LiClO4 Solid Polymer Electrolytes Filled

9196

J. Phys. Chem. C 2010, 114, 9196–9206

Structure and Mobility of PEO/LiClO4 Solid Polymer Electrolytes Filled with Al2O3 Nanoparticles Susan K. Fullerton-Shirey† and Janna K. Maranas* Department of Chemical Engineering, The PennsylVania State UniVersity, UniVersity Park, PennsylVania 16802 ReceiVed: July 13, 2009; ReVised Manuscript ReceiVed: February 21, 2010

The mechanism for improved ionic conductivity in nanoparticle-filled solid polymer electrolytes containing polyethylene oxide [PEO], LiClO4, and Al2O3 is investigated using differential scanning calorimetry [DSC], dielectric spectroscopy, small-angle neutron scattering [SANS], and quasi-elastic neutron scattering [QENS]. We measure samples with ether oxygen to lithium ratios ranging from 14:1 to 8:1 and Al2O3 nanoparticle concentrations ranging from 5 to 25 wt %. The Tg and pure PEO crystal fraction are unaffected by nanoparticle addition, and SANS reveals nanoparticle aggregation, with the extent of aggregation similar in all samples regardless of LiClO4 or Al2O3 concentration. Despite the similarity between samples, nanoparticles improve conductivity at all temperatures, but only at the eutectic concentration (ether oxygen to lithium ratio of 10:1). Our QENS results indicate that a rotation is present in both filled and unfilled samples at all concentrations and is consistent with the rotation of (PEO)6:LiClO4, a channel-like structure that is more conductive than the amorphous equivalent. The rotation becomes more restricted in the presence of nanoparticles. Introduction Rechargeable lithium-ion batteries based on solid polymer electrolytes [SPEs] offer many advantages over batteries designed with liquid or gel electrolytes. The solid polymer is less flammable than a liquid electrolyte, and potentially hazardous short-circuits caused by dendrite growth through the liquid phase are eliminated. In addition to safety benefits, the solidstate nature of the electrolyte permits the design of a light and flexible battery, because a heavy, rigid casing is not required to contain a liquid. Furthermore, a battery based on this design would be more environmentally friendly, reducing end-of-life disposal issues. Despite the advantages, SPEs suffer from one paramount problem: the room-temperature conductivity is insufficient to power a portable device. Accordingly, SPEs have been modified in many ways to improve conductivity. One such modification that retains the solid-state property of the electrolyte while increasing the conductivity is the addition of nanoparticle fillers. Micrometer-sized ceramic fillers were used over 25 years ago to improve the mechanical properties of SPEs,1 but improved conductivity was not observed until 1995 when the size of the filler was reduced to the nanometer scale.2 Since that time, a variety of studies have appeared demonstrating conductivity enhancements in SPEs by a wide range of ceramic materials including TiO2,3,4 SiO2,5-8 Al2O3,2-5,9-12 ZnO,13 BaTiO3,14 PbTiO3,15 and LiNbO3.15 Although conductivity enhancements have been observed for a variety of lithium salts and nanoparticle fillers, the room-temperature conductivity remains too low for practical application. Furthermore, the mechanism by which the nanoparticles enhance conductivity remains unclear, making it difficult to determine which modifications might be successful a priori. Lithium-ion [Li+] transport through SPEs without nanoparticles has been well-studied. Polyethylene oxide [PEO] is the polymer of choice for this application, owing to its flexible * Corresponding author. E-mail: [email protected]. † Present address: Department of Electrical Engineering, Center for Nano Science and Technology, University of Notre Dame, Notre Dame, IN 46556.

backbone and ability to solvate lithium ions. The number of PEO ether oxygen atoms required to solvate a Li+ depends on the lithium concentration and the anion identity. Lithium ions move through the polymer host by breaking and re-forming complexes with ether oxygens, and this transport mechanism is facilitated by the segmental motion of the polymer. It is generally accepted that Li+ mobility occurs most effectively through the amorphous phase, and most research efforts are aimed at improving conductivity through this phase. However, this view has been challenged by results demonstrating greater conductivity through fully crystalline SPEs than through the amorphous equivalent.16 In this case, low molecular weight PEO (1000 g/mol) and LiSbF6 were used, and the crystalline structure was described as pairs of PEO chains forming cylindrical tunnels. The reported structure contains six ether oxygens per Li+, with lithium ions located inside the tunnels and anions outside the tunnels. Multiple tunnels align side by side, with a column of anions between. The authors believe that directed ion transport through the tunnels is responsible for the improved conductivity. Although the tunnels are maintained at high molecular weight,16 conductivity decreases sharply due to tunnel misalignment.17 Long-range ordering of the crystalline tunnels is required for lithium ions to move from one electrode to the other for battery operation, but controlling the morphology with this precision has not yet been achieved. Although Li+ transport through SPEs without nanoparticles has been well studied, the mechanism by which nanoparticles enhance transport in filled systems remains unclear. Because Li+ transport and polymer segmental motion are coupled in amorphous SPEs, nanoparticles could increase charge-carrier mobility by increasing the segmental motion of the polymer. One way to evaluate polymer mobility is by measuring the glass transition temperature [Tg]. Tg is a broad transition, and we therefore regard a change in Tg of greater than five degrees to be a significant change when nanoparticles are added. With this in mind, the Tg can increase,6,2 decrease,10,18,19 or remain unchanged9,5,11,20 with nanoparticle addition for a variety of PEObased SPEs at similar salt and nanoparticle concentrations. These

