www.sciencemag.org/content/36
2/6419/1144/suppl/DC1
Supplementary
Material
s for
Room
-temperature cycling of metal fluoride electrode
s: Liquid electrolytes for
high-
energy fluoride
ion cells
Victoria K. Davis, Christopher M. Bates, Kaoru Omichi, Brett M. Savoie, Neboj
ša Mom
č
ilovi
ć
,
Qingmin Xu, William J. Wolf, Michael A. Webb, Keith J. Billings, Nam Hawn Chou,
Selim Alayoglu, Ryan K. McKenney, Isabelle M. Darolles, Nanditha G. Nair
, Adrian Hightower,
Daniel Rosenberg, Musahid Ahmed, Christopher J. Brooks, Thomas F. Miller III,
Robert H. Grubbs, Simon C. Jones*
*Corresponding author.
Email:
simon.c.jones@jpl.nasa.gov
Published
7 Dec
em
ber
2018,
Science
362
, 1144
(2018)
DOI:
10.1126/science.
aat7
070
This PDF file includes:
Materials and Methods
Supplementary Text
Figs. S1 to S20
Tables S1 to S4
Caption for data S1
References
Other supplementary material for this manuscript includes the following:
Data
S1 (
zipped
folder
)
2
Materials and Methods
Materials
All compounds and solvents were purchased from Sigma
-Aldrich, Alfa Aesar,
TCI America, Strem, and/or Fisher and used as-
received. NMR solve
nts were purchased
from Cambridge Isotope Laboratories, Inc. Anhydrous diethyl ether was obtained via
elution through a solvent column drying system
(
27
)
and degassed with argon prior to
use. Solvents used for solvent screening and electrolyte formulations
with
dry
fluoride
salts were dried over 4Å molecular sieves in either a dryroom or an argon filled
glovebox. H
2
O content (ppm) was monitored via Karl Fisher titration until solvents were
anhydrous (H
2
O ≤
16
ppm).
Electrode materials used included: bismuth
foil
(Alfa Aesar,
1mm thick, 99.999%), lead foil
(Alfa Aesar, 0.1mm thick, 99.998%), cerium foil (Alfa
Aesar
, 0.62
mm thick,
99.9%),
calcium foil (American Elements, 2mm thick, 99.9%),
Super P carbon black
(SP; MTI Corporation, TIMCAL Graphite & carbon Super P,
Conductive carbon black), and poly(vinylidene fluoride) (PVDF;
Alfa Aesar
).
Materials Synthesis
Fluoride Salt Synthesis (Np
1
F, Two Steps
).
Trimethylneopentyla
mmonium
iodide
was prepared from the addition of potassium carbonate (94.1 g, 681 mmol, 2.67
eq) to a 1 L Erlenmeyer flask containing a magnetically stirred solution
of methyl iodide
(51 mL, 819 mmol, 3.21 eq)
and neopentylamine (30 mL, 255 mmol, 1.0 eq)
in
absolute
ethanol (
400
mL)
. After magnetic stirring at room temperature for 21 hours, the mixture
was suspended
in ethanol (1.25 L), filtered, and the solvent
was removed in vacuo.
The
resulting solid was then suspended in dichloromethane (4 L) and filtered. Solvent was
again removed in vacuo
and the solid recrystallized from isopropanol (450 m
L). The
resulting crystals contained 1.6 mol% isopropanol via
1
H NMR. Removal of this trace
solvent was accomplished by dissolving in water (55 mL) and washing with hexane (3 x
50 mL). Drying in vacuo
yielded 48.53 g of white solid (74%).
1
H-NMR (300 MHz,
CD
3
CN
) δ
3.38
(2
H,
s, N
+
C
H
2
C(CH
3
)
3
), 3.21
(9 H, s,
N
+
(C
H
3
)
3
), 1.17
(9 H, s,
N
+
CH
2
C(C
H
3
)
3
).
13
C-NMR (
126
MHz, CD
Cl
3
) δ
76.71, 56.03, 33.73, 30.16.
HRMS (EI)
calcd
. for C
8
H
20
N [H
+
] 130.16, found
m/z: 130.1596.
Trimethylneopentyla
mmonium fluoride
was prepared using
an adapted procedure
from
the literature
(
19
). Silver oxide (14.559 g, 62.8 mmol, 1.5 eq) was added to a
solution of
trimethylneopentylammonium
iodide (10.772
g,
41.
9 mmol, 1 eq)
in
deionized water (225 mL) in a
n aluminum foil covered 500 mL round bottom flask. After
stirring for 1 h, ion-
exchange of iodide for hydroxide was quantitatively achieved
, as
evidenced by the lack of precipitate upon aliquot addition
to a solution of silver nitrate in
concentrated hydroc
hloric acid. The suspension was
filtered and immediately titrated
with aqueous hydrofluoric acid (0.5 wt %). HF was added dropwise and the titration
stopped at pH 7.96 (calc. endpt = 8.08). Most of the water was removed under reduced
pressure at 60 °C. The
solution was further dried by azeotrope with
bench-
grade
isopropanol (x3)
at 35 °
C. To remove trace silver residue, the solution was
filtered
(25
mm wheel filter, 0.45 μm PTFE membrane). The solution was transferred into a side arm
round bottom flask, and
the residual water was removed by azeotrope with dry
isopropanol (x5) under high vacuum (
~50 mTorr) at 100 °C for five days until
Np
1
F
remained as a
white
powder, whose F
-
singlet peak appears downfield from -
75 ppm
3
(versus DF
2
-
normalized to -
147 ppm) vi
a
19
F NMR.
