Control of the polarization of ferroelectric capacitors by the concurrent action of light and adsorbates

. Ferroelectric perovskites hold promise of enhanced photovoltaic efficiency and photocatalytic activity. Consequently, the photoresponse of oxide ferroelectric thin films is an active field of research. In electrode/ferroelectric/electrode devices, internal charge in the ferroelectric, free charge in the electrodes, and buried adsorbates at interfaces combine to screen the ferroelectric polarization and to stabilize the polar state. Under illumination, photo-induced carriers and photo-dissociated adsorbates may disrupt the screening equilibrium, modifying the switchable polarization and altering its expected benefits. Here, we explore the photoresponse of BaTiO 3 thin films in a capacitor geometry, focusing on the effects of visible illumination on the remanent polarization. By combining ferroelectric and X-ray photoelectron spectroscopy, we discover that photoreaction of charge-screening H 2 O-derived adsorbates at the buried metal-ferroelectric Pt/BaTiO 3 interface has an unexpected pivotal role, enabling a substantial modulation (up to 75%) of the switchable remanent polarization by light. These findings illustrate that the synergy between photochemistry and photovoltaic activity at the surface of a ferroelectric material can be exploited to tune photoferroelectric activity.

ABSTRACT.Ferroelectric perovskites hold promise of enhanced photovoltaic efficiency and photocatalytic activity.Consequently, the photoresponse of oxide ferroelectric thin films is an active field of research.In electrode/ferroelectric/electrode devices, internal charge in the ferroelectric, free charge in the electrodes, and buried adsorbates at interfaces combine to screen the ferroelectric polarization and to stabilize the polar state.Under illumination, photo-induced carriers and photo-dissociated adsorbates may disrupt the screening equilibrium, modifying the switchable polarization and altering its expected benefits.Here, we explore the photoresponse of BaTiO3 thin films in a capacitor geometry, focusing on the effects of visible illumination on the remanent polarization.By combining ferroelectric and X-ray photoelectron spectroscopy, we discover that photoreaction of charge-screening H2O-derived adsorbates at the buried metalferroelectric Pt/BaTiO3 interface has an unexpected pivotal role, enabling a substantial modulation (up to 75%) of the switchable remanent polarization by light.These findings illustrate that the synergy between photochemistry and photovoltaic activity at the surface of a ferroelectric material can be exploited to tune photoferroelectric activity.
The non-volatile polarization of ferroelectric materials and their piezoelectric response have been long investigated and exploited in memory elements, sensors and actuators. 13] Those ideas are at the root of the renewed interest in the photoresponse of ferroelectrics and the emerging field of photoferroelectrics. [4][5][6][7][8][9][10][11] The typically large band gaps of ferroelectric oxides and their limited photovoltaic efficiency have been challenging bottlenecks for their implementation in photovoltaics.However, in recent times, the observation of larger than band-gap open circuit photo-voltages in ferroelectric thin films [12][13][14][15] or photocurrent switchable by electric fields, 16 and the fresh experimental report of engineered ferroelectric oxides with band gaps in the visible range [17][18] have greatly enhanced the prospects for this class of photovoltaics.
Ferroelectric materials are characterized by the presence of spontaneous polarization (P) and concomitant surface charges.Due to energy considerations, typically ferroelectric materials break up into polar domains with different polarization directions.In the case of the prototypical tetragonal phase of BaTiO3 (BTO), domains may form up/down patterns, with the correspondingly positive/negative surface charges, along the tetragonal c-axis direction.However, to stabilize large ferroelectric domains, the surface charge should be properly screened.Screening may be accomplished externally (by charges provided by metallic electrodes, adsorbates, etc) or internally (by cationic non stoichiometries, oxygen vacancies, space charge regions, etc). 190][31][32][33][34][35] By the same token, ferroelectric polarization can determine surface properties providing polarization-sensitive adsorption sites for physisorbed and chemisorbed adsorbates, [36][37][38][39] dissociation, [40][41] oxidation/reduction processes, [42][43][44] and different surface terminations. 38,45 nder suitable illumination, ferroelectric materials may display radically different ferroelectric properties.Three main mechanisms are at play.First, photoinduced electronhole pairs within the ferroelectric may provide internal screening, thus reducing the surface potential contrast; [46][47] second, filling pre-existing charge-traps can modify domain pinning and imprint; 48 third bulk photovoltaic effects can drive free carriers to surfaces, enhancing or suppressing polarization. 49verall, detailed knowledge exists on the role of adsorbates and the photoresponse of free ferroelectric surfaces.However, practical photovoltaic devices require the fabrication and use of metal/ferroelectric/metal structures with capacitor configuration.In such structures, metallic electrodes and buried adsorbates at metal/ferroelectric interfaces come into play, disturbing screening mechanisms and thus electric field distributions in the devices, whose impact on device photoresponse is largely unknown.In seminal papers, Land and Peercy [50][51] and Dimos and Warren [52][53][54][55] reported that under UV irradiation, the polarization loops of some ferroelectric capacitors, made from polycrystalline samples, displayed remarkable changes most notably a dramatic reduction of the polarization and a modification of imprint.Although it was early proposed that the observed effects were related to charge trapping at grain boundaries, 52 this model was challenged by the observation of similar effects in BTO single crystals, 56 suggesting that point bulk defects may play a role.Kholkin 57 also reported a reduction of piezoresponse of PZT films under UV and suggested that a reduction of the depletion layer in the semiconducting PZT, and the concomitant inhomogeneous electric field inside the film, could stimulate the growth of domains of opposite polarity and a reduction of the measured polarization.Gruverman 58 reported the intriguing observation that the polarization loops of PbTiO3 capacitors and PbTiO3 bare films shift oppositely under UV illumination, suggesting a pivotal and distinct role of dielectric layers at ferroelectric surfaces and interfaces on the interaction between photoinduced carriers and polarization charges.
Here we aim to go a step further by addressing the role of adsorbates on the polarization of metal/ferroelectric/metal capacitors and on its sensitivity to sub-band gap photon irradiation (3.06 eV).Polarization loops of BTO capacitors were measured using semitransparent Pt electrodes deposited either on fresh or H2O-exposed BTO surfaces.We found that the remanent polarization is reduced under illumination, and the effect is more pronounced if the BTO surface has been exposed to H2O prior to electrode deposition.In an accompanying set of experiments, we have characterized a series of BTO films with electrodes deposited after exposing the samples to ambient conditions.Consistently, it is found that all films display a noticeable reduction of the remanent polarization.X-ray photoelectron spectroscopy shows that the concentration of OH - adsorbates varies (decreases) with photon illumination and correlates with the observed photosensitivity Pr, suggesting that buried adsorbates at the metal/ferroelectric interface are instrumental in the photoresponse.These findings indicate that embedded adsorbates in tandem with photocarriers play a dominant roles and determine the polarization of ferroelectric.

