.Current and voltage across the heater were measured to calculate the average heat flux provided to the heater during boiling; the good insulation properties of the support and the glue ensured that only a negligible amount of power was dissipated by conduction.The maximum temperature measurement error is calculated as 0.6 K.The complete uncertainty analysis is described by Liu et al. [10].Temperature signal was also connected to a safety system that cat off the power when an abrupt increase of temperature occurred, indicating that the CHF value was reached.Above the heater, a metal grid made of 7 stainless-steel rods of 1 mm diameter and 5 mm spaced was placed at 6 mm from the grounded surface located at the heater level; this grid configuration produces a nearly uniform electrostatic field in the boiling region allowing the vapor to escape.A high DC of 15 kV was applied producing an electric field with average value of 2.5 × 10 6 V/m.
The test cell is a 2.3 L aluminium box that contains the boiling section and is filled with liquid.Two lateral windows allow to observe the boiling process using a high-speed camera (XIMEA xiQ-MQ022MG-CM, up to 450 fps) and a LED as backlight illumination.The cell is externally insulated with foam.The temperature of the liquid inside the cell is monitored by two temperature sensors (AD590) placed in different locations.During the experiments, the temperature of the liquid was kept within ±0.5 °C of the setpoint.
The pressure control system consists of a compensating bellows connected to the test cell.The liquid fills one side of the bellows chamber, while the other side contains air.Air can be injected from a high-pressure storage (2 bar) or ejected in the ambient operating two electro-valves.With this subsystem the pressure can be adjusted compensating liquid volume variations when temperature changes.
The temperature control system is a water closed loop operated by a pump.Water flows in a heating/cooling chamber below the test cell and in a heat exchanger on whose sides electric heater pads (200 W) and 8 Peltier cells are arranged.A PID control system switches the heaters to heat and the Peltier cells to cool depending on the test cell liquid temperature.A stirrer inside the test cell ensures the homogenization (not during the experiments).
Finally, a National Instruments (NI) cDAQ-9174 data acquisition device was used to collect the signals of the sensors.Images from the camera were collected to observe the boiling patterns.

Tests at saturation
In figure 2 the performances of the two fluids are showed.It can be noticed that the CHF of FC-72 is slightly higher than the Novec 649, while the heat transfer coefficient is better for Novec.Similar trends were obtained by Forrest et al. [1] with the same fluids boiling on a Nickel wire.Even if almost the same superheat is required, Novec 649 has a lower saturation temperature, resulting in a lower wall temperature.The theoretical values of CHF calculated with Zuber-Kutateladze correlation [11] are reported.Measured values without electric field are predicted within 6% error, which is the same order of magnitude of the uncertainty.At saturation, the effect of the electric field on the FC-72 is to increase the CHF of about 8%, while for Novec 649 remains the same.However, electric field has a strong effect on the heat transfer coefficient of Novec 649 for low heat fluxes.Such effect is not present for FC-72.Finally, the VDI correlation [12] is used to predict the slope of the heat transfer coefficient (using 0.82 as exponent).It is in good agreement for both fluids in the absence of the electric field and for the FC-72 with the electric field, while electric field induces an enhancement effect for Novec 649.

Effect of subcooling
In figure 3 the boiling curves at different subcooling levels (saturation, 5 K and 25 K) are compared for the two fluids.For the FC-72 the CHF always increases with subcooling; moreover, the wall superheat (that in turn determines the heat transfer coefficient) does not change significantly with subcooling level or electric field.On the other hand, for the Novec 649 the presence of the electric field produces a decrease of the wall superheat especially at low heat fluxes and high subcooling, indicating that the combination of subcooling and electric field enhances the heat transfer coefficient.Concerning the CHF, while close to saturation there is no significant enhancement, a clear increase is observed with 25 K of subcooling.

Critical Heat flux
The CHF data are collected in figure 4. Generally, the CHF of FC-72 is higher than Novec 649 in the absence of electric field.The effect of the electric field for FC-72 is a delay of the boiling crisis, for all the subcooling levels considered.On the other hand, small or negligible enhancement of CHF is observed with Novec 649 at low subcooling levels.However, increasing the subcooling, the enhancement capability of the electric field increases, and the critical heat fluxes of both fluids is comparable.

