We use 1H, 2H, and 7Li nuclear magnetic resonance to investigate local and diffusive dynamics of LiCl-7H2O and LiCl-7D2O solutions in pristine and functionalized silica nanopores in a component-selective manner. Recently, we showed that the solution dynamics become slower when the diameter of the pristine pores is reduced. Here, we determine the effects of (aminopropyl)triethoxysilane and dye surface functionalizations on the motions of the water molecules and lithium ions from ambient temperatures down to the glass transition. The local and diffusive solution dynamics are similar in both functionalized pores but, on average, slower than in pristine pores with comparable diameters. When the exchange between different confinement regions is sufficiently slow at reduced temperatures, bimodal water and lithium dynamics may be observed. We attribute this bimodality to bulk-like motion in the pore centers and slowed-down motion at the pore walls. For the lithium ions, a bimodality observed in the pristine pores is absent in the functionalized ones. We conjecture that the steric hindrance and electrostatic interactions associated with the grafted functional groups interfere with the formation of a defined electric double layer, while the enhanced surface roughness and unequal charge distribution result in overall slower dynamics. Thus, the nature of the walls is an important parameter for the solution dynamics. Thereby, in situ measurements of the pH value inside the silica pores using the grafted dye molecules reveal that observed changes in the pH value in response to the surface functionalization are of limited relevance for the water reorientation.

The transport of dissolved ions through nanopores is of paramount relevance for a large variety of processes in nature and technology. Prominent examples include ion channels in cellular membranes, ion-exchange membranes for desalination, or nanofluidic devices. In view of this importance, the dependence of the transport properties on the pore characteristics received considerable scientific attention. Not only the sizes of the ions and the energetics of their hydration shells but also the geometry, charge, and functionalization of the confinement were proposed as critical parameters.1–5 In these studies, very narrow carbon nanotubes received considerable attention because of their ion selectivity and enhanced flow, which is associated with weak electrolyte–wall interactions and, hence, impaired by tube modifications.5–7 In sufficiently wide pores, velocity profiles and structural variations across the channels add to the complexity.2 Considering this challenging situation, it is worth determining the equilibrium structures and dynamics of liquid electrolytes in nanopores.

For this purpose, many studies focused on the relevant case of silica pores. At electrolyte–silica interfaces, the dissolved ions compensate the negative charge of the silica surface, which is associated with the ionization of the silanol groups and, hence, depends on their surface density and the pH value.8 This situation is described by the Gouy–Chapman–Stern model, which predicts an electric double layer (EDL) comprising an inner layer of more or less hydrated counterions, usually referred to as Stern layer, and an outer layer, or, equivalently, diffusive layer.9–11 However, the microscopic structure of the EDL is still elusive. It is expected to depend on the type and concentration of the ions, the properties of the surface, and the experimental conditions, e.g., the pH value.12–15 Likewise, silica nanopores affect the dynamics of electrolytes. For LiCl solutions, it was shown that the combined effects of confinement and salt reduce the mobility of water.16,17 Moreover, it was found that the glass transition temperature Tg is higher in pores with diameters of a few nanometers than in the bulk.18 Similarly, a slowdown of diffusion was reported for other salt solutions in severe silica confinements.19,20 Besides the pore diameter, the chemical structure and surface charge of the pore walls determine the dynamics of ions in the vicinity, and as a result, the mobility may even be higher in the confinement than in the bulk.21–23 

Despite this valuable information, the current understanding of ion dynamics in nanopores is far from being complete. In particular, it is still elusive in which way the dynamics of the dissolved ions and their hydration shells depend on the diameter of the pores and the properties of their walls. This lack of knowledge pertains especially to mesopores with diameters of a few nanometers where strong concentration and mobility gradients can exist across the pores, e.g., due to EDL formation.24–28 In addition, the dynamics of aqueous systems may generally depend on the pH value,29 e.g., owing to a different relevance of the Grotthuss mechanism. Therefore, it is of wide-ranging interest to determine to which degree different dynamics of aqueous systems in the bulk and in confinements result from the fact that the pH values can be largely different in these situations, as a result of acidic or basic groups on the inner surfaces. Nuclear magnetic resonance (NMR) spectroscopy is a powerful tool to address these questions because it allows for component-selective studies of dynamics in nanopores over broad dynamic ranges.30,31

Recently, we combined 1H, 2H, and 7Li NMR to determine the dynamical behaviors of aqueous LiCl solutions in the bulk32 and in silica pores with various diameters in the range of 2.8–5.4 nm.31 The main focus was on the glass-forming LiCl-7H2O and LiCl-7D2O solutions. We ascertained the local dynamics of the lithium ions and the water molecules from room temperature down to the glass transition by combining spin-lattice relaxation (SLR), line-shape analysis (LSA), and stimulated-echo experiment (STE) studies, which allowed us to determine correlation times τ over a range of more than ten orders of magnitude. Moreover, we characterized the diffusive motion of the components on length scales of several hundreds of nanometers by self-diffusion coefficients D available from static field gradient (SFG) measurements. This broad NMR approach revealed that the silica nanopores cause a slowdown of the electrolyte dynamics on all length scales, which is stronger at lower temperatures and in narrower pores and is more prominent for the lithium ions than the water molecules. Furthermore, bimodal decays of correlation functions for sufficiently wide confinements suggested that bulk-like ion dynamics in the pore centers can be distinguished from significantly retarded ion dynamics at the pore walls, possibly in a Stern layer.

Here, we study the dependence of electrolyte dynamics on the properties of the pore walls. For this purpose, we exploit the capabilities of our 1H, 2H, and 7Li NMR approaches to compare the dynamical behaviors of LiCl-7H2O and LiCl-7D2O solutions in pristine and functionalized silica mesopores. In general, a modification of the silica surfaces may affect the dynamics of the salt solutions in different ways. Specifically, the functionalization is expected to alter the surface charge and, hence, the electrolyte-silica interactions, to cause a sterical hindrance for EDL formation, and to change the pH value inside the confinement. Two modified silica confinements are considered. First, SBA-15 silica is functionalized with (aminopropyl)triethoxysilane (APTES). Second, the thus functionalized material is further modified by adding a thiazol-based dye, which enables in situ measurements of the local pH value inside the silica pores.33 Thus, our functionalization strategy yields insights not only into the effects of the surface chemistry on the dynamics of embedded electrolytes but also into the role of the pH value, an aspect that was largely ignored in previous studies of confined aqueous systems.

To investigate local electrolyte dynamics, we perform 2H (I = 1) and 7Li (I = 3/2) NMR experiments in homogeneous magnetic fields. In these studies, the nuclear quadrupole moments interact with the electric field gradients at the nuclear sites, giving rise to the observed quadrupolar frequencies,34,35
ωQ(θ,ϕ)=±δ23cos2θ1ηsin2θcos2ϕ.
(1)
Here, δ and η are the anisotropy and asymmetry parameters of the associated quadrupolar interaction tensor, respectively, and the angles θ and ϕ describe the orientation of this tensor with respect to the applied magnetic field B0. In our 2H NMR studies, the interaction tensor is nearly axially symmetric and the distinguished principal axis is aligned with the orientation of the O–D bond. In the 7Li NMR approaches, it mirrors the charge distribution in the environment of the lithium ion. Hence, the fluctuations of the quadrupolar frequencies inform about the rotational dynamics of the water molecules and about structural changes around the lithium ions, which arise when the ions move to a different environment and/or the hydration shells are regrouped.
Considering that water or ion fractions n with distinguishable dynamics may coexist in the studied samples, we fit the buildup of the nuclear magnetization after saturation, M(t), with a sum of up to two stretched exponential contributions,
M(t)=nmn1exptT1,nβ1,n+ms.
(2)
Here, mn denotes the equilibrium magnetization of the species n; T1,n and β1,n are the relaxation time and stretching parameter of the associated SLR step, respectively; and ms considers minor imperfections in the saturation. When distributions of correlation times exist, SLR will be nonexponential unless an exchange between dynamically distinguishable fractions restores the ergodicity during the magnetization buildup. Such exchange can result from molecular diffusion and/or spin diffusion.
The SLR times T1,n depend on the spectral densities J2,n(ω), which describe the fluctuations of the quadrupolar frequencies of a given spin species as a result of molecular dynamics. In temperature-dependent measurements, the relaxation time T1,n is minimum when the correlation time τn of the probed motion is of the order of the inverse Larmor frequency of the observed nucleus, explicitly, for ω0τn ≈ 0.6 and, hence, for molecular dynamics in the nanosecond regime. In our 2H SLR studies, J2,n(ω) describes the reorientation of the O–D bonds, and the anisotropy and asymmetry parameters have defined values of δ ≈ 2π · 160 kHz and η ≈ 0.32 Under such circumstances, exponential SLR steps (β1,n = 1) can be analyzed using the relation36,
1T1,n=215δ2J2,n(ω0)+4J2,n(2ω0).
(3)
To consider the complex molecular dynamics of our samples,31 we assume a Cole–Cole (CC) spectral density,
J2,nccω=ω1ωτnccαnsinπ2αn1+2ωτnccαncosπ2αn+ωτncc2αn.
(4)
It is characterized by the correlation time τncc and the width parameter αn and proved to be suitable in previous studies on confined aqueous liquids.37–46 Further details of our SLR analysis can be found in previous work.32 
Correlation times, which are independent of the shape of the spectral density, can be extracted from the 2H SLR results at high temperatures, where ω0τ ≪ 1 applies and the buildup of the magnetization is exponential. Under such circumstances, mean correlation times ⟨τ⟩ are available from the single T1 value according to
τ=321T1δ2.
(5)
In the 7Li SLR studies, we refrain from a determination of temperature-dependent correlation times because δ and η do not have defined values for this nucleus.

