A nanofluidic ion regulation membrane with aligned cellulose nanofibers

See allHide authors and affiliations

Science Advances  22 Feb 2019:
Vol. 5, no. 2, eaau4238
DOI: 10.1126/sciadv.aau4238


The advancement of nanofluidic applications will require the identification of materials with high-conductivity nanoscale channels that can be readily obtained at massive scale. Inspired by the transpiration in mesostructured trees, we report a nanofluidic membrane consisting of densely packed cellulose nanofibers directly derived from wood. Numerous nanochannels are produced among an expansive array of one-dimensional cellulose nanofibers. The abundant functional groups of cellulose enable facile tuning of the surface charge density via chemical modification. The nanofiber-nanofiber spacing can also be tuned from ~2 to ~20 nm by structural engineering. The surface-charge-governed ionic transport region shows a high ionic conductivity plateau of ~2 mS cm−1 (up to 10 mM). The nanofluidic membrane also exhibits excellent mechanical flexibility, demonstrating stable performance even when the membrane is folded 150°. Combining the inherent advantages of cellulose, this novel class of membrane offers an environmentally responsible strategy for flexible and printable nanofluidic applications.


Ion-regulating nanofluidic membranes have been intensively used in desalination (1, 2), osmosis energy generation (3, 4), ion/molecular separation (57), and ionic circuits (812). Because of the interactions between solvated ions and the inner channel walls, ion transport in a nanoscale-confined, surface-charged membrane substantially differs from bulk behavior (1315). High ionic conductivity, which can be achieved through the nanofluidic effect, is essential for applications such as ionic circuitry and nanofluidic membranes. Typically, silicon-based materials are fabricated via lithography or templating to form aligned one-dimensional nanoscale channels for nanofluidic membranes, but the material is often brittle or suffers from performance degradation upon bending or folding (8, 16), rendering it challenging for flexible applications. Cost-effective nanoporous polymer membranes have also found great success in the industry, but these materials are not sustainable, and the three-dimensional tortuous nanoporous structure limits the ionic conductivity (17, 18). Two-dimensional materials, such as boron nitride, graphene, and MoS2, have shown excellent ionic conductivity and advantageous properties for use in nanofluidic devices, but these materials require time-consuming and expensive bottom-up fabrication methods (14, 1922). Consequently, advancing technologies that rely on efficient ion regulation necessitate the continued search for materials with high-performance nanoscale channels that can be obtained on a massive scale.

Cellulose is the most abundant and sustainable material in nature and is an attractive candidate for a wide range of applications, especially those related to eco-friendly membranes (23, 24) and fluidic devices (2530). Here, we demonstrate highly efficient and tunable ion regulation using a cellulose membrane that is composed of aligned nanochannels. These cellulose nanofibers are exposed after extraction of intertwined lignin and hemicellulose from the natural wood (Fig. 1, A to C) (31). Because of the dissociation of the surface functional groups (32), the charged cellulose nanofiber surface can attract layers of counter ions adjacent to the fibers, with an exponentially decaying ion concentration toward the center of the channel (Fig. 1D). The interface-dominated electrostatic field surrounding the cellulose nanofibers provides surface charge–governed ion transport along the fiber direction, enabling desirable ionic separation.

Fig. 1 Ion transport within the aligned cellulose nanofibers.

(A) A tree trunk containing cellulose nanofibers. (B) Schematic of the removal of intertwined lignin and hemicellulose from natural wood to make the nanofluidic membrane. (C) A nanofluidic membrane that inherits the nanofiber alignment direction from natural wood, as marked. (D) Schematic of the low tortuosity nanofluidic membrane. Photo Credit: T.L., University of Maryland, College Park. Permission granted.

The surface charge and geometry of the nanochannels can also be easily tuned to modify the ionic conductivity of the membrane. Owing to the abundance of the functional groups on the cellulose nanofibers, the surface charge density can be tuned via chemical stimuli. In this work, we demonstrate a high surface charge density of −5.7 mC m−2 after converting the hydroxyl groups to carboxyl groups, which was greater than previously reported values (8, 13, 20, 21, 33). In addition, we were able to attain large tunability of the channel size up to an order of magnitude. In this manner, a high surface charge–governed ionic conductivity of ~2 mS cm−2 was observed at a KCl concentration of less than 10−2 M.

