You are on page 1of 11

RESEARCH ARTICLE

Rising Stars www.advmat.de

A Nanoconfined Water–Ion Coordination Network for


Flexible Energy-Dissipation Devices
Yuan Gao,* Mingzhe Li, Chi Zhan, Haozhe Zhang, Mengtian Yin, Weiyi Lu,*
and Baoxing Xu*

1. Introduction
Water–ion interaction in a nanoconfined environment that deeply constrains
spatial freedoms of local atomistic motion with unconventional coupling Confining water in nanoenvironments has
yielded a wide variety of unprecedented
mechanisms beyond that in a free, bulk state is essential to spark designs of a
properties beyond the free bulk phase.[1–5]
broad spectrum of nanofluidic devices with unique properties and Particularly, with a sub-10 nm confinement
functionalities. Here, it is reported that the interaction between ions and dimension approaching the molecular size,
water molecules in a hydrophobic nanopore forms a coordination network the enormous surface-to-volume ratio sig-
with an interaction density that is nearly fourfold that of the bulk counterpart. nificantly promotes the interaction between
the confined water molecules and the solid
Such strong interaction facilitates the connectivity of the water–ion network
nanochannel,[6] substantially altering the
and is uncovered by corroborating the formation of ion clusters and the properties of water, for example, ultra-fast
reduction of particle dynamics. A liquid-nanopore energy-dissipation system transport properties,[7–11] extreme phase-
is designed and demonstrated in both molecular simulations and experiments transition temperature,[12] anomalously low
that the formed coordination network controls the outflow of confined dielectric constant,[13] and suppressed os-
electrolytes along with a pressure reduction, capable of providing flexible cillation of surface energy.[14] Compared
with confined sole water molecules, intro-
protection for personnel and devices and instrumentations against external
ducing ions may alter the coordination of
mechanical impact and attack. water molecules,[15,16] leading to substan-
tially different structural and dynamic prop-
erties of confined water.[17,18] For example,
the reactivity of CO2 (aq) solutions can be
notably enhanced by the nanoconfinement with graphene and
Y. Gao stishovite (SiO2 ), which is correlated to mineral carbonation deep
State Key Laboratory of Digital Manufacturing Equipment and Technology in the Earth.[19] The presence of ions leads to an unexpectedly
Huazhong University of Science and Technology
Wuhan 430074, China severe metastable state of liquid outflow from a hydrophobic
E-mail: yuan_gao@hust.edu.cn nanopore.[20–22] Understanding the water–ion interactions and
Y. Gao, H. Zhang, M. Yin, B. Xu the associated out-of-trend phenomena is the foundation for
Department of Mechanical and Aerospace Engineering exploring nanofluidics-based devices in salt desalination,[23–25]
University of Virginia sensing,[26–28] water disinfection,[29] and energy conversion.[30–33]
Charlottesville, VA 22904, USA
E-mail: bx4c@virginia.edu
However, the perturbation effects of ions on the intermolecu-
M. Li, C. Zhan, W. Lu
lar structure of water in aqueous solutions remain controver-
Department of Civil and Environmental Engineering sial so far regarding the effect of ions on the structure of hy-
Michigan State University drogen bonds (HBs). For example, atomistic simulations and
East Lansing, MI 48824, USA experiments show that the presence of ions will strengthen or
E-mail: wylu@egr.msu.edu degrade the hydrogen-bonding networks and affect the struc-
W. Lu tural, physical, and chemical properties,[34–36] according to which
Department of Chemical Engineering and Materials Science
Michigan State University they can be categorized as “structure makers” and “structure
East Lansing, MI 48824, USA breakers”.[37] Neutron diffraction experimental measurements
reveal that the effect on HBs is ion-specific.[38] By contrast, Ra-
The ORCID identification number(s) for the author(s) of this article man spectra and femtosecond mid-infrared pump-probe mea-
can be found under https://doi.org/10.1002/adma.202303759 surements challenge this opinion that the effect of ions is ex-
© 2023 The Authors. Advanced Materials published by Wiley-VCH tremely local (within the first solvation shell), and the free water
GmbH. This is an open access article under the terms of the Creative molecules retain bulk properties.[39,40]
Commons Attribution-NonCommercial-NoDerivs License, which permits Here, we report a water–ion coordination network comprising
use and distribution in any medium, provided the original work is
properly cited, the use is non-commercial and no modifications or water–water, water–ion, and ion–ion interactions in the lithium
adaptations are made. chloride (LiCl) solution confined in a hydrophobic nanopore. Its
interaction intensity is a nearly fourfold bulk counterpart and is
DOI: 10.1002/adma.202303759

Adv. Mater. 2023, 35, 2303759 2303759 (1 of 11) © 2023 The Authors. Advanced Materials published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmat.de

Figure 1. Water–ion coordination network subjected to nanoconfinement. a) Atomistic snapshot of LiCl aqueous solution confined in a hydrophobic
silica nanopore. Water molecules are not explicitly visualized. The hydrogen-bonding network is highlighted by the mesh. b) Distribution of solid–liquid
interaction energy density ESL with various LiCl weight percentages. c) Water–ion coordination network with various LiCl weight percentages. d) Number
fractions of HB (yellow), WI (blue), and II (navy) interactions (circles) and degree of connection DOC (stars) as functions of the LiCl weight percentage.
The DOC of the bulk electrolyte (hollow stars) is presented as a reference. The uncertainty is smaller than symbols. e,f) Connection density 𝜌c of each
type of interaction along the radial direction in confined water and 40% LiCl solution (symbols), accompanied by values obtained in bulk liquids at 300 K
and 1 atm (dashed lines). The shaded regions indicate the error.

