WorksPhysics and Astronomy
Disentangling Strain and Ti3+ Contributions to the Anomalous Hall Effect in Epitaxial RuO2 Films
Abstract
The anomalous Hall effect (AHE) reported in epitaxial RuO2/TiO2 has been attributed to a strain-stabilized magnetic state, but strain can be entangled with Ru-Ti intermixing and interfacial charge redistribution, which can produce Ti3+ and potentially localized moments. Here, we disentangle these contributions using Ti-alloyed RuO2 heterostructures in which abundant Ti3+ states are retained while epitaxial strain is independently changed. The strained and relaxed films have nearly identical Ti/Ru compositions and comparable Ti3+ fractions, yet a pronounced nonlinear AHE appears only in the coherently strained film, while the relaxed heterostructure exhibits an almost linear Hall response. Spectroscopic ellipsometry further shows that the AHE-active strained state is accompanied by a reconstruction of the itinerant electronic response, including enhanced metallicity and longer carrier relaxation times. These results show that Ti3+ formation alone is insufficient to generate the anomalous Hall state and identify epitaxial strain, through its modification of the itinerant RuO2 electronic structure, as the dominant control parameter.
*Corresponding authors: [email protected]; [email protected]
Introduction
Epitaxial interfaces provide a powerful route to stabilize electronic and magnetic states absent in bulk materials [1, 2, 3, 4, 5], but the same interface can simultaneously impose lattice strain, drive charge transfer, and promote atomic reconstruction. Separating these coupled effects is therefore essential for identifying the microscopic origin of an interfacial phase. This challenge is particularly acute for , which has emerged as a prominent candidate for altermagnetism [6, 7]. Epitaxial exhibits a broad range of magnetic responses [8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42], despite bulk being largely regarded as nonmagnetic [43, 44, 45, 46, 47, 48, 49]. In particular, fully strained ultrathin ( nm) grown by hybrid molecular beam epitaxy (MBE) remains metallic to atomic-scale thickness and exhibits symmetry-breaking or magnetic signatures in optical second-harmonic generation [19], magneto-optical measurements [19], spin-resolved angle-resolved photoemission spectroscopy (ARPES) [39], polarized neutron reflectometry (PNR) [50], and the anomalous Hall effect (AHE) [20]. Epitaxial strain has been proposed as the driving mechanism [19, 20, 38, 39], but the interface also provides an alternative microscopic route: charge redistribution, and potentially Ru-Ti intermixing, can create states that introduce additional d electrons and may generate localized magnetic moments [51, 52, 53, 54]. A central unresolved question is therefore whether the AHE reflects a strain-reconstructed itinerant state of or instead originates from Ti-related interfacial electronic reconstruction.
In this Letter, we design an experiment that separates these two possibilities. We first use angle-dependent hard X-ray photoemission spectroscopy to establish the presence of at the interface. We then intentionally alloy Ti into (), creating a high density of states throughout the ultrathin conducting layer, while independently tuning its epitaxial boundary condition by inserting a strain-relaxed buffer. The resulting strained and relaxed films retain nearly identical compositions and populations, providing a direct test of whether is sufficient to produce the AHE. Despite comparable Ti populations, the AHE remains robust in the coherently strained film but is strongly suppressed with strain relaxation. Spectroscopic ellipsometry further shows that strain simultaneously reconstructs the itinerant carrier response. These experiments collectively establish epitaxial strain, rather than the mere presence of Ti states, as the dominant control of the anomalous Hall state in ultrathin RuO/TiO-based heterostructures.
