Cation–polymer interactions drive water expulsion and deswelling in n-type ladder organic mixed conductors

MainOrganic mixed ionic-electronic conductors (OMIECs) enable key applications in bioelectronics1, optoelectronics2,3, energy harvesting and/or storage4,5,6 and neuromorphic computing7,8 due to their ability to simultaneously transport ionic and electronic charges9. The functional properties of OMIECs are governed by ion–polymer interactions, which dictate swelling, charge transport and material or device stability10,11,12,13. A molecular-level understanding of these interactions is crucial, as effective material or device optimization relies on comprehensive structure–property relationships. However, the mechanisms governing counterion uptake, charge mobility and morphology evolution during electrochemical doping remain poorly understood.A defining characteristic of OMIECs is their ability to uptake solvated ions, which influences their electronic, electrochemical and mechanical properties. Swelling is typically considered integral to OMIEC functionality, as counterion uptake expands the polymer network, facilitating charge compensation and enhancing volumetric capacitance14,15,16. Yet, the complexity of electrochemical doping is not fully described by swelling alone. Beyond simple volumetric expansion, solvated counterions dictate charge transport17,18, hydration dynamics or levels19 and morphology evolution20, ultimately shaping the material’s performance in devices such as actuators21, electrochemical transistors22, batteries23 and metasurfaces24,25.Even though the mechanisms governing ion uptake in OMIECs have been extensively studied, comparatively little is known about how ion–polymer interactions (and their impact on device operation) evolve during electrochemical doping. In particular, decoupling ion uptake and solvation shell movement has proved difficult, and ion injection together with water influx is the generally accepted convention. To enable directed optimization for a range of applications, a holistic understanding of ion–polymer interactions governing material function during electrochemical doping is required, which includes the role of water.Here we leverage a multimodal operando characterization approach to reveal how strong cation–polymer interactions influence charge localization, hydration dynamics and morphology evolution in OMIECs. Using poly(benzimidazobenzophenanthroline) (BBL)—a ladder-type, side-chain-free, electron-transporting OMIEC with highly reversible doping behaviour8,26—we demonstrate that at high doping levels, protic cations enhance hydrogen bonding, promoting charge localization and disrupting ion hydration, ultimately leading to mass and thickness reduction. Unlike conventional mass loss due to ion ejection27,28,29,30, we show that this behaviour is driven by water expulsion rather than ion removal, as schematically depicted in Fig. 1a. This molecular-scale structural transformation challenges existing models of ion uptake and charge compensation in OMIECs. Furthermore, deswelling at high doping density is potentially beneficial for long-term electrochemical stability and can enable new applications in tunable metasurfaces. To elucidate this process, we used an array of advanced operando characterization and simulation techniques, providing a comprehensive picture of cation–polymer interactions governing device function during electrochemical doping. Our findings establish a mechanistic understanding of how electrolytes influence ion–polymer interactions, charge localization and morphological response, offering new insights for the application of OMIECs in electrochemical energy devices, metalenses, neuromorphic computing and bioelectronics.Fig. 1: Decreasing mass and thickness with extensive doping of BBL.a, Schematic illustration of deswelling on cation injection. CE, counter electrode; RE, reference electrode; WE, working electrode. b, The measured Δf (left axis) and ΔD (right axis) for the third to eleventh overtones from EQCM-D measurements on BBL submerged in 0.1 M NH4Cl. The applied bias is decreased stepwise from −0.5 V to −0.2 V versus Ag/AgCl with 0 V applied between steps. Bottom panel includes the Δmass from the Sauerbrey equation (light red) and viscoelastic (VE) modelling (dark red) relative to the dry mass of the BBL film. While the VE model provides higher mass values, both methods show identical trends. The intermittent 0-V steps already show considerable Δmass due to the swelling on initial doping cycles. c, Summary of the Δmass from EQCM-D (Supplementary Figs. 2 and 3) at each voltage step (from the Sauerbrey equation) for 0.1 M NaCl (blue) and 0.1 M NH4Cl (red). Error bars denoting standard deviation (5 min) and demineralized water (>15 min) and dried under N2. Drop-cast 2H NMR films detached during soaking and were dried in air by pressing them onto lint-free cloth to obtain free-standing flakes. For GIWAXS, BBL was spin-coated from a 10 mg ml−1 MSA solution (1,000 rpm, 60 s) onto silicon substrates and rinsed in deionized water. For operando GIWAXS, films