Nanoscale

PAPER

Cite this: DOI: 10.1039/c9nr10565d

Received 13th December 2019,Accepted 23rd March 2020

DOI: 10.1039/c9nr10565d

rsc.li/nanoscale

Monovalent sulfur oxoanions enable millimeter-long single-crystalline h-WO3 nanowire synthesis†

Guozhu Zhang,a Chen Wang,a Wataru Mizukami,b Takuro Hosomi,a

Kazuki Nagashima,a Hideto Yoshida, c Kentaro Nakamura,a Tsunaki Takahashi,a

Masaki Kanai,a Takao Yasui, d Yuriko Aoki,b Yoshinobu Babad andTakeshi Yanagida *a,b

Here, we discuss a misunderstanding regarding chemical capping, which has intrinsically hindered the

extension of the length of hexagonal (h)-WO3 nanowires in previous studies. Although divalent sulfate

ions (SO42−) have been strongly believed to be efficient capping ions for directing anisotropic h-WO3

nanowire growth, we have found that the presence of SO42− is highly detrimental to the anisotropic

crystal growth of the h-WO3 nanowires, and a monovalent sulfur oxoanion (HSO4−) rather than SO4

2−

only substantially promotes the anisotropic h-WO3 nanowire growth. Ab initio electronic structure simu-

lations revealed that the monovalent sulfur oxoanions were preferentially able to cap the sidewall plane

(100) of the h-WO3 nanowires due to the lower hydration energy when compared with SO42−. Based on

this capping strategy, using the monovalent sulfur oxoanion (CH3SO3−), which cannot generate divalent

sulfur oxoanions, we have successfully fabricated ultra-long h-WO3 nanowires up to the millimeter range

(1.2 mm) for a wider range of precursor concentrations. We have demonstrated the feasibility of these

millimeter-long h-WO3 nanowires for the electrical sensing of molecules (lung cancer biomarker:

nonanal) on flexible substrates, which can be operated at room temperature with mechanical flexibility

with bending cycles up to 104 times due to the enhanced textile effect.

IntroductionUnderstanding the selectivity of the crystal growth directionusing capping molecules is important for spatially designinginorganic/organic nanomaterials for various applicationsincluding catalysts,1–5 sensors6–10 and energy devices.11–16 Thechemical capping method usually utilizes structure-directingagents, which preferentially cap a specific crystal face to obtainthe selectivity of the crystal growth direction.17–24 Since suchchemical capping occurs during a dynamic crystal growthprocess under relatively harsh conditions, it is frequentlydifficult to obtain the exact physical pictures of cappingbehaviors.25–27 For example, hexagonal tungsten oxide (h-WO3)nanowires are often hydrothermally synthesized by introdu-

cing a capping molecule. Sulfate salts, such as Na2SO4, K2SO4

and (NH4)2SO4, are typically employed as the structure-directing agents.28–37 During the crystal growth process, SO4

2−

has been suggested as an efficient capping molecule, whichpreferentially caps the sidewall plane (100) of the h-WO3

nanowires.30,31 Indeed, the addition of these sulfate saltsremarkably increases the aspect ratio of the WO3 nanowires upto 625.35 However, further increase in the SO4

2− concentrationfrequently results in nanowires as well as other complex nano-structures, including bamboo-like structures,36 and even trig-gers the transformation of the crystal phase into pyrochlore-phase microcrystals from hexagonal-phase nanowires.37 Theaddition of sulfate ions during WO3 nanowire growth mustcause more complex molecular phenomena, which cannot beexplained by the existing SO4

2− capping model.30,31

Here, we discuss a long-standing misunderstanding of tra-ditional SO4

2−-based chemical capping models for h-WO3

hydrothermal nanowire synthesis. We experimentally andtheoretically reveal that the presence of SO4

2− is detrimental toh-WO3 hydrothermal nanowire growth, and the monovalentsulfur oxoanion (HSO4

−) rather than SO42− only substantially

promotes the anisotropic h-WO3 nanowire growth. By preciselycontrolling the amount of monovalent sulfur oxoanionsduring crystal growth, we have successfully fabricated ultra-

†Electronic supplementary information (ESI) available. See DOI: 10.1039/c9nr10565d

aInstitute for Materials Chemistry and Engineering, Kyushu University, 6-1 Kasuga-Koen, Kasuga, Fukuoka, 816-8580, Japan. E-mail: yanagida@cm.kyushu-u.ac.jpbInterdisciplinary Graduate School of Engineering Sciences, Kyushu University, 6-1 Kasuga-Koen, Kasuga, Fukuoka, 816-8580, JapancThe Institute of Scientific and Industrial Research, Osaka University, 8-1 Mihogaoka,Ibaraki, Osaka 567-0047, JapandGraduate School of Engineering, Nagoya University, Furo-cho, Chikusa-ku, Nagoya464-8603, Japan

This journal is © The Royal Society of Chemistry 2020 Nanoscale

Publ

ishe

d on

23

Mar

ch 2

020.

Dow

nloa

ded

by T

he U

nive

rsity

of T

okyo

on

4/14

/202

0 1:

46:3

3 PM

.