10.1021/jp906608p  2010 American Chemical Society Published on Web 05/03/2010

Structure and Mobility of Nanoparticle-Filled SPEs results appear independent of salt or nanoparticle identity and do not correlate with increasing or decreasing conductivity. In the case of PEO/LiClO4/R-Al2O3, the Tg has been reported to both decrease10 and remain unchanged11 with nanoparticle addition. Therefore, if nanoparticles influence polymer mobility, the effect cannot be evaluated by Tg measurements alone. One way to measure the molecular-level mobility of a polymer is with quasi-elastic neutron scattering [QENS]. Unlike Tg, mobility is directly measured over time scales relevant to this problem (pico- to nanoseconds). QENS has been used to measure the mobility of a nanoparticle-filled SPE based on the copolymer, 3PEG (trihydroxy poly(ethylene oxide-co-propylene oxide)).21 The results showed that the addition of TiO2 nanoparticles (21 nm diameter) slowed the mobility of the polymer. They estimate that a layer extending 5 nm from the particle surface was immobilized, and the dynamics of the polymer outside this region were unaffected. This observation leads to another possible mechanism for increasing conductivity: nanoparticles decrease polymer mobility by immobilizing ether oxygen atoms on the particle surface, thereby reducing the number of ether oxygens coordinated with lithium ions. For ether oxygen atoms to be attracted to the surface of the nanoparticle, the surface must have acidic sites (electron accepting). Several studies have focused on how the surface chemistry of nanoparticles influences the conductivity,22,10,9 indicating that nanoparticles with acidic surface chemistry increase the conductivity more than those with basic or neutral surface chemistries. The surface chemistry can notably affect conductivity even within different polymorphs of the same ceramic material. For example, it is known from molecular dynamics simulation that the surface of R-Al2O3 is terminated by aluminum atoms (acidic sites), whereas γ-Al2O3 is terminated by aluminum and oxygen atoms (acidic and basic sites).23 A conductivity study of SPEs containing these two polymorphs of alumina show that R-Al2O3 nanoparticles improve conductivity more than γ-Al2O3,11 consistent with the observation that nanoparticles with acidic sites yield higher conductivity values. It is also possible that nanoparticles increase conductivity by promoting cation-anion dissociation at the nanoparticle surface.22 If the nanoparticles are well-dispersed, confinement may impact both the extent of crystallization24 and polymer mobility.25,26 A nanoparticle-filled SPE could be viewed as a confined, capped system (substrate on both sides), provided the nanoparticles are well-dispersed. In fact, such a connection has been made between capped thin films and nanocomposites.27 However, the extent of aggregation is rarely reported for nanoparticle-filled SPEs. One SPE where the extent of nanoparticle aggregation has been characterized is the 3PEG copolymer sample described above. Small-angle neutron scattering [SANS] revealed that nanoparticles were not welldispersed, but instead aggregate into fractal structures two nanoparticles wide and at least 100-400 nm long.28 In this case, we would not expect polymer dynamics to be affected due to strong aggregation in the system. However, this structure suggests another possible way that nanoparticles could increase Li+ mobility: by creating percolating pathways for lithium ions to travel faster than they would through the bulk. Nanoparticles could also increase conductivity by decreasing the fraction of nonconducting crystalline structures; however, this cannot be the exclusive mechanism, because nanoparticles increase conductivity above the melting point2,3,7,12,29,30 and in SPEs that never crystallize.6,9 For this reason, we focus on the temperature range at or above the eutectic temperature, where we study the polymer dynamics

J. Phys. Chem. C, Vol. 114, No. 20, 2010 9197

Figure 1. Phase transitions observed from our DSC (O) and conductivity (2) measurements superimposed on the phase diagram for PEO/LiClO4 found in ref 33. EO:Li represents the ratio of PEO ether oxygen to lithium ions.

in semicrystalline and fully amorphous PEO/LiClO4, filled with γ-Al2O3 nanoparticles. We choose γ-Al2O3 for its neutral surface chemistry, allowing us to focus on polymer mobility and confinement. Nanoparticle size and extent of aggregation are evaluated using SANS and field-emission scanning electron microscopy [FE-SEM]. We also report thermal properties and ionic conductivity. Instead of preparing samples at one lithium concentration and varying the nanoparticle concentration, we vary both the lithium and nanoparticle concentrations to determine whether the nanoparticle’s effectiveness depends on the amount of lithium. Specifically, we measure nanoparticlefilled SPEs at ether oxygen to lithium ratios of 8:1, 10:1, and 14:1, where 10:1 is the eutectic point for PEO/LiClO4 (Figure 1). The nanoparticle concentration is varied from 0 to 25 wt % at each lithium concentration. According to the phase diagram in Figure 1, PEO/LiClO4 can form three crystalline phases: pure PEO, (PEO)6:LiClO4, and (PEO)3:LiClO4. (We will use a “/” to indicate SPE systems and a “:” to indicate crystalline complexes.) Pure PEO will crystallize immediately on cooling, whereas the (PEO)6:LiClO4 phase recrystallizes on the order of days.31 This initial study focuses on the pure PEO crystalline phase, and so we thermally treat our samples to prevent the formation of the (PEO)6:LiClO4 complex. PEO/LiClO4 has been previously measured using QENS by us31 and others;32 however, a nanoparticle-filled SPE based on this system has not been measured by either QENS or SANS. To complement our results, we also measure ionic conductivity using broadband dielectric spectroscopy [BDS] and characterize the Tg and crystal fraction using DSC. Experimental Details Sample Preparation. We prepared unfilled solid polymer electrolytes at ether oxygen to lithium ratios of 4:1, 8:1, 10:1, 14:1, 30:1, and 100:1 using the same procedure described in ref 31. To prepare filled samples, we added Al2O3 nanoparticles (Alfa Aesar, diameter 11 ( 3 nm) to the 8:1, 10:1, and 14:1

9198

J. Phys. Chem. C, Vol. 114, No. 20, 2010

Fullerton-Shirey and Maranas

TABLE 1: Heat Treatment Conditions before and during Measurementa premeasurement

intermediate

measurement

technique

T (°C)

time (h)

environment

time (min) at 22 °C

T (°C)

time (min)

environment

SANS QENS DSC BDS

100 100 100 100

24 24 24 24

vacuum vacuum vacuum vacuum

10 10 10 30

75, 50, 22 75, 50, 22 -90-200 22-100

60 60 N/A 10

N2 N2 N2 N2

a The premeasurement and measurement times represent how long the sample was held at each temperature. The temperatures in the “measurement” column are provided in the order the measurements were made. The intermediate time represents how long the sample was handled at room temperature between premeasurement and measurement.

samples in concentrations of 5, 10, and 25 wt %. We also prepared pure PEO and PEO + 10 wt % Al2O3 (no LiClO4) samples. The Al2O3 nanoparticles were first dried in a vacuum oven at 120 °C for 24 h and then added to a mixture of PEO, LiClO4, and anhydrous acetonitrile. The solutions were covered and mixed for 24 h, followed by 1 h of sonication. After sonication, we removed the covers, allowing the solvent to evaporate while mixing. We dried all samples in a vacuum oven at 75 °C for 5 days. After drying, the samples were hot-pressed at 100 °C to the appropriate thickness for BDS, SANS, and QENS measurements, as described in ref 31. Thermal Treatment. We thermally treat our samples so that pure PEO is the only crystalline complex present in the nanoparticle-filled samples. We previously demonstrated that 3 days are required for the (PEO)6:LiClO4 phase to form at room temperature.31 Because we measure our samples within several hours after cooling from 100 °C, we do not expect any contribution from the (PEO)6:LiClO4 complex. We discuss the individual heat treatments in each section and provide a summary in Table 1. Thermal Analysis and Conductivity Measurements. We measured the Tg and crystal fraction of all samples with a TA Instruments Q1000 differential scanning calorimeter, calibrated with an indium standard. Sample weights were 8-10 mg, and measurements were performed with a heating rate of 10 °C/ min and a cooling rate of 5 °C/min. We calculated the PEO crystal fraction based on the perfect heat of fusion for PEO, 203 J/g.34 We used BDS to measure the ionic conductivity, using the same procedure described in ref 31. Neutron Scattering. We performed neutron scattering measurements using three instruments at the NIST Center for Neutron Research in Gaithersburg, MD. The NG-3 SANS instrument was used to measure structure, and the dynamics were measured using the Disc-chopper time-of-flight spectrometer [DCS] and the high-flux backscattering spectrometer [HFBS]. Within a SANS experiment, the coherent contribution gives information on structure, whereas the incoherent contribution adds a structureless background. Contrast arises from a scattering length density difference between Al2O3 nanoparticles and amorphous PEO/LiClO4. In QENS, mobility is detected because scattering is dominated by the incoherent scattering of hydrogen atoms, reflecting the self-motion of PEO. The details of the SANS and QENS measurements are the same as those described in ref 31. We reduced and analyzed the SANS data with macros developed at NIST using IGOR Pro software.35 The QENS data was reduced using DAVE, a data analysis software developed at NIST.36 Results and Discussion Conductivity as a Function of Nanoparticle and LiClO4 Concentration. We measure the conductivity during a heatcool-heat cycle, and the results of the second heating are

