Operando NMR Visualization of Ion Dynamics in PEDOT:PSS

While organic mixed ionic/electronic conductors are widely studied for various applications in bioelectronics, energy generation/storage, and neuromorphic computing, a fundamental understanding of the interactions between the ionic and electronic carriers remains unclear, particularly in the wet state and on electrochemical cycling. Here, we show that operando NMR spectroscopy can selectively probe and quantify ion and water movement during the doping/dedoping of poly(3,4-ethylene dioxythiophene) poly(styrene sulfonate) (PEDOT:PSS) films, the most widely used organic mixed conductor. Na + ions near or within the PEDOT-rich domains experience an anisotropic environment resulting from the underlying partial PSS chain orientation in the polymer films, giving rise to a distinct quadrupolar splitting in the 23 Na NMR spectrum. Operando 23 Na NMR studies reveal a linear correlation between the quadrupolar splitting and the charge stored in the film, which is interpreted in terms of the roles that the Na + ions at the PEDOT/PSS interfaces play in charge balance and electric double layer formation. The observed correlation is quantitatively explained by a competitive binding model, in which holes on the PEDOT backbone are bound to PSS, the hole concentration changes during doping/dedoping inducing variations in the Na + binding percentage at the PEDOT/PSS interfaces. The Na + -to-electron coupling efficiency, measured via 23 Na NMR intensity changes, varies noticeably depending on the cycling history of the film. Operando 1 H NMR spectroscopy confirms that water molecules accompany the ions that are injected into/extracted from the films. These findings shed light on the working principles of organic mixed conductors and demonstrate the utility of operando NMR spectroscopy in revealing structure-property relationships in electroactive polymers. and the electric field gradient (EFG) tensor at the nucleus being a sensitive measure of the anisotropy of the Na + environments 28,29 in heterogenous polymer films. Operando 23 Na NMR studies show that the quadrupolar splitting, extracted from the quadrupolar lineshape, is negatively correlated with the number of ions stored in the film, the correlation being explained by the injection/extraction of Na + near PEDOT/PSS interfaces. The Na + -to-electron coupling efficiency is quantified via integration of the operando NMR signals. The methods demonstrated here open further possibilities to understand the mechanisms, capacity retention and degradation of mixed ionic and electronic conducting polymers.


Introduction
The design and optimisation of organic materials with mixed ionic and electronic conductivity is key to enabling the next generation of bioelectronic devices, where these materials mediate the coupling between solid-state electronics and biological media for advanced healthcare monitoring and disease treatment. 1,2 Organic mixed conductors are also used as active electrodes 3 , electrode coating layers 4 as well as polymer binders in supercapacitors and lithium ion batteries, 5 to improve the electrode conductivity and minimize side-reactions.
Moreover, these materials draw significant interest in the facile-to-fabricate, and hence lowcost electronic circuits, that are needed to realise the Internet of Things. 6,7 Finally, the recent development of neuromorphic devices based on organic mixed conductors holds great promise in new computing architectures with improved energy efficiency and biocompatibility. 8,9 Poly(3,4-ethylene dioxythiophene) doped with poly(styrene sulfonate) (PEDOT:PSS) is the most ubiquitous of all organic mixed ionic/electronic conductors. It is a degenerately doped ptype semiconductor, in which holes on the PEDOT chains are compensated by sulfonate anions on the PSS chains. Typical films contain excess PSS and the majority of the sulfonate groups are compensated by cations such as Na + or H + . PEDOT:PSS films can be dedoped electrochemically under the influence of a voltage that extracts holes through a metal electrode and injects cations from an electrolyte to compensate the "free" SO 3groups. Application of the opposite voltage leads to further doping of the film, via the extraction of cations and injection of holes to compensate the free SO 3groups. This exchange between holes and cations as the compensating species for SO 3takes place at PEDOT/PSS interfaces throughout the bulk of the film and manifests itself as a volumetric capacitance. 