Dictating anisotropic electric conductivity of a transparent copper nanowire coating by the surface structure of wood

Abstract

In this article, a robust, air-stable, flexible and transparent copper (Cu) nanowire (NW) network coating on the surface of the wood is presented, based on a fusion welding of the Cu NWs by photonic curing. Thereby, an anisotropic conductivity can be achieved, which is originating from the structural organization of the wood body and its surface. Furthermore, the Cu NWs are protected from oxidation or wear by a commercially available paraffin wax—polyolefin, which also results in surface water repellency. The developed processing steps present a facile and flexible routine for applying Cu NW transparent conductors to abundant biomaterials and solve current manufacturing obstacles for corrosion-resistant circuits while keeping the natural appearance of the substrate. It may open a venue for more extensive utilization of materials from renewable resources such as wood for electronic devices in smart buildings or mobility applications.

1. Introduction

Transparent conductors made of metal nanowire (NW) network are emerging as a scalable but low-cost alternative to conventional indium tin oxide (ITO). Such conductors may find application in next-generation bendable monitors [1], flexible touchscreen technologies [2] and electronic device integration [3,4]. The fundamental principle of the combination of transparency and conductivity is the assembly of NWs into an interlaced network for electron transport with a spacing that allows for the transmission of light [5]. Compared to the traditional transparent semiconductor, the metallic NW network has the advantage of unifying flexibility and full-spectrum transparency. Copper (Cu) [6,7] and Ag [8–10] NWs are state-of-the-art materials for the fabrication of next-generation flexible transparent conductors. However, one challenge in the fabrication of transparent metal NW conductors is to form a stable junction between two NWs for a long-distance electron transport on thermally sensitive soft materials, as high-temperature annealing under vacuum or in a protecting atmosphere is typically used processes [5]. Xiong et al. [11] used an electroless-welding method by dipping the Ag NW network-coated PET film into the solution of silver ammonia and glucose to reduce the contact resistance between Ag NWs. A promising approach was presented by Garnett et al. [12], who used a light-induced plasmonic nanowelding technique to join the NWs together without damaging the thermally sensitive substrates. A combination of flashlight and UV-C was also used to weld Ag NWs coating on the surface of PET [13]. Compared with silver, copper is 1000 times more abundant. However, the practical application of Cu NW networks is severely restricted because of the fast degradation of Cu NWs under ambient conditions. Therefore, a current research focus is on improving the resistance of Cu NWs against oxidation [14,15]. Rathmell et al. [16] coated Cu NWs with a nickel shell, which increased the stability of these cupronickel NWs in transparent conducting films. More recently, a similar concept was applied by Niu et al. [17], who coated the surface of Cu NW with a few atomic layers of gold via epitaxial overgrowth. Xu et al. [18] deposited a graphene layer on the top of Cu NW networks via chemical vapour deposition, which resulted in Cu NW networks with antioxidant properties. However, the chemical vapour deposition can only be carried out on thermostable substrates. Hence, a simple strategy that allows for fusion weld the NWs under mild conditions and improves the air and moisture stability of the Cu NW networks is needed. The flexible transparent conducting electrodes made of Cu NWs hold potentials for the application in various electronic devices. For instance, Zhong et al. [19] used intense pulsed light and roll-to-roll wiping processes to prepare patterned Cu NW electrodes, which were successfully implemented into a phosphorescent organic light-emitting diode and a flexible transparent heater.

Future bio-economies require the utilization of renewable biomaterials for the design of multi-functional products [20]. Particular attention is paid to manufacturing electric conductive materials from biomass for developing advanced material concepts and reducing energy consumption [21,22]. For this purpose, obtaining sufficient surface conductivity is the crucial step for the successful utilization of biomass [23–26]. Jung et al. [27] used cellulose nanofibril paper as a substrate for the decoration with electrical components, which were biodegradable by fungi. Takahashi et al. [28] printed flexible electrodes for touch sensor devices onto films consisting of clay and wood components cross-linked by polymers. However, there are still fundamental challenges that need to be addressed: firstly, how to decorate the surface of temperature-sensitive biomaterials with conductive metallic NW thin films and secondly, how to use the unique surface topography of the biomaterials to affect the conductive behaviour. This particularly applies to the wood with its anisotropic properties and high variability in structural and chemical composition.

