Supporting information for:

Scalable and Continuous Water Deionization by

Shock Electrodialysis

Sven Schlumpberger,† Nancy B. Lu,† Matthew E. Suss,†,¶ and Martin Z.

Bazant∗,†,‡

†Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge,

Massachusetts 02139, United States
‡Department of Mathematics, Massachusetts Institute of Technology, Cambridge,
Massachusetts 02139, United States
¶Present Address: Faculty of Mechanical Engineering, Technion Israel Institute of
Technology, Technion City, Haifa 3200003, Israel

E-mail: bazant@mit.edu

1 Electroosmotic Flow Estimation

2 To estimate the magnitude of the electroosmotic flows in our system, we start with the
3 Helmholtz-Smoluchowski formula for the electroosmotic flow velocity

ζE
uEOF = (1)
µ

4 where  is the permittivity of water (F/m), ζ is the zeta potential (V), E is the electric
5 field (V/m), and µ is the viscosity of water (Pa*s). We estimate that E = j/σ, where j is
6 the current density (A/m2 ) and σ is the bulk conductivity of the solution (S/m). We also

S1
 7 convert the velocity into a flow rate by assuming that current flows only in the transverse
8 direction in our system (i.e. directly from anode to cathode), which is a crude approximation
9 of the true boundary layer structure of the shock, but might capture the correct scalings.
10 This means that we can simply multiply by the cross-sectional area A to obtain the flow rate

ζAj ζI
QEOF = = (2)
µσ µσ

11 Since we know the magnitude of the zeta potential, we can use this formula to estimate
12 electroosmotic flow. The best way to see the effect of electroosmotic flow in our system is
13 to compare the theoretical water recovery to the obtained data. Water recovery is defined
14 as Rw = Qd /Q, where Qd is the flow rate of fresh water and Q is the total flow rate. As
15 we see in the data, Qd changes as a function of current. Assuming that the bulk EO flow
16 is not completely hindered by pressure-driven back flow, we estimate the fresh water outlet
17 flow rate as Qd = αQ + QEOF , where α is the water recovery at zero current, which depends
18 directly on the placement of the splitter and hence is a known parameter. We can then
19 obtain a new formula for water recovery:

Qd αQ + QEOF ζI
Rw = = =α+ (3)
Q Q µσQ

20 The second term is the rescaling of current that was used in figure 3(d) in the paper and
21 this formula (with a prefactor in front of the second term to allow fitting) was used to fit
22 the water recovery data.

23 Method Details

24 Before assembly, the membranes were cut into roughly 2.5×1.5 cm rectangles and then
25 treated chemically to remove any impurities and to activate them. They were first soaked in
26 3% (v/v) hydrogen peroxide at 80◦ C for an hour, rinsed with deionized (DI) water, and then

S2
27 soaked in 0.5 M sulfuric acid at the same temperature for another hour, followed again by
28 rinsing with DI water. Lastly, they were soaked in DI water at 100◦ C for one more hour and
29 then stored in DI water prior to use. The frit was glued into its acrylic frame using Devcon
30 2 Tonr Epoxy from McMaster-Carr before assembly.
31 During assembly, the membranes were trimmed to the appropriate size using a razor
32 blade. The gasketing material was used to seal the device and house the electrode channels
33 that had been cut into the gaskets. The electrode channels were open and pressurized during
34 operation using downstream pressure tubing in order to hold the membranes flat against the
35 frit. The splitter was also made from teflon gasketing material that was compressed against
36 the end of the frit using the outlet port plate. The device was assembled using five 1.5 in.
37 8-32 and four 2 in. 6-32 316 stainless steel bolts (McMaster-Carr) that were wrapped with
38 electrical tape to minimize corrosion. The 8-32 bolts were tightened to 25 in-lbs of torque
39 and the 6-32 bolts were tightened during operation until no more leaks were observed.

40 Additional Figures

41 Please note that the errorbars in all figures for ion removal, concentration, and energy/current
42 efficiency represent one standard deviation (for 4-8 realizations) about the mean and that
43 the errorbars in all figures for water recovery represent a combination of the range about the
44 mean and an estimate for maximum bias due to the measurement method.

S3
 a) 100

80
Ion Removal (%)

60

40

20
10 mM 38 µL/min NaCl
10 mM 76 µL/min NaCl
10 mM 152 µL/min NaCl
0
0 5 10 15 20
Average Current Density (A/m2)

b) 80
Water Recovery (%)

70

60

50

10 mM 38 µL/min NaCl
40 10 mM 76 µL/min NaCl
10 mM 152 µL/min NaCl

0 5 10 15 20
2
Average Current Density (A/m )

Figure S1: All data shown are for a feedwater concentration of 10 mM NaCl. a) Plot of the
percentage of salt removed versus the applied current density. As can be seen, the fraction of
ions removed increases as the flow rate decreases. b) Plot of water recovery versus the applied
current density. We observe that water recovery also increases as the flow rate decreases,
which we hypothesize is due to the contribution of electroosmotic flow to the total flow.

S4
 a) 100

Ion Removal (%) 80

60

40

20
1 mM 76 µL/min NaCl
10 mM 76 µL/min NaCl
100 mM 76 µL/min NaCl
0
0 1 2 3 4
Current (I/(z+c+eQ))

b) 75

70

65
Water Recovery (%)

60

55

50

45
1 mM 76 µL/min NaCl
40 10 mM 76 µL/min NaCl
100 mM 76 µL/min NaCl
35
0 0.5 1 1.5 2 2.5 3 3.5 4
Current (I/(z+c+eQ))

Figure S2: All data shown are for a feedwater flowrate of 76 µL/min. a) Plot of the per-
centage of salt removed versus the applied current normalized by the rate of positive charge
advection into the device. As can be seen, all three curves approximately overlap. b) Plot
of the water recovery versus the applied current normalized by the rate of positive charge
advection into the device. Water Recovery also does not differ significantly between the
three concentrations.

