# Fileset

[HZO_Operando_SPring-8.pdf](https://mdr.nims.go.jp/filesets/7a11fcab-2023-4ba5-9125-f6e58864a20b/download)

## Creator

[W. Hamouda](https://orcid.org/0000-0002-7055-7264), [Y. Yamashita](https://orcid.org/0000-0003-0994-8095), [S. Ueda](https://orcid.org/0000-0001-9425-0614), [S. Matzen](https://orcid.org/0000-0002-4244-6516), [O. Renault](https://orcid.org/0000-0002-0683-9590), [F. Mehmood](https://orcid.org/0000-0002-9530-380X), [T. Mikolajick](https://orcid.org/0000-0003-3814-0378), [U. Schroeder](https://orcid.org/0000-0002-6824-2386), [N. Barrett](https://orcid.org/0000-0002-8228-0805)

## Rights

This article may be downloaded for personal use only. Any other use requires prior permission of the author and AIP Publishing. This article appeared in W. Hamouda, Y. Yamashita, S. Ueda, S. Matzen, O. Renault, F. Mehmood, T. Mikolajick, U. Schroeder, N. Barrett; Electron trapping/detrapping at oxygen vacancies and imprint evolution in La-doped Hf0.5Zr0.5O2 ferroelectric capacitors probed by hard x-ray photoelectron spectroscopy. Appl. Phys. Lett. 3 November 2025; 127 (18): 182902 and may be found at https://doi.org/10.1063/5.0288835.[In Copyright](http://rightsstatements.org/vocab/InC/1.0/)

## Other metadata

[Electron trapping/detrapping at oxygen vacancies and imprint evolution in La-doped Hf0.5Zr0.5O2 ferroelectric capacitors probed by hard x-ray photoelectron spectroscopy](https://mdr.nims.go.jp/datasets/85a9022f-7928-465b-a2d8-d7f2aadcaec0)

## Fulltext

Electron trapping/detrapping at oxygen vacancies and imprint evolutionin La-doped Hf0.5Zr0.5O2 ferroelectric capacitors probed by hard X-rayphotoelectron spectroscopyW. Hamouda,1, a) Y. Yamashita,2 S. Ueda,2, 3 S. Matzen,4 O. Renault,5 F. Mehmood,6, b) T. Mikolajick,6 U.Schroeder,6 and N. Barrett11)SPEC, CEA, CNRS, Université Paris-Saclay, CEA Saclay, 91191 Gif-sur-Yvette, France2)Research Center for Electronic and Optical Materials, National Institute for Materials Science (NIMS), Namiki,Tsukuba 305-0044, Japan3)Synchrotron X-ray Station at SPring-8, National Institute for Materials Science, Sayo, Hyogo 679-5148,Japan4)C2N, CNRS, Paris-Saclay University, 91120 Palaiseau, France5)Univ. Grenoble Alpes, CEA, Leti, F-38000 Grenoble, France6)NaMLab gGmbH/TU Dresden, Nöthnitzer Str. 64a, 01187 Dresden, Germany(*Correspondence should be addressed to nick.barrett@cea.fr)(Dated: 13 October 2025)We evaluate the correlation between polarization-dependent oxygen vacancy profile and imprint in technologically rel-evant TiN/La doped Hf0.5Zr0.5O2/TiN ferroelectric capacitors using hard X-ray photoelectron spectroscopy (HAXPES)with in-situ biasing. The concentration of double positively charged oxygen vacancies (V..O) was inferred from theintensity of reduced hafnium (Hf3+, (5d1)) in the Hf 3d5/2 spectra, relative to fully oxidized (Hf4+, (5d0)). HAX-PES was performed using two photon energies to discriminate V..O near the top interface from those averaged overthe whole film. The evolution of imprint with time and the extracted activation energy of ∼0.1 eV strongly supporta fast, low-barrier electronic process such as shallow charge trapping/detrapping, rather than field induced vacancydrift. On the ∼30 min timescale after polarization reversal, we propose that the V..O deduced from HAXPES reflects apolarization-driven snapshot of the defect charge state (VxO↔ V..O), rather than a change in the physical vacancy densitypreviously reported. Following polarization reversal, electrons redistribute between vacancy levels and Hf 5d orbitals.This is correlated with Hf4+ ↔ Hf3+ transitions, leading to 0 ↔ 2+ vacancy charge state modifying the local elec-tronic environment and screening behavior, particularly near interfaces. From the Schottky barrier height change dueto polarization reversal, an effective screening length of 0.04 Å was estimated for the top interface. The