1
H-NMR (4
00 MHz, CD
3
CN
) δ
3.34
(2 H,
s, N
+
C
H
2
C(CH
3
)
3
), 3.26 (9 H, s,
N
+
(C
H
3
)
3
), 1.18 (9 H, s, N
+
CH
2
C(C
H
3
)
3
).
13
C-NMR (
100
MHz, CD
3
CN
) δ
76.35, 54.96, 33.41, 29.65.
19
F-NMR (376
MHz,
CD
3
CN
) δ
-74.29
(s,
F
-
), -147.00 (t,
D
F
2
-
).
1
H NMR and
19
F NMR spectra are shown in
fig. S
17.
Fluoride Salt Synthesis (Np
2
F, Five Steps
).
N
-(2,2-dimethylpropyl)
-2,2-
dimethylpropanamide
was prepared following a report
by
Anderson and coworkers
(
28
)
.
A 100 mL round bottom flask equipped with a stir bar was charged
with n
eopentylamine
(10
mL,
85.5
mmol, 1.23
eq.) , triethylamine (12 mL, 85.5 mmol, 1.23 eq.), and
chloroform (70 mL), and cooled to 0 °C. Pivaloyl chloride (8.6 mL, 69.5 mmol, 1 eq.)
was added drop wise and the resulting solution refluxed (70 °C
) for 4 hr. Upon cooling to
room temperature, the organic layer was rinsed with deionized water (3x), brine (1x),
dried over sodium sulfate, and filtered. T
he solvent was removed in vacuo
to yield an
orange solid (1
1.4
g, 96
% yield).
1
H-NMR (300 MHz, CD
Cl
3
, 20 °C)
δ
5.66
(1 H,
bs,
O=CN
H
), 3.05 (2 H, d
, NC
H
2
C(C
H
3
)
3
), 1.21 (9 H, s, O=CC(C
H
3
)
3
), 0.90 (9 H, s,
NCH
2
C(C
H
3
)
3
).
13
C-NMR (
126
MHz, CD
Cl
3
) δ
178.32, 50.32, 38.93, 32.09, 27.77,
27.26. HRMS (EI)
calcd. for C
10
H
21
NO [H
+
] 172.2882, found
m/z: 172.1700.
N
-(2,2-
dimethylpropyl)
-2,2-
dimethylpropan-
1-amine
was prepared using a
procedure
adapted from
the literature
(
29
)
. In a flame-
dried, three
-neck flask equipped with a stir
bar, lithium aluminum hydride (
6.919 g, 182
mmo
l, 1.5 eq) was suspended in a 5:
6 (vol)
diethyl ether:dibutyl ether mixture (220
mL)
and cooled to 0 °C
.
N
-(2,2-
dimethylpropyl)
-
2,2-
dimethylpropanamide
(20.701
g, 121
mmol, 1.0 eq)
was added to the flask and
stirred
for
30
minutes
. The solution was then refluxed for 42
h (120 °C
). The mix
ture was
cooled to room temperature, quenched with deionized water, and filtered. The filtrate was
treated with concentrated
hydrochloric acid until acidic
, and water (400 mL) was added
to fully dissolve the solid.
The water layer was washed with diethyl e
ther
(3x 250 mL
),
treated with concentrated
sodium hydroxide
solution
until basic, and extracted with
diethyl ether (3x
100 mL). The organic layer was dried
over sodium sulfate
, filtered, and
the solvent mostly
removed at 40 °C (
no vacuum
; the amine is vol
atile
). The resulting
product was isolated as a slightly yellow clear ether
eal solution
(47.717
g,
36.1 wt%,
91% yield).
1
H-NMR (
500 MHz, CD
Cl
3
)
δ
2.37
(4 H, s, N(C
H
2
C(CH
3
)
3
)
2
), 0.
94
(18 H, s,
N(CH
2
C(C
H
3
)
3
)
2
).
13
C-NMR (
126
MHz, CDCl
3
) δ
63.53, 31.96, 27.76. HRMS (EI)
calcd
. for C
10
H
23
N [H
+
] 158.18, found
m/z: 158.1908.
N
-(2,2-
dimethylpropyl)
-
N
,2,2-
trimethylpropan
-1-amine was prepared
using a
procedure
adapted
from t
he literature
(
28
).
The
ether
eal solution of
N
-(2,2-
dimethylpropyl)
-2,2
-dimethylpropan-
1- amine
(16.1 g in diethyl ether (36.1 wt%), 112
mmol, 1 eq) was cooled to 0 °C and f
ormic acid (
11.2
mL, 29
7 mmol,
2.65 eq)
was added
dropwise.
Formaldehyde (
8.50 mL (aq. 37 wt%
), 145
mmol
, 1.30
eq)
was added and the
mixture refluxed at 60
°C
for 22 h.
Concentrated hydrochloric acid was added until an
acidic pH was reached. The solvent was removed in vacuo
at 55 °C, yielding a peach-
colored solution. Concentrated sodium hydroxide solution
was added until a basic pH
was r
eached.
The aqueous l
ayer was extracted with diethyl ether
(3x
150 mL
). The
organic layer was
dried
over
sodium sulfate,
filtered,
and the solvent
mostly removed at
50
°C (
no
vacuum
; the amine is volatile
). The resulting product was isolated as a slightly
yellow
clear ether
eal solution
(27.454
g,
64.1 wt
%, 98% yield).
Characterization
information was found to correlate with literature values
(
28
)
.
1
H-NMR (300 MHz,
4
CDCl
3
, 20 °C) δ
2.30
(3 H, s, NC
H
3
), 2.19 (4 H, s, N(C
H
2
C(CH
3
)
3
)
2
), 0.88 (18 H, s,
N(CH
2
C(C
H
3
)
3
)
2
).