Results and Discussion
We first investigated the presence of Ti at RuO/TiO interface. Angle-dependent hard X-ray photoemission spectroscopy (HAXPES) on a 1.4 nm RuO/TiO(110) sample (Fig. S1 [55]) provides depth-sensitive access to the buried interface. Although conventional laboratory XPS does not resolve a Ti component because of its more limited probing depth [56], HAXPES reveals a small but clear Ti spectral weight at grazing take-off angles. Quantitative analysis is consistent with a nm-thick interfacial region containing % Ti. Importantly, modeling of the Ti/Ti ratio gives comparably good agreement for two distinct microscopic scenarios: Ti confined to an interfacial TiO layer or Ti incorporated in an interfacial TiRuO layer. The spectroscopy therefore establishes Ti formation but does not uniquely determine whether it arises from charge transfer or local Ru-Ti intermixing. In bulk, the electron affinity of insulating TiO ( eV) [57] and the work function of metallic RuO ( eV) [58, 59] would suggest a sizable interfacial barrier, whereas ultrathin-interface electrostatics, including strain-induced dipoles and band bending [58], can substantially modify the local potential landscape. Prior multislice electron ptychography of hybrid-MBE-grown RuO/TiO did not resolve appreciable oxygen vacancies or Ru-Ti intermixing within experimental resolution [20, 57]. The microscopic location of Ti therefore remains open, making it essential to determine experimentally whether Ti itself is sufficient to account for the AHE.
We therefore constructed a controlled test of the Ti hypothesis by growing two epitaxial heterostructures with contrasting strain states by hybrid MBE (see Supplemental Material [55]) [19, 20, 38, 39, 56]. Ti alloying intentionally creates abundant Ti-derived reduced states throughout the ultrathin -based conducting layer; it is used here not as an exact microscopic replica of the buried interface, but as a stringent test of whether a large concentration is sufficient to generate the AHE. Strain was changed independently by inserting a buffer, yielding /2 nm / (110) and /2 nm /12 nm / (110). The corresponding bulk lattice mismatches are summarized in Fig. S2 [55, 60, 61]. The same 2 nm layer was included directly beneath in both structures to keep the immediate interface chemistry as similar as possible. Without , is coherently constrained by (110); with , the relaxed buffer imposes a substantially different epitaxial boundary condition. For , density functional theory (DFT) predicts that the strain associated with (2.6% tensile along and 5.5% tensile along ) is nonmagnetic, whereas full coherency to (−4.7% compressive along and 2.3% tensile along ) stabilizes a magnetic ground state [19, 20, 38, 39]. Thus, the two heterostructures are designed to hold Ti chemistry approximately fixed while changing the lattice state expected to control magnetism.
Figure 1 establishes that the two heterostructures differ primarily in their epitaxial boundary condition rather than alloy composition. X-ray diffraction (XRD) - scans confirm phase-pure (110)-oriented growth in both samples, with the diffraction peak appearing only in the buffered structure. Pronounced thickness oscillations in XRD and X-ray reflectivity (XRR), together with Kiessig fringes extending to in Figure 1(b), indicate smooth surfaces and well-defined interfaces. XRR fitting gives 2.9 nm /2.5 nm / (110) and 1.7 nm /1.8 nm /12 nm / (110). Importantly, both thicknesses remain within the ultrathin regime in which coherently strained / has previously exhibited an AHE [20]. More decisive for the present comparison is the alloy composition. The fitted mass densities () are essentially identical, 5.732 and 5.721 g cm without and with , respectively (Figure 1d). Using , gives and 0.46. Thus, within the sensitivity of XRR, the two films have nearly identical Ti/Ru ratios, strongly constraining composition as an explanation for any qualitative difference in Hall response.

Figure 1. Structural and compositional characterization of epitaxial Ti-alloyed RuO heterostructures. (a) XRD – scans of TiRuO films grown on TiO (110) without (top) and with (bottom) a SnO buffer, showing phase-pure (110) orientation. (b) XRR data (symbols) and model fits (solid lines). (c) Bulk mass densities () of RuO and TiO used for composition estimation. (d) extracted from XRR fitting and corresponding Ti compositions (), determined using a linear mixture relation (see text).
