Introduction

The topological (Dirac/Weyl) semimetal, a class of materials with the low-energy electronic excitations described by Dirac/Weyl electron of relativistic theory, exemplifies an active frontier of modern condensed matter science. Triggered by the research on three-dimensional Dirac/Weyl semimetals, various topological semimetals such as the nodal-line semimetal have been successively identified, and the scope of Dirac/Weyl electron in solid is expanding even beyond the scheme of relativistic particle in high-energy physics1,2. A distinguished feature of Dirac/Weyl electron with massless or small effective mass character is its extremely high mobility, yielding a variety of unusual quantum transports. The representative examples are the room temperature quantum Hall effect in graphene3 or chiral anomaly in the Weyl semimetal4. Beyond the single-particle physics, currently, there is a growing interest in the strong-correlation effect on Dirac/Weyl electrons. In the magnetic quantum limit, wherein all electrons occupy the lowest Landau level (LL), the effect of Coulomb interaction is crucial, which offers a novel opportunity to study collective phenomena of relativistic electrons. Indeed, the bulk fractional quantum Hall effect, valley ordering, or nematic quantum liquid have been argued for bithmuth5,6,7. Recent research proposes the ordering of Weyl electron at extremely high magnetic fields above 80 T in the Weyl semimetal TaAs8.

Another route to realize the strong-correlation effect of the highly mobile Dirac/Weyl electrons is to utilize correlated electron materials. So far, only a few candidates of correlated Dirac/Weyl semimetals have been experimentally identified. Among them, the 5d-transition-metal oxide is a promising class of materials, where the interplay between electron correlation and relativistic spin–orbit interaction yields the Dirac/Weyl semimetal coupled to the Mott physics or magnetism9,10. Indeed, the first prediction of magnetic Weyl semimetal has been done for the pyrochlore-type iridates, wherein signatures of correlated Weyl electron have been identified as anomalous magneto-transport phenomena9,10,11,12. However, the strong electron correlation usually decreases the electron mobility along the promotion of electron localization, and thus quantum transport of relativistic electron would be hardly observed in this class of materials.

Contrary to this naive expectation, however, we report here for the strongly correlated Dirac semimetal CaIrO3 that the combination of electron correlation and spin–orbit coupling cooperatively yields highly mobile electrons with the mobility exceeding 60,000 cm2 V−1 s−1, the largest among the oxide semiconductors, as well as the unique quantum oscillation with giant positive magnetoresistivity ratio of 5500%.

Results

Material

Perovskite AIrO3 (A = Ca, Sr) crystallizes in an orthorhombic perovskite structure with the GdFeO3-type lattice distortion as shown Fig. 1a. The electronic state nearby the Fermi energy (EF) is mainly composed of nearly half-filled jeff = 1/2 multiplet of 5d orbitals of Ir4+ ion, leading to a semimetallic state with a few electron and hole pockets13,14,15,16,. The first principles calculation argues that this material is a nodal-line semimetal protected by the nonsymmorphic crystalline symmetry, wherein the conduction band and valence band cross along a closed line in momentum space as depicted in Fig. 1b13. For SrIrO3 in the thin film form, the electron pocket with Dirac-like dispersion has been identified by the angle-resolved photoemission spectroscopy, and the Dirac node positions below EF by more than 50 meV17. A striking feature of this material is the electron correlation can be effectively controlled by the one-electron bandwidth. Indeed, the modest reduction of one-electron bandwidth by the dimensional control or chemical substitution in SrIrO3 causes the semimetal-to-antiferromagnetic insulator transition18,19,20. The one-electron bandwidth can also be controlled by the GdFeO3-type lattice distortion (or Ir-O-Ir bond angle distortion) via replacing Sr with Ca; the effective electron correlation is enhanced approximately by 20% in CaIrO3 compared with SrIrO3, as measured from the increase of the bond angle distortion in the perovskite21. In this context, CaIrO3 would be located in even closer proximity to the Mott transition, offering a rare opportunity to study the Mott criticality of Dirac electron. Nevertheless, the hallmark of Dirac electron such as the highly mobile electron showing the quantum transport has not been detected. One major obstacle is the difficulty in preparing the perovskite crystal structure of CaIrO3, which is a metastable form at ambient pressure. To overcome this issue, we have employed a high-pressure synthesis technique and succeeded in synthesizing high-quality single crystals of perovskite CaIrO3.

