Superwetting of TiO2 by light-induced water-layergrowth via delocalized surface electronsKunyoung Leea, QHwan Kima, Sangmin Ana, JeongHoon Anb, Jongwoo Kima, Bongsu Kima, and Wonho Jhea,1

aInstitute of Applied Physics, Department of Physics and Astronomy, Seoul National University, Gwanak-gu, Seoul 151-747, Korea; and bPark Systems, Iui-Dong906-10, Suwon 443-270, Korea

Edited by Alenka Luzar, Virginia Commonwealth University, Richmond, VA, and accepted by the Editorial Board March 9, 2014 (received for reviewOctober 9, 2013)

Titania, which exhibits superwetting under light illumination, hasbeen widely used as an ideal material for environmental solutionsuch as self-cleaning, water–air purification, and antifogging.There have been various studies to understand such superhydro-philic conversion. The origin of superwetting has not been clarifiedin a unified mechanism yet, which requires direct experimentalinvestigation of the dynamic processes of water-layer growth.We report in situ measurements of the growth rate and heightof the photo-adsorbed water layers by tip-based dynamic forcemicroscopy. For nanocrystalline anatase and rutile TiO2 we ob-serve light-induced enhancement of the rate and height, whichdecrease after O2 annealing. The results lead us to confirm thatthe long-range attraction between water molecules and TiO2,which is mediated by delocalized electrons in the shallow trapsassociated with O2 vacancies, produces photo-adsorption of wateron the surface. In addition, molecular dynamics simulations clearlyshow that such photo-adsorbed water is critical to the zero contactangle of a water droplet spreading on it. Therefore, we concludethat this “water wets water” mechanism acting on the photo-adsorbed water layers is responsible for the light-induced super-wetting of TiO2. Similar mechanism may be applied for better un-derstanding of the hydrophilic conversion of doped TiO2 or otherphoto-catalytic oxides.

Understanding the underlying mechanism of the light-inducedsuperwetting of TiO2 is of recent scientific interest (1, 2)

and also crucial for industrial applications. The phenomenon hasbeen attributed to production of surface radical groups (1, 3–5)or photo-oxidation of hydrophobic surface contaminants (6–8),but its origin remains controversial (9, 10) in part because con-ventional contact-angle analysis alone cannot fully account forthe complex surface wettability (11). To address the fundamentalmechanism responsible for superwetting, one has to go beyondthe contact-angle measurement that reflects the specific surfaceproperties and to probe the temporal growth dynamics of thewater layer itself adsorbed on TiO2 under wide-spectrum illu-mination of visible and near-infrared (NIR) in addition to UVlight. The dynamic atomic force microscopy (12) that employsthe stiff and sensitive tip based on the quartz tuning fork (QTF)oscillator (13, 14) is an ideal solution to realize the experimentalchallenge of direct measurement of the growth rate and height,which is inaccessible to the cantilever-based friction force mi-croscopy owing to jump-to-contact instability (1).

Results and DiscussionIn Situ Measurement of Photo-Adsorbed Water Layers. Let us firstdiscuss direct measurement procedures for the growth dynamicsof the UV-induced water layers on the rutile TiO2 (R-TiO2)sample (Figs. S1–S3). When the laterally oscillating tip approachesthe surface without UV light, the capillary-condensed water me-niscus (15) [called “bare meniscus” (BM)] forms in the tip-surfacenano-gap (Fig. 1 A, i), at point E0 (red arrow, Fig. 1B). When theamplitude slightly decreases owing to the viscous damping bythe water nano-meniscus, the tip is immediately retracted untilthe meniscus is ruptured (Fig. 1B, Inset). The corresponding

effective damping force (black curve in Fig. 1C) is calculated bythe damped harmonic equation of the tip (13, 14).When R-TiO2 is irradiated with UV laser right after BM is

formed, the oscillation amplitude decreases drastically to D0(Fig. 1B). The subsequent hysteresis persists until the fartherrupture point R (Fig. 1B). This indicates UV light inducesa sharp increase in the interfacial water volume associated withthe growing water layer (Fig. 1 A, ii). The red curve in Fig. 1Cshows the damping force during the UV-induced growth. As theretraction speed increases to 1.8 (2.3) nm/s, the respectiveresponses are given by the green (blue) curve in Fig. 1 B and C,where the maximum damping occurs at the farther distance D1(D2). As observed, faster retraction produces smaller damping,whereas R is invariant within ±0.5 nm. This indicates the waterlayer grows faster than 2.3 nm/s, whereas its final height is con-stant (further discussed below).For accurate measurements of the growth rate and height of