Here, we report on an electric conductivity of a transparent Cu NW coating dictated by the surface structure of wood. The development of anisotropic conductivity on the wood surface will allow for making advanced electronic devices governed by the existing natural resource. Compared with commonly used lithography or magnetic sputtering with a mask, the process reported in this study is energy saving and low cost. By photonic curing, only the NW network is heated in less than a millisecond, while the substrate is not, which makes the process very well suited for thermally sensitive substrates such as paper and wood, which can provide mechanical flexibility compared to the typically used conductive oxides such as ITO.

2. Material and methods

2.1. Materials

Nickel acetylacetonate (Ni(acac)₂, 95%) and Oleylamine (OAm, approximate C18 content 80–90%) were purchased from ACROS Organics; copper (II) chloride dehydrate (CuCl₂·2H₂O, ≥99%), hexane (anhydrous, greater than 95%), tetrahydrofuran (THF, ACS reagent, ≥99%) and toluene (anhydrous, 99.8%) were purchased from Sigma-Aldrich. All chemicals were used as received without further purification. Whatman cellulose acetate membranes with an average pore size of 0.45 µm and a diameter of 47 mm were used for Cu NW network formation. Larch (Larix sp.) wood was cut into a dimension of 50 mm × 50 mm × 5 mm in longitudinal × radical × tangential directions.

2.2. Synthesis of copper nanowires

Cu NWs were synthesized according to the previous publication by Guo et al. [29]. In a typical synthesis, CuCl₂·2H₂O (0.8 mmol) and Ni(acac)₂ (0.4 mmol) were dissolved in OAm in a three-neck round-bottom flask connected to the Schlenk flask by magnetic stirring at 80°C with nitrogen flowing. After 15 min, the resulted clear blue solution was heated up to 180°C and kept at this temperature for 3 h under stirring. The reddish solution was then cooled down to room temperature. The hexane solution was added followed by ultrasonic treatment for 15 min. The Cu NWs were collected by centrifugation (9000 r.p.m., 5 min) and repeatedly washed with toluene using dispersion–centrifugation cycles to remove the excess OAm. The Cu NWs were finally dispersed in 50 ml of toluene. The concentration of Cu NWs was calculated to be approximately 5 mg ml⁻¹.

2.3. Mechanical and electrical measurement

The electronic and mechanical properties of an individual Cu NW were measured in an FEI Quanta 600 FFG scanning electron microscope (SEM) with two nanomanipulators (Kleindiek MM3A) installed in the microscope sample chamber. Each nanomanipulator is equipped with a chemically etched tungsten (W) tip with a diameter of approximately 100 nm. A fibre composed of Cu NWs was attached to a fine platinum wire. A W tip was first manipulated to attach to the Cu NWs on the edge of the platinum wire and then pulled out. The process was repeated until an individual NW could be picked up. The selected Cu NW was connected to this W tip through electron beam-induced deposition (EBID) of amorphous carbon. Then, another W tip is manipulated to approach this Cu NW and a frequency tuneable AC voltage is applied across the W tip and Cu NW. The forced resonance of this Cu NW is then monitored on the SEM screen. AC voltage is a time (t)-dependent voltage (V_(t) = V_d cos(ωt)), which caused a time-dependent force and dynamic deflections. Tuning the angular frequency ω = 2πf allowed the NW to be resonantly excited, which caused large-amplitude deflections. The vibration amplitude could be observed from SEM. According to Poncharal et al. [30], the frequencies are found from the following equation in the condition of carbon nanotube:

This equation results from the Bernoulli–Euler analysis of cantilevered elastic beams [32], which has been proved to be eligible for estimating the effective bending modulus of carbon nanotube [31] and ZnO NW [33]. In this equation, d is the outer diameter, d_i is the inner diameter, E_b is the elastic modulus, ρ is the density and β is a constant equal to 1.875. E_b is also called effective bending modulus in this experiment, which is an intrinsic characteristic of the measured material. In solid Cu NW, d_i is equal to zero. As a result, E_b can be calculated as the following equation [31]:

After that, the freestanding NW was fixed between the two probes, and was further adjusted to be horizontally oriented by changing the relative positions of the two probes, which allowed for measuring the length and diameter of the Cu NW through SEM imaging.