results reveala dominant role of polarization-dependent electron trapping/detrapping, partially screening ferroelectric polarizationcharges, thereby stabilizing one remanent state over the other.Keywords: ferroelectric Hf0.5Zr0.5O2, HAXPES, imperfect screening, depolarizing field, polarization imprint, oxygenvacancy (VO), charge trapping/detrappingResearch on hafnia-based ferroelectric capacitors (Fe-CAPs) has been intensified since the discovery of ferroelec-tricity in HfO2-based materials due to their potential appli-cations for low-power non-volatile memories and memristivedevices1–5. In this context, the main reliability parameters areendurance, retention, and imprint6.Field cycling is usually necessary to fully unpin the ferro-electric domains and reach optimal polarization. Enduranceis the number of cycles before fatigue sets in with closure ofthe memory window, increase in leakage current and finallyhard breakdown. Retention measures the ferroelectric statestability over time, typically at 85◦C for consumer electron-ics. Reported time-scales of∼105s must be extrapolated to 10years for commercial applications7–9.Oxygen vacancies (VO) are known to impact endurance andretention9–12. In addition to stabilizing the polar orthorhombica)Present address: Helmholtz-Zentrum Berlin für Materialien und Energie,Hahn-Meitner-Platz 1, 14109 Berlin, Germanyb)Present address: GlobalFoundries Dresden, Wilschdorfer Landstr. 101,01109 Dresden, Germanystructure at modest concentrations and providing conductiv-ity paths at higher concentrations, they act as electron trapswith shallow charged levels (V..O) located typically 0.3–0.5eV below the conduction band minimum (CBM), and deepneutral levels (VxO) positioned 2-3 eV below the CBM9,12,13.Retention is also closely connected to imprint which mani-fests itself as a time and temperature dependent shift of thepolarization-voltage hysteresis loop along the voltage axis,destabilizing one state with respect to the other. The incom-plete screening of the polarization bound charges results in adepolarizing field (Edep) in the ferroelectric layer and a fieldat the ferroelectric/electrode interface. These fields result in atime-dependent redistribution of charge via a variety of mech-anisms, including bulk charge transport, charge injection atthe interfaces, trapping/detrapping at defect sites, giving riseto the imprint field, Eimp. In addition, asymmetric electrodesor interface chemistry will result in asymmetric polarizationscreening further modulating Eimp14.To describe the polarization screening at the metal interface,an effective screening length, λeff, is used15,16. λeff is gen-erally regarded as the width of the screening charge distribu-tion in the metal electrode and therefore depends on the elec-2FIG. 1. (a) Schematic of the analyzed stack highlighting the probingdepth using two distinct photon energies (b) Optical image (top) andcross section (bottom) of a single device dedicated for HAXPES within-situ biasing.trode material, the defect state of the interfacial region i.e. thespecific interface chemistry and the polarization magnitude.It can be deduced from the Schottky barrier height (SBH)change (∆ϕ(B,n)) (in eV) due to polarization reversal15,17:∆ϕ(B,n) = 2λeffPr/ε0 (1)where Pr stands for the remanent polarization value and ε0for the vacuum permittivity.Laboratory X-ray photoelectron spectroscopy (XPS) canprovide direct quantitative information on defects and bandalignment. By assuming that each VxO releases two electronsreducing two Hf ions from 4+ to 3+ state, the double pos-itively charged V..O concentration can be estimated from theI(Hf3+)/ [(I(Hf3+) + I(Hf4+)] intensity ratio. The shift in thecore level binding energy can be used to estimate ∆ϕ(B,n) andthus λeff18,19. X-rays produced by synchrotron radiation ina hard X-ray photoelectron spectroscopy (HAXPES) experi-ment