water layers, we first illuminate UV light before BM is formedand then measure the time elapsed between illumination points(Ei’s) and contact points (Mi’s) where the approaching tip con-tacts the growing water interface via the formation of waternano-meniscus. Time resolution is 100 ms, ∼1.5 times longerthan the intrinsic relaxation-time constant of QTF (Materials andMethods, Fig. S4). With farther Ei (i = 1–5 in Fig. 1D), theelapsed time becomes longer but the growth rate stays constantat 3.4 ± 0.2 nm/s, which justifies our growth-rate measurementscheme. Notice that for any E6 much earlier than E5 all of theapproach-retraction curves converge to the orange curve (Fig.1D), which evidences the formation of “final meniscus” (FM)(Fig. 1 A, iii) at the fixed point F, irrespective of the tip speed.

Significance

TiO2, which is chemically stable, harmless, and inexpensive, hasbeen widely used for industrial applications. Recently, TiO2-coated materials, exhibiting superwetting under sunlight, havebeen developed for environmental solutions. However, themechanism responsible for superwetting of TiO2 is still in con-troversy despite many studies. We clarified its origin by per-forming tip-based in situ measurements of the growth dynamicsof the photo-adsorbed water layers, as associated with delo-calized surface electrons. Combined with molecular dynamicssimulations, we provided conclusive clues that the “water wetswater” process promotes water adsorption on the water layers,producing superwetting.

Author contributions: W.J. designed research; K.L., S.A., J.A., and W.J. performed re-search; Q.K. performed molecular dynamic simulations; K.L. and J.A. performed Kelvinprobe force microscopy measurement; K.L., Q.K., J.K., B.K., and W.J. analyzed data; andK.L. and W.J. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. A.L. is a guest editor invited by the EditorialBoard.1To whom correspondence should be addressed. E-mail: whjhe@snu.ac.kr.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1319001111/-/DCSupplemental.

5784–5789 | PNAS | April 22, 2014 | vol. 111 | no. 16 www.pnas.org/cgi/doi/10.1073/pnas.1319001111

Dow

nloa

ded

by g

uest

on

May

27,

202

1

Precise measurement of the height of the fully grown waterlayers by F is slightly complicated owing to the finite length ofFM itself. Nonetheless, one can still obtain the ultimate heightby F − E0, the difference between two meniscus-formation pointsfor FM and BM, assuming their radii of curvature (rc in Fig. 1A)are similar. This assumption is valid because the maximum dampingforce of FM (orange curve, Fig. 1E) is only ∼20% higher than thatof BM (black curve, Fig. 1C), indicating their similar sizes. Forexample, in Fig. 1E, the height (F − E0) is measured as 10 ± 0.5 nmat 200 μW/mm2. Meanwhile, the elapsed time between Ei’s and Di’s(maximum-damping points) was measured constant at ∼2.8 s for allof the growing layers (i = 1–5) in Fig. 1 C and E, which implies ittakes ∼2.8 s until full growth (note that maximum damping occurswhen the water layer is fully grown; Materials and Methods). In-terestingly, the elapsed time (∼2.8 s) times the measured growthrate (∼3.4 nm/s) yields ∼9.5 nm, which agrees well with the mea-sured height (∼10 nm), validating our in situ measurements. Noticethat the good linear fits versus Di’s (dashed red lines, Fig. 1 C and

E) suggest the damping originates from the adhesion of the waterlayer (Materials and Methods).Fig. 2A shows that the ultimate height of the fully grown

water layers increases with the UV intensity (up to ∼20 nm at∼500 μW/mm2) but is almost independent of relative humidity(RH). Fig. 2B presents the increase of growth rate versus theintensity as well as RH. These results indicate the light-induced,long-range attractive interaction contributes dominantly to photo-adsorption of water, which is supported by the following obser-vations. (i) The results of Fig. 2A confirm the water-layer growthis the long-range, light-induced process. (ii) The higher growthrate at the higher intensity and RH (Fig. 2B) suggests the exis-tence of increasingly stronger (i.e., long-range) attraction betweenwater molecules and R-TiO2, which accelerates photo-adsorp-tion. (iii) Water adsorption disappears for grazing incidence oflight, implying other ambient effects such as vibrational excitationof water molecules are negligible.

