Professional Documents
Culture Documents
ContinuousElecrochemicalRefrigeration NatEn
ContinuousElecrochemicalRefrigeration NatEn
ContinuousElecrochemicalRefrigeration NatEn
net/publication/358516569
CITATIONS READS
3 213
3 authors, including:
Aravindh Rajan
Palo Alto Research Center
10 PUBLICATIONS 37 CITATIONS
SEE PROFILE
All content following this page was uploaded by Aravindh Rajan on 17 February 2022.
Zero-global-warming-potential cooling technologies can mitigate the climate change effects attributed to the use of conven-
tional vapour compression refrigeration. In this work, we conceptualize an electrochemical refrigeration cycle and demonstrate
a proof-of-concept prototype in continuous operation. The refrigerator is based on the Brayton cycle and draws inspiration
from redox flow battery technologies. A peak coefficient of performance of 8.09 was measured with a small temperature drop
of 0.07 K. A peak cooling load of 0.934 W with a coefficient of performance of 0.93 was measured, however, with only a mod-
est measured temperature drop of 0.15 K. This is still much lower than the theoretical maximum temperature drop of 2–7 K for
these electrolytes. This work could inspire research into high-cooling-capacity redox-active species, multi-fluid heat exchanger
design and high-efficiency electrochemical refrigeration cell architectures.
D
riven by rising regional temperatures1, growing regional electrical work from a temperature difference. We develop the
populations2 and accelerating urbanization3, the global Brayton Electrochemical Refrigerator (BECR), which achieves sus-
cooling demand is expected to triple by 2050 (refs. 4,5). tained electrochemical refrigeration by employing three key fea-
In a business-as-usual scenario, nearly all of this demand will be tures. First, we identify two complementary half-cell reactions (that
met using hydrofluorocarbons employed in vapour compres- is the triiodide/iodide and ferricyanide/ferrocyanide) that maximize
sion units6. Hydrofluorocarbons have global warming poten- cooling and stability. Second, we develop modified flow battery
tials several times higher than that of carbon dioxide, and electrochemical cell architectures that facilitate continuous cooling
vapour-compression-associated emissions are projected to be as by increasing single-pass electrochemical conversion and reducing
high as 9 GtCO2 equivalents per year (that is, 15% of CO2 equivalent electrochemical overpotential. Third, we base our technology on the
emissions) by 2050 (refs. 7,8). Brayton cycle as opposed to the Stirling cycle (Supplementary Note
One facet of the solution is the discovery and development of 1), which is frequently used by TREC systems. Additionally, unlike
zero-global-warming-potential alternative cooling technologies. In the single-half-cell thermogalvanic architectures, our dual-half-cell
principle, it is possible to take a heat engine, like a thermally regen- electrochemical system decouples ionic and heat transport, allowing
erative electrochemical cycle (TREC)9, with sufficiently low internal for access to higher current densities at lower Joule heat12,20. Finally,
irreversibility and induce refrigeration via the reverse operation with we choose the Brayton cycle rather than the Stirling cycle because the
an external work input. While TRECs have been widely researched isothermal heat absorption process in a Stirling cycle competes with
for power generation10–12, the converse, a continuous electrochemi- the advecting electrolyte, which concomitantly exchanges heat with
cally driven refrigerator, has not been previously demonstrated, to the surroundings. This coupled heat and mass transfer problem for the
the best of our knowledge. Electrochemically driven refrigeration has Stirling cycle is challenging; the Brayton cycle, which also has its own
been tenuously hypothesized and modelled in the past13–16. Duan et al. challenges, circumvents this specific coupled problem by using a ther-
developed a numerical model for a continuous electrochemical refrig- mally insulated electrochemical system and separate heat exchang-
erator based on the Stirling cycle, validated the open-circuit voltage ers. With the Brayton cycle, all of the electrochemical cooling lowers
and overpotentials at various states of charge and current densities and the temperature of the electrolytes, and the heat absorption process
used the numerical model to predict cooling17. A Stirling-type cycle from the refrigerated space is accomplished with a conventional heat
inherently makes demonstration of electrochemical refrigeration exchanger, separately. Detailed analyses on the system dynamics of
difficult as the entropic heat absorption is split between the moving both the BECR and the Stirling Electrochemical Refrigerator will be
electrolyte, heat transfer across the electrochemical cell boundaries provided in two forthcoming manuscripts (A. Rajan et al., manuscript
and the thermal capacitance of the bespoke system. Recently, McKay in preparation). Together, these features allow the BECR cycle devel-
et al. demonstrated evidence of electrochemical cooling in a ther- oped in this work to achieve a peak cooling load of 0.934 W and a peak
mogalvanics architecture18. However, the steady-state cooling load coefficient of performance (COP) of 8.09. The maximum temperature
reported was ~0.12 mW, and the electrolyte was heated to 50 °C to drop obtained is ~0.15 K at peak cooling load, while the theoretical
reduce the activation overpotential. Furthermore, the maximum pos- maximum temperature drop is ~2–7 K, depending on the electrolyte.
sible temperature span of this incarnation is limited by the trade-off
between the higher limiting current density and the thermal inertia Concept of the BECR
afforded by a higher electrolyte flow-rate. The BECR scheme (Fig. 1b) employs two separated redox-active elec-
In this study, we demonstrate a technology that is the refrigera- trolytes in a closed system comprising two electrochemical cells, two
tion analogue of a TREC10–12,19 relying on the thermogalvanic effect, circulation pumps and two heat exchangers. Refrigeration is achieved
which utilizes the entropy change of redox reactions to generate by circulating the electrolytes through these components, where they
George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA, USA. 2Orca Sciences, Kirkland, WA, USA.