S5
 a) 100

80
Ion Removal (%)

60

40

10 mM 76 µL/min Na2SO4
20 10 mM 76 µL/min KCl
10 mM 76 µL/min KNO
3
10 mM 76 µL/min NaCl
0
0 5 10 15 20
Average Current Density (A/m2)

b) 70

65
Water Recovery (%)

60

55

50

45
10 mM 76 µL/min Na2SO4
10 mM 76 µL/min KCl
40 10 mM 76 µL/min KNO3
10 mM 76 µL/min NaCl
35
0 5 10 15 20
2
Average Current Density (A/m )

Figure S3: All data shown are for a feedwater concentration of 10 mM of the respective
electrolyte. a) Plot of the percentage of salt removed versus the applied current density.
As can be seen, the fraction of ions removed is similar for the three monovalent binary
electrolytes but is lower for Na2 SO4 , which is due to the fact that the sulfate ion has a
higher charge (and the fact that the concentration of Na+ is actually 20 mM). b) Plot of
water recovery versus the applied current density. We observe that water recovery does not
seem to depend much on the specific ion type or ion charge.

S6
 3
a) 10

Energy Consumption (kWh/m3) 2

10

1
10

0 10 mM 76 µL/min Na2SO4
10 10 mM 76 µL/min KCl
10 mM 76 µL/min KNO3
10 mM 76 µL/min NaCl
−1
10 100 mM 76 µL/min NaCl
1 mM 76 µL/min NaCl
10 mM 38 µL/min NaCl
−2 10 mM 152 µL/min NaCl
10
0 1 2 3 4 5 6 7
Current (I/(z c eQ))
+ +

b)
)
thermo limit

3
10
Energy Consumption (E/E

10 mM 76 µL/min Na2SO4
10 mM 76 µL/min KCl
10 mM 76 µL/min KNO3
10 mM 76 µL/min NaCl
100 mM 76 µL/min NaCl
1 mM 76 µL/min NaCl
10 mM 38 µL/min NaCl
2 10 mM 152 µL/min NaCl
10
0 1 2 3 4 5 6 7
Current (I/(z c eQ))
+ +

Figure S4: a) Plot of the total energy consumption of the SED prototype per volume of fresh
water produced, as a function of the normalized current that was applied. We can observe
that the energy consumption depends mostly on the electrolyte concentration. b) The same
energy consumption normalized by the thermodynamic energy requirement S1 for each data
point plotted against the normalized current.

S7
 60
a) 10 mM 76 µL/min Na2SO4
10 mM 76 µL/min KCl
10 mM 76 µL/min KNO3
50
10 mM 76 µL/min NaCl
Current Efficiency (%) 100 mM 76 µL/min NaCl
1 mM 76 µL/min NaCl
40 10 mM 38 µL/min NaCl
10 mM 152 µL/min NaCl

30

20

10
0 1 2 3 4 5 6 7
Current (I/(z+c+eQ))

b)
(CE − CEinfty) (%)

1
10

10 mM 76 µL/min Na2SO4
10 mM 76 µL/min KCl
10 mM 76 µL/min KNO3
10 mM 76 µL/min NaCl
100 mM 76 µL/min NaCl
1 mM 76 µL/min NaCl
10 mM 38 µL/min NaCl
0 10 mM 152 µL/min NaCl
10
0 1 2 3 4 5 6 7
Current (I/(z+c+eQ))

z eQ (c −c )
Figure S5: (a) Plot of the current efficiency (defined as CE(%) = + d nI
+,0 +,d
∗100%, where
n is the number of repeat units, which is 1 in this case) of the SED prototype, as a function
of the normalized current that was applied. We can observe that the current efficiency tends
to be higher at lower applied current. (b) Plot of the difference of the current efficiency
and the lower bound that the current efficiency seems to be approaching (set at 9.69% via
fitting). We can observe that the current efficiency appears to be exhibiting exponential
behavior as a function of current.

S8
 100

80
Ion Removal (%)

60

40

20
10 mM NaCl 76 7L/min ultra fine pores
10 mM NaCl 76 7L/min fine pores
0
0 1 2 3 4
Current (I/(z c eQ))
+ +

Figure S6: Plot of data for a feedwater concentration of 10 mM NaCl and a feedwater
flowrate of 76 µL/min for two different porous media. The two porous media are made of
the same material and are supplied by the same company. The only difference is the pore
size. The ultra fine material has a pore size of 0.9-1.4 µm, whereas the fine material has a
pore size of 4-5.5 µm. We can observe that ion removal decreases as the pore size increases. S2

S9
45 References

46 (S1) Yip, N. Y.; Elimelech, M. Thermodynamic and Energy Efficiency Analysis of Power
47 Generation from Natural Salinity Gradients by Pressure Retarded Osmosis. Environ.
48 Sci. Technol. 2012, 46, 5230–5239.

49 (S2) Dydek, E. V.; Bazant, M. Z. Nonlinear dynamics of ion concentration polarization in

50 porous media: The leaky membrane model. AIChE J. 2013, 59, 3539–3555.

S10