View Article OnlineView Journal

long WO3 nanowires with a length up to 1.2 mm. We alsodemonstrated the applicability of these millimeter-long h-WO3

nanowires for the electrical sensing of molecules (biomarkerin breath: nonanal) on flexible substrates. The millimeter-longWO3 nanowires exhibited superior mechanical flexibility anddurability due to the enhanced textile effect.

Results and discussionFirst, we will discuss the effects of both a capping moleculesource Na2SO4 concentration (CNS) and a tungsten precursor(polytungstic acid) concentration (CW) on WO3 nanowiregrowths. The CNS was controlled from 0 mM to 50 mM, andCW ranged from 0.1 mM to 54.4 mM (the growth resultswithout capping agent (CNS, 0 mM) are shown in the ESI,Fig. S1†). Fig. 1a shows the variation map of fabricated nano-structures as a function of both concentrations, respectively.Detailed SEM images are shown in the ESI, Fig. S2 and S3.†Here, oxalic acid (H2C2O4) is utilized to guarantee thatsufficient amounts of tungsten precursor colloidal particlescan be generated in solution for subsequent hydrothermalgrowth (Fig. S4†).30 The other experimental details can be seenin the Experimental section. The concentration mapping datacan be classified into three regions named “R1” to “R3”. Eachincludes (R1) no visible crystals when CW is below 1 mM, (R2)nanowires and nanorods structures with hexagonal WO3

crystal structures, or (R3) microplate structures with the pyro-

chlore WO3 crystal phase. The typical capping agent (Na2SO4)concentration-dependent SEM and XRD data are shown inFig. 1b and c (TEM characterizations of the as-grown nano-wires are shown in the ESI Fig. S5†). The first type of transitionis from R1 to R2 (type 1), which appears when CW increases.This transition is not observable above 10 mM of CNS. Thesecond type of transition is from R1 to R3 (type 2) andemerges when CW increases with higher CNS. The third type oftransition is from R2 to R3 (type 3), which is observed above∼1 mM of Cw when CNS is increasing. According to the WO3

nanostructure growth results in Fig. 1a, the h-WO3 nanowiregrowth rate mapping data are plotted in Fig. 1d. Interestingly,we found that long h-WO3 can only be formed near the bound-ary regions of these three regions. Significantly long WO3

nanowires with lengths of up to 500 μm and a diameter of100 nm (ESI Fig. S6†), are obtained within this boundary.Thus, the control of both CNS and CW critically determines theemergence of fabricated nanostructure forms, crystal phases(hexagonal or pyrochlore phases), and the fabrication of longWO3 nanowires. More importantly, traditional SO4

2−-basedchemical capping models seem to not be appropriate toexplain these experimental trends. Based on a previously pro-posed model for the mechanism of WO3 nanowire formation,SO4

2− must play an important role as a capping molecule forthe sidewall of nanowires.30,31 However, the trend in Fig. 1,where the increasing CNS is a capping molecule source, is notconsistent with this traditional SO4

2− capping model.30,31 Thisis because the increase in CNS resulted in the transition

Fig. 1 (a) Variation in the morphology and crystal structure (R1: × no visible crystal, R2: ● nanowire, R3: ◇ microplate, and ◈ mixture of nanowiresand microplates). (b) Typical SEM images of Na2SO4 concentration (CNS)-dependent WO3 nanostructures. (c) XRD characterization of the WO3 nano-structures from (b). (d) h-WO3 nanowire growth rate data.

Paper Nanoscale

Nanoscale This journal is © The Royal Society of Chemistry 2020

Publ

ishe

d on

23

Mar

ch 2

020.

Dow

nloa

ded

by T

he U

nive

rsity

of T

okyo

on

4/14

/202

0 1:

46:3

3 PM

.

View Article Online

(type 3) from nanowires to microplates. Thus, this contradic-tive result demonstrates the inability of the existing conven-tional model based on SO4

2− capping only for WO3 anisotropicnanowire crystal growth.

In order to explain the inconsistency between the experi-mental trend in Fig. 1 and the existing traditional SO4

2−

capping model, we considered the presence of HSO4− because

HSO4− rather than SO4

2− can be the dominant species undercertain conditions (e.g. low pH range), HSO4

− ↔ SO42− + H+,

pKa = 1.99.38 Fig. 2b shows the calculated concentrations oftwo different charged states of sulfur oxoanions (HSO4

− andSO4

2−) from Na2SO4 when varying CNS with the fixed value(2.72 mM) of Cw. The calculations were performed by usingcommercial software (Visual MINTEQ) (ESI Fig. S7†). For thesesimulations, the ion concentrations were calculated by usingexperimentally measured values of pH (ESI Fig. S8†) and thetemperature for the crystal growth experiments. With increas-ing CNS, SO4

2− tends to be the major species because theincrease in CNS results in an increase in pH to promote thedeprotonation of HSO4

− (Fig. 2b). On the other hand, HSO4−

tends to be dominant for the lower CNS concentration rangebelow 20 mM, as also highlighted in the plotted data of theconcentration difference data (ΔC = CHSO4

− ↔ CSO42−) between

HSO4− and SO4

2−. ΔC data exhibits a maximum of around7 mM of CNS. Interestingly, this concentration range is consist-ent with the appearance of long WO3 nanowires in Fig. 1. Atthe higher CNS range where SO4