For 7Li, the quadrupolar interaction affects the resonance frequencies of the two satellite transitions (STs) |±32|±12 between the Zeeman levels but, in first order, not that of the central transition (CT) |+12|12. Thus, when molecular dynamics is sufficiently slow at low temperatures, the 7Li NMR spectra of disordered materials consist of a broader ST line, which reflects the distributed quadrupolar frequencies, and a narrower CT line, which is only affected by the weaker dipolar couplings of the 7Li spins.35,47 When the temperature is increased, the 7Li NMR spectra start to narrow as soon as the correlation times of the lithium ion dynamics become similar to the inverse widths of the ST and CT lines in the static case. In 7Li LSA, we determine the temperature at which the width of the CT line has decreased to half of its static value ΔωCT and assign a correlation time of τp = 1/ΔωCT, yielding information about dynamics in the microsecond regime.

In 2H and 7Li STE studies, we investigate slow motions in the range of 10−4 to 101 s by correlating the quadrupolar frequencies ωQ during two short evolution times te, which are separated by a longer and variable mixing time tm. Specifically, we record the correlation functions34,35,47,48
F2cc(tm)cosωQ(0)tecosωQ(tm)te,
(6)
F2ss(tm)sinωQ(0)tesinωQ(tm)te,
(7)
where the brackets ⟨⋯⟩ denote the ensemble average. We study F2cc(tm) in 2H NMR and F2ss(tm) in 7Li NMR.
For analysis, we again consider the possible bimodality of water and ion dynamics and fit the STE data (ii = cc, ss), if required, with a sum of two decays,
F2ii(tm)=npn1FDn(tm)+FRnii(tm).
(8)
Here, pn is the contribution of species n. We describe the STE decays due to molecular dynamics with stretched exponential or, equivalently, Kohlrausch (K) functions,
Dn(tm)=exptmτnkβnk,
(9)
which are characterized by the correlation times τnk and the stretching parameters βnk. Moreover, we consider the SLR damping of the STE decays at long mixing times by
Rnii(tm)=exptmT1,niiβ1,nii.
(10)
Finally, F accounts for a possible residual correlation owing to immobile particles or anisotropic motion.
For the 2H STE analysis of F2cc(tm), we exploit that the SLR damping can be determined in independent SLR studies, and hence, Rncc(tm) can be fixed by using T1,ncc=T1,n and β1,ncc=β1,n. Moreover, for reasons discussed below, we suppose that one water species (n = 1) shows isotropic reorientation so that F = 0 for the used te values, while the dynamics of the other (n = 2) occurs outside the experimental window, i.e., D2(tm) = 1. With these assumptions, Eq. (8) simplifies to
F2cc(tm)=p1cD1(tm)R1(tm)+cR2(tm).
(11)
Hence, apart from the amplitude factor p, there are three fit parameters, explicitly, the Kohlrausch parameters τ1k and β1k and the fraction c of the constant contribution. In the 7Li STE studies of F2ss(tm), the SLR damping Rss(tm) is not critical in the studied temperature range because the 7Li SLR is slower and exponential.
To compare SLR, LSA, and STE findings for complex molecular dynamics, it is useful to calculate peak correlation times τp,n from the respective fit results.43–46 They correspond to the peak positions of the related dynamic susceptibilities and depend less on the shape of the distribution of correlation times than rate or time averages. While τp,n=τncc for the CC function, the peak correlation time is available from the Kohlrausch parametrization of the STE decays based on43 
τp,nτnk=1.7850.871βnk0.029(βnk)2+0.114(βnk)3.
(12)
Moreover, we expect that our LSA approach also essentially yields peak correlation times. By contrast, the SLR analysis in the high-temperature limit provides us with mean correlation times ⟨τ⟩.
Self-diffusion coefficients D can be measured in SFG NMR experiments, which utilize a magnetic field with a static gradient g.49 In our case, the diffusion is not free but restricted by the cylindrical MCM-41 and SBA-15 pores. Therefore, SFG NMR probes molecular displacements along the pore axes and, hence, one-dimensional (1D) diffusion,31,46,50,51 leading to STE decays,49,52
S(tm,te)0πexpq2tdDcos2ϑsinϑdϑ.
(13)
Here, td=tm+23te is a diffusion time and q = gγte can be considered as a generalized scattering vector, which sets the length scale of the diffusion measurement. Moreover, ϑ denotes the angle between the pore axis and the field gradient and the integral takes into account the powder average over the random orientations of the silica nanoparticles in our samples. In our SFG setup, diffusion on length scales ∼0.1–10 μm can be probed.53 

In the 7Li SFG studies, we record STE decays S(tm) for several fixed evolution times te and determine the diffusion coefficients D from global fits with Eq. (13), supplemented by a 7Li SLR damping factor exp(−tm/T1). In our 1H SFG approaches, we use the Hahn-echo (HE) sequence, where the mixing time tm is absent, and record S(te). To correct the HE decays for any contributions apart from diffusion, we divide these data by Shom(te), which is obtained from identical HE experiments in a homogeneous field of the same strength. Then, the diffusivities D can be obtained by fitting the thus corrected HE decays with Eq. (13) and setting tm = 0.

The used MCM-41 silica materials with pore diameters of d = 2.8 nm and d = 3.7 nm were previously synthesized and characterized.31,44 The SBA-15 compound with d = 5.4 nm was purchased from Sigma-Aldrich and described in a previous study.50 Likewise, the SBA-15 modified with APTES groups was synthesized and characterized in prior work31 and references therein. It was not obtained by grafting to the purchased SBA-15 material but prepared via co-condensation. The grafting density amounts to ∼3 APTES groups per nm2. Here, it is denoted as 6.8 nm-A. To determine the pH value inside the porous material and to analyze the effect of a considerably functionalized pore on ion and water dynamics, this SBA-15 APTES material was further modified by adding a ratiometric thiazol-based dye, explicitly, 5-methoxy pyridilthiazole, in a previous study.33 A grafting sensitivity of ∼1 dye molecules per nm2 was obtained. We refer to this modified silica framework as 5.8 nm-A+D. Relevant information about the pristine and modified confinements is summarized in Table I.

TABLE I.

Pore diameters and specific volumes of the used pristine and modified MCM-41 and SBA-15 silica materials as obtained from nitrogen adsorption measurements in previous studies.31,44,50,54 The SBA-15 material functionalized with APTES groups is indicated by the letter “A,” whereas the SBA-15 APTES further functionalized with the thiazol-based dye is labeled by “A+D.”

TypeNameDiameter (nm)Volume (cm3/g)
MCM-41 2.8 nm 2.8 0.70 
MCM-41 3.7 nm 3.7 0.89 
SBA-15 5.4 nm 5.4 0.58 
+APTES 6.8 nm-A 6.8 0.74 
+APTES + dye 5.8 nm-A+D 5.8 0.49 
TypeNameDiameter (nm)Volume (cm3/g)
MCM-41 2.8 nm 2.8 0.70 
MCM-41 3.7 nm 3.7 0.89 
SBA-15 5.4 nm 5.4 0.58 
+APTES 6.8 nm-A 6.8 0.74 
+APTES + dye 5.8 nm-A+D 5.8 0.49 

For the sample preparation, we followed the same protocol as in previous studies.31,32 Anhydrous LiCl, dried for 1–2 days at 10−5 mbar, was dissolved in distilled H2O or D2O (99.9% D, purchased from Sigma-Aldrich) at a molar water to salt ratio of R = 7, which is close to the eutectic composition of those mixtures.55,56 The LiCl-7H2O and LiCl-7D2O solutions were then degassed 6–8 times using the freeze–pump–thaw-procedure prior to adding them to the porous silica materials.

The mesoporous materials were dried at 10−5 mbar for 1–2 days before loading them with the aqueous salt solutions. The pore filling degree was set to 90% of the pore volume to avoid excess solution outside the silica confinements. To prepare the NMR samples, the loaded silica materials were filled in borosilicate glass tubes, which were flame-sealed immediately afterward. The NMR samples were stored for several days prior to the measurements to allow for equilibration and confirmed to be leak proof during all measurements by repeated weighing.

Except 5.8 nm-A+D, the pristine and functionalized silica materials were loaded with LiCl-7H2O or LiCl-7D2O to enable 1H NMR studies of water diffusion and 2H NMR studies of water reorientation, respectively. The dye-functionalized pores were only filled with LiCl-7D2O due to a very limited amount of the sample material.

To study the influence of the various nanoconfinements on the pH value, we dispersed different amounts of the pristine or modified silica particles in LiCl-7D2O solution and measured the pH value of the remaining excess solution outside the silica pores using a pH meter. The results for the different silica materials and mass fractions are presented in Table II. Further details and results of these measurements can be found in the supplementary material. The pH value of the LiCl-7D2O before particle loading is 6.35. The addition of pristine silica particles, due to the presence of the acidic silanol groups at the inner surfaces, lowers the pH value of the solutions (see Table II). This effect is stronger for high than for low mass fractions of silica particles due the different total contact area between the solution and the inner surfaces. For a given mass fraction, the MCM-41 samples have lower pH values than the SBA-15 samples because the former have larger specific surface areas than the latter and, hence, feature more of the acidic silanol groups per volume. Furthermore, the pH value is lower for the pristine than the functionalized silica particles. This effect can be rationalized by the fact that the acidic silanol groups are partly replaced by the functional groups, which contribute basic amino groups (APTES) and pyridine rings (dye). In addition, our measurements showed that the pH values remained unchanged for several hours after sample preparation. To determine the difference of the pH values of the excess solution outside the pores and the confined solution inside the pores, we used the dye functionalizations for fluorescence measurements (see below).

