Colossal magnetoresistance from spin-polarized polarons in an Ising system Authors: Ying-Fei Li, Emily M. Been, Sudhaman Balguri, Chun-Jing Jia, Mira B. Mahendru, Zhi-Cheng Wang, Yi Cui, Su-Di Chen, Makoto Hashimoto, Dong-Hui Lu, Brian Moritz, Jan Zaanen, Fazel Tafti, Thomas P. Devereaux, Zhi-Xun Shen Published: 2024 ABSTRACT SignificanceIn pursuit of exotic transport properties in quantum materials, colossal magnetoresistance (CMR) stands out as a beacon for both high intellectual intrigue and critical technological applications. Here, by systematically investigating the photoemission spectroscopy of EuCd2P2, we propose an uncharted mechanism toward CMR arising from the scattering of polarons at the ferromagnetic cluster boundaries. In a broader view, our results exemplify how multiple cooperative interactions can give rise to extraordinary behaviors in condensed matter systems. The traditional framework for understanding magnetoresistance encompasses a Kondo–Ruderman–Kittel–Kasuya–Yosida (RKKY) transition in heavy fermion systems (1, 2), a ferromagnetic-paramagnetic phase transition in concert with structural distortions in manganates (3–5), or a topological phase transition in Dirac semimetals (6–9). However, the recently discovered colossal magnetoresistance (CMR) in a generic family of materials, particularly in EuCd2P2 (Fig. 1B) (10–16), in the absence of these traditional features, points to a new mechanism. Furthermore, the unconventional CMR effect enriches the physics in the EuM2X2 (M = Cd, In, Zn; X = P, As) family where the strong correlations and topology are being actively explored (17–19). Overview of the electronic structure: the unique p-f mixing low-energy states. (A1–A3) Angle-resolved photoemission spectroscopy (ARPES) spectra along the Γ-K cut at 25 K superimposed with DFT-calculated band structures at kz=0 for EuCd2As2 (A1), EuCd2P2 (A2), and SrCd2P2 (A3). (B) Temperature dependence of the resistivity for EuCd2As2 and EuCd2P2. The Inset, scaled by 103, highlights less pronounced magnetoresistant peak of EuCd2As2 compared to that of the EuCd2P2. TN and TMR denote the magnetic phase transition temperature and resistivity maximum temperature, respectively. Detailed transport characterizations are presented in SI Appendix, section 4. (C) Spectral weight variations of the localized 4f bands and the dispersive p bands near the Eu 4d→4f resonant transition. The spectral weights are extracted from the photoemission spectra on EuCd2P2. Inset exhibits the crystal structure with arrows indicating the A-type antiferromagnetic ground states below TN in both EuCd2As2 and EuCd2P2. (D1 and D2) The low-energy DFT band structure along K-Γ-A for EuCd2As2 (D1) and EuCd2P2 (D2). The color map shows Eu 4f orbital weight projected onto the bands. (E1 and E2) Orbitally projected density of states (DOS) for EuCd2As2 (E1) and EuCd2P2 (E2) at low energy. Insets show the integrated projected-DOS for pnictogen p orbitals and Eu 4f orbitals normalized by the total DOS with the integration window marked as the shaded areas. Previous explanations of the CMR in Eu2Cd2P2 include an insulator-metal transition below TMR (13) and the free-electron scattering at the ferromagnetic boundary (14). Zhang et al. proposed a dual-phase model anchored on the observed band reconstruction at low temperatures, and consequently attributed the resistivity increase above TMR to a nonmagnetic gaped phase and the subsequent decrease below TMR to an additional EF-crossing band in the antiferromagnetic metallic phase (13), in a similar spirit to EuO (20). However, the proposed scenario is potentially insufficient to explain the following: i) the resistivity at T≪TMR is similar to that at T≫TMR and still falls outside of the metallic region (10, 12, 13), unlike the dramatic change over twelve decades in EuO (20); ii) the extracted gap in angle-resolved photoemission spectroscopy (ARPES) spectra is insufficient to account for the resistivity increase above TMR; iii) the bad metal behavior in transport at T≫TMR suggests a deviation from a gapped system (details in SI Appendix, section 4); and iv) the disconnect between