<?xml-model href='http://www.tei-c.org/release/xml/tei/custom/schema/relaxng/tei_all.rng' schematypens='http://relaxng.org/ns/structure/1.0'?><TEI xmlns="http://www.tei-c.org/ns/1.0">
	<teiHeader>
		<fileDesc>
			<titleStmt><title level='a'>Thermal analysis of thermoelectric active cooling including external thermal resistances</title></titleStmt>
			<publicationStmt>
				<publisher>AIP Publishing</publisher>
				<date>12/11/2023</date>
			</publicationStmt>
			<sourceDesc>
				<bibl> 
					<idno type="par_id">10528039</idno>
					<idno type="doi">10.1063/5.0176286</idno>
					<title level='j'>Applied Physics Letters</title>
<idno>0003-6951</idno>
<biblScope unit="volume">123</biblScope>
<biblScope unit="issue">24</biblScope>					

					<author>Nicolas Marquez Peraca</author><author>Qing Zhu</author><author>Junichiro Kono</author><author>Geoff Wehmeyer</author>
				</bibl>
			</sourceDesc>
		</fileDesc>
		<profileDesc>
			<abstract><ab><![CDATA[Thermoelectric active cooling uses nontraditional thermoelectric materials with high thermal conductivity, high thermoelectric power factor, and relatively low figure of merit (ZT) to transfer large heat flows from a hot object to a cold heat sink. However, prior studies have not considered the influence of external thermal resistances associated with the heat sinks or contacts, making it difficult to design active cooling thermal systems or compare the use of low-ZT and high-ZT materials. Here, we perform a non-dimensionalized analysis of thermoelectric active cooling under forced heat flow boundary conditions, including arbitrary external thermal resistances. We identify the optimal electrical currents to minimize the heat source temperature and find the crossover heat flows at which low-ZT active cooling leads to lower source temperatures than high-ZT and even ZT→+∞ thermoelectric refrigeration. These optimal parameters are insensitive to the thermal resistance between the heat source and thermoelectric materials, but depend strongly on the heat sink thermal resistance. Finally, we map the boundaries where active cooling yields lower source temperatures than thermoelectric refrigeration. For currently considered active cooling materials, active cooling with ZT&lt;0.1 is advantageous compared to ZT→+∞ refrigeration for dimensionless heat sink thermal conductances larger than 15 and dimensionless source powers between 1 and 100. Thus, our results motivate further investigation of system-level thermoelectric active cooling for applications in electronics thermal management.]]></ab></abstract>
		</profileDesc>
	</teiHeader>
	<text><body xmlns="http://www.tei-c.org/ns/1.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xlink="http://www.w3.org/1999/xlink">
<div xmlns="http://www.tei-c.org/ns/1.0"><p>The increasing demand for high-performance computing clusters and artificial intelligence supercomputers has led to modern computer chips reaching peak powers on the order of 1000 W. <ref type="bibr">1,</ref><ref type="bibr">2</ref> These emerging technologies are imposing stringent requirements on efficient cooling solutions capable of dissipating large powers. <ref type="bibr">[3]</ref><ref type="bibr">[4]</ref><ref type="bibr">[5]</ref> In recent decades, thermoelectric (TE) devices have been explored for integrated circuit cooling. <ref type="bibr">[6]</ref><ref type="bibr">[7]</ref><ref type="bibr">[8]</ref> In traditional TE cooling, the goal is to achieve source temperatures lower than that of the heat sink, i.e., refrigeration. Materials with low thermal conductivity (j) and high power factors (PF &#188; ra <ref type="bibr">2</ref> , where a is the Seebeck coefficient and r is the electrical conductivity), that is, high ZT &#188; ra 2 T=j, are preferred for refrigeration since the coefficient of performance (COP) of a TE module increases with increasing ZT. <ref type="bibr">[9]</ref><ref type="bibr">[10]</ref><ref type="bibr">[11]</ref> Prior TE modeling works <ref type="bibr">[12]</ref><ref type="bibr">[13]</ref><ref type="bibr">[14]</ref><ref type="bibr">[15]</ref><ref type="bibr">[16]</ref><ref type="bibr">[17]</ref><ref type="bibr">[18]</ref><ref type="bibr">[19]</ref><ref type="bibr">[20]</ref><ref type="bibr">[21]</ref><ref type="bibr">[22]</ref><ref type="bibr">[23]</ref><ref type="bibr">[24]</ref><ref type="bibr">[25]</ref> focused on thermal models spanning across multiple layers of complexity, aspiring to understand under what circumstances TEs could outperform traditional air-cooling