RESULTS AND DISCUSSION
Experiments were designed to modify the concentration of adsorbed species in a given film and explore the photosensitivity of its polarization.With that purpose, we exposed freshly BTO thin films grown by Pulsed Laser Deposition on La2/3Sr1/3MnO3(LSMO)-buffered (acting as bottom electrode) SrTiO3(001) (STO) substrates to water steam.The steam treatment consists of placing the film, at 100 o C, in a jet of water vapor (800 cm 3 •min −1 ) at atmospheric pressure for 10 hours.
After the steam treatment, the surfaces of the films were gently dried by a jet of N2 gas for 1 min to eliminate water residues.After subsequent growth of the Pt electrodes, we recorded the photoresponse of the remanent polarization, and it was compared to that of virgin (un-treated) samples whose electrodes were deposited just after BTO growth.In Figure 1a, we show the P-E loops of a virgin 109.5 nm Pt/BTO/LSMO/STO(001) capacitor.
As shown in Figure 1a, the loop recorded under illumination shows only minor variations when compared with the loop recorded in dark.The seemingly small increase of polarization is due to the larger leakage (photovoltaic) current under illumination.After this control experiment, we performed the aforementioned steam treatment, and subsequently deposited new Pt electrodes.In Figure 1b, we show the corresponding P-E loops recorded in dark and under illumination.Comparing the loops in dark before and after steam treatment, a small reduction of polarization and a sizeable increase of coercive field are observed, in agreement with the formation of a OH - layer at the BTO surface (Figure 1b, inset), further supported by dielectric measurements (see Supporting Information Figure S1), and by the small shift of P-E loops measured after steam treatment (see Supporting Information Figure S2).More relevant is the observation that the remanent polarization recorded under illumination is reduced compared with that recorded in dark.
Interestingly, the P-E loops recorded again in the dark and under illumination using the original electrodes after the sample has been steam-treated [Figure 1c] are virtually identical to the fresh contacts of Figure 1a.Therefore, the Pt electrode protects the underneath BaTiO3 surface from additional H2O adsorption [Figure 1c, inset].Comparison of data in Figure1a, b and c shows that water-derived adsorbates at the Pt/BTO interface are a key ingredient for the observed photoresponse.Note that, whereas the effect of steam treatment on remanent polarization value measured in dark conditions is small, the effect of steam treatment on the variation of polarization under illumination is large.Thus, the effect of water products at the interface is intriguingly relevant only under illumination conditions.
XPS spectra of the BTO/LSMO//STO film were recorded before [Figure 1d] and after [Figure 1(e)] the steam treatment.The spectra are dominated by a main O1s peak (I0) located at about 528.8 eV, characteristic of lattice O 2− (OL) ions in BTO 59 .Extending towards larger binding energies, a shoulder in the O1s peak is apparent.Previous XPS analyses of BTO 38,40,47,[59][60][61] and other perovskites 62 have shown that this feature results from the contribution of differently bonded oxygen species at the BTO surface.Three components occur at about: (I) 530 eV, (II) 531.5 eV and (III) 533 eV.Peak I (530 eV) has been attributed to protonated lattice oxygen (OL-H + ) 40,47 and/or BaCO3 60 .Peak II (531.5 eV) has been attributed to hydroxyl OH − groups bounded to surface cations 40,47 (most likely Ti 4+ 26, 40 ) and peak III (533 eV) is commonly assigned to oxygen hydroxyl groups either bounded to Ba 2+ (Ba-(OH)2) 59 or oxygen vacancies (probably present in our films, see Supporting Information Figures S3-4), 40 or physisorbed water H2O molecules. 62In brief, components II+III are related to the presence of OH -or H2O at the surface, and component I is related to a protonated surface and/or BaCO3.It can not be disregarded the hydrogen incorporation into the BaTiO3 lattice by steam treatment.7][68] None of those conditions are employed here and thus hydrogen insertion should not significantly contribute.This is confirmed by the absence of relevant light absorption at lower photon energy (1.94 eV and 2.38 eV) 69 in our samples.Earlier experiments 64 reported hydrogen-related bands at about 2 eV.Therefore, we exclude a relevant role of hydrogen in the observed effects.
