See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/265337801

On the complex interplay of crystallinity and surface area effects on Li-ion

intercalation and pseudocapacitive storage properties of nanocrystalline
anatase

Article in Journal of Power Sources · November 2014

DOI: 10.1016/j.jpowsour.2014.08.050

CITATIONS READS

17 298

3 authors:

Micah Jeremy Sussman Amrita Yasin

McGill University Swansea University
11 PUBLICATIONS 62 CITATIONS 12 PUBLICATIONS 237 CITATIONS

SEE PROFILE SEE PROFILE

George P. Demopoulos
McGill University
372 PUBLICATIONS 9,317 CITATIONS

SEE PROFILE

All content following this page was uploaded by Micah Jeremy Sussman on 07 May 2019.

The user has requested enhancement of the downloaded file.

 Journal of Power Sources 272 (2014) 58e67

Contents lists available at ScienceDirect

Journal of Power Sources

journal homepage: www.elsevier.com/locate/jpowsour

On the complex interplay of crystallinity and surface area effects on

Li-ion intercalation and pseudocapacitive storage properties of
nanocrystalline anatase
Micah J. Sussman, Amrita Yasin, George P. Demopoulos*
Materials Engineering, McGill University, 3610 University Street, Montreal, QC, H3A 0C5, Canada

h i g h l i g h t s g r a p h i c a l a b s t r a c t

! Effect of surface area and crystallinity

on lithium storage mechanism was
investigated.
! Increased crystallinity increases the
diffusivity of lithium ions during
lithiation.
! Storage mechanism changes from
diffusion to pseudocapacitance with
rate increase >4C.
! When diffusion-based storage is
predominant increases in polariza-
tion resistance occur.
! Increased polarization causes break-
down of electrolyte and electrode
failure.

a r t i c l e i n f o a b s t r a c t

Article history: Lithium insertion in anatase is studied by considering simultaneously the effects of crystallinity and
Received 22 July 2014 surface area on storage mechanism, namely intercalation vs. pseudocapacitive. 6 nm anatase crystallites
Received in revised form with high surface area (222 m2 g"1), synthesized via a novel continuous aqueous process and annealed at
11 August 2014
different temperatures (200 # C, 300 # C, and 400 # C), were electrochemically tested. By annealing, crys-
Accepted 13 August 2014
Available online 21 August 2014
tallinity was increased, while surface area decreased allowing for the investigation of the contributions of
each toward lithiation and delithiation behavior. The as-synthesized and best-annealed (at 300 # C)
samples were compared at various rates. At low rate (1C and 2C) the annealed anatase had higher
Keywords:
Lithium-ion batteries
reversible capacity than the as-synthesized, due to increased diffusion-based storage. At high rate
Energy storage however, as-synthesized anatase had much higher capacity due to increased surface area and ability to
Titanium dioxide store Li-ions pseudocapacitively. Increased crystallinity leads to increased diffusivity as determined by
Anatase electrochemical impedance spectroscopy, explaining why the 300 # C anneal had higher capacity at low
Annealing rate. Long-term cycling at high rate, however, showed that reliance on diffusion-based storage (inter-
Diffusion calation) in the case of the 300 # C anneal causes increased polarization that leads to electrode perfor-
mance decline. These results point to the importance of simultaneous nanoanatase property
(crystallinity and surface area) optimization for achieving stable performance.
© 2014 Elsevier B.V. All rights reserved.

* Corresponding author. Tel.: þ1 514 398 4755; fax: þ1 514 398 4492.
E-mail address: [email protected] (G.P. Demopoulos).

http://dx.doi.org/10.1016/j.jpowsour.2014.08.050
0378-7753/© 2014 Elsevier B.V. All rights reserved.
 M.J. Sussman et al. / Journal of Power Sources 272 (2014) 58e67 59