strongly coupled with the interaction correlation time and the for- of 12.0 nm and a radius of r0 = 1.57 nm. The hydrogen bonding
mation of local ion clusters and bound water. Further results re- network is visualized by lines and surface mesh, and the water
veal that, unlike the localized effects in bulk liquids, the nanocon- molecules are not explicitly shown. HBs are identified with ge-
fined network globally suppresses the rotational and translational ometric criteria described in the Experimental Section, and the
dynamics of the particles. With both molecular dynamics simu- criteria are consistent in all MD simulations. The discontinu-
lation and experiments, we have demonstrated that these coordi- ous HB network suggests that the presence of Li+ and Cl− re-
nation mechanisms can be utilized to design a reusable, flexible sists the formation of HBs by separating water molecules. The
liquid-nanopore energy-dissipation system. The formation of the water–ion and ion–ion interactions, highlighted by the arrows,
water–ion network effectively governs the outflow process of con- can be indicated by the significant solvation shells in partial ra-
fined liquid, attenuating the pressure level during the unloading dial distribution functions (RDFs) presented in Figure S1 (Sup-
process of impact and henceforth alleviating the secondary dam- porting Information). Figure 1b compares the planar distribution
age for the system’s reusability. of the interaction energy between the solid nanopore and the con-
fined liquid ESL in solutions with different LiCl weight percent-
2. Results ages of 0 (water), 10%, 20%, and 40%, where negative and posi-
tive values represent cohesive and repulsive interactions, respec-
2.1. Water–Ion Coordination Network in a Hydrophobic tively. The distributions are similar for various LiCl weight per-
Nanopore centages, with the cohesive interaction more pronounced close to
the wall of the nanopore (black circles), which corresponds to the
Figure 1a presents an atomistic snapshot of LiCl aqueous solu- solid–liquid interface layer with higher particle densities due to
tion confined in a hydrophobic silica nanopore with a length L the confinement effect (Figure S2, Supporting Information). The

Adv. Mater. 2023, 35, 2303759 2303759 (2 of 11) © 2023 The Authors. Advanced Materials published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmat.de

distribution is more random at a location further from the solid– 2.2. Cohesion and Dynamics in the Water–Ion Network
liquid interface. In addition, a higher LiCl weight percentage
promotes the overall solid–liquid interaction, as quantitatively Figure 2a exhibits the interaction energy decoupling analyses on
summarized by the solid–liquid interaction energy density 𝛾 SL in the water–ion network, where negative values represent cohe-
Figure S3 (Supporting Information). These results show that the sive interactions (see Experimental Section). With an increased
ions contribute to the cohesion of confined liquid even though LiCl weight percentage, the water–water interaction energy de-
they block the formation of the HB network. They also suggest clines while the water–ion and ion–ion interaction energy is el-
that using only the HB network is insufficient to reflect the inter- evated, consistent with the evolution of interaction fractions nc
action between confined liquid particles, different from that in a in Figure 1d. Consequently, the overall interaction energy is pro-
free, bulk counterpart.[40] moted, suggesting that the introduction of ions enhances the co-
The water–ion and ion–ion interactions are determined ac- hesion of the confined solution, which agrees well with the in-
cording to the RDFs (Figure S1, Supporting Information, also see creasing overall degree of connection (Figure 1d). The stronger
Experimental Section) and expressed as imaginary connections cohesion due to a larger population of ions leads to prolonged cor-
in complement to the HBs in Figure 1c. Similar to HBs, these relation time 𝜏 c for each type of interaction, as demonstrated by
connections are more robust than non-bonded interactions, and Figure 2b. The correlation time was determined via the autocorre-
their breaks and reconstructions are strongly associated with the lation function S(𝜏) (inset of Figure 2b, also see Experimental Sec-
thermodynamics of water molecules and ions. With a higher LiCl tion), and a more rapid reduction in S(𝜏) corresponds to a shorter
weight percentage, the number of HBs (yellow) is reduced while 𝜏 c . More autocorrelation functions are presented in Figure S9
the quantities of water–ion (WI, blue) and ion–ion (II, navy) inter- (Supporting Information). Apparently, 𝜏 c values of ion–ion and
actions rise due to the solvation of ions.[41] Representative snap- water–ion connections are higher than water–water connections
shots with different LiCl weight percentages support these results dominated by HBs, reflecting stronger interaction strengths. The
(Figure S4, Supporting Information). Figure 1d further charac- 𝜏 c of Li+ –H2 O is longer than that of Cl− –H2 O due to higher hy-
terizes the coordination network with the fraction of each type dration energy.[43] Extensive calculations suggest that an elevated
of connection nc (circles, left vertical axis). The network mainly temperature can reduce 𝜏 c (Figure S10, Supporting Information).
contains WI interactions for wt% > 20, and the II interactions oc- The comparison between the confined liquid (circles) and the
cupy less than 10% of all interactions. Given the opposite trend bulk counterpart (squares) suggests that the confinement effect
of the increase in the numbers of WI and II and the reduction in mainly promote 𝜏 c of WI and HB. Such an effect is more pro-
HBs with a higher population of ions, the degree of connection, nounced with a higher ion population. Moreover, 𝜏 c of HB, WI,
DOC = (NHB + NWI + NII )∕(0.5N(N − 1)), is defined to quantify and II interactions negatively depends on the confinement ra-
the overall connectivity in the confined liquid and plotted as stars dius r0 and the ionic radius of the cation, as demonstrated in
in Figure 1d (right vertical settings). In the definition NHB , NWI , Figure S11 (Supporting Information).
and NII represents the total number of HB, WI, II interactions, The extended correlation time for WI and II interactions by
and the denominator 0.5N(N − 1) is the maximum number of a higher LiCl weight percentage is associated with the ion clus-
interactions in a system with N particles. The DOC in confined ters, as illustrated in Figure 2c. The cluster with a larger size
liquid (solid stars) is ≈1.25 times that in the bulk counterpart at (more ionic members) contributes more to both WI and II in-
300 K and 1 atm due to the confinement effect. Further calcu- teractions, and the interaction energy of a cluster with a specific
lations suggest that the water–ion coordination network can be size is independent of the LiCl weight percentage. The formation
observed with various confinement radii r0 and types of cations. of clusters and enhanced ion–ion correlation agree well with pre-
A larger r0 and ionic radius of the cation can reduce the value of vious experimental studies.[44,45] However, a higher weight per-
DOC (Figure S5, Supporting Information). In addition, a higher centage increases the population of ionic pairs or clusters, as pre-
confinement temperature T can suppress DOC, as demonstrated sented in Figure 2d, facilitating the cohesion of the network. To
in Figure S6 (Supporting Information). understand the enhanced HB correlation time by the presence of
Figure 1e,f demonstrates the radial distribution of the connec- ions, Figure 2e,f compares the potential of mean force (PMF, see
tion density 𝜌c of HB, WI, and II in the confined water and 40% Experimental Section) with respect to intermolecular geometric
LiCl solution (symbols), accompanied by the 𝜌c values in bulk properties, namely the molecular separation rOO and pair angle
water and solution (dashed lines). For water, the confinement ef- 𝜃 OOH (the atomistic model in Figure 2e), in confined water and
fect promotes the formation of HBs mainly in the interface layer, LiCl solution. In both cases, a notable potential valley is located
where the particle density is higher (Figure S2, Supporting In- on the bottom-left of the graph, which corresponds to HBs. Two
formation), as suggested by the peak at r/r0 ≈ 0.65 that notably PMF barriers on the top and right of the HB region respectively
surpasses the dashed line. In the presence of ions (Figure 1f), represent the rotational and translational dynamics needed to de-
the confinement effect on HBs is still more pronounced in the stroy a hydrogen bond in the system. In the presence of ions,
interface layer. However, the confinement effect largely facili- the saddle region becomes narrower, suggesting promoted HB
tates the WI formation outside the interface layer, promoting strengths. More PMF contours are presented in Figure S12 (Sup-
𝜌c by several times. Coordination analyses indicate that this ob- porting Information). A higher temperature can expand the sad-
servation is correlated to the larger ionic clusters at r/r0 < 0.4 dle region and decrease the PMF potential valley, reducing the
(Figure S7, Supporting Information), which possess greater sol- strength of HBs (Figure S12 Supporting Information).
vation shells.[42] The same observation is obtained with various The correlation time of interactions in the water–ion network
LiCl weight percentages, as shown in (Figure S8, Supporting is related to the dynamics of the particles. Figure 3a illustrates
Information). the relationship between the rotational correlation time 𝜏 rot of