1
✉e-mail: shannon.yee@me.gatech.edu
n 1 4
tio e– B A
Cation flux
ec
cell cell
Isentropic reaction, ΔSrxn > 0
Cold side HX
e– A B
Cation flux
3
TC e– 2
3
QC
A′ B′
on
pti
absor TC
t
hea Colder
olar
Isom electrolytes
2
Entropy
Fig. 1 | Concept for the BECR. a, Temperature–entropy diagram for the BECR cycle showing the nature of the four constituent processes that convert the
electrolytes sequentially and cyclically from states 1 to 4. QH and QC are the rejected heat and the absorbed heat, respectively. The area enclosed within
the cycle is the electrical work input required to drive the cycle. b, Schematic showing the essential components operating in a continuous scheme. Two
redox-active electrolytes shown in purple (A + e− ⇌ A′) and yellow (B′ + e− ⇌ B) are pumped (shown using solid triangles) through four components, each
corresponding to one process shown in a and marked by the associated electrolyte states. An external power supply consumes power PIN and drives current
I through an adiabatic cold (electrochemical) cell, driving a net electrochemical reaction A + B → A′ + B′ with ∆Srxn > 0 isentropically. The entropic heat
required for the electrochemical reaction is absorbed from the lattice of the electrolytes, lowering the temperature of the electrolytes exiting the cold cell.
A cold side heat exchanger (HX) absorbs heat (represented by squiggly arrows) from a cold source TC devoid of any electrochemical reaction. The power
supply also drives the opposite electrochemical reaction, A′ + B′ → A + B with ∆Srxn < 0, in the adiabatic hot cell. In this cell, entropic heat liberation raises
the temperature of the electrolytes exiting the hot cell. The hot side HX rejects absorbed heat into a hot sink TH. Current continuity in the electrochemical
cells is preserved by an ion-exchange membrane that allows for an ionic current, shown to be cationic in this schematic.
are subject to the processes shown in the temperature–entropy dia- as ∆Srxn/nF = αCELL, where αCELL is the cell temperature coefficient,
gram in Fig. 1a. For tractability, Fig. 1a features the averaged state of n is the number of electrons transferred, and F is the Faraday con-
both electrolytes at a specific point in the scheme. The electrolytes stant. The cell temperature coefficient is the combination of con-
are driven from state 1 to state 2 in the cold electrochemical cell isen- tributions from both electrolytes, αCELL ≈ αC – αA, where subscripts
tropically using an electric current. The direction of this current is C and A denote the cathode and anode, respectively11. To develop
such that the entropy change of the electrochemical reaction is posi- the proof of concept, several half-cell reactions were first screened
tive (∆Srxn > 0), leading to a decrease in electrolyte temperature. The (Supplementary Table 3) to find appropriate half-cell reactions,
reduction (A + e− → A′; e−, electron) and oxidation (B → B′ + e−) reac- using a non-isothermal temperature coefficient measurement
tions both require entropic heat, which is sourced from the electrolyte technique (Supplementary Note 2). We ultimately chose the aque-
lattice, since the cell is thermally insulated. A membrane keeps the ous I− 3 /I (αI >
−
0) and ferricyanide/ferrocyanide (FCN3−/FCN4−,
two electrolytes separate and shuttles ions in response to the electric αFCN < 0) half-cell reactions due to the relatively high αCELL ≈ αI – αFCN
field created by the electron transfer between the half-cell reactions to value (~2 mV K–1) and the low crossover of both anionic
maintain charge neutrality. These ions must ideally each be a specta- redox-active species when combined with a cation-exchange mem-
tor ion that is introduced into the electrolyte as a counterion or via a brane, such as Nafion, as employed in this work. The state of charge
supporting electrolyte. The electrolytes then absorb heat from the cold (SOC)-dependent values of αI, αFCN and αCELL (measured using the
source at a temperature TC in the cold side heat exchanger, thereby isothermal temperature coefficient measurement technique shown
heating the electrolytes from state 2 to state 3. In the hot electro- in Supplementary Note 3) are shown in Fig. 2a. The SOC is defined
chemical cell, the driven electrochemical reaction is opposite to that as the normalized electrons per unit volume of electrolyte that is
of the cold electrochemical cell, that is, ∆Srxn < 0. In this case, both the available for electron transfer in the direction of the cold cell elec-
reduction (B + e− → B′) and oxidation (A → A′ + e−) reactions gener- trochemical reaction (ΔSrxn > 0). The maximum number of elec-
ate entropic heat, which is rejected into the electrolyte. Therefore, the trons per volume is determined by the solubility of the limiting
electrolytes heat up within the hot electrochemical cell, traversing reagent, in this case, potassium ferrocyanide at 0.6 M.
