A new magnetic phase in the nickelate perovskite TlNiO$_3$

TL;DR

This paper reports the discovery of a previously unknown magnetic phase in TlNiO₃ that exists above its Néel temperature, observed through NMR and muon spin rotation, and is influenced by external magnetic fields.

Contribution

It provides experimental evidence for a second magnetic phase in TlNiO₃ above the known Néel temperature, a novel finding in nickelate perovskites.

Findings

Existence of a second magnetic phase in TlNiO₃ above T_N.

The new phase persists up to T_N* = 202 K.

The phase is suppressed by a 1 T magnetic field.

Abstract

The RNiO$_3$ perovskites are known to order antiferromagnetically below a material-dependent N\'eel temperature $T_\text{N}$. We report experimental evidence indicating the existence of a second magnetically-ordered phase in TlNiO$_3$ above $T_\text{N} = 104$ K, obtained using nuclear magnetic resonance and muon spin rotation spectroscopy. The new phase, which persists up to a temperature $T_\text{N}^* = 202$ K, is suppressed by the application of an external magnetic field of approximately 1 T. It is not yet known if such a phase also exists in other perovskite nickelates.

Equations

$${\nu }_{\text{edge}}=\frac{\gamma }{2\pi }(B_0+B_{\text{int}}),$$

$$\phantom{,}S_{i}(t)=N_{0i}\,\big{[}1+\operatorname{Re}\,A(t)\,\mathrm{e}^{\mathrm{i}\phi_{i}}\big{]}\,\mathrm{e}^{-\frac{t}{\tau_{\mu}}}+B_{i}\quad,$$

$$A(t)=A\,\Big{[}\tfrac{2}{3}\cos(2\pi\nu^{(1)}t)\mathrm{e}^{-\lambda_{\text{x}}^{(1)}t}+\tfrac{1}{3}\mathrm{e}^{-\lambda_{\text{z}}t}\Big{]}+B\,\mathrm{e}^{-\lambda_{\text{DC}}t}+C$$

$$A(t)=a^{(0)}{\mathrm{e}}^{-{\lambda }_{\text{z}}t}+\sum _{i=1}^4{a}^{(i)}\cos (2\pi {\nu }^{(i)}t){\mathrm{e}}^{-{\lambda }_{\text{x}}^{(i)}t}+C,$$

Full text

Although first synthesized already in 1970 Demazeau et al. (1971), the rare-earth nickelates RNiO3 have been the subject of intense research efforts during the last decade due to their peculiar metal–insulator transition at $T_{\text{MI}}$ and subsequent antiferromagnetic (AFM) order between Ni spins below a Néel temperature $T_{\text{N}}$ Medarde (1997); Catalan (2008). By substituting rare-earth ions R3+ with different radii, the perovskite lattice can be distorted continuously. This distortion affects the magnetic couplings of Ni spins by modifying the Ni–O–Ni superexchange angle. A phase diagram of RNiO3, mostly based on previously published results, is shown in Fig. 1. Despite recent theoretical Mazin et al. (2007); Park et al. (2012); Subedi et al. (2015) and experimental Ruppen et al. (2015) progress, the nature of the paramagnetic (PM) insulating phase is still unclear. The AFM order below $T_{\text{N}}$ is characterized by a propagation vector $k_{\text{pc}}=(\nicefrac{{1}}{{4}},\nicefrac{{1}}{{4}},\nicefrac{{1}}{{4}})$ with respect to the (pseudo-)cubic unit cell of the ideal perovskite structure. Thus, a period of the magnetic structure comprises four Ni sites along each pseudocubic crystal axis. Previous experimental reports suggested a collinear “up-up-down-down” structure García-Muñoz et al. (1992, 1994); Alonso et al. (1999a), while others argued for a non-collinear spiral spin configuration Fernández-Díaz et al. (2001); Scagnoli et al. (2006, 2008); Bodenthin et al. (2011). The AFM phase is predicted to be ferroelectric Cheong and Mostovoy (2007); Giovannetti et al. (2009), but this has not yet been confirmed experimentally. Since the magnitude of the ferroelectric polarization is very different for the two candidate spin arrangements Giovannetti et al. (2009), investigations of the magnetism of RNiO3 are crucial for the understanding of their possible multiferroicity. Nuclear magnetic resonance (NMR) is a powerful technique to locally probe magnetic behavior. However, to our knowledge, the only previous NMR work involving a nickelate perovskite was a 139La-NMR study of LaNiO3 Sakai et al. (2002), which is a PM metal at all temperatures.