spectroscopy and transport in that the TMR does not align with the temperature at which the band reconstruction feature emerges (details in SI Appendix, section 2.D). Alternatively, Sunko et al. observed the formation of the in-plane ferromagnetic clusters near TMR in the resonant elastic X-ray scattering and magneto-optical polarimetry (14). Consequently, they attribute the enhancement of resistivity near TMR to the free-electron scattering at the boundary of ferromagnetic clusters. The evidence of ferromagnetic clusters from these photon probes is further supported by detailed analysis on magnetic transport measurements (16); and they will likely play an important role in the physics of the material. Usachocv et al. further performed LDA+U calculations of the electronic structure, aiming at addressing the physics within the framework which also only involves electrons scattering at magnetic domains (16). However, this conventional spin-fermion coupling scenario fails to adequately account for the substantial resistivity change either, as this Ising-like picture would typically give a much smaller magnetoresistance (21) than that in EuCd2P2 (10). Compounding the complexities is the discrepancy in the magnitude of resistivity and their temperature dependence across samples, which, despite comparable magnetoresistance percentage, differs by both three orders of magnitude and the slope in high-temperature regime in two reports (13, 14). Our research focuses on more metallic samples, consistent with (10, 14) (SI Appendix, Fig. S20). In this study, we utilize ARPES and density functional theory (DFT) calculations to decipher these puzzles. We provide spectral results that have close correspondence with the transport, ensuring their relevance to the CMR physics in the bulk material. We then show the spectral lineshapes that indicate the strong presence of polaron formation, which in turn can explain the amplified CMR at TMR from scattering at ferromagnetic cluster interfaces, as well as the bad metal behavior at T≫TMR. Using resonance soft X-ray photoemission and density functional theory, we show that the origin of the enhanced polaron physics in Eu2Cd2P2 versus an other material Eu2Cd2As2 in the same family as coming from strong f state mixing into the p bands near the Fermi level. Our finding unified competing pictures for CMR in these materials and provided microscopy ingredients to engineering future CMR materials. More generally, this result provides further support for the general theme of extreme properties arising from cooperative interactions of different degrees of freedom in complex materials. Results We initiate our discussion by a comprehensive comparison of the ARPES spectra, the band structure calculated by the DFT, and the orbitally projected density of states in EuCd2As2, EuCd2P2, and the nonmagnetic counterpart, the SrCd2P2, respectively (details in SI Appendix, section 1). While EuCd2As2 may potentially exhibit a topologically nontrivial band structure under a magnetic order with proper symmetry (17, 22), the electronic structure of EuCd2P2 is topologically trivial (23) with a discernible gap near the chemical potential μ (Fig. 1 A and B), ruling out the possibilities of a topological phase transition as a potential candidate for CMR. Compared to the SrCd2P2, the ARPES spectra of EuCd2As2 and EuCd2P2 reveal two main feature: the localized Eu 4f states within the −1.5 eV to −1 eV energy range and the dispersive pnictogen p bands extending toward μ (Fig. 1A1 and A2). Unlike the coexisting itinerant electrons and localized f electrons at low energy in heavy fermion systems (1), the dispersive p bands solely dominating the low-energy states indicates a departure from the traditional Kondo-RKKY picture in EuCd2P2. We then focus on the low-energy p states and present a temperature-dependent set of ARPES spectra near TMR, the corresponding momentum distribution curves (MDCs) at a binding energy of 150 meV, and the temperature-varied MDC differential plots in Fig. 2A–C, respectively. At face value, we observe completely incoherent spectra with consistently suppressed low-energy spectral