1. Introduction While all of this work has focused on the intercalation proper-
ties of anatase at both high and low rate, the contributions from
Various nanophases of TiO2 have been evaluated as active ma- faradaic (bulk insertion) and non-faradaic (surface-area controlled
terial for anodes in lithium ion batteries [1e3]. Research focusing capacitance) have not been clearly distinguished. The ability of
on each of the naturally occurring phases (rutile, anatase, and nanoparticles to deliver good Li-ion storage performance over
brookite) [4e8], as well as synthetic TiO2(B) [9,10], have all shown longer cycling ranges at high rate is usually attributed to shortened
that TiO2 has promise as a replacement for the current commer- diffusion length [26,27]. Wang et al. [26] synthesized porous
cially used graphitic materials. Part of this is ascribed to the pseu- nanoparticles that showed excellent cycling stability at high rate
docapacitive storage that is seen in TiO2, not associated with (~100 mAh g"1 after 100 cycles at 20C), and though their particles
graphite [11e15]. had a high specific surface area, they ascribed the high capacity and
Pseudocapacitance presents itself in TiO2 as reversible surface retention to shortened diffusion paths. The possibility of pseudo-
storage [16]. However, this is only a partial contribution to the ca- capacitive storage associated with the high rate was never
pacity. This contribution is usually ascribed to the amount of sur- mentioned.
face area associated with active material particles [11,17]. The other Presently, to the best of our knowledge, there is no published
contributing storage mechanism is diffusion-based lithium inser- work focusing on the effects of both crystallinity and surface area
tion and extraction that is accompanied by crystallographic phase on high rate Li-ion storage in nanoscale anatase. While Kang et al.
change as per the following reaction: [22] mentioned the low crystallinity of their particles as a reason
for poor performance, they did not quantify this, nor did they
! " compare to the crystallinity of their annealed samples. The other
TiO2 þ x Liþ þ e" ⇔Lix TiO2 (1)
previously mentioned works assumed that their nano-anatase was
highly crystalline, leaving questions as to the effect that crystal-
Insertion of lithium into TiO2, and anatase in particular, can linity has had on the performance of their material. Further, as the
theoretically reach x ¼ 1 in LixTiO2, which translates to anatase investigated is usually high surface area and thus a small
335 mAh g"1, however, practical electrochemical lithiation occurs crystallite size, it seems that is it simply assumed that the diffusion
to x ¼ 0.5 (168 mAh g"1) [18]. Part of this has to do with the fact that length is the primary reason for increased capacity without any
during diffusive insertion anatase undergoes a phase change from mention of the surface area-based storage.
tetragonal TiO2 to orthorhombic Li0.5TiO2 and then to tetragonal Recent research in our laboratory focusing on the forced hy-
LiTiO2 [19,20]. Belak et al. [20] explained this transition as a full drolysis of aqueous TiCl4 solution [28], has led to the discovery that
particle transition, determining that a particle can either be TiO2 or ~6 nm anatase crystallites can be produced and recovered via fast
Li0.5TiO2 or LiTiO2 but not any two at one time, whereas Borghols (30 min residence time), low temperature (at 80 # C) reaction car-
et al. [19] stated that a layer of LiTiO2 diffuses 4 nm deep within the ried in a continuous-stirred reactor at pH 3 [29]. The nanoparticles
TiO2 particle making the x in LixTiO2 a function of particle depth, produced possess high surface area, but are poorly crystalline giv-
thus by shrinking the particle size, from bulk to nano (~7 nm), the ing an opportunity to understand the role crystallinity and surface
fraction of the particle that was converted to LiTiO2 would increase area play in lithium intercalation and pseudocapacitance upon
to 100%. This, however, was in direct disagreement with Jiang et al. annealing. It is thus the goal of this work to report on the electro-
[21] who demonstrated that for 7 nm particles (as compared to chemical energy storage properties of a new type of anatase
30 nm particles) lithiation by diffusion was only part of the process. nanoparticles synthesized via a high throughput continuous
The 7 nm particles (commercially synthesized) showed shorter aqueous solution process and to elucidate the effect that both
diffusion plateaus than their 30 nm counterparts, which indicated crystallinity and surface area play on the mechanism of Li-ion
that surface-based storage played a role as important in the storage storage and cyclability at low and high rate.
of charge as diffusion-based storage [21]. As well, Kang, et al. [22]
showed that the optimal particle size range was 8e25 nm, deter- 2. Experimental section
mined by annealing 5 nm particles synthesized via solvothermal
methods. The 8 nm particle material had the best capacity even at 2.1. Synthesis and characterization
high rate (20C), namely 75 mAh g"1, by contrast their as-
synthesized crystallites (5 nm) performed extremely poorly, Anatase nanoparticles were synthesized using a continuous
something that was ascribed to poor crystallinity. However, they aqueous process developed in our laboratory [29], based on a
did not investigate further these differences to identify the un- modification of an earlier batch method developed by Charbonneau
derlying causes. et al. [28] In a typical synthesis run, a 0.1 M TiCl4(aq) solution was
Further, Belak, et al. [20] also explained that, from a diffusion continuously pumped into a stirred reactor preheated to 80 # C and
stand point, by changing the aspect ratio, rather than just shrinking neutralized with NH4OH or NaOH to pH ¼ 3. After the attainment of
the particle, facile insertion of lithium could occur leading to higher steady-state operation at 30 min mean residence time, a colloidal
capacity even at high rate. Synthesis of non-spherical particles has nanoanatase suspension was produced and separated by centrifu-
provided increasing lithiation capacity. Chen and Lou [23] produced gation. The material was washed 3 times in deionized water and
nanosheets via a hydrothermal method using titanate isoprop- once with anhydrous ethanol, then dried at 80 # C overnight. The
oxide, that achieved 120 and 112 mAh g"1 at 10C and 20C, resulting solid was mechanically ground prior to its use and
respectively, however their anatase saw large capacity fading at characterization.
high rate. In the meantime Shim et al. [24], who synthesized high As-prepared anatase samples were annealed at 200 # C, 300 # C,
surface area (95 m2 g"1) nanowires, achieved 200 mAh g"1 at 1C and 400 # C for 24 h to increase crystallinity. The samples were
and 115 mAh g"1 at 20C. The short diffusion distance allowed for characterized by X-ray Diffraction (XRD), performed with a Bruker
facile lithiation at high rate allowing more capacity even with a D8 Discover Diffractometer equipped with a copper target
surface area lower than that of the nanoparticles characterized by (lCuKa ¼ 0.15406 nm), to identify the crystal phases present. The
Jiang et al. Finally Bresser et al. [25] had impressive intercalation degree of crystallinity was determined by comparing the integrated
results with nanorods 3e4 nm in diameter, achieving 250 mAh g"1 area under the broad (amorphous) hump and sharp crystalline
at C/10 and 130 mAh g"1 at 10C rate. peak corresponding to the (101) anatase crystal plane. Further,
60 M.J. Sussman et al. / Journal of Power Sources 272 (2014) 58e67

electrodes cycled 500 times at high rate were examined using XRD.
The electrodes were removed from the cells, washed in dimethyl
carbonate (DMC) and dried under argon. They were then mounted
on glass microscope slides with carbon tape. Brunauer-Emmett-
Teller (BET) surface area was used to determine particle size and
surface area, using a Micromeritics TriStar 3000 apparatus. All
samples weighed approximately 0.1 g and were degassed for 24 h at
80 # C prior to nitrogen physisorption.
A Philips CM200 Transmission Electron Microscope (TEM) was
used at 200 kV to determine particle morphology and crystal
structure. TEM specimens were prepared by depositing particles on
a copper grid. Selected-Area Electron Diffraction (SAED) for crystal
structure analysis was conducted using a double-tilt specimen
holder (2 different rotation axes with a maximum tilt angle of
~20# ).

