Enhanced Conductivity via Homopolymer-Rich Pathways in Block Polymer-Blended Electrolytes

Abstract

The optimization of ionic conductivity and lithium-ion battery stability can be achieved by independently tuning the ion transport and mechanical robustness of block polymer (BP) electrolytes. However, the ionic conductivity of BP electrolytes is inherently limited by the covalent attachment of the ionically conductive block to the mechanically robust block, among other factors. Herein, the BP electrolyte polystyrene-block-poly(oligo-oxyethylene methacrylate) [PS-b-POEM] was blended with POEM homopolymers of varying molecular weights. The incorporation of a higher molecular weight homopolymer additive (α > 1 state) promoted a “dry brush-like” homopolymer distribution within the BP self-assembly and led to higher lithium salt concentrations in the more mobile homopolymer-rich region, increasing overall ionic conductivity relative to the “wet brush-like” (α < 1 state) and unblended composites, where α is the molecular weight ratio between the POEM homopolymer and the POEM block in the copolymer. Neutron and X-ray reflectometry (NR and XRR, respectively) provided additional details on the lithium salt and polymer distributions. From XRR, the α > 1 blends showed increased interfacial widths in comparison to their BP (unblended) or α < 1 counterparts because of the more central distribution of the homopolymer. This result, paired with NR data that suggested even salt concentrations across the POEM domains, implied that there was a higher salt concentration in the homopolymer POEM-rich regions in the dry brush blend than in the wet brush blend. Furthermore, using ⁷Li solid-state nuclear magnetic resonance spectroscopy, we found a temperature corresponding to a transition in lithium mobility (T_{Li mobility}) that was a function of blend type. T_{Li mobility} was found to be 39 °C above T_g in all cases. Interestingly, the ionic conductivity of the blended BPs was highest in the α > 1 composites, even though these composites had higher T_gs than the α < 1 composites, demonstrating that homopolymer-rich conducting pathways formed in the α > 1 assemblies had a larger influence on conductivity than the greater lithium ion mobility in the α < 1 blends.

Graphical Abstract

INTRODUCTION

Lithium-ion batteries are some of the most widely used energy storage devices because of their high energy densities and their ability to be recharged for hundreds of cycles.^1,2 In most commercial lithium-ion battery systems, the electrolyte is a flammable liquid with a thin polymer membrane as a separator.³ Because of the lack of overall mechanical robustness combined with the flammability of the liquid electrolyte, current batteries are subject to catastrophic failure as a result of, for example, separator fracture or lithium dendrite penetration.^2-5 To improve the mechanical robustness of the electrolyte and mitigate these issues, replacing the liquid electrolyte/separator system with homopolymer or block polymer (BP) electrolytes has been examined.^6-15 However, an appropriate balance between conductivity and mechanical robustness is required to develop safe and commercializable lithium-ion batteries.^1,6,7,16,17

Because the majority of lithium-ion conductivity in a polymer electrolyte occurs by segmental chain motion of the polymer and inter- and intra-chain hopping of the ions, one method to enhance the lithium-ion mobility is to decrease the glass transition temperature (T_g) of the conducting component of the polymer electrolyte.⁵ Homopolymer electrolytes, such as poly(ethylene oxide) (PEO) doped with a lithium salt, have high ionic conductivity, as they are composed of entirely ion-conducting material and are not covalently bound to a less-mobile high-T_g block.^18,19 However, as many homopolymer electrolytes are engineered to have low T_g, their shear modulus is insufficient to prevent mechanical failure.^20,21 BP electrolytes typically consist of ion-conducting polymer covalently attached to a high-T_g block, such as polystyrene (PS).^17,22,23 In a BP electrolyte, as a result of the phase-separated domains, the ionic conductivity and shear modulus are decoupled, enabling simultaneous optimization of the two properties.^17,22-24 Nonetheless, the ionic conductivity remains limited in these BP electrolyte systems.^18,25 Creating a homopolymer-blended BP electrolyte is a promising strategy to achieve a more desirable balance of ionic conductivity (enhanced by the homopolymer additives) and mechanical robustness (maintained by the BP nanostructure and high-T_g component), though it is essential to understand the role of homopolymer molecular weight and volume fraction on the homopolymer and salt distributions (and ultimately the ion transport) in these composite electrolytes.

Non-salt-doped model systems composed of a homopolymer A blended with a BP A-b-B have been examined both theoretically and experimentally to understand the distribution of homopolymer within the BP matrix and the morphology of the resulting system.^26-31 It has been established that the ratio of the homopolymer molecular weight to the molecular weight of the corresponding block in the BP (α = M_{n,A in homopolymer}/M_{n,A in BP}) and the homopolymer volume fraction in the blend (φ_H) affect the morphology and distribution of homopolymer in the self-assembled, composite system.^26-30 In the limit of high α, the homopolymer forms its own domain, termed “dry brush”, whereas in the limit of α = 0, the homopolymer has a distribution similar to that of the A block in the BP, termed “wet brush”.^26,27,30 More specifically, from self-consistent field theory, the α > 1 case leads to homopolymer distributions, which are dependent only on the brush overlap parameter, w_OL (a system-specific parameter), that approach symmetric hyperbolic tangents centered around the middle of the domain.²⁷ The α < 1 case, however, leads to homopolymer distributions that are a function of both the homopolymer molecular weight and w_OL, in which the homopolymer tends to distribute more evenly at lower molecular weights.²⁷ For higher φ_H, both the breadth and the maximum concentration of the homopolymer distribution increases.^26,31 These theoretically predicted trends have been verified experimentally in lamellae-forming systems with neutron reflectometry (NR) experiments on PS-b-polyisoprene/deuterated PS (PS-b-PI/dPS) and PS-b-poly(methyl methacrylate)/dPS blends across a variety of α, φ_H, and M_n,BP values.^26,30 Manipulating the blend characteristics to achieve a desired homopolymer distribution is useful in engineering the next generation of composite systems.