from state 3 to state 4. Finally, the electrolytes return to state 1 from While the temperature coefficient represents the entropic driv-
state 4 by rejecting heat to a hot sink at temperature TH in the hot side ing force that leads to cooling, the heat capacity of the electrolyte
exchanger. We detail the system dynamics of the BECR (A. Rajan et al., represents the thermal inertia. The volumetric heat capacities of the
manuscript in preparation) and Stirling Electrochemical Refrigerator electrolytes were measured at the 50% SOC configuration using a
systems, and their material level considerations, elsewhere21 and will calorimeter (Supplementary Note 4) to be 2.49 ± 0.38 MJ m−3 K−1
investigate them in future work. for aqueous I− 3 /I and 3.58 ± 0.78 MJ m K for aqueous FCN /
− −3 −1 3−
Selection and characterization of electrolyte refrigerants tions, since we do not expect the heat capacities of the electrolytes
The BECR efficiency and cooling load are governed by the entropy to change substantially due to the small intra-electrolyte species
change of the combined electrochemical reaction, which is given conversion occurring with electron transfer. The SOC-dependent
2
298
294
Direction of cold cell
–1 reaction
3I– + 2FCN3– I3– + 2FCN4–
∆Srxn > 0
292
–2
I3–/I–
I3–/I– FCN3–/FCN4– Cell FCN3–/FCN4–
–3 290
0 20 40 60 80 100 0 20 40 60 80 100
SOC (%) Final SOC (%)
Fig. 2 | Electrolyte characterization. a, Temperature coefficients of the individual electrolytes (non-isothermal measurement) and the full cell (isothermal
measurement) as a function of the SOC. We choose the direction of the reaction occurring in the cold electrochemical cell to be the SOC reference. The
dashed lines are linear fits that describe the temperature coefficients as a function of the SOC. b, The linear fit from a and the specific heat capacity of
the electrolyte are used to calculate the expected drop in electrolyte temperature as function of reaction extent along the isentropic path for a closed
electrochemical cell. The initial configuration starts at 90% SOC with the electrolytes in thermal equilibrium at 300 K. For this calculation, it is assumed
that the ion-exchange membrane has infinite thermal resistance. Details are provided in Supplementary Note 5.
temperature coefficients and heat capacities can be used to pre- and a graphite serpentine flow field to increase the residence time,
dict the maximum isentropic temperature difference as a function and therefore the degree of electrochemical conversion of the elec-
of reaction extent, as depicted in Fig. 2b, achievable in the absence trolyte. Electrolyte flow-rates were measured using optical tachom-
of thermodynamic irreversibilities (that is, inter-electrolyte and eters (Supplementary Note 7). Eight different electric current inputs
exo-cell heat transfer, and Joule heating), within a closed electro- (up to 4 A) were sourced with a galvanostat to the electrochemi-
chemical cell. Supplementary Note 5 details how Fig. 2b was gen- cal cells for 20 minutes to achieve a steady state (with 1 minute of
erated. In general, the temperature difference may be increased by open-circuit data measurement on either side of the sourced cur-
alternative half-cell reactions with larger temperature coefficients, rent domain; Fig. 3b). When the current is applied, the electrolyte
lower specific heat capacity and higher solubility of the limiting exiting the cold cell is driven to a lower SOC, decreasing the outlet
reagent. To obtain the maximum isentropic drop in temperature temperatures of both electrolytes using the entropy of the reactions.
for a given flow-rate, electrode volume and operating current in an Conversely, the electrochemical reaction in the hot cell increases the
open system, the electrochemical cell must adhere to the tenants of SOC (the direction of current I is reverse in the hot cell relative to
a plug flow reactor (that is, without any heat or mass dispersion)22. the cold cell), in turn increasing the outlet temperature over time.
In such an architecture, for a given pair of electrolytes, the isentropic Since the electrochemical cells are in series and the flow-rates of
drop in temperature increases with higher electrolyte residence time the electrolyte streams in both cells are equal, the SOC values of
and operating current. Additional thermodynamic cycle details, the electrolytes entering both cells remain invariant after the steady
the BECR figure of merit and the predicted performance will be state has been achieved (when the voltages in the cells, VHOT and
addressed in forthcoming theoretical work (A. Rajan et al., manu- VCOLD, saturate). The cells reach the thermal steady state (dTi/dt = 0;
script in preparation). Additionally, we confirmed that the I− 3 /I //
−
Ti, temperature; t, time; subscript i, any of the electrolyte outlet tem-
Nafion//FCN /FCN system possessed the required lifetime for
3− 4−
peratures like C2 or H4, as seen in Fig. 3a) at ~7 minutes and elec-
the proof-of-concept experiment by cycling it at ~16 mA cm–2 within tric steady state (dVi/dt = 0; Vi, voltage; subscript i, HOT or COLD)
a potential window far away from water-splitting regimes (between at ~2 minutes. At the thermal steady state, the difference between
–0.6 V and 0.6 V) for 100 cycles over 200 hours (Supplementary Note the electrolyte outlet and inlet temperatures is equal to the net heat
6). For perspective, we operate the proof-of-concept refrigerator for generation within the corresponding electrode. We used the differ-
only 50 cycles and 20 hours, observing steady, continuous cooling. ence of the values of this temperature differential at the steady state
(t = 21 minutes) and at the onset of the current (t = 1 minute) as the
BECR proof of concept steady-state temperature response of the electrolyte to the heat pro-
The proof-of-concept experiment (Fig. 3a) and the schematic in Fig. cess occurring in the electrode (Fig. 3c).