2.2. Electrochemical measurements

All electrochemical testing was performed via a BioLogic VSP

Fig. 1. X-ray diffraction patterns for as-prepared anatase (A0; I), and anatase samples
Potentiostat/Galvanostat/EIS system using Swagelok-type half-cells
annealed at 200 # C (A200; II), 300 # C (A300; III), and 400 # C (A400, IV).
assembled in an argon-filled glovebox. The as-prepared and
annealed anatase samples were used as active materials for the
working electrodes, while lithium foil was used as the counter is described in detail in the Supporting information section of this
electrode. Electrode sheets were fabricated using 80% active ma- paper). Table 1 shows the key physical characterization data for the
terial, 10% acetylene black, and 10% poly(vinylidene) fluoride (PVDF, different anatase nanomaterials. A0 is the least crystalline (70.1%),
HSV 900, Solvay) binder dissolved in 1-methyl 2-pyrrolidone (NMP, while A400 is the most (90.5%). Using Scherrer's equation, the
Sigma Aldrich) solvent. The resulting slurry was then doctor bladed crystallite size was determined for each sample. As with the
onto aluminum foil sheets and dried at 120 # C for 24 h. The elec- percent crystallinity, the crystallite size increases as the anneal
trolyte used in this work was a solution of 1 M LiPF6 in EC:DMC:DEC temperature increases.
(2:2:1 by weight) mixed solution (BASF). In addition to Scherrer's equation used for nanocrystallite size
Galvanostatic charge/discharge testing was performed on the determination that is known to be associated with error [31], the
cells with cutoff voltages at 3 and 1 V. A specific current of samples were characterized using Brunauer-Emmett-Teller (BET)
168 mA g"1 (corresponding to 1C rate) was used for both charging surface area and transmission electron microscopy (TEM) tech-
and discharging. Cells were first discharged to insert Li-ions into niques. The BET results (Table 1) show that as the anatase is
pristine TiO2, and then charged. Testing at 1C rate was carried out annealed it undergoes a decrease in surface area from 222.6 m2 g"1
over 100 cycles for as-prepared anatase only. For all samples, (A0) to 99.6 m2 g"1, as expected, which corresponds to an increase
cycling at 10C rate was performed for 100 cycles (and again for 500 in crystallite size from 6.37 to 14.24 nm. The shape of the isotherm
cycles for as-prepared and 300 # C annealed anatase). Rate capa- associated with the anatase samples is indicative of a non-porous
bility was performed on all samples at specific currents repre- nanosolid, implying that the crystallite size calculated from the
senting 1C, 2C, 5C, 10C, 20C, and back to 1C. Each sample was cycled BET surface area measurements should better represent the true
10 times at 1C then 5 times at the subsequent rates. size and not the Scherrer method calculated size. This is corrobo-
Electrochemical Impedance Spectroscopy (EIS) was performed rated with the TEM micrographs, seen in Fig. 2, which show that the
on each sample to understand the electrode behavior and diffusion anatase crystallites are non-porous spherical nanoparticles. Parti-
at different states of charge (SOC). The technique used alternating cles in the micrograph of A0 measure approximately 6 nm in
current (AC) impedance, in which AC modulation of known ampli- diameter, while A200, A300, and A400, measure approximately 9,
tude (5 mV) across a frequency range (0.5 mHze1 MHz) is used to 12, and 15 nm in diameter respectively. SAED was also performed
measure the “in phase” and “out of phase” components. The diffu- on each sample to verify the phase of the particles as anatase
sion coefficient for lithium ion transport was calculated using the (Fig. S2).
Warburg impedance as determined from EIS measurements [30]. The decrease in surface area seen in the annealed samples
should have an effect on the non-faradaic storage associated with
3. Results and discussion TiO2 [21,32]. By lowering the surface area, there will be less surface
sites for Li-ions to store. However, by increasing the crystallinity,
3.1. Nanoanatase characterization diffusion should proceed more rapidly as the structure is more
ordered, thus increasing the capacity seen at high rate. Annealing
Fig. 1 shows the X-ray diffraction (XRD) patterns of the syn-
thesized anatase, both as-prepared and after annealing at 200 # C,
300 # C, and 400 # C (referred to as A0, A200, A300, and A400, Table 1
respectively). The XRD patterns for all of the samples show a main Key physical characterization values for the different anatase nanomaterials.
peak at 2q z 25.3# , which is in good agreement with the peak Sample Percent Crystallite sizea Surface area BET crystallite size
corresponding to the (101) plane of anatase. The other peaks seen name crystallinity [nm] [m2g"1] [nm]
in the diffraction patterns correspond to anatase, thus there is no A0 70.1% 9.578 222.6 6.37
evidence of impurity peaks suggesting the formation of a single A200 74.3% 10.322 158.7 8.93
phase. Percent crystallinity was determined mathematically by A300 89.5% 12.513 120.1 11.8
comparing the amorphous area with the crystalline area under the A400 90.5% 15.062 99.6 14.24

pattern (individual patterns are available in Fig. S1 and the method a

Based on Scherrer method.
 M.J. Sussman et al. / Journal of Power Sources 272 (2014) 58e67 61

Fig. 2. Transmission electron micrographs of A0 (a), A200 (b), A300 (c), and A400 (d) anatase samples.

