Photoresponse dynamics in amorphous-LaAlO /SrTiO interfaces 3 3 Emiliano Di Gennaro,1,∗ Ubaldo Coscia,2 Giuseppina Ambrosone,1 Amit Khare,1 Fabio Miletto Granozio,1 and Umberto Scotti di Uccio1 1Dipartimento di Fisica, Univ. di Napoli Federico II and CNR-SPIN, Compl. Univ. di Monte S. Angelo, Via Cinthia I-80126 Napoli (Italy) 2Dipartimento di Fisica, Univ. di Napoli Federico II and CNISM Unit´a di Napoli, Compl. Univ. di Monte S. Angelo, Via Cinthia I-80126 Napoli (Italy) (Dated: January 15, 2015) 5 The time-resolved photoconductance of amorphous and crystalline LaAlO3/SrTiO3 interfaces, 1 bothhostinganinterfacial 2-dimensionalelectron gas,isinvestigatedunderirradiation byvariable- 0 wavelengths, visible or ultraviolet photons. Unlike bare SrTiO3 single crystals, showing relatively 2 small photoconductance effects, both kinds of interfaces exhibit an intense and highly persistent photoconductance with extraordinarily long characteristic times. The temporal behaviour of the n extra photoinduced conductance persisting after light irradiation shows a complex dependence on a interface type(whether amorphous or crystalline), sample history and irradiation wavelength. The J experimental results indicate that different mechanisms of photoexcitation are responsible for the 4 photoconductance of crystalline and amorphous LaAlO3/SrTiO3 interfaces undervisible light. We 1 propose that the response of crystalline samples is mainly due to the promotion of electrons from the valence bands of both SrTiO3 and LaAlO3. This second channel is less relevant in amorphous ] l LaAlO3/SrTiO3, where the higher density of point defects plays instead a major role. e - Keywords: Interfaces,2DElectronGas,Photoconductivity, Oxides r t s . t The interface between the band insulators the oxygen vacancies are formed close to the interface, a LaAlO (LAO) and SrTiO (STO) can host a 2- duringsamplefabrication,becausethe oxidationofLAO m 3 3 dimensional electron gas (2DEG). In the seminal partially occurs by taking O atoms from the STO sub- - paper by Ohtomo and Hwang,[1] an epitaxial LAO strate. [11] Furthermore, oxygen vacancies can produce d n film was grown on a single crystal (001) STO substrate a 2DEG even when an ER cannot take place, as for in- o with single TiO termination. In this heterostructure, stanceintheinterfacebetweenoxygen-reducedSTOand 2 c the alternating LaO and AlO planes carry opposite vacuum.[12,13]Aspectaculardemonstrationofsuchbe- 2 [ charges, leading to a net polarization of the LAO film, haviorinLAO/STOinterfaceswasprovidedbythe work 1 in contrast with the non-polar state of STO. Such polar of Chen, et al., [14] showing that the interface between v discontinuity was indicated as a source of instability, crystallineSTO andamorphousLAOcanbe conductive. 4 resulting in the injection of electrons from the topmost Recently,theroleofoxygenvacanciesininterfacesbased 7 LAO layers into the Ti3d states at the interface, thus on either a crystalline (c), or an amorphous (a) LAO 3 giving rise to the interfacial 2DEG. [1] The above layers on STO was further clarified by observing the ef- 3 described ”electronic reconstruction” (ER) model is fects of thermal and ion irradiation treatments on both 0 . confirmed by ab initio computations[2–4] and provides kinds of samples.[15, 16] The results suggest that in c- 1 a natural explanation of some well established exper- LAO/STO the conductivity can be ascribed to both ER 0 imental results, as the need of TiO -terminated STO and oxygen vacancies, depending on the sample growth 5 2 substrates and a LAO film critical thickness of 4 unit conditions, whereas in a-LAO/STO only oxygen vacan- 1 : cells to achieve interface conductivity.[5, 6] Experiments cies play a role. v and theoretical calculations also agree on the bending Thesensitivityofc-LAO/STOtothelightwasdiscov- i X of the electronic bands of STO close to the interface, ered severalyears ago.[17–19] More recently, the investi- r as an effect of the local electric field. Such bending gation of the transport properties of irradiated samples a determines a triangular quantum well that confines the revealed its potential interest for both applicative and 2DEG within a few nanometers. [4, 7, 8] fundamental physics.