In the present work, we present the first NMR investigation of an insulating member of the RNiO3 family, in the form of a combined NMR and muon-spin rotation ($\mathrm{\mu}$SR) study of the magnetic properties of TlNiO3. This material is known to have the same qualitative behavior as the analogous rare-earth nickelates Kim et al. (2002). From an NMR perspective, TlNiO3 is unique among the RNiO3 compounds because both 203Tl and 205Tl nuclei are excellent for NMR. In addition to the previously known AFM phase below $T_{\text{N}}=104$ K, our measurements reveal a previously unknown magnetically-ordered phase between $T_{\text{N}}$ and $T_{\text{N}}^{*}=202$ K. An applied magnetic fields of 1 T is sufficient to suppress the static magnetic order between $T_{\text{N}}$ and $T_{\text{N}}^{*}$. Nevertheless, hysteretic dynamics in this temperature range are still detected by NMR, even in applied fields of several teslas (see below).

*Experimental details. *The synthesis of polycrystalline TlNiO3 is described in Ref. Kim et al. (2001). We confirmed the 99% purity of our powder sample by X-ray diffraction. DC magnetometry data taken at various applied fields display a kink at $T_{\text{N}}=104$ K, consistent with the previously known AFM transition (see Fig. 2).

At low fields, a small ferromagnetic contribution with a Curie temperature $T_{\text{C}}=25$ K is observed, most likely due to the presence of residual Ni(OH)2 from the synthesis Kim et al. (2001). (Note that a similar feature can be seen in the susceptibility data shown in Refs. Kim et al. (2001, 2002) as well.) Because of the large magnetic moment of this ferromagnet, an impurity concentration of 1% by mass is sufficient to explain the substantial magnetic response in our data. Due to the small mass fraction, this impurity is irrelevant for $\mathrm{\mu}$SR. Moreover, since this impurity does not contain thallium, it does not affect our 203Tl- and 205Tl-NMR measurements, either.

Nuclear magnetic resonance. **203Tl- and 205Tl-NMR spectra were acquired using a standard spin-echo pulse sequence. Because of the very large linewidth, the NMR signal in the AFM phase was acquired by sweeping the frequency and integrating the Fourier-transformed spin-echo. The transverse relaxation time $T_{2}$ was measured by varying the delay $\tau$ between radio-frequency pulses in the spin-echo sequence and fitting the resulting NMR amplitudes to an exponential decay $M(\tau)=M_{0}\exp(-\nicefrac{{2\tau}}{{T_{2}}})$. Due to the rapid relaxation and decoherence rates in the PM phase ($T_{1}\approx 20$ $\mu$s, $T_{2}\approx 16$ $\mu$s), the spin-echo intensities are low and substantial signal-averaging is required to improve the signal-to-noise ratio of the NMR measurements. In addition to standard NMR experiments in an applied field, we performed complementary zero-field NMR investigations in the magnetically-ordered phase.

Two stable thallium isotopes, 203Tl and 205Tl, occur naturally with abundances of 29.5% and 70.5%. Both nuclei have spin $\nicefrac{{1}}{{2}}$ and gyromagnetic ratios of $\nicefrac{{\gamma_{203}}}{{2\pi}}=24.3216$ MHz/T and $\nicefrac{{\gamma_{205}}}{{2\pi}}=24.5603$ MHz/T, respectively. Due to the high abundances and gyromagnetic ratios, both nuclei are well suited for magnetic resonance measurements. In TlNiO3, the resonance signals of both 203Tl and 205Tl were detected at a frequency shift of $K=1.03$% at room temperature.