13
C-NMR (
126
MHz, CDCl
3
) δ
74.68, 48.25, 33.51, 28.91. HRMS
(EI) calcd. for C
11
H
25
N [H
+
] 172.2065, found
m/z: 172.2072.
Dimethyldineopentylammonium
iodide
was prepared from a procedure
adapted
from
the literature
(
19
). A solution of
N
-(2,2-
dimethylpropyl)
-
N
,2,2-
trimethylpropan-
1-amine
in Et
2
O (64 mass %, 17.07g, 99.6 mmol, 1 eq), methyl iodide (19 mL, 305 mmol, 3.1 eq),
and acetonitrile (85 mL) were added to a 500 mL round bottom flask equipped with a stir
bar and refluxed for 5 days. The solvent was then rem
oved in vacuo at 45 °C, and the
product was recrystallized from isopropanol to yield 23.826 g off
-white crystals (76%
yield).
1
H-NMR (
300 MHz, CD
Cl
3
) δ 3.
67
(4 H, s, N
+
(C
H
2
C(CH
3
)
3
)
2
), 3.47
(6 H, s,
N
+
(C
H
3
)
2
), 1.25
(18 H, s, N
+
(CH
2
C(C
H
3
)
3
)
2
).
13
C-NMR (
126
MHz, CDCl
3
) δ
77.50,
54.23, 34.07, 30.59. HRMS (EI) calcd. for C
12
H
28
N [H
+
] 186.22, found m/z: 186.2222.
Dimethyldineopentylammonium
fluoride
was prepared from a
procedure
adapted
from the
literature
(
19
)
. Silver oxide (24.616 g, 106 mmol, 1.5 eq) was added to a
solution of the iodide salt (22.158
g,
70.7 mmol, 1 eq)
in deionized water (330 mL) in an
aluminum foil covered 500 mL round bottom flask. After stirring for 1 h, ion-
exchange of
iodide for hydroxide was quantitatively achieved
, as evidenced by the lack of precipitate
upon aliquot addition
to
a solution of silver nitrate in concentrated hydrochloric acid. The
suspension was filtered and immediately titrated with aqueous hydrofluoric acid (0.5 wt
%). HF was added dropwise and the titration stopped a
t pH 7.96 (calc. endpt = 8.08).
Most of the water was removed under reduced pressure at 60 °C. The solution was
further dried by azeotrope with bench
-grade isopropanol (x3) at 35 °C. To remove trace
silver residue, the solution was micron filtered (25 mm w
heel filter, 0.45 μm PTFE
membrane). The solution was transferred into a side arm round bottom flask, and the
residual water was removed by azeotrope with dry isopropanol (x5) under high vacuum
(~50 mTorr) at 100 °C for 5 days until a pale yellow powder re
mained, whose F
-
singlet
peak appears downfield from -
75 ppm (versus DF
2
-
normalized to -
147 ppm) via
19
F
NMR.
1
H-NMR (3
00 MHz, CD
3
CN
, 20 °C) δ
3.48
(4 H,
s, N
+
(C
H
2
C(CH
3
)
3
)
2
), 3.34 (6 H,
s, N
+
(C
H
3
)
2
), 1.19 (18 H, s, N
+
(CH
2
C(C
H
3
)
3
)
2
).
13
C-NMR (
100
MHz, CD
3
CN
) δ
77.77,
53.77, 34.06, 30.20.
19
F-NMR (282
MHz, CD
3
CN
, 20 °C) δ
-72.87
(s,
F
-
), -147.00 (t,
D
F
2
-
).
1
H NMR and
19
F NMR spectra are shown in fig. S
18.
Copper Nanoparticle Synthesis
.
Hydrazine hydrate (50-
60%, 3 mL, 17.66
M)
was added to a stirring solution of hexadecyltrimethylammonium
bromide
(CTAB)
(0.68
g, 1.87 mmol) and citric acid monohydrate (0.08 g, 0.38 mmol) in deionized water (75
mL
) under argon at 23 °C. The solution was allowed to age for 20 minutes under argon.
Ammonium hydroxide (0.5 mL, 14.5
M) was added to a solution of c
opper
(II)
nitrate
hemipentahydrate
(0.465 g, 2 mmol) and CTAB
(0.68 g, 1.87 mmol) in deionized water
(75 mL). The copper precursor solution was immediately poured into the hydrazine
solution and this mixture was stirred under argon for 2 hours. Copper nanoparticles were
isolated via centrifuge (12,000 rpm, 5 min). T
he supernatant was discarded and the
copper nanoparticles were washed with ethanol (10 mL) twice. The product identity was
confirmed via pXRD
(fig. S
19A)
. The copper nanoparticles have ~50 nm diameter, as
determined via TEM imaging (
fig. S
19B)
.
Cu@LaF
3
Cor
e-Shell Nanoparticle Synthesis
.
Copper nanoparticles were
prepar
ed as described above
. Once copper nanoparticles were isolated via centrifuge
(12,000 rpm, 5 min), the supernatant discarded, and the copper nanoparticles washed with
5
water
(30 mL) twice, the Cu nanoparticles were
re-dispersed
in
deionized
water
(150 mL)
and stirred under argon. Hydrazine hydrate (50
-60%, 3 mL, 17.66 M) was added to the
Cu nanoparticles and stirred for 10 minutes.
