Crystalline Electric Field as a Probe for Long-Range Antiferromagnetic Order and Superconducting State of ${\mathrm{CeFeAsO}}_{1-x}{\mathrm{F}}_x$

We use inelastic neutron scattering to study the crystalline electric field (CEF) excitations of ${\mathrm{Ce}}^{3+}$ in ${\mathrm{CeFeAsO}}_{1-x}{\mathrm{F}}_x$ ($x=0$, 0.16). For nonsuperconducting CeFeAsO, the Ce CEF levels have three magnetic doublets in the paramagnetic state, but these doublets split into six singlets when the Fe ions order antiferromagnetically. For superconducting ${\mathrm{CeFeAsO}}_{0.84}{\mathrm{F}}_{0.16}$ ($T_c=41\mathrm{K}$), where the static antiferromagnetic order is suppressed, the Ce CEF levels have three magnetic doublets at $\hslash \omega =0$, 18.7, 58.4 meV at all temperatures. Careful measurements of the intrinsic linewidth $\Gamma$ and the peak position of the 18.7 meV mode reveal a clear anomaly at $T_c$, consistent with a strong enhancement of local magnetic susceptibility ${\chi }^{\prime \prime }(\hslash \omega )$ below $T_c$. These results suggest that CEF excitations in the rare-earth oxypnictides can be used as a probe of spin dynamics in the nearby FeAs planes.

Figure 1

Summary of the CeFeAsO crystal or magnetic structure and CEF levels determined from our inelastic neutron scattering experiments. (a) The Fe spin ordering in the CeFeAsO chemical unit cell. (b) The Fe spins in CeFeAsO with respect to the Ce positions. The Fe moments lie in the $a\mathrm{\text{−}}b$ plane along the $a$ axis and form an antiferromagnetic collinear spin structure. (c) ${\mathrm{Ce}}^{3+}$ CEF levels in CeFeAsO for temperatures above and below the Fe AF Néel temperature of $T_N=140\mathrm{K}$ [6]. The arrows indicate possible transitions. (d) ${\mathrm{Ce}}^{3+}$ CEF levels in superconducting ${\mathrm{CeFeAsO}}_{0.84}{\mathrm{F}}_{0.16}$ at low temperature.

Figure 2

Temperature dependence of the CEF excitations in CeFeAsO and ${\mathrm{CeFeAsO}}_{0.84}{\mathrm{F}}_{0.16}$ integrated over $0<Q<4{\AA }^{-1}$ and our model determination of the CEF levels. (a) Ce CEF magnetic excitations in absolute units after subtracting the LaFeAsO phonons. The squares to the left denote the data with $E_i=30.8\mathrm{meV}$ to show the splitting. The solid line is our model fit with parameters listed in Table . The inset shows the raw DCS data integrated over $0<Q<2.2{\AA }^{-1}$ in arbitrary units as a function of temperature without the LaFeAsO phonon subtraction. (b) Ce CEF excitations at 200 K; solid line is our model calculation. (c–d) CEF excitations and their temperature dependence for ${\mathrm{CeFeAsO}}_{0.84}{\mathrm{F}}_{0.16}$. The solid lines are model calculations. Inset in (c) shows that the 0.7 meV excitation seen in CeFeAsO is missing in ${\mathrm{CeFeAsO}}_{0.84}{\mathrm{F}}_{0.16}$. Error bars represent 1 standard deviation.

Figure 3

(a) Raw $S(Q,\omega )$ spectra of CeFeAsO at 60 K and $E_i=30\mathrm{meV}$ on MERLIN. (b) Ce CEF excitations after subtraction of the LaFeAsO background. (c) Temperature dependence of the 18.7 meV excitations. The peaks around 18 meV at 7 K are fitted with 2 Lorentzians, the widths of which were fixed for fittings at all higher temperatures. Although this ignores the broadening of linewidths upon warming and gives illusionary finite peak separation above $T_N$, the net splitting solely caused by molecular field can be monitored by abrupt change of this separation which happens below 140 K, the $T_N$ of Fe as shown in (e).

Figure 4

Temperature dependence of the $\hslash {\omega }_1=18.7\mathrm{meV}$ Ce CEF excitation for ${\mathrm{CeFeAsO}}_{0.84}{\mathrm{F}}_{0.16}$. (a) MERLIN measurements using $E_i=30.8\mathrm{meV}$ at different temperatures. The instrumental energy resolution is 2.1 meV at elastic position (horizontal bar). (b) The peak position as a function of temperature for the 18.7 meV CEF level. The solid squares are MERLIN data while light squares are FANS data; both show a clear anomaly at $T_c$. (c) The intrinsic linewidth $\Gamma (T)$ as a function of temperature. The solid line shows the expected ${\Gamma }_n(T)$ assuming noninteracting Fermi liquid. $\Gamma (T)$ deviates from ${\Gamma }_n(T)$ near $T_c$. (d) $\Gamma (T)/{\Gamma }_n(T)$ shows a clear anomaly near $T_c$, consistent with (c).