[20–26] Photoconductivity was re- It was soon realized that, beside the ER model, other ported to be moderate when samples were illuminated mechanisms may explain the presence of free electrons by visible light, and it became considerably larger above populating of the quantum well. In particular, each aphotonenergythresholdofabout3.25eV(382nm),[20] oxygen vacancy introduced in STO acts as an electron correspondingtotheindirectbandgapofSTO.Astriking donor, thus explaining the observed dependence of the featureofc-LAO/STOisthepersistenceofthephotocur- carrier number on the oxygen partial pressure during rents after restoring the dark conditions; the character- the LAO film growth.[9, 10] It was also suggested that istic times range from 10 to 104 seconds or even more, depending on the sample. In presence of point defects, bareSTOcanalsoshowpersistentphotoconductivity,[27] resemblingtheclassicallattice-relaxationofdopedGaAs, ∗ Correspondingauthor: [email protected] CdZn,CdS,etc.[28]Thecaseofc-LAO/STOmayinstead 2 beassociatedwith,e.g.,GaAs/AlGaAsheterostructures, TABLE I. Transport data at room temperature in dark con- wherethepersistentphotoconductivityisascribedtothe dition for all samples. spatial separation of the electron-hole pair under the ef- fect of the local electric field.[29, 30] Sample d (nm) V (V) Id (A) Rd (Ω) Inthis paperweinvestigatethe photoresponseofboth C1 2.4 1 23.2 µ 43 k a- and c-LAO/STO.Our aim is to provide a wider char- A1 2.4 1 15.6 µ 64 k acterization of the interfaces based on amorphous LAO, A2 1.2 30 3.5 n 8.6 G thatarecertainlylessexplored,andtocomparetheirbe- S1 - 500 2 p 250 T haviortothatofcrystallineinterfaces. Potentially,thisis averyinterestingissue,becausethecrucialdifferencebe- tween the two systems consists in the amount of oxygen vacancies that, acting as point defects, may play a key role in the photoresponse. In the following, we present data regarding two samples of a-LAO/STO with differ- ent thickness, one c-LAO/STO and one bare STO sub- strate. We stress that, according to the mentioned atti- tude of LAO to oxidize at the expense of STO, and due to the differentgrowthconditions we adopted,the inves- tigated a-LAO/STO samples contain much more point defects, i.e., oxygen vacancies, than c-LAO/STO. Fur- ther details on the fabrication are given in the Method section. It turned out that the time evolution of the photoconductance, during the exposition to radiation of different wavelengths in the visible-ultraviolet (VIS-UV) region, and after having turned the light irradiation off, shows clear differences between different types of sam- ples. Weanalyzethepossiblemicroscopicmechanismsof FIG. 1. Resistance vs. temperature plots of C1, A1 and A2 samples carrier photogeneration, and propose that the different behavior is ascribed to different band diagrams for both structures. This is discussed in terms of naive diagrams for a- and c-LAO/STO. R(T)plot)areobserved,inagreementwithliterature.[14, 15] Basically,A1 shares with the crystalline interface C1 quite similar transport properties in dark. RESULTS AND DISCUSSION To measurethe samplesphotoresponsewe selectthree different wavelengths: 365 nm (3.4 eV, above the indi- rectbadgapofSTO);400nm(3.1eV,belowthe indirect The four samples used in transport and photo- badgap of STO); 460 nm (2.7 eV, well below the indi- transport measurements are labelled as follows: C1 - a conducting 6 u.c. (thickness d≃2.4 nm) c- rect badgap of STO). We define the photocurrent Iph(t) as the difference between the current I(t), measured at LAO/STO heterostructure; time t during the light illumination, and the current in A1 - a conducting a-LAO/STO interface, with an esti- mated LAO thickness d≃2.4 nm (similar to C1); darkconditionsbeforetheirradiation,Id. Then,thepho- A2 - an insulating a-LAO/STO interface, with an esti- toconductance is given by σph(t) = IphV(t), V being the mated LAO thickness d≃1.2 nm; voltage drop across the two probes. The time depen- S1 - a bare TiO -terminated STO single crystal. dent photoconductance of C1, A1, A2, and S1 at three 2 different wavelengths is reported in fig. 2. In dark, S1 exhibits a very high resistance (250 TΩ at roomtemperature,closetothelimitofourmeasurement The photoresponsedynamics of allthe measuredsam- setup), characteristic of fully oxidized, non doped STO plesshowarelativelysteepincrease,followedbyaslower