has been suggested to lead to high-rate capacity and cycling per- were obtained at lower rate or with particles having a unique
formance in TiO2 as well as other materials [26,33,34]. In TiO2 the morphology [35e37]. Sudant et al. obtained a 165 mAh g"1
increase in capacity due to high crystallinity is postulated by Wang reversible capacity for 6 nm nanoparticles with 223 m2 g"1 surface
et al. to be due to a more ordered phase transition from tetragonal area, however, this was obtained at C/20 rate, and after 50 cycles
TiO2 to orthorhombic Li0.5TiO2 [26]. However, this may have also was only at 90% capacity retention [35]. Zhang et al. [37] achieved
come from the fact that they used 10 nm mesoporous nano- 172 mAh g"1 but were using mesoporous hollow spheres as their
particles, allowing for an increase in surface sites available for active material. In all these studies the authors invoked an inter-
lithium storage rather than increased diffusion-based storage. This calation type of storage arguing the benefits of nanostructuring in
is deduced from the voltage profiles associated with the lithiation terms of a short diffusion length for lithium within the particle.
of their particles at various rates [21]. As the rate increased, the However, as already mentioned earlier pseudocapacitive storage
diffusive plateau was reduced until, at high rate, the profile con- may also play a role in this system especially at high rate (10C), a
sisted of an initial voltage drop then a sloping line that could be central issue of research focus in the present work. In order to fully
attributed to pseudocapacitive behavior [14]. understand the relationship between crystallinity and surface area
of nanoanatase on type of Li-ion storage, all of the samples were
3.2. Investigation of storage mechanisms cycled at 10C rate for 100 cycles (Fig. 3). The Figure shows that A0
achieves the highest capacity after 100 cycles at 115 mAh g"1. While
3.2.1. Galvanostatic charge/discharge the trend in capacity mirrors that of A0, A200 has a much lower
The charge/discharge performance of A0 was evaluated to un- initial capacity and after 100 cycles, its final capacity is 90 mAh g"1.
derstand the lithiation behavior of the as-prepared nanoanatase. A300 has a different trend in capacity than A0 and A200. After the
Galvanostatic cycling was carried out (Fig. S3) between 3 V (charge initial cycle, the capacity retention is higher. The initial cycle has a
limit) and 1 V (discharge limit). During the formation cycle, A0 lower reversible capacity than A0 (121 mAh g"1 compared to
achieved a specific capacity of 230 mAh g"1 after lithiation, but the 134 mAh g"1), but the retention of the next cycles is much higher. In
delithiation capacity was 195 mAh g"1. This is in good agreement the following cycles, while there is some capacity fading, the
with the work of Jiang et al., who achieved similar lithiation and reversible capacity stabilizes allowing A300 to achieve a capacity of
delithiation capacities for 6 nm anatase nanoparticles [21]. How- 105 mAh g"1 after 100 cycles. A400 again shows the same trend as
ever, after the initial lithiation and delithiation, there is capacity A300, but with a much lower initial capacity of 85 mAh g"1. After
fading before the capacity stabilizes. The coulombic efficiency is 100 cycles, A400 achieves only 65 mAh g"1, the least of the 4
almost 100% over 100 cycles leading to a specific reversible capacity samples.
of ~150 mAh g"1. In other works focusing on 6 nm anatase nano- As the crystallinity increases, there is an initial decrease in ca-
material, a higher reversible capacity is reported, but these values pacity, associated with the loss of surface area. The crystallinity has
62 M.J. Sussman et al. / Journal of Power Sources 272 (2014) 58e67

Fig. 3. Specific capacity and coulombic efficiency data for all samples cycled 100 times at 10C rate (C ¼ 168 mAh g"1) between 3 and 1 V.

only increased by ~4%, while the surface area has decreased by nanoanatase materials achieve the same reversible capacity,
~30% as we move from the as-prepared material to that annealed at although their mechanisms of Li-ion storage appear to differ with
200 # C. The significant loss of surface area is apparently the cause of A0 dominated by surface storage (pseudocapacitive) and A300 by
the capacity loss, as the trend in the capacity follows the same as A0 diffusion storage (intercalation).
implying that surface-based storage dominates in the lower tem- By examining the voltage profiles of A0 and A300 at each rate
perature (lower crystallinity) nanoanatase. A300, however shows a (Fig. 4b and c), the mechanisms associated with lithium storage can
20% increase in crystallinity as compared to A0, and while the be deduced [14]. The diffusion-based lithiation (intercalation) of
surface area has decreased significantly, it seems that in this case, anatase occurs in multiple steps [20,25]. In order to understand the
the increased crystallinity makes up for the loss of surface area various mechanisms involved in the process we focus on the
pointing to the increasing importance of diffusion-based storage. plateau and tail of the voltage profile. In Fig. 4b and c, the voltage
Further, with the change in the trend of capacity, there is an profile at low rate has a voltage drop to 1.75 V, where the Li-ions
apparent difference between the lithiation of A0 and A300. In insert into the tetrahedral interstitial sites forming a solid solu-
contrast to A300, A400 shows limited increases in crystallinity, tion of a-Li0.1TiO2 [20]. At 1.75 V a plateau occurs representative of
with a major loss in surface area. The capacity trend follows that of bi-phasic diffusion within the TiO2 crystal structure. Here, the
A300, but with the loss in surface area, the capacity decreases anatase undergoes a shift from tetragonal TiO2 to orthorhombic b-
significantly. Li0.5TiO2 [18,20], as the Li-ions diffuse from the tetrahedral to the
As A0 and A300 showed similar capacities after 100 cycles at octahedral sites in the anatase. The diffusion processes end with a
10C, they were compared further in terms of rate capability sloping tail until 1.5 V, which is indicative that b-Li0.5TiO2 is the
(Fig. 4a). The first 10 cycles (at 1C) show that A0 has a slightly stable structure [20]. Below 1.5 V the storage occurs in the surface
higher formation cycle reversible capacity (195 mAh g"1) compared sites [21]. Upon charging (delithiation), the surface sites are the first
to A300 (189 mAh g"1). However, after the 3rd cycle, A300 has a to delithiate, indicated by the curved region up to 2 V, where
higher reversible capacity as it undergoes less capacity fading at another diffusion plateau occurs as Li-ions are removed from the
low rate compared to A0. At 2C, A300 maintains, slightly, a higher lattice once again. In both of these cases, the amount of capacity
capacity as compared to A0. However, at 5C, A0 shows a higher attributed to surface storage will be dictated by the surface sites
capacity than A300. Interestingly, the difference between A300 and available based on the surface area of the particles [38].
A0 at both 2C and 5C is 5 mAh g"1. As the specific current increases, The voltage profile for A0 has a short plateau during its lith-
the difference in capacity increases further to 50 mAh g"1 at 20C. iation, even at low current, as compared to A300, which has half of
However, when the current is brought back down to 1C, A300 again its profile occurring in the plateau. As previously mentioned, the
has a higher capacity than A0. The shift in capacity difference in- plateau indicates diffusion-based intercalation storage. When
dicates that there is some difference in the storage mechanism that comparing the two, the majority of the storage in A0 comes from
the two nanoanatase materials lithiate/delithiate. A0 and A300 surface area-controlled mechanisms, while in A300 it is based
were cycled at 1C rate 100 times to compare low rate performance primarily on diffusion control. While intuitively it can be assumed
(data shown in Fig. S3). As observed in the rate capability study, that this has to do with the difference in both crystallinity and
here again there is a much smaller initial capacity drop, followed by surface area between the two, it is pertinent to examine the dif-
consistently higher capacity after the third cycle for the A300 ference in diffusivity between A0 and A300 at various points in the
crystalline anatase material in comparison to the poorly-crystalline lithiation to understand the dependence of capacity on the storage
as-prepared A0 nanoanatase. While the increase in capacity is not mechanism.
significant, as cycle numbers increase, the difference in capacity
increases showing a higher capacity retention as compared to A0. 3.2.2. Electrochemical impedance spectroscopy
The higher capacity is likely due to increased diffusion-based lith- The diffusivity of Liþ in TiO2 was determined through electro-
iation. As the rate increases, the mechanism associated with the chemical impedance spectroscopy (EIS). After cycling 5 times at 1C
lithiation of A0 is less rate-limited apparently due to predominant rate, the electrodes were lithiated to various states of charge (SOC)
surface-based storage. There exists a rate threshold (approximately corresponding to 0, 25, 50, 75, and 100% capacity. This capacity
occurring at 4C rate as shown in Fig. S4) at which A0 and A300 translates to a fraction x in LixTiO2 corresponding respectively to
 M.J. Sussman et al. / Journal of Power Sources 272 (2014) 58e67 63

