High-energy and low-cost membrane-free chlorine flow battery

Grid-scale energy storage is essential for reliable electricity transmission and renewable energy integration. Redox flow batteries (RFB) provide affordable and scalable solutions for stationary energy storage. However, most of the current RFB chemistries are based on expensive transition metal ions or synthetic organics. Here, we report a reversible chlorine redox flow battery starting from the electrolysis of aqueous NaCl electrolyte and the as-produced Cl2 is extracted and stored in the carbon tetrachloride (CCl4) or mineral spirit flow. The immiscibility between the CCl4 or mineral spirit and NaCl electrolyte enables a membrane-free design with an energy efficiency of >91% at 10 mA/cm2 and an energy density of 125.7 Wh/L. The chlorine flow battery can meet the stringent price and reliability target for stationary energy storage with the inherently low-cost active materials (~$5/kWh) and the highly reversible Cl2/Cl− redox reaction.

which R is the radius of liquid particle, L is the cell height, mp is particle mass, is density, c is concentration, v is volume of the particle. X<<Y*Re (Re=Reynold's number), thus the relative motion between the two fluids is negligible 1 . The biphasic flow was treated as a phase consisting of NaCl/H2O and CCl4 with a specific volume ratio. To determine the physical properties of the pseudo-single phase in the porous electrode, saturation (Sj) was introduced for phase j (j=NaCl/H2O or CCl4). The Sj is the volumetric ratio of phase j in the void volume of the porous electrode, which was measured by displacement experiment upon flow (see next section). The Qj is the volumetric flow rate of phase j in the cell, and the flow rate vj is defined as follows: The flow rate of the pseudo-single phase is: The overall concentration C of species i ( i= Cl2) in the pseudo-single phase in the porous electrode is: The effective diffusivity of Cl2 in the pseudo-single phase is corrected as the harmonic average of that in the aqueous and organic phase, considering disordered mixture 2 : Since Cland Na + are only present in the aqueous phase, the diffusivity and conductivity will only be discussed as in the aqueous phase. The diffusivity and conductivity of Cl -, Na +, and Cl2 were modified using Buggerman correlation 3 to account for the tortuosity of the porous electrode.
Due to the significantly larger size of the reservoir than the flow cell, the steady-state model was developed at the designated state of charge. Nernst-Plank equation was solved in both the porous electrode and NaCl/H2O to account for the species distribution. The porous electrode was treated as the superimposition of the mixed liquid phase and the carbon network as in porous electrode theory 4,5 . Since the reaction only occurs at the NaCl/H2O-carbon interface, the reaction area is corrected by the fractional wettability Xj of NaCl/H2O in the electrode. In the organic phase, only Cl2 is soluble, and the transport of Cl2 obeys Fick's law.
The negative electrode is grounded, the cell potential equals the sum of the equilibrium potential of the positive electrode and the overpotential of both electrodes. The equilibrium potential of the positive electrode is determined by the Nernst equation and the local concentration of Cland Cl2 in the aqueous phase.
The overpotential is related to the local current density by Bulter-Volmer kinetics: The total current density on the electrode: There are two charge carriers: electrons and ions in the porous electrode, the electrical and ionic currents are denoted as ie and iion, respectively. The corresponding potentials are ∅ and ∅ .

The volume percentage of CCl4 and NaCl/H2O in the porous electrode
The wettabilities of the porous electrode to NaCl/H2O and CCl4 was characterized by displacement experiment: A dry porous carbon electrode was weighted (M0), and then we allowed a NaCl/H2O stream flow through the electrode and obtained the weight of the NaCl/H2O soaked electrode (M1). Afterward, we allowed a stream of carbon tetrachloride to flow through the NaCl/H2O soaked electrode for liquid displacement, then weighed the electrode and got the value M2. This weight should equal the summation of M0 add Ma1 (remained aqueous solution after CCl4 flow through the electrode) add MCCl4 (the weight of CCl4 into the pores of electrode). The Ma2 is the weight of the aqueous solution replaced by the CCl4. Here we assume the volume of the aqueous solution replaced by the carbon tetrachloride is equal. We can calculate that the Ma1=9.3mg and 0.007cm 3 MCCl4= 21.7 mg and 0.0137 cm 3 . So, the volume ratio of the aqueous solution to the carbon tetrachloride during the flowing process is 0.007/0.0137=0.51. The volume percent of the aqueous solution in the porous electrode is 33.8%, the carbon tetrachloride is 66.2%.        The peak at 540 cm -1 in Fig. S5C-d is assigned to be chlorine 11 , and the peak at 500 cm -1 and 540 cm -1 in Fig. S5C-b are also assigned to chlorine 12 , which indicates that trace chlorine dissolves in the NaCl aqueous solution although the solubility is very low. However, the Raman spectrum (Fig.S5C-c) of NaCl aqueous solution after 60 minutes of discharging process does not show an obvious Cl2 peak, which indicates CCl4 stops the dissolution of Cl2 into NaCl/H2O.