weight, with no resemblance to that of a Landau quasiparticle. Consequently, our analysis is directed toward higher binding energy, and we nevertheless observe a further spectral weight suppression around the TMR, which becomes increasingly apparent after the background subtraction, spectral weight integration, and normalization in Fig. 2 D and E. As expected from transport-spectroscopy correspondence, the spectral weight mirrors the anomalous resistivity change near TMR and is absent in the nonmagnetic counterpart, the SrCd2P2 (Fig. 2E), highlighting its significant link to CMR. Such spectroscopic transport correspondence was not observed in previous photoemission studies (13, 16). Spectral weight suppression as the spectroscopic correspondence to the colossal magnetoresistance (CMR). (A) ARPES spectra of EuCd2P2 below, at, and above TMR in the Left, Middle, and Right panels, respectively. The Right panel also displays the three bands fitted from MDCs (α in red, β in orange, and γ in blue). A parabolic fit to the β band dispersion highlights its Fermi level-crossing character (white dashed line). (B) MDCs at a binding energy of 150 meV (marked by the red dashed line in A) at various temperatures. Spectral weight at TMR undergoes pronounced suppression. (C) MDCs in (B) subtracted by the one at 9 K. (D) The temperature dependence of the integrated spectral weight at various binding energies. The red shaded area in (B) indicates the integration area. The spectral weight suppression is more pronounced at lower binding energies. (E) The temperature dependence of integrated spectral weight in EuCd2P2 and SrCd2P2 along with the negative excessive resistivity ρ(2K)−ρ(T). The spectral weight in EuCd2P2 exhibits a profile akin to anomalous resistivity change and is absent in SrCd2P2. Error bars of spectral weight indicate the variations at different binding energies in (D). Error bars of temperature indicate the difference between the thermal diode reading and the actual sample temperature. The colored background indicates the antiferromagnetism (AFM), short-range ferromagnetic clusters (FM cluster), and paramagnetism (PM) from low to high temperatures. To further discern the essential elements contributing to CMR, the striking difference in transport between EuCd2P2 and EuCd2As2 provides a valuable comparative basis. Hence, we compare the ARPES spectra of EuCd2P2 and EuCd2As2 in Fig. 3. Consistent with the DFT results, the ARPES spectra exhibit three sets of dispersive bands in both materials, labeled as α, β, and γ, respectively (Fig. 3A1 and B1). Notably, the spectral weight of the α band in EuCd2P2 shows a marked reduction in spectral weight relative to that in EuCd2As2, a disparity that is more pronounced in the second-momentum-derivatives (Fig. 3A2 and B2). This results in a clear divergence between single-peak-like and multi-peak-like (MDCs, Fig. 3A3 and B3). Unlike the manganates or cuprates where the low-energy spectra are dominated by a single band, the multiband nature here introduces complexities and necessitates spectral simulation for interpretation. Utilizing the DFT-calculated band structure, we present the simulated ARPES spectra with small and large imaginary parts of self-energy (Σ″) for EuCd2P2 (Fig. 3A4 and A5) and for EuCd2As2 (Fig. 3B4 and B5), respectively, after careful treatments of the photoemission matrix elements and the kz resolution (details in SI Appendix, section 3). The simulated MDCs are juxtaposed with the experimental results in Fig. 3A6 and B6, where the single-peak-like (multi-peak-like) scenario signifies strong (weak) interactions in EuCd2P2 (EuCd2As2). As Σ″ captures the many-body interactions in the spectral function, the difference in Σ″ points to more pronounced interactions in EuCd2P2 compared to that in EuCd2As2. Low energy spectra in the polaronic limit. (A1 and A2) ARPES spectra of EuCd2P2 (A1) and EuCd2As2 (A2) along the K-Γ-K cut at 25 K. Three pnictogen p bands (α, β, and γ) are fitted from momentum distribution curves (MDCs). (B1 and B2) Second-momentum-derivative of (A1 and A2). Red arrows indicate the α band with suppressed spectral weight in