1b are alike, save for the heat exchangers. In the proof-of-concept The steady-state temperature response was used to evaluate
prototype, one heat exchanger regulates the temperature of each the electrolyte-specific cooling load (Fig. 4a,b) and rejected heat
half-cell electrolyte at the inlets of the hot and cold cells to 27 °C (Fig. 4c,d) using the heat capacity and flow-rate of the electrolytes.
(that is, TH = TC = 27 °C), thereby, maintaining constant inlet tem- The FCN3−/FCN4− electrolyte outperforms its counterpart because
peratures to clearly demonstrate that cooling is occurring from |αFCN| > |αI| and because it has a lower overpotential at 4 A, where
only the electrochemical reactions and is not an artefact of ambi- the entropic cooling effects of the triiodide reduction are negated
ent (room) temperature fluctuations. The electrochemical cells have by Joule heating. The larger Joule heating associated with the I− 3 /I
−
acrylic housing to increase thermal resistance to the surroundings reaction arises from its higher activation overpotential, attributed to
a b
TC1 Galvanostat
Peristaltic TC3 a
pump Peristaltic
pump
Cold cell
c
TH1 d
TH3
VCOLD = Va – Vb
VHOT = Vc – Vd
Hot cell
TH2
TH4 TE HX
TE HX
Hot cell
c d
0.2 0.1
TC1 TC2 TC3 TC4 ΔT21 ΔT43
0
0
Ti(t) – Ti(t = 0) (K)
–0.1
1.5 A 1A
–0.2 –0.2
ΔTij(t) – ΔTij(t = 0) (K)
0.5 0.1
ΔT21 ΔT43
0
0
–0.1
1.5 A TH1 TH2 TH3 TH4 2.5 A
–0.5 –0.2
0.3 0.1
ΔT21 ΔT43
Voltage (V)
0
0.1
–0.1
VCOLD VHOT
1.5 A 3.5 A
–0.1 –0.2
0 2 4 6 8 10 12 14 16 18 20 22 0 2 4 6 8 10 12 14 16 18 20 22
Time, t (min) Time, t (min)
Fig. 3 | BECR proof of concept. a,b, Schematics showing the thermo-hydraulic (a) and electrical (b) connections of the experimental set-up. Thermoelectric
(TE) modules regulate the temperature of the electrolytes at both electrochemical cells. The electrolyte temperatures are monitored at the points marked
by the variable resistor symbol. These temperatures, T, are labelled as C or H for cold or hot cell, and numbered 1 through 4 for I3−/I− electrolyte inlet,
I3−/I− electrolyte outlet, FCN3−/FCN4− electrolyte inlet, and FCN3−/FCN4− electrolyte outlet respectively. Voltage across the cold cell VCOLD (and hot cell
VHOT) is given by the potential difference at points a and b (and c and d) c, Representative data set for 1.5 A showing the time variation of the electrolyte
temperatures at the inlets and outlets of the electrochemical cells and the cell voltages. d, Time variation of the cold cell outlet and inlet temperature
differential for three representative current values. Here, ∆Tij is the difference between the cold cell electrolyte outlet and inlet temperatures and is defined
as TCi – TCj. The dark grey regions are when no current is sourced, and the indicated current is sourced in the light grey domain. In b and c, the lines and
associated shaded regions represent the mean value and standard deviation of the indicated parameter over six repetitions.
the large solvent reorganization energy of the covalent bond cleavage thermodynamic and kinetic origins. Even though the cells operate
that accompanies the triiodide reduction23. The electrolyte-specific in the ohmic regime (Supplementary Fig. 16) at the higher current
cooling loads were combined to obtain the total cooling load, Q̇C values, the Joule heating scales with I2, increasing the irreversibility.
(Fig. 4e). The COP, β, was evaluated using β = Q̇C /PIN (Fig. 4e), Furthermore, T2 and T4 (the average temperature of the electrolytes
where PIN = I(VHOT + VCOLD). We choose to neglect the circulating exiting the cold cell and hot cell respectively, as shown in Fig. 1a)
pumping power; an estimate of the pumping power based on Darcy’s decrease and increase respectively with increasing current. This leads
law is provided in Supplementary Note 8. The BECR can achieve a to heat transfer processes that are increasingly deviant from the iso-
target cooling capacity by achieving a high degree of electrochemical thermal heat transfer characteristic of the ideal Carnot cycle. The
conversion at low electrolyte flow-rate. Therefore, unlike flow batter- measured cooling load and COP (evaluated above) are in good agree-
ies, the BECR is not bound by the trade-off between high pumping ment with the predicted values, calculated using an energy balance.