Photo-Adsorption by Delocalized Shallow-Trap Electrons. Now oneneeds to clarify the physical processes that result in such a light-induced, long-range attraction. It is well known that when theelectron-hole pairs are created by UV absorption subsequentcharge transfer to the TiO2 surface generates electrons trappedin the intrinsic defects of TiO2 associated with O2 vacancies (16).We have confirmed the UV-generated electrons are indeeddelocalized in the shallow traps by using Kelvin probe forcemicroscopy (KPFM) for anatase TiO2 (A-TiO2) and R-TiO2(SI Text and Fig. S5). Meanwhile, the homogeneous electric fieldsproduced by the delocalized electrons result in the field-inducedattraction between polar water molecules and TiO2, which wassuggested as a possible source for the long-range attractive in-teractions in TiO2 (17). For unified understanding of these twoseparately known processes (i.e., delocalized electrons and photo-adsorption), one has to investigate the roles of the shallow elec-tron-trapping states associated with O2 vacancies in the light-induced water adsorption process.For this purpose, we have performed photo-adsorption ex-

periments both with and without O2-annealing treatment onA-TiO2 (Fig. S6) as well as R-TiO2, under wide laser irradiationcovering visible, NIR, and UV spectra. Fig. 3A shows, forvarious values of intensity, the ultimate height of water layerson A-TiO2 decreases sharply at ∼445-nm wavelength, stays nearlyconstant until ∼635 nm, and then gradually diminishes before∼980 nm. With O2 treatment, however, the height on the O2-treated A-TiO2 (O2) becomes almost half compared with the

Fig. 1. UV-induced growth of water layer on R-TiO2. (A) Schematics of light-induced water adsorption on R-TiO2. (i) The capillary-condensed BM of wateris formed in the tip-surface nano-gap without laser. (ii) UV laser is illuminatedright after BM is formed, producing the growing water layer. (iii) The FM isformed in the gap between the tip and the fully grown water layer, obtainedwhen laser is illuminated before BM is formed. (B) The tip-oscillation ampli-tude versus the tip-sample distance (z) for three different tip-retractionspeeds. The tip-approach speed is constant ∼1.15 nm/s, time interval betweentwo successive approaches is 200 ms, RH is ∼50 ± 0.2%, ambient temperatureis 21.5 ± 0.1 °C, and UV intensity is ∼200 μW/mm2. The BM is formed at E0 (redarrow). (C) The associated damping force is obtained by the equation of thetip motion that incorporates the tip amplitude (shown in B) and the corre-sponding phase signal (Fig. S4C). (D) The approach-retraction curves for var-ious illumination points Ei (i = 1,. . .,5) and the respective contact points Mi

between the tip and water layer. The tip approaches and retracts at ∼1.15nm/s. The growth rate is measured constant ∼3.4 ± 0.2 nm/s. The FM is formedat the identical contact point F when E6 is at any point farther from F, whichrepresents the fully grown water layer. Notice that the difference, F − E0, isthe ultimate height of water layer (A, iii). (E) The damping force for the datain D, derived in a similar manner (Fig. S4D). The good linear fits in C and E(dashed red lines) indicate the damping force results from the adhesion ofwater layers (Materials and Methods).

Fig. 2. In situ measurement of the growth rate and height of water layer. (A)The height of the fully grown water layer versus the intensity of UV laser (∼365nm) for several values of RH. Measurements were made at 20 different positionson the R-TiO2 surface. The height is determined mainly by the intensity, butalmost independent of RH. Notice that the height keeps increasing with furtherincrease of intensity, but thermal drift, distorting the approach-retraction curves,limits accurate measurements. (B) The corresponding growth rate of water layersincreases with intensity as well as with RH, in contrast to the case of height.These results indicate the existence of the long-range, light-induced attractiveinteractions between water vapors and R-TiO2.

Lee et al. PNAS | April 22, 2014 | vol. 111 | no. 16 | 5785

APP

LIED

PHYS

ICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

May

27,

202

1

nontreated A-TiO2 and falls more rapidly near ∼800 nm (Fig.3B). The relatively suppressed photo-adsorption for A-TiO2 (O2)is observed most clearly in the UV region despite generally largeUV absorption for both films. This means that UV absorptionalone cannot produce photo-adsorption and also that O2annealing opposes the long-range attractive interaction. Noticethat O2 annealing results in a decreased number of trapping sitesassociated with oxygen vacancies (18) as well as decreased op-tical absorption in A-TiO2, as in Fig. 3D (16). Therefore, thereduced photo-adsorption in the UV region can be attributedto the reduced oxygen vacancies.In this regard, the photo-adsorption behaviors of A-TiO2 and