Beyond the homopolymer distribution, Matsen developed phase diagrams for these A-b-B/A blends using self-consistent field theory for a variety of values of φ_H and α.²⁹ Many of the classical phases, such as lamellae and hexagonally-packed cylinders, were found, but phases that are less typical in equilibrated BP systems, such as hexagonally-perforated lamellae, also were noted.²⁹ Furthermore, blends with homopolymer incorporated at relatively high volume fractions (φ_H = 0.4, f_A = 0.5, χN = 12) maintained a lamellar morphology, which suggests that a large phase window exists for performing reflectometry-based studies.²⁹ Finally, recent work on homopolymer blends has suggested that the addition of low molecular weight homopolymers promotes rapid annealing of thin films and stabilizes a bicontinuous microemulsion phase (which may have promising ion conducting properties) in the case of ternary A-b-B/A/B blends.^32-34 In developing a salt-doped homopolymer-blended BP electrolyte system, parameters like α, φ_H, and the salt-doping ratio, r, can be used to manipulate the combined salt and polymer density profiles such that the lithium salt is doped at ideal concentrations in a given component in the composite electrolyte.

In an A-b-B/A-blended BP electrolyte system, the distribution of salt also is of interest because the ionic conductivity of an electrolyte is highly dependent on the salt concentration such that there exists a salt concentration that leads to a maximum conductivity. The salt concentration that promotes maximum conductivity depends on the polymer system because of the local stresses in the ion-conducting component.³⁵ For example, in PEO-based electrolytes doped with lithium bis(trifluoromethane)sulfonimide, the ionic conductivity is maximized at a salt-doping ratio (molar ratio of ethylene oxide monomer segments in the PEO to lithium ions) of [EO]:[Li]_σ,max = 9:1, whereas, in PS-b-PEO-based electrolytes doped with the same salt, the maximum conductivity is at [EO]:[Li]_σ,max = 5:1.¹² Because conductivity is affected by a number of interrelated factors including morphology, alignment, and local salt concentration, it is important to deconvolute these influences using a combination of bulk conductivity and lithium mobility measurements.^35-37 It is anticipated that if the local concentration of salt is controlled such that the most mobile region of a polymer electrolyte is doped at [EO]:[Li]_σ,max, the conductivity of that particular electrolyte would be maximized.

The salt distribution in a salt-doped BP has been elucidated using a variety of techniques, including energy-filtered transmission electron microscopy (EF-TEM),³⁵ X-ray photoelectron spectroscopy (XPS) with C₆₀⁺ etching,³⁸ and NR.³⁹ Though there has not been complete agreement between these studies, these differences possibly can be attributed to differences in BPs (PS-b-PEO in the EF-TEM studies vs PS-b-poly(oligo-oxyethylene methacrylate) [PS-b-POEM] in the XPS with C₆₀⁺ etching and NR studies). For the case of a homopolymer-blended BP electrolyte to be described herein, it is necessary to understand the salt distribution and correlate it to the homopolymer distribution because the lithium salt may interact preferentially with the BP or homopolymer.

In this work, we determined the salt and polymer distributions and used these distributions to understand the bulk conductivity trends in BP/homopolymer-blended electrolytes doped with lithium trifluoromethanesulfonate (lithium triflate) salt. The spatial distributions of the lithium triflate salt were elucidated, as a function of homopolymer molecular weight and lithium salt concentration, using NR. As described previously, the natural scattering length density (SLD) contrast between the lithium salt and the two polymer blocks promoted the determination of the salt distribution.³⁹ The same samples were probed with X-ray reflectometry (XRR) to establish the polymer interfacial profile of the electrolyte blends. Blends with α < 1 had lower interfacial widths than their unblended counterparts and significantly lower interfacial widths than blends with α > 1. Using a strong-segregation theory (SST) approach, the dependences of the effective Flory–Huggins interaction parameter, χ_eff, and the statistical segment length of the POEM + salt, b_POEM, on the lithium salt concentration were calculated as a function of α and [EO]:[Li] in the blended electrolytes. By employing the combined XRR and NR profiles, the salt concentration in the POEM homopolymer was calculated, and the blend with α > 1 had a significantly higher lithium salt concentration within the homopolymer than its α < 1 counterpart.

The abovementioned structural insights then were combined with results from bulk electrolyte studies. The thermal properties were evaluated using differential scanning calorimetry (DSC); electrolytes with α < 1 had a lower T_g,POEM than their unblended or α > 1 counterparts. The lithium mobility also was examined by ⁷Li solid-state nuclear magnetic resonance (NMR) spectroscopy, and the electrolytes with α < 1 had the highest lithium mobility, followed by the unblended and α > 1 examples. Thus, the lithium mobility was related to the mobility of the POEM, which is defined as the difference between the operating temperature, T, and T_g,POEM. A transition temperature that corresponds to the lithium mobility, termed T_{Li mobility}, was established for each blend, and T_{Li mobility} was correlated directly to the T_g,POEM of the blends. Interestingly, despite the enhanced lithium mobility in the α < 1 blend, the ionic conductivity was highest in the α > 1 blend. This enhancement was attributed to the lower χ_eff and the more centralized homopolymer distribution in the α > 1 blend. Furthermore, the unblended and α < 1 blend conductivities fell onto a single curve once differences in the T_g,POEM and volume fraction of conducting block were considered, but the α > 1 blend conductivities remained higher. Overall, this combined analysis was able to elucidate an alternate method to enhance the conductivity independently from the polymer segmental chain motion.

MATERIALS AND METHODS

Materials.