TABLE II.

pH values of the LiCl-7D2O solution for low (∼0.08 wt. %) and high (∼5 wt. %) mass fractions of the studied pristine and functionalized silica materials. For these mass fractions, the volume of the solution is about 1430 times and 20 times the total pore volume, respectively. The pH measurements were performed 8 h after the sample preparation. The results for intermediate silica mass fractions and shorter waiting times can be found in the supplementary material. The limited amount of the dye-functionalized silica material prohibited pH measurements at high mass fractions.

SamplepH (low fraction)pH (high fraction)
2.8 nm 5.96 ± 0.02 2.97 ± 0.003 
3.7 nm 6.14 ± 0.03 2.86 ± 0.01 
5.4 nm 6.06 ± 0.03 3.20 ± 0.11 
6.8 nm-A 6.23 ± 0.05 4.67 ± 0.02 
5.8 nm-A+D 6.66 ± 0.30 ⋯ 
SamplepH (low fraction)pH (high fraction)
2.8 nm 5.96 ± 0.02 2.97 ± 0.003 
3.7 nm 6.14 ± 0.03 2.86 ± 0.01 
5.4 nm 6.06 ± 0.03 3.20 ± 0.11 
6.8 nm-A 6.23 ± 0.05 4.67 ± 0.02 
5.8 nm-A+D 6.66 ± 0.30 ⋯ 

The NMR experiments and setups were described in previous works.31,32 Briefly, the NMR studies in homogeneous B0 fields were performed at Larmor frequencies ω0 of 2π · 91.2 MHz (1H), 2π · 46.1 MHz (2H), and 2π · 62.9 MHz (7Li). The saturation-recovery sequence in combination with a solid-echo detection were employed in the SLR studies, and appropriate pulse sequences and phase cyclings were used for the STE experiments.34,35,47,48 The SFG measurements were performed utilizing a custom-made superconducting magnet, which features two coils in anti-Helmholtz geometry to produce a magnetic field with strong static gradients up to 180 T/m.57 The 7Li diffusivities were acquired utilizing the STE pulse sequence and a sample position where the magnetic field and field gradient amount to 3.8 T and 72.8 T/m, respectively. The 1H self-diffusion coefficients were measured using the HE sequence, B0 = 2.2 T, and g = 105 T/m. The 2H diffusivities were extracted from STE experiments at B0 = 3.8 T and g = 140.3 T/m. In all NMR studies, the temperature was controlled via liquid-nitrogen cryostats, allowing for a temperature accuracy better than ±1 K and a temperature stability better than ±0.5 K.

The above discussed pH measurements revealed that the pH values of the excess solutions outside the pores depend on the type of the mesoporous material, while this approach did not provide direct information about the pH value inside the silica pores. This information can be obtained by using the dye molecules covalently linked to the silica frameworks in the 5.8 nm-A+D material as reporters,33 which is one of the reasons for the following studies of the dynamical behavior inside this confinement. The in situ pH measurements are described in more detail in a previous study33 and the supplementary material. The approach exploits that the fluorescence signals of the protonated and nonprotonated states of the dye molecule can be resolved in the spectrum. Hence, the ratio of the signal intensities yields the relative relevance of these states, and an analysis in dependence on the pH value of the solution provides access to the pKa value of dye-functionalized silica pores.

In Fig. 1, we show the pH-dependent ratio of the dye molecules in the protonated and nonprotonated states when linked to the solution-filled silica matrix and dispersed in a bulk solution, respectively. We see that the curve is shifted to lower pH values for dyes being located in the confinement of nanopores. Both in the confinements and the bulk, the presented pH-dependent ratio of the protonated and nonprotonated states of the dyes is comparable to the ratios reported in previous studies using similar33,58 or different59–61 types of pH-reporter dyes. Following a previous study33 and performing sigmoidal fits, we find a change in the pKa value from pKa = 4.1 for the free dye in a bulk solution to pKa = 3.3 for the bound dye in the 5.8 nm-A+D material. Using this dye, a shift by ∼1 pH unit was also observed for a different solution.33 These experimental findings are consistent with theoretical results for protonation equilibria in nanoconfinement.62–64 Therefore, we expect for the studied silica frameworks that the pH value of the solutions inside the confinements is about one unit smaller than that of the solutions outside, which is readily accessible and reported in Table II. Under the observed conditions, the available silanol and amino groups are largely ionized. Thus, the inner surfaces of the pristine silica are negatively charged, while those of the functionalized silica feature an uneven charge distribution with, e.g., Si–O and NH3+ groups.65,66

FIG. 1.

Ratio of the intensities of the signals from the protonated and nonprotonated states of the dye molecule as a function of the pH value. Results for the dye molecules linked to the silica framework (solid symbols) and dispersed in a bulk D2O solution (open symbols) are compared. The solid lines are sigmoidal fits yielding pKa = 3.3 in the confinement and pKa = 4.1 in the bulk.

FIG. 1.

Ratio of the intensities of the signals from the protonated and nonprotonated states of the dye molecule as a function of the pH value. Results for the dye molecules linked to the silica framework (solid symbols) and dispersed in a bulk D2O solution (open symbols) are compared. The solid lines are sigmoidal fits yielding pKa = 3.3 in the confinement and pKa = 4.1 in the bulk.

Close modal

Differential scanning calorimetry (DSC) measurements yield the glass transition temperatures Tg of the confined LiCl-7D2O solutions. Figure 2 shows the thermograms of heating runs at 10 K/min. In the wider pristine and modified silica pores with diameters of ∼6 nm, we find clear glass transition steps. The glass transition temperatures Tg are indicated by vertical lines and amount to ∼139 K (see Table III), in agreement with the value for the LiCl-7D2O bulk solution.32 Moreover, there are minor melting peaks at ∼204 K for these confinements (see the supplementary material). In the narrower pristine pores, the glass transition differs from the bulk behavior and a melting peak is absent. Tg shifts to higher temperatures by ∼6 K in the pores with d = 3.7 nm, while the glass transition step becomes blurred and possibly bimodal with onsets near 142 and 148 K in the narrowest pores (d = 2.8 nm), suggesting inhomogeneous solution dynamics.

FIG. 2.

DSC thermograms revealing the glass transition steps of LiCl-7D2O in the studied pristine and modified silica confinements. The measurements were performed with a heating rate of 10 K/min. Sloped baselines were subtracted from the measured signals. The curves are shifted vertically for better visibility. The dotted lines indicate the onset glass transition temperatures Tg.

FIG. 2.

DSC thermograms revealing the glass transition steps of LiCl-7D2O in the studied pristine and modified silica confinements. The measurements were performed with a heating rate of 10 K/min. Sloped baselines were subtracted from the measured signals. The curves are shifted vertically for better visibility. The dotted lines indicate the onset glass transition temperatures Tg.

Close modal
TABLE III.

Onset glass transition temperatures Tg and melting temperatures Tm obtained for the confined LiCl-7D2O solutions from DSC heating runs at a heating rate of 10 K/min.

SampleTg (K)Tm (K)
2.8 nm 141.6/147.7 ⋯ 
3.7 nm 145.1 ⋯ 
5.4 nm 139.2 204.1 
6.8 nm-A 139.3 203.7 
5.8 nm-A+D 138.9 203.3 
SampleTg (K)Tm (K)
2.8 nm 141.6/147.7 ⋯ 
3.7 nm 145.1 ⋯ 
5.4 nm 139.2 204.1 
6.8 nm-A 139.3 203.7 
5.8 nm-A+D 138.9 203.3 

In order to determine the local dynamics of the LiCl-7D2O solution in the pristine and modified silica pores, we analyze the 2H and 7Li SLR. Figure 3 shows the normalized buildup of the 2H and 7Li magnetizations for various pore diameters and characteristic temperatures. For 2H, we observe a single SLR step at high and low temperatures, but a bimodal buildup at intermediate temperatures in the range 150–200 K. The bimodality indicates that two water species with different rotational dynamics can be distinguished on the milliseconds time scale of the magnetization buildup. Therefore, we fit the experimental data at high and low temperatures with a single SLR step and those at intermediate temperatures with a superposition of two SLR steps. The latter fits yield the time constant T1,f/s and stretching parameter β1,f/s for the fast/slow SLR component. For 7Li, a single SLR step is found for all samples and temperatures. However, the single exponential behavior at high temperatures evolves into a pore-diameter dependent stretched exponential behavior upon cooling. In order to analyze the dependence of the local dynamics on the pore diameter and functionalization in more detail, we compare the fit parameters obtained from these 2H and 7Li SLR analyses. The SLR times T1,n are shown in Fig. 4 and the stretching parameters β1,n are displayed in Fig. 5.

FIG. 3.

Normalized magnetization buildup curves, M(t)/M0, for LiCl-7D2O in the studied silica confinements: (a) 2H and (b) 7Li saturation recovery. The lines are fits with Eq. (2).

FIG. 3.

Normalized magnetization buildup curves, M(t)/M0, for LiCl-7D2O in the studied silica confinements: (a) 2H and (b) 7Li saturation recovery. The lines are fits with Eq. (2).

Close modal
FIG. 4.

Temperature dependence of the SLR times for LiCl-7D2O in the bulk (solid lines)32 and in silica pores with the indicated diameters and functionalizations: (a) For 2H, fast (T1,f, solid symbols) and slow (T1,s, open symbols) SLR steps can be distinguished at intermediate temperatures. The upper left inset shows the mean correlation times ⟨τ⟩ determined from the SLR data according to Eq. (5). (b) For 7Li, a single SLR step is found for all samples and temperatures. In both panels, insets focus on the minimum regions for exemplary samples. The numbers specify the temperatures of the minima, as obtained from parabolic fits (dashed lines). The results for the pristine pores were obtained in previous work.31 

FIG. 4.