The current density was measured between the two probes by applying an electrical field scanning by using a Keithley 4200 semiconductor characterization system. The resistance of the Cu NW between the two W tips as shown in figure 1d is 473 Ω based on the measurement resulted shown in figure 1f. The resistivity (μ) of the Cu NW is calculated to be 2.6 × 10⁻⁷ Ω m according to the following equation:

Figure 1.

(a) The SEM image of the Cu NWs. (b) SAED pattern acquired from an individual Cu NW and the TEM image of this Cu NW is shown in (c). (d) SEM image of the electric-field-induced resonance of a Cu NW. (e) SEM image of electrical measurement of a Cu NW. (f) The measured I–V curve for the Cu NW in (e). (g) The I–V curve of the Cu NW after the formation of ohmic contact.

2.4. Fabrication of transparent conductive film

Different amounts of Cu NW suspensions (in the range from 0.1 to 1.0 ml) were diluted by 50 ml of toluene via sonication. The Cu NWs were assembled into a thin film by filtering the diluted suspension onto a cellulose acetate membrane under vacuum. The membrane with the thin Cu NW film was then placed onto a water surface. The thin Cu NW film was separated from the membrane and floated on the water. The freestanding thin Cu NW film was then deposited on the wood veneer. Rapid photonic curing with the pulsed light from xenon flash lamp was used to improve the junctions between the Cu NWs, while limiting the thermal loading of the substrate. The photonic curing was performed under nitrogen atmosphere with a voltage of 900 v for 80 µs.

2.5. Air stability and water-repellent coating

Parafilm (0.25 g, PM-996, 4 inch × 125 feet Roll, Bemis North America) was dissolved in the solution of THF (20 g) by sonication at approximately 70°C for 20 min. The clear solution (1 ml) was then drop casted onto the wood surface with or without Cu NW coating by drying at ambient atmosphere for 24 h.

2.6. Stability test

The conductive thin films made of 0.8 ml Cu NW suspension, which were coated onto the wood surface, were used for the air and thermal stability tests. Samples with or without Parafilm coating were placed in an oven at a temperature of 130°C. The resistance was measured every 30 min. Mechanical stability was tested on a sample coated with a 0.4 ml Cu NW suspension. The sample with a surface area of 3 × 2 cm² was faced to a 300-mesh sandpaper, loaded with a mass of 100 g. The mechanical resistance was recorded every 10 cm of abrasion.

2.7. Flexibility test

The Cu NW network was deposited onto the surface of a wood veneer with a thickness of 0.2 mm for the flexibility test. The conductive veneer was bent following the curvature of a glass tube with different radii, 3.95 cm, 2.90 cm, 2.15 cm, 1.20 cm and 0.5 cm, respectively. The electric resistance was measured on the sample in the curved condition.

2.8. Further characterization

Two lines of silver conductive paint (Structure Probe, Inc.) with a distance of 2 cm were applied on the surface of Cu NW thin film as the electrodes. For the samples with the water-repellent coating, the silver electrodes were applied before the drop casting of the Parafilm THF solution. The current–voltage (I–V) characteristic measurements were carried out on the probe station by pressing two metal needles on each electrode. Static contact angles were measured on a Drop Shape Analyzer-DSA100 (Krüss GmbH, Germany) at ambient conditions. Precisely, 8 µl of deionized water was placed onto the wood surface, and then photographs were taken every 5 min. A voltage between −0.75 and 0.85 V was applied by a Keithley617 system. The microstructure of the Cu NW thin film was analysed by SEM (FEI Quanta 200F). The transparency of the thin Cu NW film was indirectly measured by the colour change of the wood with and without the Cu NW film coating. The samples were tested with a colorimeter. This device measures or compares colours based on the CIE 1976 L*a*b* colour space. In one measurement, the instrument generates three indexes: the brightness L, the relation of red to green a and the relation of yellow to blue b. The total colour change can be calculated by the following equation: . Each sample was measured three times on the back of the wood veneer and three times on the front side with the Cu NWs. Afterwards, the ΔE values could be calculated for the different concentrations of Cu NWs.