We next verify that the two structures impose contrasting lattice states. In-situ reflection high-energy electron diffraction (RHEED) acquired immediately after growth (Fig. S3 [55]) remains sharp and streaky for the film grown directly on [Figure 2(a)], consistent with a smooth coherently constrained surface. The -buffered film instead shows broadened, diffuse streaks [Figure 2(b)], consistent with lattice relaxation. Reciprocal-space maps (RSMs) around the asymmetric (310) and (332) reflections [Figure 2(c)-Figure 2(f)] make this contrast quantitative. Around (332), the unbuffered peak is aligned in-plane with along [1], whereas the buffered structure shifts toward the bulk position. The (310) map shows the same trend along []: the unbuffered film shares the in-plane reciprocal vectors, while the buffered heterostructure approaches the bulk lattice parameter. Because the ultrathin contribution cannot be separately resolved in the buffered RSM, the exact alloy lattice parameter is not directly extracted.

Figure 2. Strain state and Ti valence analysis of Ti-alloyed heterostructures. (a,b) Schematic structures and corresponding RHEED patterns along the azimuth for fully strained (without buffer) and strain-relaxed (with buffer) films. (c,d) RSMs around the (310) reflection. The film peak aligns with the substrate along in-plane without the buffer, indicating coherent strain, while the buffered sample approaches the bulk lattice. (e,f) RSMs around the (332) reflection confirming strain along the direction. (g,h) Ti XPS spectra with and components. Both samples exhibit comparable fractions.
The crucial control is that this change in lattice state does not appreciably change the Ti valence population. Ti X-ray photoelectron spectra [Figure 2(g) and Figure 2(h)] are well described by components assigned to (~456.5 eV) and (~458.0 eV). Quantitative fitting gives fractions of 45.50% in the coherently strained film and 43.45% in the relaxed film. Thus, the two structures contain nearly identical, substantial populations despite their markedly different strain state. Because the ~1.5 nm inelastic mean free path of Al K photoelectrons is comparable to the film thickness, the measured signal is dominated by the layer. Combined with the essentially identical alloy densities and -0.46 compositions in Figure 1, these measurements establish the controlled comparison needed to distinguish -related effects from those of strain.
The Hall response then provides a direct discrimination between the two mechanisms (Figure 3). A pronounced nonlinearity in the field-dependent Hall resistivity, , develops only in the coherently strained film below 5 K, with a lower onset temperature than that reported for fully strained RuO [20], whereas the relaxed film remains nearly linear over the full temperature range. At 1.8 K, decomposition of into ordinary and anomalous contributions (see Supplemental Material [55]) yields a sizable AHE only for the strained film [Figure 3(c)]; in the relaxed heterostructure the anomalous component is strongly suppressed [Figure 3(d) and S4 [55]]. The key point is the controlled contrast: changing the strain state switches the anomalous Hall response while the Ti concentration changes only from 45.5% to 43.5% and the fitted alloy density remains essentially unchanged. Ti is therefore not sufficient to generate the observed anomalous Hall state. Instead, the strain selectivity is consistent with calculations in which the TiO-imposed distortion stabilizes a magnetic RuO electronic structure [19, 20, 38, 39]. The AHE conductivity of strained TiRuO is 0.35 S cm, approximately one order of magnitude smaller than in fully strained RuO/TiO of comparable thickness [20]. Rather than enhancing the AHE, strong Ti incorporation therefore appears to weaken the RuO-derived response. This behavior is also distinct from that reported in doped TiO heterostructures, where the reported AHE conductivity is much smaller (– S cm), decreases on cooling, becomes undetectable below 100 K, and is accompanied by pronounced ferromagnetic hysteresis [61].

Figure 3. Strain-dependent anomalous Hall effect in Ti-alloyed RuO2 heterostructures. (a,b) measured at various temperatures for coherently strained (a) and strain-relaxed (b) films. A pronounced nonlinearity develops only in the strained state. (c,d) Decomposition of at 1.8 K into ordinary (OHE) and anomalous (AHE) Hall contributions. Symbols denote data; lines show the total fits and extracted OHE (gray) and AHE (purple) components.