All chemicals were stored in an argon-filled glovebox after purification and before use. Styrene (99%, stabilized, Acros Organics) and oligo-oxyethylene methyl ether methacrylate (OEM, > 99%, stabilized, Sigma-Aldrich, average molar mass = 500 g mol⁻¹) were purified by passage through a basic alumina column. The styrene was dried further by distilling it from calcium hydride. OEM and styrene were degassed via three freeze–pump–thaw cycles. Copper bromide [Cu(I)Br, 98%, Acros Organics] was purified by stirring in acetic acid for 30 min, filtering, washing twice with cold ethanol, and drying under dynamic vacuum. Anhydrous methanol (99.8%, Sigma-Aldrich) was stirred with calcium hydride, degassed via three freeze–pump–thaw cycles, and distilled. Ethyl 2-bromoisobutyrate (EBiB, 98%, Sigma-Aldrich), N,N,N′,N″,N″′-pentamethyldiethylenetriamine (PMDETA, 99%, Sigma-Aldrich), and anisole (>99%, Fisher Scientific) were degassed via three freeze–pump–thaw cycles. Dry tetrahydrofuran (THF, >99%, optima, Fisher-Scientific) was degassed by purging with argon and dried on solvent purification columns. Lithium triflate (99.995%, Sigma-Aldrich) was dried at 150 °C for 48 h.

Polymer Synthesis and Characterization.

PS-b-POEM was synthesized according to literature protocols.⁴⁰ POEM homopolymers were made by atom transfer radical polymerization with EBiB as the initiator. In a typical POEM polymerization, in an argon-filled glovebox, Cu(I)Br (0.1 mmol), PMDETA (0.1 mmol), and anisole were mixed to ensure full complexation of copper and ligand. Separately, OEM (2–10 mmol), EBiB (0.1 mmol), and anisole were combined in a round-bottom flask. To initiate the reaction, the copper solution was added to the reaction vessel, and then, the flask was sealed. The reaction conversion was monitored by ¹H NMR (Bruker AV600III) spectroscopy with CDCl₃ (0.03 v/v % TMS) as a solvent, and the reaction was stopped by cooling with liquid nitrogen and then opening the reaction to air. The reaction mixture was purified by passage through a neutral alumina column to remove the copper and precipitated into mixtures of petroleum ether/diethyl ether/2-propanol (1/1/0.1 by volume). Size-exclusion chromatography (Viscotek VE2001) with THF as the eluent (1.0 mL min⁻¹) and PS standards (1780–205 000 g mol⁻¹) was used to determine the molecular weight of the PS block, along with the dispersity of the POEM homopolymers and PS-b-POEM. The volume fraction of POEM in the PS-b-POEMs and the molecular weights of the POEM homopolymers were determined by ¹H NMR spectroscopy in CDCl₃ (0.03 v/v % TMS). The different PS-b-POEMs and POEMs, and their molecular properties, are listed in Table 1. All polymers were dried at 120 °C for 48 h before storage in a 0 °C freezer in an argon-filled glovebox.

Table 1.

Polymer Characteristics of Polymers Used in This Study^a

polymer	Mn (g mol−1)	Mn,POEM (g mol−1)	dispersity (Ð)	fPOEM
PS-b-POEM-17.8	43 100	17 800	1.19	0.38
PS-b-POEM-23.9	42 200	23 900	1.21	0.53
POEM-8	8000	8000	1.13	1
POEM-24.1	24 100	24 100	1.08	1

^aThe volume fraction of POEM is notated as f_POEM and is calculated using the bulk densities of the constituent homopolymers.

Blend fabrication was performed in an argon-filled glovebox, unless otherwise noted. PS-b-POEM/POEM blends (Table 2) were formulated by dissolving the desired PS-b-POEM and POEM in dry THF and allowing the solution to stir for at least 3 h. All blends were prepared with 20 wt% homopolymer. The blends were sealed in a drying chamber and dried under dynamic vacuum for 12 h. The homopolymer concentration was verified by ¹H NMR spectroscopy.

Table 2.

Blend Combinations Used in This Study^a

polymer blend	BP	homopolymer	φH	α
Unblended-17.8	PS-b-POEM-17.8	N/A	0
Wet brush-17.8	PS-b-POEM-17.8	POEM-8	0.2	0.45
Dry brush-17.8	PS-b-POEM-17.8	POEM-24.1	0.2	1.35
Unblended-23.9	PS-b-POEM-23.9	N/A	0
Wet brush-23.9	PS-b-POEM-23.9	POEM-8	0.2	0.33
Dry brush-23.9	PS-b-POEM-23.9	POEM-24.1	0.2	1.01

^aφ_H is the homopolymer volume fraction in the blend, and α is the ratio of the molecular weight of the POEM homopolymer to the molecular weight of the POEM block in the BP. Blends with α > 1 were expected to exhibit dry-brush-like behavior, and blends with α < 1 were anticipated to show wet-brush-like behavior.

Polymer Film Preparation.

Solutions of Unblended-17.8, Wet brush-17.8, Dry brush-17.8, and lithium triflate solutions were prepared in an argon-filled glovebox by dissolving each material in dry THF at approximately 5 wt% and stirring for at least 3 h. The polymer and salt stock solutions were mixed gravimetrically and stirred for 3 h to achieve the desired [EO]:[Li], ranging from 24:1 and 8:1. 15 wt% methanol was added to the electrolyte solutions to enable formation of higher-quality thin films;^38,39 in these cases, the final polymer/salt/THF/methanol solutions were ~3.5 wt% polymer.

Silicon wafer substrates (〈100〉 orientation, Wafer World, Inc.) were rinsed with toluene three times and cleaned in an ultraviolet–ozone oven (model 342, Jelight Co., Inc.). Films were flow coated⁴¹ to cast the salt-doped BPs and BP blends on the cleaned silicon substrates. Films with shallow thickness gradients (70 mm long, 25.4 mm wide, between 90 and 150 nm thick) were cast and commensurate regions corresponding to 2.5 or 3.5 times the domain spacing (d) were used for NR and XRR experiments. The films were dried under vacuum overnight at room temperature before being annealed under dynamic vacuum for 6 h at 135 °C (for Unblended-17.8 and Wet brush-17.8) or 120 °C (for Dry brush-17.8). Film thicknesses were measured with spectral reflectometry using a Filmetrics F20-UV interferometer.