probing the same core levels increase significantly thephotoelectron inelastic mean free path and thus the probingdepth with respect to laboratory based XPS20. This allowsstudying the ferroelectric layer in realistic device structures ina non-invasive manner17,18,21,22.Recent studies have highlighted the role of elec-tronic charge injection and/or trapping/detrapping at themetal/ferroelectric interface as a key contributor to imprintformation at room temperature14. This charge movement isdriven by the built-in fields Edep and the interfacial electricfield, however, the kinetics depend on the activation energy,i.e. the potential barrier to each mechanism. In the case ofcharge trapping/detrapping at point defects such as VO, theequilibrium concentration of the latter is determined by thetotal energy of the system in a given polarization state. In thevicinity of electrodes, low-permittivity interfacial layers (IL)generate increased electric fields, further modulating chargeredistribution and resulting in time-dependent imprint fields,as discussed in the “fluid imprint” model23–26.In this work, we investigate the imprint and its underlyingmechanisms in fully woken-up TiN/La:Hf0.5Zr0.5O2/TiN fer-roelectric capacitors using HAXPES combined with electricalmeasurements. While the V..O signature in HAXPES has pre-viously been used to assess static oxygen vacancy concentra-FIG. 2. (a) Endurance plot and (b) I - V, (c) P - V responses of pris-tine (black) and woken-up, after 106 cycles (blue) TiN/HZLO/TiNcapacitors.tion, our results reveal that on short time scale after polariza-tion reversal (30 minutes at room temperature), polarization-dependent variations in the V..O (Hf3+) signal can also reflectthe number of electrons trapped at VO sites. The extracted ac-tivation energy of∼0.1 eV for the polarization imprint and thepolarization-dependent variation in Hf3+/Hftot photoemissionsignal indicate a fast, low-barrier electronic process ratherthan slower VO migration.TiN/La:Hf0.5Zr0.5O2(HZLO)/TiN ferroelectric capacitorswere fabricated on p-doped Si(100) substrates. The La con-centration is 2.3 mol%, achieved by modulating the La/(Hf+ Zr) atomic layer deposition cycle ratios and growth ratesaccordingly. Crystallization of the 10 nm HZLO film in thedesired polar orthorhombic phase was achieved by rapid ther-mal annealing at 500◦C for 20 s in a N2 atmosphere afterthe deposition of the 12 nm TiN top electrode and verifiedby grazing incidence X-ray diffraction%8. 100×100 µm2 de-vices were then patterned for HAXPES analysis with in-situapplied bias. Figure. 1 shows a schematic cross-section of thecapacitor stack and an optical image of the device. For bias-ing, capacitors are connected via bonding pads to gold con-tacts on the sample holder. Capacitors were cycled ex-situ towoken-up states using an aixACCT TF 1000 ferrotester using100 kHz square pulses with±3 V amplitude. Hysteresis loopswere measured at various cycling stages using±3 V triangularpulses at 1 kHz (Fig. 2). In the pristine state, a typical pinched,antiferroelectric-like hysteresis curve is observed. Cycling to106 cycles results in the unpinching of the hysteresis, an in-crease of the 2Pr value from 16 to 35 µC.cm−2 and a 1.1 - 1.3V coercive voltage (VC).For the imprint experiment, woken-up capacitors were po-larized in P↑ (pointing toward the top electrode) and P↓ (to-ward the bottom electrode) states. The capacitors were thenheld for different durations at room temperature. The current-voltage (I-V) and the resulting polarization-voltage (P-V)3FIG. 3. Positive (red) and negative (blue) time-dependent imprint corresponding to P↑ and P↓ stored states from 1min to 3hrs at (a) roomtemperature and (b) at different temperatures allowing the extraction of the activation energy. The dotted line in (a) indicates the time of theHAXPES analysis. The Arhennius plot is shown