If strain controls the AHE through the itinerant RuO electronic structure, the AHE-active state should also carry a distinct metallic response. Spectroscopic ellipsometry provides such evidence [62, 63]. Figures 4(a) and 4(b) show the real () and imaginary () parts of the dielectric function measured along [10] and [1]. All films exhibit an epsilon-near-zero (ENZ) energy, (), where , characteristic of a metallic carrier response [64]. In the coherently strained film, shifts to higher energy along both crystallographic directions and the directional splitting is reduced. The increase in corresponds to a larger effective plasma frequency, consistent with increased carrier density and/or reduced effective mass, while the reduced splitting signals a less anisotropic itinerant response. The accompanying evolution of is constrained by the Kramers-Kronig relations. The loss function, , provides an independent view Figure 4(c): the plasmon peak also shifts upward and becomes less anisotropic under strain. Quantitative fitting (see Supplemental Material [55]) shows that both the carrier scattering time and plasmon relaxation time increase in the strained film Figure 4(d), consistent with the decrease in resistance R from 526–705 in the relaxed structures to 144–314 in the fully strained structures. Thus, the same epitaxial constraint that activates the AHE also reorganizes the itinerant carrier dynamics, supporting an origin tied to the strain-reconstructed metallic state rather than to a fixed population of localized Ti moments [20, 62, 63].

Figure 4. Strain-dependent optical response and carrier dynamics. (a,b) Real () and imaginary () dielectric functions from spectroscopic ellipsometry measured along [] and [1]. (c) Loss-function spectra with Lorentzian fits, showing the strain-dependent plasmon response. (d) Summary of , , , , and at 300 K.
In summary, we disentangle two effects that are normally inseparable at the RuO/TiO interface: epitaxial strain and Ti-related electronic reconstruction. HAXPES establishes the presence of Ti at the interface, but Ti-alloyed RuO heterostructures show that even substantially higher Ti concentrations do not by themselves account for a robust AHE. With composition and Ti fractions held nearly fixed, the anomalous Hall response appears only under coherent epitaxial constraint and is strongly suppressed upon strain relaxation. The AHE-active state is simultaneously characterized by a higher-energy, less anisotropic plasmon response, longer carrier relaxation times, and lower resistance. These results identify strain-driven reconstruction of the itinerant electronic structure, rather than Ti formation alone, as the dominant control of the anomalous Hall state. More broadly, they demonstrate how chemically matched but lattice-distinct heterostructures can separate competing interfacial mechanisms in correlated quantum materials.
Acknowledgements
Film synthesis and structural characterizations (S.G.J. and B.J.) were supported by the U.S. Department of Energy through grant Nos. DE-SC0020211, and (partly) DE-SC0024710. Transport, and ellipsometry (at UMN) were supported by the Air Force Office of Scientific Research (AFOSR) through Grant No. FA9550-21-1-0025 and FA9550-24-1-0169. Film growth was performed using instrumentation funded by AFOSR DURIP awards FA9550-18-1-0294 and FA9550-23-1-0085. S.G.J and B.J also benefitted from the support from the Air Force Office of Scientific Research Multi University Research Initiative (AFOSR MURI, Award No. FA9550-25-1-0262). S.S., U.M.J., A.X.G, S.L., and T.L. acknowledge support from AFOSR MURI, Award No. FA9550-25-1-0262. Parts of this work were carried out at the Characterization Facility, University of Minnesota, which receives partial support from the NSF through the MRSEC program under Award No. DMR-2011401. S.G.J also acknowledges support from Hankuk University of Foreign Studies Research Fund. S.A.C. was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, Materials Sciences and Engineering Division, Synthesis and Processing Science program (FWP 10122 at PNNL). The angle-dependent HAXPES was carried out at Diamond Light Source on beamline I09.
Notes
The authors declare no conflict of interest.
Data availability
The data are available from the authors upon reasonable request.
References
- [1]S. A. Chambers, Surf. Sci. Rep. 39, 105 (2000).
- [2]F. Hellman et al., Rev. Mod. Phys. 89, 025006 (2017).
- [3]H. Y. Hwang, Y. Iwasa, M. Kawasaki, B. Keimer, N. Nagaosa, and Y. Tokura, Nat. Mater. 11, 103 (2012).
- [4]J. Mannhart and D. G. Schlom, Science 327, 1607 (2010).