Neutron Reflectometry.

NR experiments were conducted on the multi-angle grazing-incidence k-vector (MAGIK) instrument⁴² at the National Institute of Standards and Technology (NIST) Center for Neutron Research. Neutrons with an incident wavelength of 5 Å were used, and for each sample, scans of Q_z from 0 to 0.0877 Å⁻¹ were recorded with a step size of 0.0004 Å⁻¹. To minimize the warping of the silicon substrates, elastic clamps were used with an aluminum sample holder, and a borated aluminum mask was added between the neutron source and the films to minimize scattering from the sample holder, clamps, and silicon wafers. The reflectometry profiles were reduced using NIST’s reductus software and analyzed with repeating lamellae models in the refl1D software program.^43,44

X-ray Reflectometry.

XRR experiments were conducted on a Rigaku Ultima IV X-ray diffractometer. A Cu Kα X-ray source (λ = 1.54 Å) was used, and a thin, parallel X-ray beam was scanned over incident angles (θ) and detection angles (2θ) of 0° < 2θ < 3° with a step size of 0.004° and a scan rate of 0.5° min⁻¹. The beam size was set such that the critical edge of the thin film sample was captured. Rigaku GlobalFit software was used to fit the XRR profile with repeating lamellae models.

Bulk Electrolyte Fabrication.

Lithium salt solutions of lithium triflate and dry THF were allowed to stir for at least 3 h. Electrolytes were created by dissolving the desired amount of BP (PS-b-POEM-23.9) or BP blend (Wet brush-23.9 and Dry brush-23.9) in dry THF, stirring for at least 3 h, and adding the requisite lithium salt solution gravimetrically. The BP and BP blend electrolyte solutions were mixed for at least 3 h before they were sealed in drying chambers and dried under dynamic vacuum for about 12 h. Once fully dry, the electrolytes were stored in an argon-filled glovebox until testing. This sample preparation method was used for all small-angle X-ray scattering (SAXS), DSC, ⁷Li solid-state NMR spectroscopy, and alternating current (AC) impedance spectroscopy specimens.

Small-Angle X-ray Scattering.

All salt-doped SAXS samples were prepared in an argon-filled glovebox by sealing between two Kapton films in a homemade sample holder to avoid absorption of moisture. SAXS was conducted on a Xenocs instrument with a sealed-tube X-ray source (Cu Kα, λ = 1.54 Å) operating at 2.0 kW with a 2D detector and 2000 mm sample-to-detector distance. The path tubes were held under dynamic vacuum to reduce scattering from air, and sample temperatures were controlled using a Linkam HFSX350-CAP stage under dynamic vacuum. All SAXS samples were annealed at 150 °C for 2 h, cooled to 120 °C for 4 h, and cooled further to 30 °C for 4 h under dynamic vacuum, and SAXS profiles were acquired at each temperature. The data reported herein were taken at 120 °C. All twodimensional scattering data were azimuthally integrated, resulting in plots of scattered intensity vs scattering vector, q.

Differential Scanning Calorimetry.

DSC experiments were performed on a TA Instruments Discovery DSC with an RCS90 cooling accessory. Samples were loaded in an argon-filled glovebox and hermetically sealed in aluminum pans. Three heating/cooling cycles between −80 and 150 °C at 5 °C min⁻¹ were used with a nitrogen atmosphere. Baseline calibration was conducted with sapphire disks, and temperature and cell constant calibrations were determined using an indium standard. The reported T_g was determined from the midpoint of the inflection in the third heating trace. The second and third heating were compared to ensure reproducibility upon heating.

⁷Li NMR Spectroscopy.

⁷Li NMR data were collected using a 300 MHz DMX300 Bruker NMR spectrometer with a 4 mm probe using the solid echo (π/2-τ-π/2-τ-Acq) and single π/2 pulse sequences with ¹H decoupling. Solid echoes were used for T < 47 °C. For T > 47 °C, solid echoes and single pulse excitation sequences yielded indistinguishable spectra. Each sample (16–29 mg) was sealed in a Kel-F rotor insert in a dry glovebox, placed into a ZrO₂ rotor, and loaded into the NMR probe that was purged with dry air (−100 °C dew point) to ensure that samples did not encounter moisture. Typical NMR experimental parameters were 116.64 MHz Larmor frequency, 3.5 μs π/2 pulse, 20 μs τ delay, 80 kHz ¹H continuous wave decoupling, 1 or 5 μs dwell time, 16 384 receiver gain, 4096 data points, 61 440 zero filled points, and 100 Hz apodization. Temperature was controlled to within ±1 °C and was calibrated using a lead nitrate ²⁰⁷Pb NMR standard.⁴⁵ For the variable temperature experiments, each sample was cooled to the lowest temperature (−70 °C) first, equilibrated for 15 min, and then scanned. Samples were equilibrated at subsequently warmer temperatures for 15 min prior to scanning. Experiments were performed at temperatures ranging from −70 to 137 °C.

AC Impedance Spectroscopy.

Salt-doped BPs and BP blends were pressed under elevated temperatures (150 °C) into disks under vacuum in an argon-filled glovebox and placed into a homemade test cell on a Linkam HFS91 CAP stage as described previously.⁴⁶ To perform the ionic conductivity measurements, the test cell was connected to a Princeton Applied Research PARSTAT 2273 frequency response analyzer, and all measurements were taken under vacuum. Samples were heated to 120 °C for 2 h to promote good contact with the aluminum electrodes and then cooled to 20 °C. The impedance measurements were taken upon heating between 20 and 150 °C at 10 °C increments, holding at each temperature for 10 min. At a given temperature, two measurements were acquired and averaged. The AC frequency range and voltage amplitude were 0.1–1 MHz and 10 mV, respectively. The touchdown point in the Nyquist plot was defined as the bulk resistance, R, and the ionic conductivity, σ, was calculated as σ = L/(RA), in which L is the sample thickness and A is the area between the electrolyte and electrode.