Temperature dependence of the SLR times for LiCl-7D2O in the bulk (solid lines)32 and in silica pores with the indicated diameters and functionalizations: (a) For 2H, fast (T1,f, solid symbols) and slow (T1,s, open symbols) SLR steps can be distinguished at intermediate temperatures. The upper left inset shows the mean correlation times ⟨τ⟩ determined from the SLR data according to Eq. (5). (b) For 7Li, a single SLR step is found for all samples and temperatures. In both panels, insets focus on the minimum regions for exemplary samples. The numbers specify the temperatures of the minima, as obtained from parabolic fits (dashed lines). The results for the pristine pores were obtained in previous work.31 

Close modal
FIG. 5.

Temperature-dependent stretching parameters β1 characterizing the (a) 2H and (b) 7Li SLR steps of LiCl-7D2O solution in silica pores with the indicated diameters and functionalizations. For the bimodal 2H SLR at intermediate temperatures, the stretching parameters of the fast (β1,f) and slow (β1,s) steps are shown as solid and open symbols, respectively. The results for the pristine pores were obtained in previous work.31 The solid lines are the results for the bulk LiCl-7D2O solution.32 

FIG. 5.

Temperature-dependent stretching parameters β1 characterizing the (a) 2H and (b) 7Li SLR steps of LiCl-7D2O solution in silica pores with the indicated diameters and functionalizations. For the bimodal 2H SLR at intermediate temperatures, the stretching parameters of the fast (β1,f) and slow (β1,s) steps are shown as solid and open symbols, respectively. The results for the pristine pores were obtained in previous work.31 The solid lines are the results for the bulk LiCl-7D2O solution.32 

Close modal

For 2H, we see in Fig. 4(a) that, at high temperatures, the SLR times of both modified pores are similar and shorter than those of the pristine pores with a comparable diameter (5.4 nm). To investigate the corresponding dependence of water dynamics on the pore characteristics, we exploit that Eq. (5) allows us to determine mean correlation times ⟨τ⟩ from the SLR results near ambient temperatures. In the upper left inset, we observe that the results for the modified pores resemble those for the pristine 2.8 nm pores, while ⟨τ⟩ is shorter in the pristine 3.7 and 5.4 nm pores and even shorter in the bulk solution. Thus, in addition to the pore diameter, the chemistry of the pore walls notably affects the water reorientation in the confined LiCl-7D2O solution. Overall, the mean correlation times differ by a factor of ∼5 between the functionalized and bulk samples. At intermediate temperatures, ∼150–200 K, 2H SLR is bimodal. T1,f continues the high-temperature behavior and exhibits a minimum, whereas T1,s is about two orders of magnitude slower and increases monotonously upon cooling. Near Tg, T1,f rapidly approaches T1,s so that both components can no longer be distinguished at even lower temperatures. The lower right inset focuses on the minimum region, where the correlation times are on the order of 1 ns. It reveals that the T1,f minimum is located at somewhat higher temperatures and values in the functionalized and narrowest (2.8 nm) pores, indicating that the water fraction causing the fast 2H SLR step exhibits slightly slower and more heterogeneous reorientation in these confinements.

Figure 5(a) confirms that the 2H SLR is single exponential (β1 = 1) down to ∼200 K. In the intermediate temperature range, we observe that the fast SLR step is exponential, i.e., β1,f ≈ 1, down to ∼180 K, where the stretching parameter starts to decrease, whereas the slow SLR step is nonexponential (β1,s ≈ 0.6). Below ∼150 K, there is a single nonexponential SLR step characterized by β1 ∼ 0.7–0.8. These findings imply that two water fractions with distinguishable reorientation dynamics exist and cause bimodal SLR at intermediate temperatures, while exchange processes between these fractions occur on the longer time scale of the magnetization buildup and average the different SLR behaviors at high and low temperatures. We expect that water diffusion across the pore volume is sufficiently fast to restore the ergodicity above ∼200 K, while spin diffusion leads to an exchange between different magnetization components below ∼150 K. Thus, the mean correlation times ⟨τ⟩ obtained from the SLR data near ambient temperatures reflect an average over the dynamical behaviors in different pore regions. In the intermediate temperature range, the exponentiality of the fast SLR step indicates a liquid-like behavior of the associated water fraction and enables quantitative SLR analysis (see below), while the nonexponentiality of the slow SLR step implies that the contributing water fraction does not explore a significant number of different local environments on the time scale of the magnetization buildup.

For 7Li in Fig. 4(b), T1 decreases with the decreasing pore diameter for the pristine pores at high temperatures, implying a slowdown of the local dynamics, as was discussed in more detail in a previous study.31 By contrast, the functionalized samples do not show a consistent picture in this temperature range. In particular, we currently do not have a clear explanation for the relatively short T1 times of the 5.8 nm-A+D sample at ambient temperatures. Below 250 K, two sets of samples can be distinguished. The T1 minimum is located at higher temperatures and values in the modified and narrow (2.8 nm) pristine pores than in the wider pristine pores (3.7 and 5.4 nm), indicating that the environments of the lithium ions change more slowly and more heterogeneously in the former confinements. Specifically, the correlation times amount to 0.6/ω0 = 1.5 ns at the indicated temperatures of the respective T1 minima. In addition, the former set of samples with modified/narrow pores shows longer SLR times than the latter set with wider pristine pores at low temperatures. In Fig. 5(b), we see that the 7Li SLR is essentially exponential for the modified pores and the 2.8 nm pores in the whole temperature range. For the 3.7 and 5.4 nm samples, β1 ≈ 1 holds down to ∼170 K, while the stretching parameter decreases upon further cooling.

A comparison of the 2H and 7Li SLR results for confined and bulk32 LiCl-7D2O solutions enables further insights. For 2H, we find that T1,f and β1,f are similar to the bulk data, while T1,s and β1,s are very different. Therefore, we attribute the fast SLR step to water molecules in the pore centers, which show a bulk-like structural (α) relaxation and the slow SLR step to water molecules near the pore walls featuring retarded dynamics. For both water fractions, the results weakly depend on the type of the confinement. As aforementioned, the chemical exchange between the water fractions is slow on the time scale of the magnetization buildup below ∼200 K so that it is possible to distinguish two 2H SLR steps, while the T1 values reflect an average of the diverse water dynamics in all pore regions at higher temperatures. Therefore, we propose that the difference between the high-temperature T1 values of the confined and bulk solutions largely results from the contributions of slowed dynamics near the pore walls to the observed average. The spin diffusion is faster for 7Li than for 2H because of the different gyromagnetic ratios and, hence, dipolar couplings. Therefore, we expect that bimodal SLR is not observed for 7Li because, in addition to ion diffusion, spin diffusion leads to an effective exchange of magnetization. However, the small 7Li stretching parameters β1 ∼ 0.75 in the pristine pores with diameters of 3.7 and 5.4 nm below ∼160 K, which are found neither in the other pores nor in the bulk, indicate that the efficiency of these exchange mechanisms does not suffice to fully restore the ergodicity, suggesting a particularly significant inhomogeneity of the lithium ion dynamics in these confinements.

Altogether, the 2H and 7Li SLR of LiCl-7D2O are very similar in both functionalized pores. If the confinement size was the decisive factor, the results for the modified pores should best match those for the widest pristine pores with a diameter of 5.4 nm. However, this does not comply with our observations. Explicitly, the 2H and 7Li data for the functionalized samples largely agree with those for the narrowest pristine pores with a diameter of 2.8 nm, while they differ from those for the wider pristine confinements with d = 3.7 nm and d = 5.4 nm. Thus, the SLR results suggest that the surface chemistry significantly affects the solution dynamics, explicitly, that the steric hindrance and the positive charges associated with the functional groups result in slower confinement-averaged electrolyte motion.

Next, we perform 7Li LSA. In Fig. 6(a), we display the 7Li NMR spectra of LiCl-7D2O in the 5.8 nm-A+D pores. The broad low-temperature spectra, which comprise a characteristic superposition of Gaussian CT and ST lines, evolve into narrow Lorentzian high-temperature spectra when the temperature is increased from below 140 K to above 200 K. This motional narrowing indicates that the molecular dynamics in the solution crosses the microseconds time scale of 7Li LSA. The results for the other studied confinements are qualitatively similar. For further analysis, we fit the spectra with a weighted superposition of a broader Gaussian and a narrower pseudo-Voigt function, which allows us to follow the evolution from the Gaussian to the Lorentzian shape. In Fig. 6(b), we show the temperature-dependent width of the CT line for the confined and bulk32 LiCl-7D2O solutions. The motional narrowing occurs at similar temperatures for both functionalized pores and for the 2.8 nm pristine pores, at somewhat lower temperatures for the 3.7 and 5.4 nm native pores, and at even lower temperatures for the bulk solution. This shift indicates that the renewal of the lithium ionic environments is slower in the former than in the latter set of confinements, in agreement with the above SLR results. The correlation times obtained from this 7Li LSA will be discussed together with the results of the other methods below.

FIG. 6.

(a) 7Li NMR spectra of LiCl-7D2O in the 5.8 nm-A+D matrix at the indicated temperatures. (b) Temperature-dependent full width at half maximum of the CT lines in the 7Li NMR spectra of LiCl-7D2O in the bulk32 and in various confinements. The results for the pristine pores were obtained in previous work.31 The lines are guides for the eye.

FIG. 6.