Figure 2 (A and B) shows various configurations of the cellulose membranes after lignin and hemicellulose removal of the natural wood. When dried in air under ambient temperature, a densified structure can be obtained with closely packed cellulose nanofibers, potentially due to hydrogen bonding and van der Waals forces (34). Scanning electron microscopy (SEM) images in Fig. 2 (C and D) show the parallel-packed cellulose fibers aligned with the wood growth direction, while Fig. 2E shows a view of the cellulose nanofiber ends. The hydrophilic and highly aligned cellulose nanofibers render the membrane highly efficient in fluidic transport. To demonstrate, one end of a sample 2 cm by 1 cm by 200 μm in size was immersed into a water solution colored with ink to measure the liquid uptake. A fluidic rate of >1 mm s−1 was observed (fig. S1), twice higher than the previously demonstrated values obtained with randomly oriented cellulose fibers (0.5 mm s−1) (25, 35). Small-angle x-ray scattering (SAXS) verified the alignment of the cellulose membrane down to the molecular level, in which an elliptical pattern was observed for the cellulose membranes both without (dry) and with (wet) water (Fig. 2F). The dry membrane appears white, which is indicative of the complete removal of lignin (fig. S2) (31). Upon immersion in water, a broadband transmittance of >65% was obtained from 400 to 1100 nm (Fig. 2G), and letters of the text beneath the membrane became visible. The transparency change is mainly attributed to the matched optical refractive indexes of cellulose (1.48) (36) and water (typically ~1.50).

Fig. 2 Characterizations of the delignified wood.

Photos of various configurations of the (A) nanofluidic cellulose membrane, including (B) a cellulose cable wrapped around a rod. The arrows indicate the nanofiber alignment direction. Scale bars, 1 cm. (C) Side view SEM image of the cellulose membrane and (D) the aligned cellulose nanofibers at higher magnification. (E) Top view SEM image showing the tips of the cellulose nanofibers. (F) The elliptical shape of the diffraction pattern in SAXS for the dry and wet membrane, indicating the molecular level alignment of cellulose. (G) The transmittance of the dry and wet cellulose membrane.

We investigated the nanofluidic performance of the cellulose membrane using the ionic conductivity setup shown in Fig. 3A (see Methods). Carboxyl groups have a greater tendency to dissociate into negatively charged carboxylates (32), leading to a higher charge density and therefore a higher negative zeta potential in deionized water. Therefore, using a previously described method (37), we applied TEMPO (2,2,6,6-tetramethylpiperidine-1-oxyl) oxidation to the cellulose to convert the primary hydroxyl groups on the surface chains of the cellulose crystallites into carboxyl groups (37), as illustrated in Fig. 3B. The resulting oxidized cellulose membrane exhibited a higher zeta potential of −78 mV, compared with −45 mV for just delignified cellulose (Fig. 3b). The respective ionic conductivity of these materials in KCl solutions is shown in Fig. 3C. The ion transport behavior in both cellulose membranes exhibited a conductivity plateau orders of magnitude higher than that of the bulk solution for concentrations below ~10−2 M. Within the surface-governed ion transport region, a conductivity as high as ~2 mS cm−1 was obtained for the oxidized membrane compared with 1.1 mS cm−1 for the unmodified counterpart, indicating the effectiveness of modifying the surface functional groups to tune the ion transport behavior. Using Eq. 1, we estimated the surface charge of the membranes based on the zeta potentialEmbedded Image(1)in which σ is the surface charge, ε is the dielectric constant, ε0 is the permittivity of vacuum, ζ is the zeta potential, and λd is the Debye length, which was −3.2 and −5.7 mC m−2 for the as-made cellulose and oxidized cellulose, respectively (38). With the estimate of the surface charge for these samples, the overall conductivity trend can be fitted using the following equationEmbedded Image(2)in which Z is the cation valence, μ+ is the cation mobility, μ is the anion mobility, C is the ion concentration, NA is the Avogadro’s number, and h is the channel diameter of the nanofluidic cellulose system (39). As shown in Fig. 3C, the fit of Eq. 2 agrees well with the experimental data of the ionic conductivity versus KCl concentration and allows us to calculate a channel diameter of ~2 nm for both cellulose membranes. We attribute the difference in the value of the ionic conductivity plateaus for each membrane to the difference of the surface charge densities between the materials.

Fig. 3 Ionic conductivity measurement with chemical modifications and physical densifications.

(A) Ionic conductivity measurement setup. (B) Zeta potential of the cellulose fibers and oxidized/surface-charged cellulose under neutral pH with a concentration of cellulose approximately 0.1%. (C) An ionic conductivity test with KCl solution for the cellulose membrane before and after oxidization. The oxidized cellulose exhibits an increased ionic conductivity plateau due to the higher surface charge. (D) Ionic conductivity of the undensified cellulose and densified cellulose membrane in KCl solution.