linear tail. This is representative of the Warburg impedance (ZW),

which translates directly to the diffusivity of Liþ into TiO2.
The equivalent circuit in Fig. S5 was used to fit the spectra in
Fig. 5. Re corresponds to the electrolyte resistance, but also includes
the uncompensated bulk resistances from the separator and
counter electrode. The C1R1 parallel element has been explained to
represent the SEI film, electrode roughness, and inhomogeneity of
the lithiation reaction on the surface for both TiO2 and other ma-
terials [30,39]. However, SEI formation in TiO2 does not normally
occur for cutoff voltages above 1 V and are a result of the decom-
position of the electrolyte [25]. Instead, the roughness and in-
homogeneity of lithiation are represented by the parallel element.
The other parallel element is a Randall circuit comprised of the
double layer capacitance, Cdl, the charge transfer resistance, Rct, and
the Warburg impedance, ZW. This circuit describes the two time
constants (two semicircles) and linear tail displayed in the EIS
spectra.
A table of the values associated with the various circuit elements
(Table S1) can be found in the Supporting information along with
plots of the trends in the data throughout the lithiation of both A0
and A300 (Fig. S6). For both A0 and A300, Re remains essentially
constant. This is expected as the electrolyte and separator remain
constant, and the counter electrode is lithium metal, thus is
considered an infinite lithium supply, and therefore will not have
major resistance fluctuation. The capacitive part of the C1R1
element remains constant as well, in contrast the resistive portion
of the element undergoes a significant increase from 0% lithiation
to 25% lithiation for both A0 and A300. However, the resistance
drops again at 50% lithiation, though A0 does not see as significant
of a decrease as A300. Due to the inhomogeneity of the crystal
structure, there are fewer distinct lithiation zones (points) devel-
oping during the diffusion-based storage in the bi-phasic plateau of
the voltage profile. At 50% lithiation, the resistance has decreased
for both samples but A0 decreases only slightly. This may be again
due to the lower crystallinity. As the lithiation mechanism shifts
from diffusion-based to surface area control [21], the in-
homogeneity still plays a significant role in the storage of lithium
within the lattice. Only at 75% and full lithiation, where the ma-
jority storage occurs through surface area control, does the resis-
tance here drop to its original level.
The charge transfer resistance and double layer capacitance of A0
and A300 show contrasting trends. For A300, charge transfer
resistance is constant at all points during lithiation, the double layer
capacitance, however, sees an initial drop, at 25% lithiation, which
corresponds to the insertion of ions into the lattice, but then an
exponential increase in capacitance at 75% lithiation. A0, on the
other hand, has a major increase in charge transfer resistance at 75%
lithiation and only a minor increase in capacitance leading to a
plateaued value at full lithiation. This indicates that for A0, there is
Fig. 4. Rate capability data for A0 and A300 showing the reversible capacity at 1C, 2C, pseudocapacitive storage in the surface layers, as indicated by the
5C, 10C, 20C and back at 1C (a) and the corresponding voltage profiles for A0 (b) and increased charge transfer resistance, most likely due to repulsive
A300 (c).
forces from LieLi interaction [40,41]. For A300, however, the lattice
more easily intercalates lithium within the fully developed crystal-
line structure, thus as it approaches the full extent of lithium
x ¼ 0, 0.125, 0.25, 0.375, and 0.5. The impedance was measured insertion, the Li-ions would tend to align on the surface rather than
from 1 MHz to 0.5 mHz, after the cell had reached the open circuit inserting into surface layers. This causes a build up of charge, indi-
voltage (OCV). The Nyquist plots for A0 and A300 at the various cated by the increase in capacitance. However, it is likely that charge
stages of lithiation can be seen in Fig. 5. The Nyquist plots can be is not entering the surface layers as the charge transfer resistance
broken down into two parts, a semi-circular part and a linear part. remains constant, thus with lower surface area, the contribution
The spectra exhibit either two semi-circles or a depressed semi- from pseudocapacitance decreases. In A0, because the ions can both
circle between 1 MHz and 4 Hz. This is indicative of two time insert into the disordered surface layers as well as align on the
constants, as shown by the equivalent circuit in Fig. S4. Based on the surface, the pseudocapacitive contribution is much greater, whereas
values associated with the capacitance and resistance of the first for A300 is much lower due to reduced surface area.
time constant, the two semi-circles can coalesce and form a single, The Warburg impedance describes the diffusion-based re-
depressed semicircle. From 4 Hz to 0.5 mHz, the spectra exhibit a sistances within the electrode from which the Li-ion diffusion
64 M.J. Sussman et al. / Journal of Power Sources 272 (2014) 58e67

Fig. 5. Nyquist plots for A0 (a) and A300 (b) at 0, 25, 50, 75, and 100% lithiation. Zoomed in view of high frequency impedance for A0 (c) and A300 (d).