EuCd2P2. (C1 and C2) MDCs from the energy window marked by the grey arrow in (A1 and A2). MDCs of EuCd2P2 show a single-peak line shape in (A3), contrasting with the multi-peaked MDCs of EuCd2As2 in (B3). (D1, D2, E1, and E2) Simulated ARPES spectra with small and large imaginary parts of self-energy (Σ″) in (D1) and (E1) for EuCd2P2 and in (D2) and (E2) for EuCd2As2, respectively. (F1 and F2) MDCs at a binding energy of −0.4 eV from the simulated ARPES spectra with small (red) and large (blue) Σ″ juxtaposed with the experimental MDC (black) for EuCd2P2 (F1) and EuCd2As2 (F2), respectively. The MDC with large (small) Σ″ agrees with the experimental one qualitatively in EuCd2P2 (EuCd2As2) and is highlighted accordingly. We explore the origin of these enhanced interactions. We first exclude the possibility of collective spin excitations, as the single-peak MDCs in EuCd2P2 persist up to 150 K, well above the magnetic transition and deep into the paramagnetic phase (details in SI Appendix, section 2.C). A more likely candidate for these persistently enhanced interactions is the electron–phonon (e-ph) coupling, further corroborated by several factors: i) a “bad metal” behavior at T≫TMR in transport (Fig. 1B) (10), ii) a diminished spectral weight at low energy devoid of Landau quasiparticles (Fig. 2A), iii) a broadened spectral dispersion aligned with the DFT results, indicative of strongly polaronic limit (5, 24, 25), and iv) the Gaussian-enveloped energy distribution curves due to Frank–Condon broadening (SI Appendix, Fig. S19) (26). Consequently, we highlight the polaronic interactions for the itinerant p electrons as a critical factor in the emergence of the CMR. A unique p-f mixing in the low-energy states could serve as the microscopic origin of the polaronic interactions. We show the orbitally projected band dispersion and the density of states for EuCd2As2 and EuCd2P2 in Fig. 1D1, D2, E1, and E2, respectively. We note a substantial mixing of polaronic 4f states into pnictogen p states near μ (Fig. 1D1 and D2), which is more pronounced in EuCd2P2 (Fig. 1E1 and E2) and echoes the comparatively larger polaronic interactions in EuCd2P2 as discussed above. The prominent p-f mixing is further supported experimentally by the fact that the spectral weight of the pnictogen p bands follows the same trend as that of the Eu 4f bands across the Eu 4d→4f resonant transition (Fig. 1C), details in SI Appendix, section 2.A). Remarkably, the p-f mixing not only enhances the polaronic interactions on low-energy states in the phonon channel but also enables the conducting p electrons carrying partial magnetic moments from f electrons, facilitating a direct coupling to the magnetic clusters in the spin channel. As such, we suggest the presence of spin-polarized polarons as the conducting electrons in EuCd2P2. Discussion With the establishment of spin-polarized polarons, we now discuss the mechanism for the resistivity enhancement near TMR. Primarily, the CMR in EuCd2P2 originates from the scattering of conducting electrons at the boundaries of ferromagnetic clusters near TMR and dissipates in (anti)ferromagnetic phase (T ≪ TMR) or paramagnetic phase (T ≫ TMR), mirroring the time-reversal symmetry breaking in ref. 14 and aligning with the decoherence of electrons reflected by spectral weight suppression near TMR (Fig. 2E). However, this classical Ising picture involving near-free conducting electrons typically yields ∼5% magnetoresistance (21), insufficient to account for >5,000% resistivity changes in EuCd2P2. To reconcile such dramatic difference, we show the critically helpful role of strong e-ph coupling by considering a tunneling process for spin-polarized polarons across a magnetic cluster boundary. In this process, the tunneling current I in the small voltage limit is given by:I=4eV∑k,p∑σ,σ′Tk,pσ,σ′2∫dxπAσL(k,x)Aσ′R(p,x)∂f∂x, where e is the charge of electron, V is the voltage, Tk,pσ,σ′ is the tunneling matrix element between the momentum-spin channel (k,σ) and (p,σ′), and f is the Fermi–Dirac distribution with spatial dependence. AσL,R is the spectral function on either side of the ferromagnetic