illustrated in Figure 2. Before we describe the influence of nanoparticles on conductivity, we examine some general trends as a function of temperature. The conductivity of the 8:1 sample decreases smoothly with decreasing temperature, whereas the 10:1 and 14:1 samples experience a sharp decrease at 50 and 60 °C, respectively. This temperature dependence is consistent with the phase diagram (Figure 1) where pure PEO crystallizes at 50 °C for the 10:1 concentration (the eutectic point) and 60 °C for 14:1. Decreasing conductivity in the presence of pure crystalline PEO demonstrates that the crystal structures block pathways for ion conduction and/or decrease polymer mobility in amorphous domains. Decreased mobility occurs when LiClO4 is expelled into the nearby amorphous domains during pure PEO crystallization. In addition, a thermal hysteresis is observed at 10:1 and 14:1, a consequence of holding the samples long enough at each temperature for pure PEO to partially recrystallize. Nanoparticles affect conductivity most significantly at the eutectic concentration (10:1), yielding the highest conductivity at 5 wt % Al2O3. The nanoparticles have no effect at increased LiClO4 concentration (8:1), and little effect at decreased concentration (14:1). We notice that conductivity is normally maximized at salt concentrations near the eutectic concentration in unfilled SPEs with a eutectic point (e.g., 11:1 for PEO/ LiTFSI37,38 and 22:1 for PEO/LiAsF633). When nanoparticles are added, our data reveal that nanoparticles have a stronger effect on conductivity near the eutectic composition. This has not been reported previously, because nanoparticle-filled SPEs are normally measured only at the lithium concentration that yields the maximum conductivity. Thermal Analysis. Our conductivity data suggest that nanoparticles do not improve conductivity by affecting crystallization, because the shape of the conductivity curve is the same below the melting point with and without nanoparticles (inset of Figure 2B). We use DSC to more accurately evaluate the influence of nanoparticles on crystallization and to determine the Tg dependence. The crystal fraction [Xc] and Tg are reported in Table 2, where the crystal fraction represents the fraction of the sample occupied by pure PEO lamellae. The data indicate that nanoparticles have little effect on the Tg or extent of pure PEO crystallization, whereas LiClO4 has a strong effect on both. The influence of LiClO4 on Tg is discussed in a previous publication.31 The fact that neither Tg nor Xc vary with nanoparticle addition may seem surprising because nanoparticles often influence both properties. The invariance may result from polymer/surface interactions. For example, nanoparticles will influence Tg so long as the polymer does not wet the surface of the particle.39 PEO likely wets the Al2O3 surface, because of interactions between ether oxygens and acidic sites on the nanoparticle surface. The invariance of Tg and Xc may also result from nanoparticle aggregation. When aggregated, only a fraction of the nanoparticle surface area is available to interact with the polymer. Techniques for dispersing nanoparticles, such as sonication, are

Structure and Mobility of Nanoparticle-Filled SPEs

Figure 2. Conductivity as a function of temperature at LiClO4 concentrations of (A) 8:1, (B) 10:1, and (C) 14:1. Inset in (B) shows the first heating scan at 10:1 for the unfilled and 5 wt % nanoparticle [NP] samples.

sometimes used in nanoparticle-filled SPE studies, and sometimes no special techniques are reported. There appears to be no correlation between conductivity enhancement when steps are taken to disperse nanoparticles and when they are not, suggesting that either nanoparticle dispersion is unimportant, or the dispersion techniques are ineffective. It is difficult to evaluate the effectiveness of the dispersion techniques, because the extent of aggregation is not often characterized. We use FESEM and SANS to characterize the extent of nanoparticle aggregation in our SPEs. Nanoparticle Aggregation. FE-SEM images are illustrated in Figure 3 for filled and unfilled samples with (A and C) and without LiClO4 (B and D). Notice that image C does not contain

J. Phys. Chem. C, Vol. 114, No. 20, 2010 9199 nanoparticles, verifying that the white features in the other images are heavily aggregated nanoparticles. The extent of aggregation appears similar with or without LiClO4 (A versus B), although the sample without LiClO4 is clearer. By enlarging a section of the PEO + 10 wt % Al2O3 sample, we observe a small population of spherical clusters (less than 50 nm), along with a population of elliptical clusters that are highly polydisperse. The long axis of the elliptical clusters appears to have an average size of approximately 100 nm. It is possible that nanoparticle clusters of this size are stable structures, because we observe clusters of this size not only in the PEO melt but also in solutions of acetonitrile and toluene, where the cluster size is measured using dynamic light scattering. The FE-SEM images are useful for characterizing aggregation on length scales larger than tens of nanometers, but the presence of well-dispersed individual nanoparticles cannot be resolved. Furthermore, the FE-SEM images represent one snapshot of one small section of our sample, whereas SANS represents an average of many snapshots and allows us to detect and quantify the amount of dispersed nanoparticles and nanoparticle clusters of size less than 200 nm. We indicate the size range measured by SANS with a circle on the FE-SEM image D. We will refer to aggregates within the size range measured by SANS as “clusters” and those outside the SANS window as “aggregates”. We expect to see a contribution at high q representing the size of individual nanoparticles, and a strong contribution at low q resulting from nanoparticle aggregation. The SANS data for filled samples are illustrated in Figure 4, including the pure nanoparticle powder. We know that the pure powder is highly aggregated, so the fact that the other samples look similar supports extensive aggregation. As expected, the scattered intensity increases with increasing nanoparticle concentration; however, the shape of the data remains unchanged, indicating that the extent of aggregation is unaffected over the size range measured by SANS. The extent of aggregation is also unaffected by LiClO4 concentration, indicating that charge is not accumulating on the nanoparticle surface over this LiClO4 range. We point out that crystallization does not contribute to the scattering in these samples, because they are measured at 80 °C where the polymer is completely amorphous. Furthermore, we know that the scattering arises from nanoparticles and not from other sources, such as LiClO4 clustering, because the scattered intensity of the filled samples is more than 2 orders of magnitude larger than the scattered intensity of the unfilled samples. The scattering from the unfilled samples is illustrated and discussed in detail in a previous publication.31 Three features appear in all the SANS data: (1) a shoulder at q ) 0.08 Å-1, (2) strong scattering at intermediate q with a slope of -3, and (3) a low q downturn for q values less than 0.005 Å-1. The shoulder at q ) 0.08 Å-1 represents individual nanoparticles and can be fit to a model that describes polydisperse hard spheres. The slope of -3 at low q indicates the presence of dense aggregates with rough surfaces,40 consistent with our FE-SEM images. The low q downturn likely represents the size of the smallest clusters, although it could also indicate the presence of multiple scattering from the nanoparticles. One variable that can indicate multiple scattering is the intensity of the transmitted neutron beam. The unfilled and filled samples have equivalent transmissions at the same LiClO4 concentration and sample thickness. However, data for the filled samples reveal a low q downturn, whereas the unfilled data do not.31 The fact that this feature is absent in the unfilled data, combined with the fact that the transmission does not decrease when nanoparticles are added, suggests that it is unlikely for the low