A-TiO2 (O2) in the visible and NIR regions described in Fig. 3Bagree qualitatively with the optical absorbance spectra in Fig. 3D,except above ∼900-nm wavelength, where significant NIR ab-sorption exists for both samples. Notice that the growth rate ofA-TiO2 (O2) is also decreased by ∼25% (∼80%) at ∼405 nm(∼635 nm) compared with A-TiO2. This indicates the decreasedlong-range attraction is also associated with the reduced oxygenvacancies. In particular, beyond ∼1,000 nm, photo-adsorption isfurther suppressed (Fig. 3B). This can be attributed to the factthat such NIR absorption, which depletes the shallow electron-trapping states that are located at ∼0.5 eV below the conduction-band edge (19), results in the diminished adsorption.Furthermore, we observed increased photo-adsorption in the

supplementary dynamic force microscopy measurement on theH2-annealed A-TiO2 (SI Text and Fig. S7A) compared with A-TiO2, which is clearly associated with the increased oxygen va-cancies and optical absorbance of H2-annealed TiO2 (Fig. S7B).Nonetheless, photo-adsorption is still suppressed beyond ∼1,000nm despite much more significant NIR absorption than A-TiO2. Itis probably because the optical absorption excites the electrons at

the trapping sites to the conduction band and thereby depletes thetrapped electrons.However, for R-TiO2 we observe different adsorption behav-

iors with respect to A-TiO2, which shows the unique resonancenear ∼780 nm (Fig. 4A). Fig. 4 B and C present the character-istics of R-TiO2 with and without O2 treatment. The measuredheight in the UV region is almost half decreased (Fig. 4B) as inthe A-TiO2 case, and the growth rate is also reduced by ∼45%(∼70%) at ∼405 (∼780) nm compared with the nontreatedR-TiO2 (Fig. 4C). Nonetheless, Fig. 4B does not reflect the bulkproperty of optical absorbance (Fig. 4D), where the effect of O2

treatment is not noticeable below ∼800 nm. This indicates theenhanced photo-adsorption on R-TiO2 peaked near ∼780 nm isalso correlated with the surface O2 defects, as detailed below.The visible and NIR optical bands of TiO2 are known to

contribute to the intrinsic defect states associated with O2 va-cancies that yield such traps as the F-type color centers and Ti3+

sites (16, 20, 21). Regarding the F-type center, optical bandsconsist of three centers: pure oxygen vacancy (F2+), single-elec-tron-trapped oxygen vacancy (F+), and two-electron-trappedoxygen vacancy (F). The calculated optical band of R-TiO2 is at∼760 nm for F+-center and ∼1.7 μm for F-center (22), whereasthe F2+-center state is even closer to the conduction band. No-tice that the energetic depth of F- and F2+-centers roughlymatches the defect states that are related to the shallow electrontraps (23) including Ti3+ (20, 21). Therefore, photogeneratedelectrons from the F+-centers, unlike those excited from theF-centers, migrate and stay captured in the shallow trapping sites(16), which probably contributes to the observed resonance near∼780 nm on R-TiO2, as discussed for the roles of the shallowtraps in photo-adsorption on A-TiO2.

Fig. 3. Photo-adsorption spectra of O2-annealed and nontreated A-TiO2.(A) The height of light-induced water layers versus the light wavelength,measured at more than 20 different locations on the A-TiO2 surface. Thewavelength ranges from ∼365 nm to ∼980 nm by using 12 laser diodes and3 diode-pumped solid-state lasers, and the intensity varies from 100 μW/mm2

to 500 μW/mm2. Photo-adsorption spectra is observed in the UV, visible,and NIR regions below ∼1,000 nm. (B) The layer height is decreased afterO2-annealing treatment on A-TiO2 (at 500 °C for 2 h), which is attributed tothe reduced number of the defect sites associated with O2 vacancies. (C) Thegrowth rate of water layers shows similar decrease for the O2-annealedA-TiO2. (D) UV-visible-NIR absorbance spectra for both A-TiO2 samples, con-verted from diffusive reflectance spectra. Suppressed photo-adsorption de-spite significant photo-absorption beyond ∼1,000 nm indicates the depletionof the shallow electron-trapping states resulting from NIR absorption.

Fig. 4. Photo-adsorption spectra of O2-annealed and nontreated R-TiO2. (A)The height of water layers versus the wavelength, measured at more than 20different positions on R-TiO2. Photo-adsorption is observable in the UV,visible, and NIR regions, with a local maxima at ∼780 nm. (B) Comparison ofthe height before and after O2 annealing on R-TiO2 (at 900 °C for 2 h). Theenhanced photo-adsorption near 780 nm is attributed to the electronscaptured in the shallow trapping sites, which are photogenerated from theF+-type color centers. (C) Comparison of the growth rate before and after O2

treatment on R-TiO2. (D) Optical absorbance spectra for both R-TiO2 samples,converted from diffusive reflectance spectra. Diminished photo-adsorptiondespite significant NIR absorption above ∼1,000 nm can be explained interms of the shallow electron-trapping states associated with O2 vacancies,similarly to the A-TiO2 case.

5786 | www.pnas.org/cgi/doi/10.1073/pnas.1319001111 Lee et al.