10 The morphology of PEDOT:PSS films is controlled via its deposition from dispersions containing colloidal particles with PEDOT-rich cores and PSS-rich coronas (Figure 1a-c). 11 These particles flatten out when spin-casted onto a flat substrate, forming oblate-spheroid, or "pancake-shaped" PEDOT-rich domains, embedded in a PSS-rich matrix. Cross-sectional atomic force microscopy images reveals that the PEDOT-rich domains are typically 20-30 nm in diameter and 4-6 nm in height, 12 containing comparable amounts of PEDOT and PSS as revealed by X-ray scattering 13 , with the excess PSS giving rise to the PSS-rich matrix. Films, therefore, show anisotropic electronic (through interconnected PEDOT-rich domains) and ionic (predominantly within the PSS-rich matrix) conduction pathways. [13][14][15][16] PEDOT:PSS has been traditionally used in optoelectronics as a hole-conducting layer. 17 While electronic charge carrier transport in this material in the dry-state is reasonably well understood, the ion-polymer interactions and ion-electron coupling within the conjugated polymer in the wet-state and during device operation have not been studied with equal depth. 10,18,19 This is mainly due to challenges in monitoring molecular and electronic structure in a heterogeneous polymer matrix and in differentiating between electronic and ionic charge carriers using traditional electrical techniques alone. It is of interest to differentiate and quantify each process, as the competing absorption of Na + and H + in PEDOT:PSS induces a local pH change, which may affect the stability of the polymer or the viability of cells in biological interfaces. 20,21 Various characterization techniques have been reported to analyse the ion injection/extraction process during dedoping/doping. Electrochemical quartz crystal microbalance with dissipation monitoring (EQCM-D) has been used to measure the change in mass per injected cation, 22 however, an accurate estimate of the ionic-to-electronic coupling efficiency from this measurement requires independent knowledge concerning both the nature of the ions and the number of water molecules in their hydration shells involved in the electrochemical processes. Ion transport in conjugated polymers has been probed by the moving front method using time-lapse optical micrographs, 19 which revealed fast cation drift mobility in PEDOT:PSS films. This method, however, cannot differentiate between species (e.g., Na + from H + ), nor can it quantify the relative strength of coupling between different charge species. New, quantitative techniques are needed to understand ion injection and transport, their coupling to electronic conductivity, and polymer film structure.
Nuclear magnetic resonance (NMR) spectroscopy is a non-invasive and nuclei-specific technique that can provide both quantitative atomic and dynamic information. 23-26 Operando NMR spectroscopy in particular has been applied to study various battery materials to provide structural and dynamic information concerning transient species. 27 Here, operando 23 Na and 1 H NMR spectroscopies are used to understand the ion and water movement within the PEDOT:PSS film. The film was cross-linked with 1 wt % 3-glycidoxypropyltrimethoxysilane (GOPS) to prevent the film dissolving in the aqueous electrolyte. We exploit the quadrupolar nature of the 23 Na nucleus, the quadrupolar interaction between the nucleus and the electric field gradient (EFG) tensor at the nucleus being a sensitive measure of the anisotropy of the Na + environments 28,29 in heterogenous polymer films. Operando 23 Na NMR studies show that the quadrupolar splitting, extracted from the quadrupolar lineshape, is negatively correlated with the number of ions stored in the film, the correlation being explained by the injection/extraction of Na + near PEDOT/PSS interfaces. The Na + -to-electron coupling efficiency is quantified via integration of the operando NMR signals. The methods demonstrated here open further possibilities to understand the mechanisms, capacity retention and degradation of mixed ionic and electronic conducting polymers.