were lifted off in water and transferred to gold-coated porous silicon substrates.CVCV was performed using a three-electrode setup comprising the BBL-coated working electrode, a Pt mesh counter electrode and an Ag/AgCl pellet reference electrode. Measurements were carried out using a BioLogic SP200 potentiostat at a scan rate of 50 mV s−1 unless otherwise stated.EQCM-DTi/Au-coated quartz crystals (QSX 338, Biolin Scientific) were coated with BBL (see above) and mounted in a QSense module (Biolin Scientific) filled with electrolyte of interest. QCM data (frequency and dissipation shifts) were recorded using a QSense Analyzer (Biolin Scientific). Electrochemistry was performed using the Ti/Au-coated quartz crystal as the working electrode, a Pt film counter electrode, and an Ag/AgCl reference electrode (Dri-Ref-2SH, World Precision Instruments). A potentiostat (µAutolab, Metrohm) was used to apply the bias. Measurements typically consisted of 8 CV cycles (0.05 V s−1) from the lowest stable potential to +0.2 V versus Ag/AgCl, followed by a stepwise increase in applied bias from this lowest stable potential to +0.2 V in steps of 0.1 V for 60 s, with intermittent voltage holds at 0 V versus Ag/AgCl for 60 s. Qsense Dfind software (Biolin Scientific) was used to calculate the Δmass from the Sauerbrey equation (third overtone) or through viscoelastic modelling using the Dfind Broadfit algorithm.EC-AFMGlass substrates were cleaned by successive sonication in acetone, deionized water and isopropyl alcohol, and dried with N2. Cr (5 nm) and Au (50 nm) layers were thermally deposited, after which BBL was spin-coated from MSA solution, rinsed with deionized water and dried under N2. The films were patterned by photolithography and plasma etching. EC-AFM measurements were performed on a Dimension Icon XR (Bruker) using an electrochemical cell with a Pt wire counter electrode and an Ag/AgCl pellet reference electrode. Imaging was conducted in off-resonance mode using silicon nitride probes having a nominal tip radius of 20 nm and a spring constant of 0.7 N m−1. Films were initialized by eight CV cycles at 0.05 V s−1, after which measurements were performed by stepping the potential from negative to positive, allowing at least 60 s of equilibration before imaging. Image distortions were corrected using a first-order plane fit, and the film thickness was obtained as the height difference between the two dominant features (gold substrate and polymer film) measured from the peak-to-peak distance of Gaussians fitting the raw data histograms.GIWAXSGIWAXS measurements were performed at the 11-BM CMS beamline at NSLS-II with a Pilatus 300K detector. A beam energy of 13.5 keV was used for all measurements. The distance between the sample and the detector was calibrated using a BeH standard. For ex situ measurements, an angle of incidence of 0.14° was used with a 10 s exposure time. For operando measurements, an electrochemical cell (Supplementary Fig. 31) was placed in a sealed chamber with Kapton windows, where humid N2 was continuously flowing to reduce oxygen side reactions. The whole chamber was kept in vacuum to minimize air scatter. The X-ray scattering data were reduced in Python using the open-source pyFAI and pygix modules44,45, following the general procedure reported in refs. 25,46. The out-of-plane (qz) and in-plane (qxy) linecuts were taken from cake slices from χ = −20° to 20° and 65° to 87° with respect to the qz direction, respectively. Small-angle scattering between χ = −3° and 3° was excluded from the out-of-plane linecuts. Scattering peaks were fitted to Gaussian–Lorentzian lineshapes using the lmfit package in Python v.3.9. Electrochemical measurements were taken on a Gamry potentiostat in a three-electrode setup with a leakless Ag/AgCl reference pellet (eDAQ) and a platinum wire counter electrode. The electrolytes (0.1 M NaCl and 0.1 M NH4Cl) were degassed before cell assembly.Operando IR absorption spectroscopyA back-etched silicon wafer (universal ATR element, IRUBIS) was sputter-coated with indium tin oxide while at 300 °C, yielding ~20 nm indium tin oxide with a sheet resistance of ~220 Ω per square. This wafer was subsequently covered with BBL as described above. The sample was loaded into a spectroelectrochemical cell (Jackfish J1, PIKE Technologies) and mounted atop the Veemax III accessory (PIKE Technologies) for use with the Fourier transform IR spectrometer (Spectrum 3, PerkinElmer). The spectroelectrochemical cell was filled with the chosen electrolyte. A Pt wire counter electrode and an Ag/AgCl reference electrode (in saturated KCl with glass frit, PN 162-4723 PIKE Technologies) were submerged in the electrolyte to complete the three-electrode setup. Spectra were recorded using a liquid nitrogen-cooled mercury cadmium telluride detector and with the Veemax III at 55°. Measurements spanned 4,000–450 cm−1, with a 1 cm−1 datapoint spacing, and included 50 repeats. Bias was applied for 80 s (SP200 potentiostat, BioLogic) during spectral acquisition, followed by a 30-s dedoping step at 0 V versus Ag/AgCl. Pristine BBL films