crystals. The deposition of a very thin amorphous LAO transient behavior. However, the intensity is dramati- overlayer (sample A2) determines a noticeable change: cally different for each case. In table II we report the the room temperature resistance in dark is lower by 4 value of photoconductance after 300 s of illumination, orders of magnitude with respect to S1 (Table I). σpmhax , for each sample and wavelength. At 365 nm, σmax is of the order of µS for A1 and C1 samples, However, A2 still shows an insulating behavior, i.e., a ph while it is 3 and 6 orders of magnitude smaller for A2 negative slope in the R(T) plot (fig. 1), suggesting that and S1, respectively. Conversely, the photosensitivity thefabricationprocessintroducedsomeoxygenvacancies in the structure, but their amount is still insufficient to Ψph(t) = σpσhd(t) = IpIhd(t) is maximum in S1, lower in determine the formation of a mobile 2DEG. When the A2 and even lower in A1 and C1. amorphous LAO film is thicker (A1), a dramatic dropof These preliminary observations allow to outline the resistance and a metallic behavior (positive slope in the general framework. A set of donor states (possibly dif- 3 tions,thesestatesarecompletelyfilledintheconducting interfaces,sothatallthe electronspromotedbythelight 1.0 C1 365 nm a) must contribute to the conduction. Instead, they are (at 0.8 S) 400 nm leastpartially)empty inA2. Therefore,a fractionofthe 0.6 460 nm (ph 0.4 photo-generated electrons may occupy these states and may not contribute to the electrical conduction of A2, 0.2 i.e., it may not be detected in the photoresponse mea- 0.0 surements. 1.2 A1 b) Since the same photon flux is adopted at all wave- S) 1.0 length, and considering that each process of photo- (ph00..68 excitation generates one free carrier, it follows that the 0.4 variation of photoconductance with wavelength just re- 0.2 flects the variation of photogeneration efficiency. The 0.0 observation that the maximum photocondutance σmax ph 6.0 A2 c) systematicallyincreasesatdecreasingtheradiationwave- nS) 4.5 length is then easily understood, as higher energy pho- (ph3.0 tonscanexcitedeeperstates;and,inparticular,photons at 365 nm can, in principle, promote electrons from the 1.5 STO valence band (VB) to the conduction band (CB). 0.0 The ratio between the values of σmax at 365nm and ph 1.0 S1 d) 460nmishigherintheinsulatingsamples(S1,A2),than S) 0.8 in the conducting samples (A1, C1). Still in the same p (ph0.6 spirit of a na¨ıve interpretation, this suggests that the 0.4 trapstatesformanarrowband,closetothevalenceband 0.2 edge, in S1 and in A2, so that low energy photons can 0.0 hardlypromoteelectronstotheCB.Thetrapstatesform 0 250 500 750 1000 1250 1500 instead a broad band in A1, and still broader in C1. In Time (s) other terms, the effective optical gap at the interface is FIG. 2. a), b), and c) Photoconductance vs. time of C1, A1 always smaller than the band gap of bulk STO; and it andA2interfaces;d)photoconductancevs. timeofbareSTO issmallerinc-LAO/STOinterfacesthanina-LAO/STO (sampleS1). Thephotonenergy at 365nm(3.4 eV)isabove ones. the indirect STO gap; at 400 nm (3.1 eV) and 460 nm (2.7 eV) it is below this threshold. All samples were illuminated for the same time (for 300 s) at equal photon flux (∼ 2x1014 TABLEII.Photoconductanceandphotosensitivityvaluesaf- photons cm−2s−1) for each wavelength. ter300 s of illumination for thethree different wavelengths. Sample λ(nm) σmax (nS) Ψmax ph 365 991 0.043 ferentforS1,A1,A2andC1)canbeexcitedbyphotons. C1 400 524 0.023 The photoresponse is directly related to the number ∆n 460 267 0.012 of electrons promoted into the conduction band. In the 365 1285 0.082 steady-stateregime,∆nisdeterminedbythebalancebe- A1 400 285 0.018 tween the excitation and the recombination rates. The 460 75 0.005 enhancement of the photoconductance in the sequence 365 6.3 54 S1,A2,andC1/A1thensuggeststhattherecombination A2 400 0.666 5.7 ratebecomeslowerandlowerwhentheLAOthicknessin- 460 0.296 2.5 creases. Underthispointofview,ourresultsallowtoex- tendtothecaseofa-LAO/STOapreviousinterpretation 365 8.8·10−4 220 of the c-LAO/STO photoresponse.