Adv. Mater. 2023, 35, 2303759 2303759 (3 of 11) © 2023 The Authors. Advanced Materials published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmat.de

Figure 2. Interaction energy and correlation time of connections in the water–ion network. a) Overall (hollow stars) and decoupled interaction energy of
HB (yellow), WI (blue), and II (navy) interactions (circles) in the water–ion network as functions of the LiCl weight percentage. Shaded regions indicate
uncertainties. b) Correlation time 𝜏 c of interactions as functions of the LiCl weight percentage. The results of bulk water and electrolyte are presented
for reference. c) Water–ion and ion–ion interaction energies contributed by ions as a function of the cluster size. d) Probabilities of ions that participate
in clusters of various sizes. e,f) Potential of mean force with respect to the molecular separation rOO and pair angle 𝜃 OOH in confined water and 40 wt%
LiCl solution.

water molecules (hollow circles, see Experimental Section) and Dz can be partially understood by the negative dependence on
the LiCl weight percentage, which was obtained with the rota- the cluster size, as shown in Figure 3d. More calculations illus-
tional correlation functions (Figure S13, Supporting Informa- trate that a higher temperature promotes both the rotational and
tion). A higher weight percentage suppresses the rotational dy- translational dynamics of the particles (Figure S15, Supporting
namics (longer 𝜏 rot ), which corresponds well to the extended 𝜏 c of Information).
HBs (Figure 2d). The mechanism can be understood by catego-
rizing water molecules into bound (participate in an ion solvation
shell) and free water molecules (not participate in any solvation 2.3. Outflow Control of Confined Solution by Water–Ion Network
shells). The 𝜏 rot of bound water molecules exhibits a stronger pos-
itive dependence on the weight percentage than the free water. In Immersing a hydrophobic nanopore in a liquid environment
addition, a higher population of ions, ionic pairs, and clusters lifts can form an energy-dissipation system (Figure S15, Supporting
the portion of bound waters (squares), and the overall rotational Information), which relies on the infiltration–outflow process
dynamics is dominated by bound water molecules at high LiCl to attenuate external mechanical impact.[46] In this section, we
weight percentages. demonstrate that the water–ion interaction mechanisms can be
To investigate the translational dynamics of water, Figure 3b used to suppress the electrolyte outflow. Owing to the hydropho-
presents the effect of LiCl weight percentage on the diffusion co- bic nature of the nanopore, no water molecules or ions can spon-
efficient in the axial (z-) direction, Dz , of water molecules, de- taneously infiltrate the nanopore. An external pressure loading
termined by the mean square displacement (Figure S14, Sup- can be exerted by applying a quasistatic displacement-controlled
porting Information). The diffusion of water molecules is also loading on the piston to compress the system, as illustrated by
restricted by the ions, stabilizing the water–water connections. Figure 4a. Once the pressure reaches the infiltration pressure
Similar to rotational dynamics, translational dynamics is domi- Pin , the capillary resistance of the hydrophobic nanopore is over-
nated by bound water molecules with high LiCl weight percent- come, and substantial intrusion occurs. In the current study, all
ages, although the difference between bound and free molecules Pin values were taken as the mean pressure value of the linear
is not significant. Figure 3c demonstrates that the Dz of ions also part in the P –ΔV/V0 curve upon loading, where V0 is the ini-
declines with a higher ion population. The decreasing trend of tial volume of the system. A higher ion population promotes the

Adv. Mater. 2023, 35, 2303759 2303759 (4 of 11) © 2023 The Authors. Advanced Materials published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmat.de

Figure 3. Dynamics of confined water molecules and ions. a) Rotational correlation time 𝜏 c (left vertical axis) and the portion of bound water (right
vertical axis) as functions of the LiCl weight percentage in the confined solution. b,c) Axial diffusion coefficient Dz of water molecules and ions as a
function of the weight percentage. d) Dz of ions in various cluster sizes. The shaded region represents the uncertainty. The uncertainty is smaller than
symbols when not presented.