23
Na NMR reveals two Na + environments and partial orientation in the polymer leading to characteristic lineshapes that depend both on the Na + local environment and on any motion the Na + ions undergo. 30 When sodium ions hop rapidly between sites with random orientations, the quadrupolar interaction is averaged to zero, and a single 23 Na NMR resonance is expected. This is the case for the PSS film, indicating that (i) the distribution of EFG tensor orientations which the sodium ions experience is isotropic and (ii) that the motion between the sites occupied by Na + in the film is rapid. When the sodium ions hop between environments with a non-random distribution of EFG tensors (an anisotropic medium), the quadrupolar interaction is no longer completely averaged. If all Na + ions sample the same anisotropic EFG distribution within the NMR timescale (ergodicity), this leads to a triplet of central and two satellite signals with a fixed intensity ratio of 3: 4: 3 and a frequency separation between the satellites, ∆ , depending on the film orientation 30,31 as follows: where 〈 ̅ 〉 is the residual quadrupole coupling constant (RQCC) and is the angle between the magnetic field B 0 and the vector perpendicular (i.e., normal) to the film. ∆ , as defined here, corresponds to twice the quadrupole splitting between the central line and each of the satellite peaks.
The triplet observed for this particular PEDOT:PSS film ( Figure 1d) is typical for an anisotropic film structure with non-vanishing RQCC. However, the observed integral ratio in the triplet for this particular film is 0.22 : 0.56 : 0.22 (Figure 1d) and the relative intensity of the central peak increases as more NaCl solution is added to the polymer film ( Figure S1). The additional intensity (comprising 29% of the total 23 Na content for this film) is assigned to both Na + ions in the free electrolyte on top of the film and in the isotropic regions of the swollen film including PSS-rich and GOPS-containing region. This environment is referred to as "Na + in PSS" below, but strictly, it represents all the Na + ions located in isotropic media.
When the cast PEDOT:PSS film is rotated inside the NMR coil about the coil axis, the frequency separation, ∆ , between the satellites changes as a function of the film orientation according to Eq. 1 yielding a value for 〈 ̅ 〉 of 4 kHz. The results indicate that the principal component of the (averaged) EFG tensor is also normal to the film providing strong evidence for a partial orientation of the polymer (pancake) nanostructure with respect to the substrate, as schematically illustrated in Figure 1c. The NMR spectrum taken of a wet PEDOT:PSS polymer film scraped of the glass substrate and randomly packed into an NMR tube gives a typical quadrupolar "powder" pattern, from which a similar 〈 ̅ 〉 value is extracted ( Figure S1).
The results clearly indicate that the Na + adsorption sites in the isotropic PSS-rich domains versus the anisotropic domains can be differentiated and quantified via 23 Na NMR spectroscopy. The anisotropic domains likely comprise both the PEDOT/PSS "pancakes" and nearby PSS polymer chains with partial ordering that originates from the film microstructure.  Operando NMR experiments were then performed to monitor the Na + ion movement in response to electrical doping/dedoping of the film, using a symmetric cell configuration similar to that developed to study supercapacitors 32 with two pieces of mass-and area-balanced Operando 23 Na NMR studies were performed by biasing the electrochemical cell to a constant voltage for 40 min, while 1D 23 Na NMR spectra were continuously recorded to monitor the Na + ion injection/extraction into/from the observed electrode ( Figure 3). The current flowing through the cell decays during each of the voltage-hold steps, but even after 40 mins has not yet reached a steady state in the thick film used here, particularly on stepping the voltage from -0.6 to +0.6 V with respect to the electrolyte. Under negative bias, which leads to dedoping of the film, the 23 Na NMR quadrupolar splitting becomes larger (Figure 3c). The trend is opposite upon doping the film (i.e. on applying +0.6 V). A negative, close-to-linear correlation between the charge and satellite peak separation was found (Figure 3c-iii).