[23] The argument is S1 400 9.8·10−5 25 also qualitatively confirmed by the different duration of 460 1.2·10−5 3 the tails after turning the illumination off; in fact, the recovery rate increases in the sequence C1/A1, A2, S1. Inordertogetamoreaccuratedescriptionofthe pho- However, this enhancement may also have an alterna- toresponse,itisnecessaryto discussthe dynamicbehav- tive explanation. It hasbeen proposed[2] thatthe deep- ior of the signals. We first start addressing the issue of est electronic states in the quantum well of c-LAO/STO the increase of photoconductance as a function of the interfaces have low or null mobility. It is reasonable to illumination time (see fig. 2). The plots of σ (t) are ph assume that the same also holds for a-LAO/STO, since characterizedby a steep onset followedby a slowertran- the electronic properties of the 2DEG are mainly deter- sient. Thisbehavioriswelldescribedbyabi-exponential mined by the STO crystal in both cases. In dark condi- growthfunctionwithtwocharacteristicrisetimes τ and 1 4 τ , relatedto the fast and slow components respectively: of the fast and slow component of the signal, i.e., the A 2 parameter. This difference is marked by the dependence σ (t)=σ A 1−exp − t + of A vs. radiation wavelength for the two samples (fig. ph inf (cid:26) (cid:20) (cid:18) τ (cid:19)(cid:21) 3c). 1 t (1−A) 1−exp − (1) (cid:20) (cid:18) τ2(cid:19)(cid:21)(cid:27) TABLEIII.Bestfittingparametersforthethethreedifferent illumination wavelengths. where σ is the saturation value of the photoconduc- inf tance and A represents the relative weight of the fast Sample λ(nm) σinf (nS) A τ1(s) τ2(s) component. 365 990 0.78 0 82 C1 400 553 0.53 9 147 460 295 0.19 24 140 1.0 a) 365 1420 0.33 1.3 166 0.8 A1 400 322 0.56 4 228 inf0.6 460 74 1 40 – /ph0.4 365 nm 400 nm In order to address this issue, we first remind that the 0.2 460 nm 0.0 C1 selected wavelengths correspondto photon energies that span from slightly above to well below the indirect STO 1.0 b) gap. In sample C1, the fast component is progressively 0.8 reduced with increasing the wavelength, confirming the inf0.6 previous observations for crystalline samples.[23] At 460 /ph0.4 nm, well below the gapthreshold, the fast channel is ac- tually almost suppressedand the signalrise is veryslow. 0.2 A1 The opposite behavior is observed in sample A1, where 0.0 the relative amplitude of the fast component increases 0 50 100 150 200 250 300 at increasing wavelength. At 460 nm, the slow channel Time (s) is totally suppressed, so that the signal is well described by a simple exponential growth with one characteristic 1.0 c) time τ . Such different response of C1 and A1 undoubt- 1 0.8 edlyindicatesadifferentunderlyingmechanismofphoto- A 0.6 excitation, which distinguishes a- from c-LAO/STO. 0.4 We consider now the delicate issue of the microscopic 0.2 A1 description of carriers photo-excitation. The modelling C1 of the dynamics in terms of band diagrams is neither 0.0 straightforward nor unambiguous; however, we believe 360 380 400 420 440 460 that the main physics is captured by the naive sketches (nm) of fig. 4. The right panel (fig. 4b) is based on published FIG. 3. a) Photoresponse of C1 and b) photoresponse of A1 diagramsfor c-LAO/STO;[23,26,31] the left panel(fig. at 365 nm (black), 400 nm (red), and 460 nm (green). The 4a)describesinsteada-LAO/STO.Therightsidesofthe data are normalized to the asymptotic value σ ; solid lines band diagramsfor a-LAO/STOand c-LAO/STOdeeply inf are fit curves. c) Dependence of A vs. radiation wavelength differ the one from each other and are in agreement λ for both samples. with those already reported in [16]. For a-LAO/STO, we draw, as an ansatz, a scheme qualitatively similar We focus now the attention on the metallic samples, to the one for bulk LAO crystal. For c-LAO/STO, in- C1 andA1 samples. The fitting parametersarereported stead, we adopt the standard ER model, with the bands in Table III, and the fits are shown in fig. 3. In our ex- of LAO bent upwards by the macroscopic electric field. perimental conditions, the maximum photoconductance Theleftpartofthediagramsofbotha-andc-LAO/STO scales with light intensity, i.e., the optical pumping is shows the quantum well that is formed at the interface, below the threshold at which the response falls in the within STO, and that confines the mobile electrons in saturationregime. This circumstance allows the connec- the 2DEG. The point defects, which we mainly identify tionbetween the dynamics ofthe photoconductance and with oxygen vacancies, add to the bands a distribution the photogeneration mechanism of carriers. On this ba- of trap states, forming a broad in-gap band. Such states sis, we interpret the elongation of the fast time τ at are mainly localized in a region close to the interface, 1 increasingwavelengthasaneffectofthedecreaseofpho- because, as previously mentioned, they are formed dur- togeneration efficiency. This mechanism is common to ing the film growth step.