2− is the dominant sulfuranion, pyrochlore WO3 microplates emerge as shown in Fig. 1.As such, in comparison with the experimental data in Fig. 1,the calculated data for sulfur oxoanions proposes a model thatexplains the experimentally observed anisotropic WO3 nano-wire growth in terms of the presence of HSO4

− rather than thepreviously proposed SO4

2− as the structure-directing cappingspecies. This model is based on the experimental results thatlong WO3 nanowires can be formed for the HSO4

− dominant

concentration range (i.e. ΔC > 0). In order to further validatethis capping model for the wider range of experimentsemployed, we calculated the populations of two sulfur oxoa-nions via measuring the pH values for the wide ranges of Cw

and CNS. The measured pH values at the growth temperature,when varying both Cw and CNS, are shown in Fig. S5.† Theincrease in Cw results in a decrease in the pH value due to theincreased concentration of tungstic acid (H2WO4). Withincreasing CNS, the pH value increases because Na2SO4 acts asa buffer in acidic solution due to its weak basicity.39 Based onthese pH values, we can calculate the populations of two sulfuroxoanions (ESI Fig. S9†). ΔC/Cw values are plotted in Fig. 2cbecause the capping efficiency of HSO4

− on tungsten oxidenuclei is related to ΔC/Cw. As can be seen, the calculated ΔC/Cw data show the maximum concentration range where longWO3 nanowires were observed in Fig. 1b. Thus, there is acertain correlation between the presence of HSO4

− in a solu-tion and the appearance of long WO3 nanowires.

The results above imply that HSO4−, rather than SO4

2−,should be able to preferentially cap a sidewall of h-WO3 (100)nanowires, enhancing the anisotropic nanowire crystal growth.To further validate this model, we intentionally changed theequilibrium concentration of HSO4

− by reducing the growthsolution pH from 1.68 to 1.12 with 1 M NaOH. The totalamount of Na2SO4 was controlled at 20 mM. As speculated, theequilibrium ion concentration of HSO4

− increased with thedecreasing pH value via protonation (ESI Fig. S10†). Althoughthe pH range is relatively narrow, 1.12–1.68, the effect on thefabricated WO3 nanostructure morphology is significant (ESIFig. S11†). At the pH value of 1.68, only the microplate mor-phology with the pyrochlore WO3 crystal phase (ESI Fig. S12†)is observable. Decreasing pH values strongly enhance the an-isotropic hexagonal WO3 nanowire growth (ESI Fig. S12†).These results provide further experimental support for amodel based on HSO4

− rather than the previously suggested

Fig. 2 (a) Schematic illustration of ionization for Na2SO4 in solution. (b) Calculated concentrations of two different charged states of sulfur oxoa-nions (HSO4

− and SO42−) from Na2SO4 when varying CNS. (c) Calculated ΔC/CW data when varying CW and CNS.

Nanoscale Paper

This journal is © The Royal Society of Chemistry 2020 Nanoscale

Publ

ishe

d on

23

Mar

ch 2

020.

Dow

nloa

ded

by T

he U

nive

rsity

of T

okyo

on

4/14

/202

0 1:

46:3

3 PM

.

View Article Online

SO42− as an efficient capping species for anisotropic WO3

nanowire crystal growth.To obtain a better understanding of the selective capping

behaviours of HSO4− or SO4

2− on WO3 nanostructure surfaces,we performed ab initio-density functional theory (DFT) simu-lations. Under gas-phase conditions, the binding energy of asingle HSO4

− or SO42− onto the WO3 crystal plane can be esti-

mated by considering the difference in the total energies ofthe combined and the isolated systems with the followingformula:

EBinding ¼ EHSO4"=SO4

2"þWO3 " ðEHSO4"=SO4

2" þ EWO3Þ:

However, in the aqueous solution, HSO4− or SO4

2− isstrongly hydrated with the surrounding water molecules.40,41

Therefore, the calculations must consider such hydrationeffects to compare the capping effect of HSO4

− or SO42− on the

WO3 surfaces. As illustrated in Fig. 3a, the competitionbetween ion adsorption and hydration determines whether thecapping effect of molecules on solid surfaces within anaqueous solution occurs. Consequently, the net bindingenergy of capping molecules (HSO4

− or SO42−) may be formu-

lated as follows:

ΔENet binding ¼ EBinding " αEHydration

where αEHydration is the hydration energy loss after the mole-cules bind to the WO3 surface (α, hydration energy lossratio42). The net binding energy (ΔE) of the capping molecules(HSO4

− or SO42−) on the hexagonal WO3 nanowire top (001)

and sidewall (100) planes are calculated by the methoddescribed in the Experimental section and ESI.† Fig. 3b showsthe calculated net biding energy of both capping molecules(HSO4

− and SO42−) on the sidewall (100) plane of h-WO3 (calcu-

lation data of the (001) plane is shown in ESI, Fig. S17†). Ascan be seen, the net binding energy for HSO4

− on the (100)plane is negative, whereas SO4

2− exhibits a positive net

binding energy due to the higher stability of hydration(Table 1). In addition, the net binding energies of both HSO4