We further demonstrated the effect of channel geometry on the ionic conductivity by preparing an undensified cellulose membrane (see Methods) and comparing its performance with the densified sample. The undensified samples exhibited an ionic conductivity plateau of 0.2 mS cm−1, one order lower than that of the densified sample (Fig. 3D). The fit of the undensified cellulose membrane conductivity results indicated a channel diameter of around 20 nm, which is about 10 times larger than that of the densified cellulose membrane (channel diameter ~2 nm). The mechanical properties of the densified and the undensified samples are shown in fig. S3. We obtained the tensile strengths of 58 ± 5 MPa and 9 ± 2 MPa for the densified and the undensified samples, respectively.

To explore the use of this cellulose nanofiber membrane as an ion regulation device, we demonstrated the ionic rectification effect of the material acting as a flexible transistor with electrical gating, in which the cellulose membrane can preferentially accumulate ions that have the opposite charge as the channel walls. Silver paste was painted on the membrane to act as the gating metal (Fig. 4A), and a 10−6 M KCl solution was used as the liquid electrolyte. The gating voltage was controlled by a Keithley 2400 power source, while the ionic current–voltage characteristics were recorded. When the gating voltage was negative, the local concentration of K+ should further increase under the gate, which will contribute to a large cationic current density. Meanwhile, positive gating will repel K+ and lead to an even lower current density than the neutral gating condition (Fig. 4B). Figure 4C shows the ionic currents measured under different gating potentials from −2 to 2 V. The ion conductivity under Vg = −2 V was about one order of magnitude higher than the value under Vg = 2 V and equivalent to that of 10−2 M KCl, indicating an efficient accumulation of positive ions with negative gating (Fig. 4D). The device exhibits a negligible electrical gate leakage current, which was measured to be below the noise floor of the Keithley 2400.

Fig. 4 A cellulose-based ionic transistor.

(A) Schematic of a freestanding cellulose nanofluidic transistor with a painted metal contact for gating. (B) Schematic of the gating effect on the ion distribution within the cellulose membrane. (C) Current-voltage characteristics of the cellulose nanofiber membrane with different gating voltages from −2 to 2 V. (D) Characterization of the transistor using varied gating voltages. Inset: Semi-log plot of ionic current versus gating voltage. (E) Photo image of a flexible and biocompatible cellulose nanofiber ribbon. (F) Ion conductivity shows minimal change upon folding.

The cellulose membrane is flexible and even foldable. Figure 4E shows a ribbon of the membrane that can be twisted and wrapped around a finger. To observe how folding affects the ionic conductivity performance, we used a membrane 2 cm by 2 mm by 1 mm in size and recorded the current under an applied voltage of 0.5 V as we folded the material. The ionic conductivity under a concentration of 10−6 M KCl exhibited minimal changes upon folding, with no notable performance degradation for a folding angle of up to 150° (Fig. 4F). We have also evaluated the membrane’s stability in conductivity and mechanical strength (figs. S4 and S5). After several wet-dry cycles, the membrane shows no notable degradation of the performance.


In this work, we report a permselective nanofluidic membrane featuring aligned cellulose nanofibers directly derived from wood. The removal of lignin and hemicellulose introduces numerous open nanochannels among the cellulose nanofibers. Via structural engineering, the diameter of the nanochannels can be tuned from 2 to 20 nm. A high surface charge density of −5.7 mC m−2 was obtained after chemical modification of the cellulose functional groups. The long-range ordered arrays of the charged nanofibers led to highly efficient ion transport in the centimeter-long devices. A high ionic conductivity plateau of ~2 mS cm−1 was obtained, which is orders of magnitude higher than that of the bulk KCl solution at the same concentrations. In addition, we used electrostatic gating to further increase the ionic conductivity of 10−6 M KCl electrolyte to a value that is equivalent to the ionic conductivity of 10−2 M KCl. This work demonstrates the use of wood-derived cellulose membranes toward foldable, scalable, and high-performance nanofluidic devices.


Materials and chemicals

American Basswood was purchased from Walnut Hollow Company. Sodium hydroxide (NaOH) and sodium sulfite (Na2SO3) (both purchased from Carolina Biological Supply Company) and hydrogen peroxide (H2O2, 30% solution; EMD Millipore Corporation) were used for lignin extraction of the Basswood. The deionized water (type I deionized water; purity, >18 megohm·cm) from ChemWorld was used to further remove the residual ions from the delignified wood.