Both a solution of l
anthanum nitrate
hexa
hydrate (
0.433 g, 1
mmo
l)
in water (15 mL), and a solution of sodium fluoride
(0.042 g, 1 mmol) in water (15 mL) were simultaneously injected into the copper
nanoparticle solution over a period of 5 minutes via syringe pump (3 mL/min).
The
mixture was then stirred for an additional 10 minutes under argon. The core
-shell
material was isolated via centrifuge (12,000 rpm, 5 min.), the supernatant was discarde
d,
and the remaining core
-shell nanoparticles were washed with ethanol (10 mL) twice.
Peaks corresponding to both LaF
3
and met
allic copper were exhibited by pXRD
(fig.
S20)
. Core
-shell product identity
was confirmed via ICP
-MS of the powder (Cu:La:F
[atomic %
] = 77.5:6.8:15.7), EDX micro analysis (Cu:La:F
[at%
] = 94.7:1.6:3.7)
, and
EDS elemental mapping (Cu:La:F[at%] = 94.7:1.6:3.7)
. The copper core has a 50 nm
diameter with a 5 nm
-thick LaF
3
shell, as determined via TEM imaging
(Fig. 3A)
.
Cu
-La
F
3
Thin
-Film Preparation.
80 nm of copper
(Cu sputtering target)
was
deposited onto a 5x20
mm
area, 1 mm thick
glassy carbon
(GC)
substrate
via DC
sputtering
: 100 W; 3 mTorr; 63A/min sputtering rate
. The Cu
-coated substrate remained
in the chamber to cool down. Then,
4.5 nm of lanthanum
fluoride (
LaF
3
sputtering target)
was
deposited on
top
of
the copper thin
-film
via RF sputtering
: 100 W; 3 mTorr;
10Å
/min sputtering rate
. The coated substrate was then cut into 5x5 mm strips for
electrochemical testing. Results
from
electrochemi
cal testing
and XPS
are shown in
Fig.
3H, Fig. 3I, and fig. S16.
Electrode Fabrication.
Bismuth foil, lead
foil, cerium foil, and calcium foil were
cut into thin strips for use in three
-electrode cells. Copper nanoparticles or Cu@LaF
3
core
-shell nanoparticles were made into a paste with PVDF and/or SP, pressed into
stainless steel mesh, and dried under vacuum prior to three
-electrode assembly. The Cu
-
LaF
3
thin
-film was made as described above and assembled into a three
-electrode cell.
Electrochemical testing details are described in the following section.
Electrochemical
Testing
Electrolyte
Ionic Conductivity
Studies.
Ionic conductivities for a number of
anhydrous
Np
1
F and Np
2
F solutions
were
investigated
by AC impedance spectroscopy
using a VersaSTAT potentiostat
. Measurements
were
acquired between 100 mHz and 1
MHz using an air
-free glass conductivity cell including a Teflon ring sealing the solution
between two parallel Pt electrodes. The Pt electrodes are separated by ~1 cm, and
the
cell constant
was
determined before each exper
iment by measuring the conductivity of
an
aqueous potassium chloride (
0.1M
) solution.
Thermal control was
provided by a Tenney
TUJR chamber, with the sample allowed to reach thermal equilibrium before
measurement (as determined by observation of no change in the impedance spectrum
over time).
Electrolyte
Voltage Window Determination
.
Fluoride electrolyte solutions
were
investigated by linear sweep voltammetry
using a Bio
-Logic VMP2 potentiostat
to
determine their
electrochemical/
voltage
stability
window
usi
ng a 1 mV/s scan rate. A
Pt
working electrode, Pt auxiliary,
and non-
aqueous Ag
+
/Ag (MeCN) reference electrode,
with Ar purge
, were
employed for
these studies
. V
oltage windows
were determined by
two methods: (i) the
J
cut
-off
method, using a limiting curre
nt of 100 μA/cm
2
(Fig. 1I and
6
fig. S
5A), and (ii) the linear fit method (
30
)
, where voltage limits are defined as the
intersection between linear fits of the I-
V curves before and after the onset for electrolyte
decomposition (fig. S
5, B to D
, summary data
in E
).
Solid
-Electrolyte Interphase Formation.
1
H
,1
H
,2
H
,2
H
-perfluorooctyltriethoxy-
silane (FOTS; 0.25 M) was added to 0.75 M Np
1
F/BTFE electrolyte. This solution
mixture was used in a three-
electrode set
-up with a Ce or Ca working electrode, Pt wire
counter electrode, and silver wire quasi
-reference electrode (see table S
4). FOTS grafting
to Ce or Ca anodes was achieved by cyclic voltammetry (CV) and monitored in situ
via
electrochemical impedance spectroscopy (EIS). CV parameters: the potential was varied
between
-0.8V and +0.7 V vs Li
+
/Li for 5 cycles using a scan rate of 100 mV/s. EIS
parameters:
AC impedance spectroscopy m
easurements
were
acquired between
0.1
Hz
and
0.2
MHz
. A Bio
-Logic VMP2 potentiostat was used to alternate between CV and EIS
electro
chemical tests. The initial test was CV followed by EIS, followed by CV, etc.
Tests alternated from CV to EIS repeatedly until ten total electrochemical tests had been
performed on the Ce or Ca anode
. Confirmation of SEI formation on the Ce or Ca anode
was
achieved via ex situ
XPS measurements. An example data set collected using a Ca
anode is shown, where odd numbered test
s correspond to CV data (fig. S8A) and even
numbered tests correspond to EIS data (fig. S8
B). XPS analysis of the Ca surface with
FOTS g
rafted is also shown as an examp
le (fig. S8
C).
Three-
Electrode
Assembly and Cycling Details.