submerged in electrolyte were used as the background IR spectrum, and the differential transmittance spectra reported were referenced to this background.Operando ellipsometrySpectroscopic ellipsometry was measured with an RC2 Mueller Matrix ellipsometer (J.A. Woollam) using an electrochemical ellipsometry cell (redox.me). At a 70° incident angle, the ellipsometer response of a layer of BBL on a Cr/Au-coated Si substrate was recorded under varying applied bias. Baseline measurements were performed on the empty cell (with substrate) and on the cell filled with electrolyte to determine the electrolyte refractive index and the specifications of the substrate used (including Au/Cr). Fitting was performed using CompleteEase (J.A. Woollam).Terahertz spectroscopyTerahertz time-domain spectroscopy was performed using an ultrafast Ti:sapphire amplified laser system (Astrella, Coherent; 35-fs pulse duration, 800-nm centre wavelength, 1-kHz repetition rate, 6 W). Terahertz pulses were generated by optical rectification of part of the fundamental beam in a ZnTe crystal and directed onto the sample using two off-axis parabolic gold mirrors. The transmitted terahertz pulses were refocused by two additional off-axis parabolic mirrors onto a second ZnTe crystal for detection, where they overlapped spatially and temporally with a gate pulse derived from the fundamental beam, and electro-optic sampling was used. The terahertz pulses induced a birefringent refractive index in the detection ZnTe crystal, altering the polarization of the gate pulses from linear to slightly elliptical. The elliptically polarized gate beam was passed through a quarter-wave plate to render it nearly circularly polarized. The two orthogonal polarizations were split by a Wollaston prism and focused onto a balanced photodiode, which measured the intensity difference between them, proportional to the electric field. By changing the time delay between the terahertz and gate pulse using a micrometre translation stage, the terahertz electric field was measured at different time delays. The resulting terahertz pulses had a duration of ~1 ps and a spectral bandwidth of 0.1–2.5 THz.For terahertz spectroscopy on electrochemically doped samples, polymer films were deposited between two parallel Au electrodes (5-mm spacing) and mounted in an electrochemical cell. A thin layer of electrolyte (0.1 M NaCl or NH4Cl in water) was applied, and an Ag/AgCl pellet electrode served as the reference. The Au electrodes were short-circuited during the terahertz measurements. Frequency-dependent complex conductivity spectra were obtained by measuring the transmitted terahertz electric field through the electrolyte–doped polymer–substrate and a reference transmission through the electrolyte–substrate or electrolyte–neat polymer–substrate. Fourier transforms of both time-domain signals yielded the experimental complex transmission coefficient, which was fitted to a theoretical transmission coefficient accounting for electric-field-induced changes, pulse propagation through each layer, transmission at each interface and multiple reflections within the polymer film to obtain the complex conductivity spectra40. The complex terahertz conductivity spectra are analysed using the Drude–Smith model to extract charge-carrier densities and charge mobility contributing to short-range, nanometre-scale transport41. In this phenomenological model, the complex conductivity is given by:$$\mathop{\sigma }\limits^{ \sim }(\omega )=\frac{{\varepsilon }_{0}{\,\omega }_{{\rm{P}}}^{2}\tau }{(1-i\omega \tau )}\left[1+\frac{{c}_{1}}{1-i\omega \tau }\right]$$where ωP is the plasma frequency, ε0 the vacuum permittivity, τ the scattering time and c1 the localization parameter. The effective short-range mobility was calculated as \({\mu }_{{\rm{eff}}}=\frac{e\tau }{{m}^{\ast }}\,(1+{c}_{1})\), where m* is the effective mass. The short-range mobility reflects charge motion over nanometre distances during the ~1-ps duration of the terahertz pulse.Operando 2H NMRThe operando NMR cell consisted of a gold mesh (1.3 × 0.5 cm2, 0.004 mm thick; Sigma GF14297430) and two free-standing BBL and PEDOT:PSS films (0.6 × 0.5 cm2) separated by a borosilicate glass fibre (Whatman, GF/A). PEDOT:PSS (Clevios PH1000, Heraeus Holding) was mixed with 1 wt% 3-glycidoxypropyltrimethoxysilane (Sigma-Aldrich), filtered through a 0.45-µm polyvinylidene fluoride filter and heated at 70 °C for 12 h. The resulting self-standing films were rehydrated in deionized water, rinsed with deionized water (1 h soaking) three times and stored in D2O. 0.1 M NaCl or NH4Cl in D2O was used as the electrolyte. An Ag/AgCl-coated silver wire (0.127 mm diameter; Sigma) served as a pseudo-reference electrode. The cell was configured to detect only one electrode at a time, with the BBL electrode positioned within the NMR coil. Further details of the cell design can be found in refs. 32,47. CV was performed using a Biologic SP-100 potentiostat to check the quality of the operando cell before and after the experiments. Operando NMR electrochemical measurements were performed