heat sinks or vapor-compression refrigerators. <ref type="bibr">12,</ref><ref type="bibr">15,</ref><ref type="bibr">26</ref> The overarching conclusion has been that TE materials such as Bi 2 Te 3 , SiGe, and PbTe with a ZT $ 1 cannot compete with traditional refrigeration systems in terms of COP. <ref type="bibr">12,</ref><ref type="bibr">15,</ref><ref type="bibr">27,</ref><ref type="bibr">28</ref> Nevertheless, TEs have found niche applications in the market due to their small size, robustness, and simplicity. <ref type="bibr">[29]</ref><ref type="bibr">[30]</ref><ref type="bibr">[31]</ref> For instance, TEs can be found in portable coolers, climate-control seat systems, and in laser diodes for temperature control and wavelength stabilization. <ref type="bibr">[30]</ref><ref type="bibr">[31]</ref><ref type="bibr">[32]</ref><ref type="bibr">[33]</ref> In 2015, Zebarjadi <ref type="bibr">34</ref> proposed a TE cooling mechanism-hereto referred to as TE active cooling-that considers a heat source warmer than the heat sink and uses TE materials with high j and high PF, but relatively low-ZT. Active cooling uses both the Fourier and Peltierinduced heat flows to transfer energy in the natural gradient direction of heat transfer, consequently acting as an active heat sink. <ref type="bibr">35</ref> This mechanism was recently experimentally implemented by Adams et al. <ref type="bibr">36</ref> using correlated (CePd 3 ) and magnon-drag metals (Co) as the p-and n-type materials of an active cooling device, respectively.</p><p>The recent discoveries in this field of research has led to new works looking for novel and low-cost materials suitable for active cooling applications-both experimentally and computationally. <ref type="bibr">35,</ref><ref type="bibr">[37]</ref><ref type="bibr">[38]</ref><ref type="bibr">[39]</ref><ref type="bibr">[40]</ref> However, previous active cooling studies have imposed constanttemperature boundary conditions in their thermal models and have ignored external thermal resistances. Here, external thermal resistances refers to every thermal resistance in series with the TE material, including interfacial, contact, constriction/spreading resistances and the thermal resistance of the physical heat sink itself. For TE refrigeration, it has been shown that not accounting for these parameters significantly impacts the performance of the TEs and leads to higher source temperatures than those of an optimally designed TE system. <ref type="bibr">14,</ref><ref type="bibr">16,</ref><ref type="bibr">17</ref> In addition, for most applications, the hot-and cold-side temperatures of a TE material are a response to the input conditions and not know a priori. <ref type="bibr">14,</ref><ref type="bibr">20</ref> In this paper, we perform an analysis of TE active cooling under forced heat flow boundary conditions, including arbitrary external thermal resistances. We consider a TE active cooler with device-level Seebeck coefficient (S), thermal conductance (K), and electrical resistance (R) [see Fig. <ref type="figure">1(a)</ref>]. In what follows, the nondimensionalized variables are indicated with a superscript asterisk. We find an analytical solution for the optimal electrical current J &#195; min that minimizes the heat source (chip) temperature, T 0 [see Fig. <ref type="figure">1(b)</ref>]. We identify the lower crossover heat flow &#240;Q &#195; 0 &#222; crossover at which ZT &#188; 0.1 active cooling yields lower chip temperatures than ZT &#188; 1 TE refrigeration. The parameters J &#195; min and &#240;Q &#195; 0 &#222; crossover are independent of the dimensionless heat source-side thermal conductance (G &#195; 0 ) but depend on the dimensionless heat sink-side thermal conductance (G &#195; 1 ). Moreover, we quantify the temperature difference DT &#195; between a baseline ZT ! &#254;1 TE refrigerator system and an active cooling device. For currently considered active cooling materials, we find that active cooling with ZT &lt; 0.1 is advantageous compared to ZT ! &#254;1 refrigeration for G &#195; 1 &gt; 15 and 1 &lt; Q &#195; 0 &lt; 100. Our results pave the way toward further studies of active cooling in high-heat flux applications.