Due to the distribution of orientations of the intrinsic magnetic field, the NMR powder pattern broadens by two orders of magnitude in the AFM phase. Two cusps, marked by vertical dashed lines in Fig. 3, are seen in the NMR spectrum, which correspond to the broadened upper edges of antiferromagnetic NMR powder patterns from two magnetically-inequivalent Tl-sites. The edges originate from crystallites whose internal magnetic field is aligned parallel to the applied field Yamada and Sakata (1986). They occur at frequencies

[TABLE]

where $B_{0}$ is the applied magnetic field, and $B_{\text{int}}$ is the local internal field. This internal field is static on the typical timescale of an NMR experiment, which is at least $10^{-4}$ s. Since the relative difference of the gyromagnetic ratios $\nicefrac{{\gamma_{205}-\gamma_{203}}}{{\gamma_{205}}}\approx 1$% is much smaller than the relative width of the powder pattern, the contributions from 203Tl and 205Tl cannot be distinguished in the magnetically-ordered phase. Hence, when mentioning Tl-NMR spectra in the AFM phase, we refer to the superposition of 203Tl- and 205Tl-NMR spectra.

By applying Eq. 1 to the NMR spectrum acquired at 50 K and $B_{0}=2$ T, shown in Fig. 3, we extract the two internal magnetic fields $B_{\text{int}}\approx 0.6$ T and 1.0 T. The zero-field (ZF) NMR frequency $\nu_{\text{{ZF}}}=\nicefrac{{\gamma}}{{2\pi}}B_{\text{int}}$ corresponds to $B_{0}=0$ in Eq. 1. Since $\nu_{\text{ZF}}$ is proportional to the local magnetic field, its temperature-dependence reflects the order parameter of a magnetically-ordered phase.

By following the temperature-dependence of the two ZF-NMR frequencies, as plotted in Fig. 4, we identify two magnetic phases — the previously reported phase below $T_{\text{N}}=104$ K Kim et al. (2001, 2002) and a new phase which persists up to $T_{\text{N}}^{*}\sim 200$ K. We were unable to establish the exact transition temperature from ZF-NMR alone, because the NMR signal-to-noise ratio deteriorates significantly at low resonance frequencies, close to the transition.

Measurements performed in $B_{0}=3.5$ T did not reveal any static magnetic order above $T_{\text{N}}$. However, as shown in Fig. 5, the dephasing rate of the nuclear spins $T_{2}^{-1}$ shows a significant hysteresis between $\sim 125$ K and 200 K. This almost coincides with temperature range where we find the new magnetic phase in zero field, between $T_{\text{N}}$ and $T_{\text{N}}^{*}$. Below 100 K, both the dephasing rate $T_{2}^{-1}$ and the relaxation rate $T_{1}^{-1}$ (not shown) follow an Arrhenius law $T_{i}^{-1}=\tau_{i}\,\exp(-\nicefrac{{\Delta}}{{k_{\text{B}}T}})+C_{i}$ with an energy gap $\nicefrac{{\Delta}}{{k_{\text{B}}}}=82\pm 5$ K for the magnetic excitations.

*Muon-spin rotation. *Muon-spin rotation ($\mu$SR) experiments were performed in TlNiO3 powder in zero applied field between 5 K and ambient temperature using the General Purpose Spectrometer (GPS) at the Swiss Muon Source (S$\mu$S) at the Paul Scherrer Institut. In analogy to ZF-NMR, the zero-field $\mathrm{\mu}$SR frequencies reflect the local internal magnetic field (at the muon stopping site), hence providing a measure of the order parameter in the magnetically-ordered phase. Unlike NMR, $\mu$SR can detect low-frequency signals without loss of amplitude, as can be seen in Fig. 6. This allows for the determination of the new phase transition temperature $T_{\text{N}}^{*}=202$ K in zero field.