Operando 23 Na NMR shows voltage-dependent changes in Na + concentration
On applying a negative bias, Na + injection is expected if Na + is the major ionic charge compensating species. The total 23 Na NMR peak integral grows by approximately 1-2 % during PEDOT dedoping (0V→ -0.6V), but it drops more noticeably by 3.6% during doping (-0.6V→ +0.6V) as shown in Figure 3b-iii. The 23 Na NMR spectra were deconvoluted to separate the isotropic and anisotropic components as demonstrated in Figure 1e (shown in Figure 3b-iv and v, respectively), a voltage switch from -0.6V to +0.6V resulting in decrease of the signal from Na + ions in the PEDOT-rich (anisotropic) region by 2.5 %, the integral of Na + in PSS-rich region and electrolyte dropping more rapidly with time but by a smaller amount of 1.1 % (see Figure S5 for further analysis). The absolute intensity of the 23 Na NMR signal was calibrated to convert the NMR integral into the number of ions ( Figure S6), the overall integral change corresponding to 8.110 -7 moles of Na + , while the number of electrons removed from the polymer is 0.097 C (corresponding to 1.010 -6 moles of electrons, that is one hole in 28 to 41 EDOT units, and an estimated doping level of 2.4 -3.5% per EDOT unit), leading to a Na + -toelectron coupling efficiency of 81 % on doping.
When the voltage applied to the film increases stepwise from +/-0.2 V to +/-0.6 V, alternating between positive and negative voltages (Figure 4a), the resulting quadrupolar splitting also increases proportionally, increasing first at a voltage hold of -0.2 V and then decreasing when the voltage is stepped to +0.2 V. A negative linear correlation between charge and splitting was again observed as shown in Figure S14b in SI. The voltage hold time used however are much shorter, and the system no longer has time to equilibrate. This is particularly pronounced on switching from negative to positive voltages, since in contrast to the positive to negative step, no 0 V hold step was applied. Consequently, more charge is injected into the film than removed, which is accompanied by a steady increase in the 23 Na intensity. Similar trends are observed in Figure 4b where a lower biasing voltage of +/-0.4 V is used. 23 Na NMR integral changes observed for the films in Figure 4a and b were analysed, and the corresponding Na +to-electron coupling efficiency varies between 63 and 135 % depending on the exact cycling history of the film ( Figure S7).
A thinner (0.35 mm thick) PEDOT:PSS film was prepared in the same way to more closely mimic films in real devices and its operando NMR data is shown in Figure 4c. Both the charge and 23 Na quadrupolar splitting change much more rapidly in response to the alternating voltage than seen for the 1.15 mm thick film, and the same negative trend in quadrupolar splitting vs. charge is seen ( Figure S9g). but also of water. 34 The volume expansion pushes water out of the area sampled by the NMR coil, reducing the intensity of the 4.8 ppm signal. Note that each 1 H NMR spectrum takes about 48 s to acquire and the rapid change of 1 H NMR spectra as a function of voltage indicates that the hydration speed of the film is within one or two minutes, which is faster than many other common hydrogels. 35 An initial (inconclusive) attempt to monitor the anion movement by operando 35 Cl NMR is shown in Figure S16 in the SI. The signal-to-noise of the 35 Cl NMR spectra was poor due to the low sensitivity of 35 Cl nucleus, and the operando data do not show noticeable changes.

Discussion:
Physical Interpretation of 23 Na NMR quadrupolar splittingcharge correlation The asymmetric charge distribution around the Na + ions is measured here via 23 Na NMR  (Figures 3 and 4).
A key feature of PEDOT:PSS is that the amount of SO 3available for Na + binding is controlled by the number of holes on PEDOT, i.e. the level of doping of the film. By assuming that the binding of holes to SO 3 -(as illustrated in Figure 1a) is strong (and dictated by local charge balancing/neutrality) whereas the Na + binding to SO 3is weaker, an expression for the satellite separation ∆ and thus 〈 ̅ 〉 as a function of the charge (Q) delivered to PEDOT:PSS can be readily derived (see SI section 6 for derivation and assumptions): with K Na the binding equilibrium constant between Na + and SO 3 - Figure 4 and Figure S14).