</p><p>We solve the following three 1D, steady-state energy balance equations for a TE material, assuming a constant rate of heat transfer (Q 0 ) on the heat source side, a constant heat sink (T 3 ) temperature, and including arbitrary external thermal resistances between the heat source and the heat sink sides [1=G 0 and 1=G 1 , respectively, see Fig. <ref type="figure">1</ref></p><p>(1)</p><p>(2)</p><p>Here, S, K, and R are the device Seebeck coefficient, device thermal conductance, and device thermal resistance, respectively. For a TE module consisting of N p-and n-type thermocouples of length L p;n and cross-sectional area A p;n ;</p><p>An=Ln , where a p;n ; j p;n , and q p;n are the Seebeck coefficient, thermal conductivity, and electrical resistivity of the individual p-n legs, respectively. Q 0 is the rate of heat transfer into the system, T 0 and T 3 are the heat source (chip) and heat sink temperatures, T 1 and T 2 are the unknown TE heat source-side and heat sink-side temperatures, I is the driving current, and G 0 and G 1 are the thermal conductances between the chip and the TE and between the TE and the heat sink, respectively [see Fig. <ref type="figure">1(b)</ref>]. We assume that S, K, and R are temperature-independent material properties; therefore no Thomson effect is present. The driving current I is taken to be positive when positive charge flows from the n-to the p-type TE leg, in which case heat is absorbed from the chip and liberated at the heat sink (cooling). <ref type="bibr">42</ref> To non-dimensionalize these equations, we start by dividing them by the crossover power under no external thermal resistances [i.e., S 2 T 2 3 2R , see Fig. <ref type="figure">1(a)</ref>], which gives</p><p>Here, we have defined the following dimensionless parameters: ). Nontraditional TE materials with low-ZT (high-K, solid black line) are suitable for active cooling <ref type="bibr">[34]</ref><ref type="bibr">[35]</ref><ref type="bibr">[36]</ref><ref type="bibr">[37]</ref><ref type="bibr">[38]</ref><ref type="bibr">[39]</ref><ref type="bibr">[40]</ref> applications in high-Q 0 flow cases, where the heat source is warmer than the heat sink. The crossover power, equal to S 2 T 2 3 =2R when no external thermal resistances are included, marks the transition point between regimes. (b) Thermal resistance network used in the thermal modeling of this work. Here, arbitrary heat source-side and heat sink-side thermal resistances, 1=G 0 and 1=G 1 , respectively, account for any finite external thermal resistances present in the system.</p><p>The device thermal conductance K only appears in the denominator of Eq. (12). Therefore, changing ZT 3 while holding the other dimensionless variables fixed can be easily interpreted as directly changing K without modifying S or R. The explicit solution to Eqs. ( <ref type="formula">4</ref>)-( <ref type="formula">6</ref>) is</p><p>The G &#195; 0 dependence only appears in the denominator of the first term in Eq. ( <ref type="formula">13</ref>). Therefore, any temperature differences or derivatives of this equation are independent of G &#195; 0 , which is an important observation for the discussion in the next sections. In the supplementary material, we show how the general solutions for T &#195; 0 and the dimensionless current (J &#195; ) reduce to previously reported active cooling solutions when there are no external thermal resistances. The baseline values of G &#195; 0 &#188; 4 and G &#195; 1 &#188; 46 used in some calculations in the following were estimated from previously reported TE measurements, <ref type="bibr">13,</ref><ref type="bibr">17</ref> with dimensional values shown in Table <ref type="table">S1</ref>.</p><p>Figure <ref type="figure">2</ref>(a) plots the dimensionless chip temperature T &#195; 0 as a function of J &#195; for different values of Q &#195; 0 . Since it is desirable to operate the TE device at optimal electrical current, we begin by minimizing T &#195; 0 with respect to J &#195; by calculating @T &#195; 0 @J &#195; &#188; 0. When arbitrary external thermal resistances are included, the solution is a fourth-order polynomial in J &#195; , and only one of the four roots is real and positive, which we define as J &#195; min (see the supplementary material). Gray vertical lines in Fig. <ref type="figure">2</ref>(a) correspond to J &#195; min for each Q &#195; 0 . In general, J &#195; min is a function of G &#195; 1 ; ZT 3 , and Q &#195; 0 and needs to be optimized in each case. From here on, we use this value of J &#195; in Eq. ( <ref type="formula">13</ref>), which reduces the number of system parameters to be optimized from 4 to 3, assuming Q &#195; 0 is given. Figure <ref type="figure">2</ref>(b) shows T &#195; 0;min plotted as a function of ZT 3 for four