9200



TABLE 2: Tg and Xc Values for All Samplesa Tg (°C) concentration

0 wt % NP

pure PEO 100:1 30:1 14:1 10:1 8:1 4:1

-55.0 ( 2.2 -35.7 ( 4.0 -21.8 ( 2.2 -15.0 ( 2.0 -27.8 ( 2.3 -17.2 ( 2.1 -8.5 ( 1.9

5 wt % NP

10 wt % NP

Xc (%) 25 wt % NP

-55.0 ( 3.1 -17.6 ( 1.9 -25.9 ( 2.5 -18.2 ( 2.2

-13 ( 3.0 -27.2 ( 2.4 -16.6 ( 2.6

-14.5 ( 2.3 -25.1 ( 2.1 -19.8 ( 3.0

0 wt % NP 7 64 54 31 2 0 0

5 wt % NP

10 wt % NP

25 wt % NP

76 27 0 2

30 0 1

30 1 0

a Samples are heated from room temperature to 200°C, cooled to -90 °C, and reheated to 200 °C. Data from the first and second heating scans are similar, indicating that pure PEO recrystallizes on the time scale of the DSC measurement. The Tg value represents the midpoint of the transition, and the error bars define the range of the transition.

Figure 3. FE-SEM images of (A) PEO + LiClO4 + 10 wt % Al2O3, (B) PEO + 10 wt % Al2O3 (no LiClO4), and (C) PEO + LiClO4 (no NP). Image D represents a subsection of the sample imaged in B. The ether oxygen to lithium ratio is 10:1 for the samples that contain LiClO4 (A and C). The circle on image D represents the size scale measured by SANS.

q downturn to arise from multiple scattering introduced by nanoparticles. Using the FE-SEM image as a guide, we fit the data with a model describing ellipsoids at low q41 and polydisperse spheres representing individual nanoparticles at high q.35 We could not fit the data using only models that described spheres. With these two contributions, the fit slightly misses the data between q ) 0.012 and 0.025 Å-1. This is corrected by including polydisperse spherical clusters at a size intermediate between the primary particles and the ellipsoidal clusters. The fit reveals that 25% of the nanoparticles are dispersed, 3% are part of spherical clusters, and 11% are part of elliptical clusters. The remaining 61% are aggregates larger than the size scale measured by SANS. These values correspond well with the FE-SEM image, although we cannot estimate the population of individual nanoparticles from the FE-SEM image D. The primary particle diameter returned by the fit is 4.8 ( 1.3 nm where the error represents one standard deviation of a Gaussian distribution. We tested various other distributions, such as log-normal and Schultz, and the polydispersity and quality of the fit were unaffected. The size of the particles returned by SANS (4.8 ( 1.3 nm) is approximately half the size quoted by the manufacturer (11 nm ( 3 nm). This discrepancy has been observed previously, and it could possibly be explained by considering the techniques used to measure the particle size.42

The manufacturer used gas absorption, where nanoparticle clustering could lead to overestimating the nanoparticle size. In contrast, nanoparticle clustering would only contribute to low q scattering in the SANS measurement, which is outside the q range representing the primary particle size. The second population of spheres (intermediate q) has an average diameter of 22 ( 11 nm, corresponding to clusters which are 2-6 nanoparticles wide. One would expect these clusters to be approximately spherical due to the small number of nanoparticles involved. In contrast, the clusters contributing to the low-q downturn are shaped like ellipsoids, with an average size of 43 × 140 × 200 nm3. Unlike the spherical features, we use a monodisperse model to represent the elliptical feature. This is reasonable because the size of the ellipsoids detected by SANS is close to the maximum spatial scale accessible by our measurements (200 nm). Taken together, the fits suggest that a wide range of particle clusters are present, and that when the clusters are larger, they tend to be elliptical rather than spherical. All fitting parameters are provided in Table 3, and the fit lines are illustrated in Figure 4. The cluster size and relative fraction of each population do not change with nanoparticle concentration. Given this level of nanoparticle aggregation, it is not surprising that the dynamics measured by Tg are unaffected by nanoparticle addition. We would only expect nanoparticles to


J. Phys. Chem. C, Vol. 114, No. 20, 2010 9201

Figure 4. SANS data for all of LiClO4 and nanoparticle [NP] concentrations at 80 °C. All the data except PEO + 10 wt % NP are shifted on the y axis for clarity by the factors given to the right of the data (14:1 by 100, 10:1 by 1E5, 8:1 by 1E8, and pure Al2O3 NP by 1E9). The lines through the data represent fits to a model describing two populations of polydisperse spheres and one population of monodisperse ellipsoids.

influence Tg due to confinement if the distance from the surface of one nanoparticle to the next was less than the Rg/2.43,44 Rg is the polymer radius of gyration and is equal to 27 nm for our samples. Even if the 100 nm aggregates were well-dispersed, the aggregate spacing would be 155 nm at 10 wt % nanoparticle loading, much larger than Rg/2. Thus far, we have observed that Al2O3 nanoparticles do not influence the Tg or crystal fraction, nor does the nanoparticle concentration affect the extent of aggregation within the window of SANS. Even though our DSC results indicate that polymer mobility is unaffected by nanoparticles, it is possible that