power and high porous electrode-specific surface area. The precip- Finally, we emphasize that while this proof-of-concept dem-
itous drop in COP with increasing current (shown in Fig. 4e) has onstration of continuous electrochemical refrigeration is made
Cooling
COP, β
1.2 8
FCN3–/FCN4–
6 0.6
0.9 6
0.6 4
0.3 2 3 0.3
Cooling
0 0
High Joule heat FCN3–/FCN4–
–0.3 –2 0 0
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
Operating current, / (A) Operating current, / (A)
Operating current, / (A)
current. Blue regions indicate a positive cooling load where the entropic heat absorption dominates the Joule heat and the electrolytes are capable of
cooling. Red regions indicate a negative cooling load due to excessive Joule heating. c,d, Electrolyte-specific rejected heat for the I−
3 /I (c) and FCN /
− 3−
FCN4− (d) electrolytes as a function of the operating current. e, Measured and predicted values of the COP and total cooling load as a function of the
operating current. Markers and error bars (visible if bigger than the marker) represent the mean value and standard deviation of the quantity across six
repetitions. The lines and shaded bounds of the predicted COP (too small to be visible) and cooling load represent the same. Dashed lines serve as guides
for the measured values.
possible by appreciable electrolyte redox conversion, substantial the technologies (Fig. 5b). Through this lens, the entropic driving
improvement can still be made by increasing the residence time force for electrochemical refrigeration is more than double that of
of the electrolytes. The first Damköhler number, given by the ratio vapour compression.
of the reaction rate and the advective molar transport rate, for this Table 1 details the device COP and temperature drop achieved
work is ~0.08, which amounts to ~8% conversion for a plug flow by experimental works corresponding to the previously mentioned
reaction with first-order kinetics22. There is substantial scope for refrigeration technologies. We chose popular refrigerants used in
improvement by operating at higher currents (increased rate of other refrigeration technologies in Fig. 5 and Table 1 simply to pro-
reaction) or decreasing the volumetric flow-rate of the electrolytes vide context to our work. Research within each of these technolo-
(decreased molar influx). Of course, the latter must be done keep- gies may have led to other refrigerants that demonstrate superior
ing in mind that decreasing the electrolyte flow-rate to arbitrarily performance. For example, a rotary magnetocaloric refrigerator
low values will decrease the cooling load, and hence there exists an that employs LaFeSiH as the refrigerant has shown a temperature
optimum electrolyte flow-rate for a given current that provides the increase of 11 K and a COP of 1.9 (ref. 26). Table 1 shows that sub-
maximum electrolyte-specific cooling load. As future motivation, stantial work must be done to improve the temperature drop of
Kim et al. reported a twofold increase in the temperature coefficient our technology, while maintaining the COP, to improve its com-
of FCN3−/FCN4− by employing a 15 wt% methanol–water solution mercial viability. We believe that this can be done through addi-
as the solvent24. Yamada et al. increased the temperature coefficient tional research into high-temperature-coefficient half-cell reactions
of I−
3 /I to 2 mV K using a supramolecular scheme . These works
− –1 25
with high solubility, low-freezing-point solvents with low specific
could be immediately adopted into the BECR cycle to improve the heat capacities and electrochemical cell architectures that can drive
cooling load several fold. isentropic reactions to completion. Higher-temperature-coefficient
half-cell reactions with high solubility have larger entropy changes
Discussion and conclusions per unit volume of electrolyte, and thereby increase an electro-
We have compared the inverse gravimetric heat capacities and the lyte’s capacity for cooling, in addition to delaying the onset of
gravimetric maximum possible entropy change of existing refrig- temperature-induced precipitation. Low-vapour-pressure solvents
erants with those of our electrochemical refrigerants (Fig. 5a). (for example, ionic liquids) and low-freezing-point solvents (for
The maximum possible entropy change and the specific heat rep- example, alcohols) allow for the BECR cycle to operate in a larger
resent the entropic driving force and the thermal inertia towards parameter space (operating current and electrolyte flow-rate) with
cooling. With existing aqueous redox chemistries, we found that little threat of excessive internal pressurization and/or the electro-
electrochemical refrigeration compares well against vapour com- lyte freezing. Lower electrolyte-specific heat elicits a greater temper-
pression, suggesting that it could someday be an appreciable ature response per unit electrochemical reaction. Electrochemical
zero-global-warming-potential refrigeration technology. A more cell architectures that drive the reactions towards completion
compelling argument for electrochemical refrigeration can be made (higher first Damköhler number) isentropically (minimal over-
upon comparing the entropy change per unit carrier, ∆SCARRIER, for potential) can generate higher electrolyte temperature drops. The
a b
102 80
Solid to Solute to
gas solute
MC (Gd5Si2Ge2) Electron to
electron
60
101
Liquid solution
TE (Bi2Te3) to gas
ΔSCARRIER /R
Alternative
C–1 (kg K kJ–1)
solvents
40 Liquid to
MH (LaNi5)
gas
Solid to
VC (R140a) solid
100
20
Alternative
redox couples
EC
(Fe2+/3+//FCN3–/4–) Ab
(NH3 – water)
10–1 0
0 0.5 1.0 1.5 2.0 MC TE VC MH Ab EC
–1 –1
ΔS (kJ kg K )
Fig. 5 | Comparison of refrigeration technologies. a, Electrochemical refrigeration (EC) compared against popular refrigerants used in magnetocaloric
(MC), thermoelectric (TE), vapour compression (VC), metal hydride (MH) and absorption (Ab) refrigeration along the axes of the maximum possible
gravimetric entropy change (ΔS) and the inverse of the gravimetric specific heat capacity (C−1). b, Comparison of the entropy per unit carrier normalized
by the universal gas constant, R, for the same technologies along with the corresponding physical nature of the transformations. Data are compiled from
references listed herein33–38,27 and are summarized in Supplementary Table 4.