Fig. 1
figure 1

Unusually large transport mobility in perovskite CaIrO3. a The crystal structure and b the sketch of Dirac-like dispersion near the line node (blue line). The primitive vectors a, b, and c are defined in the orthorhombic notation with space group Pbnm. c Temperature dependence of the resistivity. The inset is the photograph of single-crystalline CaIrO3. The red bar denotes the length of 100μm. d The Hall conductivity σxy vs. B (B = μ0H, with μ0 the vacuum permittivity) at various temperatures. The reciprocal field of peak or dip as indicated by triangles corresponds to the averaged transport mobility μtr. e, f show the μtr and density (n3D) of highest mobility carrier as a function of temperature. The temperature dependence of electrical conductivity σxx is also shown. g The μtr of typical high-mobility oxide semiconductors and metals21, 32,33,34,35,36,37,38,39,40 as well as the heterostructure of (Mg,Zn)O/ZnO41 and LAO(LaAlO3)/STO(SrTiO3)42. The carrier densities of heterostructures are derived by assuming the effective channel thickness of 10 nm

Full size image

Transport measurements

Figure 1c shows the temperature dependence of resistivity of CaIrO3. The resistivity shows a metallic temperature dependence above 150 K, whereas a notable peak emerges at around 20 K. The sign of Hall conductivity σxy indicates that the dominant carrier is electron type (Fig. 1d). At 0.12 K, a dispersion-type profile with a peak and dip is observed around ±0.14 T, but diminishes as temperature increases. By employing the semiclassical Boltzmann formula (Supplementary Note 1), we extracted the density (n3D) and mobility (μtr) of electrons. As shown in Fig. 1e, the mobility is enhanced with decreasing temperature and reaches 62,000 cm2 V−1 s−1 at 0.12 K, which is exceptionally high among the typical bulk oxide semiconductors (see Fig. 1g). The enhancement of mobility is in parallel to that of electrical conductivity, while the carrier density is nearly temperature independent below 20 K (Fig. 1f). Therefore, it is likely that the peak of resistivity can be understood from the counter-balance of carrier density and mobility, and the highly mobile electrons govern the transport at low temperatures.

Figure 2a shows the magnetoresistivity at 0.12 K measured in varying magnetic field (B) and its direction within the ac plane. For B||a (θ = 0°), the resistivity initially increases up to 2 T, decreases during 2–9 T, and exhibits two orders of magnitude increase above 9 T. By tilting the magnetic field from the a-axis, the giant magnetoresistivity above 9 T is suppressed and the resistivity moderately increases up to 14 T. Moreover, clear Shubnikov–de-Haas (SdH) oscillations are observed below 10 K. For B||a, two SdH oscillations are observed during 1–3 and 2–9 T. The latter (high-field oscillation) diminishes as θ increases, while the former (low-field oscillation) becomes clear at θ = 90° (B||c). Figure 2b, c shows the oscillation components after subtraction of the non-oscillating background for high-field oscillation (B||a) and low-field oscillation (B||c), respectively. We plot the LL fan diagram to extract the oscillation frequency BF and phase shift φ, following the Lifshitz–Onsager quantization rule BF/B = n − φ (inset to Fig. 2a). Here, the peak and valley positions, which are consistent with the second derivative of ρxx ( = dρxx2/dB2), are assigned to integer and half-integer, respectively. The linearity of the fan plot up to the quantum limit may be a consequence of small Zeeman splitting. First, we analyzed the high-field oscillation for B||a. The extracted BF and φ are 11.2T and −0.9, respectively. The corresponding extremal area of Fermi surface (FS) is determined to be SF = 1.0 × 10−3 Å−2 and the carrier density estimated from SF is 2.1 × 1017 cm−3. Because of the absence of oscillation above 9 T and the small extremal area of FS (BF = 11.2 T), it is likely that 9 T is sufficient to reach the quantum limit, where electrons of this FS start to occupy the lowest LL. We extracted the cyclotron mass mc, Fermi velocity vF, and Dingle temperature TD according to the Lifshitz–Kosevich formula,