−

and SO42− on the (100) plane are positive, indicating that they

are unable to adsorb on the (100) plane in the water phase. Asa result, the nanowire sidewall (100) planes are hindered fromfurther growth, while the growth of the top (001) planes isselectively facilitated. These results are also consistent withprevious data, in which the hydration energy of SO4

2−

(−1080.2 kJ mol−1)43 is reported to be four times higher thanthat of HSO4

− (−285.1 kJ mol−1).42 These simulations alsoindicate that in addition to the hydration effects, the lessercoulombic repulsion effect of the monovalent sulfur oxoanionas compared to the divalent sulfate ion might increase theadsorbed ion surface density, which also promotes a cappingeffect. Thus, these ab initio-DFT simulations highlight thatHSO4

− preferentially adsorbs onto the (100) plane of h-WO3,while SO4

2− is strongly hydrated rather than being on the (100)plane of h-WO3. This strong hydration effect of SO4

2− is well-known as the Hofmeister series.44,45 In addition, this stronghydration effect can be experimentally confirmed in Fig. 1a,where the critical nucleation concentration decreases with CNS

increasing above 5 mM of the CNS due to the strong hydrationeffect of SO4

2−. These simulation results give an interpretation

Fig. 3 (a) Crystal structure of hexagonal WO3. (b) Calculated binding energies of both sulfur oxoanions (HSO4− and SO4

2−) on the (100) plane of thehexagonal WO3 in the vapour and water phases. The coordination structures were simulated by DFT calculations.

Table 1 Binding energies, ΔEBinding and ΔENet_binding, of differentcapping species on h-WO3 (001) and (100) planes

Cappingspecies

Binding energy (kJ mol−1)

In the vapor phaseΔEBinding

In the water phaseΔENet_binding

(001) (100) (001) (100)

SO42− −448.8 −618.4 74.9 60.3

HSO4− −297.7 −315.3 56.1 −64.9

Paper Nanoscale

Nanoscale This journal is © The Royal Society of Chemistry 2020

Publ

ishe

d on

23

Mar

ch 2

020.

Dow

nloa

ded

by T

he U

nive

rsity

of T

okyo

on

4/14

/202

0 1:

46:3

3 PM

.

View Article Online

for a model based on a HSO4− capping mechanism during the

hexagonal WO3 nanowire growth trend in Fig. 1.The above experimental and theoretical data consistently

highlight that the presence of monovalent HSO4− rather than

the previously proposed divalent SO42− essentially promotes

anisotropic WO3 nanowire growth. This suggests that suppres-sing the formation of divalent sulfur oxoanions in a solutionmust be able to solve the detrimental effect above due to thepresence of SO4

2−. However, the controllability of the concen-tration of HSO4

− is substantially limited due to the occurrenceof protonation or deprotonation in solution as long as weemploy sulfate salts including Na2SO4, K2SO4 and (NH4)2SO4.To overcome this essential difficulty, we employed sodiummesylate (CH3SO3Na) as a capping molecule source instead ofNa2SO4. This is because the corresponding CH3SO3

− anion hasa structure in which one of the hydroxyl groups on HSO4

− isreplaced by a methyl group,46 and therefore CH3SO3

− does notdissociate further into a divalent anion (Fig. 4a). Fig. 4b showsthe applicability of the present method for the wider concen-tration range in comparison with Fig. 1b. Interestingly, theeffect of CH3SO3Na is significant, as can be seen in the SEM

images (Fig. 4c). Note that only at the highest concentration ofCH3SO3Na, the WO3 microplate emerges due to the decompo-sition of CH3SO3

− into SO42− in the temperature range

employed.47 The long nanowire morphology with the hexag-onal phase WO3 can be maintained even on increasing theconcentration of CH3SO3Na (CMS) up to 50 mM (ESI Fig. S13and S14†), which is in sharp contrast to the case for NaSO4,where the pyrochlore crystal phase WO3 microplates emerge asshown in Fig. 1a. Since this method using CH3SO3Na does notsuffer from the presence of SO4

2− or varying CMS, the controll-ability of fabricating long WO3 nanowires is much superior tothat of the conventional Na2SO4-based method. By utilizingCH3SO3Na as a capping molecule source, the length of thenanowires reached up to 1.2 mm, as shown in Fig. 4d (thehigh-magnification SEM image is shown in Fig. S15†).Surprisingly, these ultra-long nanowires can be assembled intocentimeter-long nanowire yarn during growth (Fig. 4e andVideo S1†). Compared with the reported WO3 nanowiresobtained through the hydrothermal method, the as-grownnanowires in this work showed the highest length and aspectratio (ESI Table S1 and Fig. S16†). Thus, the charged state of

Fig. 4 (a) Schematic illustration of ionization for CH3SO3Na in a solution. (b) The effect of CH3SO3Na concentration (CMS) on the fabricated WO3

nanostructure morphology (SEM images). (c) Variations in the morphology and crystal structure on varying CW and CMS. (d) Millimeter-long h-WO3

nanowire when controlling CW (2.18 mM) and CMS (6 mM). (e) Manipulation of the as-grown h-WO3 nanowire yarn (assembled from ultra-longnanowire).

Nanoscale Paper

This journal is © The Royal Society of Chemistry 2020 Nanoscale

Publ

ishe

d on

23

Mar

ch 2

020.