Bismuth
,
lead
,
copper,
Cu@LaF
3
, Cu
-LaF
3
thin
-film,
calcium, or cerium electrodes were employed as the
working electrode in a standard three
-electrode cell. Platinum wire was used as the
counter electrode and a
silver wire in 0.01 M AgTOf/MPPy
-TFSI was used as the non
-
aqueous
pseudo-
reference electrode.
Specific details about each battery
, electrolyte
composition,
and cycling parameters used are listed
, along
with ICP
-MS data of the
electrolyte solution after cycling (table
S4) . Electrochemical charge and discharge
cycling was carried out using a
VersaSTAT MC potentiostat.
Instrumentation
Nuclear magnetic resonance (NMR) spectra were obtained using either a Mercury
Plus 300, Varian 400 MR, Inova 500, or Bruker 400
NMR spectrometer. Chemical shifts
for protons are reported in parts per million downfield from tetramethylsilane and are
referenced to residual proti
o-solvent
in the NMR solvents: CDCl
3
(δ 7.26), CD
3
CN (δ
1.96). Data are represented as follows: chemical shift, integration, multiplicity (s =
singlet, d = doublet, sep = septet, m = multiplet, br = broad), coupling constants in Hertz
(Hz), and assignment. Mass spectrometric data were obtained at the Calt
ech Mass
Spectrometry Facility.
Pulsed
-field gradient spin
-echo (PFG
-SE)
1
H and
19
F NMR
experiments were
performed on a Varian 500 MHz spectrometer with auto-
x pfg broadband probe
interfaced with a workstation equipped with VnmrJ software (v 4.2). In an A
r filled
glovebox, a 5 mm NMR tube was charged with 400 μL of a 0.75 M solution of Np
1
F (in
BTFE or 3:1 BTFE:DME) or Np
2
F (in BTFE) and sealed with a rubber septum and
secured using Teflon tape. The NMR tube was removed from the glovebox and flame
sealed.
The sample was then loaded into the spectrometer and DOSY spectra were
recorded (unlocked in pure
proti
o-solvent) at the desired temperature (5 –
40 °C,
7
increments of 5 °C). The temperature of the probe was calibrated using a methanol
standard.
Inductively
coupled plasma mass spectrometry (ICP
-MS) was operated by
Laboratory Testing, Inc. using Thermo iCap-
Q-Mass Spec. 3 mL of electrolyte was
collected after each electrochemical test. Samples were diluted by water up to 10 mL,
such that 10-
30% electrolyte wa
s included. Before ICP
-MS analysis, samples were
shaken to thoroughly mix both the organic and water layers.
Powder X
-ray diffraction (pXRD) patterns were collected using Bruker D8
ADVANCE instrument with X
-ray generator of 40 kV and 40 mA. Post electroche
mical
testing samples were loaded into a sealed home-
made cell with a Be
window to avoid air
and moisture. Parameters of pXRD scans were in the range of 10 to 90° 2θ with 0.027 2θ
step
-size and a count time of 12 sec/step. pXRD pattern
s of Cu@LaF
3
and Cu precursor
powders were collected in air.
Transmission electron microscopy (TEM) and high resolution TEM (HR
-TEM)
images were collected using
an FEI Tecnai F20 operating at 200 kV. Energy dispersive
spectroscopy (EDS) was performed using
an
image
-corrected F
EI Titan3TM G2 60-
300
operating at 300 kV, equipped with a Super
-X four quadrant detector. The post
electrochemical testing samples were dispersed in anhydrous n
-hexane (Aldrich) in a
glovebox (H
2
O < 0.5 ppm). 20 μL of colloidal suspension (1 mg/1 mL) was
drop-
cast
onto a nickel TEM grid with holey carbon substrate. Samples were vacuum dried for two
day before being transferred to the TEM in air. For as-
synthesized copper nanoparticles
and Cu@LaF
3
nanoparticles, the TEM samples were dispersed in ethanol and
drop
-cast
on a nickel grid.
XPS
depth profile analyses were performed
by Nanolab Technologies
using
a K
-
Alpha
TM
+ X
-ray Photoelectron Spectrometer (XPS) System manufactured by Thermo
Fisher Scientific, Inc. Samples were not exposed to X
-rays until
the
measurement was
started to minimize
the
chance of degradation. X
-rays are monochromatic Al K
α
1486 eV
(8.3383
Å). The etch rate of thermally grown SiO
2
was
used as a rough measure of etch
depth. The argon ion etch crater size was
4 x 2 mm
with an
X-ray be
am size of
0.4 mm
.
The Ar
+
etch
ing was performed with an etch
rate for SiO
2
of
0.8 Å/sec
. Depth profiles
were obtained with an Ar
+
beam voltage of 0.5 kV
, angle of incidence of 30 degrees, and
an etch rate for SiO
2
of
0.8 Å/sec
.
STEM pictures and EELS spectra were obtained by using a Jeol2100F
microscope equipped with a GIF Tridiem Gatan EELS spectrometer. EELS maps were
recorded at 120 kV, and EELS point spectra were taken at 200 kV accelerating voltage.
Probe size was 1.5 nm for the mapping and 0.7 nm
for the point acquisition. Entrance and
exit angles of the electron beam w
ere
12 mrad. Energy resolution was 1.0 eV as
measured from the full width half maximum of zero loss peak in vacuum. All EELS
spectra were obtained between 390 eV and 1000 eV with 0.3 eV energy steps and 1 sec
exposures. Elemental analysis was carried out by using the standard Gatan/EELS
software assuming power law for pre
-edge background, and a Hartree
-Slat
er model for
quantification. For the analysis of La M
5,4
edge spectra, first t
he pre
-edge background
was removed, then two sigmoidal functions of the form 1/(1+e
-x
), one at each of the M
5
and M
4
edges, of the same amplitude as the edge jump were subtracted from the data.