Dow

nloa

ded

by g

uest

on

May

27,

202

1

Superwetting by “Water Wets Water” Mechanism. Now we discussthe effects of the photo-adsorbed water layers on superwetting.To address the responsible mechanism, we have performed mol-ecular dynamics (MD) simulations to investigate the combinedeffects of our two main observations (i.e., delocalized surfaceelectrons and light-induced photo-adsorption). Here one mayconsider two possible mechanisms that lead to hydrophilic con-version: the electrowetting process (24, 25) and “water wetswater” process (26). In our simulation, we assume the homoge-neous surface is free of preadsorbed water clusters that are de-pendent on RH (27) and hydration layer that is observed onmetal oxides in solution (28, 29). Note that the porosity of films(Fig. S1) as well as the atomic level morphology of the surfacecan affect the contact angle in the heterogeneous surface (30,31). However, the contact angle on the homogeneous surfaceallows rather straightforward comparison between experiment andsimulation results (32), and many previous hydrophilic conversionexperiments on TiO2 were based on the homogeneous surfaces,which makes our assumptions probable.For electrowetting simulation, we apply the uniform vertical

electric field E and observe that the contact angle decreases asE increases from 0.3 to 0.9 V/nm, which corresponds to thesurface potentials estimated by our KPFM study (Fig. S5) butsaturates within ∼1 ns for all three facets (Fig. 5A). For morerealistic treatment of the delocalized surface electrons in the Ti3+

shallow traps, extra charge of ∼0.01e is assigned to each Ti atom,

so that total charge of ∼115.52e is distributed on the TiO2 layer.Here ∼7 V/nm field is applied on the surface by putting anotherTiO2 film, having +115.52 charge, at ∼30 nm separation. Asshown in Fig. 5B, the water droplet propagates extremely slowlycompared with Fig. 5A (and also Fig. 5C) even if E is an order ofmagnitude higher. Therefore, Fig. 5 A and B indicate that elec-trowetting contributes to slight reduction of the contact angle, butnot to superwetting.For the “water wets water” process, however, we now simulate

dynamic change of the contact angle for the water droplet in thepresence of a ∼1.1-nm-thick water layer on each surface, butwithout any E field. Fig. 5C clearly shows that the dropletbecomes completely wet on the thin water layer of three facets,as evidenced by rapid decrease of the initial contact angle to zerowithin ∼2 ns (note that the rather large error bar is due to co-existence of the layered region and precursor film). Conse-quently, one can conclude that superwetting of TiO2 results fromthe strong water–water attraction. When the nanodroplet is lo-cated on the 1 or 2 monolayer (ML) water film, superwettingdoes not occur because the droplet does not spread much andretains its shape with a decreased contact angle as shown in Fig.5D. This result has been discussed in previous studies on the roleof interfacial water in the adsorption process: “Water does notwet a water monolayer” (26, 32). However, when the waterdroplet is released on the 4 ML water, the droplet interactsdominantly with the topmost fourth layer and finally spreads out

Fig. 5. MD simulation of water–water attraction for superwetting of TiO2. MD simulation is performed on the stable rutile (110) [R(110)], rutile (101) [R(101)],and anatase (101) [A(101)] surface, which provides numerical estimate of the contact angle of a water droplet spreading on the photo-adsorbed water layers.(A) In compliance with the decreased surface potential ∼300 mV for A-TiO2 during UV illumination in KPFM (Fig. S5), the uniform external field E is appliedalong the z-direction with its magnitude 0.3 V/nm, 0.6 V/nm, and 0.9 V/nm for each surface. The contact angle is slightly decreased with E but is saturatedwithin ∼1 ns. (B) For more realistic charge distribution of the delocalized electrons (discussed in the text), each Ti atom is assigned by extra charge −0.01e.Then the maximum propagation radius Rmax of the water droplet increases slightly with time, but at a much lower speed compared with A despite an order ofmagnitude larger E, which excludes the electrowetting origin of superwetting. (C) The contact angle of the water nanodroplet decreased rapidly down tozero within ∼2 ns on the ∼1-nm-thick (4 ML) photo-adsorbed water layer in the absence of E, which evidences the strong “water wets water” mechanismresponsible for superwetting of TiO2. (D) The spreading of the nanodroplet on the 1, 2, 3, and 4 ML of water are simulated to confirm the condition ofsuperwetting for the three surfaces. Note that when the contact angle of the droplet decreases smaller than 10° the droplet begins to meet its periodic imagefor the above system size. However, the 10° threshold angle is sufficient to reproduce the light-induced superwetting of TiO2 (39). Therefore, we define theoccurrence of superwetting at the contact angle when the droplet meets the periodic image.