A model with weak binding of Na + to the SO 3groups that are not bound to the holes is supported by a number of observations including: first, the quadrupolar splitting increases when the concentration of the NaCl solution increases ( Figure S4), the higher concentration of Na + ions shifting the binding equilibrium towards increased Na + -SO 3binding and secondly, 〈 ̅ 〉 tracks the charge almost instantaneously, doping of PEDOT almost instantaneously reducing the number of SO 3binding sites available for Na + . It should be kept in mind that while this binding model considers holes to be tightly bound to SO 3groups, the holes themselves can still be mobile. This is not inconsistent with the energy level structure of PEDOT:PSS. Holes in degenerately doped organic semiconductors occupy a transport manifold that originates from highest occupied molecular orbitals of the conjugated system and is broadened due to positional and electrostatic disorderthe latter arising in part by the strong dipolar interactions between holes and dopants (SO 3in this case). 39,40 As a result, holes close to the Fermi level will be highly mobile, while those at the tail of the manifold near the band gap will be essentially trapped. However, as the film is dedoped and the Fermi level approaches the edge of the transport manifold, these holes will also become mobile albeit with a lower mobility. 41 The results presented here show that the Na + ions themselves are also mobile and in rapid equilibrium with the SO 3groups and thus can readily respond to maintain local charge neutrality as the holes move.

Ion dynamics
The curves of 〈 ̅ 〉 vs. the hold time at different voltages can be fit with an exponential decay function using the equation: A × (− ) + , yielding values of that vary from 20-400 s depending on film thickness ( Figure S9 and S10). The average -value is 355 s for the thick film and 22 s for the thin film. A self-diffusion coefficient of the Na + ion in the films was measured by 23 Na pulsed field gradient (PFG) NMR to be 1.13×10 -5 cm 2 s -1 ( Figure S11), which corresponds to a drift mobility of 4.40×10 -4 cm 2 s -1 V -1 that it is in the same order of magnitude of the drift mobility reported by Stavrinidou et al. 19 Assuming a 1D ion diffusion process, the root of the mean-squared displacement of the Na + ions during the time periods of is 0.90 mm for thick film and 0.23 mm for thin film, the values of which are comparable to the film thicknesses (1.15 mm and 0.35 mm, respectively). This analysis suggests that the charge injection and the resultant change in 〈 ̅ 〉 do reflect the ion redistribution process within the film.
The NMR intensity changes seen on doping (Figure 3b iv-vi and Figure S5) provide insight concerning the ion dynamics within the film. When holes are injected, this is immediately detected by 23 Na NMR spectroscopy as fewer ions now bind on average to the PSS/PEDOT interfaces and the quadrupolar splitting decreases. However, the intensity of the anisotropic component of signal does not respond as quickly, as it takes a finite time for the displaced ions to drift out of the pancake (anisotropic) region. Charge neutrality must, however, be achieved and the more highly mobile (but more distant) Na + ions and possibly protons are the first to leave the film, as manifested by the faster NMR intensity decay for the Na + signal from the PSS region. The general migration of the (hydrated) ions in and out of the film is also seen via the changes in 1 H signal intensity.

Origin of the hysteresis
The discussion above develops a model that predicts a linear correlation between charge and splitting; however, the experimental data shows a degree of hysteresis and some deviation from the simple linear correlation (Figure 3c and Figure S14), likely due to other parameters that are not considered within this model. The first set of phenomena to consider, are the parasitic reactions that are seen during voltage-holding periods, particularly at higher voltages. Faradaic process that is presumed to be linear in time (as shown in Figure S9, S10). The contribution to the charge from Faradaic (parasitic) processes drops when the films are only charged to +/-0.2 and 0.4 V, suggesting that the parasitic reactions can be mitigated by using appropriate voltage windows. One possible source of these parasitic reactions is the reduction of dissolved oxygen by dedoped PEDOT 0 to form H 2 O 2 . 21,42 Other factors include possible changes in the PSS binding site local environment that might occur during doping/dedoping, resulting in changes in  Since the measured residual value of , 〈 ̅ 〉, is also affected by the shape of the PEDOT-rich domains (as illustrated in Figure   S12), the hysteresis may be caused by microstructural changes of the PEDOT:PSS (quantified via the order parameter, S) during electrochemical biasing, such as film swelling due to changes in water content. Other phenomena might be important: for example, Paulsen et al. 43 recently reported an unsymmetric rate of structural change of PEDOT:PSS during doping and dedoping by operando X-ray scattering, ascribing the transient structural behaviour to the complex polaron-bipolaron dynamics that affects electronic charge carrier dynamics.