Fig. S6. Difference of solvation energy of Cl2 in (A) CCl4 and (B) saturated NaCl solution calculated by Ab initio Molecule Dynamic.
The blue balls represent the Cl atom in Cl2, and the green, brown, red, white, and yellow balls represent the Cl, C, O, H, and Na atoms, respectively. The result indicates that Cl2 in CCl4 does not spontaneously transfer to NaCl solution due to a positive Gibbs free energy. The reason is due to the unfavorable displacement of Clfrom Na + when Cl2 enters NaCl solution. The AIMD calculation was performed with the VASP package [13][14][15] . The ion-electron interaction is described with the Projector Augmented Wave (PAW) method, and the exchange-correlation energy is described by the functional of the Perdew-Burke-Ernzerhof (PBE) form of the generalized gradient approximation (GGA) 16 -18 . Plane wave energy cut-off of 350 eV is chosen, and a minimal Г-centered 1 x 1 x 1 k-point grid is used. All molecular dynamics simulations were performed in the NVT ensemble using a Nosé−Hoover thermostat. Each system was heated to 300 K, equilibrated for 10.0 ps, and then simulated for 10.0 ps to get the average free energy. The visualization of the structures is made by using VESTA software 19 .      Fig. S10B shows the rate performance of carbon-coated NaTi2(PO4)3. Even at the rate of 315 C, the specific capacity can remain 80 mAh g -1 (65% at 9C).    The Zn//Cl2-CCl4 flow battery also delivers a good electrochemical performance. Its operation voltage is around 1.9V at 50 mA/cm 2 , higher than the NaTi2(PO4)3||Cl2 battery system (around 1.676V at the same current density). According to the cost evaluations for all-vanadium flow batteries, the active materials take up 37% of the total cost, and the membranes take up ~30% of the total cost, electrolyte storage takes up 8%, the pump and heat exchange takes up 3%, and other operational cost takes up 21% 35 . Thus, only active material and membrane costs were compared here as other costs are unavailable for most systems. The cost/kWh includes only the electrolyte and active materials used for cell assembly. The prices for mineral commodities: NaCl (~$40/ton), Zn ($2090.34/ton), I2 ($18500-$32000/ton), Br2 ($4400/ton), Li metal ($121600/ton), V2O5 ($5568-$9300/ton) were collected from National Minerals Information Center 36 . The prices for other commodities: ZnI2 ($5-10/kg), ZnBr2 ($3-5/kg), CCl4 ($1000/ton) LiFePO4 ($8100/ton) were collected from Alibaba. The price of NH4H2PO4 ($300/ton), TiO2 ($1000/ton), Na2CO3 ($200/ton) for NaTi2(PO4)3 anode were also collected from Alibaba 37 . The prices for organic redox couples were projectiles for mass production provided by the corresponding references. The unit cost of Nafion membranes 3838 and LATP electrolyte 39 were listed separately.

Supplementary Notes 2 Evaluation of chlorine permeation from the battery system
The cell is a closed system with the Cl2-CCl4 in a glass container assembled under atmospheric pressure. Therefore, the leakage could potentially occur at the seal through either chemical corrosion or permeation driven by the pressure difference across the sealing gasket. The sealing gasket for the bottle and the tubing for Cl2-CCl4 flow consists of a Viton fluoroelastomer. The Viton fluoroelastomer is resistant to chlorine corrosion. It is also resistant to swelling by halogenated solvents, demonstrating less than 10% volume expansion 40 in a wide temperature range between -20 o C to 200 o C. Thus, we only consider the leakage through Cl2 permeation.
Cl2 permeation rate was calculated with equation S2 41 , which depends on the difference of Cl2 partial pressure ( 2 ) inside and outside the gasket. The partial pressure of Cl2 for Cl2-CCl4 at different concentrations can be calculated with Raoult's law for the binary mixture. The 2 inside the container is 111.22 kPa when concentration of Cl2 is 0.184 mol/mol CCl4 (saturation) at 20 o C . At 50 o C, the 2 of saturated Cl2-CCl4 increases to 116.0 kPa (0.088 mol Cl2/mol CCl4) ( Table S6). The 2 outside the container is assumed to be 0 kPa. The permeation flux of Cl2 through the Viton gasket (F) into the atmosphere calculated through equation S2 is between 1.24 to 19.03 mL/day (3.72-57.1 mg/day) at 20 o C and 2.08 to 30.60 mL/day (5.1-75 mg/day) at 50 o C (The variations are caused by the different permeation coefficients reported for different fluoroelastomers 42 ). At these leakage rates, it requires 0.5 to 20 days at 20 o C or 0.3 to 15 days at 50 o C to reach the permissible exposure limit (PEL) of chlorine ( 3mg/ m 3 43 ) in an unventilated 1 m 2 * 2 m storage space (around the size of a fume hood). Thus, it is very unlikely to be exposed to chlorine higher than the PEL with the appropriate gaskets, chlorine sensor, and in a ventilated environment.  45 2 = 2 × 2 = 116.0 ( 2 ≤ 0.088) F = K A(p Cl2,in − p Cl2,out ) d F= permeation flux of the gas (mL/day) ; K= permeation coefficient=60-1000 (20 o C), 100 -2000 (50 o C) (mL/100 in 2 /day/atm 42 ) ; A= gasket surface area= (1.5 cm*1.5 cm-0.5 cm* 0.5 cm)*π*2 =12.56 cm 2 ; d= gasket thickness= 1 cm; p Cl2,in = partial pressure of Cl2 inside the container; p Cl2,out = partial pressure of Cl2 outside the container (S2)