infiltration pressure Pin , supported by more detailed calculations the P–ΔV/V0 relationship starts to return to the loading curve in
(Figure S16, Supporting Information). At the end of the curve Figure 4a, and significant reductions in both Nin and the pressure
plateau, the liquid completely infiltrates the nanopore and forms difference are observed in Figure 4b (stage II). The bi-stage vari-
the water–ion network discussed in previous sections, and the ation in the pressure difference Pin − P during the unloading is
system becomes incompressible. Overloading the system will observed for various LiCl weight percentages and temperatures
cause a rapid increase in pressure. It should be noted that the (Figure S18, Supporting Information). The slower outflow rate of
concentration of ions confined in the nanopore is reduced by the LiCl solution leads to ≈900 residual particles at the end of the
less than 10% due to the accumulation effects at both ends of unloading (ΔV/V0 = 0), suggesting an incomplete outflow.
the nanopore (Figure S17, Supporting Information). As discussed above, a higher population of ions and a lower
Quasistatic unloading is conducted by moving the piston in temperature enhance the connectivity and correlation times of
the opposite direction. In the beginning, the unloading P–ΔV/V0 connections of the network, which lead to a slower outflow rate.
curves coincide with the loading histories before the system re- To further highlight this fact, Figure 4d,e compares the evolution
covers from the overloaded stage. The outflow process is initiated of the confined particle number 𝜌N and connection densities 𝜌C
when ΔV/V0 drops below the volume leading to the complete of water and 26 wt% LiCl solution during the outflow. Although
infiltration, for example, ΔV/V0 = 0.08 for water. Meanwhile, the high intensities of 𝜌N locate in the solid–liquid interaction
the P–ΔV/V0 evolution starts to deviate from the loading curve. layer for both water and LiCl solution, the water–ion interactions
The pressure drops below Pin , suggesting the internal cohesion (Figure 1f) in the solution significantly strengthen the network
of the liquid particles is acting against the unloading. According in a larger volume. Consequently, the outflow can be reduced by
to the pressure balance, the equivalent “stretch loading” exerted the introduction of ions. As presented in Figure S19 (Support-
on the confined liquid is the difference between the infiltration ing Information), a higher LiCl weight percentage can reduce
pressure and the real-time pressure, highlighted in Figure 4b the outflow rate and extend the outflow process. The degrada-
(symbols). We also track the number of liquid particles inside the tion of the water–ion network is more pronounced at higher r/r0
nanopore Nin (curves) during the outflow process, and the result due to the significant surface stress at the solid–liquid interface
shows that the liquid cohesion is resisting the outflow, suggested (Figure S20, Supporting Information). More detailed analyses
by a slow reduction in Nin for both water and 26 wt% LiCl solu- demonstrate that the fraction of connections does not vary during
tion (stage 1). Nevertheless, the mild outflow degrades the water– the outflow (Figure S21, Supporting Information).
ion network, which can be visualized in Figure 4c, where the The bi-stage outflow process can be characterized by two fea-
solid-like necking behavior is observed for both types of liquid. tures: a constant outflow rate during stage 1, R1 , and the de-
Further unloading breaks down the network and the cohesion of lay time for the substantial outflow on stage 2, t2 (Figure S22,
the liquid, removing the resistance to outflow. Correspondingly, Supporting Information). According to Figure 4f, a general

Adv. Mater. 2023, 35, 2303759 2303759 (5 of 11) © 2023 The Authors. Advanced Materials published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmat.de

Figure 4. Outflow control of liquid in hydrophobic nanopores by the water–ion network. a) External pressure loading P on the liquid–hydrophobic
nanopore system as a function of the normalized volume change ΔV/V0 , where V0 is the initial volume of the system. The liquid can be either water or
LiCl solution. b) Number of confined particles Nin (curves) and the pressure difference between the infiltration pressure and unloading pressure Pin –P
symbols) as functions of the system volume change ΔV/V0 during the outflow process. c) Representative snapshots of water–ion networks of water (I
and II) and LiCl solution (III and IV) during the outflow process. d,e) Distributions of the confined particle density 𝜌N and connection density 𝜌C in water
and LiCl solution during the outflow process. f) General relationship between R1 and PRR. g) General relationship between t2 and MPD. Shaded regions
near curves indicate uncertainty.

relationship can be found between R1 and the pressure reduction details), SEM images of the nanopore, information on surface
rate (PRR), defined as the negative slope of the P–ΔV/V0 curve treatment, and the distribution of the pore size. According to
during stage 1 of outflow, as marked in Figure 4a. A higher PRR the SEM images, irregular nanopores are interconnected in a 3D
reflects the stronger cohesion of the water–ion network, which manner.[47] The average radius of nanopores r0 is ≈3.87 nm, close
limits the dynamics of particles and hinders the outflow. The re- to that in MD simulations, as presented in Figure 5b. The in-
lationship includes the outflow with various LiCl weight percent- filtration pressure is increased in the presence of ions, and the
ages and at various temperatures. In addition, the enhanced con- MPD is enhanced owing to the strengthened water–ion coordi-
nection of the network also promotes the internal pressure that nation network, which indicates a reduced outflow. These obser-
the confined liquid can endure, delaying the substantial outflow vations agree well with simulations. Additional P–ΔV/V0 curves
in stage 2. Figure 4g demonstrates the general relationship be- obtained with various LiCl weight percentages and temperatures
tween t2 and the maximum pressure difference (MPD), defined are given in Figure S24 (Supporting Information), which sup-
as the maximum difference between the infiltration pressure Pin port the same trend. To examine the completion of outflow, cyclic
and the external pressure, which is also marked in Figure 4a. loading tests were performed, as shown in Figure 5c. For a sys-
Both PRR and MPD positively depend on the LiCl percentage but tem with water, the pressure loading histories are almost iden-
reduce with T (Figure S23, Supporting Information). tical in the three successive cycles, indicating the complete out-
To validate the mechanism and tunable outflow investigated flow of the confined water.[22] However, for systems with 26 wt%
with atomistic simulations, we have conducted quasistatic com- LiCl solution, hysteresis occurs in the second and third load-
pression experiments on liquid–silica gel systems. Figure 5a ing cycles, which suggests that the attenuated outflow leads to
presents the experimental setup (see Experimental Section for residual liquid at the end of the unloading. This observation is

Adv. Mater. 2023, 35, 2303759 2303759 (6 of 11) © 2023 The Authors. Advanced Materials published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmat.de

Figure 5. Liquid infiltration and outflow experiments on LiCl solution/hydrophobic silica nanopores. a) Schematic of the experiment setup, SEM photo of
silica nanoporous particles, and the surface treatment of nanopores. The nanoporous media displays interconnected irregular pores. The contact angle
image confirms the hydrophobicity of nanopores after surface treatment. The width of nanopores exhibits a narrow distribution. b) Relationship between
externally applied quasistatic pressure and reduced volume change ΔV/V0 in experiments, where V0 is the initial system volume. c) Pressure–volume
loading curves in cyclic loading tests.