different values of Q &#195; 0 . In this plot, changing ZT 3 at fixed Q &#195; 0 is equivalent to varying K while holding the dimensional variables S, R, and Q 0 fixed for each value of Q &#195; 0 . For low values of Q &#195; 0 , increasing the figure of merit ZT 3 provides refrigeration below ambient temperature, as shown for Q &#195; 0 &#188; 0:1 (blue curve). However, as Q &#195; 0 increases, lower values of ZT 3 yield lower chip temperatures than high values of ZT 3 (red, black, and purple curves, respectively), which is the basis for TE active cooling. Notice that for the red, black, and purple curves T &#195; 0;min &gt; 1 for any ZT 3 , indicating that the chip temperature is always greater than the ambient temperature, as expected for active cooling. For Q &#195; 0 &#188; 1 and G 0;1 ! &#254;1, the T &#195; 0;min ! 1 limit is recovered (see the supplementary material), in agreement with the Fig. <ref type="figure">1</ref>(a) sketch showing</p><p>. The qualitative trends shown here also hold in the case where G &#195; 0 &gt; G &#195; 1 , as shown in Fig. <ref type="figure">S2</ref>. Figure <ref type="figure">3</ref> plots T &#195; 0;min as a function of Q &#195; 0 for ZT 3 &#188; 1 (dashed lines) and ZT 3 &#188; 0:1 (solid lines) and three different values of G &#195; 1 . This plot is a quantitative version of the sketch shown in Fig. <ref type="figure">1</ref>(a), now considering the effects of heat sink thermal resistances. Here, Q &#195; 0 increases while ZT 3 is held constant at either 0.1 or 1; from Eqs. ( <ref type="formula">8</ref>)-( <ref type="formula">12</ref>), this scenario could correspond to an increase in the dimensional heat flow Q 0 at fixed S and R while considering a large K (low ZT 3 ) or small K (high ZT 3 ). We define the dimensionless lower and upper crossover powers &#240;Q &#195; 0 &#222; crossover and &#240;Q &#195; 0 &#222; max as the intersection points between these two curves. Only the &#240;Q &#195; 0 &#222; crossover intersection is shown in Fig. <ref type="figure">3</ref>, while the behavior of &#240;Q &#195; 0 &#222; max is explored below. Figure <ref type="figure">3</ref> shows that active cooling with ZT 3 &#188; 0:1 yields a lower chip temperature than </p><p>G &#195; 1 &#188; 46 and approaches 1 for the case of infinite G &#195; 0;1 , representing no thermal resistances. &#240;Q &#195; 0 &#222; crossover is only a function of G &#195; 1 , as shown in the inset of Fig. <ref type="figure">3</ref>.</p><p>The physical meaning of the crossover points shown in Fig. <ref type="figure">3</ref> can be understood by referring to the simpler case with no external thermal resistances. Figure <ref type="figure">1</ref>(a) shows that crossover power in the noexternal resistance case represents the scenario where there is no temperature difference across the TE material, meaning that the Fourier heat flow contribution K&#240;T 1 &#192; T 2 &#222; in Eqs. ( <ref type="formula">2</ref>) and (3) vanishes for all K. The physical interpretation of the crossover in Fig. <ref type="figure">3</ref> including external resistances is more complicated because the crossover conditions do not typically correspond to the case when T &#195; 1 &#188; T &#195; 2 , but rather simply differentiate the regions where high K is preferable to low K for mitigating the temperature rise.</p><p>Figure <ref type="figure">3</ref> shows that at smaller G &#195; 1 and ZT 3 &#188; 1, the chip temperature is larger than the ambient even for small Q &#195; 0 near 0.1, meaning that the high-K (low-ZT 3 ) active cooling device is more advantageous for heat removal at smaller values of Q &#195; 0 , compared to the noresistance scenario. Beyond G &#195; 1 &#188; 15; &#240;Q &#195; 0 &#222; crossover stays within 5% of unity, and we recover the crossover of Fig. <ref type="figure">1</ref>(a) (S 2 T 2 3 =2R) for negligible thermal resistances. In dimensional units, using representative values from prior TE measurements reported in Table <ref type="table">S1</ref>, for G 1 $ 0:5 W/K and ZT 3 &#188; 0:1, active cooling is more advantageous than ZT 3 &#188; 1 refrigeration for powers above &#240;Q 0 &#222; crossover $ 30 W.