(a) 7Li NMR spectra of LiCl-7D2O in the 5.8 nm-A+D matrix at the indicated temperatures. (b) Temperature-dependent full width at half maximum of the CT lines in the 7Li NMR spectra of LiCl-7D2O in the bulk32 and in various confinements. The results for the pristine pores were obtained in previous work.31 The lines are guides for the eye.

Close modal

Information about slow electrolyte dynamics at low temperatures is available from STE approaches. Figure 7 shows exemplary 2H STE data. In panel (a), we compare F2cc(tm) of confined and bulk LiCl-7D2O solutions at ∼150 K. We see that the correlation functions decay more slowly and more stretched in the silica confinements than in the bulk solution. Again, the slowdown depends on the pore size and functionalization. The increased stretching indicates a stronger heterogeneity of the water reorientation in the confinements than in the bulk, consistent with the above 2H SLR results. In panel (b), the temperature dependence of F2cc(tm) is shown for the 5.8 nm-A+D sample. Clearly, the correlation functions shift to shorter times upon heating. In addition, we observe that a bimodality evolves from the strong stretching of the decays when the temperature is increased and the SLR damping of the STE decays becomes less effective. Such bimodality is expected based on the two-step 2H magnetization buildup in the relevant temperature range, which we attributed to coexisting faster and slower water fractions near the pore centers and the pore walls, respectively. Closer analysis reveals that the fast STE decay occurs prior to the onset of SLR, while the slow STE decay agrees with the slow SLR step. Thus, the reorientation dynamics of the fast water fraction is accessible, whereas that of the slow water fraction occurs outside the time window of the experiment. To consider this situation, we fit the 2H STE decays with Eq. (11). Averaging the fit results at the studied temperatures, we find that the slow/immobile water fraction is larger in the functionalized pores and the 2.8 nm pores (c = 0.2–0.3) than in the 3.7 and 5.4 nm pores (c = 0.1–0.2). For the pristine pores, this finding is consistent with the expectation that the fraction of slow molecules near the pore walls is higher for smaller pore diameters.

FIG. 7.

Correlation functions from 2H STE studies on bulk32 and confined LiCl-7D2O solutions: (a) F2cc(tm) for the modified and pristine silica frameworks and the bulk solution (stars) at ∼150 K together with the corresponding SLR damping functions (dashed lines). (b) F2cc(tm) for the 5.8 nm-A+D sample at various temperatures. The evolution time was set to te = 5 µs in all measurements. In both panels, the solid lines are fits with Eq. (11). All STE decays are normalized to F2cc(0)=1 based on the fit results.

FIG. 7.

Correlation functions from 2H STE studies on bulk32 and confined LiCl-7D2O solutions: (a) F2cc(tm) for the modified and pristine silica frameworks and the bulk solution (stars) at ∼150 K together with the corresponding SLR damping functions (dashed lines). (b) F2cc(tm) for the 5.8 nm-A+D sample at various temperatures. The evolution time was set to te = 5 µs in all measurements. In both panels, the solid lines are fits with Eq. (11). All STE decays are normalized to F2cc(0)=1 based on the fit results.

Close modal

In Fig. 8, we compile F2ss(tm) correlation functions of confined and bulk LiCl-7D2O from 7Li STE experiments. For the pristine silica pores, the F2ss(tm) data were analyzed in some detail in previous work.31 The analysis revealed monomodal 7Li STE decays for d = 2.8 nm and clearly bimodal ones in the 3.7 and 5.4 nm pores. To rationalize these findings, it was conjectured that a Stern layer forms at the silica walls of the wider pores and gives rise to a slow but not immobile fraction of lithium ions, while this effect is suppressed by severe confinement. Here, a single 7Li STE decay is observed in both functionalized silica frameworks. It is hardly affected by SLR damping so that a determination of correlation times is straightforward. In panel (b), this finding is confirmed for the 5.8 nm-A+D sample in a broader temperature range. We conclude that, like severe confinement, the present functionalizations suppress the formation of a Stern layer, most probably, as a consequence of the steric hindrance and positive charges associated with the grafted functional groups.

FIG. 8.

Correlation functions from 7Li STE studies on bulk32 and confined LiCl-7D2O solutions: (a) F2ss(tm) for the modified and pristine silica frameworks and the bulk solution (stars) at ∼150 K together with the corresponding SLR damping functions (dashed lines). (b) F2ss(tm) for the 5.8 nm-A+D sample at various temperatures. The evolution time was set to te = 20 µs in all measurements. In both panels, the solid lines are fits with Eq. (8). A single decay with F ≈ 0.0 is sufficient to interpolate the data for the functionalized and 2.8 nm samples, whereas a weighted superposition of two decays is required to fit the results for the 3.7 and 5.4 nm samples.31 All STE decays are normalized to F2ss(0)=1 based on the fit results.

FIG. 8.

Correlation functions from 7Li STE studies on bulk32 and confined LiCl-7D2O solutions: (a) F2ss(tm) for the modified and pristine silica frameworks and the bulk solution (stars) at ∼150 K together with the corresponding SLR damping functions (dashed lines). (b) F2ss(tm) for the 5.8 nm-A+D sample at various temperatures. The evolution time was set to te = 20 µs in all measurements. In both panels, the solid lines are fits with Eq. (8). A single decay with F ≈ 0.0 is sufficient to interpolate the data for the functionalized and 2.8 nm samples, whereas a weighted superposition of two decays is required to fit the results for the 3.7 and 5.4 nm samples.31 All STE decays are normalized to F2ss(0)=1 based on the fit results.

Close modal

We move on from local dynamics to diffusive dynamics. Specifically, we measure self-diffusion coefficients D in 1H and 7Li SFG NMR. Figure 9 shows exemplary results. In panel (a), we depict 1H HE decays S(te) for LiCl-7H2O in the 6.8 nm-A pores. The decay shifts to longer times when the temperature is decreased, indicating a slowdown of water diffusion, strictly speaking, of proton diffusion. In panel (b), 7Li STE decays S(tm) for LiCl-7D2O in the 5.8 nm-A+D pores are presented. The signal decreases at shorter mixing times tm when the evolution time te is extended, in agreement with the expectation for diffusive motion. Considering the fact that the displacements are restricted to cylindrical pores, we follow previous SFG studies31,46,50,51 of diffusion in the pristine silica materials and fit the 1H and 7Li SFG data with the 1D diffusion model [see Eq. (13)].

FIG. 9.

(a) 1H SFG HE decays S(te) for LiCl-7H2O in the 6.8 nm-A framework at the indicated temperatures. To correct for any damping, which is not caused by proton diffusion, the data are divided by the HE decays Shom(te) obtained in a homogeneous magnetic field of the same strength. The solid lines are fits with the 1D diffusion model. (b) 7Li SFG STE decays S(tm) of LiCl-7D2O in the 5.8 nm-A+D framework for the indicated evolution times te at 240 K. The solid lines are global fits with the 1D diffusion model. The dashed line is the SLR function at this temperature.

FIG. 9.

(a) 1H SFG HE decays S(te) for LiCl-7H2O in the 6.8 nm-A framework at the indicated temperatures. To correct for any damping, which is not caused by proton diffusion, the data are divided by the HE decays Shom(te) obtained in a homogeneous magnetic field of the same strength. The solid lines are fits with the 1D diffusion model. (b) 7Li SFG STE decays S(tm) of LiCl-7D2O in the 5.8 nm-A+D framework for the indicated evolution times te at 240 K. The solid lines are global fits with the 1D diffusion model. The dashed line is the SLR function at this temperature.

Close modal

The resulting self-diffusion coefficients D are plotted in Fig. 10. The 1H diffusivities in the modified 6.8 nm-A pores resemble those in the pristine 2.8 nm pores. The diffusion coefficients in both confinements weakly differ from those in the bulk. By contrast, a slowdown by about one order of magnitude was reported for the diffusion of pure water in the identical 2.8 nm pores.50 At the highest studied temperatures, the diffusivities appear to be even higher in the pores than in the bulk. However, this finding should be treated with great caution since, due to the enhanced mobility, an appreciable fraction of water molecules reaches the pore ends during the diffusion time at higher temperatures, and hence, they may exit the pores and undergo fast transport in the interparticle volume, as was discussed in some detail in previous work.50 Because the limited amount of the dye-functionalized material inhibits 1H SFG measurements for this confinement, we perform additional 2H SFG studies on the confined LiCl-7D2O solutions. We find that the 2H diffusivities are similar in both functionalized pores at 280 K but lower than the 1H diffusivities in the studied confined and bulk LiCl-7H2O solutions. Considering that at least a part of the difference between the 1H and 2H diffusivities results from the isotope effect, we conclude that the 5.8 nm-A+D confinement does not cause a strong slowdown of water diffusion either. Like the 1H diffusivity, the 7Li diffusivity is very similar in the 6.8 nm-A and 2.8 nm pores and only a factor of ∼2 lower than in the bulk. By contrast, the 7Li diffusivity is reduced by about one order of magnitude in the 5.8 nm-A+D pores. The origin of this strong slowdown of the lithium ion diffusion in the dye-functionalized material, which is not accompanied by slower local dynamics, is not completely clear. We speculate that the bulky dye molecules protrude into the interior of the pores and sterically hinder the diffusion of the highly hydrated lithium ions, i.e., they act as bottlenecks for the long-range ionic motion.

FIG. 10.

(a) 1H diffusion coefficients (striped symbols) for LiCl-7H2O in the bulk (asterisk),32 in the pristine 2.8 nm pores,31 and in the modified 6.8 nm-A pores together with 2H diffusion coefficients (solid symbols) for LiCl-7D2O in the 6.8 nm-A and 5.8 nm-A+D frameworks. (Plus symbol) denotes 17O diffusion coefficients in the bulk.67 (b) 7Li diffusion coefficients for LiCl-7D2O in various confinements and in the bulk. The results for the bulk solutions (stars) and the pristine confinements were obtained in previous studies.31,32 The solid lines are fits of the bulk data.32 

FIG. 10.