Table 1 | Comparison of device COP and temperature drop of some refrigeration technologies
Technology Refrigerant COP Temperature drop (K) Comment Reference
Electrochemical I3 /I //FCN /FCN
− − 3− 4−
8.09 0.07 Peak COP This work
3 /I //FCN /FCN
I− 0.93 0.15 Peak cooling load
− 3− 4−
Vapour compression R410a 3.1 6–8 Heating, ventilation and air conditioning 27
R134a 3.5 10–40 Refrigeration 28
Magnetocaloric Gadolinium 0.8 15.4 29
Thermoelectric Bismuth telluride 0.72 14 30
Absorption Ammonia–water 0.45 27 31
Metal hydride Lanthanum nickel 0.26 30 32
operating current and electrolyte flow-rate influence the COP and reagent grade and were used without purification. MilliQ grade (19.7 MΩ cm)
cooling load, and they must be rationally optimized according to the ultrapure deionized water was used to prepare all the solutions. Non-aqueous
thermodynamic, kinetic and physical properties of the electrolytes. solvents were used without drying.
In conclusion, we have demonstrated a continuous electro-
chemical refrigerator that is inspired by the reverse Brayton cycle. It Electrochemical cells. The electrochemical cells were designed to maintain
an adiabatic environment and increase the residence time of the electrolytes
achieves a peak COP of 8.09 with a temperature drop of 0.07 K, and within the cell. Some 0.5-inch-thick acrylic endplates and a serpentine channel
a peak cooling load of 0.934 W (at 14.6 mW cm–2, normalized by the made of resin-impregnated graphite (Graphite Store, MW001204) were used.
cross-sectional area of the porous electrodes). We acknowledge that Supplementary Fig. 13 shows an exploded view of the cold and hot electrochemical
only a modest temperature drop of 0.15 K at peak cooling load was cells. The acrylic housing sandwiches the serpentine, porous graphite electrodes
observed (Fig. 3c), but theoretically it could have achieved a tem- (AvCarb, G600A) and cation-exchange membrane (Nafion 211). Eight M6 bolts
were used to hold the assembly together. Electrolytes were introduced into the
perature drop of 2–7 K (Fig. 2b) if we were able to achieve a greater cells using polyvinylidene fluoride (PVDF) adaptors (McMaster-Carr, 5533K411).
degree of electrochemical conversion. The system demonstrates The current collectors were thin copper strips that were placed in between the
cooling for a total of ~18 hours without any sign of deterioration of acrylic housing and serpentine graphite channel. The size of the porous electrodes
the system performance. (and therefore the rest of the assembly) was chosen using a conservative estimate.
A simple energy balance affords the following equation for small temperature
changes: V̇CV ΔT = IαT∞ − I2 R. Here, we assume that the majority of the
Methods overpotential occurs due to the ohmic overpotential from the membrane
Materials. All chemicals were purchased from commercial suppliers resistance. We use some conservative values for the flow-rate V̇ (1 ml s–1), heat
(Sigma-Aldrich, VWR International, The Science Company), were of at least capacity CV (3.5 MJ m–3 K–1), temperature coefficient α (1.5 mV K–1) and membrane
Heat exchangers. We opted for an active heat exchanger (Supplementary Fig. 14) Calculation of equilibrium-time cell resistance. The equilibrium-time cell
rather than a passive heat exchanger to eliminate ambient temperature fluctuation resistance values during the proof-of-concept experiment were calculated using the
and maintain a constant electrochemical inlet cell temperature; this allows us expression below, where i denotes the values corresponding to the hot or cold cell:
to unequivocally demonstrate electrochemical cooling decoupled from ambient
temperature fluctuations of the heat exchangers. The active heat exchange was Vi (t = 21 minutes) − Vi (t = 1 min)
Ri =
performed using a thermoelectric module attached to a cold plate, which had I
a tube for the electrolyte to flow. Since the electrolytes must always be kept
from mixing with each other, we used two active heat exchanger modules (TE Calculation of measured and predicted cooling loads. The measured total
Technology, CP-061HT). We embedded a 1/8-inch-outer-diameter stainless-steel cooling load was calculated by the sum of the electrolyte-specific products of the
tube (chosen for its corrosion resistance to the triiodide/iodide and ferricyanide/ volumetric specific heat (CV,e), volumetric flow-rate (V̇e) and equilibrium-time
ferrocyanide redox couples) into two 1/4-inch-thick copper blocks, which were drop in temperature (ΔTij):
mounted to the thermoelectric heat exchanger. A 1/4-inch-thick neoprene ∑
Q̇C,measured = V̇e CV,e ΔTij
rubber was use to insulate the thermoelectric module and copper block from the
e
environment. The parts of the stainless-steel tubing that were in contact with the
ambient air were liberally covered in a polymer as insulation. The temperature Subscript e denotes electrolyte and subscripts i and j denote the RTDs located
was set and regulated by sourcing power to the thermoelectric modules using a at the electrolyte outlet and inlet on the cold cell, respectively. The predicted total
programmable temperature controller (SRS, PTC10) that employed two PTC440 cooling load was calculated using an energy balance:
TEC drivers. The temperature feedback to the closed proportional–integral–
derivative loop was provided by thermistors (OMEGA, item no. 44031) that Q̇C,predicted = IαT − I2 RCOLD
monitored the temperature of the electrolytes exiting the heat exchangers. The
programmable temperature controller used a PTC320 card to acquire inputs from T̄ is the volume-averaged temperature within the cold cell and RCOLD is the
the thermistors located at the outlets. resistance of the cold cell at steady-state. Since the maximum drop in temperature
in this work is ~0.2 K, we assume that the cell is isothermal at 27 °C.