dynamics are influenced on time scales shorter than those measured by DSC. We have previously used QENS to measure mobility over pico- to nanosecond time scales of the unfilled SPE (PEO/LiClO4).31 This study revealed a process unrelated to Tg, which we attribute to the restricted rotation of (PEO)6: LiClO4 remnants in the liquid phase. It is possible that this rotation is affected by nanoparticles; thus, we investigate the dynamics of nanoparticle-filled samples with QENS. Polymer Dynamics as a Function of Nanoparticle and LiClO4 Concentration. We use QENS to measure PEO mobility in the presence of LiClO4 and nanoparticles. Hydrogen has the largest incoherent cross section of any element, and the only species in our SPE that contains hydrogen is PEO; therefore, only the mobility of PEO is detected. We measure samples with ether oxygen to lithium ratios of 8:1, 10:1, and 14:1 at 50 and 75 °C, and pure PEO at 75 °C. Measurements are made at all nanoparticle concentrations for the 10:1 sample, and at zero and 10 wt % for the 8:1, 14:1, and pure PEO samples. At 50 °C, the degree of crystallization depends on the sample composition. The 8:1 sample is at the melting point of pure PEO, and we therefore expect crystalline nuclei but no significant growth. The sample is below the melting point of (PEO)6:LiClO4, where crystalline nuclei can be present, but sufficient time has not elapsed for significant crystal growth. The 10:1 sample is at the eutectic composition at 50 °C, and because both crystalline phases are at their melting point, we expect only crystalline nuclei of pure PEO and (PEO)6:LiClO4 without crystal growth. The 14:1 sample is below the melting point of both pure PEO and (PEO)6:LiClO4. The sample is semicrystalline because of the fast crystallization kinetics of pure PEO, whereas (PEO)6:LiClO4 will exist primarily as nuclei. QENS detects changes in energy and momenta of scattered neutrons, reflecting the temporal and spatial dependence of the mobility of atoms in the sample. The number of neutrons scattered as a function of energy and spatial scale is given by the incoherent structure factor, S(q,ω). The connection between data from different instruments and the numerical treatment of the stretched exponential are better suited to the time domain. Thus, we inverse Fourier transform the data to the selfintermediate scattering function, S(q,t), illustrated in Figure 5

TABLE 3: Fit Parameters for a Model Describing Two Populations of Polydisperse Spheres and a Population of Monodisperse Ellipsoidsa

sample

NP volume fraction in bulk sample

dispersed nanoparticles

spherical clusters

elliptical clusters

aggregates

dispersed NP (vol %)

NP diameter (nm)

clustered NP (vol %)

cluster diameter (nm)

clustered NP (vol %)

cluster size (nm3)

aggregated NP (vol %)

NP powder

0.2

55 ( 0.1

5.0 ( 1.3

10 ( 0.05

22 ( 11

20 ( 0.15

42 × 110 × 200

15.0

8:1 5% NP 8:1 10% NP 8:1 25% NP

0.017 0.036 0.094

29 ( 0.4 27 ( 0.14 32 ( 0.06

4.4 ( 1.5 4.8 ( 1.2 4.8 ( 1.4

4.1 ( 0.02 2.7 ( 0.01 3.2 ( 0.01

22 ( 11 22 ( 11 22 ( 11

13 ( 0.06 8 ( 0.05 10 ( 0.07

42 × 140 × 200 42 × 155 × 200 44 × 142 × 200

53.9 62.3 54.8

10:1 5% NP 10:1 10% NP 10:1 25% NP

0.017 0.036 0.094

24 ( 0.3 20 ( 0.1 29 ( 0.04

4.6 ( 1.4 4.8 ( 1.2 5.0 ( 1.2

3.0 ( 0.1 2.4 ( 0.06 4.3 ( 0.09

22 ( 11 22 ( 11 22 ( 11

11 ( 0.05 7 ( 0.05 14 ( 0.07

46 × 146 × 200 41 × 134 × 200 42 × 136 × 200

62.0 70.6 60.7

14:1 5% NP 14:1 10% NP 14:1 25% NP

0.017 0.036 0.094

24 ( 0.3 27 ( 0.2 22 ( 0.1

4.8 ( 1.3 4.8 ( 1.3 4.8 ( 1.3

3.0 ( 0.02 4.0 ( 0.06 2.3 ( 0.01

22 ( 11 22 ( 11 22 ( 11

10 ( 0.06 12 ( 0.04 9 ( 0.03

42 × 145 × 180 42 × 138 × 180 45 × 140 × 180

63.0 57.0 66.7

PEO + 10% NP

0.036

27 ( 0.1

5.0 ( 1.25

2.7 ( 0.02

22 ( 11

12 ( 0.08

44 × 150 × 200

58.3

a The total volume fraction of nanoparticles in each sample is provided in the first column. The size and vol % of the individual nanoparticles and those in clusters are given, along with the vol % of aggregates larger than the length scale measured by SANS. The error in the individual nanoparticle and cluster diameters represents one standard deviation from the average, and is calculated by multiplying the diameter and the polydispersity values returned by the model.

9202



Figure 5. S(q,t) versus time at (A) 75 °C and (B) 50 °C for q ) 1.04 Å-1. The order of samples in the legend corresponds to the order of the data, and the lines represent fits as described in the text.

Figure 6. S(q,t) for 10:1 and all nanoparticle concentrations at q ) 1.04 Å-1: (A) 75 °C and (B) 50 °C. The lines represent fits as described in the text.

for all samples with 0 and 10 wt % nanoparticles at 75 °C (A) and 50 °C (B) and q ) 1.04 Å-1. As discussed below, the data are fit as received from the inverse Fourier transform. The HFBS data in Figure 5 are scaled to present the data on a continuous curve. The scaling accounts for the fact that the decay is measured in two pieces (a sum) whereas the decay can only be split in this way if it represents distinct processes well separated in time. This treatment is for display purposes only and does not influence the fitting procedure or the resulting fit parameters. Several general features are apparent from the QENS data in Figure 5. As we reported in an earlier publication,31 polymer mobility decreases with increasing LiClO4 concentration in the liquid phase (75 °C), consistent with previous observations of PEO-based amorphous SPEs.32,45 The slow mobility is attributed to the coordination of ether oxygens with Li+ ions. When nanoparticles are added, they slow the polymer mobility but only at the highest lithium concentration, 8:1, and the effect appears stronger at 50 °C compared to 75 °C. The explanation for this behavior is not immediately obvious. Based on the conductivity data above, we might expect nanoparticles to influence the mobility of the 10:1 sample if polymer mobility is the mechanism by which conductivity improves. Specifically, the 5 wt % sample would be influenced the most, because the conductivity is the highest at this concentration. We illustrate S(q,t) for all the 10:1 samples in Figure 6, demonstrating that

mobility is not influenced at any nanoparticle concentration or temperature. Considering the mobility and structure results, it is unclear how nanoparticles improve conductivity at the eutectic concentration. A difference does exist in the QENS data at a concentration of 8:1, and we therefore evaluate these results in more detail. To better characterize mobility, we fit the data to a stretchedexponential equation and examine the spatial dependence of the fitting parameters. We chose the Kolraush-Williams-Watts [KWW] stretched-exponential equation, which is generally applicable for describing segmental relaxation and rotation in polymers.46

[ ( τ(q,t T) ) ]

S(q, t) ) EISF + (1 - EISF) exp -

β(q,T)

The fitting parameters include the polymer relaxation time, τ, the distribution of relaxation times, β, and the elastic incoherent structure factor, EISF. Two processes are captured by the DCS data. The initial decay that occurs on time scales less than 2 ps has been observed for other polymers and is associated with cage vibrations and torsional librations, and the decay over time scales longer than 2 ps is associated with segmental motion. We determine that


Figure 7. Attempted fit of DCS and HFBS data without the second stretched exponential.

fits to the segmental process (KWW1) are improved by including the initial decay associated with vibrations (KWWvib), as this more accurately describes the region where the two processes are mixed. Thus, we fit the DCS data with the product of two stretched exponentials (KWWvib KWW1). Over times longer than 2 ps, the DCS segmental process (KWW1) and the HFBS data cannot be fit to a single process, as illustrated in Figure 7. Consequently, we use more than one KWW expression to fit the data. The HFBS data are fit to the sum of two stretched exponentials, KWW1 + KWW2, thus assigning fit parameters for KWW2 only. The presence of an additional process is consistent with the second, slower process observed by QENS in other SPEs: unfilled PEO/LiClO4,31 PEO/LiTFSI,32 and PEO/ LiBETI,47 where TFSI and BETI stand for N(SO2C2F3)2 and N(SO2C2F5)2.