Discussion
Despite FC-72 and Novec 649 present similar thermophysical properties and the boiling performances are comparable, there are some important differences in terms of electric field effects that were presented.Such aspects are associated to differences in the boiling patterns that were observed during the tests with the high-speed camera.
In figure 5 the boiling patterns of the two fluids in the same conditions are compared.FC-72 shows no significant differences between the two cases (figures 5c and 5d).Bubbles are elongated vertically due to the electric stresses (like the bubble at the left corner of figure 5d) but the large amount of vapor makes them to coalesce into the columns.Being the boiling pattern similar, also similar heat transfer coefficients were observed (figure 2).
The boiling pattern of FC-72 is very similar to the one of Novec 649 in the absence of electric field (figures 5a and 5c).However, a great difference appears when the electric field is applied: bubbles are vigorously pushed towards the heater and spreads laterally (figure 5b).In the authors opinion, this lateral agitation and spread of vapor outside the heater boundaries is the reason why the heat transfer coefficient is enhanced, especially at lower heat fluxes, when the amount of vapor is less.The shape of single bubbles is less elongated and more similar to a bullet (see figure 5a on the left).Lower bubbles density allows the electric field to distribute and penetrate maximizing the electric stresses at the interface [8,11] and producing forces that repel each other [13].On the other hand, when there is a large amount of vapor at high heat fluxes, bubbles coalesce in columns cancelling the previous benefits.Electric forces tend to push the bubbles to the heater, until the vapor reaches the proximity of the grid where there is a high gradient of the electric field, that induces vapor to escape towards the bulk liquid [14].However, Novec 649 vapor is subjected to higher electric stresses that obstacles the vapor escaping at low subcoolings, reducing the CHF enhancement (figure 4).Another proof of that is provided in figure 6. Increasing the subcooling, the amount of vapor below the grid is reduced.Vapor bubbles are always pushed towards the heater and laterally, but when the subcooling is higher (figures 6a and 6b) and bubbles are small and far from the grid, a reduction of the bubbles size is observed in the presence of the electric field.
High subcooling means higher temperature gradients, that in turn produce electroconvection.This fact, together with the improved penetration of the electric field (no vapor that shields it) can explain the CHF and heat transfer coefficient enhancement of Novec 649 at 25 K subcooling.
In summary, Novec 649 and FC-72 behave differently in the presence of an electric field.According to Table 1, the electric permittivity of the two fluids is similar.However, the electrical resistivity of Novec is three orders of magnitude lower than FC-72.This seems to be the most important difference together with the GWP.Electric resistivity and dielectric permittivity influence the relaxation time of the electric charges, and the resulting electric behaviour of the fluid: the relaxation time for the two fluids is τ = 155 s for FC-72 and about τ = 0.2 s for Novec 649.It is possible to estimate from the highspeed videos the bubbles detachment, which occurs between 0.1 and 1 seconds; this can be considered as the range of the characteristic time of the system tc.Thus, while FC-72 can be considered an insulating fluid (verified by [14]), the Novec 649 behaves as a leaky-dielectric.In a leaky-dielectric, free charges accumulate at the interfaces modifying the field; free charges usually dominate the electric force, and their effect is stronger than dielectrophoresis and electrostriction forces.

Conclusions
A comparison of the pool boiling performances at ambient pressure and degassed conditions of the two dielectric fluids FC-72 and Novec 649 was performed.Results showed that the fluids are mostly comparable in the absence of the electric field, confirming previous works available in literature.Zuber-Kutateladze and VDI correlations are verified for Critical Heat Flux and heat transfer coefficient respectively.Increasing the subcooling level produces an increase of the CHF.Indeed, a difference is noted when the electric field is applied by means of a grid electrode: bubbles motion and agitation is quite more intense for Novec 649 with respect to FC-72, leading to a reduction of the wall superheat in the first part of the nucleate boiling curve (the isolated bubbles regime).Noteworthy, an accurate measure of electric conductivity for Novec 649 is currently not available in literature.However, according to the manufacturer data and the experimental observations, we concluded that the Novec 649 behaves as a leaky-dielectric.
In conclusion, based on experimental evidence the Novec 649 is a suitable candidate to replace FC-72 in boiling applications and showed good potentialities for the enhancement through an external electric field -especially if combined with high subcooling levels.A better measure of the electrical conductivity, a detailed analysis of the electric forces and of the influence of the electrode shape will be performed.

Figure 2 .
Figure 2. Boiling curves and heat transfer coefficients comparison at saturation with and without electric field.The VDI correlation is also showed [3].

Figure 3 .
Figure 3. Boiling curves of Novec 649 and FC-72 with and without electric field and at different subcooling levels.

Figure 4 .
Figure 4. Critical Heat Flux for both fluids as a function of subcooling level in the absence and presence of the electric field.

Figure 5 .
Figure 5.Comparison of the boiling patterns in saturation conditions.(a) Novec 649 in the absence of electric field.(b) Novec 649 in the presence of the electric field.(c) FC-72 in the absence of electric field.(d) FC-72 in the presence of the electric field.

Figure 6 .
Figure 6.Comparison of the boiling patterns of Novec 649.(a) 25 K subcooling in the absence of electric field.(b) 25 K subcooling in the presence of the electric field.(c) 5 K subcooling in the absence of electric field.(d) 5 K subcooling in the presence of the electric field.€ Saturation conditions in the absence of electric field.(f) Saturation conditions in the presence of the electric field.
If τ ≫ t c the