[15] Due to the different fab- both C1 and A1. Instead, a striking difference between rication procedure, the density of vacancies is lower in the samples is found in terms of the relative amplitude c-LAO/STO, and as a first approximation we neglect 5 it. The sketches show some possible channels of pho- 2. In c-LAO/STO, the excitation by low energy pho- tons is a slow process, that promotes to the quan- tum wellelectrons fromthe LAO VB;the fastpro- cess requires, instead, high energy photons, and consistsintheionization(possiblyphononassisted) of STO VB states. 3. In a-LAO/STO, on the contrary, the excitation by low energy photons is a fast process, promot- ing to the quantum well electrons from the in-gap trap states and the slow process (promoting elec- tron from the LAO VB) requires high energy pho- tons. Furthermore, the direct excitation of STO VB states in a-LAO/STO may trigger a complex energycascadewithinthe broadin-gapband. This effectwouldalsoindirectlycontributetoslowdown FIG.4. Sketchofthebandstructureofa)a-LAO/STOandb) thedynamicsofexcitationbyhighenergyphotons. c-LAO/STO. The grey dotted region depicts the in-gap trap states of STO. For clarity, the bending of the STO bands is We stress that our data would be hardly explained by emphasizedinbothpicturesand,inc-LAO/STO,thebandof assuming that all the relevant transitions are due to the in-gapstatesisomitted. Theslow(s)andfast(f)mechanisms excitation of one type of trap states, i.e., oxygen vacan- of photogeneration are represented byarrows. cies only. In particular, this would be at odds with the higherphotoconductanceinducedbylowenergyphotons togeneration, labeled as s or f to indicate slow and fast in c-LAO/STO than in a-LAO/STO (see, e.g., data of dynamics, respectively. The main channel of fast exci- σmax in table II). In fact, the oxygenvacancy states cer- tation is the local (on-site) promotion of electrons from ph tainly form a broader band in a-LAO/STO than in c- initial states with O-character to final states of the CB LAO/STO. with Ti-character; this process may actually involve in- termediate states, such as excitons (i.e., localized Ti3+ We finally add some comments on the bending of c-LAO bands (fig. 4b). In contrast to first- states),[32] which rapidly decay into the CB. The slow principle computations, [2–4] the bending was not processis here identifiedwith the promotionofelectrons observed in some x ray photoemission spectroscopy from LAO states with O-character to STO states with (XPS) measurements.[34–36] In other XPS works (e.g., Ti-character. These transitions are important, because [37]), as well as in cross-sectional scanning tunnelling they can involveinitial and final states with non negligi- spectroscopy,[38]itappearedwithreducedslopewithre- bleoverlap. Inthecaseofc-LAO,forinstance,theclosest spect to expectations. In [35], the authors propose two OatomofLAOisactuallyintheapicalsiteofaTiatom effects (that in our view may concur) to explain the dis- of the STO side. A similar consideration can be applied crepancy between theory and experiment: a) during the also to the case of a-LAO. However, there is a relevant XPS experiment, the sample is heavily far from equilib- difference between the twocases. All the VB-edge states riumduetoxrayphotoexcitation;b)thesurfacecharge- of LAO share the same energy in a-LAO/STO. Instead, trapstatesactasdonors,sothatevenatequilibriumthe as the sketch in fig. 4b qualitatively shows, the upturn bandbending ofc-LAOis attenuated. Bothmechanisms of the bands in c-LAO is so steep, that the VB edge is alreadyliftedby≃0.2-0.3eVattheapicalsites. Thefar- mayalsoaffectthephotoresponseandgivequantitatively relevant corrections. However, we believe that the