Least
-square fittings of the M
5
and M
4
peaks were carried out by constraining amplitude,
loss energy and FWHM. Refined amplitudes were used to calculate the M
5
/M
4
ratios.
8
Samples were transferred to a nitrogen glove bag and dispersed in anhydrous n-
hexane
(Aldrich). 20 μL of colloidal suspension (1 mg/1 mL) drop
-
cast
onto a Ni TEM grid with
holey carbon substrate. Samples were vacuum dried 2 days before TEM analysis.
Samples
were
transferred to the TEM holder in air.
Computational
Methods
and Calculations
LAMMPS was used to perform all molecular dynamics simulation
s (
31
)
. All
simulations used a one fs integration time
step, Velocity
-Verlet integration, and periodic
boundary conditions. Long-
range electrostatics were modelled using the particle
-particle
-
particle
-mesh (PPPM) algorithm (
32
)
and Lennard
-Jones interactio
ns were truncated at
14
Å. All simulations were initialized from diffuse configurations containing at least
1500 atoms, using a cubic grid to place solvent molecules in random orientations without
overlaps. The simulations were first relaxed in the NVE ensemble with restrained atomic
displacements of 0.1 Å
per time
step for 30 ps, followed by a 1 ns NPT equilibration
where the temperature was linearly increased from 100 K to 298 K to condense the
simulations. The simulations were further equilibrated at 298
K for 2 ns in the NPT
ensemble, prior to performing ion insertions for the solvation free energy calculations. In
the NPT simulations, the Nosé
-Hoover thermostat and barostat were employed using the
modified form proposed by Martyna, Tobias, and Klein as implemented in LAMMPS
(
33
)
. For the radial distribution functions (RDFs) reported in the main text, individual
ions were randomly inserted into the pre
-equilibrated solvent simulations, allowed to
further equilibrate in the NPT ensemble for 1 ns, then the RDFs were generated from an
additional 10 ns of production data.
The radial distribution function
for
F
–
(Fig. 1D) was
calculated to characterize the strength of its
interaction with the
α
–CX
2
(X = H or F)
moiety of the indicated solvent. In the case of BTFE, there is a large probability of F
–
interaction about 2Å from the H atom of the
α
–CH
2
group; for diglyme, the
corresponding probability is considerably reduced. BPFE shows a very small probability
of
F
–
interaction with the F
-containing backbone over all separations.
The protocols for
the free-
energy simulations are described in detail in a dedicated section below.
Since several of the solvents presented in this study are novel, suitable fo
rce-
fields
were unavailable
from the existing literature
. Therefore, all solvent force fields in this
study were parameterized on the basis of density functional theory (DFT) quantum
chemistry calculations, using the B3LYP
-D3/def2-
TZVP level of theory computed via
the Orca software package (
34
)
. Following a previously described approach (
35
), the
solvent force fields were parameterized using the OPLS force-
field function form (
36
)
,
except that 1
-4 pairwise interactions were excluded in the non-
bonded intera
ction
computation. In brief, bond, angle, and dihedral force
-field terms were derived from
potential energy curves computed for internal degrees of freedom for each molecule in
vacuum, optimizing the other degrees of freedom as a function of the mode scan.
The
resulting energy curves were self
-consistently fit to obtain the corresponding force
-
constant parameters and equilibrium displacement parameters in the force field. All bond,
angle, and dihedral modes for the ions were taken from OPLS (
36
).
For all so
lvents and
ions,
Lennard-
Jones parameters were taken from the universal force field (UFF) (
37
)
and
partial charges were obtained from CHELPG calculations (
38
)
performed on the
optimized geometries of the respective molecules.
9
Quantum chemical calculations
were used to characterize the partial charge
distribution in BTFE and diglyme (
fig. S4). BTFE exhibits larger partial positive charges
(0.12) on the hydrogen atoms of the
α
–CH
2
moiety within the F
–
solvation structure than
diglyme (0.01)
, as BTFE has two electron
-withdrawing groups flanking the
α
–CH
2
moiety. Fig
ure
1E depicts the innermost solvation shell of F
–
in liquid BTFE
(as
described in the main text)
. The solvation shell for the Np
2
+
cation in BTFE is more
complex, but qualitativel
y, the
β
-CF
3
groups on BTFE appear to be the most prevalent in
the Np
2
+
solvation structure (f
ig. S
12).
Input files for all simulations are supplied and
provide full details of the employed force
-field parameters and protocols for the
simulations (Data Fil
e S1
).
Thermodynamic integration was used to calculate the ion
-specific solvation free
energies in each solvent. Scaled Lennard
-Jones (U
LJ
) and Coulomb (U
C
) potentials were
used to introduce the ion-
solvent potential energy terms gradually
in t
o the solvent
-only
potential energy terms (U
S
) with
where λ
LJ
is a linear scaling parameter for the solvent
-ion Lennard-
Jones interactions
(U
LJ
) and
U
C
(
λ
C
) =
U
LJ
(1
) +
λ
C
U
C
퐞퐞퐞퐞
.