Dow

nloa

ded

by T

he U

nive

rsity

of T

okyo

on

4/14

/202

0 1:

46:3

3 PM

.

View Article Online

the capping molecules critically determines the degree of an-isotropic WO3 nanowire growth.

Finally, we show the applicability of the fabricated long hex-agonal WO3 nanowires towards molecule-sensing devices withrobust mechanical flexibility. We employed nonanal as themodel target molecule (as a lung cancer biomarker inbreath48). Fig. 5a shows photographs of the as-grown ultra-long hexagonal WO3 nanowire film and the fabricated flexiblenanowire film sensors on polyethylene naphthalate (PEN) sub-strate. The details of the fabrication of nanowires and flexiblesensor devices, including the nanowire transfer process, canbe seen in the Experimental section. Note that the WO3 nano-wire films strongly stick to the PEN substrate, evenin situations of mechanical bending, for up to 104 cycles (ESIFig. S23†). Fig. 5b shows typical data of the sensing responseto 2.67 ppm nonanal (the gas concentration was calibrated byGC-MS, ESI Fig. S18†) at room temperature. To study the effectof the nanowire length on the flexible sensor characteristics,the data for relatively long WO3 nanowires (average ∼200 μm,containing millimeter-long nanowire) and relatively short WO3

nanowires (average ∼15 μm) with similar diameters (∼100 nm)are shown in the figure. Both nanowire sensor devices exhibi-

ted sensing responses to the nonanal molecules, and the longnanowire-based sensor presented a greater sensing perform-ance as compared with that of the short nanowire. This sensorresponse can be recovered to the initial state on removingnonanal (ESI Fig. S19 and S20†). To the best of our knowledge,so far, this is the first report of nonanal electrical sensing atroom temperature.49,50 Since the diameter of the WO3 nano-wire cannot be fully depleted at room temperature51 and thereare no metal–semiconductor contact barriers formed in thenanowire sensor device (Fig. S21†), the sensing performance ispredominately controlled by the nanowire junctions.52 Assuch, introducing nonanal decreases the nanowire junction re-sistance of the sensors. Also, ultraviolet–visible absorptionspectra and photoluminescence spectra (Fig. S22†) provideclear evidence that the long h-WO3 nanowires contain a greaterdensity of oxygen vacancy defects as compared with the shortnanowire.53 Thus, greater nonanal sensing performance wasobserved in the long nanowire-based sensor device beforemechanical bending.

As is known, the implementation of micro- or nanoscalestructural forms of inorganic semiconductor materials in flex-ible electronics relies critically on their ability to be bent

Fig. 5 (a) Schematic illustration of the h-WO3 nanowire-based molecular sensor for lung cancer biomarker detection (upper), and the structure ofthe as-fabricated nanotextile sensor device (lower) (scale bar 2 mm). Response curves of the (b) ultra-long and (c) short h-WO3 nanowire-basedsensors to 2.67 ppm of nonanal, tested with and without mechanical bending at room temperature. (d) and (e) Images of the long and short h-WO3

nanowire-based sensors after 2000/500 cycles of mechanical bending. (f ) The response of the ultra-long h-WO3 nanowire-based sensor to2.67 ppm of nonanal under different bending curvature radii and (g) bending cycles.

Paper Nanoscale

Nanoscale This journal is © The Royal Society of Chemistry 2020

Publ

ishe

d on

23

Mar

ch 2

020.

Dow

nloa

ded

by T

he U

nive

rsity

of T

okyo

on

4/14

/202

0 1:

46:3

3 PM

.

View Article Online

repeatedly without fracturing.54 Herein, we mainly focused onthe bendability of the nanowire-based sensor device. Fig. 5band c show that the mechanical bending process strongly inhi-bits the sensor response only for devices composed of rela-tively short nanowires but not for those with relatively longnanowires. It was found that the long WO3 nanowire sensordevice demonstrated only slight sensing performance degra-dation (8.6% of the initial value) after being subjected to 2000cycles of bending (Fig. 5b), while the short WO3 nanowiresensor device showed the rapid degradation of the sensing per-formance to nonanal, even after bending for 500 cycles(Fig. 5c). As shown in Fig. 5d and e, the film degradation withcracks is responsible for this discrepancy between the twonanowire sensor devices. The mechanical bending process forover 500 times resulted in a crack within the nanowire filmscomposed of relatively short nanowires. However, suchmechanically induced film cracks are much less observable forsensor devices composed of relatively long nanowires. This issolely due to the enhanced textile effect of longer nanowiredevices when compared with short nanowires. Specificallyspeaking, due to the intrinsically high-aspect-ratio and spatialentanglement of the long nanowires, the constructed nanowirenetwork can prevent the formation of the cracks when strainsare applied, thereby reducing the dependence of the resistancechange on the strain.55 We further investigated the effects ofthe bending strain and the cycle endurance on the sensing per-formance. Fig. 5f shows the results of bending curvatureradius (strain)-dependent sensing responses, implying that thesensor performances exhibit slight degradations when the cur-vature radius is lower than 2.1 mm. Even for bending with acurvature radius as small as 0.5 mm (high bending strain), thesensor still retained good sensing properties (36.9% degra-dation). The short nanowire-based sensor presented dramaticperformance degradation as the curvature radius went lowerthan 20 mm, and almost lost all the performance with a curva-ture radius of 0.5 mm. The bending cycle endurance data isshown in Fig. 5g. There was no significant sensing perform-ance degradation for the long WO3 nanowire-based flexiblesensor for up to 104 cycles (ESI Fig. S23†). The short nanowire-based sensor showed poor mechanical bending performance.These results highlight the superior molecule-sensing pro-perties with excellent mechanical flexibility due to theenhanced textile effects of long WO3 nanowires.