$$\Delta \rho _{xx}{\mathrm{/}}\rho _{xx} \propto \exp \left( { - x_{\mathrm{D}}} \right)\frac{x}{{\sinh \left( x \right)}}\cos 2{\mathrm{\pi }}\left( {B_{\mathrm{F}}{\mathrm{/}}{B}{\mathrm{ + }}\varphi } \right),$$
(1)

with xD = 2π2kBTD/ħωc and x = 2π2kBT/ħωc. The estimated mc, vF, and TD are 0.31 ± 0.04m0, with m0 the free electron mass, 6.9 ± 0.6 × 104 m s−1, and 3.5 K, respectively (Supplementary Fig. 3a). We similarly analyzed the low-field oscillation for B||c. The oscillation frequency is extracted to be BF = 3.2 T corresponding to the SF = 3.0 × 10−4 Å−2, which is nearly one-third of that for high-field oscillation. Moreover, mc, vF, TD, and φ are determined to be 0.12 ± 0.04m0, 8.7 ± 1.2 × 104 m s−1, 4.5 K, and −0.3, respectively. The extracted parameters for both oscillations are summarized in Supplementary Table 2.

Fig. 2
figure 2

Shubnikov–de-Haas (SdH) oscillations and giant positive magnetoresistivity. a The angular dependence of resistivity ρxx at 0.12 K. The SdH oscillation for 2–9 T is clear for B||a, while that for 1–3 T is clear even for B||c. The giant positive magnetoresistivity is observed above 9 T nearby B||a. Inset shows the Landau index plot of the oscillation for B||a and B||c and illustration of measurement geometry for B and electrical current I (||a). The tilting angle θ is defined as zero for B||a. b, c Oscillatory component at various temperatures for B||a and for B||c. d, e Illustration of band dispersion nearby the line node described in kb–kc plane and ka–kb plane. The red and blue line denotes the cross-section of inner- and outer-Fermi surface (FS), respectively. The SdH oscillation of outer-FS is not visible for B||c, which is indicated by dashed lines

Full size image

Calculations based on density functional theory and dynamical mean field theory

To understand the observed transport properties, we carried out the calculation using the density functional theory combined with the dynamical mean field theory (DMFT) to calculate a realistic electronic structure subject to the Coulomb interaction Ueff and relativistic spin–orbit interaction (see Methods). The band structure with Ueff = 0 eV is displayed in Fig. 3a. The bands lying from −0.4 to 0.5 eV dominantly consist of jeff = 1/2 state. Highly dispersing bands along X-U-Z line and R-T-Y line are crossing EF, yielding electron pockets. Specifically, the Dirac-like linearly dispersing bands are seen nearby the U-point (ka = 0, kb = π, kc = π); the band crossing constitutes a closed loop encircling the U-point on the kb = π plane (Fig. 3c) at −0.2 eV. Interestingly, the two bands are dispersing in nearly parallel in a wide energy regime, yielding two FS neighboring each other. On the other hand, the bands along R-T-Y, Z-Γ-Y, and Y-S-R lines show the parabolic dispersion. Figure 3b shows the band structure with Ueff = 2.0 eV. The bandwidth of jeff = 1/2 state is renormalized into the energy regime from −0.2 to 0.2 eV while keeping the presence of line node. Notably, both the electron and hole pockets shrink compared with the case of Ueff = 0 eV with keeping the charge neutrality. Indeed, bands around Y-S-R line are pushed down and the hole pocket around S-point vanishes. Figure 3d–g shows the magnified view nearby the U-point. With increasing Ueff, the bandwidth is significantly renormalized and vF is reduced from 2.0 × 105 m s−1 at Ueff = 0 eV to 8.0 × 104 m s−1 at Ueff = 2.0 eV (Supplementary Figure 8). Consequently, the energy of the line node relative to EF is squeezed from −0.2 eV at Ueff = 0 to −0.03 eV at Ueff = 2.0 eV. This large renormalization is a precursory phenomenon of the Mott transition. The systematic evolution of band structure upon increasing Ueff suggests the remarkable scenario that the Fermi energy is most closely tuned to the Dirac node at the Mott criticality.