in the inset of (b).FIG. 4. Polarization dependent Hf 3d5/2 spectra at 6 keV (a,c) and 8 keV (b,d) photon energy in woken up capacitors (e) Estimated oxygenvacancy concentration (V..O) as a function of probing depth for P↑ and P↓ remanent states.curves were measured with a triangular pulse train (±3 V/10µs) so that the initial stored state is always maintained afterthe measurement.HAXPES measurements were conducted on the BL15XUbeamline of SPring-827. Photon energies of 6 and 8 keVwere selected using a double-crystal Si (111) monochroma-tor (DCM) and a post-monochromator channel-cut Si (333)crystal. This gives a total energy resolution at room temper-ature of 250 and 280 meV, respectively, determined by ex-tracting the 16/84% width of the Fermi edge of a groundedgold reference sample, also used for binding energy calibra-tion. Photoelectrons were detected using an hemispherical an-alyzer (VG Scienta, R4000) with an emission angle (θ ) of35◦ with respect to the sample surface normal. The samplingdepth using Hf 3d5/2 photoelectrons is defined as 3λcosθ(corresponding to 95% of total intensity), with λ representingthe kinetic energy-dependent inelastic mean free path (IMFP)of the selected photoelectrons through the top TiN electrode.The IMFP were calculated using the Tanuma, Powell, andPenn (TPP-2M) algorithm implemented in the QUASES soft-ware28 to be ∼6 and 8.5 nm using 6 and 8 keV excitationenergies, yielding sampling depths of 15±0.8 (6 keV) and21±1.1 nm (8 keV). Analysis of the spectra was performedusing CasaXPS software29. A Shirley background was used tosimulate the secondary electron background and pseudo-Voigt(Gaussian/Lorentzian with 70% Lorentzian character) func-tion was used to fit the spectra. In-situ polarization switchingwas performed using a triangular pulse (±3 V/10 µs) witha KEYSIGHT 33512B waveform generator. HAXPES mea-surements were recorded ∼30 min after the pulse applicationin short-circuit conditions.Figure 3(a) shows the evolution of imprint voltage ((V+C4- V−C )/2, where V+C and V−C are the coercive voltages) withtime for the P↑ and P↓ states at room temperature, measuredover a 3 hour period. The imprint field evolves toward +0.4MV.cm−1 for P↑ and -0.5 MV.cm−1 for P↓ within the first 30min, coinciding with the timing of the HAXPES measure-ments. The imprint increases over time, reflects the time-dependent built-in internal field (Eimp) that differentiates thetwo remanent states and stabilizes the previously programmedstate. Figure 3(b) presents the temperature dependence of theimprint evolution for five different temperatures from roomtemperature to 150◦C . The extracted activation energy fromthe Arrhenius plot of ln(imprint slope) versus 1/T is 0.12 eV(inset) is significantly lower than typical migration barrier foroxygen vacancies (∼1 eV)30. This strongly indicates that thedominant mechanism behind the initial imprint formation isnot bulk V..O migration but a much faster process. Specifically,we suggest that the initial imprint evolution is due to chargetrapping and detrapping of electrons at VO sites.HAXPES was used to evaluate the Hf3+ distribution as afunction of polarization state in the woken up state in or-der to evaluate the V..O concentration at a characteristic time(30 minutes) during the imprint evolution shown in Fig. 3(a).Figure 4 shows the polarization dependent Hf 3d5/2 spectraat (a,c) 6 and (b,d) 8 keV photon energies. Spectra are fit-ted by two peaks: Hf4+ (gray) and Hf3+ (black)24,31. Thetwo peak fit is detailed in Supplemental Material. We es-timate the V..O concentration from the intensity ratio using1/8×I(Hf3+)/I(Hftot)32. The results are plotted as a functionof the photon energy in Fig. 4(e) for P↑ and P↓ states. At thetop interface the V..O concentration is higher in the P↓ state(0.81%) than in the P↑ state (0.65%) whereas deeper into thefilm the situation is reversed, 