The time-domain histogram of detector $i$, $S_{i}(t)$ is given by

[TABLE]

where $A(t)$ is the asymmetry, $N_{0i}$ are normalized signal count rates, $\phi_{i}$ are phase offsets, $B_{i}$ are background count rates, and $\tau_{\mu}=2.2$ $\mu$s is the muon lifetime Yaouanc and Dalmas de Réotier (2011).

First, the $\phi_{i}$ are obtained from a calibration measurement in the PM phase in a transversely applied field of 3 mT. Second, the $N_{0i}$ and $B_{i}$ are fitted for each dataset using Eq. 2 and assuming $A(t)$ to be constant, which is a good approximation at long times $t\gtrsim 0.3$ $\mu$s. Then, the parameters of $A(t)$ are estimated by means of a global maximum-likelihood fit of the detector histograms. The Python Python Software Foundation package iminuit imi is used to perform the fits and determine the standard errors using the Minos algorithm James and Roos (1975).

Above $T_{\text{N}}$, the expression

[TABLE]

is used to fit the $\mu$SR asymmetry Yaouanc and Dalmas de Réotier (2011); Dalmas de Réotier and Yaouanc (1997). This expression accounts for two muon stopping sites: one where the mean local magnetic field is non-zero and one where it is zero; at both sites exponential relaxation occurs by fluctuations of the magnetic field. Exponential relaxation was also found in previous $\mu$SR experiments on other RNiO3 perovskites García-Muñoz et al. (1995, 2006).

Below $T_{\text{N}}$, the asymmetry is modeled using

[TABLE]

which describes the $\mu$SR signal from four magnetically-inequivalent muon sites experiencing non-zero local fields and exponential relaxation Yaouanc and Dalmas de Réotier (2011); Dalmas de Réotier and Yaouanc (1997). A comparison between the experimental asymmetry and our fits is shown in Fig. 6. Since the $\mu$SR signal dephases within the first 0.2 $\mu$s, the Fourier-transformed spectrum is strongly broadened. Thus it is impossible to resolve oscillations whose frequencies differ by less than a few MHz. This is the reason why we could establish only two frequencies at 100 K, and only three frequencies at 5 K and 15 K. In the corresponding fits, a rapidly relaxing DC component was added phenomenologically to improve the fit convergence.

The temperature-dependence of the frequencies $\nu^{(i)}$ plotted in Fig. 7 clearly shows the two phase transitions at $T_{\text{N}}=104$ K and $T_{\text{N}}^{*}=202$ K. Additionally, one can see that there are four magnetically inequivalent muon sites below $T_{\text{N}}$, giving rise to four different oscillation frequencies. The physical implication of these four muon sites is an open question. Since there are three inequivalent O2--sites in TlNiO3 Kim et al. (2002), a tentative explanation of the three frequencies appearing below $T_{\text{N}}$ is that they originate from muons bound to the three different oxide ions. We have not been able to explain the approximately linear temperature-dependence of the lowest $\mu$SR frequency. However, this may be related to the occurence of a magnetically ordered phase embedded within another ordered phase.

*Conclusion. *Our NMR and $\mu$SR data reveal the presence of a not-yet-reported magnetically-ordered phase in TlNiO3 between $T_{\text{N}}=104$ K and $T_{\text{N}}^{*}=202$ K in low magnetic fields and confirm the previously-known AFM phase below $T_{\text{N}}$. Both phases clearly show static magnetic order on the timescales of NMR and $\mu$SR. Due to the strong broadening of both the NMR and $\mu$SR spectra, a distinction between short-range and long-range order is not possible,. Future $\mu$SR experiments on TlNiO3 are intended to map the phase boundary of the newly reported magnetic phase as a function of applied magnetic field and temperature, and to explore the significance of the four $\mu$SR frequencies identified below $T_{\text{N}}$. Further measurements on other compounds will show whether this new phase is unique to TlNiO3 or a universal feature of the RNiO3 family.