ConclusionsIn summary, we have found that the monovalent sulfur oxoa-nions (HSO4

− and CH3SO3−), rather than previously proposed

SO42−, substantially enhance the anisotropic nanowire growth

of h-WO3. In fact, the presence of SO42− hinders the aniso-

tropic crystal growth of WO3. The ab initio-density functionaltheory (DFT) simulations revealed that the monovalent sulfuroxoanions are more effectively able to cap the sidewall plane(100) of hexagonal WO3 nanowires due to the lower hydrationenergy as compared with SO4

2−. Based on this capping strategy

using CH3SO3−, which can generate only monovalent sulfur

oxoanions, we successfully achieved ultra-long of WO3 nano-wires up to the millimeter range (1.2 mm). Furthermore, wedemonstrated the feasibility of these millimeter-long WO3

nanowires for the electrical sensing of molecules (lung cancerbiomarker: nonanal) on the flexible substrates, which can beoperated at room temperature with the mechanical flexibilityeven bending cycles up to 104 times due to the enhancedtextile effect.

ExperimentalFabrication of WO3 nanowires

h-WO3 nanowires were fabricated in two steps, which includethe polytungstic acid sol synthesis and nanowire growth. Forthe polytungstic acid sol synthesis, 4 g of tungsten powder (W)was dissolved in 30 g of hydrogen peroxide solution (H2O2,30 wt% in H2O) in an ice bath under stirring. After the Wpowder was completely dissolved, a transparent solution wasobtained. Pt foil was then immersed in the solution to catalyti-cally remove the excess H2O2. When no bubbles were gener-ated, the as-synthesized polytungstic acid sol was diluted withdistilled water to 100 g. Subsequently, 12.5 g of polytungsticacid sol (2.72 mM W) was transferred into a 100 mL capacity ofautoclave and 37.5 g distilled water was added to further dilutethe solution. Then, 0.5625 g (6.25 mM) oxalic acid (H2C2O4)was added to the mixture and NaOH solution (1 mol L−1) wasused to adjust the pH of the mixture solution to 1.68. Next,0.5 g (3.5 mM) of sodium sulfate (Na2SO4) was added to thegrowth solution. After the complete dissolution of Na2SO4, theTeflon-lined high-pressure autoclave was sealed and trans-ferred to the sintering furnace and maintained at 200 °C for24 hours. WO3 nanowires were finally obtained after centrifu-gation of the products and washing with distilled water toremove the excess ions.

Structural characterizations

The structural characterization of the grown WO3 nanowireswas performed using X-ray diffraction (Rigaku, SmartLab, 45kV, 40 mA), field emission scanning electron microscopy(JEOL, JSM-7610F) equipped with energy-dispersive X-ray spec-troscopy (EDS) and transmission electron microscopy (JEOL,JEM-ARM200F, 200 kV).

Molecule sensing procedure

The as-grown WO3 nanowires were washed several times withwater and added to the polyethylene naphthalate (PEN) sub-strate. The residual water was removed By heating the PENsubstrate at 100 °C for 15 min. After the removal of the water,the WO3 nanowires were stuck to the PEN substrate andformed a uniform nanowire film. The radiofrequency (RF)sputtering technique was used to deposit the platinum (Pt)electrode with a stainless steel mask. To enhance the adhesionbetween the substrate and the Pt electrodes, a thin layer(∼5 nm) of titanium (Ti) was sputtered ahead of the Pt. The

Nanoscale Paper

This journal is © The Royal Society of Chemistry 2020 Nanoscale

Publ

ishe

d on

23

Mar

ch 2

020.

Dow

nloa

ded

by T

he U

nive

rsity

of T

okyo

on

4/14

/202

0 1:

46:3

3 PM

.

View Article Online

sensing performance toward nonanal was carried out by ahome-made chamber combined with the Keithley 4200A-SCSParameter Analyzer (TEKTRONIX, INC.) at room temperature.The resistance of the WO3 nanowire device under N2 wasrecorded as the baseline (RN2

). The resistance upon exposure tononanal gas (produced by bubbling liquid nonanal solution)was noted as Rgas. Sensitivity (S) was defined as ΔR/RN2

, ΔR =(RN2

− Rgas). Sensitivity degradation (ΔS) was defined as (Sflat −Sbending)/Sflat. All the sensing tests were carried out under a DCbias voltage of 1 V. Nonanal gas was produced by bubblingliquid nonanal solution with high-purity N2, and its concen-tration was calibrated by gas chromatography-mass spec-troscopy (GCMS, Shimadzu GCMS-QP2010 Ultra) with a temp-erature control system (GL Sciences, OPTIC-4). The flow ratesof the dry air and the nonanal gas were controlled at 200 sccmduring the sensing performance tests.