3. Results and discussion

The as-synthesized Cu NWs were dispersed in 50 ml of toluene solution by sonication for coating preparation. Figure 1a displays the SEM image of the Cu NWs deposited onto a Si substrate, where pure NW with an average diameter of approximately 50 nm and a length in the range of 5–10 µm can be observed. Some NWs are curved, indicating the flexibility of these Cu NWs. Figure 1b displays the selective area electron diffraction (SAED) acquired from an individual Cu NW, from which superposition of the diffraction spots is observed. The green, blue, yellow, red and purple sketched diffraction patterns correspond to the [112], [001], [101], [011] and [114] zone axes of face-centred cubic system, indicating a twinned structure. The transmission electron microscopy (TEM) image of this Cu NW is shown in figure 1c, from which the diameter of the Cu NW is measured to be approximately 53 nm. The diffraction pattern together with the corresponding TEM image indicates that the Cu NW grows along direction.

The effective bending modulus of the NW is determined by the electric-field-induced resonance method [20,31]. A W tip is first manipulated to pick up an individual Cu NW from the edge of a platinum wire, where many Cu NWs have been attached. The selected Cu NW was connected to this W tip through EBID of amorphous carbon. Then, another W tip is manipulated to approach this Cu NW and a frequency tuneable AC voltage is applied across the W tip and Cu NW (figure 1d). The forced resonance of this Cu NW is determined by in situ SEM to be at a frequency of ∼3305 kHz. This equation results from the Bernoulli–Euler analysis of cantilevered elastic beams [32], which has been proved to be eligible for estimating the effective bending modulus of carbon nanotube [31] and ZnO NW [33]. After measuring the resonant frequency, the Cu NW was horizontally oriented with the help of another W tip. Thus, the length (∼2.3 µm) and diameter of this Cu NW (∼53 nm) can be accurately measured through SEM images, and its bending modulus is then calculated to be approximately 49.8 GPa (electronic supplementary material). Compared with bulk Cu with a Young's modulus in the range of 110–128 GPa, our measured Cu NWs are softer owing to the twinned structure [29]. Simulations reported by Li et al. [34] indicated that the twin-boundary spacing resulted in softening in nano-twinned copper. The nanoparticle attached to the Cu NW, which can be observed in figure 1d, may also affect the obtained bending modulus. The effective bending modulus controls the capacity of NW for deformation, which implies their application as the building block of flexible electrodes that follow the unique topography of the substrate.

By connecting Cu NWs between two W tips through EBID (figure 1e), the conductivity is successfully measured with a Keithley 4200 semiconductor characterization system. In the initial stage, a nonlinear current–voltage (I–V) curve with tiny current was measured (figure 1f). An ohmic contact was then formed, leading to a linear I–V curve. The organic ligands on the surface of Cu NW and possible thin tungsten oxide layer on the W probes can be cleared by Joule heating during the I–V measurement, which may lead to the formation of ohmic contacts between W probes and the NW (figure 1f). Therefore, the resistivity of the Cu NW was calculated to be 2.6 × 10⁻⁷ Ω m⁻¹, which indicates that the Cu NW has a comparable resistivity to that of bulk Cu (1.71 × 10⁻⁸ Ω m⁻¹ at approximately 300 K).

To assemble the Cu NWs in a continuous thin film, the Cu NW suspension was firstly diluted by n-hexane and then filtered through a cellulose acetate membrane by vacuum (figure 2a). The membrane with the thin Cu NW film on top was then placed on a water surface. After several minutes, the Cu NW thin film separated from the membrane, which swelled after absorbing water. The thin Cu NW film floating on water surface was then collected by the targeted substrate, wood in this work. Photonic curing using pulsed light from a xenon flash lamp was applied to weld the interlaced Cu NWs to achieve a conductive network. Since the pulsed light generates a sufficiently high temperature only for milliseconds on the surface, the method can be even applied to thermal sensitive substrates, which allows for substituting traditional glass or ceramic substrates.

Figure 2.

(a) Schematic illustration of the fabrication process of conductive and hydrophobic wood surface. (b) SEM image of the Cu NW network-coated wood surface. The blue arrow points to a concave lumen channel and the red arrow points to the ridge formed by the cell walls. (c) High-resolution SEM acquired from the ridge of the cut-opened wood cell wall with Cu NW network coating.