(a) 1H diffusion coefficients (striped symbols) for LiCl-7H2O in the bulk (asterisk),32 in the pristine 2.8 nm pores,31 and in the modified 6.8 nm-A pores together with 2H diffusion coefficients (solid symbols) for LiCl-7D2O in the 6.8 nm-A and 5.8 nm-A+D frameworks. (Plus symbol) denotes 17O diffusion coefficients in the bulk.67 (b) 7Li diffusion coefficients for LiCl-7D2O in various confinements and in the bulk. The results for the bulk solutions (stars) and the pristine confinements were obtained in previous studies.31,32 The solid lines are fits of the bulk data.32 

Close modal
To compare our findings for the local and diffusive dynamics of the confined LiCl solutions, we employ the Stokes–Einstein–Debye (SED) relation
DτD=29RH2.
(14)
Here, RH is the hydrodynamic radius and the prefactor 29 results from the facts that NMR probes rank l = 2 correlation functions and that the SED treatment assumes isotropic rotational diffusion. We employ the SED relation to calculate “diffusion correlation times” τD from the self-diffusion coefficients D for comparison with the correlation times τp obtained from our SLR, LSA, and STE approaches to the local dynamics. In this analysis, we use the hydrodynamic radii determined in previous studies,31,32,50 explicitly, RH = 1.05 Å for the lithium ions and RH = 1.35 Å for the water molecules. These values are somewhat smaller than expected based on the particle sizes, suggesting that a complex transport mechanism leads to some retardation of the diffusive motion relative to the local dynamics.45,51 The results for τD will be discussed in Sec. IV G.

Figure 11 shows the peak correlation times τp of water reorientation in the confined and bulk32 LiCl-7D2O solutions from the 2H SLR and STE studies. The confinement effects are moderate in the whole temperature range. Specifically, the correlation times from the SLR and STE approaches to the confined solutions are similar to the α relaxation times of the bulk solution, which follow a Vogel–Fulcher–Tammann (VFT) temperature dependence.68 Moreover, the rotational correlation times τp resemble the diffusion correlation times τD of water. In particular, both time constants exhibit a similar temperature dependence, implying that the rotational motion and the translational diffusion of water are coupled for the LiCl-7D2O solutions in all studied confinements, as anticipated in the SED treatment. By contrast, evidence for a rotation-translation decoupling was reported for pure water in silica pores.50 

FIG. 11.

Correlation times of water dynamics in bulk32 (stars) and confined LiCl-7H2O or LiCl-7D2O solutions from 1H and 2H NMR. Specifically, the peak correlation times τp from the SLR (checkered symbols) and STE (solid symbols) studies on LiCl-7D2O solutions are shown together with the diffusion correlation times τD (striped symbols) from the SFG measurements on LiCl-7H2O, which were calculated from the 1H self-diffusion coefficients D using the SED relation [see Eq. (14)], and a hydrodynamic radius of RH = 1.35 Å, as determined in a previous study on neat water in silica pores.50 The crossed circle and horizontal bar mark the correlation time τp = 100 s obtained from the glass transition temperatures of the bulk32 and confined LiCl-7D2O solutions, respectively. The solid line shows correlation times of water reorientation in bulk LiCl-7.3H2O from dielectric spectroscopy.68 The inset is an enlarged view of the SLR results at temperatures in the vicinity of the T1,f minimum together with the bulk data as solid line.

FIG. 11.

Correlation times of water dynamics in bulk32 (stars) and confined LiCl-7H2O or LiCl-7D2O solutions from 1H and 2H NMR. Specifically, the peak correlation times τp from the SLR (checkered symbols) and STE (solid symbols) studies on LiCl-7D2O solutions are shown together with the diffusion correlation times τD (striped symbols) from the SFG measurements on LiCl-7H2O, which were calculated from the 1H self-diffusion coefficients D using the SED relation [see Eq. (14)], and a hydrodynamic radius of RH = 1.35 Å, as determined in a previous study on neat water in silica pores.50 The crossed circle and horizontal bar mark the correlation time τp = 100 s obtained from the glass transition temperatures of the bulk32 and confined LiCl-7D2O solutions, respectively. The solid line shows correlation times of water reorientation in bulk LiCl-7.3H2O from dielectric spectroscopy.68 The inset is an enlarged view of the SLR results at temperatures in the vicinity of the T1,f minimum together with the bulk data as solid line.

Close modal

In Fig. 11, a closer inspection of the peak correlation times reveals a very minor dependence of the water reorientation on the functionalization and diameter of the silica pores. To present the effects in more detail, the inset focuses on the τp results from the SLR analysis. We see that the water reorientation in both functionalized pores is similar and slightly longer than in the bulk. Above, we found somewhat larger differences between the mean correlation times ⟨τ⟩ of the confined and bulk LiCl-7D2O solutions [see Fig. 4]. This apparent discrepancy can be explained when we consider that our results imply bulk-like and slowed-down water dynamics in the pore center and at the pore wall, respectively, and that the different SLR analyses are differently affected by this distribution. Specifically, fast exchange interferes with a discrimination of both fractions in the SLR analysis at higher temperatures and, hence, ⟨τ⟩ reflects an average over the respective dynamics. By contrast, the water fractions are distinguishable at lower temperatures and the τp values result from the SLR analysis of the faster, bulk-like one in the pore center, but receive little contributions from the slower one. At low temperatures, the 2H STE correlation times show a weaker temperature dependence than expected based on the glass transition temperatures Tg obtained from our DSC experiments. As was discussed in our previous studies on LiCl solutions,31,32 this finding suggests that 2H NMR starts to probe the common water-related secondary (ν) relaxation of aqueous systems when the α relaxation becomes very slow near Tg, in harmony with findings for other water mixtures.46,69–71

In Fig. 12, we display the peak correlation times τp of lithium ion dynamics in the bulk and confined LiCl-7D2O solutions. The 7Li SLR and LSA results confirm that the renewal of the local lithium ionic environments is slightly faster in the 3.7 and 5.4 nm samples than in the 2.8 nm, 5.8 nm-A+D, and 6.8 nm-A samples. The diffusion correlation times τD are consistent with the peak correlation times τp, except for the dye-functionalized pores. The diffusion data extend the correlation times to higher temperatures and follow the VFT temperature dependence of the α relaxation in the bulk solution. Consistently, our previous study on the pristine pores concluded that local and diffusive dynamics are also coupled for the lithium ions.31 For the 5.8 nm-A+D pores, τD is about an order of magnitude longer than expected from the time scale of the local dynamics. As aforementioned, this discrepancy may indicate that the grafted dye molecules protrude into the pore interior and form obstacles for the long-range transport of the lithium ions. At low temperatures, the 7Li STE results reflect the above discussed bimodal lithium ion dynamics in the wider pristine pores, which was recently rationalized by nearly bulk-like and strongly slowed-down lithium ion fractions in the pore center and in a Stern layer, respectively.31 In the 5.8 nm-A+D and 6.8 nm-A pores and the pristine 2.8 nm pores, a clear bimodality is not observed. Rather, the monomodal STE decay exhibits correlation times, which are significantly longer than those of the bulk solution, in harmony with the relatively slow dynamics in the functionalized and narrow pores observed in our other 2H and 7Li NMR studies and the glass transition temperatures Tg from the DSC experiments.

FIG. 12.

Correlation times of lithium ion dynamics in bulk32 (stars) and confined LiCl-7D2O from 7Li NMR. Specifically, the peak correlation times τp from SLR (open symbols), LSA (dotted symbols), and STE (solid symbols) studies are shown together with the diffusion correlation times τD (striped symbols) from SFG measurements. The latter were calculated from the 7Li self-diffusion coefficients D using the SED relation [see Eq. (14)] and a hydrodynamic radius of RH = 1.05 Å, as determined in a previous study on the bulk solution.32 The crossed circle and the horizontal bar mark the correlation time τp = 100 s obtained from the glass transition temperatures of the bulk32 and confined LiCl-7D2O solutions, respectively. The solid line shows correlation times of bulk LiCl-7.3H2O from dielectric spectroscopy.68 

FIG. 12.

Correlation times of lithium ion dynamics in bulk32 (stars) and confined LiCl-7D2O from 7Li NMR. Specifically, the peak correlation times τp from SLR (open symbols), LSA (dotted symbols), and STE (solid symbols) studies are shown together with the diffusion correlation times τD (striped symbols) from SFG measurements. The latter were calculated from the 7Li self-diffusion coefficients D using the SED relation [see Eq. (14)] and a hydrodynamic radius of RH = 1.05 Å, as determined in a previous study on the bulk solution.32 The crossed circle and the horizontal bar mark the correlation time τp = 100 s obtained from the glass transition temperatures of the bulk32 and confined LiCl-7D2O solutions, respectively. The solid line shows correlation times of bulk LiCl-7.3H2O from dielectric spectroscopy.68 

Close modal

We combined various 1H, 2H, and 7Li NMR methods to investigate local and diffusive motions of aqueous LiCl solutions in pristine and functionalized silica pores. For the functionalization, APTES groups were linked to the silica walls in one sample and this material was further modified by a thiazole-based dye in another sample. In the whole range from ambient temperatures down to the glass transition, we observed moderate but systematic confinement effects. Our previous study on pristine mesoporous silica revealed that the solution dynamics are slower in narrower pores.31 Here, we found that, in addition to the pore size, the properties of the pore walls affect the solution dynamics. For the most part, similar dynamics were observed in both modified pores. However, the mobility of the water molecules and lithium ions was smaller in the functionalized pores than in the pristine pores with a comparable diameter of 5.4 nm.