1.01% and 0.40% for P↑ and P↓states, respectively. The ∼0.4% V..O difference between thetop interface and the bulk (6 vs. 8 keV) indicates a signifi-cant profile of V..O across the stack, not necessarily expectedin woken-up devices33–35.The V..O gradient for P↑ and P↓ derived from HAXPES(Fig. 4(e)) is qualitatively consistent with the direction of Edepand the measured Eimp (imprint) set by the remanent polariza-tion (E(z) = - dφdz ), but not necessarily indicative of large-scalevacancy drift over the 30 min timescale, suggesting that theobserved V..O-related signal modulation may originate primar-ily from a modulation of their charge state rather than physicalmigration.At 6 keV, when polarization points downward, resultingin negative bound charge at the top interface, electrostaticconsiderations would normally favor more positively chargedoxygen vacancies36. However, the higher V..O content at thetop interface in P↓ (compared to P↑) cannot simply be ex-plained on such a short timescale by V..O migration, since thelatter has an activation energy about one order of magnitudehigher than the measured value (∼0.1 eV). The combinedHAXPES and electrical measurements indicate that the V..Osignal reflects polarization-dependent modulation of the va-cancy charge state and the associated valency of neighboringHf ions. For P↓, electron detrapping from pre-existing neutralvacancies (VxO) converts them into doubly ionized (V..O), form-FIG. 5. Schematic illustration of the polarization-dependent redistri-bution of trapped electrons and the resulting depth-dependent oxygenvacancy charge state in La-doped Hf0.5Zr0.5O2 capacitors under (a)P↑ and (b) P↓ remanent polarization. Over the 30-minute timescaleof this study, the ionic drift response is expected to be limited. In-stead, the polarization discontinuity at the interfaces enhances chargeredistribution. In P↑ state electrons are trapped at V..O near the top in-terface and detrapped from neutral VXO at the bottom interface. Theopposite process occurs in the P↓ state. (c) Suggested exchange be-tween V..O-Hf3+ and VxO-Hf4+ defects by trapping/detrapping.ing V..O-Hf3+ complexes via transfer of the released electronsonto neighboring Hf 5d orbitals (5d1). This increases the mea-sured Hf3+ fraction. For P↑, on the contrary, electron trappingat V..O results in the formation of VxO-Hf4+ complexes by de-populating the Hf 5d orbitals (Hf3+ (5d1)→ Hf4+ (5d0)), i.e.electron transfer from Hf3+ to V..O to form an Hf4+-VXO com-plex. This reduces the Hf3+ signal and weakens the contri-bution from vacancy-related electronic screening. In both po-larization states, the balance between V..O-Hf3+ and VxO-Hf4+complexes near the top electrode is determined by the totalenergy of the system, V..O having a lower energy near to neg-ative bound polarization charge. Edep contributes to providinga driving force for redistribution of charge, but the V..O kineticsare limited by the high activation energy. In this picture, elec-tron trapping/detrapping processes are essentially local, anddo not imply long-range ”conduction-like” electron transportacross the ferroelectric film.The 8 keV spectra, which probe the entire ferroelectricthickness, show a markedly higher V..O fraction for P↑ thanP↓ (Fig. 4e). If the total vacancy population remains con-stant on the ∼30 min timescale, these changes are attributedto negative bound polarization charge at the top interface (P↓)promoting detrapping and vacancy ionization, while positivebound charge at the bottom interface (P↓) favor vacancy neu-tralization. The opposite applies for P↑ at both interfaces andthe scenario is consistent with the low activation energy ex-tracted from imprint kinetics, pointing to a fast, low-barrierelectronic mechanism rather than slow ionic drift.The Supplemental Material includes a quantitative depth-profile simulation of V..O based on Hf3+ fractions from theexperimental 6 and 8 keV measurements. Modeling the 105nm film as five, 2 nm sublayers, and assuming a linear elec-trostatic potential profile within the HZLO film with magni-tudes and sign derived from the experimental imprint resultspresented in Fig. 3(a), shows that the 0.6% difference in theaveraged 8 keV measurement between P↑ and P↓ translatesinto a much larger contrast in the vicinity of the bottom inter-face: ∼1.2% difference (∼1.20% for P↑ and ∼0.01% for P↓).The predicted depth concentration profile reinforces the pro-posed mechanism: electron detrapping from VXO near negativepolarization charges boosts V..O-Hf3+ fraction, while electrontrapping at V..O near positive polarization charges favors VxO-Hf4+, reducing V..O for better screening efficiency. Indeed, V..Oprefer energetically being in proximity to negative polariza-tion charge36.Previous studies under more static conditions (typicallyseveral days to weeks after poling) have demonstrated thatthe Hf3+/Hftot ratio can be used to accurately quantify thefixed V..O concentration changes induced by electrode mate-rial, doping, or annealing conditions24,37,38, and to evaluateV..O drift over longer timescales39,40. Our findings indicatethat this spectroscopic signature, on shorter timescales (here30 min after poling), also reflects a snapshot of charge occu-pancy variation, under Edep. The ∼0.1 eV activation energystrongly suggests that shallow electron trapping/detrappingdominates the early imprint evolution, consistent with priorfindings7,9. Shvilberg et al30 reported a diffusion coefficientof 10−18 cm2.s−1 for oxygen vacancies with an activation en-ergy of 1 eV, corresponding to an average migration distanceof only 1 nm over 3 hrs. One must also consider the possi-bility of electron injection from the electrode under negativebias. In order to assess the possibility of charge injection, wehave calculated the SBH from the core level spectra.The SBH is the offset between the conduction band edge ofHZLO and the Fermi level of the metal electrode. Figure 6shows the bias induced displacements of the Hf 3d5/2. Thespectra shift towards EF by 300 meV when switching from P↑to P↓. The grounded Ti 1s spectra from the top electrode arealso included providing the binding energy reference and, asexpected, shows no shift, confirming the metallic nature.FIG. 6. Polarization-induced band shifts near top interface as de-tected on the Ti 1s and Hf 3d5/2 emission lines at 6 keV.The SBH deduced from Fig. 6 are 2.55 eV for P↑ and 2.85eV for P↓, respectively (see Supplemental Material). Themeasured top SBH and the imprint-derived internal field af-ter 30 min (∼0.4-0.5 MV.cm−1) imply that, at 0 V, the in-terfacial layer field is ∼2.0–2.5 MV.cm−1 (εHZLO/ε IL∼5), in-sufficient for significant thermionic or Fowler–Nordheim in-jection. We can therefore exclude charge injection as theprincipal origin of the imprint change on the experimentaltimescale. As shown in Fig. 2, the measured current around1.5 V is dominated by switching and dielectric contributions,while the leakage remains very low and does not increase sig-nificantly even up to 3 V. This suggests that any fresh chargeinjection or field-assisted transient tunneling is restricted tothe ms time window of the polarization pulse and likely setsthe initial trap occupancy, but cannot explain the imprint evo-lution with time7. At longer timescales (tens of minutes at 0V), the relevant mechanism is therefore charge transfer/back-transfer between oxygen vacancy states and neighboring Hf5d orbitals (Hf4+ ↔ Hf3+), rather than ongoing injection orlong-range transport across the film.The change in SBH with polarization, ∆ϕ(B,n) allows ac-cessing the effective screening length via Eq.1. Using 2Pr=35µC.cm−2 and the measured ∆ϕ(B,n)= 0.3 eV, the λeff is 0.04 Å.The value is similar to the typical effective screening length