</p><p>Figure <ref type="figure">4</ref> quantifies the temperature difference between a baseline ZT 3 ! &#254;1 TE refrigeration system and an active cooling device. We define a dimensionless temperature difference DT &#195; &#188; T &#195; 0;baseline &#192;T &#195; 0;min . Here, we are comparing two systems with identical external thermal conductances G &#195; 0 and G &#195; 1 and have chosen T &#195; 0;baseline to be the temperature of a ZT 3 ! &#254;1 TE refrigerator and T &#195; 0;min the temperature of an active cooling device with dimensionless figure of merit ZT 3 . Taking ZT 3 to infinity for finite G &#195; 1 corresponds to taking a device with thermal conductance K &#188; 0. DT &#195; does not depend on G &#195; 0 since we are calculating a temperature difference [see Eq. ( <ref type="formula">13</ref>)]. In the DT &#195; &gt; 0 (outperformance) regions, low-ZT 3 active cooling leads to lower source temperatures than ZT 3 ! &#254;1 refrigeration. This improvement compared to even the ZT 3 ! &#254;1 limit is possible, since in active cooling, the heat source temperature is larger than ambient [see Fig. <ref type="figure">1(a)</ref>], and the system operates in forced heat transfer mode. <ref type="bibr">43</ref> The thermal energy from the hot heat source and work applied to the system are dissipated at the cold heat sink, consistent with the second law of thermodynamics, consequently making active cooling more advantageous for heat removal than refrigeration. <ref type="bibr">43,</ref><ref type="bibr">44</ref> Gray areas in Fig. <ref type="figure">4</ref> correspond to DT &#195; &lt; 0 and represent regions where there is no benefit in using active cooling over TE refrigeration (underperformance). White solid lines are the DT &#195; &#188; 0 boundaries in each case.</p><p>Figure <ref type="figure">4</ref> shows that the outperformance regions compared to the ZT 3 ! &#254;1 limit shrink as G &#195; 1 decreases: in active cooling, the excess power that is transferred to the TE material for Peltier cooling must be dissipated by the heat sink, and a lower G &#195; 1 makes this task more 1 and Q &#195; 0 fixed is equivalent to taking the thermal conductance K to 0 at fixed power factor. In the regions where DT &#195; &gt; 0, active cooling yields lower chip temperatures than ZT 3 ! &#254;1 TE refrigeration; this improvement is possible because the chip temperature is warmer than the ambient temperature. Gray areas correspond to DT &#195; &lt; 0 and represent regions where there is no benefit in using active cooling over refrigeration. White solid lines are the DT &#195; &#188; 0 boundaries in each case. DT &#195; is independent of G &#195; 0 .</p><p>FIG. <ref type="figure">3</ref>. Crossover between low-ZT 3 active cooling and high-ZT 3 refrigeration. The ZT 3 &#188; 1 (dashed lines) and ZT 3 &#188; 0:1 (solid lines) curves intersect at the dimensionless lower &#240;Q &#195; 0 &#222; crossover (black arrows) and upper &#240;Q &#195; 0 &#222; max (not shown) crossover powers, marking the transition between the active cooling and refrigeration regimes. For &#240;Q &#195; 0 &#222; max &gt; Q &#195; 0 &gt; &#240;Q &#195; 0 &#222; crossover , active cooling with low-ZT 3 yields a lower chip temperature than high-ZT 3 refrigeration. Vertical and horizontal gray lines corresponds to Q &#195; 0 &#188; 1 and T &#195; 0;min &#188; 1, respectively. Inset: &#240;Q &#195; 0 &#222; crossover is independent of G &#195; 0 and shifts to lower values as G &#195; 1 decreases.</p><p>challenging. For example, Fig. <ref type="figure">4</ref>(a) shows that a dimensionless temperature difference DT &#195; &gt; 2:5 compared to ZT 3 ! &#254;1 can be achieved for ZT 3 &lt; 0:1 and Q &#195; 0 between 20 and 300. In addition, we can obtain the relative temperature reduction by calculating jT &#195; 0;baseline &#192; T &#195; 0;min j= T &#195; 0;baseline &#188; DT &#195; =T &#195; 0;baseline . For DT &#195; $ 2 and Q &#195; 0 &#188; 10, we obtain a DT &#195; =T &#195; 0;baseline $ 63% relative temperature reduction if active cooling is selected over ZT 3 ! &#254;1 TE refrigeration.</p><p>Figure <ref type="figure">5</ref> investigates the DT &#195; &#188; 0 boundaries of the outperformance regions (white solid lines in Fig. <ref type="figure">4</ref>) as a function of G &#195; 1 . For an application with specific Q &#195; 0 ; ZT 3 , or G &#195; 1 requirements, this plot provides the range of values needed to design an active cooling device capable of remaining competitive against best-case scenario TE refrigeration. We find that for currently considered active cooling materials, such as Cu-Ni alloys <ref type="bibr">35</ref> (green horizontal dashed line) and carbon nanotube fibers <ref type="bibr">37</ref> (CNTFs, blue horizontal dashed line), active cooling with ZT &lt; 0.1 is advantageous