sim- thertheinitialstatesarelocated,thelargeristheenergy plequalitativepicturethatweproposewouldnotchange lift. also taking into account these extra channels. The sketch makes no claim to completeness; more ex- We turnnow to the issue ofthe signaldecayafter illu- citation mechanisms can be at play and also be quan- mination (fig. 5). As a first comment, we note that this titatively relevant (e.g., the transitions from the defect process is affected by the sample history. Depending on states of STO in c-LAO/STO, such as photoexcitations the durationandonthe intensity ofpreviousexpositions of carriers from defect-induced deep-level traps as sug- tothelight,aswellasonthewavelength,weobservedif- gested in [33], or from any other kind of surface trap ferentvaluesofthefinalbaseline,indicativeofmetastable state). However,the minimum set of transitions that we states. The further relaxation toward the initial equilib- showinthediagramsissufficienttoqualitativelyexplain rium state in dark may be extremely long. We did not several experimental observations: exploresystematicallythisfeatureinthiswork. However 1. The reduction of the optical gap in a-LAO/STO it appears that the recoverytowardthe metastable state (withrespecttotheindirectSTOgap)isexplained is always a process longer than the signal rise one. as an effect of oxygen vacancies; in c-LAO/STO, As mentionedbefore,the increaseofthe characteristic instead, the reduction of the optical gapis a result decay time with respect to the rise transient is indica- of the band bending of LAO. tive of a hysteretic behavior that can be attributed to 6 LAO/STO. a) ) 1 365 nm C1 The transient photoconductance of all samples shows 0 (t 400 nm a bi-exponential growth,under illumination, followedby h p 460 nm a long decay when the light is turned off. In spite of ph the qualitative similarities, the photoconductance dy- namics reveals a deep difference between a-LAO/STO andc-LAO/STO.Inc-LAO/STO,thefastchannelofthe 0,1 photoresponse becomes progressively less relevant at in- b) creasing photon wavelength; the reverse takes place in ) 1 A1 (t0 a-LAO/STO, where it is the slow channel loosing rela- ph tive weight at increasing wavelength. /h We propose a different mechanism of excitation for a- p LAO/STOandc-LAO/STO,whichnaturallystemsfrom the different band diagrams of the two structures. In- deed,theopticalexcitationofa-LAO/STOinvolvestran- 0,1 sitions from either in-gap trap states due to oxygen va- 300 500 700 900 1100 1300 1500 cancies (fast channel) or the LAO VB (slow channel). Time (s) Thefirstchannelisactivatedbyphotonsinawide range FIG.5. Signaldecayafterswitchingofftheilluminationfora) ofenergy,whilethesecondisrestrictedtoabove-gappho- C1andb)A1samples. Thesamedataoffig. 2arenormalized tons. The excitation of c-LAO/STO also involves two to thecorresponding σmax values and plotted in log scale. ph channels. ThefirstincludesthetransitionsfromtheSTO VB (fast channel) and is restricted to high energy pho- tons; the second includes the transitions from the bent the following mechanisms: a) the activation of a slow LAO VB band (slow channel), and is activated even by energy cascade through intermediate defect states, and low energy photons. b) the drift of photo-generatedholes under the action of Finally, a-LAO/STOandc-LAO/STOalsohavea dif- the macroscopic electric field in the region close to the ferent behavior of permanent photoconductance when interface. As previously discussedwhen commenting the lightis turned off, indicating that the physicaldifference signal rise dynamics, the former mechanism may have between the structures also determines different dynam- comparativelylarger weight in a-LAO/STOafter UV ir- ics of relaxation. radiation. Inthatcasethedeeplevelsinthebroadin-gap band are excited, and part of them might be long-lived. The latter mechanismmay instead affect the recoveryof METHOD c-LAO/STO after visible light irradiation, if some holes createdintheLAOregionarepushedtowardthe sample surface by the electric field that acts in the polar layer. The LAO thin films are grown by a RHEED-assisted Pulsed Laser Deposition technique on TiO -terminated 2 (001) STO single crystal substrates, resorting to a KrF CONCLUSION excimerlaserbeamwithafluence of1.5J/cm2 ontarget and a repetition rate of 1 Hz. The crystalline sample ◦ We investigated the transient photoconductance of C1 is grown at 720 C with a partial pressure of oxygen amorphous-LAO/STO interfaces, under monochromatic PA(O2)=1x10−2mbar. Afterthedeposition,thesample radiationin the UV and Vis region, and comparedthese iscooledin1htoroomtemperatureinthesamepressure systems with a crystalline-LAO/STO structure and a condition. TheamorphoussamplesA1andA2aregrown bare STO crystal. In all the samples, the photoconduc- at room temperature with a PA(O2) = 1x10−5mbar. tance assumes the highest value at the shortestexplored Allthesamplesarecontactedbyultrasonicallybonded wavelength (365 nm, corresponding to above-gap exci- aluminiumwires. Theresistanceasafunctionofthetem- tation). At longer wavelengths, the photoconductance perature R(T) is recorded in dark condition in a closed significantly decreases in c-LAO/STO. The reduction is cyclerefrigerator,usingthe standardfourprobe Vander even stronger in conducting a-LAO/STO, and dramatic Paw configuration for C1 and A1, and a two probe con- in the insulating a-LAO/STO sample and in the bare figuration for A2 and S1, due to their high resistance STO crystal. This observation indicates that the opti- values. In all the photoconductance measurements, the cal gap of the interfaces is smaller than the band gap current across the samples is measured at fixed voltage of bulk STO; and that it is the smallest in the case inthetwoprobeconfiguration,byresortingtoaKeithley of c-LAO/STO. This effect would be hardly explained 6487 picoammeter. All the values of applied voltage V, by assuming that the photogeneration takes place in all dark current Id, and resistance in dark Rd, are reported samples by exciting the same states (e.g. in-gap states in Table I. due to oxygen vacancies), since vacancies certainly gen- In order to investigate the time evolution of the pho- erate a broader defects band in a-LAO/STO than in c- toconductance, the whole sample surface (∼5×5 mm2) 7 is uniformly illuminated for 300 s with monochromatic is measured by a radiometer (Laser Precision RK-5720 radiation obtained by means of a Xenon lamp and of Power Radiometer) equipped with a silicon probe. The band-pass interference filters (FWHM = 10 ± 2 nm). light intensity was attenuated by means of neutral den- Threewavelengthsareselected: 365nm(3.4eV);400nm sity filters in order to obtain a photon flux of ∼ 2x1014 (3.1 eV); 460 nm (2.7 eV). The power density radiation photons cm−2s−1 for the three wavelengths. [1] Ohtomo A. & Hwang H. Y. A High-Mobility Electron 104, 261603–261603 (2014). GasattheLaAlO3/SrTiO3Heterointerface. Nature 427, [17] Huijben,M. InterfaceEngineeringforOxideElectronics: 423–426 (2004). Tuning Electronic Properties by Atomically Controlled [2] Popovi´c, Z. S., Satpathy, S. & Martin, R. M. Origin Growth. Ph.D. thesis, Universityof Twente (2006). of the Two-Dimensional Electron Gas Carrier Density [18] Huijben,M.etal. Electronicallycoupledcomplementary at the LaAlO3/SrTiO3 Interface. Phys. Rev. Lett. 101, interfaces between perovskite band insulators. Nature 256801 (2008). Mat. 5, 556–560 (2006). [3] Pentcheva, R. & Pickett, W. E. Avoiding the Polariza- [19] Thiel, S. P. Study of Interface Properties in tion Catastrophe inLaAlO3/SrTiO3(001) throughPolar LaAlO3/SrTiO3 Heterostructures. Ph.D. thesis, Univer- Distortion. Phys. Rev. Lett. 102, 107602 (2009). sitat Augsburg(2009). [4] Stengel,M. First-principles modelingof electrostatically [20] Irvin,P.etal.Rewritablenanoscaleoxidephotodetector. doped perovskite systems. Phys. Rev. Lett. 106, 136803 Nature Phot. 4, 849–852 (2010). (2011). [21] Rastogi,A.,Pulikkotil,J.,Auluck,S.,Hossain,Z.&Bud- [5] Nishimura,J.,Ohtomo,A.,Ohkubo,A.,Murakami,Y.& hani, R. Photoconducting state and its perturbation by Kawasaki,M.Controlledcarriergenerationatapolarity- electrostatic fields in oxide-based two-dimensional elec- discontinued perovskite heterointerface. Jap. Jour. of tron gas. Phys. Rev. B 86, 075127 (2012). Appl. Phys. 43, L1032 (2004). [22] Tebano, A., Fabbri, E., Pergolesi, D., Balestrino, G. & [6] Thiel, S., Hammerl, G., Schmehl, A., Schneider, C. W. Traversa,E.Room-TemperatureGiantPersistentPhoto- & Mannhart, J. Tunablequasi-two-dimensional electron conductivity in SrTiO3/LaAlO3 Heterostructures. ACS gases in oxide heterostructures. Science 313, 1942–1945 Nano 6, 1278–1283 (2012). (2006). [23] Di Gennaro, E. et al. Persistent photoconductivity in [7] Basletic M. et al. Mapping the Spatial Distribution 2d electron gases at different oxide interfaces. Adv. Opt. of Charge Carriers in LaAlO3/SrTiO3 Heterostructures. Mat. 1, 834–843 (2013). Nature Mat. 7, 621–625 (2008). [24] Lu, H.