consistent with MD simulations (Figure 4a,b). Moreover, both reduction performance is directly related to the strength of water–
PRR and MPD measured in experiments increase with a higher ion networks and can be characterized by the PRR proposed in
salt weight percentage but decline with temperature (Figure S25, Figure 4f. In other words, a higher PRR corresponds to more
Supporting Information), which agrees well with MD simula- rapid pressure alleviation during the unloading process, which
tions (Figure S26, Supporting Information). can significantly relieve the secondary damage. As for an experi-
mental demonstration, we have fabricated cushion pads that in-
corporate liquid-nanopore systems with water and 26 wt% LiCl
2.4. Application Demonstration in Designing Flexible electrolyte (Figure 6c) and evaluate their energy-dissipation per-
Energy-Dissipation Device with Water–Ion Network formance under typical flat-on-flat impact conditions. Figure 6d
presents typical acceleration (proportional to the applied pres-
We demonstrate a conceptual design that the liquid-nanopore sure) histories of the cushion pads during an external impact.
systems comprising water–ion networks can be utilized to de- While the acceleration histories of both systems are highly con-
sign a reusable, flexible energy-dissipation device, as presented sistent during the loading process, the acceleration of the LiCl
in Figure 6a. The liquid-nanopore system is enclosed in a thin- system declines at a faster rate than the system with sole water, re-
film carrier made of soft materials. Due to the flexibility of sulting in a more rapid reduction of acceleration (in g) to the safe
liquid and soft materials, the device can provide conformal range. To quantify the improvement in the protective capability,
attachment to the human body. Under an external mechan- we evaluate the impact Severity Index (SI) (NOCSAE DOC ND
ical impact, the pressure P exerted on the system increases 001–17m23) via SI = ∫ Acc2.5 dt, where Acc represents the accel-
(Figure 6b, navy curves) and triggers the infiltration of elec- eration in units of gravity (g). As compared in Figure 6e, the SI for
trolytes. The infiltration process dissipates part of the mechanical the system with water is 300.4 ± 11.7, whereas the ion-containing
energy by overcoming capillary resistance and friction between system shows a much-reduced SI of 251.2 ± 8.6. This result sug-
the liquid and the nanopore,[48] which accompanies the impact gests that a higher PRR of the outflow improves impact miti-
loading. gation performance. To guide the design of high-performance
For protection devices with fully elastic behavior, P in the un- energy-dissipation systems, a scaling law has been established
loading process is expected to follow its behavior during the im- between the PRR and the solid–liquid interface energy density
pact loading (Figure 6b, blue curves), maintaining a high level 𝛾 SL and the radius of nanopores r0 (Figure 6f,g), which is based
for a large portion of the loading–unloading cycle. However, for on the similarity between the confined liquid outflow and the
devices based on liquid-nanopore systems, the bi-stage outflow tensile mechanical deformation of solid nanowires (Figure S27,
process will occur (Figure 4b), significantly reducing the pres- Supporting Information).[49] A higher solid–liquid interface en-
sure P (Figure 6b, yellow curves) and, thus, the associated sec- ergy density 𝛾 SL can limit the rotational and translational dynam-
ondary damage during the unloading process. Such pressure- ics (Figure S28, Supporting Information). A larger PRR of the

Adv. Mater. 2023, 35, 2303759 2303759 (7 of 11) © 2023 The Authors. Advanced Materials published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmat.de

Figure 6. Application demonstration in designing electrolyte–hydrophobic nanopore flexible energy-dissipation device. a) Schematic of the conceptual
design: a flexible energy-dissipation device enabled by the liquid-nanopore system. b) Evolution of the pressure of the external loading P during the
loading–unloading cycle with the energy-dissipation device (yellow curves) and without the device (i.e., fully elastic behavior, blue curves). c) Illustration
of the fabrication process and optical image of flexible cushion pads. d) Acceleration history of cushion pads containing liquid-nanopore systems with
water and 26 wt% LiCl electrolyte during the flat-to-flat impact. e) Comparison of SI of the cushion pads with water and 26 wt% LiCl electrolyte. A lower
value of SI represents better protection performance. f,g) Scaling laws of the pressure reduction rate PRR in simulations and experiments, where 𝛾 SL
and r0 denotes the solid–liquid interaction energy density and the radius of nanopores, respectively.

energy-dissipation device can be achieved with a higher 𝛾 SL and thus delays the outflow process. The bi-stage outflow process can
lower r0 , capable of readily preventing the damage during the un- be characterized by the features in the pressure loading histories,
loading process. namely the PRR, and maximum pressure difference, via general-
ized relationships. Furthermore, we have demonstrated that the
3. Conclusion water–ion interaction and transport mechanisms can be adopted
to design a reusable, flexible liquid-nanopore energy-dissipation
We have reported the formation of water–ion coordination net- device. The outflow attenuates the mechanical releasing rate and
works by investigating the interaction between the lithium and reduces the pressure level during the unloading process of ap-
chlorine (LiCl) ions and water molecules confined in a silica hy- plied impact. These findings provide immediate application guid-
drophobic nanopore. The network possesses a higher degree of ance to designing nanofluidic-based flexible protection systems
connectivity and interaction energy than the hydrogen bonding with adjustable performance and responsive properties to exter-
network in the confined environment. Comprehensive analyses nal mechanical stimuli by leveraging the formed water–ion net-
of the dynamics of the system show that a higher population of works in a nanoconfinement environment.
ions extends the correlation time of the interaction between dif-
ferent species and restricts the rotational and translational dy-
namics of particles. Its correlations with both the formation of 4. Experimental Section
ion pairs and clusters and the participation fraction of bound wa-
Atomistic Modeling and Computational Methods: The atomistic model
ter in the solvation shells of ions are also elucidated. We have contained a cylindrical rigid silica nanopore with a radius r0 of 1.57 nm
illustrated that these interaction mechanisms can be utilized to and a length L of 12.0 nm. The nanopore was immersed in a liquid reser-
alter the outflow of the confined liquid. The water–ion coordina- voir composed of water or LiCl aqueous solution. The liquid reservoir pos-
tion network enhances the cohesion of the confined liquid and sessed a dimension of 7 nm (x-direction) × 7 nm (y-direction) × 25 nm

Adv. Mater. 2023, 35, 2303759 2303759 (8 of 11) © 2023 The Authors. Advanced Materials published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmat.de