Wood is composed of elongated cells. The surface of wood consists of concave lumen channels and ridges formed by the (double) cell walls (electronic supplementary material, figure S1). The SEM image shows a unique Cu NW network pattern directed by the wood surface topography (figure 2b; electronic supplementary material, figure S2), which was further approved by X-ray diffraction measurements (electronic supplementary material, figure S3). The NW network coated the ridges (wood cell walls) from the cut-open cells, while the valleys (cut-open lumen) were hardly coated. Electronic supplementary material, figure S4 schematically illustrates the mechanism for the selective deposition of the Cu NW network on the ridges. Cu NWs have a weak interaction with each other before photonic curing because the organic surfactant covers the NW surface. The Cu NW network floating on the water surface is rather loose (electronic supplementary material, figure S4a) and will deform according to the topography of wood when deposited onto the wood surface. It will break in the middle above the valleys during the deformation because there is no support underneath (electronic supplementary material, figure S4b). As a result, the Cu NW network does not cover the bottom of the valleys (electronic supplementary material, figure S4c). Figure 2c shows a high-magnification SEM image acquired from a cell wall that is coated with Cu NWs. It demonstrates that the NWs interlace with each other to form a network for electron transport, while there is plenty of space between the NWs allowing for light transmission. As a result, the aesthetic appearance of wood can well be preserved (electronic supplementary material, figures S5 and S6).

To study the electronic properties of the Cu NW network-coated wood surface, the sheet resistance was measured in different orientations on the wood surface coated with different amounts of Cu NWs. As shown in figure 3a, a high conductivity was achieved along the longitudinal direction (figure 3b). A sheet resistance of 95.3 Ω cm⁻² was observed when the Cu NW network was made of 0.4 ml Cu NW suspension. The sheet resistance decreased to 3.4 Ω cm⁻² when the amount of Cu NW suspension used to prepare Cu NW network was increased to 1 ml. A comparison of the resistances of the samples treated by photonic curing and oven treatment indicates that the photonic curing results in higher conductivity (electronic supplementary material, table S1). Furthermore, the photonic curing has the advantages of a short treatment time and prevention of thermal damage of the substrate. To demonstrate the high conductivity along the longitudinal direction, a LED lamp was illuminated by applying an electronic field in parallel to the longitudinal direction of wood, as shown in the inset of figure 3a. However, the wood surface is almost insulated along the radial direction. The current density was in the range of 10 × 10⁻⁶ mA cm⁻² (figure 3c), once the electrode was drawn in parallel to the longitudinal direction of wood as shown in figure 3d. This anisotropic conductivity of the flexible Cu NW network was dictated by the unique wood surface topography, as the non-coated lumen areas act as strong resistance (figure 2b). It should be noted that also the measured sheet resistance along the longitudinal direction is larger than that of a Cu NW network deposited on a flat substrate such as glass, because the connectivity of the Cu NW network is dictated by the rough and inhomogeneous wood topography.

Figure 3.

(a) The conductivity measured along the longitudinal direction (b) of the wood surface coated with Cu NW network made of 0.4, 0.5, 0.7, 0.8, 0.9 and 1.0 ml of Cu NW suspension. (c) The conductivity measured perpendicular to the longitudinal direction (d) of the wood coated with Cu NW network made of 0.4, 0.6, 0.8, 0.9 and 1.0 ml of Cu NW suspension.

The deterioration of Cu NW networks upon oxidation after exposing to the ambient air is a primary limiting factor for their durability. In a simple approach, we produced a hydrophobic coating from Parafilm, a paraffin wax–polyolefin thermoplastic blend, which is widely used in the laboratories as a stretchable vapour or a gas barrier [35]. It has already been used as a hydrophobic layer for the rapid fabrication of paper-based microfluidic devices [36]. The Parafilm dissolved in THF was dropped onto the wood surface as a transparent, antioxidation and hydrophobic coating as indicated in figure 2a and as shown in electronic supplementary material, figure S7. The Parafilm solution can form a continuous coating on the whole wood surface after the evaporation of the THF (electronic supplementary material, figure S8). This hydrophobic layer does not only decrease the problematic liquid water uptake [37], but also protects the Cu NW network from oxidation or mechanical damage. Figure 4a displays the contact angle of a water drop on the Cu NW-coated wood surface with an additional layer of Parafilm. The water drop is not absorbed by the wood substrate, and the contact angle is higher than 100° throughout the measurement, which indicates that the modified wood surface is hydrophobic. The small decrease in the contact angle is mainly owing to the evaporation of water.

Figure 4.