of0.07 Å calculated for the Au/PMN-PT interface17. The lowervalue obtained here may be due to the screening contributionof the oxynitride interface layer and to the effectiveness of thecharge redistribution at the two TiN/HZLO interfaces due tothe density of V..O24.In conclusion, HAXPES was used to probe the link betweenV..O profile and imprint in fully woken-up HZLO capacitors.Imprint values measured within 30 minutes and the low acti-vation energy (∼0.1 eV) indicate a fast electronic mechanism,primarily charge trapping/detrapping at pre-existing oxygenvacancies. The polarization-dependent modulation of the V..Odepth profile reflects redistribution of electronic charge with-out physical vacancy motion, stabilizing polarization via in-ternal field modulation. While vacancy drift is absent on thistimescale, slower ionic migration may contribute over longerdurations, leading to permanent imprint shifts. Very fast in-jection may occur during the switching pulse (ms range) andlikely sets the initial trap occupancy. Future time-resolvedphotoemission with higher temporal resolution could sepa-rate these contributions over timescales from 1 to 105 s. Theobserved polarization-dependent Schottky barrier height sup-ports the imperfect screening model, underscoring the impor-tance of controlling V ..O density and interface quality for im-proved reliability.The Supplemental Material includes a quantitative depth-profile simulation of V..O based on Hf3+ fractions from theHAXPES data and a description of the method used to de-termine the Schottky barrier height.This project has received funding from the EuropeanUnion’s Horizon 2020 research and innovation programmeunder grant agreement 780302 3εFERRO. The synchrotronradiation experiments were performed with the approval ofthe Japan Synchrotron Radiation Research Institute (JASRI,proposal no. 2020A4908). We acknowledge access to theNanofabrication platform at CEA/SPEC. The data that sup-6port the findings of this study are available from the corre-sponding author upon reasonable request.1T. S. Böscke, J. Müller, D. Bräuhaus, U. Schröder, and U. Böttger, Ap-plied Physics Letters 99, 102903 (2011).2W. Banerjee, A. Kashir, and S. Kamba, 18, 2107575 (2022).3H. Mulaosmanovic, E. T. Breyer, S. Dünkel, S. Beyer, T. Mikolajick, andS. Slesazeck, Nanotechnology 32, 502002 (2021).4E. Covi, H. Mulaosmanovic, B. Max, S. Slesazeck, and T. Mikolajick,Neuromorphic Computing and Engineering 2, 12002 (2022).5T. Mikolajick, S. Slesazeck, M. H. Park, and U. Schröder, MRS Bulletin43, 340 (2018).6R. Alcala, F. Mehmood, P. Vishnumurthy, T. Mittmann, T. Mikolajick,and U. Schröder, 2022 IEEE International Memory Workshop, IMW2022 - Proceedings , 31 (2022).7S. Lancaster, P. D. Lomenzo, M. Engl, B. Xu, T. Mikolajick, U. Schröder,and S. Slesazeck, Frontiers in Nanotechnology 4, 939822 (2022).8F. Mehmood, M. Hoffmann, P. D. Lomenzo, C. Richter, M. Mater-ano, T. Mikolajick, and U. Schröder, Advanced Materials Interfaces 6,1 (2019).9P. Yuan, G-Q. Mao, Y. Cheng, K-H. Xue, Y. Zheng, Y. Yang, and P. Jiang,Nano Res. 15, 3667 (2022).10M. Pešić, F. P. G. Fengler, L. Larcher, A. Padovani, T. Schenk, E. D.Grimley, X. Sang, J. M. LeBeau, S. Slesazeck, U. Schröder, and T. Miko-lajick, Advanced Functional Materials 26, 4601 (2016).11A. Jan, T. Rembert, S. Taper, J. Symonowicz, N. Strkalj, T. Moon, Y. S.Lee, H. Bae, H. J. Lee, D. H. Choe, J. Heo, J. MacManus-Driscoll,B. Monserrat, and G. Di Martino, Advanced Functional Materials 33,2214970 (2023).12D. R. Islamov, V. A. Gritsenko, T. V. Perevalov, V. A. Pustovarov, O. M.Orlov, A. G. Chernikova, A. M. Markeev, S. Slesazeck, U. Schröder,T. Mikolajick, and G. Y. Krasnikov, Acta Materialia 166, 47 (2019).13A. Chouprik, D. Negrov, E. Y. Tsymbal, and A. Zenkevich, Nanoscale13, 11635 (2021).14C. Zacharaki, P. Tsipas, S. Chaitoglou, L. Bégon-Lours, M. Halter, andA. Dimoulas, Applied