coefficient can be obtained [30]. The diffusivity of Li-ions in TiO2 significantly, and by 100% lithiation, the diffusivity is the same as at
was calculated using the equation derived by Ho et al. [42]: 0% lithiation.
With the increase in crystallinity (A300), there is an increase in
diffusivity of Li-ions in anatase. This is in good agreement with
#$ %$ %&2 Wang et al. [5Wang2012a], as the more ordered crystal can more
1 VM dE
DLi ¼ (2) easily change from tetragonal to orthorhombic. By increasing
2 SFA dx
crystallinity, the ions diffuse more effectively as evidenced by the
higher diffusion coefficient determined for A300, increasing the
where VM is the molar volume of the active material, S is the contact
amount of possible storage due to bi-phasic diffusion. However, the
area between surface and electrolyte, F is Faraday's constant, A is a
reliance on diffusion as the major mechanism for charge storage
pre-exponential factor calculated from the Warburg impedance,
poses a problem for cyclability as in more detail is discussed later.
and dE/dx is the slope obtained from a plot of lithium content vs.
Fig. 4b and c shows an increase in polarization between lithiation
OCV. The plots used to calculate the diffusion coefficient are seen in
and delithiation. The polarization is caused by the activation energy
the Supporting information of this paper. Fig. S7a shows the plot of
associated with the change in crystal structure during the insertion
inserted lithium fraction (x in LixTiO2) vs. OCV for the various
and removal of Li-ions. As the current is higher (lithiation and
lithiation states. The slope of the line provides the value of dE/dx for
delithiation at faster rate) there is more resistance to phase change.
A0 and A300. The pre-exponential factor, A, changes with state of
At higher rate, as diffusion occurs, the voltage difference between
charge and diffusivity so it was determined by plotting u"1/2 vs Z00
lithiation and delithiation diffusion increases significantly.
(Fig. S7b and c) for frequencies between 10 and 0.5 mHz at each of
the various lithiation states for A0 and A300.
The important values and calculated diffusivity at each lithiation 3.2.3. Diffusivity effects on pseudocapacitance
point are summarized in Table S2, while a plot of lithiation state vs. Having determined that enhanced crystallinity in A300 induced
diffusivity for A0 and A300 can be seen in Fig. S8. The initial by annealing led to increased Li-ion diffusivity begs the question as
diffusivity for both A0 and A300 is approximately 10"16 cm2 s"1. At to how this affects lithiation at different charging/discharging rate.
25% lithiation, the voltage profile shows that bi-phasic diffusion Differential capacity (DQ/DV) analysis allows for a more detailed
dominates. The diffusivity increases for both A0 and A300, how- investigation into the contributions of diffusion and surface area
ever, A300 sees an increase of 2 orders of magnitude based storage to capacity by translating the plateaus associated
(10"14 cm2 s"1), while A0 only sees an increase of an order of with diffusion into peaks [14]. For a voltage profile showing pure
magnitude (10"15 cm2 s"1). At 50% lithiation, both samples are no diffusion-based storage, sharp peaks at the oxidation and reduction
longer in the bi-phasic plateau, and diffusivity decreases potentials will be seen while the rest of the plot will be
 M.J. Sussman et al. / Journal of Power Sources 272 (2014) 58e67 65

approximately at zero. When DQ/DV is greater than zero, but not a which may be explained by the lower contribution of diffusive
peak, it can be assumed that there is some sort of pseudocapacitive (intercalation type) storage. By relying less on lithium diffusion,
charge storage. In the case of TiO2 this happens in the voltage range and instead storing in surface sites, A0 does not undergo a large-
after reduction and before oxidation. Fig. 6 shows the DQ/DV curves scale phase change, therefore experiencing less resistance.
for A0 and A300 at various rates, as well as a comparison between Further, at 10C, there is a stark contrast between the DQ/DV
the two at 1C and 10C. Comparing the curves in Fig. 6a and b, we curves of the A0 and A300 nanoanatase materials (Fig. 6d). Aside
note the peaks, which correspond to diffusion-based capacity are from the increased polarization, the contribution from pseudoca-
extremely tall for A300, and that pseudocapacitive storage is almost pacitive storage is almost zero for A300, especially during dein-
non-existent. Further, as the rate increases, the distance between tercalation. A300, however, has a higher diffusion peak in the
the peaks increases, representing an increase in polarization be- deintercalation half of the curve. It is postulated that because of
tween intercalation and deintercalation. While both A0 and A300 decreased surface area, A300 has fewer sites for pseudocapacitive
have intercalation and deintercalation peaks at approximately 1.7 storage; as a consequence the bulk of its reversible capacity comes
and 2.0 V, respectively, at 1C rate, A0 sees only 0.2 V overall increase from diffusion-based storage. However, with the insertion of
in polarization at 10C, while the increase for A300 is 0.6 V. lithium into the lattice, and the phase-change associated with that
When a closer, side-by-side examination of the DQ/DV curves insertion, the stresses on the crystal structure of A300 may
for A0 and A300 at 1C rate is performed (Fig. 6c), a second peak can augment significantly more than in the case of A0 that relies mainly
be seen at 1.5 V. While the peak is just a hump for A0, it is clearly on pseudocapacitive storage, an issue that can impact the material's
seen in A300. This hump is associated with the insertion of lithium cyclability as discussed next.
as the stable b-Li0.5TiO2 phase, as translated from the voltage pro-
file. Theoretically, this would be followed by another plateau where 3.2.4. Impact of mode of storage on cyclability
lithium would insert as a second bi-phasic region, this time con- In order to determine the effect of mode of storage on nano-
sisting of b-Li0.5TiO2 and g-LiTiO2 [20]. However, as the DQ/DV plots anatase crystal structure and capacity retention, A0 (predominantly
show, this stable single phase diffusion is followed by surface surface-based storage) and A300 (predominantly diffusion-based
storage. At 1C, the capacitive contribution to capacity is approxi- storage) were cycled at 10C for 500 cycles. Fig. 7 shows that for
mately the same during the lithiation. During delithiation, there is the 1st 100 cycles, A0 and A300 exhibit macroscopically similar
less capacitive contribution in A300, leading to the sharp peak seen capacity retention behavior. A300 shows a stable capacity that is
in Fig. 6a (Fig. 6c has been truncated to show the capacitive con- slightly higher than A0, then decreases to just below it. After 100
tributions). Interestingly, the polarization between intercalation cycles, however, the capacity of A300 drops dramatically until it
and deintercalation is smaller for A0 than A300, even at low rate, reaches a capacity of approximately 45 mAh g"1. A0, however,