The comparison of data in Figure 1d and Figure 1e indicates that the position of the main I0 peak (OL in BTO) is virtually identical before and after the treatment.More interestingly, the highenergy shoulder of the main O1s peak shows an important increase after steam treatment.These spectra were analyzed to determine the different (I0, I, II, and III) contributions and their evolution due to steam treatment using suitable deconvolution procedures (see Methods).We used these data as proxy for their relative abundance.Supporting Information Table S1 summarizes the relative areas under the I0, I, II, and III peaks before and after steam treatment and the relative changes of remanent polarization upon illumination (defined as Pr = [Pr(dark) − Pr(illumination)]/Pr(dark), where Pr is the remanent polarization).Data in Supporting Information Table S1 show that the major difference in the spectra of Figures 1d,e is related to the II+III components, which significantly increase after steam treatment.As II and III O1s peaks are ascribed to the presence of H2O-related adsorbates, these results demonstrate that the steam treatment promotes their presence, and they are likely responsible for the enhanced Pr photoresponse.If one assumes that adsorbed molecular H2O is removed during electrode (Pt) deposition, it follows that the enhanced photoresponse after the steam treatment is due to the increased presence of chemisorbed OH − groups at the BaTiO3 surface, underneath the newly-grown Pt electrodes, after treatment.Although the AFM images collected after steam treatment show an increased presence of adsorbate-related products, no changes can be seen in the XRD pattern (see Supporting Information Figure S5).We repeated the steam treatment experiment in a Pt/BTO/LSMO/DSO and in a thinner Pt/BTO/LSMO/STO capacitor and we obtained similar results (see Supporting Information Figure S6).This points to adsorbate-induced modifications only at the surface.
To confirm the robustness of the fundamental role of buried adsorbates on the photosensitivity of polarization in ferroelectric capacitors, a series of BTO samples of thicknesses 36.5, 73 and 109.5 nm were grown on LSMO-buffered STO and DyScO3 (DSO) substrates and left in ambient conditions for several days before the growth of Pt electrodes.First, we collected XPS spectra of all these BTO films (shown in Supporting Information Figure S7).Focusing on the evolution of the O1s spectra for different films, the relative contribution of each component (I, II, and III and I0) to the whole O1s spectrum was determined (see Methods in Supporting Information) and used as measure of their relative abundance.The extracted relative fractions of I0, I, II, III contributions to the O1s peak and the corresponding (I0 and I) values for Ba3d5/2 are summarized in Supporting Information Table S2.The most important trend is the observed increase of the (II+III) ((OH − /H2O)) contribution when increasing film thickness for films grown on both STO and DSO substrates, although more pronounced on DSO.We ascribe the more prominent (II+III) ((OH − /H2O)) contribution on thicker films and films grown on DSO to an increase of films' effective surface area, as revealed by roughness measurements performed by AFM (see Supporting Information Figure S8).
The P-E loops of all BTO/LSMO//STO and BTO/LSMO//DSO samples were collected in dark and under illumination (see Supporting Information Figure S9 and the corresponding Pr values included in Supporting Information Table S2).In Figure 2 we plot the relative intensity of II+III (OH − /H2O) components, extracted from the XPS spectra, versus the corresponding ΔPr.Data in Figure 2 show that ΔPr increases when the relative fraction of (II + III) (OH − /H2O) increases.
Importantly, this trend holds for BTO films for all thicknesses and both substrates.For completeness, the II+III areas and Pr of the steam treated sample of Figure 1 is also included in  When not visible, error bars are smaller than the data point size.