Near ambient temperatures, our 1H and 2H NMR studies informed about the average dynamics of the water molecules in the LiCl-7H2O and LiCl-7D2O solutions, respectively. Specifically, the SLR analysis yielded mean correlation times and the SFG approach delivered self-diffusion coefficients. The results revealed that the reorientation and diffusion of water are, on average, factors of ∼2–5 slower in the functionalized pores than in the bulk solution. Furthermore, a comparison of the SLR and SFG data implied that the close relation of the local and diffusive motions predicted by the SED relation is retained in the modified confinements. This finding is in harmony with our previous results for the LiCl solutions in the pristine pores31 but in contrast to an SED breakdown reported for pure water therein.50 

At reduced temperatures, 2H SLR and STE experiments indicated a bimodality of water dynamics in both the pristine and the modified pores. For the faster water fraction, we observed that the reorientation is marginally slowed down and more heterogeneous in all studied confinements but overall still bulk-like. Therefore, we argued that this weakly affected water species resides in the pore center. Considering that the pH measurements, including the analysis of the fluorescence signals of the wall-grafted dye molecules, indicated more acidic conditions in the pristine pores than in the functionalized ones and, in particular, in the bulk, the overall weak confinement effect implies that the water reorientation hardly depends on the pH value in the covered range. It should be explored in future work to which degree this finding can be generalized to other confined aqueous systems, in particular, to the important case of confined water. For the slower water fraction, a quantitative analysis of the dynamics was not possible. Following previous studies,31,72 we conjectured that this water species is located at the pore walls, but detailed structural information is not available. It is tempting to speculate that it is related to an EDL at the silica surface. However, this hypothesis is not supported by the finding that the bimodal behavior is not affected by the functionalization.

The 7Li SLR and LSA studies did not provide evidence for a bimodality of the lithium ion dynamics at sufficiently high temperatures. This indicates that there is a permanent exchange of lithium ions between all confinement regions and, hence, a Stern layer, if existent, is not truly immobile. Moreover, the 7Li SLR and LSA data showed that the renewal of the lithium ionic environments is somewhat slower in the functionalized pores than in the bulk solutions. The corresponding temperature shift amounts to less than 10 K. Consistent with our findings for the pore-averaged water dynamics at high temperatures, the local lithium ion motion in pristine pores with narrower (2.8 nm) rather than comparable (5.4 nm) diameters matches that in the functionalized pores, implying a significant role of the surface chemistry. 7Li SFG diffusometry revealed a particularly strong effect of the functionalization. The diffusion coefficients of the lithium ions are approximately an order of magnitude lower in the dye-functionalized pores, which may suggest that these bulkier functional groups protrude into the interior of the pores and from obstacles for the long-range transport of the highly hydrated ions. All these observations indicate that various pore parameters synergistically determine the local and diffusive dynamics of the LiCl solutions.

A recent 7Li STE approach argued that bimodal lithium ion dynamics near the glass transition is caused by the formation of a Stern layer with reduced mobility.31 Explicitly, a bimodality was observed in the 3.7 and 5.4 nm pristine pores but not in the narrower 2.8 nm ones. Here, we found monomodal rather than bimodal 7Li STE decays in the modified pores. Therefore, we conjecture that, in addition to a severe geometrical restriction, a functionalization of the silica surfaces can suppress Stern-layer formation. Specifically, we expect that the grafted APTES groups impede the formation of a defined surface layer of lithium ions for mainly two reasons. First, they constitute steric obstacles, which increase the surface roughness, and second, their amino moieties are ionized at the relevant pH values, and hence, the largely homogeneous negative surface charge of the pristine silica is replaced by an uneven charge distribution comprising SiO and NH3+ groups on the functionalized walls.

Altogether, our results indicate that confinements have multiple effects on the dynamics of aqueous salt solutions. The interactions of the salt solutions with the pore walls cause concentration and mobility gradients. Therefore, pore-averaged dynamical behaviors depend on the degree and range of these inhomogeneities. The diameter of the pores determines the relative relevance of bulk-like dynamics in the inner confinement regions and wall-affected dynamics in the outer ones. A functionalization of the walls alters the structure and dynamics of the confined solution. Here, we showed that adding a steric hindrance and, thus, increasing the surface roughness and a variation and diversification of the surface charge have notable effects. However, future works need to further improve our understanding of the highly complex and heterogeneous properties of salt solutions in functionalized nanopores before these effects can be used to control and tailor the transport properties.

See the supplementary material for further pH measurements, including details on the in situ characterization in fluorescence measurements using the grafted dye molecules, and for further DSC measurements.

Financial support from the LOEWE project iNAPO funded by the Ministry of Higher Education, Research and the Arts (HMWK) of the Hessen state is gratefully acknowledged. H. S. Varol acknowledges funding by a Career Bridging Grant of the Technische Universität Darmstadt. The authors thank Professor Herbert Plenio from the Chemistry Department of the Technische Universität Darmstadt for giving us the opportunity to use his fluorescence spectrometer.

The authors have no conflicts to disclose.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