퐒퐒퐒퐒
where λ
C
is a linear scaling parameter for the solvent
-ion Coulomb interactions. The
potential in eq. S1 was implemented using standard λ
–dependent soft
-core Lennard
-Jones
potentials, as implemented in LAMMPS with n = 1 and
α
LJ
= 0.5. The potential in eq. S2
was i
mplemented by scaling the charges on the ion by λ
C
. The total solvation free
-energy
was obtained by
The brackets in eq. S3 indicate an ensemble average, and the approximation has been
made that the P
Δ
V contribution to the free energy change can be safe
ly neglected. The
integrals in eq. S3 were evaluated numerically using the trapezoidal rule, with λ
LJ
and λ
C
incremented in steps of 0.2 (
eleven
steps total, six
for the Lennard-
Jones phase and six
for the electrostatics, less one redundant step connecting
the two phases). The system was
allowed to equilibrate for 250 ps at each λ
-
step, then an additional 250 ps of dynamics
were used for calculating the necessary derivatives. The derivatives in eq. S3 were
calculated by finite
-difference. At endpoints, forw
ard or backward finite
-difference was
used
;
at all other points the central difference was used with a λ
–step of 0.01 to evaluate
the derivative. In the case of the polyatomic cations, an additional free energy
contribution associated with removing the int
ramolecular electrostatics must be
computed. Free
-energy perturbation was used to evaluate this contribution from a ten ns
MD trajectory of the individual cations in vacuum. The reported
Δ
G
TI
errors were
estimated by bootstrap resampling (5 million samples
).
The pKa of acetonitrile, propionitrile, and BTFE were calculated according to
U
LJ
�
λ
LJ
�
=
U
S
+
λ
LJ
U
LJ
퐞퐞퐞퐞
.
퐒퐒퐒퐒
∆
퐺퐺
푇푇푇푇
=
∫
〈
푑푑
푈푈
퐿퐿퐿퐿
푑푑
휆휆
퐿퐿퐿퐿
〉
1
0
푑푑
휆휆
퐿퐿퐿퐿
+
∫
〈
푑푑
푈푈
퐶퐶
푑푑
휆휆
퐶퐶
〉
1
0
푑푑
휆휆
퐶퐶
퐞퐞퐞퐞
.
퐒퐒퐒퐒
10
where
Δ
G
-H,i
is the free energy of deprotonation for species i, R is the ideal gas constant,
T is 298 K, and a and b are calibration constants obtained by a least
-squares fit of eq. S4
to experimental pKa values for reference solvents (1,2,3,4,5-
pentamethylcyclopenta
-1,3-
diene, pKa = 26.1; cyclopenta
-1,3
-diene, pKa = 18.0; dimethyl 2-
(trifluoromethyl)propanedioate, pKa = 10.8; dimethyl 2-
methylpropanedioate, pKa =
18.0; dimethyl propanedioate, pKa = 15.9; 1,1,1,3,3,3-
hexafluoro-
2-
(trifluoromethyl)propane, pKa = 11.0; m
ethane, pKa = 56.0; and acetonitrile, pKa = 31.3)
(
39–41
).
Δ
G
-H,i
was calculated by first geometry optimizing the neutral and deprotonated
species in vacuum, followed by additional geometry optimization using the COSMO
polarizable continuum model with a dielectric of 46.7 to match the DMSO reference
solvent. After optimizing the geometry, a frequency calculation was performed to obtain
the zero
-point energy correction to the free energy. The difference in zero-
point energy
corrected single point energies y
ielded
Δ
G
-H,i
for each species. All quantum chemistry
calculations were performed at the B3LYP
-D3/ma
-def2
-TZVP level of theory.
Calculation of solvent pKa values for acetonitrile and propionitrile (fig. S3A) correlate
well with those reported in the litera
ture (
42
).
푝푝
퐾퐾
푎푎
,
푖푖
=
푎푎
∆
퐺퐺
−
퐻퐻
,
푖푖
푅푅푇푇푅푅푅푅
(
10
)
+
푏푏
퐞퐞퐞퐞
.
퐒퐒퐒퐒
11
Supplementary
Text
Solvent
Screening with
Np
1
F and Long-
Term
St
ability
of
the
F
-
Ion in
Non
-Aqueous
Solutions
All solvent screening experiments in this study were carried out inside an argon
filled glovebox (H
2
O ≤ 10 ppm).
Solvents were purcha
sed
commercially
and
dried over
4Å molecular sieves until anhydrous, as measured via Karl Fisher titration. Purity of all
such
-treated
solvent
s were
confirmed via NMR spectroscopy
prior to solvent screening.
Solvent screening was carried out by dissolving anhydrous
Np
1
F in the
anhydrous solvent
until the solution was saturated. Weight
s (in grams) of 5 mL oven-
dried scintillation
vial
s, solvent, and Np
1
F were recorded
using an analytical balance inside the glovebox,
enabling saturation concentrations (M) of Np
1
F
in the solvent
subsequently
to be
determined
. An aliquot of the saturated solution was then pipetted into an oven-
dried
NMR tube
containing
0.5 mL of CD
3
CN NMR solvent
, sealed, and then brought out of
the glovebox for
1
H and
19
F NMR spectroscopy.
Char
acterization of the two reactions, (i)
between C
H
3
CN and F
-
to form
H
F
2
-
, and (ii) between C
D
3
CN and F
-
to form
D
F
2
-
, is
well
-established in the literature (
43
)
. All solvents
screen
ed were
expected to exhibit a
triplet peak in the
19
F NMR from
D
F
2
-
(
δ
= -147.0
ppm;
J
= 18 Hz). Because
H
F
2
-
and
D
F
2
-
do not undergo fast exchange with each other on the NMR timescale
(
17, 43
)
,
spectra
that
showed
a new triplet peak in the
1
H NMR from
H
F
2
-
(
δ
= 16 ppm;
J
= 1
21
Hz) and
/or
a doublet peak in the
19
F NMR from
H
F
2
-
(
δ
= -146.6
ppm;
J
= 1
21
Hz)
were
considered to be indicative of F
-
reaction with the solvent
being screened
.