Finally, 23 Na NMR integral changes with biasing show a dependence on cycling history. When parasitic (Faradaic) process is deducted from the observed charge, some experiments show Na + -to-electron coupling efficiencies of close to 98 % ( Figure S6). However, this strong coupling efficiency is not always seen, some experiments showing a greater increase in Na NMR integral than expected based on this charge passed ( Figure S7a). We ascribe these observations to competitive binding of Na + vs. H + to the sulfonate groups in PSS, the H + ions hoping more rapidly and likely responding more quickly to changes in voltage bias. The anions may also play a role, which cannot be excluded until better signal-to-noise ratio 35 Cl NMR spectra can be acquired.

Conclusions:
23 Na and 1 H NMR spectroscopy were used to study ionic gating of PEDOT:PSS aqueous films.
Via 23 Na NMR, we differentiate between two sodium environments. The first involves Na + in the PSS-rich matrix, which gives rise to a single peak indicating that (i) the Na + ions are highly mobile and (ii) the SO 3moieties on the PPS chains are randomly oriented in the film. The second environment is assigned to Na + ions in and nearby the PEDOT-rich regions and gives rise to a triplet in the 23 Na NMR spectrum. This triplet indicates that the ions experience a residual EFG (and thus non-zero residual quadrupolar coupling constant, 〈 ̅ 〉) that depends on the orientation of the film to the static magnetic field, confirming that the film comprises highly ordered microstructures -or "pancakes" -oriented parallel to the film substrate.
A negative linear correlation between 〈 ̅ 〉, and the charge stored within the film, was observed in the operando 23 Na NMR measurements. This was interpreted via a competitive binding and was rinsed with DI water and soaked in 1M NaCl solution.

NMR cell assembly
The photo of the symmetric cell used for NMR measurement are shown in Figure 2b. It The electrolyte was 1 M NaCl in DI water. One of the electrodes is placed inside the NMR coil to be monitored. The details about the cell designed can be found in Ref. 27 Cyclic voltammetry (CV) was performed using a Biologic SP-100 potentiostat to check the quality of the symmetric cell before and after operando NMR experiments.

Operando solid-state NMR
NMR experiments were performed using a Bruker Avance spectrometer operating at a magnetic field strength of 7.05 T, corresponding to 1 H and 23 Na Larmor frequencies of 300 and 79.4 MHz, respectively. In-house designed static probes with automatic tuning (ATMC) and matching capabilities and a double resonance (DR) probe (along with connections for an external battery cycler) were used for the measurements. The NMR coil used in this study was a 5-turn silver coated Cu solenoidal coil with a diameter of 11 mm. The electrochemical cell was aligned such that the electrodes were parallel to the applied field for operando measurement. In order to maximise the signal-to-noise ratio for the time-restricted measurement, single-pulse experiments were used for both 1 H and 23 Na NMR experiments.
Operando 23 Na NMR data of the thin PEDOT:PSS film (0.35 mm) were acquired on a DR probe using a 90° pulse corresponded to 10 μs (mutation frequency of 25 kHz) at 100 W with 256 transients and a recycle delay of 0.25 s (around 1 min per spectrum). The signal-to-noise ratio of the thick PEDOT:PSS film (1.15 mm) was much better than thin film, thus, only 32 transients were used (around 10 s per spectrum). The spin-lattice (T 1 ) relaxation of the 23 Na NMR signals were measured to be 33 ms using saturation recovery pulse sequence. The recycle delay of 0.2 s used here is sufficient-long for quantitative information. voltage) to cells. The NMR spectra were processed in Bruker Topspin software and further peak deconvolution and data processing was performed with R scripts.