(z-direction), containing ≈100 000 atoms. The liquid reservoir was sealed number of oxygen atoms in a shell between rOO and rOO + drOO from
with a rigid wall and a piston movable in the z-direction. The two ends a specific O atom if the pair angle 𝜃 OH existed between 𝜃 OH + d𝜃 OH ,
of the nanopore in the z-direction were 8 and 4 nm from the piston and to the same number if the molecules were noninteracting, which was
wall, respectively. Periodic boundary conditions were adopted in the planar 2𝜋𝜌N sin 𝜃OH d𝜃OH rOO 2 dr
OO . In this expression, 𝜌N denotes the number
(x- and y-) directions, and a nonperiodic boundary condition was adopted density of the confined water. In the calculation, the step differences for
in the axial (z-) direction. A schematic of the model was presented in radius d𝜃 OH and angles 𝜃 OH were set as 0.01 Å and 0.4°, respectively. At
Figure S13 (Supporting Information). The water molecules were described least 10 000 frames of molecular coordinates during 0.5 ns were utilized
by the SPC/E model that reproduced reliable dynamics and structural in calculations.
properties.[50] Similar numerical results could be obtained by the modified Dynamics of the Confined Water–Ion Network: The rotational dynamics
TIP3P model (Figure S29, Supporting Information).[51] The non-bonded of confined water molecules were studied with the rotational autocorre-
interactions between water molecules, ions, and the nanopore were mod- lation function coupled with the second-order Lagrange Polynomials:[58]
eled by the Lennard-Jones potential.[46,52] The particle–particle–particle– C𝜇 (𝜏) = 12 ⟨3[𝜇(t)𝜇(t + 𝜏)]2 − 1⟩t , where 𝜇 is the dipole orientation vector.
mesh with a root mean of 10−4 was utilized to capture the Coulomb inter- The rotational correlation time 𝜏 rot could be estimated as the time when
action between charged particles. C𝜇 (𝜏) was reduced to 1/e. The translational dynamics of the confined water
All molecular dynamics simulations were conducted with LAMMPS.[53] molecules and ions was investigated with the mean-square displacement,
The time step was set as 0.5 fs. A temperature ≈2000 K was applied to the MSD, calculated via MSD (𝜏) = ⟨|ri (t + 𝜏) − ri (t)|2 ⟩. The effect of the cen-
systems containing ions under the canonical (NVT) ensemble for 4 ns ter of mass was not considered since no obvious drifting was observed
until the distribution of ion concentration was homogenous in the liquid (Figure S30, Supporting Information).
reservoir. Next, the system was relaxed at the target temperature between Material Preparation: The liquid-nanopore system used in the qua-
300 and 360 K for 1 ns. The position of the piston in the z-direction was sistatic experiment contained a hydrophobic nanoporous media and LiCl
adjusted by a rigid-body algorithm[54] to achieve a zero initial pressure P aqueous solutions. The raw porous media was a hydrophilic nanoporous
on the piston, and the process took another 1 ns. To establish the con- silica gel (Supelco 53 698, Sigma-Aldrich, Inc.). The material was re-
fined water–ion network, a quasistatic displacement loading in the nega- ceived in powder form with a particle size ranging from 40–75 μm. The
tive z-direction was applied to the piston. The outflow process was trig- hydrophilic nanopore surface was converted to hydrophobic by anchor-
gered by moving the position in the positive z-direction. Considering the ing a thin layer of n-octyldimethylchlorosilane (SIO6711.0, Gelest Inc.)
thermodynamics noise of the system,[7] the results of pressure were aver- on the nanopore wall.[59] In brief, a mixture containing 1 g of silica gel
aged using the data in 10 000 adjacent frames. In addition, each outflow and 40 mL of anhydrous toluene was stirred for 3 h at 90 °C. 10 mL of
simulation was repeated 5 times to study the properties of pressure re- n-octyldimethylchlorosilane and 1 mL of pyridine was added to the mix-
duction rate and maximum pressure difference. The planar distribution of ture at room temperature, after which the mixture was stirred for 24 h
the interaction energy ESL was obtained by calculating the potential en- at 95 °C. The treated silica gel was then filtered, thoroughly washed with
ergy contributed by local water-nanotube and ion-nanotube atomic pairs ethanol, and dried. The nanopore size was measured to be 3.87 nm by
per volume in each 0.2 nm × 0.2 nm × 12.0 nm bin distributed in the the porosimetry method.[60] The contact angles of liquid droplets on the
x–y plane. The solid–liquid interaction energy density 𝛾 SL was evaluated treated silica gel surface were measured by a goniometer (Model 290,
by summing the potential energy between the nanopore and all confined Ramé–Hart Instrument Co.) (Figure S31, Supporting Information). The
electrolyte particles divided by the surface area of the nanopore. At least aqueous solutions used in the liquid-nanopore specimen included DI wa-
20 000 frames during 1.0 ns in the simulation were utilized to calculate ter, 13 wt% LiCl solution, and 26 wt% LiCl solution. In order to prepare the
both the ESL distribution and the 𝛾 SL value. liquid-nanopore specimen, 0.1 g of treated silica gel and 0.6 mL of aque-
Interaction Energy Decomposition: The interaction energy between dif- ous solution were injected into a stainless-steel testing cell. The specimen
ferent species of the water–ion network was computed by post-processing was placed in a vacuum (≈3 kPa) for 1 h to eliminate the air amount in the
the coordinate of the confined water molecules and ions. The Coulomb liquid-nanopore system before being completely sealed. The cross-section
interaction between charged species was approximated by pairwise inter- of the testing cell, A, was 126 mm2 and the length of the specimens, l, was
action potential to decompose the interaction between different species. 6.0 ± 0.1 mm.
A cutoff of 2 nm for Coulomb interaction was adopted to ensure an error Quasistatic Compression Experiment: The liquid-nanopore specimen
of less than 5% in the total potential energy compared to the original sys- sealed in the testing cell was compressed by an Instron machine (Model
tem modeled with the particle–particle–particle–mesh algorithm. At least 5982, Instron, Inc.) equipped with an environmental chamber (Model
10 000 frames of coordinate data during the equilibrium state of the sys- 3119–609, Instron, Inc.) at a temperature of 300, 325, 350, and 370 K.
tem were used to calculate the mean and uncertainty values. A constant loading speed of 2 mm min−1 was selected to ensure system
Analyses of Water–Ion Coordination Networks: A hydrogen bond equilibrium during the liquid infiltration–outflow process. The specimen
formed between two adjacent water molecules (water–water connection) displacement d and the applied force F gradually increased with the ex-
was identified by two geometric criteria associated with the molecular ternal loading. As all the nanopores were filled with liquid, the force re-
structure: i), the distance between the two oxygen atoms of the two sponse became linear, and a peak force was set to trigger the unloading
molecules rOO was less than 0.35 nm; ii), the angle between the oxygen– process. Once the peak force was reached, the upper compression platen
hydrogen covalent bond of one water molecule and the line determined by was moved back at the same rate. This quasistatic loading–unloading cycle
the two oxygen atoms of the two water molecules 𝜃 OOH was smaller than was consecutively repeated three times for each specimen. For each type
30°.[55] A water–ion or ion–ion connection was determined based on their of liquid-nanopore system, at least three specimens were tested. The sys-
pairwise separation. If an ion or water participates in the solvation shell of tem pressure built into the specimen was calculated as P = F/A, where A
another ion (RDFs in Figure S1, Supporting Information), the connection denotes the surface area of the sample, and the system volumetric change
could be identified. The visualization was performed with OVITO.[56] The was determined by ΔV/V = d/l. The PRR was calculated from the negative
probe sphere radius in the surface mesh visualization was set as 0.35 nm, slope of the P–ΔV/V0 curve, the same as that in MD simulations, and the
the same as rOO . MPD was defined as the difference between the system pressures at the
The autocorrelation function of a connection was calculated via S(𝜏) = midpoints of the liquid infiltration plateau and liquid outflow plateau in
⟨h(𝜏)h(0)⟩∕⟨h(0)2 ⟩, where h(𝜏) is one when the connection is up to time the P–ΔV/V0 curve.
𝜏.[57] Otherwise, h(𝜏) = 0. The correlation time was estimated as the value Fabrication of Cushion Pads and Dynamic Impact Test: The cushion
of 𝜏 when S(𝜏) declines to 1/e. pads were fabricated by sandwiching the liquid-nanopore system in be-
The potential of Mean Force Analyses: The potential of mean tween two flexible plastic films (Clear Vinyl Sheet 20 Ga, Marine Vinyl Fab-
force with respect to rOO and 𝜃 OOH was calculated via PMF = ric Inc.) with pre-molded shapes. The cushion pads with a diameter of
−kB Tg(rOO , 𝜃OOH )), where kB is the Boltzmann constant. The distribu- 46 mm and a height of 8 mm were then welded together using the radio-
tion function g(rOO ,𝜃 OOH ) was evaluated as the ratio of the average frequency (RF) welding technique. The incorporated liquid-nanopore