Initial screening of Np
1
F revealed three broad classes of organic solvents (as
described in the main text). Examples of class (b) solvents include nitriles such as
acetonitrile (ACN), 2
–methoxyacetonitrile, 3–
methoxypropionitrile (MeOPN), and
pyridines such as 2,6–difluoropyridine, whereas, examples of class (c) solvents include
propionitrile (PN), 3
–fluorobenzonitrile, and 1–
methyl
–1–propylpyrrolidinium
bis(trifluorosulfonyl)imide (MPPy
–TFSI).
Upon determining PN and BTFE to be
excellent solvents for
stable solvation of
the
F
-
ion,
two
J. Young NMR tubes were prepared containing anhydrous
Np
1
F/PN
and
Np
1
F/BTFE solution
s respectively
(both
wit
hout
CD
3
CN
NMR solvent)
and sealed in
inert atmosphere
. These J.
Young tubes were stored on the benchtop at room temperature
for
over 140 days while monitoring the l
ong-
term stability of
the
F
-
ion
via
1
H and
19
F
NMR (fig. S3
B). To our surprise, the initial NMR for both Np
1
F/PN and Np
1
F/BTFE
showed
minor
traces of HF
2
-
. The initial F
-
present in these samples was normalized to
100%, relative to the trace HF
2
-
. Over time, the %HF
2
-
increases, concurrent with a
small,
but
observable decrease in %F
-
.
Wh
ile the reaction between F
-
and CH
3
CN are known, it was necessary to explore
whether F
-
is a strong enough base to deprotonate PN or BTFE. Computational methods
were used to calculate
the pKa of acetonitrile, PN, and BTFE, as described above
. The
presence
of HF
2
-
was never observable in these solutions when CD
3
CN NMR solvent was
used (fig. S3, C and D). Although the reaction of F
-
with CD
3
CN is well
-characterized
(
43
), the analogous behavior of PN and BTFE has not been described in the literature. To
explor
e whether the trace HF
2
-
observed arises from reactivity with PN or BTFE, a set of
control NMR experiments was carried out using deuterated PN (
d
5
-PN) and CD
3
CN (fig.
S3E). In the CD
3
CN solution, DF
2
-
is observed as expected, arising due to deprotonation
12
of the acidic CD
3
groups by F
-
and no HF
2
-
is detected. T
he reaction of F
-
with CD
3
CN,
therefore, must dominate over reaction with any trace protic impurities
(due to
the
vast
excess of CD
3
CN)
. For the
d
5
-PN solution, the opposite behavior is observed: F
-
does not
react with the
d
5
-PN solvent to give DF
2
-
, but instead reacts with an unidentified protic
impurity (present in trace quantities) to form a minor amount of HF
2
-
. In light of this, we
conclude that F
-
is not a strong enough base to abstract deuterium from
d
5
-PN under these
conditions.
Deuterated BTFE is not commercially available, and our own attempts to synthesize
d
4
-BTFE were unsuccessful. Computational methods were used to calculate the pKa of
acetonitrile, PN, and BTFE, as described above. The calculated values for acetonitrile and
PN (fig. S
3A) are in excellent agreement with the literature (
42
). Comparing the
calculated pKa values of acetonitrile, PN, and BTFE, BTFE appears to be significantly
less acidic than PN, and should therefore, be even less reactive as a proton donor to F
-
.
Hence, we conclude that the traces of HF
2
-
observed in Np
1
F/PN, Np
1
F/
d
5
-PN, and
Np
1
F/BTFE solutions (in the absence of CD
3
CN) arise from small amounts of unknown
protic impurities present in the system, and are not derived from deprotonation of the
bulk solvent.
Overall,
the F
-
ion
is chemically stable for
a long
period
when stored at
room
-temperature in anhydrous, non-
aqueous liquid solution (e.g. PN or BTFE) under
inert atmosphere.
Solution Properties
of
Ionic
Motion in
Fluoride
-Ion
Electrolytes
To fully characterize the ionic properties in liquid solution, pulsed
–field gradient
spin–echo (PFG
–SE)
1
H and
19
F
NMR and AC impedance measurements were carried
out for three electrolyte formulations (tables S1 to S3). An Arrhenius plot of self
–
diffusivit
y coefficients (D
Np
+
+ D
F-
) reveals a higher activation energy for the Np
1
F/BTFE
electrolyte over the Np
2
F/BTFE electrolyte (fig. S
13A), whereas activation energies for
ionic conduction are comparable for both electrolytes (fig. S
13B). The transport numbers
for Np
+
cations (
t
+
) and F
-
anions (
t
-
) were
calculated using eq. S5 (
44
)
:
where
D
n
is the self
-diffusion coefficient (in m
2
/s) for the indicated ion, as determined via
pulsed-
field gradient spin
-echo (PFG
-SE)
1
H and
19
F NMR.
The degree of
ion dissociation (
α
) wa
s calculated using eq. S
6 (
45
)
:
푡푡
+
=
퐷퐷
푁푁푝푝
+
(
퐷퐷
푁푁푝푝
+
+
퐷퐷
퐹퐹
−
)
,
푡푡
−
=
퐷퐷
퐹퐹
−
(
퐷퐷
푁푁푝푝
+
+
퐷퐷
퐹퐹
−
)
퐞퐞퐞퐞
.
퐒퐒퐒퐒
α
=
σ
AC
σ
nmr
퐞퐞퐞퐞
.
퐒퐒퐒퐒