RESULTS

To probe the enhancement in conductivity for the homopolymer-blended BP composite electrolytes, the structural, thermal, and electrochemical properties were examined. As a result of the large parameter space (α, φ_H, M_n,Bp, M_n,H, etc.), finding an appropriate balance of the homopolymer distribution, salt distribution, and overall ion conducting properties is challenging. To this end, we investigated the salt and polymer distributions in parallel-lamellae-forming thin films by NR and XRR, respectively. Next, we proxied the polymer and lithium mobilities by DSC and ⁷Li NMR spectroscopy, respectively, to probe the effect of homopolymer molecular weight (and thus α) on thermal properties and ion environment. Finally, we measured the conductivity of the blended electrolytes and compared the effect of the salt and polymer distributions to the ion mobility to elucidate the primary conduction mechanisms in these composite electrolytes.

The neutron and X-ray SLDs (ρ_neutron and ρ_X-ray, respectively) for the component materials used in this study are listed in Table 3. The ρ_{neutron,Li triflate} is higher than either of the component polymers, and ρ_neutron,POEM is less than ρ_neutron,PS; thus, as salt is added, a contrast match condition can be found When ρ_{neutron,POEM+Li triflate} = ρ_neutron,PS, as described previously for salt-doped BP electrolyte systems.³⁹

Table 3.

Neutron and X-ray SLDs⁴⁷

material	neutron SLD (ρneutron) (×106 Å−2)	X-ray SLD (ρX-ray) (×106 Å−2)
PS	1.41	9.61
POEM	0.78	11.30
Li triflate	3.08	15.86

NR profiles of films of Unblended-17.8, Wet brush-17.8, and Dry brush-17.8 doped with [EO]:[Li] ratios of 12:1 were collected on the MAGIK reflectometer (colored data points, Figure 1). At least one Bragg peak was identifiable in all three profiles, though they were much more prominent in Wet brush-17.8 (where, a second order peak, marked with a black arrow, also appears) than in either of the other films. As the homopolymer content changed, the interfacial roughness was affected, which was particularly illustrated by the rapid decay of the Kiessig fringes in the Dry brush-17.8 profile because of increased mixing between blocks. Additional profiles of films with different salt doping ratios ([EO]:[Li] = 24:1, 8:1, and 18:1) are provided in Figures S1, S2, and S3, respectively.

Figure 1.

NR profiles (colored data points) and model fits (black lines) for Unblended-17.8 (green), Wet brush-17.8 (orange), and Dry brush-17.8 (purple) with [EO]:[Li] = 12:1. The black arrow indicates the second-order peak in the Wet brush-17.8. Error bars and uncertainty represent one standard error. We note that the χ² reported in this manuscript is a measure of the statistical fitting of a model to the reflectometry data (using the normalized χ² test) and is not related to the Flory–Huggins parameter, χ_eff.

NR profiles were fit to repeating multilayer lamellae models, the outputs of which are shown in Figure 2; a representative fit at [EO]:[Li] = 12:1 is presented for Unblended-17.8, Wet brush-17.8, and Dry brush-17.8. Multilayer models for the other salt concentrations are included in Figures S1-S3. The ρ_neutron profiles are plotted as a function of the depth through the film in Figures 2 and S1-S3. To reduce the number of fitting parameters, the following constraints were applied. The value of ρ_neutron,PS was held close (within 0.05 × 10⁶ Å⁻²) to the value for the neat PS, as it has been established that lithium salts are not found in the PS domains.^35,38,39 Additionally, for each respective sample, all interior PS and interior POEM + salt domains had similar model parameters (layer thicknesses, SLDs, and interfacial widths). This approach allowed us to determine average properties of the interior region, separate from the interfacial POEM and surface PS layers. This route was chosen (as opposed to a multilayer model with more degrees of freedom) to identify more physically relevant solutions, as opposed to those that just minimized χ². In all cases, there was a reduced SLD at the polymer/substrate interface, likely due to small amounts of solvent trapped at the substrate.

Figure 2.

SLD profiles for Unblended-17.8, Wet brush-17.8, and Dry brush-17.8, at [EO]:[Li] = 12:1, as a function of distance from the substrate surface.

The same films (Unblended-17.8, Wet brush-17.8, and Dry brush-17.8) were probed using XRR to quantitively determine the domain spacing (d) and interfacial roughness (f_int) between the PS and POEM domains; see Figure 3. As expected, the d of the blended BP electrolytes generally was higher than that of the unblended samples (Figure 3a) across all salt concentrations. Furthermore, for the blended samples, the wet brush had a higher d than the dry brush, consistent with results from other systems with low φ_H.^30,48 For all samples, d increased linearly with increasing r = [Li]/[EO], and all blend types had similar linear scaling. The f_int (Figure 3b) was defined as two times the average interfacial width (t_int) between PS and POEM domains, normalized by the domain spacing, which provided a measure of the relative amount of the lamellar period contained within the interfacial region between domains. f_int is defined in eq 1

$$f_{{}_{\text{int}}}=\frac{2t_{{}_{\text{int}}}}{d}=\frac{2(2\pi {)}^{1∕2}\delta }{d}$$ (1)

in which δ is the roughness parameter from the Rigaku GlobalFit software.^39,49,50 f_int generally decreased as r increased, and Dry brush-17.8 had the highest interfacial widths across all salt concentrations. Additional XRR profiles and fits for all samples are shown in Figure S4.

Figure 3.