Adv. Mater. 2023, 35, 2303759 2303759 (9 of 11) © 2023 The Authors. Advanced Materials published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmat.de

system contained 2 g of nanoporous media and 6 g of water or 26 wt% LiCl [6] H. Wang, Z. L. Liu, J. Lao, S. Zhang, R. Abzalimov, T. Wang, X. Chen,
aqueous solution. A nanoporous media with a lower Pin was selected to Adv. Sci. 2022, 9, 2104697.
match the required working pressure of the cushion pad. The dynamic im- [7] M. Ma, F. Grey, L. Shen, M. Urbakh, S. Wu, J. Z. Liu, Y. Liu, Q. Zheng,
pact behavior of the cushion pads was characterized by a lab-customized Nat. Nanotechnol. 2015, 10, 692.
drop tower apparatus (inset in Figure 6d).[61] Before each test, the cushion [8] Y. Wu, H. Ling, Y. Qian, Y. Hu, B. Niu, X. Lin, X.-Y. Kong, L. Jiang, L.
pad was placed on the flat base of the drop tower. As the test was triggered, Wen, ACS Nano 2022, 16, 11092.
a 2.3 kg drop mass fell freely and impacted the cushion pad. The impact [9] K. Wu, Z. Chen, J. Li, X. Li, J. Xu, X. Dong, Proc. Natl. Acad. Sci. USA
speed was maintained at 3.0 m s−1 by adjusting the free drop height. An 2017, 114, 3358.
accelerometer (353B03, PCB Group, Inc.) was fixed to the drop mass to [10] A. Schlaich, J. Kappler, R. R. Netz, Nano Lett. 2017, 17, 5969.
measure the acceleration history, which was simultaneously recorded by a
[11] G. Hummer, J. C. Rasaiah, J. P. Noworyta, Nature 2001, 414, 188.
high-speed digitizer (PXIe-5105, NI Corp.) at the sampling rate of 106 sam-
[12] V. Kapil, C. Schran, A. Zen, J. Chen, C. J. Pickard, A. Michaelides, Na-
ples s−1 .
ture 2022, 609, 512.
[13] L. Fumagalli, A. Esfandiar, R. Fabregas, S. Hu, P. Ares, A. Janardanan,
Q. Yang, B. Radha, T. Taniguchi, K. Watanabe, Science 2018, 360, 1339.
Supporting Information [14] Q. Yang, P. Sun, L. Fumagalli, Y. Stebunov, S. Haigh, Z. Zhou, I.
Grigorieva, F. Wang, A. Geim, Nature 2020, 588, 250.
Supporting Information is available from the Wiley Online Library or from [15] R. C. Remsing, I. G. McKendry, D. R. Strongin, M. L. Klein, M. J. Zdilla,
the author. J. Phys. Chem. Lett. 2015, 6, 4804.
[16] Z. Li, R. P. Misra, Y. Li, Y.-C. Yao, S. Zhao, Y. Zhang, Y. Chen, D.
Blankschtein, A. Noy, Nat. Nanotechnol. 2023, 18, 177.
Acknowledgements [17] K. Zhou, Z. Xu, Nano Lett. 2020, 20, 8392.
[18] H. Li, J. S. Francisco, X. C. Zeng, Proc. Natl. Acad. Sci. USA 2015, 112,
Y.G. and M.L. contributed equally to this work. This research was sup- 10851.
ported by the National Science Foundation Directorate for Engineering Di- [19] N. Stolte, R. Hou, D. Pan, Nat. Commun. 2022, 13, 5932.
vision of Chemical, Bioengineering, Environmental and Transport Systems [20] L. Guillemot, T. Biben, A. Galarneau, G. Vigier, É. Charlaix, Proc. Natl.
(Grant #1805451). The computations were conducted by utilizing the Ex- Acad. Sci. USA 2012, 109, 19557.
treme Science and Engineering Discovery Environment (XSEDE) through [21] Y. Gao, M. Li, H. Zhang, Y. Zhang, W. Lu, B. Xu, Matter 2022, 5, 266.
allocation TG-MCH210002, which was supported by the National Science [22] Y. Gao, M. Li, Y. Zhang, W. Lu, B. Xu, Proc. Natl. Acad. Sci. USA 2020,
Foundation (grant number ACI-1548562).
117, 25246.
[23] V. P. Kurupath, S. K. Kannam, R. Hartkamp, S. P. Sathian, Desalination
2021, 505, 114978.
Conflict of Interest [24] M. Heiranian, A. B. Farimani, N. R. Aluru, Nat. Commun. 2015, 6,
8616.
The authors declare no conflict of interest. [25] Q. Lu, W. Shi, H. Yang, X. Wang, Adv. Mater. 2020, 32, 2001544.
[26] N. Liu, Y. Jiang, Y. Zhou, F. Xia, W. Guo, L. Jiang, Angew. Chem. 2013,
125, 2061.
[27] J. Zhang, L. Zhang, Z. Li, Q. Zhang, Y. Li, Y. Ying, Y. Fu, Small 2021,
Data Availability Statement 17, 2101665.
The data that support the findings of this study are available from the cor- [28] S. Ghosh, A. Sood, N. Kumar, Science 2003, 299, 1042.
responding author upon reasonable request. [29] M. S. Mauter, I. Zucker, F. Perreault, J. R. Werber, J.-H. Kim, M.
Elimelech, Nat. Sustainability 2018, 1, 166.
[30] Y. Wu, T. Zhou, Y. Wang, Y. Qian, W. Chen, C. Zhu, B. Niu, X.-Y. Kong,
Y. Zhao, X. Lin, Nano Energy 2022, 92, 106709.
Keywords [31] Y. Qiao, L. Liu, X. Chen, Nano Lett. 2009, 9, 984.
energy dissipation, hydrophobic nanopores, nanofluidics, outflow, water– [32] A. Siria, P. Poncharal, A.-L. Biance, R. Fulcrand, X. Blase, S. T. Purcell,
ion networks L. Bocquet, Nature 2013, 494, 455.
[33] Z. Zhang, L. Wen, L. Jiang, Nat. Rev. Mater. 2021, 6, 622.
[34] S. A. Roget, K. A. Carter-Fenk, M. D. Fayer, J. Am. Chem. Soc. 2022,
Received: April 22, 2023
Revised: July 3, 2023 144, 4233.
Published online: July 6, 2023 [35] J. Peng, D. Cao, Z. He, J. Guo, P. Hapala, R. Ma, B. Cheng, J. Chen,
W. J. Xie, X.-Z. Li, Nature 2018, 557, 701.
[36] R. Fuentes-Azcatl, M. C. Barbosa, J. Phys. Chem. B 2016, 120, 2460.
[37] Y. Marcus, Chem. Rev. 2009, 109, 1346.
[1] E. Secchi, S. Marbach, A. Niguès, D. Stein, A. Siria, L. Bocquet, Nature [38] R. Leberman, A. Soper, Nature 1995, 378, 364.
2016, 537, 210. [39] J. D. Smith, R. J. Saykally, P. L. Geissler, J. Am. Chem. Soc. 2007, 129,
[2] A. Schlaich, E. W. Knapp, R. R. Netz, Phys. Rev. Lett. 2016, 117, 048001. 13847.
[3] K. V. Agrawal, S. Shimizu, L. W. Drahushuk, D. Kilcoyne, M. S. Strano, [40] A. W. Omta, M. F. Kropman, S. Woutersen, H. Bakker, Science 2003,
Nat. Nanotechnol. 2017, 12, 267. 301, 347.
[4] L. Salvati Manni, S. Assenza, M. Duss, J. J. Vallooran, F. Juranyi, S. [41] M. Wang, S. Liu, H. Ji, T. Yang, T. Qian, C. Yan, Nat. Commun. 2021,
Jurt, O. Zerbe, E. M. Landau, R. Mezzenga, Nat. Nanotechnol. 2019, 12, 3198.
14, 609. [42] H. Chen, E. Ruckenstein, J. Phys. Chem. B 2015, 119, 12671.
[5] M. Neek-Amal, F. M. Peeters, I. V. Grigorieva, A. K. Geim, ACS Nano [43] J. Teychené, H. Roux-de Balmann, L. Maron, S. Galier, J. Mol. Liq.
2016, 10, 3685. 2019, 294, 111394.