Lee et al. PNAS | April 22, 2014 | vol. 111 | no. 16 | 5787

APP

LIED

PHYS

ICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

May

27,

202

1

to produce superwetting for all of the three facets. Notice thatthe fourth layer is found unordered and has almost the samesurface tension value as the bulk water, whereas the first layershows a very ordered and solid-like property (SI Text and Figs. S8and S9). Therefore, the ∼1-nm water layer (or 4 ML), which ismuch thinner than the measured water thickness obtained by thelight-induced photo-adsorption, is still thick enough to com-pletely wet the nanodroplet. The larger microsized water drop-lets naturally take a longer time to be superwetted (for example,more than an hour under UV intensity less than 10 μW/mm2)(2, 10), which can speed up at higher light intensity and RH, asdemonstrated in the present study. Considering the photo-adsorbed layers are determined dominantly by the amount ofoxygen vacancies despite differing wettability of the facets (33)and superwetting is also observed for water layers thicker than3∼4 ML regardless of the choice of TiO2 surface, the “water wetswater” mechanism may be generally satisfied for different facets.In conclusion, we have presented conclusive evidences that

the “water wets water” mechanism associated with the photo-adsorbed water layers, produced by the long-range attractionowing to the delocalized surface electrons, is responsible forsuperwetting on TiO2. The hydrophilic conversion observed inthe doped TiO2 (34) or other photo-catalytic n-type metal oxideswith oxygen vacancies (35) such as ZnO and WO3, which gen-erate delocalized electrons, may be better understood by a simi-lar photo-adsorption process.

Materials and MethodsTiO2 Film. Nanocrystalline anatase (A-TiO2, blue) and rutile (R-TiO2, white)films were prepared by depositing TiO2 paste onto the Ti substrate viadoctor blading, followed by annealing at 500 °C (A-TiO2) and 900 °C (R-TiO2)for 6 h in air. Their surface characteristics and crystal structures wereobtained by field-emission scanning electron microscopy (Fig. S1), X-raydiffraction (Fig. S2), and X-ray photoemission spectroscopy (Fig. S3, A-TiO2),respectively. The TiO2 films are tightly glued to the sample holder andloaded to the atomic force microscope system in a sealed metal chamber.The RH inside the chamber is controlled by flowing pure N2 gas through thebubbler filled with deionized water (>18.3 MΩ). The RH is continuouslymonitored by a sensor (SHT75, specified accuracy <1.8%; Sensirion) placednear the sample holder.

QTF Probe. A silica tip (curvature radius ∼25 nm) was glued to one prong ofthe QTF oscillator (Fig. S4A). The QTF probe allows noncontact (free-oscil-lation amplitude ∼0.55 nm) and sensitive force-gradient (∼0.01 N/m) de-tection at variable height and speed owing to its high stiffness (∼27,000Nm−1) and quality factor ðQ∼ 7,000Þ at the resonant frequency ðf0 ∼ 32 kHzÞ

in ambient conditions (Fig. S4B). The relaxation time constant ð2Q=2πf0Þ is∼68 ms. During DFMmeasurement, the fluctuation of the piezo displacementis controlled within ∼0.02 nm, and the thermal drift is measured less than∼1.2 Å/min, so that the distance-control precision is high enough for reliableand repeatable experiments.

Adhesion Force of Water Layers. The tip-sample adhesion force is derived bythe Laplace force (15) that acts on the contact area A= πðr sinϕÞ2 (Fig. 1 A, ii),

F ≈−AγL=rc =−πrdð1+ cosϕÞγL=rc, [1]

where γL is the surface tension and d is the height of the wetted part of tip,d = rð1− cosϕÞ. If the curvature rc does not change significantly during tipretraction, F is proportional to d because 1:75< ð1+ cosϕÞ< 2:0 for the layerheight ∼10 nm. The good linear fit by 10.8(9)-Di (nm) (dashed red lines, Fig. 1C and E) indicates (i) the damping force exhibits linear behavior versus d (Eq. 1)and (ii) the line represents the maximum value of d for each Di where maxi-mum damping occurs. Notice that the ordinate 10.8(9) (nm) agrees to theultimate height ∼10 nm, which justifies the use of the damping force (Fig. 1 Cand E) as the adhesion force between tip and water layer.