Cellulose membrane preparation

Basswood was cut along the growth direction at various dimensions (typical thickness, <5 mm; length/width, <10 cm). NaOH and Na2SO3 were dissolved in deionized water at concentrations of 2.5 and 0.4 M, respectively. The wood slices were boiled in the solution for 10 hours. Then, the wood slices were immersed in boiling H2O2 solution (30%) until completely white. The resulting wood membrane was then rinsed in deionized water to remove the residual ions and chemicals. The effective removal of residual chemicals was verified by obtaining a low ionic conductivity of the solution (<20 μS cm−1) that had been used to wash the sample. To make the undensified membrane, the sample was subsequently treated by freeze drying. To make a densified membrane, the sample was dried in air under ambient temperature while being pressed for 5 hours. For typical TEMPO oxidation, the cellulose membrane (1 g) was soaked in 100 ml of water containing 0.016 g of TEMPO and 0.1 g of NaBr, with the pH value being adjusted to 10 by adding 0.5 M NaOH solution. Approximately 5 mmol NaClO was added to the solution to initiate the oxidation reaction. The pH value of the solution was closely monitored using a pH meter and was maintained at 10 by continuously adding 0.5 M NaOH. The reaction was terminated after 2.5 hours by adding 0.5 M HCl to lower the pH to 7. The membrane was then taken out of the solution and washed with 0.5 M HCl. Last, the cellulose membrane was solvent exchanged to acetone and then toluene and allowed to sit in air to evaporate all solvent.


An SEM (Hitachi SU-70) was used to characterize the morphologies of the samples. Compositional analysis of the cellulose membrane and original wood was carried out using high-performance liquid chromatography (Ultimate 3000; Thermo Fisher Scientific, USA). The optical transmittance from 400 to 1100 nm was characterized by ultraviolet-visible absorption spectroscopy (Lambda 35; PerkinElmer, USA) with an integrating sphere to collect all transmitted light. To determine the zeta potential of the original and TEMPO-oxidized white wood, the samples were dispersed in deionized water by ultrasonication treatment (100% amplitude, 10 min, Fisher Scientific FS 110D) to produce a 0.1 wt % (weight %) concentration. Zeta potential (ζ) of the dispersed suspension was measured using a Zetasizer Nano S90 (Malvern Instrument) without adjusting the ionic strength. The ζ value was calculated from the electrophoretic mobility using the Henry equation and Huckel approximation. The tensile strength of the wood was measured with a TA Q800 DMA system in controlled-force mode. The force increased from 1 up to 18 N at 0.01 N/min.

Current voltage measurement

A cellulose membrane was embedded into polydimethylsiloxane elastomer with two wells carved on either side to serve as reservoirs for the electrolyte solution. The cellulose nanofluidic membrane was immersed in electrolyte for at least 1 day before measurement. Two homemade Ag/AgCl electrodes were inserted into the reservoirs. BioLogic from Science Instrument was used to record the current signal while providing the voltage source at a scanning rate of 10 mV s−1.


Supplementary material for this article is available at

Fig. S1. Highly hydrophilic cellulose demonstrating directional fluidic transport.

Fig. S2. Composition results of the natural wood and the wood membrane after step I and II treatments.

Fig. S3. The mechanical tensile strength for the densified and undensified samples.

Fig. S4. Current-voltage characteristics of the membrane infiltrated multiple times with 10−5 M KCl solution.

Fig. S5. Tensile strength of the membranes that underwent different dry-wet cycles.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We acknowledge the support of the Maryland NanoCenter and its AIMLab. We acknowledge the help from J. Y. Zhu at the U.S. Forest Products Lab on the compositional analysis of original and delignified wood. Funding: This work was supported as part of the Nanostructures for Electrical Energy Storage (NEES), an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, and Basic Energy Sciences under award number DESC0001160. Author contributions: T.L. and L.H. designed the experiments. T.L., S.X.L., Z.S., and M.R. carried out the data analysis of ionic conductivity. T.L., W.K., C.C., E.H., C.J., and J.D. carried out the performance measurements and material characterization. X.Z. and R.B. carried out the mechanical tests and structural characterization. All authors contributed to the writing of the manuscript. Competing interests: T.L. and L.H. are the inventors on a patent currently pending with the University of Maryland (no. 62/789325, filed 7 January 2019). All the other authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.
View Abstract

Navigate This Article