- [5]P. Zubko, S. Gariglio, M. Gabay, P. Ghosez, and J.-M. Triscone, Annu. Rev. Condens. Matter Phys. 2, 141 (2011).
- [6]L. Šmejkal, R. González-Hernández, T. Jungwirth, and J. Sinova, Sci Adv 6, eaaz8809 (2020).
- [7]L. Šmejkal, J. Sinova, and T. Jungwirth, Phys. Rev. X 12, 040501 (2022).
- [8]I. H. Choi, S. G. Jeong, B. Jalan, and J. S. Lee, Nano Converg. 13, 1 (2026).
- [9]T. Berlijn et al., Phys. Rev. Lett. 118, 077201 (2017).
- [10]Z. H. Zhu et al., Phys. Rev. Lett. 122, 017202 (2019).
- [11]C. A. Occhialini et al., arXiv preprint arXiv:2510.13767 (2025).arxiv.org/abs/2510.13767
- [12]B. Z. Gregory et al., arXiv preprint arXiv:2510.13781 (2025).arxiv.org/abs/2510.13781
- [13]M. Hiraishi, H. Okabe, A. Koda, R. Kadono, T. Muroi, D. Hirai, and Z. Hiroi, Phys. Rev. Lett. 132, 166702 (2024).
- [14]P. Keßler et al., npj Spintronics 2, 50 (2024).
- [15]A. Akashdeep et al., Appl. Phys. Lett. 128 (2026).
- [16]O. Fedchenko et al., Sci Adv 10 (2024).
- [17]J. Liu et al., Phys. Rev. Lett. 133, 176401 (2024).
- [18]Z. Lin et al., Phys. Rev. B 111, 134450 (2025).
- [19]S. G. Jeong et al., Proc. Natl. Acad. Sci. U.S.A. 123, e2526641123 (2026).
- [20]S. G. Jeong et al., Proc. Natl. Acad. Sci. U.S.A. 122, e2500831122 (2025).
- [21]D. Wickramaratne, M. Currie, S. S. Fields, C. D. Cress, and S. P. Bennett, J. Mater. Chem. C (2025).
- [22]S. S. Fields et al., ACS Appl. Mater. Interfaces 18, 6166 (2026).
- [23]S. Karube, T. Tanaka, D. Sugawara, N. Kadoguchi, M. Kohda, and J. Nitta, Phys. Rev. Lett. 129, 137201 (2022).
- [24]H. Bai et al., Phys. Rev. Lett. 128, 197202 (2022).
- [25]Y. Guo et al., Adv. Sci. 11, 2400967 (2024).
- [26]X. Feng et al., Phys. Rev. Lett. 132, 086701 (2024).
- [27]Z. Li et al., Adv. Mater. 37, 2416712 (2025).
- [28]Y. Zhang et al., Nat. Commun. 16, 5646 (2025).
- [29]B. Sekh, H. Rahaman, R. Maddu, P. K. Mishra, T. Jin, and S. Piramanayagam, arXiv preprint arXiv:2501.12593 (2025).arxiv.org/abs/2501.12593
- [30]H. Bai et al., Phys. Rev. Lett. 130, 216701 (2023).
- [31]Z. Li, Z. Zhang, X. Lu, and Y. Xu, arXiv:2407.07447 (2024).arxiv.org/abs/2407.07447
- [32]Y. Zhang, H. Bai, L. Han, C. Chen, Y. Zhou, C. H. Back, F. Pan, Y. Wang, and C. Song, Adv. Funct. Mater. 34, 2313332 (2024).
- [33]Y. Fan, Q. Wang, D. Wang, Q. Huang, Y. Chen, L. Bai, and Y. Tian, J. Magn. Magn. Mater. 606, 172398 (2024).
- [34]C.-T. Liao, Y.-C. Wang, Y.-C. Tien, S.-Y. Huang, and D. Qu, Phys. Rev. Lett. 133, 056701 (2024).
- [35]Z. Wang et al., Phys. Rev. Lett. 133, 046701 (2024).