Porous electrode and ion-exchange membrane preparation. Prior to all
experiments, the porous electrodes were immersed in an acid bath containing Non-isothermal temperature coefficient measurements. Half-cell reactions
three parts 68% nitric acid and one part 18 M sulfuric acid. They were heated to were screened using a non-isothermal temperature coefficient measurement.
80 °C and refluxed overnight. The electrodes were washed in deionized water until To contain a variety of redox-active couples and solvent combinations, the cell
pH neutral and then heated in an oven at 400 °C for three hours. The electrodes housing was made of a 7-cm-long polytetrafluoroethylene (PTFE) cylinder with
were assembled into the electrochemical cells and flooded with electrolytes as a 1.5 cm bore across its length, to house the electrolyte solution. The electrodes
soon as possible. The as-purchased Nafion membrane was first heated in a 2 M were made of resin-impregnated graphite (Graphite Store, MW001204). After the
KOH solution at 60 °C overnight. It was then washed thoroughly with deionized electrolyte was loaded into the cell housing, the PTFE annulus was sandwiched by
water and stored in a 4 M KCl solution. After each experiment, the membrane was the graphite electrodes using neoprene gaskets and two M16 bolts that thread into
washed thoroughly with deionized water and stored in the KCl solution until the the bottom electrodes. Fibre glass washers were used to keep the graphite electrode
next experiment. electrically isolated via zinc-plated bolts. The assembled cell was then sandwiched
between two thermoelectric modules (TE Technology, CP-061HT) using a lead
Electrolytes for proof-of-concept experiment. Of all the redox-active species screw mechanism that minimizes interfacial thermal resistance by applying
employed in the proof-of-concept experiment, the species with the limiting pressure. When the experiment is started, the programmed thermoelectric
solubility was the ferrocyanide with a room temperature solubility of ~0.6 M. modules apply varying thermal gradients, as shown in Supplementary Fig. 3. The
Therefore, with reference to the cold electrochemical cell reaction, the cell is at top thermoelectric module is always maintained at a higher temperature than the
100% SOC when the ferrocyanide ion is at that value. Conversely it is at 0% SOC bottom one. The temperatures of the electrodes are continuously monitored using
when all of the ferrocyanide has been converted to ferricyanide at 0.6 M. The K-type thermocouples, and the voltage across the two electrodes is measured using
quantity of the iodide ions and triiodide ions are determined using the solubility a data acquisition unit (Agilent, 34970A). The hot electrode and cold electrode
constraint of the ferrocyanide species and by the requirement, albeit arbitrary, that are always connected to the positive and negative terminals, respectively. The
at 50% SOC, both the electrolytes have the same concentration of potassium ions. non-isothermal temperature coefficient is evaluated by the slope of the voltage
Potassium ions were used for the iodide, ferricyanide and ferrocyanide species. The change and the difference in temperature between the two electrodes. For these
triiodide ion is achieved by dissolving molecular iodine into an iodide solution, measurements, electrolytes were prepared with equal concentrations of the reduced
which causes the iodine to abstract an iodide to form the triiodide. This reaction and oxidized species.
has an equilibrium constant that strongly favours the triiodide ion, and therefore
it is assumed that no iodine is present in the solution. The proof-of-concept Isothermal temperature coefficient measurements. Once two promising half-cell
experiment was primed with electrolyte at 80% SOC, and the details are provided reactions with the appropriate temperature coefficients were shortlisted, they
in Supplementary Table 2. were assembled in an electrochemical cell with an ion-exchange membrane of
choice (Supplementary Fig. 4). This was done for two reasons: (1) to confirm that
Proof-of-concept set-up. The tubing used to connect the components together the temperature coefficients added up in a complete electrochemical cell set-up
had a 1/4 inch outer diameter and 1/8 inch inner diameter. The tubing used to and that there were no mechanisms that diminished the entropy of the half-cell
contain the ferricyanide/ferrocyanide electrolyte was a flexible polyvinyl chloride reactions, and (2) to evaluate the lifetime of the electrochemical system under
tubing, and rigid perfluoroalkoxy tubing was used to contain the triiodide/iodide varying temperature under the open-circuit condition. To test a variety of redox
electrolyte due to the electrolyte’s tendency to engage in SN2 nucleophilic attack with systems, the tenants of maximum inertness were maintained. The cell housing
non-fluorinated polymers. All adaptors and fittings were made of PVDF. At the was made of PTFE, and 2 cm through-holes were bored into the middle to house
inlet and outlet electrolyte ports of the electrochemical cells, resistive temperature the porous electrodes. Carbon felt (AvCarb, G600A) was used as the cathode and
devices (RTDs) were introduced into the electrolyte streams using PVDF tee anode material. The cathode and anode sandwiched the ion-exchange membrane
adaptors. Our previous experiments showed that it was necessary for these RTDs of choice. The current collectors were made of resin-impregnated graphite
to have a metal shield to prevent capacitive coupling to the voltage applied by the (Graphite Store, MW001204) and zinc-plated screws were screwed into them