(a) Domain spacing d, and (b) volume fraction of interface, f_int, as a function of Li triflate salt concentration, r, for Unblended-17.8 (green circles), Wet brush-17.8 (orange squares), and Dry brush-17.8 (purple diamonds) specimens. d increased and f_int decreased as r increased for all blend conditions. The error bars in (a) represent the standard deviation in layer thickness from the XRR fits, and the error bars in (b) are propagated uncertainties of the roughness parameter from XRR fits. If error bars are not visible, the error is smaller than the size of the data point.

With an enhanced understanding of the salt and polymer distributions of the blended electrolytes, bulk characterization was performed using a different (relative to the reflectometry studies) parent BP to generate Unblended-23.9, Wet brush-23.9, and Dry brush-23.9 samples doped with Li triflate at [EO]:[Li] = 16:1. Because of the higher volume fraction of POEM in PS-b-POEM-23.9, upon doping with Li triflate, a hexagonally packed cylindrical morphology was formed in all blended samples, as suggested by the peaks indexed to 1, √3, and 2 in the SAXS profiles (Figure S5). It must be noted that the ion transport behavior of BP electrolytes is affected by the nanostructured morphology; however, it has been established by Matsen that the homopolymer distribution is similar between hexagonally packed cylinders and lamellae.²⁹ Thus, it is assumed that these studies are relevant to both morphologies, and the insights gleaned from the NR and XRR studies can be used to understand the bulk characterizations herein.

First, as ionic conductivity in polymer electrolytes is related to the segmental chain motion of the polymer, we used DSC to examine the thermal transitions in the blends. The DSC traces for the three samples are shown in Figure 4, and the T_g,POEM, T_g,PS, and change in heat capacity of the T_g,POEM (Δc_P,POEM) are given in Table 4. Two transitions were clearly visible in each trace, which corresponded to T_g,POEM (−47 to −43 °C) and T_g,PS (93–95 °C), though T_g,POEM was a more prominent transition, due to the higher mass fraction of POEM. For all samples, T_g,PS was close to the predicted homopolymer T_g,PS at this molecular weight on the basis of the empirical equation suggested by Fox and Flory,⁵¹ indicating that there was little mixing between the blocks. T_g,POEM was lowest in Wet brush-23.9 and highest in Dry brush-23.9, with Unblended-23.9 falling in between. Though these differences are small, they are relative to the distinct instrument resolution, and small changes in T_g can correspond to sizable differences in the ionic conductivity (~20% increase in conductivity at 40 °C by decreasing T_g,POEM from −43.5 to −47.0 °C).⁵² Δc_P,POEM was calculated by measuring the difference in height of the extrapolated baselines above and below T_g,POEM, and Wet brush-23.9 has the highest Δc_P,POEM, whereas Dry brush-23.9 has the lowest value.

Figure 4.

DSC trace on third heating (5 °C min⁻¹, N₂) for Unblended-23.9 (green), Wet brush-23.9 (orange), and Dry brush-23.9 (purple) samples, normalized by the total blended BP electrolyte mass. The curves are shifted vertically for clarity.

Table 4.

DSC Data for the BP Blends

polymer blend	Tg,POEM(°C)	Tg,PS(°C)	ΔcP,POEM
Unblended-23.9	−45.7	93.8	0.43
Wet brush-23.9	−47.0	94.6	0.50
Dry brush-23.9	−43.5	94.6	0.36

Although the polymer thermal transitions often are considered as a reasonable proxy for the lithium mobility, it was possible to probe the lithium mobility directly using solid-state ⁷Li linewidths. At temperatures for which the lithium still behaved as if in a glass (less than −30 °C), Li–O “bonds” were in the rigid limit, and two resonances were identified: a narrow Lorentzian resonance and a broad Gaussian resonance, which were due to central (−1/2 to 1/2) and satellite (−3/2 to −1/2, and 1/2 to 3/2) transitions, respectively, in the quadrupolar nuclei. As the temperature increased above the T_g,POEM, the Li–O “bonds” broke or reoriented on <100 μs (=π/Δν) time scales [ν is the full width at half-maximum (FWHM) of the Gaussian resonance] such that both the intensity and breadth of the Gaussian curve decreased at the expense of the Lorentzian resonance, which also narrowed further upon heating. Generally, in this fast fluctuation limit, narrower peaks indicate more mobile lithium species.^53,54 An example of curves at two different temperatures is provided in Figure S6. The FWHM of the fitted Lorentzian resonances of the Unblended-23.9, Wet brush-23.9, and Dry brush-23.9 samples are shown in Figure 5. The lithium mobilities in all blend conditions were high (approaching the 100 Hz of apodization applied to the time domain signal), but the Dry brush-23.9 lithium mobility was significantly lower than the other two mobilities.

Figure 5.

FWHM of Lorentzian fit peaks from ⁷Li NMR spectra as a function of temperature for Unblended-23.9 (green circles), Wet brush-23.9 (orange squares), and Dry brush-23.9 (purple diamonds).

To determine the impact of homopolymer-blending on the ionic conductivity, AC impedance experiments were conducted on the Unblended-23.9, Wet brush-23.9, and Dry brush-23.9 samples doped with Li triflate at the salt-doping ratio of [EO]: [Li] = 16:1. The conductivity profiles as a function of inverse temperature are shown in Figure 6a. At all temperatures, Dry brush-23.9 exhibited the highest ionic conductivity, followed by Wet brush-23.9 and then Unblended-23.9. The conductivities were fit to the nonlinear Vogel–Fulcher–Tammann (VFT) equation (eq 2)

$$\sigma ={\sigma }_{\mathrm{o}}{\mathrm{e}}^{-B∕T-T_{\mathrm{o}}}$$ (2)

in which σ_o is the pre-exponential factor proportional to the number of charge carriers, B is an effective activation energy, T is the operating temperature, and T_o is a reference temperature, chosen as T_o = T_g,POEM – 50 K. By plotting conductivity as a function of T/T_g and normalizing by the volume fraction of POEM (Figure 6b), additional factors that influence conductivity can be emphasized.^17,52,55 Fitting parameters are provided in Table S2.