Ab initio electronic structure simulations

We used density functional theory (DFT) to compute theadsorption energies of SO4

2− and HSO4− on the h-WO3 (100)

and (001) planes. Different levels of theory were combined viathe ONIOM extrapolation scheme, which allowed us to con-sider the hydration effects and to refine results from pure DFTby exact-exchange hybrid DFT calculations.

Conflicts of interestThere are no conflicts to declare.

AcknowledgementsThis work was supported by KAKENHI (Grant Number: No.JP18H01831, No. JP18KK0112, No. JP18H05243, No.JP17H04927) and CAS-JSPS Joint Research Projects (Grant No.JPJSBP120187207), CREST (JPMJCR19I2) and Mirai R&D ofJapan Science and Technology Corporation (JST). This studywas also partly supported by the Cooperative ResearchProgram of “Network Joint Research Center for Materials andDevices”, “Dynamic Alliance for Open Innovation BridgingHuman, Environment and Materials in Network JointResearch Center for Materials and Devices” and the MEXTProject of “Integrated Research Consortium on ChemicalSciences”. The computation in this work was performed usingthe facilities of RIIT, Kyushu University, and theSupercomputer Center, the Institute for Solid State Physics,the University of Tokyo.

Notes and references1 T. Tachikawa, S. Yamashita and T. Majima, J. Am. Chem.

Soc., 2011, 133, 7197–7204.2 M. B. J. Roeffaers, B. F. Sels, H. Uji-i, F. C. De Schryver,

P. A. Jacobs, D. E. De Vos and J. Hofkens, Nature, 2006, 439,572.

3 R. Xu, D. S. Wang, J. Y. Zhang and Y. D. Li, Chem. – Asian J.,2006, 1, 888–893.

4 A. Trovarelli and J. Llorca, ACS Catal., 2017, 7, 4716–4735.5 H. Zhang, M. Jin, Y. Xiong, B. Lim and Y. Xia, Acc. Chem.

Res., 2013, 46, 1783–1794.6 S. Tian, F. Yang, D. Zeng and D. Xie, J. Phys. Chem. C, 2012,

116, 10586–10591.7 S. Deng, V. Tjoa, H. Fan, M. Tan, H. R. Sayle, D. C. Olivo,

M. Mhaisalkar, S. Wei, J. Sow and C. H. Sow, J. Am. Chem.Soc., 2012, 134, 4905–4917.

8 M. D. Arienzo, L. Armelao, A. Cacciamani, C. M. Mari,S. Polizzi, R. Ruffo, R. Scotti, A. Testino, L. Wahba andF. Morazzoni, Chem. Mater., 2010, 22, 4083–4089.

9 X. Han, M. Jin, S. Xie, Q. Kuang, Z. Jiang, Y. Jiang, Z. Xieand L. Zheng, Angew. Chem., Int. Ed., 2009, 48, 9180–9183.

10 M. H. Huang and P.-H. Lin, Adv. Funct. Mater., 2012, 22,14–24.

11 T. Sheng, Y. Xu, Y. Jiang, L. Huang, N. Tian, Z. Zhou,I. Broadwell and S. Sun, Acc. Chem. Res., 2016, 49, 2569–2577.

12 C. Hu, M. Li, J. Qiu and Y.-P. Sun, Chem. Soc. Rev., 2019, 48,2315–2337.

13 H. J. Yin and Z. Y. Tang, Chem. Soc. Rev., 2016, 45, 4873–4891.

14 J. E. ten Elshof, H. Y. Yuan and P. G. Rodriguez, Adv. EnergyMater., 2016, 6, 1600355.

15 H. Zhang, Q. Yun, Q. Lu, X. Zhang and C. Tan, Angew.Chem., Int. Ed., 2018, 57, 626–646.

16 C.-C. Chiang, C.-Y. Hung, S.-W. Chou, J.-J. Shyue,K.-Y. Cheng, P.-J. Chang, Y.-Y. Yang, C.-Y. Lin, T.-K. Chang,Y. Chi, H.-L. Chou and P.-T. Chou, Adv. Funct. Mater., 2018,28, 1703282.

17 Y. Sun and Y. Xia, Science, 2002, 298, 2176–2179.18 Y. Xia, Y. Xiong, B. Lim and S. E. Skrabalak, Angew. Chem.,

Int. Ed., 2009, 48, 60–103.19 J. Zeng, Y. Zheng, M. Rycenga, J. Tao, Z.-Y. Li, Q. Zhang,

Y. Zhu and Y. Xia, J. Am. Chem. Soc., 2010, 132, 8552–8553.20 P. D. Cozzoli, A. Kornowski and H. Weller, J. Am. Chem.

Soc., 2003, 125, 14539–14548.21 S. M. Lee, S. N. Cho and J. Cheon, Adv. Mater., 2003, 15,

441–444.22 M. Jin, G. He, H. Zhang, J. Zeng, Z. Xie and Y. Xia, Angew.