-L. et al. Reversible Insulator-Metal Transition [8] Cantoni, C. et al. Electron transfer and ionic displace- of LaAlO3/SrTiO3 Interface for Nonvolatile Memory. ments at the origin of the 2d electron gas at the lao/sto Scien. Rep. 3, 1–6 (2013). interface: Directmeasurementswithatomic-columnspa- [25] Lu, H.-L. et al. Photoelectrical Properties of Insulating tial resolution. Adv. Mat. 24, 3952–3957 (2012). LaAlO3/SrTiO3Interfaces.Nanoscale6,736–740(2014). [9] Huijben, M. et al. Structure-Property Relation of [26] Liang, H. et al. Giant Photovoltaic Effects Driven by LaAlO3/SrTiO3 Interfaces. Adv. Mat. 21, 1665–1677 ResidualPolarFieldwithinUnit-Cell-ScaleLaAlO3films (2009). on SrTiO3. Scien. Rep. 3, 1–7 (2013). [10] Cancellieri, C. et al. Influenceof theGrowth Conditions [27] Tarun, M. C., Selim, F. A. & McCluskey, M. D. Persis- on the LaAlO3/SrTiO3 Interface Electronic Properties. tentphotoconductivityinstrontiumtitanate. Phys.Rev. Eur. Lett. 91, 1–4 (2010). Lett. 111, 187403 (2013). [11] Schneider, C. et al. The origin of oxygen in oxide thin [28] Lang, D.&Logan,R. Large-lattice-relaxation modelfor films: Role of the substrate. Appl. Phys. Lett. 97, persistent photoconductivity in compound semiconduc- 192107–192107 (2010). tors. Phys. Rev. Lett. 39, 635 (1977). [12] Santander-Syro,A.etal. Two-DimensionalElectronGas [29] Queisser, H. & Theodorou, D. Decay kinetics of persis- with Universal Subbands at the Surface of SrTiO3. Na- tent photoconductivityin semiconductors. Phys. Rev. B ture 469, 189–193 (2011). 33, 4027 (1986). [13] Di Capua, R. et al. Observation of a Two-Dimensional [30] Kioupakis, E., Yan,Q. & Van deWalle, C. G. Interplay Electron Gas at the Surface of Annealed SrTiO3 Single ofpolarizationfieldsandaugerrecombinationintheeffi- CrystalsbyScanningTunnelingSpectroscopy.Phys.Rev. ciencydroopofnitridelight-emittingdiodes. Appl.Phys. B 86, 155425 (2012). Lett. 101, 231107–231107 (2012). [14] Chen, Y. et al. Metallic and Insulating Interfaces of [31] Rastogi, A.& Budhani,R. Solar Blind Photoconductiv- AmorphousSrTiO3-BasedOxideHeterostructures.Nano ity in Three-Terminal Devices of LaAlO3/SrTiO3 Het- Lett. 11, 3774–3778 (2011). erostructures. Opt. Lett. 37, 317–319 (2012). [15] Liu, Z. et al. Origin of the Two-Dimensional Electron [32] Zhou, K.-J. et al. Localized and Delocalized Ti 3d Car- Gas at LaAlO3/SrTiO3 Interfaces: The Role of Oxygen riersinLaAlO3/SrTiO3 SuperlatticesRevealedbyReso- Vacancies and Electronic Reconstruction. Phys. Rev. X nantInelasticX-rayScattering. Phys.Rev.B 83,201402 3, 021010 (2013). (2011). [16] De Luca, G. et al. Potential-Well Depth at Amorphous- [33] Rastogi, A., Pulikkotil, J. J. & Budhani, R. C. LaAlO3/Crystalline-SrTiO3 Interfaces Measured byOp- Enhanced Persistent Photoconductivity in δ-Doped tical Second Harmonic Generation . Appl. Phys. Lett. LaAlO3/SrTiO3 Heterostructures. Phys. Rev. B 89, 8 125127 (2014). [38] Huang, B.-C. et al. Mapping Band Alignment across [34] Chambers,S.etal.Instability,intermixingandelectronic Complex Oxide Heterointerfaces. Phys. Rev. Lett. 109, structure at the epitaxial heterojunction. Surf. Scien. 246807 (2012). Rep. 65, 317 – 352 (2010). [35] Berner, G. et al. Direct k-Space Mapping of the Elec- tronicStructureinanOxide-OxideInterface. Phys. Rev. Lett. 110, 247601 (2013). [36] Drera, G. et al. Intrinsic origin of interface states and AUTHORS CONTRIBUTION band-offset profiling of nanostructured LaAlO3/SrTiO3 heterojunctions probed by element-specific resonant spectroscopies. Phys. Rev. B 90, 035124 (2014). A.K. and E.D.G. prepared the samples, E.D.G. per- formed the cryogenic transport measurements, U.C. [37] Slooten, E. et al. Hard x-ray photoemission and density functional theory study of the internal electric field in and G.A. carried out the phototransport measurements, SrTiO3/LaAlO3 oxide heterostructures. Phys. Rev. B U.S.d.U. and F.M.G. wrote the main manuscript text. 87, 085128 (2013). All authors reviewed the manuscript.