II + III (% area)
P r

(%)
We remind here that XPS spectra give information on the exposed BTO part at the film surface, and ferroelectric characterization is performed using Pt-capped capacitors on the very same film.
Therefore, the relative I0, I, II and III contributions may not represent their actual concentration at the buried BTO surface under the Pt electrode.However, as data in Figure 2 shows that for all samples, including steam treated samples, there is a clear connection between the water-related adsorbate contribution to the XPS spectra and ΔPr, we propose that there is a close link between ΔPr and the nature and relative abundance of adsorbates.
The hysteresis loops and the XPS data confirm the presence of water and dissociated metalbonded hydroxyls OH − and probably OL-H + protons at the buried surface of ferroelectric Ptcovered BTO films.The buried adsorbates modulate the magnitude of the switchable polarization and the polarization variation under illumination (Figures 1 and 2).Therefore, it is crucial to assess any change of adsorbate characteristics under illumination.Spectra of the whole set of samples display similar effects as summarized in Figure 3b, where the relative fractions of (II+III) in dark and under illumination for all samples are plotted.Detailed observation of Figure 3b shows that the relative decrease of the (II+III) fraction in the XPS spectra is more pronounced for samples grown on DSO and for the thicker samples.Recall that the same trend was observed for the phororesponse (Pr) of films on STO and DSO substrates, which was found to be larger in the latter (Figure 2).To emphasize this observation, in Figure 3c, we plot the relative variation of the (II+III) fraction under illumination [(II+III) = [II+III(illumination)]-II+III(dark)] versus Pr.These data strongly suggest that there is a correlation between the decrease of the (II+III) fraction in the XPS spectra, related to (OH − /H2O) adsorbates, under illumination and the corresponding polarization reduction.The charge redistribution relevant role may occur at both interfaces Pt/adsorbates/BTO and BTO/LSMO.In the first case, charge accumulation at Pt/BTO interface will weaken adsorbates chemisorption energy, making the OH -and H + adsorbates less stable and favoring their oxidation/reduction to forming water molecules, with subsequent reconstruction of the dielectric interface layer at this interface [Figure 4b].In this case, OH − /H + are instrumental to favor photocarrier accumulation and polarization reduction, in agreement with the effect of water-related adsorbates at the surface on Pr.In the second case, photocarriers will accumulate at the BTO/LSMO interface and may induce the formation of a dielectric layer at the LSMO surface by charge depleting it.However, as this scenario does not require the presence of water-related adsorbates at the Pt/BTO interface, which we have shown to be relevant, we disregard it.In both cases the polarization destabilization can result from the increase of available charges resulting in a decrease of polarization or to its trapping by defects resulting in a polarization internal screening, 3 the fact that the reduction of polarization is persistent after illumination suggests that carriers trapping may play an important role in the studied system (see Supporting Information Figure S10).
The presence of a newly-formed, photo-induced dielectric interface layer of thickness d and permittivity d at the surface of a ferroelectric film has a strong impact on the polarization loops.
It is well-known that the presence of a dielectric interface layer may produce a reduction of Demonstration of the gradual introduction of the added by light dielectric layer can be inferred from the gradual reduction of Pr for increasing illumination time and light power (see Supporting Information Figure S12).