- [36]Y. Kobayashi, S. Karube, I. Sugiura, H. Narita, R. Hisatomi, Y. Shiota, and T. Ono, AIP Adv. 14, 115120 (2024).
- [37]C. Pan, S. Hu, F. Yang, D. Yang, W. Fan, Z. Shi, L. Pan, S. Zhou, and X. Qiu, arXiv:2412.18937 (2024).arxiv.org/abs/2412.18937
- [38]J. D. S. Forte, S. G. Jeong, A. Santhosh, S. Lee, B. Jalan, and T. Low, Phys. Rev. B 114, 084408 (2026).
- [39]Y. Zhang et al., Sci Adv 12, eaec2917 (2026).
- [40]S. S. Fields et al., arXiv preprint arXiv:2509.22866 (2025).DOI
- [41]A. Bose et al., Nat. Electron. 5, 267 (2022).
- [42]C. He, Z. Wen, J. Okabayashi, Y. Miura, T. Ma, T. Ohkubo, T. Seki, H. Sukegawa, and S. Mitani, Nat. Commun. 16, 8235 (2025).
- [43]T. Osumi et al., arXiv preprint arXiv:2501.10649 (2025).DOI
- [44]S. Hahn et al., Phys. Rev. B 111, 214431 (2025).
- [45]M. Wenzel, E. Uyukur, S. Rößler, M. Schmidt, O. Janson, A. Tiwari, M. Dressel, and A. A. Tsirlin, Phys. Rev. B 111, L041115 (2025).
- [46]L. Kiefer et al., J. Phys.: Condens. Matter 37, 135801 (2025).
- [47]S. Paul, G. Mattoni, H. Matsuki, T. Johnson, C. Sow, S. Yonezawa, and Y. Maeno, J. Cryst. Growth 673, 128405 (2026).
- [48]G. Yumnam et al., Cell Rep. Phys. Sci. 6 (2025).
- [49]X. Peng et al., Commun. Mater. 6, 177 (2025).
- [50]S. G. Jeong et al., arXiv preprint arXiv:2601.10518 (2026).arxiv.org/abs/2601.10518
- [51]G.-b. Han, S.-j. Hu, S.-s. Yan, and L.-m. Mei, phys. status solidi RRL 3, 148 (2009).
- [52]A. K. Rumaiz, B. Ali, A. Ceylan, M. Boggs, T. Beebe, and S. Ismat Shah, Solid State Commun. 144, 334 (2007).
- [53]D. Kim, J. Hong, Y. R. Park, and K. J. Kim, J. Phys.: Condens. Matter 21, 195405 (2009).
- [54]M. Parras, Á. Varela, R. Cortés-Gil, K. Boulahya, A. Hernando, and J. M. González-Calbet, J. Phys. Chem. Lett. 4, 2171 (2013).
- [55]See Supplemental Material at http://link.aps.org/supplemental/###/#### for additional information about the experimental methods and a detailed discussions.
- [56]S. G. Jeong et al., Sci Adv 11, eadw7125 (2025).
- [57]A. Imanishi, E. Tsuji, and Y. Nakato, J. Phys. Chem. C 111, 2128 (2007).
- [58]S. G. Jeong et al., Nat. Commun. (2026).
- [59]F. Creazzo and S. Luber, Appl. Surf. Sci. 570, 150993 (2021).
- [60]W. H. Baur and A. A. Khan, Acta Crystallogr. B 27, 2133 (1971).
- [61]A. A. Bolzan, C. Fong, B. J. Kennedy, and C. J. Howard, Acta Crystallogr. B 53, 373 (1997).
- [62]N. Nagaosa, J. Sinova, S. Onoda, A. H. MacDonald, and N. P. Ong, Rev. Mod. Phys. 82, 1539 (2010).
- [63]S. G. Jeong et al., Applied Physics Reviews 12 (2025).
- [64]R. D. Prasetiyawati, S. G. Jeong, C.-K. Park, S. Song, S. Park, T. Park, and W. S. Choi, Curr. Appl. Phys. 53, 110 (2023).