J. Phys. Chem. C, Vol. 114, No. 20, 2010 9203 Because it is difficult to assign β accurately without data covering several decades in time, we fix β for the KWW1 process at 0.6, the expected value for the segmental relaxation of pure PEO based on dielectric spectroscopy measurements with a larger time range.48 We do not fix β for the KWW2 process because we are not sure of its origin. Error is assessed using a previously established procedure.49 In brief, 500 data sets are generated based on the original S(q,t) data and associated error bars, and each data set is fit using different initial guesses for the fit parameters. We report the average value of the 500 fits, along with error bars that represent one standard deviation from the average. The error bars represent the range of each parameter that can accurately fit the data within one standard deviation while leaving the other fitting parameters unconstrained. The fits to the data are presented as a product, S(q,t) ) KWW1KWW2, in Figures 5 and 6. The fit lines are generated using the parameters obtained by fitting each data set as described above. The β values for the second process range from 0.7-0.9 at 75 °C and 0.8-1.0 at 50 °C. We report the relaxation time for the first process (τ1) as a function of spatial scale in Figure 8 at 75 and 50 °C. The spatial dependence of the relaxation times is characteristic of segmental motion and is discussed in detail in a previous publication.31 In contrast to the effect of LiClO4, nanoparticles have no effect on the segmental motion of the polymer in the presence of LiClO4 salt, indicating that nanoparticles affect conductivity by some mechanism other than increasing the segmental relaxation of PEO. Unlike τ1, the relaxation times for the second process (Figure 9) are independent of spatial scale at 50 °C, indicating a rotational process. To learn about the geometry of the rotation, and how nanoparticles influence the rotation, we examine the spatial dependence of EISF2. These data are illustrated in Figure 10 for

Figure 8. τ1 versus q at 75 °C (A-C) and 50 °C (D-F). Samples without nanoparticles are illustrated in (A) and (D) samples with and without 10 wt % nanoparticles are illustrated in (B) and (E), and samples at 10:1 are illustrated in (C) and (F) at all nanoparticle concentrations. The error bars represent one standard deviation from the average. Pure PEO data from ref 50 (×) is included in (A) at 75 °C.

9204



Figure 9. τ2 versus q at 75 °C (A-C) and 50 °C (D-F). Samples without nanoparticles are illustrated in (A) and (D), samples with and without 10 wt % nanoparticles are illustrated in (B) and (E), and samples at 10:1 are illustrated in (C) and (F) at all nanoparticle concentrations. The error bars represent one standard deviation from the average.

Figure 10. EISF2 as a function of q at 75 °C (top row) and 50 °C (bottom row) for each LiClO4 concentration with and without 10 wt % nanoparticles. Fit lines represent fits to a model describing restricted rotation over a circle with 6 jump sites and radius 3 Å.

the 8:1, 10:1, and 14:1 samples with and without 10 wt % nanoparticles at both temperatures. It is clear that the nanoparticles influence the geometry of the rotation at a lithium concentration of 8:1, accounting for the difference in S(q,t) illustrated in Figure 5. As in our study of the unfilled SPE,31 we fit the EISF2 data with a model that describes rotation with a nonuniform distribution,51 and briefly review the results here. We include

the fit lines in Figure 10 for the EISF2 data at 50 °C. Rotation occurs over a circle, where points on the circle are unequally weighted, leading to preferred angular orientations. The model includes three fitting parameters: the number of sites on the circle, the radius of the circle, and β′, which defines how strongly the angular distribution is peaked (zero means there is no preferred angular orientation.) Although PEO has no side groups


Figure 11. Angular distribution for the 8:1 sample with and without nanoparticles. The equation for the distribution is given in ref 51. The cartoons represent restricted rotation of protons (green dots) around a Li+ ion in the 6:1 structure. A proton with intense shading indicates a high probability of being located at a specific angle, and less intense shading represents a low probability.

that would rotate, the addition of LiClO4 gives rise to such structures. As mentioned in the Introduction, the 6:1 crystalline phase forms when two PEO chains form a cylinder, with Li+ ions located inside, and this structure can persist to some extent even in the liquid phase.52,53 The radius returned by our fit is equal to the radius of the structure measured by neutron diffraction (Li-H distance of ∼3 Å).53 The fact that rotation is restricted seems reasonable, considering PEO chain connectivity and the coordination of ether oxygens with Li+ ions in (PEO)6: LiClO4. The fraction of protons involved in the rotating (PEO)6: LiClO4 structure is largest at 8:1, consistent with the fact that the quality of the fit is best at this concentration, and decreases with decreasing lithium concentration. A rotation is observed at 50 °C, but τ2 depends weakly on q at higher temperature (Figure 9A). The q dependence suggests that other motions degrade the rotation before a clear rotational signature is observed. As a result, we cannot fit EISF2 versus q with a model that describes rotation at 75 °C, restricted or otherwise. When we compare the fit parameters for the rotational model between the 8:1 sample with and without nanoparticles at 50 °C, the only parameter that changes is β′. Specifically, β′ increases from 1.8 to 3.0 with nanoparticles. Physically, this means that the rotation is more restricted (i.e., specific sites on the circle are more preferred than others.) We demonstrate this by plotting the angular distribution in Figure 11 for the 8:1 sample with and without nanoparticles, where φ indicates the angular position on the circle. The distribution shows that the probability of protons rotating (60° and (120° is less likely in the presence of nanoparticles. The difference suggests that ether oxygens in the 6:1 complex coordinate with acidic sites on the nanoparticle surface, thereby further restricting the rotation of the structure. The decreased rotation induced by the nanoparticle surface could propagate along the structure due to chain connectivity, thus extending the influence of nanoparticles to protons some distance away from the surface. The attraction of ether oxygens to the nanoparticle surface is further supported by the fact that the segmental relaxation times of the filled sample without salt are slower at high q than the segmental relaxation times of pure PEO (Figure 8B). PEO coordination with the nanoparticle surface is similar to coordination with Li+ ions, and we know from the QENS data that polymer mobility decreases with increasing Li+ concentration.