Adv. Mater. 2023, 35, 2303759 2303759 (10 of 11) © 2023 The Authors. Advanced Materials published by Wiley-VCH GmbH
www.advancedsciencenews.com www.advmat.de

[44] M. Han, S. A. Rogers, R. M. Espinosa-Marzal, Langmuir 2022, 38, [53] S. Plimpton, J. Comput. Phys. 1995, 117, 1.
2961. [54] H. Kamberaj, R. Low, M. Neal, J. Chem. Phys. 2005, 122,
[45] C. Cheng, G. Jiang, G. P. Simon, J. Z. Liu, D. Li, Nat. Nanotechnol. 224114.
2018, 13, 685. [55] B. Xu, Y. Li, T. Park, X. Chen, J. Chem. Phys. 2011, 135, 144703.
[46] Y. Gao, M. Yin, H. Zhang, B. Xu, ACS Nano 2022, 16, 9420. [56] A. Stukowski, Modell. Simul. Mater. Sci. Eng. 2009, 18, 015012.
[47] M. Amabili, Y. Grosu, A. Giacomello, S. Meloni, A. Zaki, F. Bonilla, A. [57] S. K. Pattanayak, S. Chowdhuri, J. Mol. Liq. 2013, 186, 98.
Faik, C. M. Casciola, ACS Nano 2019, 13, 1728. [58] M. D. Gelenter, V. S. Mandala, M. J. Niesen, D. A. Sharon, A. J. Dregni,
[48] B. Xu, Y. Qiao, X. Chen, J. Mech. Phys. Solids 2014, 62, 194. A. P. Willard, M. Hong, Commun. Biol. 2021, 4, 338.
[49] X. Zhang, X. Li, H. Gao, Acta Mech. Sin. 2017, 33, 792. [59] M. Li, L. Xu, W. Lu, Langmuir 2020, 36, 4682.
[50] P. Mark, L. Nilsson, J. Phys. Chem. A 2001, 105, 9954. [60] A. Y. Fadeev, V. Eroshenko, J. Colloid Interface Sci. 1997, 187,
[51] D. J. Price, C. L. Brooks, J. Chem. Phys. 2004, 121, 10096. 275.
[52] M. B. Gee, N. R. Cox, Y. Jiao, N. Bentenitis, S. Weerasinghe, P. E. [61] M. Li, J. Li, S. Barbat, R. Baccouche, W. Lu, Compos. Struct. 2018,
Smith, J. Chem. Theory Comput. 2011, 7, 1369. 200, 120.

Adv. Mater. 2023, 35, 2303759 2303759 (11 of 11) © 2023 The Authors. Advanced Materials published by Wiley-VCH GmbH

You might also like