1.
A.
Siria
,
P.
Poncharal
,
A.-L.
Biance
,
R.
Fulcrand
,
X.
Blase
,
S. T.
Purcell
, and
L.
Bocquet
,
Nature
494
,
455
(
2013
).
2.
M.
Thomas
,
B.
Corry
, and
T. A.
Hilder
,
Small
10
,
1453
(
2014
).
3.
A.
Esfandiar
,
B.
Radha
,
F. C.
Wang
,
Q.
Yang
,
S.
Hu
,
S.
Garaj
,
R. R.
Nair
,
A. K.
Geim
, and
K.
Gopinadhan
,
Science
358
,
511
(
2017
).
4.
Y.
Gao
,
M.
Li
,
Y.
Zhang
,
W.
Lu
, and
B.
Xu
,
Proc. Natl. Acad. Sci. U. S. A.
117
,
25246
(
2020
).
5.
M.
Wang
,
W.
Shen
,
S.
Ding
,
X.
Wang
,
Z.
Wang
,
Y.
Wang
, and
F.
Liu
,
Nanoscale
10
,
18821
(
2018
).
6.
G.
Hummer
,
J. C.
Rasaiah
, and
J. P.
Noworyta
,
Nature
414
,
188
(
2001
).
7.
Y.
Li
,
Z.
Li
,
F.
Aydin
,
J.
Quan
,
X.
Chen
,
Y.-C.
Yao
,
C.
Zhan
,
Y.
Chen
,
T. A.
Pham
, and
A.
Noy
,
Sci. Adv.
6
,
eaba9966
(
2020
).
8.
J. P.
O’Reilly
,
C. P.
Butts
,
I. A.
I’Anson
, and
A. M.
Shaw
,
J. Am. Chem. Soc.
127
,
1632
(
2005
).
10.
D. L.
Chapman
,
London, Edinburgh, Dublin Philos. Mag. J. Sci.
25
,
475
(
1913
).
11.
O.
Stern
,
Z. Electrochem.
30
,
508
(
1924
).
12.
Y.
Qiu
,
J.
Ma
, and
Y.
Chen
,
Langmuir
32
,
4806
(
2016
).
13.
T. A.
Gmür
,
A.
Goel
, and
M. A.
Brown
,
J. Phys. Chem. C
120
,
16617
(
2016
).
14.
A. M.
Darlington
,
T. A.
Jarisz
,
E. L.
DeWalt-Kerian
,
S.
Roy
,
S.
Kim
,
M. S.
Azam
,
D. K.
Hore
, and
J. M.
Gibbs
,
J. Phys. Chem. C
121
,
20229
(
2017
).
15.
M. M.
Waegele
,
C. M.
Gunathunge
,
J.
Li
, and
X.
Li
,
J. Chem. Phys.
151
,
160902
(
2019
).
16.
E.
Mamontov
,
D. R.
Cole
,
S.
Dai
,
M. D.
Pawel
,
C. D.
Liang
,
T.
Jenkins
,
G.
Gasparovic
, and
E.
Kintzel
,
Chem. Phys.
352
,
117
(
2008
).
17.
S.
Beckert
,
M.
Gratz
,
J.
Kullmann
,
D.
Enke
, and
F.
Stallmach
,
Appl. Magn. Reson.
44
,
827
(
2013
).
18.
M. P.
Longinotti
,
V.
Fuentes-Landete
,
T.
Loerting
, and
H. R.
Corti
,
J. Chem. Phys.
151
,
064509
(
2019
).
19.
R.
Takahashi
,
S.
Sato
,
T.
Sodesawa
, and
H.
Nishida
,
Phys. Chem. Chem. Phys.
4
,
3800
(
2002
).
20.
J.
Kunetz
and
L.
Hench
,
J. Am. Ceram. Soc.
81
,
877
(
1998
).
21.
D.
Stein
,
M.
Kruithof
, and
C.
Dekker
,
Phys. Rev. Lett.
93
,
035901
(
2004
).
22.
C.
Duan
and
A.
Majumdar
,
Nat. Nanotechnol.
5
,
848
(
2010
).
23.
J.
Peng
,
D.
Cao
,
Z.
He
,
J.
Guo
,
P.
Hapala
,
R.
Ma
,
B.
Cheng
,
J.
Chen
,
W. J.
Xie
,
X.-Z.
Li
,
P.
Jelínek
,
L.-M.
Xu
,
Y. Q.
Gao
,
E.-G.
Wang
, and
Y.
Jiang
,
Nature
557
,
701
(
2018
).
24.
F. H. J.
van der Heyden
,
D. J.
Bonthuis
,
D.
Stein
,
C.
Meyer
, and
C.
Dekker
,
Nano Lett.
7
,
1022
(
2007
).
25.
J.
Lyklema
,
J. Phys.: Condens. Matter
13
,
5027
(
2001
).
26.
S. S.
Dukhin
,
G.
Kretzchmar
, and
B.
Miller
,
Dynamics of Adsorption at Liquid Interfaces
(
Elsevier
,
1995
).
27.
H.
Zhang
,
A. A.
Hassanali
,
Y. K.
Shin
,
C.
Knight
, and
S. J.
Singer
,
J. Chem. Phys.
134
,
024705
(
2011
).
28.
R.
Hartkamp
,
B.
Siboulet
,
J.-F.
Dufrêche
, and
B.
Coasne
,
Phys. Chem. Chem. Phys.
17
,
24683
(
2015
).
29.
E.
Steinrücken
,
T.
Wissel
,
M.
Brodrecht
,
H.
Breitzke
,
J.
Regentin
,
G.
Buntkowsky
, and
M.
Vogel
,
J. Chem. Phys.
154
,
114702
(
2021
).
30.
D. C.
Martínez Casillas
,
M. P.
Longinotti
,
M. M.
Bruno
,
F.
Vaca Chávez
,
R. H.
Acosta
, and
H. R.
Corti
,
J. Phys. Chem. C
122
,
3638
(
2018
).
31.
S.
Schneider
,
C.
Säckel
,
M.
Brodrecht
,
H.
Breitzke
,
G.
Buntkowsky
, and
M.
Vogel
,
J. Chem. Phys.
153
,
244501
(
2020
).
32.
S.
Schneider
and
M.
Vogel
,
J. Chem. Phys.
149
,
104501
(
2018
).
33.
R.
Brilmayer
,
M.
Brodrecht
,
C.
Kaiser
,
H.
Breitzke
,
B.
Kumari
,
J.
Wachtveitl
,
G.
Buntkowsky
, and
A.
Andrieu‐Brunsen
,
ChemNanoMat
6
,
1843
(
2020
).
34.
K.
Schmidt-Rohr
and
H. W.
Spiess
,
Multidimensional Solid-State NMR and Polymers
(
Academic Press
,
1994
).
35.
R.
Böhmer
,
K. R.
Jeffrey
, and
M.
Vogel
,
Prog. Nucl. Magn. Reson. Spectrosc.
50
,
87
(
2007
).
36.
N.
Bloembergen
,
E. M.
Purcell
, and
R. V.
Pound
,
Phys. Rev.
73
,
679
(
1948
).
37.
J.
Hedström
,
J.
Swenson
,
R.
Bergman
,
H.
Jansson
, and
S.
Kittaka
,
Eur. Phys. J.: Spec. Top.
141
,
53
(
2007
).
38.
Y.
Kurzweil-Segev
,
A.
Greenbaum
,
I.
Popov
,
D.
Golodnitsky
, and
Y.
Feldman
,
Phys. Chem. Chem. Phys.
18
,
10992
(
2016
).
39.
J. K. H.
Fischer
,
P.
Sippel
,
D.
Denysenko
,
P.
Lunkenheimer
,
D.
Volkmer
, and
A.
Loidl
,
Commun. Phys.
3
,
95
(
2020
).
40.
M.
Sattig
and
M.
Vogel
,
J. Phys. Chem. Lett.
5
,
174
(
2014
).
41.
M.
Sattig
,
S.
Reutter
,
F.
Fujara
,
M.
Werner
,
G.
Buntkowsky
, and
M.
Vogel
,
Phys. Chem. Chem. Phys.
16
,
19229
(
2014
).
42.
C.
Lederle
,
M.
Sattig
, and
M.
Vogel
,
J. Phys. Chem. C
122
,
15427
(
2018
).
43.
D.
Demuth
,
M.
Sattig
,
E.
Steinrücken
,
M.
Weigler
, and
M.
Vogel
,
Z. Phys. Chem.
232
,
1059
(
2018
).
44.
M.
Weigler
,
M.
Brodrecht
,
G.
Buntkowsky
, and
M.
Vogel
,
J. Phys. Chem. B
123
,
2123
(
2019
).
45.
D.
Demuth
,
M.
Reuhl
,
M.
Hopfenmüller
,
N.
Karabas
,
S.
Schoner
, and
M.
Vogel
,
Molecules
25
,
4127
(
2020
).
46.
M.
Reuhl
,
P.
Monnard
, and
M.
Vogel
,
J. Chem. Phys.
156
,
084506
(
2022
).
47.
R.
Böhmer
,
M.
Storek
, and
M.
Vogel
, “
Heterogeneous rotational and translational dynamics in glasses and other disordered materials studied by NMR
,” in
Modern Magnetic Resonance
, 1st ed., edited by
G.
Webb
(
Springer
,
Cham
,
2017
), pp.
1
20
.
48.
R.
Böhmer
,
G.
Diezemann
,
G.
Hinze
, and
E.
Rössler
,
Prog. Nucl. Magn. Reson. Spectrosc.
39
,
191
(
2001
).
49.
P. T.
Callaghan
,
Principles of Nuclear Magnetic Resonance Microscopy
(
Clarendon Press
,
Oxford
,
1991
).
50.
M.
Weigler
,
E.
Winter
,
B.
Kresse
,
M.
Brodrecht
,
G.
Buntkowsky
, and
M.
Vogel
,
Phys. Chem. Chem. Phys.
22
,
13989
(
2020
).
51.
M.
Reuhl
,
M.
Weigler
,
M.
Brodrecht
,
G.
Buntkowsky
, and
M.
Vogel
,
J. Phys. Chem. C
124
,
20998
(
2020
).
52.
Diffusion NMR of Confined Systems: Fluid Transport in Porous Solids and Heterogeneous Materials
, edited by
R.
Valiullin
(
The Royal Society of Chemistry
,
2017
).
53.
G.
Fleischer
and
F.
Fujara
, “
NMR, basic principles and progress
,” in
Solid-State NMR I Methods
, 1st ed., NMR Vol. 30, edited by
P.
Diehl
,
E.
Fluck
,
H.
Günther
,
R.
Kosfeld
, and
J.
Seelig
(
Springer Berlin Heidelberg
,
1994
), pp.
159
207
.
54.
M.
Brodrecht
,
H.
Breitzke
,
T.
Gutmann
, and
G.
Buntkowsky
,
Chem. - Eur. J.
24
,
17814
(
2018
).
55.
M.
Kobayashi
and
H.
Tanaka
,
Phys. Rev. Lett.
106
,
125703
(
2011
).
56.
C.
Monnin
,
M.
Dubois
,
N.
Papaiconomou
, and
J.-P.
Simonin
,
J. Chem. Eng. Data
47
,
1331
(
2002
).
58.
M.-H.
Zheng
,
J.-Y.
Jin
,
W.
Sun
, and
C.-H.
Yan
,
New J. Chem.
30
,
1192
(
2006
).
60.
J.
Han
,
A.
Loudet
,
R.
Barhoumi
,
R. C.
Burghardt
, and
K.
Burgess
,
J. Am. Chem. Soc.
131
,
1642
(
2009
).
61.
D.
Bottenus
,
Y.-J.
Oh
,
S. M.
Han
, and
C. F.
Ivory
,
Lab Chip
9
,
219
(
2009
).
62.
M.
Tagliazucchi
,
O.
Azzaroni
, and
I.
Szleifer
,
J. Am. Chem. Soc.
132
,
12404
(
2010
).
63.
F. M.
Gilles
,
M.
Tagliazucchi
,
O.
Azzaroni
, and
I.
Szleifer
,
J. Phys. Chem. C
120
,
4789
(
2016
).
64.
E.
Gonzalez Solveyra
,
R. J.
Nap
,
K.
Huang
, and
I.
Szleifer
,
Polymers
12
,
2282
(
2020
).
65.
J. M.
Rosenholm
and
M.
Lindén
,
J. Controlled Release
128
,
157
(
2008
).
66.
A.
Walcarius
,
M.
Etienne
, and
B.
Lebeau
,
Chem. Mater.
15
,
2161
(
2003
).
67.
T.
Feiweier
,
O.
Isfort
,
B.
Geil
,
F.
Fujara
, and
H.
Weingärtner
,
J. Chem. Phys.
105
,
5737
(
1996
).
68.
M.
Nakanishi
,
P.
Griffin
,
E.
Mamontov
, and
A. P.
Sokolov
,
J. Chem. Phys.
136
,
124512
(
2012
).
69.
S.
Sudo
,
S.
Tsubotani
,
M.
Shimomura
,
N.
Shinyashiki
, and
S.
Yagihara
,
J. Chem. Phys.
121
,
7332
(
2004
).
70.
S.
Capaccioli
,
K. L.
Ngai
,
S.
Ancherbak
,
M.
Bertoldo
,
G.
Ciampalini
,
M. S.
Thayyil
, and
L.-M.
Wang
,
J. Chem. Phys.
151
,
034504
(
2019
).
71.
M.
Reuhl
,
P.
Monnard
, and
M.
Vogel
,
J. Chem. Phys.
155
,
224501
(
2021
).
72.
Y.
Yao
,
V.
Fella
,
W.
Huang
,
K. A. I.
Zhang
,
K.
Landfester
,
H.-J.
Butt
,
M.
Vogel
, and
G.
Floudas
,
Langmuir
35
,
5890
(
2019
).
Published open access through an agreement with Technische Universität Darmstadt

Supplementary Material