J. Phys. Chem. C, Vol. 114, No. 20, 2010 9205 Thus, when salt is present, the ether oxygen/nanoparticle coordinations may simply replace ether oxygen/Li+ coordinations. This would account for the invariance of segmental mobility with nanoparticle addition: PEO is coordinated equally with and without nanoparticles when salt is present. The results of this study show that conductivity increases with nanoparticles at the eutectic concentration (10:1), yet the only detectable property change occurs in the dynamics of the 8:1 sample. This result is surprising because ion mobility in SPEs is normally associated with the segmental motion of the polymer host. In our case, conductivity increases without a corresponding increase in polymer motion. This could result from an increased number of mobile ions due to nanoparticle/cation or nanoparticle/anion interactions. Another possibility is that a highly conductive polymer conformation is stabilized by the nanoparticle surface, where ion motion is decoupled from polymer motion. One such conformation is the 6:1 crystalline complex, for which a decoupling of ion and polymer mobility has been demonstrated.16 Remnants of (PEO)6: LiClO4 have been observed in the liquid phase,53 and their stabilization in the presence of nanoparticles could increase conductivity without affecting polymer motion. Although it is not possible to determine a mechanism for improved conductivity based on the present data, we can suggest one possible explanation that would require further investigation. The sample with a concentration of 10:1 is at the eutectic point at 50 °C, and under these conditions we expect crystalline nuclei of both (PEO)6:LiClO4 and pure PEO in equilibrium with the liquid phase. A model that describes marginal-stability theory during solidification predicts alternating lamellae of the two coexisting phases at the eutectic, or rods of one component in a matrix of the other.54 If this occurs, pure PEO would be the complex most likely to align next to the nanoparticles, due to the lack of coordination of its ether oxygens with the salt. Thus, it is possible that a layer of pure PEO will align next to the particles and a layer of (PEO)6: LiClO4 will align next to the pure PEO. If additional layers form, they will alternate and repeat as they extend away from the nanoparticle surface. It is the stabilization of the (PEO)6:LiClO4 layer that could permit enhanced ion mobility. The rotation of this structure remains unaffected, possibly because the (PEO)6:LiClO4 layers are stabilized but not constrained to the topology of the nanoparticle/SPE interface. Unlike the 10:1 sample, only (PEO)6:LiClO4 nuclei will form in the 8:1 sample at 50 °C. In the absence of pure PEO, the (PEO)6: LiClO4 structure will align next to the surface of the nanoparticles, restricting the rotation of the conductive structure as detected by our QENS measurements. The fact that the conductivity does not improve with nanoparticles at this concentration could be due to the fact that the (PEO)6:LiClO4 structure is aligned with the nanoparticle surface and cannot provide a connected pathway for conduction. We mentioned above that although this complex can be highly conductive, it has been demonstrated that proper alignment is required. The mechanism could be similar at temperatures above 50 °C, where the structures are no longer nuclei but fluctuations of (PEO)6: LiClO4 and pure PEO. It is possible that nanoparticles serve to stabilize these fluctuating structures, accounting for the improved conductivity at the eutectic composition. The concept of concentration fluctuations influencing conductivity near a eutectic point is not new; one study demonstrates that the electrical properties in metal alloys are influenced by concentration fluctuations at the eutectic concentration.55 Conclusions We measure the thermal properties, ionic conductivity, extent of aggregation, and polymer mobility of nanoparticle-filled solid

9206


polymer electrolytes containing PEO, LiClO4, and nanoparticles of Al2O3. We thermally treat our samples so that the only crystalline phase that forms is pure PEO. Our DSC results indicate that nanoparticles do not affect the pure PEO crystal fraction. SANS reveals that nanoparticles are aggregated to a similar extent in all samples, where 25 vol % of the nanoparticles are well-dispersed, 3 vol % are part of spherical clusters approximately 2-5 nanoparticles wide, and 11 vol % are part of elliptical clusters with dimensions of ∼43 × 140 × 200 nm3. The remaining nanoparticles are part of aggregates larger than the spatial scale of the instrument (200 nm). Despite the similarity between the filled and unfilled samples, the ionic conductivity increases with nanoparticles, but only at the eutectic composition of 10:1. DSC measurements reveal no change in Tg with nanoparticle addition, and QENS measurements reveal that the segmental motion of the polymer is not influenced by nanoparticle addition. These results indicate the nanoparticles do not improve conductivity by increasing polymer mobility. A rotation is detected in both filled and unfilled samples that is consistent with the rotation of (PEO)6:LiClO4. Although we thermally treat our samples to prevent the formation of the stable crystal phase of this structure, we expect crystalline nuclei of this structure below the melting point and fluctuations above. We observe that the rotation becomes more restricted in the presence of nanoparticles at an ether oxygen to lithium ratio of 8:1, at both 50 and 75 °C. This suggests that ether oxygens in (PEO)6:LiClO4 nuclei and the (PEO)6:LiClO4 fluctuations interact directly with the acidic sites on the nanoparticle surface at this concentration. Restricted rotation of this structure is observed only at the 8:1 concentration, whereas conductivity is improved only at the eutectic concentration of 10:1. The difference between 8:1 and 10:1 is that both pure PEO and (PEO)6:LiClO4 nuclei or fluctuations can form simultaneously at 10:1, depending on the temperature. We cannot know the mechanism for improved conductivity based on this data, but one suggestion is that the nanoparticle surface stabilizes the conductive (PEO)6:LiClO4 structure, allowing them to persist long enough for conduction to occur. Acknowledgment. Financial support for this work was provided by the National Science Foundation, Polymers Program, DMR-0706402. This work utilized facilities supported in part by the National Science Foundation under Agreement No. DMR0454672. References and Notes (1) Weston, J. E.; Steele, B. C. H. Solid State Ionics 1982, 7, 75–79. (2) Krawiec, W.; Scanlon, L. G., Jr.; Fellner, J. P.; Vaia, R. A.; Vasudevan, S.; Giannelis, E. P. J. Power Sources 1995, 54, 310–315. (3) Croce, F.; Appetecchi, G. B.; Persi, L.; Scrosati, B. Nature 1998, 394, 456–458. (4) Ahn, J. H.; Wang, G. X.; Liu, H. X.; Dou, S. X. J. Power Sources 2003, 119, 422–426. (5) Johansson, P.; Ratner, M. A.; Shriver, D. F. J. Phys. Chem. B 2001, 105, 9016–9021. (6) Capiglia, C.; Mustarelli, P.; Quartarone, E.; Tomasi, C.; Magistris, A. Solid State Ionics 1999, 118, 73–79. (7) Scrosati, B.; Croce, F.; Persi, L. J. Electrochem. Soc. 2000, 147, 1718–1721. (8) Tominaga, Y.; Asai, S.; Sumita, M.; Panero, S.; Scrosati, B. J. Power Sources 2005, 146, 402–406. (9) Jayathilaka, P. A. R. D.; Dissanayake, M. A. K. L.; Albinsson, I.; Mellander, B.-E. Electrochim. Acta 2002, 47, 3257–3268. (10) Wieczorek, W.; Stevens, J. R.; Florjan˜czyk, Z. Solid State Ionics 1996, 85, 67–72. (11) Tambelli, C. C.; Bloise, A. C.; Rosa´rio, A. V.; Pereira, E. C.; Magon, C. J. Electrochim. Acta 2002, 47, 1677–1682. (12) Dissanayake, M. A. K. L.; Jayathilaka, P. A. R. D.; Bokalawala, R. S. P.; Albinsson, I.; Mellander, B.-E. J. Power Sources 2003, 119, 409–414.