(a) The contact angle of a water drop on the surface of Cu NW network and Parafilm-coated wood surface as a function of time. (b) Sheet resistance versus the bending curvature on the Cu NW network and Parafilm-coated wood veneer (the inset schematically illustrates the bending direction). (c) The air stability of the Cu NW network-coated wood surface with and without the protection of Parafilm. (d) The sheet resistance versus abrasion distance on the Cu NW network-coated wood surface with and without Parafilm protection. Please note that the difference in initial sheet resistances among deformation, heat stability and abrasion tests is caused by different amounts of Cu NWs in a unit area and the different wood substrate surfaces.

To study the effect of deformation on the conductivity of the Cu NW network, it was deposited onto the surface of a wood veneer with a thickness of 0.2 mm, which resulted in initial sheet resistance of 275 Ω cm⁻² without any bending (figure 4b). The sheet resistance decreased to 255 Ω cm⁻² as the wood veneer was bent to a curvature of 0.25 cm⁻¹. This may owe to a reorientation of the flexible Cu NW network caused by bending, which increased the connection among the Cu NWs. However, as the curvature increased to 0.83 cm⁻¹, the sheet resistance again raised to 267 Ω cm⁻², because of the separation-taking place at the contact points between the Cu NWs. This sheet resistance remained even when the veneer was flat again. When a bending curvature of 2.0 cm⁻¹ was applied to the veneer, contact points between Cu NWs were fully broken and the sheet resistance of the Cu NW network increased significantly to 9100 Ω cm⁻².

The stability of the conductivity was accessed via an accelerated treatment by putting the Cu NW network-coated wood slices with and without a Parafilm layer in an oven at a temperature of 130°C. As can be seen in figure 4c, the sheet resistance increased gradually from approximately 15 Ω cm⁻² to approximately 20 Ω cm⁻² for both specimens during the first 3 h of ageing. A notable difference was observed afterwards. The specimen without Parafilm coating exhibited a sheet resistance of approximately 85 Ω cm⁻² after 6 h treatment in the oven, while the one with the Parafilm coating still maintained a stable sheet resistance at approximately 25 Ω cm⁻². To study the stability of the conductive wood surface against mechanical damage, abrasion tests were performed on the conductive wood surface with and without a Parafilm protection layer. As shown in figure 4d, the conductivity of the coated Cu NW network was stable in the first 20 cm of abrasion, and then deteriorated afterward, which resulted in sheet resistance of approximately 3300 Ω cm⁻² at an abrasion distance of 50 cm. By contrast, the sample with Parafilm coating was by far more robust with a sheet resistance increase from 180 Ω cm⁻² to approximately 550 Ω cm⁻² after 50 cm of abrasion. Furthermore, the tape ripping off test and abrasion test shown in electronic supplementary material, figure S9 and table S2 indicate that the Parafilm coating is very robust and has strong adhesion to the wood surface. These results demonstrate that the Parafilm coating did not only render the wood surface water repellent, but also protected the Cu NW network from oxidation and mechanical damage. However, a natural wax-based coating could be an alternative to Parafilm, when a renewable coating material should be applied.

4. Conclusion

By combining photonic curing of a highly conductive and flexible Cu NW with a Parafilm solution-based coating, a highly conductive, air-stable and water-repellent coating has been established on the surface of the wood, one of the most abundant natural materials. The electrical characteristics of the Cu NW network are dictated by the unique surface topography of wood, which results in a highly anisotropic conductivity. The facile utilization of Parafilm, a common laboratory material, not only limits the problematic liquid water uptake of wood, but also preserves the Cu NW network from oxidation and mechanical damage. This concept of embedded functionality opens a new venue for the integration of thermal sensitive biomaterials as conductive substrates for electronic devices, which could be a renewable alternative to the plastic substrates traditionally used.

Data accessibility

The experimental datasets supporting this article are included in the main paper as well as the electronic supplementary material.

Authors' contributions

H.G. carried out Cu NW synthesis, developed the coating methods, the design of the study and drafted the manuscript. M.B. carried out conductivity, contact angle measurement and stability test. X.L. measured the mechanical and electrical properties of an individual Cu NW. A.W. performed photonic curing and supported conductivity measurement. Q.C. coordinated the study of the single Cu NW and helped the data analysis and revised the manuscript. I.B. contributed to conceiving and designing the study and to the drafting of the manuscript. All authors gave final approval for publication.

Competing interests

We have no competing interests.

Funding

X.L. and Q.C. were supported by NSF of China (grant nos11528407 and 61621061).