Physics Letters 117, 212905 (2020).15M. Stengel, P. Aguado-Puente, N. Spaldin, and J. Junquera, Physical Re-view B 83, 235112 (2011).16A.M. Bratkovsky, and A.P. Levanyuk, arxiv.org/abs/0801.1669 (2008).17E. Kröger, A. Petraru, A. Quer, R. Soni, M. Kalläne, N. a. Pertsev,H. Kohlstedt, and K. Rossnagel, Physical Review B - Condensed Mat-ter and Materials Physics 93, 235415 (2016).18M. Müller, P. Lömker, P. Rosenberger, M. H. Hamed, D. N. Mueller,R. A. Heinen, T. Szyjka, and L. Baumgarten, Journal of Vacuum ScienceTechnology A 40, 13215 (2022).19E. A. Kraut, R. W. Grant, J. R. Waldrop, and S. P. Kowalczyk, PhysicalReview Letters 44, 1620 (1980).20C. S. Fadley, Springer Series in Surface Sciences, Vol. 59 (2016) pp. 1–34.21J. E. Rault, G. Agnus, T. Maroutian, V. Pillard, P. Lecoeur, G. Niu,B. Vilquin, M. G. Silly, A. Bendounan, F. Sirotti, and N. Barrett, PhysicalReview B - Condensed Matter and Materials Physics 87, 155146 (2013).22Y. Matveyev, V. Mikheev, D. Negrov, S. Zarubin, A. Kumar, E. D. Grim-ley, J. M. Lebeau, A. Gloskovskii, E. Y. Tsymbal, and A. Zenkevich,Nanoscale 11, 19814 (2019).23Z. Liu, K. Toprasertpong, Z. Cai, M. Takenaka, and S. Takagi, AppliedPhysics Letters 125, 072904 (2024).24W. Hamouda, C. Lubin, S. Ueda, Y. Yamashita, O. Renault, F. Mehmood,T. Mikolajick, U. Schröder, R. Negrea, and N. Barrett, Applied PhysicsLetters 116, 252903 (2020).25P. Buragohain, A. Erickson, P. Kariuki, T. Mittmann, C. Richter, P. D.Lomenzo, H. Lu, T. Schenk, T. Mikolajick, U. Schröder, and A. Gruver-man, ACS Applied Materials and Interfaces 11, 35115 (2019).26S. Vecchi, F. M. Puglisi, P. Appelt, R. Guido, X. Wang, S. Slesazeck,T. Mikolajick, and S. Lancaster, Advanced Electronic Materials 11,2400204 (2025).27S. Ueda, Y. Katsuya, M. Tanaka, H. Yoshikawa, Y. Yamashita, S. Ishi-maru, Y. Matsushita, and K. Kobayashi, AIP Conference Proceedings1234, 403 (2010).28S. Tanuma, C. J. Powell, and D. R. Penn, Surface and Interface Analysis43, 689 (2011).29N. Fairley, “http://www.casaxps.com/,” (2020).30L. Shvilberg, C. Zhou, M. K. Lenox, B. L. Aronson, S. T. Jaszewski, E. J.Opila, N. K. Lam, C. Zhou, M. K. Lenox, B. L. Aronson, N. K. Lam,S. T. Jaszewski, E. J. Opila, and J. F. Ihlefeld, Applied Physics Letters124, 2252905 (2024).31Y. A. Matveyev, A. M. Markeev, Y. Y. Lebedinskii, A. A. Chouprik, K. V.Egorov, W. Drube, and A. V. Zenkevich, Thin Solid Films 563, 20 (2014).32W. Hamouda, A. Pancotti, C. Lubin, L. Tortech, C. Richter, T. Mikola-jick, U. Schröder, and N. Barrett, Journal of Applied Physics 127, 64105(2020).33M. Pešić, F. L. G. Fengler, L. Larcher, A. Padovani, T. Schenk, E. D.Grimley, X. Sang, J. M. LeBeau, S. Slesazeck, U. Schröder, and T. Miko-lajick, Advanced Functional Materials 26, 4601 (2016).34U. Schröder, C. Richter, M. H. Park, T. Schenk, M. Pešić, M. Hoffmann,F. P. G. Fengler, D. Pohl, B. Rellinghaus, C. Zhou, C. C. Chung, J. L.Jones, and T. Mikolajick, Inorganic Chemistry 57, 2752 (2018).35L. Azevedo Antunes, S. Obernberger, P. Schwermer, J. Hintz, R. Ganser,U. Schröder, T. Mikolajick, and A. Kersch, Journal of Applied Physics137, 1 (2025).36Y. Mi, G. Geneste, J. Rault, C. Mathieu, A. Pancotti, and N. Barrett,Journal of Physics: Condensed Matter 24, 275901 (2012).37R. Alcala, M. Materano, P. D. Lomenzo, P. Vishnumurthy, W. Hamouda,C. Dubourdieu, A. Kersch, N. Barrett, T. Mikolajick, and U. Schröder,Advanced functional materials 33, 2303261 (2023).38M. Engl, W. Hamouda, I. Häusler, S. Lancaster, L. Carpentieri, T. Miko-lajick, C. Dubourdieu, and S. Slesazeck, ACS Applied Materials & Inter-faces 17, 51468 (2025).39N. Barrett, W. Hamouda, C. Lubin, J. Laguerre, C. Carabasse, N. Vaxe-laire, J. Coignus, S. Martin, and L. Grenouillet, Applied Physics Letters125, 43502 (2024).40W. Hamouda, F. Mehmood, T. Mikolajick, U. Schröder, T. O. Mentes,A. Locatelli, and N. Barrett, Applied Physics Letters 120, 202902 (2022).