Fullerton-Shirey and Maranas (13) Xiong, H.-M.; Zhao, X.; Chen, J.-S. J. Phys. Chem. B 2001, 105, 10169–10174. (14) Sun, H. Y.; Sohn, H.-J.; Yamamoto, O.; Takeda, Y.; Imanishi, N. J. Electrochem. Soc. 1999, 146, 1672–1676. (15) Sun, H. Y.; Takeda, Y.; Imanishi, N.; Yamamoto, O.; Sohn, H.-J. J. Electrochem. Soc. 2000, 147, 2462–2467. (16) Gadjourova, Z.; Andreev, Y. G.; Tunstall, D. P.; Bruce, P. G. Nature 2001, 412, 520–523. (17) Staunton, E.; Andreev, Y. G.; Bruce, P. G. Faraday Discuss. 2007, 134, 143–156. (18) Choi, B.-K.; Kim, Y.-W.; Shin, K.-H. J. Power Sources 1997, 68, 357–360. (19) Choi, B.-K.; Shin, K.-H. Solid State Ionics 1996, 86, 303–306. (20) Quartarone, E.; Tomasi, C.; Mustarelli, P.; Magistris, A. Electrochim. Acta 1998, 43, 1321–1325. (21) Karlsson, C.; Best, A. S.; Swenson, J.; Howells, W. S.; Bo¨rjesson, L. J. Chem. Phys. 2003, 118, 4206–4212. (22) Croce, F.; Persi, L.; Scrosati, B.; Serraino-Fiory, F.; Plichta, E.; Hendrickson, M. A. Electrochim. Acta 2001, 46, 2457–2461. (23) Blonski, S.; Garofalini, S. H. Surf. Sci. 1993, 295, 263–274. (24) Dalnoki-Veress, K.; Forrest, J. A.; Massa, M. V.; Williams, A. J. Polym. Sci., Part B: Polym. Phys. 2001, 39, 2615–2621. (25) Ellison, C. J.; Torkelson, J. M. Nat. Mater. 2003, 2, 695–700. (26) Fryer, D. S.; Nealey, P. F.; de Pablo, J. J. Macromolecules 2000, 33, 6439–6447. (27) Bansal, A.; Yang, H.; Li, C.; Cho, K.; Benicewicz, B. C.; Kumar, S. K.; Schadler, L. S. Nat. Mater. 2005, 4, 693–698. (28) Karlsson, C.; Best, A. S.; Swenson, J.; Kohlbrecher, J.; Bo¨rjesson, L. Macromolecules 2005, 38, 6666–6671. (29) Singh, T. J.; Mimani, T.; Patil, K. C.; Bhat, S. V. Solid State Ionics 2002, 154, 21–27. (30) Reddy, M. J.; Chu, P. P.; Kumar, J. S.; Rao, U. V. S. J. Power Sources 2006, 161, 535–540. (31) Fullerton-Shirey, S. K.; Maranas, J. K. Macromolecules 2009, 42, 2142. (32) Mao, G. M.; Perea, R. F.; Howells, W. S.; Price, D. L.; Saboungi, M. L. Nature 2000, 405, 163–165. (33) Robitaille, C. D.; Fauteux, D. J. Electrochem. Soc. 1986, 133, 315– 325. (34) Wunderlich, B. Macromolecular Physics; Academic Press: New York, 1980; Vol. 3. (35) Kline, S. R. J. Appl. Crystallogr. 2006, 39, 895–900, Part 6. (36) The Data Acqusition and Visualization Environment [DAVE] is found at http://www.ncnr.nist.gov/dave. (37) Labreche, C.; Levesque, I.; Prud’homme, J. Macromolecules 1996, 29, 7795–7801. (38) Edman, L.; Ferry, A.; Doeff, M. M. J. Mater. Res. 2000, 15, 1950– 1954. (39) Bansal, A.; Yang, H.; Li, C.; Cho, K.; Benicewicz, B. C.; Kumar, S. K.; Schadler, L. S. Nat. Mater. 2005, 4, 693. (40) Teixeira, J. J. Appl. Crystallogr. 1988, 21, 781–785. (41) Feigin, L. A.; Svergun, D. I. Structure and Analysis by Small-Angle X-Ray and Neutron Scattering; Plenum: New York, 1987. (42) Bugnicourt, E.; Galy, J.; Ge´rard, J.-F.; Boue´, F.; Barthel, H. Polymer 2007, 48, 949–958. (43) Ellison, C. J.; Torkelson, J. M. Nat. Mater. 2003, 2, 695. (44) Fryer, D. S.; Nealey, P. F.; de Pablo, J. J. Macromolecules 2000, 33, 6439. (45) Mao, G. M.; Saboungi, M. L.; Price, D. L.; Armand, M.; Mezei, F.; Pouget, S. Macromolecules 2002, 35, 415–419. (46) Arrighi, V.; Higgins, J. S. Phys. B 1996, 226, 1–9. (47) Triolo, A.; Arrighi, V.; Triolo, R.; Passerini, S.; Mastragostino, M.; Lechner, R. E.; Ferguson, R.; Borodin, O.; Smith, G. D. Phys. B 2001, 301, 163–165. (48) Jin, X.; Zhang, S. H.; Runt, J. Polymer 2002, 43, 6247–6254. (49) Sakai, V. G.; Chen, C. X.; Maranas, J. K.; Chowdhuri, Z. Macromolecules 2004, 37, 9975–9983. (50) Saboungi, M. L.; Price, D. L.; Mao, G. M.; Fernandez-Perea, R.; Borodin, O.; Smith, G. D.; Armand, M.; Howells, W. S. Solid State Ionics 2002, 147, 225–236. (51) Be´e, M. Quasielastic Neutron Scattering - Principles and Applications in Solid State Chemistry, Biology and Materials Science; IOP Publishing Ltd: Bristol, England, 1988. (52) Mao, G. M.; Saboungi, M. L.; Price, D. L.; Armand, M. B.; Howells, W. S. Phys. ReV. Lett. 2000, 84, 5536–5539. (53) Mao, G.; Saboungi, M. L.; Price, D. L.; Badyal, Y. S.; Fischer, H. E. Europhys. Lett. 2001, 54, 347–353. (54) Langer, J. S. Phys. ReV. Lett. 1980, 44, 1023. (55) Ikeda, M.; Shibata, T.; Aoki, H.; Itami, T. J. Non-Cryst. Solids 2002, 312, 217–221.

JP906608P

LiClO4 Solid Polymer Electrolytes Filled

Recommend Documents