Figure 6.

(a) Ionic conductivity as a function of 1000/T and (b) ionic conductivity normalized by the volume fraction of total POEM as a function of T_g/T, for Unblended-23.9 (green circles), Wet brush-23.9 (orange squares), and Dry brush-23.9 (purple diamonds) samples.

DISCUSSION

These combined results highlight the competition between brush regime and mobility in blended BP electrolytes. In terms of the structure, NR and XRR were used to infer the salt and polymer distributions in the electrolyte. From NR, the results were consistent with the hypothesis that the salt distribution followed the total POEM polymer distribution in all blend conditions. The two homopolymer profiles as a function of the normalized position in the POEM domain (x/l_POEM) are shown in Figure 7, and calculations are detailed in the Supporting Information. The Dry brush-17.8 homopolymer distribution was calculated using the analytical expressions developed by Shull and Winey with the polymer characteristics calculated from the XRR data herein.²⁷ The Wet brush-17.8 homopolymer distribution, as a result of the more complex nature of wet brush distributions, was fit to the homopolymer profile of the closest values of α and Φ_homopolymer from the same study (α = 0.52 vs 0.45, Φ_homopolymer = 0.31 vs 0.38).²⁷ Using these homopolymer distributions (Figure 7a), the salt doping ratio in the homopolymer POEM was determined algebraically. From these results (Figure 7b), it is apparent that the homopolymer in the dry brush regime has a significantly higher salt concentration, reaching a minimum value of [EO]:[Li] in the center of the domain at [EO]:[Li] = 13:1 vs 20:1 in Wet brush-17.8. Therefore, because the POEM homopolymer was more centrally localized in Dry brush-17.8, the salt concentration is enhanced within the homopolymer regime.

Figure 7.

(a) Fraction of homopolymer POEM within the POEM domain and (b) [EO]:[Li] in the homopolymer POEM as a function of x/l_POEM (position in POEM domain) for Dry brush-17.8 (purple) and Wet brush-17.8 (orange) doped at [EO]:[Li] = 12:1.

The XRR results were analyzed using SST to examine how the lithium salt interacted with the BP blends. For the Unblended-17.8 samples, the parameters determined from XRR (d,f_int, and the densities of each of the components) were combined with properties of the BP (N, φ_POEM, and the statistical segment length of PS) to calculate χ_eff and b_POEM.³⁹ For the homopolymer-blended BPs, the effective N_POEM was determined by ¹H NMR spectroscopy. The calculated χ_eff and b_POEM are plotted in Figure 8. The statistical segment length of PS was assumed to be constant at b_PS = 0.68 nm, and a reference volume ν₀ = 0.668 nm³ (the volume of one POEM monomer) was used for all samples.^52,56 As the salt concentration increased, b_POEM increased for Unblended-17.8 and Wet brush-17.8; for Dry brush-17.8, there was a slight decrease in b_POEM at high salt concentrations, which was indicative of the brush architecture, as the homopolymer, which was segregated to the center of the POEM domain especially in the dry brush sample, had a higher ion concentration. Similarly, χ_eff increased with higher salt concentrations for all samples, and Dry brush-17.8 had the lowest χ_eff, while Unblended-17.8 had the highest χ_eff. We hypothesize that the different dependencies of χ_eff as a function of r was caused by the dry brush, homopolymer-rich regime, which led to a lower effective salt concentration in the BP-rich region of the self-assembled nanostructure. These differences in homopolymer blending therefore influenced the structure of the resulting electrolytes.

Figure 8.

(a) POEM statistical segment length (b_POEM) and (b) effective Flory–Huggins interaction parameter (χ_eff) as a function of r for Unblended-17.8 (green circles), Wet brush-17.8 (orange squares), and Dry brush-17.8 (purple diamonds) specimens.

The interfacial width trends determined by XRR were supported by the thermal behavior established through DSC studies. Δc_P,POEM (Table 4) was highest in Wet brush-23.9, followed by Unblended-23.9, and then Dry brush-23.9. Using the scaling from Morese-Seguela et al.,⁵⁷ it follows that the thickness of interfacial region can be estimated from eq 3

$$t_{{}_{\text{POEM}}}=\frac{L_{{}_{\text{POEM}}}}{2}\left(1-\frac{\Delta c_{{}_{\text{P,POEM}}}}{\Delta c_{{}_{\text{P,pure POEM}}}}\right)$$ (3)

in which L_POEM is a characteristic length scale of the POEM layer and Δc_{P,pure POEM} is the heat capacity change at T_g for homopolymer POEM. Because of local salt concentration fluctuations, determination of Δc_{P,pure POEM} at the comparative salt concentration of [EO]:[Li] = 16:1 is difficult, but the general trend that the highest Δc_P,POEM (such as in Wet brush-23.9) corresponded to the lowest interfacial widths is nonetheless informative and matches the trends elucidated in this work.

The combination of DSC and ⁷Li solid-state NMR spectroscopy enhanced understanding of the interplay between polymer and lithium mobilities. T_g,POEM was inversely correlated to the lithium mobility as shown in Figure 9 such that the blended sample with the lowest T_g,POEM had the highest lithium mobility. These differences in lithium mobility that occur more than 100 °C above the polymer electrolyte’s T_g highlight the need to continue to develop ion-conducting polymers with depressed T_gs as a way to improve the lithium mobility.

Figure 9.

FWHM of the Lorentzian fit at 67 °C vs T_g,POEM for the Unblended-23.9 (green circles), Wet brush-23.9 (orange squares), and Dry brush-23.9 (purple diamonds) samples.