Fig. 3
figure 3

Band structure calculated by density functional theory and dynamical mean field theory. a, b Overview of band structure with the Hubbard Ueff = 0 and 2.0 eV. The scale bar denotes the magnitude of spectral function. c Illustration of momentum space and position of Dirac line node (blue line). d–g The magnified view of band structure around the Dirac line node with different Ueff. The Dirac line node approaches the Fermi energy with increasing Ueff. The Dirac-like dispersion is significantly renormalized for Ueff = 2.5 eV, which is a precursory phenomenon of the Mott criticality. The averaged Fermi velocity is 2.0 × 105, 1.8 × 105, 8.1 × 104, and 2.0 × 104m s−1 at Ueff = 0, 1.0, 2.0, and 2.5 eV, respectively

Full size image

Fermi surface

Among the electron pockets around U-point and those around T-point, we conclude that the formers are the plausible candidates of the highly mobile electron by following reasons. First, the observed Fermi velocity is consistent with the calculated band structure. Second, the angular dependence of SF is consistent with that expected from the electron pockets around U-point (Supplementary Note 3). In this context, it is conceivable that the low (high)-field oscillation is ascribed to the inner-FS (outer-FS) inherent to the Dirac-like dispersion as illustrated in Fig. 2d, e. Assuming that the Fermi velocities of inner- and outer-bands are equivalent and that their anisotropy can be neglected, the diameter of line node is estimated to be about 0.008 Å−1, which is as small as about 1/70 of the reciprocal lattice units π/a (Supplementary Note 3). Moreover, the energy of line node is estimated to be 5–8 meV below EF. Such a nearly perfect tuning of EF to the line node and the good agreement of vF between the theory (Ueff = 2.0 eV) and experiment verify the aforementioned scenario.

Giant magnetoresistivity

Having established these results, we discuss the origin of giant magnetoresistivity observed above 9 T. Figure 4a shows the θ-dependence of resistivity at various magnetic fields. As θ increases, the resistivity at 14 T rapidly decreases; the resistivity at θ = 20° is <1/10 of that at θ = 0°. On the contrary, the resistivity at lower fields moderately increases as a function of θ. Figure 4b shows the temperature dependence of resistivity at θ = 0° (B||a). With decreasing temperature, the resistivity increases at 14 T, but shows a metallic behavior below 8 T. The close similarity between the angular and temperature dependence of giant magnetoresistivity and those of high-frequency oscillation (Supplementary Figure 3 and Figure 5) suggests that the highly resistive state is inherent to the quantum limit of the outer-FS. Under sufficiently high magnetic fields in the quantum limit, the magnetic length \(l_{B} = 1{\mathrm{/}}\sqrt {eB}\) is reduced to the scale of Fermi wavelength, leading to the quasi-one-dimensional (1D) state dispersing along the field direction. The important consequence of quasi-1D confinement of Dirac electron is the gap opening or electron localization due to the strong enhancement of electron correlation or disorder effect as well as the inter-node mixing effect22,23,24,25. Indeed, the temperature dependence of resistivity at B(||a)=14 T can be described by the variable range hopping model with electron correlation or localized Tomonaga–Luttinger liquid model (Supplementary Figure 9). Thus, it is likely that the giant magnetoresistivity is ascribed to the gap-out of line node or the electron localization promoted by the magnetically induced quasi-1D confinement, which can be realized at the modest field as enabled by the nearly perfectly tuned line node. Such a quantum transport in the quantum limit has been rarely realized in existing Dirac/Weyl semimetals, which may characterize the emergent topological property of this correlated Dirac semimetal.