CONCLUSIONS
We have explored the response of the ferroelectric polarization of Pt/BTO/LSMO capacitors upon blue-laser illumination.We have observed that there is a substantial reduction of remanent polarization under illumination, which appears to be directly correlated with the relative concentration of dissociated OH -and H + adsorbates chemisorbed to the buried BTO surface.We argue that whereas adsorbates underneath the top electrodes introduce a dead layer that increases the coercive electric field in dark conditions, these adsorbates under illumination are released.The adsorbates release is ascribed to the fact that they feel less attraction to the BaTiO3 surface because BaTiO3 surface polarization is partially screened by the photoexcited carriers.The released adsorbates generate a dielectric dead layer that produces a decrease of switchable polarization.Insitu XPS measurements have indeed confirmed that the concentration of OH -is reduced under illumination, suggesting its oxidation and the formation of a dielectric layer that ultimately decreases the measured polarization.More generally, these findings illustrate how the synergy between photochemistry and photovoltaic activity at the surface of a ferroelectric material can be exploited to tune photoferroelectric activity, providing a simple recipe to manipulate it.

Samples growth.
A series of BTO (001) films of thicknesses 36.5, 73 and 109.5 nm where grown by pulsed laser deposition on SrTiO3(001) (STO) and DyScO3(110) (DSO) substrates buffered with a 30 nm La2/3Sr1/3MnO3 (LSMO) film acting as bottom electrode.Growth details were reported elsewhere. 69,72 TO and DSO have a cubic structure and produce a nominal compressive mismatch with BTO of 2.20% and 1.23%, respectively.Pt top contacts (20 nm thick, showing around 14% transparency, [73][74] and 60 x 60 m 2 width) were grown ex-situ by room-temperature sputtering on as-grown BTO surfaces, using distinct experimental protocols as described in the text.
Ferroelectric characterization.Polarization-electric field (P-E) hysteresis loops were measured in the top-top configuration thus cancelling imprint electric fields that could arise from any asymmetry between top and bottom electrodes, and allowing to obtain symmetric P-E loops. 69,75  P-E loops were measured by applying an electric bias on one Pt electrode and grounding an adjacent one, spaced by 15 m, and either in dark or under illumination with blue laser (3.06 eV, 10 W•cm −2 ) with ≈300 m diameter spot, large enough to illuminate two adjacent Pt electrodes homogeneously.The measurement frequency, selected following the protocols described elsewhere, 75 was 1kHz in all cases where leakage and other extrinsic contribution are found to be small.

X-ray photoelectron spectroscopy characterization.
The surfaces of films were analyzed by Xray photoelectron spectroscopy (XPS) collected either in dark or under illumination using the same blue laser.
Further experimental details are in Methods included in Supporting Information.

ASSOCIATED CONTENT
The following files are available free of charge.
Detailed methods, raw data extracted from XPS fittings, sketch of the experimental set-up used for the steam treatment experiments, and complementary structural, morphological, XPS, dielectric and photoelectric/ferroelectric characterization is included in Supporting Information (PDF).

Figure 2 .
Figure 2. Dependence of the change of remanent polarization with light (Pr) on the relative

Figure 4 .
Figure 4. (a) Sketch of the polarization screening in a Pt/BaTiO3 where the metallic electrode has been grown ex-situ after ambient exposure.Chemisorbed OH − and H + radicals, bonded to the polarization.Using the simple model proposed by Tagantsev and Gerra71 of a dielectric layer of thickness and permittivity (d, d) attached to a ferroelectric film of thickness t and polarization P, polarization loops can be calculated and used to inspect the effects of the dielectric layer.The dependence of P(E,(d, d)) for a broad range of (d, d) values can be found in Supporting Information FigureS11.In Figure4c, we show the P(E) loops calculated for a few illustrative values of (d, d) for a BTO film of thickness 109.5 nm as the characterized in Figure1.Data in this plot clearly illustrate that the dielectric layer reduces the measured polarization.The (d = 1 nm, d= 10) and (d = 5 nm, d= 76) values used in Figure3care selected to mimic the observed changed of polarization (Pr ≈ 60 %) (Figure1).It thus follows that the reduction of polarization under illumination can be ascribed to the formation of a dielectric layer of d within (10-76) and a thickness of (1-5) nm.The formation of this dielectric layer is compatible with the observed reduction of the OH -concentration under illumination, the charge accumulation at the Pt/BTO interface and the relevant role of water related adsorbates presence at the surface and provides a simple description of the photosensitivity of the polarization in Pt/BTO/LSMO heterostructures.