using the same Biologic VSP potentiostat.Operando 2H NMR experiments were performed on a Bruker Avance spectrometer operating at 7.05 T (2H Larmor frequencies of 40 MHz) using in-house-designed static and double-resonance probes and a six-turn silver-coated copper solenoidal coil (11 mm diameter). The electrochemical cell was aligned with the electrodes parallel to the magnetic field. Quantitative single-pulse experiments were conducted using a 10-μs (100 W) 90° pulse, a recycle delay of 0.2 s and 32 scans. A spin-lattice relaxation time of 33 ms was measured for anisotropic water with a saturation recovery pulse sequence. Thus, the chosen 0.2 s recycle delay was sufficiently long to ensure quantitative information extraction. Spectra were referenced to D2O at 4.75 ppm and processed using Bruker TopSpin v.4.0.5 and MATLAB.MD simulationIon and water uptake in BBL crystallites were modelled using MD. The system consisted of 4 layers of 10 stacked BBL tetramers surrounded by an explicit electrolyte containing 13,808 water molecules, 1,008 Na+ or NH4+ cations and 1,008 Cl− anions, corresponding to an electrolyte concentration of approximately 4 mol l−1. Periodic boundary conditions were applied, ensuring that the BBL tetramers behave as infinite polymer chains. The central BBL crystallite structure was obtained from a conformational search of the syn conformer following established procedures48,49. Interatomic interactions were described using the Dreiding force field (Materials Studio software), with atomic charges derived from the DDEC6 partition scheme applied on a periodical DFT charge density50. The density was computed on a BBL dimer repeated along the molecular axis using the projected-augmented wave method with the Perdew–Burke–Ernzerhof functional as implemented in the VASP code51,52 (see Supplementary Note 8 for details). Water molecules and ions were subsequently added to the BBL crystallite, using the well-established extended simple point charge model to describe interactions involving water molecules53.A 10 ns-long MD simulation (NPT ensemble; p = 1 atm, T = room temperature) was conducted to equilibrate the system before doping the BBL chains. The doping procedure involved adding one charge every 200 ps during a subsequent NPT simulation. A doping level of 100% (200%) is achieved after 32 (64) ns. Rather than assuming a uniform distribution of charges, atomic charge differences between the neutral and charged BBL were computed from the DDEC6 partition scheme applied on their respective periodical DFT charge densities. A fraction of the charge was assigned to each atom, proportional to these computed changes, thereby properly accounting for differences in electronegativity between atoms. During the doping simulations, atomic coordinates were saved every 200 ps for postprocessing analysis. This procedure was replicated five times for each electrolyte. All simulations, except the conformational search performed in Materials Studio, were carried out using the LAMMPS package54, with a 12.0-Å cut off for van der Waals interactions and the particle–particle–particle–mesh method for electrostatic interactions.To characterize the systems, the number of water molecules and ions within the interlayer regions, the interlayer and/or intermolecular distances, the coordination numbers between ions and the nitrogen and carbonyl groups, and the interaction energies between cations and the BBL layers were analysed. To this end, the centres of mass of each BBL chain were calculated, excluding the first and last chains of each π-stacked layer to avoid edge effects. These 32 points served as reference points to define three surfaces, which were converted into volumes by using the long axis of the BBL polymer chains as the third dimension, as illustrated in Supplementary Fig. 32. Further details on the analysis are listed below: The number of water molecules and ions in the interlayer regions was determined by counting the number of species within the three defined volumes for each of the 320 frames saved during the MD simulation. The intermolecular distance was calculated as the average distance between BBL chains in a given frame. The interlayer distance was measured as the distance between the best-fit plane going through the centres of mass of one layer and the centre of mass of the adjacent layer. Coordination numbers between ions and the carbonyl groups or nitrogens of the BBL chains were computed as the average number of oxygen (carbonyl groups) or nitrogen atoms within a radial distance of 3.0 Å (for Na+) or 3.5 Å (for NH4+), relative to the centre of mass of each cation for each frame. The choice of the radial distances was made in such a way as to select only the first coordination shell around the ions. Interaction energies were calculated as the sum of the van der Waals and electrostatic interaction energies between the two central BBL layers and the ions located in the interlayer region (excluding water molecules). These energies were then normalized by the number of ions per frame to facilitate comparison between ion types.