Fig. 4
figure 4

Angular and temperature dependence of giant magnetoresistivity. a The resistivity at various magnetic fields vs. θ at 0.12 K. b Temperature dependence of resistivity for B||a

Full size image

Discussions

We demonstrated that the electron correlation cooperative with the relativistic spin–orbit coupling yields unusually highly mobile electrons exceeding 60,000cm2 V−1 s−1 in the Dirac semimetal perovskite CaIrO3. From the analysis of SdH oscillation, the Fermi energy is nearly pinned at the Dirac line node (within ~8 meV) and the magnetic quantum limit is reached at the modest magnetic field of about 9 T. In the quantum limit, we identified a field-direction-sensitive giant positive magnetoresistivity with a ratio of about 5500% at 0.12 K and 14 T, suggesting the mass gap opening or the electron localization promoted by the quasi-1D confinement. The manifestation of highly mobile Dirac electrons in strongly correlated materials opens a new stage of research on quantum phenomena in topological materials.

Methods

Sample preparation and characterization

Single crystals of perovskite CaIrO3 were grown by solid-state reaction using the cubic-anvil-type facility. The materials were heated up to 1200 °C under 1 GPa, remained there for 10 min, and then quenched to room temperature. The high-pressure synthesis can provide the high-quality samples with the right stoichiometry (i.e., half band-filling). This is crucial to pin the Fermi energy nearby the Dirac line node. The typical size of crystal is about 0.5 × 0.5 × 0.3 mm3 as shown in the inset to Fig. 1b. The crystal orientation was determined using the X-ray diffractometer.

The crystallographic symmetry is orthorhombic Pbnm, in agreement with the previous report according to powder X-ray diffraction26. We analyzed the crystal structure on the basis of the collected data sets of single-crystal X-ray diffraction. Supplementary Figure 1 shows the typical X-ray diffraction pattern taken at 300 K. In total, the 14,436 reflections (799 independent reflections) up to sin θ/λ = 0.92 Å−1 were observed. No superlattice reflections which violate the Pbnm symmetry are discernible within the experimental accuracy. The derived structural parameters are listed in Supplementary Table 1.

Measurement of resistivity and Hall resistivity

Measurements of resistivity and Hall resistivity were performed by standard four-terminal geometry with indium electrode. The measurements above 2 K were done by using the Physical Property Measurement System (Quantum Design). The measurement below 1 K was performed by using the dilution refrigerator equipped with the 14 T superconducting magnet (Oxford instruments). We employed a standard lock-in technique with a fixed excitation current (1–10μA) at low frequency (1–10 Hz).

Electronic structure calculation

We used the Perdew–Burke–Ernzerhof exchange-correlation functional27 as implemented in the WIEN2K program28. The muffin tin radii (RMT) of 2.11, 2.00, and 1.77 bohr were used for Ca, Ir, and O, respectively. The maximum modulus for the reciprocal vectors Kmax was chosen such that RMT Kmax = 7.0 and a 10 × 10 × 10 k-mesh in the first Brillouin zone was used. We then constructed the Wannier functions for the t2g bands, using the WIEN2WANNIER29 and the WANNIER9030 codes.

With these Wannier functions, we constructed a tight-binding model with three t2g Wannier orbitals at each of four equivalent Ir sites in the unit cell. We considered the Hubbard-type onsite Coulomb interaction Ueff on the basis diagonalizing the local Hamiltonian, where Ueff is an orbital independent density–density-type interaction.

The correlation effects due to Ueff were taken in within the DMFT31, which counts the full local correlations while neglecting the nonlocal correlations, which are expected to be small in the present three-dimensional material. The impurity problem in the DMFT was solved with an exact diagonalization method, where the DMF is expressed by a finite number of bath sites. We used nine bath sites for this. Applying the Lanczos method to the finite-size Hamiltonian, we calculated the ground state and the self-energy. For real-frequency axis, we introduced an energy-smearing factor of 5 meV. This self-energy on real-frequency axis was used to calculate the spectral function plotted in Fig. 3. The calculated results show that the system persists to be metallic up to a relatively large value of Ueff. This is presumably because the semimetallic density of states of this material will not have a much benefit of the energy reduction due to the opening of the Mott gap.