Fig. 6. Differential capacity (DQ/DV) plots of A0 (a) and A300 (b) at 1C, 2C, 5C, 10C, and 20C. Comparison of DQ/DV of A0 and A300 at 1C (c) and 10C (d) rates.
66 M.J. Sussman et al. / Journal of Power Sources 272 (2014) 58e67

Fig. 7. Specific capacity and coulombic efficiency for A0 and A300 lithiated at 10C rate (1C ¼ 168 mA g"1) over 500 cycles.

remains constant until approximately 175 cycles after which its diffusion-based storage. Furthermore by lowering the charging/
capacity decreases slowly but always remaining higher than that of discharging rate polarization is reduced hence avoiding the drastic
A300 (70 mAh g"1 vs 45 mAh g"1) capacity loss due to electrolyte interaction. This was demonstrated
The breakdown in capacity exhibited by the A300 material is at by cycling the A0 material for 500 times but this time at 2C. As can
all likelihood linked may be due to the increased crystallinity be seen with the cycling data plotted in Fig. S11, a much more stable
combined with the loss of surface area. In both A0 and A300, the capacity retention is achieved with only a modest capacity fade
differential voltage profiles (Fig. S8a and b) change significantly (~20 mAh g"1) linked to typical coulombic efficiency behavior.
over 500 cycles. Where the first 100þ cycles show a distinct Further alleviation of the capacity fading problem may be achieved
diffusion plateau, though short due to the high current and fast Li- by cycling to a cutoff voltage of 1.2 V, as described by Bresser et al.
ion intercalation and deintercalation transport, the plateau even- [25]. Alternatively fabrication of anatase electrode structures that
tually disappears for both materials and the curves assume the maximize the benefits of increased surface area and crystallinity
characteristic shape of essentially sole pseudocapacitive type of without sacrificing one for the other may provide a stable and
storage [14,15]. In the case of A300, the lower surface area leads to highly reversible Li-ion storage performance.
low storage at the end of 500 cycles, while A0 retains a majority of
its capacity. The question still remains as to why this capacity loss 4. Conclusion
occurs. As a first hypothesis it was thought this to stem from a
structural change that blocks Li-ion diffusion channels. However, A new variety of phase pure nano-anatase TiO2 material has
XRD patterns of A0 and A300 before and after 500 cycles (Fig. S10) been synthesized via a scalable aqueous solution process carried
show that there is only a minor peak shift for both samples. This out in a continuous-stirred tank reactor at 80 # C. The anatase has a
could be associated with small changes in lattice parameter related small nanocrystallite size (~6 nm) and large surface area
to some (but not all) irreversible capacity loss. As no significant (222.6 m2 g"1), which translates to an excellent candidate for high
structural change was detected the other hypothesis made to rate lithiation and delithiation performance. As-synthesized
explain the capacity loss relates to possible interaction with the anatase (A0) provided a stable reversible capacity at 1C
electrolyte. (C ¼ 168 mA g"1) rate, achieving 150 mAh g"1 over 100 cycles. By
It was seen in Fig. 5b and c that with increased rate, there is a annealing, crystallinity was increased, while surface area decreased
large increase in polarization. The resistance brought on by the allowing for the investigation of the contributions of each toward
increased rates (a larger amount seen in A300 as compared to A0) lithiation and delithiation capacity. A0 and the annealed samples
may cause localized voltage drops to below the 1 V cutoff voltage. were cycled at high rate (10C) over 100 cycles, where A0 achieved
As explained by Bresser et al., a voltages below 1 V leads to 115 mAh g"1 and the sample annealed at 300 # C (A300) achieved
passivation layer formation due to interaction with the electrolyte the highest reversible capacity (of the annealed samples) of
[25]. Because of this apparent voltage instability, the electrolyte 105 mAh g"1. When these two samples were compared at various
becomes more reactive and catalyzes the formation of a passivation rates, it was found that at low rate (1C and 2C) A300 had a higher
layer, whose presumably increasing thickness is responsible for the reversible capacity than A0, due to increased diffusion-based stor-
increased capacity fading. The same phenomenon is seen during age as the major mechanism for Li-ion transport. At high rate
the rate capability testing when the rate is increased to 20C. Even however, A0 had much higher capacity due to the increased surface
though it is only for 5 cycles, the polarization is so large that a area and ability to store Li-ions pseudocapacitively.
20 mAh g"1 capacity fade from the 1st to the 5th cycle occurs, Using electrochemical impedance spectroscopy, it was deter-
however, upon returning to 1C, the polarization is essentially mined that with increased crystallinity, the diffusivity of Li-ions
removed, and the capacity fading is negligible. This phenomenon is within the lattice increases by an order of magnitude resulting in
seen only at high rate because of the large increase in polarization A300's reliance mainly on diffusion-based intercalation as storage
associated with attempting to diffuse lithium into the crystal. This mechanism. As rate increases, the increased crystallinity and
is why the capacity loss for A0 is only approximately 30 mAh g"1 as decreased surface area causes increased polarization. This mani-
compared to 60 mAh g"1 for A300. By taking advantage of pseu- fests in long-term stability loss due to apparent electrolyte reac-
docapacitance, A0 exhibits much less polarization as it relies less on tivity and breakdown. By understanding the effects of anatase
 M.J. Sussman et al. / Journal of Power Sources 272 (2014) 58e67 67