Chem., Int. Ed., 2011, 50, 10560–10564.23 S. F. Tan, G. Bisht, U. Anand, M. Bosman, X. E. Yong and

U. Mirsaidov, J. Am. Chem. Soc., 2018, 140, 11680–11685.24 C. Garcia Núñez, A. F. Braña, N. López and B. J. García,

Nano Lett., 2018, 18, 3608–3615.25 Q. Fu and X. Bao, Chem. Soc. Rev., 2017, 46, 1842–1874.26 P. Liu, R. Qin, G. Fu and N. Zheng, J. Am. Chem. Soc., 2017,

139, 2122–2131.27 T. H. Yang, Y. Shi, A. Janssen and Y. Xia, Angew. Chem., Int.

Ed., 2019, DOI: 10.1002/anie.201911135.28 X. W. Lou and H. C. Zeng, Inorg. Chem., 2003, 42, 6169–

6171.29 X. L. Li, J. F. Liu and Y. D. Li, Inorg. Chem., 2003, 42, 921–

924.

Paper Nanoscale

Nanoscale This journal is © The Royal Society of Chemistry 2020

Publ

ishe

d on

23

Mar

ch 2

020.

Dow

nloa

ded

by T

he U

nive

rsity

of T

okyo

on

4/14

/202

0 1:

46:3

3 PM

.

View Article Online

30 Z. Gu, Y. Ma and W. Yang, Chem. Commun., 2005, 3597–3599.

31 Z. Gu, T. Zhai, B. Gao, X. Sheng, Y. Wang, H. Fu, Y. Ma andJ. Yao, J. Phys. Chem. B, 2006, 110, 23829–23836.

32 Z. Gu, H. Li, T. Zhai, W. Yang, Y. Xia, Y. Ma and J. Yao,J. Solid State Chem., 2007, 180, 98–105.

33 X. C. Song, Y. F. Zheng, E. Yang and Y. Wang, Mater. Lett.,2007, 61, 3904–3908.

34 K. Huang, Q. Pan, F. Yang, S. Ni, X. Wei and D. He, J. Phys.D: Appl. Phys., 2008, 41, 155417.

35 A. Phuruangrat, D. J. Ham, S. J. Hong, S. Thongtem andJ. S. Lee, J. Mater. Chem., 2010, 20, 1683–1690.

36 J. Zhang, J. P. Tu, X. H. Xia, X. L. Wang and C. D. Gu,J. Mater. Chem., 2011, 21, 5492–5498.

37 J. Shi, G. Hu, R. Cong, H. Bu and N. Dai, New J. Chem.,2013, 37, 1538–1544.

38 D. H. Ripin, Evans pKa table, 2005.39 L. Lerner, Small-scale synthesis of laboratory reagents with

reaction modeling, CRC Press, 2011.40 R. Caminiti, G. Paschina, G. Pinna and M. Magini, Chem.

Phys. Lett., 1979, 64, 391–395.41 T. I. Yacovitch, T. Wende, L. Jiang, N. Heine and G. Meijer,

J. Phys. Chem. Lett., 2011, 2, 2135–2140.42 Y. Marcus, J. Chem. Soc., Faraday Trans., 1987, 83, 339–349.43 J. O. Bockris, M. Gamboa-Aldeco and M. Szklarczyk,

J. Electroanal. Chem., 1992, 339, 355–400.

44 Y. Zhang and P. S. Cremer, Annu. Rev. Phys. Chem., 2010,61, 63–83.

45 S. Nihonyanagi, S. Yamaguchi and T. Tahara, J. Am. Chem.Soc., 2014, 136, 6155–6158.

46 C. H. Wei and B. E. Hingerty, Acta Crystallogr., 1981, 37,1992–1994.

47 P. C. Hu and M. E. Tuvell, J. Am. Oil Chem. Soc., 1988, 65,1007–1012.

48 P. Fuchs, C. Loeseken, J. K. Schubert and W. Miekisch,Int. J. Cancer, 2010, 126, 2663–2670.

49 T. Itoh, T. Nakashima, T. Akamatsu, N. Izu and W. Shin,Sens. Actuators, B, 2013, 187, 135–141.

50 Y. Masuda, T. Itoh, W. Shin and K. Kato, Sci. Rep., 2015, 5,10122.

51 A. Labidi, C. Lambert-Mauriat, C. Jacolin, M. Bendahan,M. Maaref and K. Aguir, Sens. Actuators, B, 2006, 119, 374–379.

52 A. Rothschild and Y. Komem, J. Appl. Phys., 2004, 95, 6374–6380.

53 S. Cong, Y. Yuan, Z. Chen, J. Hou, M. Yang, Y. Su, Y. Zhang,L. Li, Q. Li, F. Geng and Z. Zhao, Nat. Commun., 2015, 6,7800.

54 A. J. Baca, J. H. Ahn, Y. Sun, M. A. Meitl, E. Menard,H. S. Kim, W. M. Choi, D. H. Kim, Y. Huang andJ. A. Rogers, Angew. Chem., Int. Ed., 2008, 47, 5524–5542.

55 S. Gong and W. Cheng, Adv. Electron. Mater., 2017, 3, 1600314.

Nanoscale Paper

This journal is © The Royal Society of Chemistry 2020 Nanoscale

Publ

ishe

d on

23

Mar

ch 2

020.

Dow

nloa

ded

by T

he U

nive

rsity

of T

okyo

on

4/14

/202

0 1:

46:3

3 PM

.

View Article Online