compared to ZT 3 ! &#254;1 refrigeration for G &#195; 1 &gt; 15 and 1 &lt; Q &#195; 0 &lt; 100 (vertical dashed lines), respectively. Common TE materials, such as BiTeSe (red horizontal dashed line), <ref type="bibr">45</ref> on the other hand, require G &#195; 1 &gt; 25 to outperform ZT 3 ! &#254;1 TE refrigeration for the same Q &#195; 0 range. This manuscript primarily focuses on changing ZT 3 by tuning the thermal conductance K at fixed power factor. If instead we increase the power factor while holding the dimensional values of K, G 1 , and Q 0 constant, this will have the effect of increasing ZT 3 , while reducing Q &#195; 0 and G &#195; 1 . These changes tend to reduce the temperature rise in the system, while also making active cooling less advantageous than traditional TE refrigeration, as the limitation of refrigeration is based on the finite power factor [see crossover point of S 2 T 2 3 =2R in Fig. <ref type="figure">1(a)</ref>]. Therefore, as expected, it is optimal to select the TE material to maximize the power factor for both TE refrigeration and active cooling applications. Prior work has also shown that thin-film <ref type="bibr">46,</ref><ref type="bibr">47</ref> and/or micro-thermoelectric devices <ref type="bibr">24,</ref><ref type="bibr">48</ref> can be advantageous for nearjunction thermal management of high heat fluxes. Based on the modeling in this work, further experimental studies integrating high power factor and high thermal conductance active cooling materials <ref type="bibr">37,</ref><ref type="bibr">40</ref> in thin-film or micro-thermoelectric format could offer opportunities to mitigate chip temperature rises if contact resistances can be made sufficiently small and fabrication can be performed in a scalable manner.</p><p>In summary, this work uses analytical solutions to the onedimensional TE equations to explore the temperature profiles and heat flows associated with TE active cooling when arbitrary external thermal resistances are included. We use this solution to obtain the range of figures of merit and dimensionless heat flows at which active cooling leads to lower source temperatures than ZT 3 ! &#254;1 TE refrigeration and to establish the lower and upper bounds of this region. These results show that for state-of-the-art materials, active cooling is advantageous compared to ZT 3 ! &#254;1 refrigeration for dimensionless heat sink thermal conductances larger than 15 and dimensionless source powers between 1 and 100 using dimensional values from previously reported TE measurements. <ref type="bibr">13,</ref><ref type="bibr">17</ref> We estimate a relative temperature reduction of at least 60% can be achieved for source powers on the order of 1000 W if active cooling is selected over TE refrigeration and the active cooling device has the same S and R as the traditional TE device. The solutions presented here can be used in the design and analysis of active cooling devices for thermal management and temperature control of high-power electronics.</p><p>See the supplementary material for a description of the electrical current optimization procedure, theoretical limiting cases of the active cooling equations, temperature calculations for the case where G &#195; 0 &gt; G &#195; 1 , and dimensional values used to obtain the base-case dimensionless G &#195; 0 and G &#195; 1 parameters.</p><p>This study was supported by the National Science Foundation under Grant No. 2230727. N.M.P would like to gratefully acknowledge Liu Idea Lab for Innovation and Entrepreneurship for their support. J.K. acknowledges support from the Robert A. Welch Foundation through Grant No. C-1509. 1 , the DT &#195; &#188; 0 boundaries (white solid lines in Fig. <ref type="figure">4</ref>) delimit the region where low-ZT 3 active cooling outperforms ZT 3 ! &#254;1 refrigeration. These boundaries are independent of G &#195; 0 and uniquely determined by G &#195; 1 for a given Q &#195; 0 and ZT 3 . Horizontal dashed lines are the ZT 3 values of a common TE material (BiTeSe, 45 red) and two candidates for active cooling applications, Cu-Ni alloys <ref type="bibr">35</ref> (green) and carbon nanotube fibers <ref type="bibr">37</ref> (CNTFs, blue), respectively. Vertical dashed lines correspond to Q &#195; 0 &#188; 1 and Q &#195; 0 &#188; 10 2 , respectively.</p></div><note xmlns="http://www.tei-c.org/ns/1.0" place="foot" xml:id="foot_0"><p>Published under an exclusive license by AIP Publishing</p></note>
		</body>
		</text>
</TEI>