MDSimulation. Simulation is performedwithGROMACS code and time step of1.0 fs in a canonical ensemble at 300 K temperature using a Berendsenthermostat. With the extended simple point charge water model used andforce-field parameters for TiO2-H2O system adapted from ref. 36, the contactangle of the droplet is calculated by methods similar to those in ref. 37.Dividing the droplet into meshes with each dimension of 0.5 nm (horizontal) ×0.25 nm (vertical), we record the coordinates (z,r) of the meshes where the waterdensity falls to 0.3 g/cm3, and then obtain the density profile by fitting the pa-rabola z = a + br + cr2, which provides the contact angle θ by tan θ = dz/drjz=0.We consider the water droplet consisting of 2,180 molecules on rutile (110) of18.8 × 19.1945 nm2, rutile (101) of 18.778 × 18.944 nm2, and anatase (101) of18.925 × 18.432 nm2, respectively. The three facets are selected from the first andsecond peaks of X-ray diffraction spectra (Fig. S2). The contact angle dependencyon the droplet size was reported previously (30, 31). We checked the contactangle variation by using the droplet consisting of 2,180, 4,000, and 8,000 watermolecules for facets of the above size as well as twice-larger size. The resultingcontact-angle variation is negligible within the error bar and does not exceed 4°from 36.2° of rutile (110), 39.7° of rutile (101), and 25.02° of anatase (101) for allof the considered systems of different sizes. We use the unit called theML (38) todefine the thickness of the water thin film; 1–4ML of water film is assumed to beadsorbed on the TiO2 surface and be equilibrated. After equilibration, thedroplet of 2,180 water molecules is located on the film and the subsequentsimulation is carried out.

ACKNOWLEDGMENTS. K.L. thanks Young S. Park for his help with samplepreparation. This work was supported by the National Research Foundationof Korea grant funded by the Korea government (Ministry of Science, ICTand Future Planning) (No. 2013-056344).

1. Wang R, et al. (1997) Light-induced amphiphilic surfaces. Nature 388:431.2. Thompson TL, Yates JT, Jr. (2006) Surface science studies of the photoactivation of

TiO2—new photochemical processes. Chem Rev 106(10):4428–4453.3. Nakamura R, Ueda K, Sato S (2001) In situ observation of the photoenhanced ad-

sorption of water on TiO2 films by surface-enhanced IR absorption spectroscopy.

Langmuir 17(8):2298–2300.4. Nosaka AY, et al. (2003) Photoinduced changes of adsorbed water on a TiO2 photo-

catalytic film as studied by 1HNMR spectroscopy. J Phys Chem B 107:12042–12044.5. Lee FK, Andreatta G, Benattar J (2007) Role of water adsorption in photoinduced

superhydrophilicity on TiO2 thin films. Appl Phys Lett 90:181928.6. Wang CY, Groenzin H, Shultz MJ (2003) Molecular species on nanoparticulate anatase

TiO2 film detected by sum frequency generation: Trace hydrocarbons and hydroxyl

groups. Langmuir 19(18):7330–7334.7. White JM, Szanyi J, Henderson MA (2003) The photon-driven hydrophilicity of titania:

A model study using TiO2(110) and adsorbed trimethyl acetate. J Phys Chem B 107(34):

9029–9033.8. Zubkov T, et al. (2005) Ultraviolet light-induced hydrophilicity effect on TiO2(110)(1 x 1).

Dominant role of the photooxidation of adsorbed hydrocarbons causing wetting by

water droplets. J Phys Chem B 109(32):15454–15462.9. Kanta A, Sedev R, Ralston J (2005) Thermally- and photoinduced changes in the water

wettability of low-surface-area silica and titania. Langmuir 21(6):2400–2407.10. Takeuchi M, Sakamoto K, Martra G, Coluccia S, Anpo M (2005) Mechanism of pho-

toinduced superhydrophilicity on the TiO2 photocatalyst surface. J Phys Chem B

109(32):15422–15428.11. Nakajima A (2011) Design of hydrophobic surfaces for liquid droplet control. NPG

Asia Mater 3:49–56.

12. Garcia R, Perez R (2002) Dynamic atomic force microscopy method. Surf Sci Rep 47:197–301.

13. Choe H, et al. (2005) Formation, manipulation, and elasticity measurement of ananometric column of water molecules. Phys Rev Lett 95(18):187801.

14. Lee M, Jhe W (2006) General theory of amplitude-modulation atomic force micros-copy. Phys Rev Lett 97(3):036104.

15. Israelachivili JN (2011) Intermolecular and Surface Forces (Academic, London), 3rd Ed,pp 456–457.

16. Kuznetsov VN, Serpone N (2009) On the origin of the spectral bands in the visibleabsorption spectra of visible-light-active TiO2 specimens analysis and assignments.J Phys Chem C 113(34):15110–15123.

17. Szczepankiewicz SH, Moss JA, Hoffmann MR (2002) Electron traps and the stark effecton hydroxylated titania photocatalysts. J Phys Chem B 106(31):7654–7658.