cluster in the form ofAk,σL,R=πZδ(x−ϵk±Jσ), where Z is the quasiparticle residue, ϵk is the energy of electrons with momentum k, and J is the spin exchange splitting. Neglecting the momentum dependence of quasiparticle residue and tunneling matrix for this localized process, the resistance across the ferromagnetic cluster wall becomesVI∝1Z2·1D, with the joint density of spin-split states D=4π|T|2∑σ,k,pδ(ϵk±σJ)δ(ϵp±σJ). Therefore, the CMR in EuCd2P2 can be conceptually decomposed into two factors. First, the formation of ferromagnetic clusters increases the tunneling resistance by a reduction in the joint density of spin-split states D under large spin splitting. Second, the quasiparticle residue Z is significantly diminished by strong e-ph coupling, which amplifies the resistivity change by 1/Z2. This scenario highlights the cooperative interactions between the spin and phonon channels that account for pronounced resistivity modulation. In summary, we propose a scenario where the CMR arises from the scattering of the spin-polarized polarons at the interfaces of ferromagnetic cluster in bad metals. In stark contrast with the magnetoresistance from the conventional spin-fermion coupling (21), it is the critically helpful role of e-ph coupling that dramatically amplifies the magnitude. Such polaronic behavior may be the characteristic of the entire EuM2X2 material family (M = Cd, In, Zn; X = P, As) with EuCd2P2 being the most extreme case, and offer insights into their strong correlations and topological properties. In a broader view, our findings echo those in the high-temperature superconductivity, where Tc is substantially enhanced by e-ph coupling (27, 28), illustrating the potential for multiple collaborative interactions to catalyze dramatic changes during phase transitions. Materials and Methods Experiments. The samples, grown by the flux method (10), demonstrated good homogeneity within the same batch in transport measurements. The resistivity measurements were thereby measured on a sample that is either the same or within the identical batch for the ARPES experiments. The ARPES measurements were mainly performed with a Scienta DA-30 analyzer with an overall energy resolution of 12 meV at Beamline 5-2, Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory. Temperature-dependent measurements are carefully controlled on sample aging by limiting the measurement times below 20 h and maintaining the vacuum better than 2×10−11 Torr below 25 K and than 4×10−11 Torr above 25 K. Photon-flux-dependent measurements were also performed on the samples to confirm the absence of the charge effect. The Fermi edge of polycrystalline gold was measured after each high statistic spectra on samples to determine the temperature-dependent chemical potential that also fluctuates <1 meV in 1 h. The samples are prepolished with large and flat surfaces to overcome the difficulties in sample cleavage. A real space mapping utilizing a small beam spot (60 μm × 40 μm diameter) at the synchrotron beamline station was carried out before taking photoemission spectra to ensure a consistent dispersion across a surface area greater than 250 μm × 250 μm. DFT Calculations. The band structures and magnetic configuration energies were calculated with DFT implemented in the Vienna ab initio simulation package (29–32), using the projector-augmented wave method (33) and the Perdew–Burke–Ernzerhof exchange-correlation functional (34). We used the europium pseudopotential with 4f electrons explicitly in the valence, with an additional Hubbard U=5 eV and spin–orbit coupling (SOC) to treat their localization properly (10, 19, 35). The DFT+U+SOC calculations were performed using the rotationally invariant DFT+U method (36). We chose the A-type antiferromagnetic as the ground state for calculating band structures. Ionic relaxation was carried out on a k-point Monkhorst-Pack grid sizes of 17×17×9 for 2-Eu supercell and 17×17×17 for 1-Sr unit cell, followed by the self-consistent field calculations using a k-point grid of 21×21×11 for the 2-Eu supercell and 21×21×21 for the single-Sr unit cell. Supplementary Material Appendix 01 (PDF)