galvanostat. In addition to the components shown in Fig. 3a, we used two sealed to facilitate robust electrical connections. The cell housing was sealed using M6
Erlenmeyer flasks for each electrolyte stream as a reservoir source (and sink) for bolts and neoprene gaskets. The electrolytes were first introduced into the cell
the inlets (and outlets) of the electrochemical cells, which served three purposes: using syringes and then air-tight sealed using redundant stop cocks and one-way
(1) they acted as gas traps that removed any air bubbles from the system, (2) they valves. This restricted any solvent motion due to osmotic pressure. After the
allowed one pump to generate two flow streams and (3) by placing the outlet (or electrochemical cell was assembled and checked for any leakage, it was placed
inlet) tubing above (or below) the liquid level, they ensured the flow of electrolytes in between the same two thermoelectric modules as described in the previous
in only one direction (eliminating any possibility of unintended back flow). K-type section. This time, however, the modules were programmed to establish the same
thermocouples were placed close to the electrochemical cells to provide evidence temperature. The temperature feedbacks that informed the proportional–integral–
that the temperature changes caused by sourcing the electrochemical current are derivative controller were disconnected from the modules and firmly affixed on
not correlated to changes in environmental temperature. A photograph of the the cathode and anode current collectors. Thermocouples were used to measure
completed set-up is shown in Supplementary Fig. 15. Peristaltic pumps (Control the temperature of both current collectors, and the open-circuit voltage of the
Company, 3385) circulated the electrolyte. The hot and cold electrochemical cells cell was monitored continuously. Keeping in line with convention, the half-cell
were connected electrically in series to the galvanostat (Biologic, VSP-300). Eight reaction with the positive temperature coefficient was connected to the positive
Electrolyte temperature as a function of reaction extent. For an arbitrary Using a pore radius of 10 μm, porosity of 0.7 and Kozeny–Carman constant
electrochemical system, the first law of thermodynamics may be written as the of 5.55, the above permeability is calculated to be 3.92 × 10–10 m2. This assumes
following expression: that the electrolyte takes the convoluted path of the serpentine graphite channels,
( ) where each serpentine flow field has a total of 20 channels. The channels are 3 mm
∑ ∑ ∑ ∑ wide and 3 mm deep. We conservatively estimate that the entire depth is occupied
dH = d Ni H̄i = H̄i dNi + Ni dH̄i = TdS + μi dNi
by the porous electrode due to the compressive stress put into the electrochemical
i i i i
cell assembly. In reality, there will be parallel pathways that reduce the pumping
demand, for example, the serpentine graphite channel area unobstructed by the
In the above expression, H is the total enthalpy, N is the number of particles, H̄
electrode or bypass electrolyte routes within the porous electrode that have very
is the partial molar enthalpy, T is temperature, S is entropy, μ is chemical potential
short path lengths from the inlet to the outlet. The volumetric flow-rate of the
and subscript i(represents
) all species participating in the electrochemical system. electrolyte is taken to be 1 ml s–1, and the dynamic viscosity of the electrolyte is
Recalling that ∂H̄i /∂T P = C̄P,i, where C̄P,i is the partial molar heat capacity at
taken to be 8.9 × 10–4 Pa s. Using these parameters, the pumping power required to
constant pressure P, and employing the concept of an average heat capacity at
drive the electrolyte through one porous electrode is estimated to be 0.4 W. Since
constant pressure P, CP, the above expression may be written as the following:
one pump drives the electrolyte through two porous electrodes in parallel, the total
∑ ∑
TdS = CP dT + H̄i dNi − μi dNi power consumed by one pump is 0.2 W. We use two pumps to drive electrolyte
i i
through two sets of porous electrodes, each in parallel; therefore, we estimate the
entire pumping power to be 0.4 W.
Finally, we recognize that by fixing the initial and final states of the reaction, we
can define a reaction coordinate. One point on the reaction coordinate will then Uncertainty and error propagation. The electrolyte-specific cooling load QC,e at
map to relative concentrations of all species in the electrochemical system. We also thermal equilibrium (acquired at time teq) is given by the following expression:
recognize that the changes in the partial molar enthalpies and chemical potential
are simply the enthalpy and Gibbs free energy of the reaction. The above equation QC,e = V̇e CV,e ΔTij (teq )
can now be written as the following for the special case of an isentropic reaction:
where V̇ is volumetric flow-rate, ∆Tij = Ti − Tj is the difference between the
CP electrolyte inlet and outlet temperatures of the cold electrochemical cell and
dS = dT + ΔSrxn (ξ) dξ = 0
T subscript e represents the electrolyte. The uncertainty in the electrolyte-specific
cooling load is then given by the following expression:
The reaction coordinate ξ is interchangeable with the SOC. Finally, the linear √(
fit from the trends in Fig. 2a and specific heat capacities may be used to integrate )2 ( ) ( )
δQC,e δ V̇e δCV,e 2 δΔTij (teq ) 2
both sides of the above equation to find the final temperature as a function of = + +
reaction extent. QC,e V̇e CV,e ΔTij (teq )