18. Sekiya T, et al. (2004) Defects in anatase TiO2 single crystal controlled by heattreatments. J Phys Soc Jpn 73(3):703–710.

19. Boschloo G, Fitzmaurice D (1999) Spectroelectrochemical investigation of surfacestates in nanostructured TiO2 electrodes. J Phys Chem B 103(12):2228–2231.

20. Wendt S, et al. (2008) The role of interstitial sites in the Ti3d defect state in the bandgap of titania. Science 320(5884):1755–1759.

21. Kowalski PM, Camellone MF, Nair NN, Meyer B, Marx D (2010) Charge localizationdynamics induced by oxygen vacancies on the TiO2(110) surface. Phys Rev Lett105(14):146405.

22. Chen J, Lin LB, Jing FQ (2001) Theoretical study of F-type color center in rutile TiO2.J Phys Chem Solids 62:1257–1262.

23. Gosh HN, Asbury JB, Lian T (1998) Direct observation of ultrafast electron injectionfrom coumarin 343 to TiO2 nanoparticles by femtosecond infrared spectroscopy.J Phys Chem B 102(34):6482–6486.

5788 | www.pnas.org/cgi/doi/10.1073/pnas.1319001111 Lee et al.

Dow

nloa

ded

by g

uest

on

May

27,

202

1

24. Daub CD, Bratko D, Luzar A (2011) Electric control of wetting by salty nanodrops:Molecular dynamics simulations. J Phys Chem C 115(45):22393–22399.

25. Yuan Q, Zhao YP (2010) Precursor film in dynamic wetting, electrowetting, andelectro-elasto-capillarity. Phys Rev Lett 104(24):246101.

26. Wang C, et al. (2009) Stable liquid water droplet on a water monolayer formed atroom temperature on ionic model substrates. Phys Rev Lett 103(13):137801.

27. Hu J, Xiao XD, Ogletree DF, Salmeron M (1995) Imaging the condensation andevaporation of molecularly thin films of water with nanometer resolution. Science268(5208):267–269.

28. Fukuma T, Higgins MJ, Jarvis SP (2007) Direct imaging of lipid-ion network formationunder physiological conditions by frequency modulation atomic force microscopy.Phys Rev Lett 98(10):106101.

29. Argyris D, Phan A, Striolo A, Ashby PD (2013) Hydration Structure at the α−Al2O3

(0001) surface: Insights from experimental atomic force spectroscopic data and at-omistic molecular dynamics simulations. J Phys Chem C 117(20):10433–10444.

30. Wang J, Bratko D, Luzar A (2011) Probing surface tension additivity on chemically het-erogeneous surfaces by a molecular approach. Proc Natl Acad Sci USA 108(16):6374–6379.

31. Bratko D, Daub CD, Leung K, Luzar A (2007) Effect of field direction on electrowettingin a nanopore. J Am Chem Soc 129(9):2504–2510.

32. Phan A, Ho TA, Cole DR, Striolo A (2012) Molecular structure and dynamics in thinwater films at metal oxide surfaces: Magnesium, aluminum, and silicon oxide surfa-ces. J Phys Chem C 116(30):15962–15973.

33. Almeida AR, Calatayud M, Tielens F, Moulijn JA, Mul G (2011) Combined ATR-FTIR andDFT study of cyclohexanone adsorption on hydrated TiO2 anatase surfaces. J PhysChem C 115(29):14164–14172.

34. Asahi R, Morikawa T, Ohwaki T, Aoki K, Taga Y (2001) Visible-light photocatalysis innitrogen-doped titanium oxides. Science 293(5528):269–271.

35. Miyauchi M, Nakajima A, Watanabe T, Hashimoto K (2002) Photocatalysis and photo-induced hydrophilicity of various metal oxide thin films. Chem Mater 14(6):2812–2816.

36. Bandura AV, Kubicki JD (2003) Derivation of force field parameters for TiO2-H2Osystems from ab initio calculations. J Phys Chem B 107(40):11072–11081.

37. Giovambattista N, Debenedetti PG, Rossky PJ (2007) Effect of surface polarity on watercontact angle and interfacial hydration structure. J Phys Chem B 111(32):9581–9587.

38. Castrillón SRV, Giovambattista N, Aksay IA, Debenedetti PG (2011) Structure andenergetics of thin film water. J Phys Chem C 115(11):4624–4635.

39. Ohler B, Langel W (2009) Molecular dynamics simulations on the interface betweentitanium dioxide and water droplets: A new model for the contact angle. J PhysChem C 113(23):10189–10197.

Lee et al. PNAS | April 22, 2014 | vol. 111 | no. 16 | 5789

APP

LIED

PHYS

ICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

May

27,

202

1