The temperature dependence of the motional narrowing of the broad ⁷Li resonance also revealed a new characteristic temperature relevant to the mobility of lithium through polymer electrolytes. The broad peak, which represents the population of ⁷Li spins involved in rigid Fi–O “bonds”, became narrow upon heating because of the breaking and fast reorientations of these Fi–O bonds, causing scrambling of the three principal components of the Li quadrupolar tensor. When these events occurred on a time scale (t) faster than the inverse of the ⁷Li linewidth (t ⪡ π/30 000 Hz ≈ 10 μs), these three components became dynamically averaged. To further investigate this dynamic process, we aimed to quantify the temperature range over which this transition occurs, and the transition temperature, termed T_{Li mobility}, was defined as the mid-point of the transition. Representative curves for Unblended-23.9, Wet brush-23.9, and Dry brush-23.9 are shown in Figure 10. Curves were empirically fit to match the sigmoidal shape of the curves by minimizing the weighted sum of squares error via eq 4

$$I=K\times {\tan }^{-1}\left(\frac{T-T_{\mathrm{o}}}{T_{\mathrm{f}}-T_{\mathrm{o}}}\right)$$ (4)

in which I is the FWHM of the Gaussian peak, K is a fitting parameter to capture the magnitude of the quadrupolar tensor in units of Hz, T is the temperature of interest, T_o is the onset of the transition, and T_f is the end of the transition. T_{Li mobility} was defined as the midpoint of T_o and T_f. From these fits, T_{Li mobility} was −6.0 °C for Unblended-23.9, −8.4 °C for Wet brush-23.9, and −4.9 °C for Dry brush-23.9. It is notable that this transition occurred at 39.0 ± 0.6 °C above T_g,POEM in all of the blends (T_g,POEM = −45.7 °C for Unblended-23.9, −47.0 °C for Wet brush-23.9, and −43.5 °C for Dry brush-23.9), which was consistent with the idea that the motion of the ethylene oxide-containing segments were responsible for breaking the Li–O bonds and, furthermore, indicates that T_{Li mobility} may be a more representative measure of the temperature at which most of the lithium ions in the electrolyte are mobile. It is expected that the difference between T_g,polymer and T_{Li mobility} is a polymer- and lithium-salt-dependent property, and examining this difference in a variety of lithium salts and polymers will be fruitful in the development of electrolyte systems.

Figure 10.

FWHM of the Gaussian fit from ⁷Li linewidth NMR spectroscopy as a function of temperature for Unblended-23.9 (green circles), Wet brush-23.9 (orange squares), and Dry brush-23.9 (purple diamonds). Fits to eq 4 are shown with gray dashed lines.

The discrepancy between the enhanced lithium mobility in Wet brush-23.9 and the enhanced ionic conductivity in Dry brush-23.9 provokes questions about the relative importance of mobility versus brush regime. As noted in Figure 6b, by normalizing by the T_g,POEM and φ_POEM Unblended-23.9 and Wet brush-23.9 collapse onto a single curve, likely indicating that their conduction mechanism is similar. It is straightforward to rationalize these similarities on the basis of the structural characterizations by NR and XRR, as the wet brush acts like a plasticized version of the unblended BP electrolyte. The VFT fits to the conductivity (Table S2), shown with the gray lines in Figure 6a, indicated that there was little difference in the B parameter, which relates to the effective activation energy. Moreover, the B parameters were all similar to tapered PS-b-POEM doped with Li triflate electrolytes at [EO]:[Li] = 15:1,⁵² suggesting a similar conduction mechanism in all of these samples. The effective ion concentration, which is related to the pre-exponential factor, σ_o, was twice as large for Dry brush-23.9 (σ_o = 2.2 × 10⁻² S cm⁻¹) than Wet brush-23.9 (σ_o = 1.1 × 10⁻² S cm⁻¹), which correlated to the homopolymer salt concentration being approximately twice as high in Dry brush-23.9 and was close to the calculated result in Figure 7b. This difference implies that the region consisting of majority homopolymer is better able to conduct ions. Thus, though lithium mobility was higher in Wet brush-23.9 than Dry brush-23.9, the higher salt concentration in the homopolymer-rich regime in Dry brush-23.9 had a stronger effect on the ionic conductivity.

CONCLUSIONS

We report a strategy to introduce homopolymer-rich conducting pathways in BP composite electrolytes to improve ionic conductivity. The lithium salt and polymer distributions were probed by NR and XRR. NR was harnessed to determine the lithium salt profiles, and the results were consistent with fits in which the lithium salt followed the POEM profile. Polymer interfacial profiles, including the domain spacings and interfacial widths, were determined using XRR; Wet brush-17.8 had the lowest interfacial widths, and Dry brush-17.8 had the highest interfacial widths across all salt concentrations. The homopolymer profiles were used to calculate the effective salt concentration in the POEM homopolymers, and, despite having the same overall salt concentration, the Dry brush-17.8 had a significantly higher maximum salt concentration in the POEM homopolymer. By combining the XRR results with SST, the dry brush-blended system had the lowest effective Flory–Huggins parameter across all salt concentrations, indicating higher compatibility between blocks. In bulk electrolyte materials, the wet brush-blended electrolytes demonstrated a depressed T_g,POEM and the highest lithium mobility, as determined by DSC and ⁷Li NMR linewidth measurements. From the ⁷Li NMR, a new relevant temperature for lithium conduction, T_{Li mobility}, was defined as the mid-point temperature between lithium-oxide bonds being completely rigid in a glassy polymer electrolyte matrix and being completely broken leaving lithium free to move amongst the polymer matrix on time scales faster than 10 μs. This transition temperature was found to follow the same trend as T_g,POEM, and for all samples probed, T_{Li mobility} was 39 °C above T_g,POEM. The ionic conductivity was probed via ac impedance spectroscopy, and despite its suppressed lithium mobility, the dry brush-blended electrolyte demonstrated the highest ionic conductivity. This enhancement was attributed to the higher salt concentration in the homopolymer-rich regime, which led to an overall increase in the fraction of mobile ions. Taken together, these insights demonstrate the complex nature of maximizing the ionic conductivity in a BP electrolyte material, as both structure and inherent mobility are key in the performance of these materials.