crystallinity and surface area on Li-ion storage, the electrode [17] T. Brezesinski, J. Wang, J. Polleux, B. Dunn, S.H. Tolbert, J. Am. Chem. Soc. 131
(2009) 1802.
fabrication process can be optimized to provide the best combi-
[18] R. van de Krol, A. Goossens, A.A. Meulenkamp, J. Electrochem. Soc. 146 (1999)
nation of properties for highly stable and reversible performance. 3150e3154.
[19] W.J.H. Borghols, D. Lutzenkirchen-Hecht, U. Haake, E.R.H. van Eck,
Acknowledgements F.M. Mulder, M. Wagemaker, Phys. Chem. Chem. Phys. 11 (2009) 5742e5748.
[20] A.A. Belak, Y. Wang, A. Van der Ven, Chem. Mater. 24 (2012) 2894e2898.
[21] C. Jiang, M. Wei, Z. Qi, T. Kudo, I. Honma, H. Zhou, J. Power Sources 166 (2007)
The authors would like to thank Dr. Fuqiang Guo for his help 239e243.
with the synthesis tests. This research is supported with a strategic [22] J.W. Kang, D.H. Kim, V. Mathew, J.S. Lim, J.H. Gim, J. Kim, J. Electrochem. Soc.
158 (2011) A59eA62.
project grant from the Natural Sciences and Engineering Research [23] J.S. Chen, X.W. Lou, Electrochem. Commun. 11 (2009) 2332e2335.
Council of Canada (NSERC) (STPGP 396562-2010) and sponsored by [24] H.-W. Shim, D.K. Lee, I.-S. Cho, K.S. Hong, D.-W. Kim, Nanotechnology 21
Hydro-Quebec's Energy Conversion and Storage department. (2010) 255706e255714.
[25] D. Bresser, E. Paillard, E. Binetti, S. Krueger, M. Striccoli, M. Winter, S. Passerini,
J. Power Sources 206 (2012) 301e309.
Appendix A. Supplementary data [26] J. Wang, Y. Zhou, Y. Hu, R. O'Hayre, Z. Shao, J. Mater. Sci. 48 (2013)
2733e2742.
[27] W. Li, F. Wang, S. Feng, J. Wang, Z. Sun, B. Li, Y. Li, J. Yang, A.A. Elzatahry, Y. Xia,
Supplementary data related to this article can be found at http:// D. Zhao, J. Am. Chem. Soc. 135 (2013) 18300e18303.
dx.doi.org/10.1016/j.jpowsour.2014.08.050. [28] C. Charbonneau, R. Gauvin, G.P. Demopoulos, J. Electrochem. Soc. 158 (2011)
H224eH231.
[29] A. Yasin, F. Guo, G. P. Demopoulos, Materials Research Society Spring Meeting
References
(2013).
[30] H. Joachin, T.D. Kaun, K. Zaghib, J. Prakash, J. Electrochem. Soc. 156 (2009)
[1] Z. Yang, D. Choi, S. Kerisit, K.M. Rosso, D. Wang, J. Zhang, G. Graff, J. Liu, A401eA406.
J. Power Sources 192 (2009) 588e598. [31] R. Delhez, T.H. Keyser, J.I. Langford, D. Louer, E.J. Mitmeijer, E.J. Sonneveld, in:
[2] C. Jiang, J. Zhang, J. Mater. Sci. Technol. 29 (2013) 97e122. R.A. Young (Ed.), The Reitveld Method, IUCr Monograph #5, Oxford University
[3] Y. Yang, H. Wang, Q. Zhou, M. Kong, H. Ye, G. Yang, Nanoscale 5 (2013) Press, 1995, pp. 132e166.
10267e10274. [32] M. Wagemaker, W.J.H. Borghols, F.M. Mulder, J. Am. Chem. Soc. 129 (2007)
[4] J.S. Chen, X.W. Lou, J. Power Sources 195 (2010) 2905e2908. 4323e4327.
[5] Y.-S. Hu, L. Kienle, Y.-G. Guo, J. Maier, Adv. Mater. 18 (2006) 1421e1426. [33] E. Hosono, T. Kudo, I. Honma, H. Matsuda, H. Zhou, Nano Lett. 9 (2009)
[6] S.K. Panda, S. Lee, W.-S. Yoon, H. Shin, J. Power Sources 249 (2014) 59e65. 1045e1051.
[7] X. Su, Q.-L. Wu, X. Zhan, J. Wu, S. Wei, Z. Guo, J. Mater. Sci. 47 (2012) [34] D. Zhan, Q. Zhang, X. Hu, G. Zhu, T. Peng, Solid State Ionics 239 (2013) 8e14.
2519e2534. [35] G. Sudant, E. Baudrin, D. Larcher, J.-M. Tarascon, J. Mater. Chem. 15 (2005)
[8] D. Dambournet, I. Belharouak, K. Amine, Chem. Mater. 22 (2010) 1173e1179. 1263e1269.
[9] A.R. Armstrong, G. Armstrong, J. Canales, R. Garcia, P.G. Bruce, Adv. Mater. 17 [36] H.-G. Jung, S.W. Oh, J. Ce, N. Jayaprakash, Y.-K. Sun, Electrochem. Comm. 11
(2005) 862e865. (2009) 756e759.
[10] R. Giannuzzi, M. Manca, L. De Marco, M.R. Belviso, A. Cannavale, T. Sibillano, [37] F. Zhang, Y. Zhang, S. Song, H. Zhang, J. Power Sources 196 (2011) 8618e8624.
C. Giannini, P.D. Cozzoli, G. Gigli, Appl. Mater. Interf. 6 (2014) 1933e1943. [38] M.J. Sussman, N. Brodusch, R. Gauvin, G.P. Demopoulos, J. Electrochem. Soc.
[11] J. Wang, J. Polleux, J. Lim, B. Dunn, J. Phys. Chem. C 111 (2007) 14925. 160 (2013) A3100eA3107.
[12] K. Zhu, Q. Wang, J.H. Kim, A.A. Pesaran, A.J. Frank, J. Phys. Chem. C 116 (2012) [39] J. Wang, Y. Zhou, Y. Hu, R. O'Hayre, Z. Shao, J. Phys. Chem. C 115 (2011)
11895. 2529e2536.
[13] B. Laskova, M. Zukalova, A. Zukal, M. Bousa, L. Kavan, J. Power Sources 246 [40] A.G. Dylla, P. Xiao, G. Henkelman, K.J. Stevenson, J. Phys. Chem. Lett. 3 (2012)
(2014) 103. 2015e2019.
[14] A.G. Dylla, G. Henkelman, K.J. Stevenson, Acc. Chem. Res. 46 (2013) [41] M.J. Sussman, M. Celikin, A. Yasin, G.P. Demopoulos, Electrochim. Acta 138
1104e1112. (2014) 215e223.
[15] V. Augustyn, P. Simon, B. Dunn, Energy Environ. Sci. 7 (2014) 1597e1614. [42] C. Ho, I.D. Raisrick, R.A. Huggins, J. Electrochem. Soc. 127 (1980) 343e350.
[16] L. Kavan, J. Solid State Electrochem. 18 (2014) 2297e2306.

View publication stats