This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Statistical and Analytical Approaches
to Finite Temperature Magnetic Properties of SmFe12 compound

Takuya Yoshioka Department of Applied Physics, Tohoku University, Sendai 980-8579, Japan ESICMM, National Institute for Materials Science, Tsukuba, Ibaraki 305-0047, Japan Center for Spintronics Research Network, Tohoku University, Sendai 980-8577, Japan    Hiroki Tsuchiura Department of Applied Physics, Tohoku University, Sendai 980-8579, Japan ESICMM, National Institute for Materials Science, Tsukuba, Ibaraki 305-0047, Japan Center for Spintronics Research Network, Tohoku University, Sendai 980-8577, Japan    Pavel Novák Institute of Physics of ASCR, Cukrovarnická, Prague 6 162 00, Czech Republic
Abstract

To investigate the magnetic properties of SmFe12, we construct an effective spin model, where magnetic moments, crystal field (CF) parameters, and exchange fields at 0 K are determined by first principles. Finite temperature magnetic properties are investigated by using this model. We further develop an analytical method with strong mixing of states with different quantum number of angular momentum JJ (JJ-mixing), which is caused by strong exchange field acting on spin component of 4f4f electrons. Comparing our analytical results with those calculated by Boltzmann statistics, we clarify that the previous analytical studies for Sm transition metal compounds over-estimate the JJ-mixing effects. The present method enables us to make quantitative analysis of temperature dependence of magnetic anisotropy (MA) with high-reliability. The analytical method with model approximations reveals that the JJ-mixing caused by exchange field increases spin angular momentum, which enhances the absolute value of orbital angular momentum and MA constants via spin-orbit interaction. It is also clarified that these JJ-mixing effects remain even above room temperature. Magnetization of SmFe12 shows peculiar field dependence known as first-order magnetization process (FOMP), where the magnetization shows an abrupt change at certain magnetic field. The result of the analysis shows that the origin of FOMP is attributed to competitive MA constants between positive K1K_{1} and negative K2K_{2}. The sign of K1(2)K_{1(2)} appears due to an increase in CF potential denoted by the parameter A20r2A_{2}^{0}\langle r^{2}\rangle (A40r4A_{4}^{0}\langle r^{4}\rangle) caused by hybridization between 3d3d-electrons of Fe on 8i8i (8j8j) site and 5d5d and 6p6p valence electrons on Sm site. It is verified that the requirement for the appearance of FOMP is given as K2<K1<6K2-K_{2}<K_{1}<-6K_{2}.

pacs:
Valid PACS appear here

I Introduction

There have been intensive studies on developing new rare-earth (RR) lean permanent magnetic materials which have strong magnetic properties comparable to those of Nd-Fe-B. Nitrogenated compounds as NdFe12N or NdFe11TiN have been considered to be candidates of such materials, and thus series of experimental and theoretical efforts have been made to figure out the magnetic properties of these materials Miyake ; Hirayama1 . SmFe12 with the ThMn12 structure (Fig. 1) is also a possible candidate and has attracted renewed interest because it exhibits excellent intrinsic magnetic properties such as uniaxial magnetocrystalline anisotropy Hadjipanayis . Although SmFe12 itself is thermodynamically unstable, it has been known that the substitution of Fe with a stabilizing element, such as Ti or V, can remove this difficulty Ohashi1 ; Ohashi2 ; Hu ; Kuno ; Schoenhoebel . In these systems, however, the saturated magnetization is reduced due to anti-parallel alignment of magnetic moments of Ti and V relative to those of Fe. Recent development of the synthesis technology made it possible to fabricate highly textured single phase samples of SmFe12 thin film Kato ; Hirayama2 ; Ogawa1 ; Ogawa2 ; Sepehri-Amin3 , and it has been shown experimentally that Co substitution for Fe enhances their magnetic properties, such as Curie temperature and magnetic anisotropy (MA) Hirayama2 . Thus, SmFe12-based systems belong to one of the most promising hard magnetic materials, and therefore to clarify the basic magnetic properties of SmFe12 is crucially important.

Refer to caption
Figure 1: (color online) Crystal structure of SmFe12 compound in ThMn12 structure. Inequivalent sites: Sm(2aa), Fe(8ff), Fe(8ii), and Fe(8jj) are shown by different-colored balls and solid lines show interatomic short contacts less than 3.2 Å.

So far many attempts have been performed for microscopic understanding of the magnetic properties of RR based permanent magnets Evans ; Toga ; Nishino ; MiyakeAkai ; Yoshioka ; Tsuchiura2 . Among them a powerful method is to combine the first-principle calculations for electronic states at the ground state with a suitable model for finite temperature properties Yamada ; Sasaki ; Miura-direct ; Miura-pl ; Kuzminlinear ; Kuzmin2nd ; Kuzminlimit ; Millev ; Kuzminmix ; Magnani . As for SmFe12, Harashima et al. (2015) Harashima , Körner et al (2016) Koerner and Delonge et al (2017) Delange performed the first-principle calculations and model analysis of magnetic properties. In the theoretical study of Sm-based intermetallic compounds, however, there remains a basic issue how to deal with the formidably strong JJ-mixing effects in Sm. This is the problem studied for a long period on the Sm-based magnets VanVleck ; Sankar ; Wijin . There are some attempts to include the JJ-mixing in the analytical form by the first-order perturbation for the crystal fields (CFs) Kuzminmix ; Magnani . However, Kuz’min pointed out that the Sm-based magnetic materials are exceptional for application of the method Kuzminmix .

We have recently developed a similar method Yoshioka ; Tsuchiura2 , in which the model parameters are calculated by the first-principles and the finite temperature magnetic properties are calculated in a statistical way, and applied it to R2R_{2}Fe14B systems. By taking into account CF parameters up to 6-th order, the model satisfactorily explained the experimental results for magnetization curves and the temperature dependence of MA constants Yoshioka ; Tsuchiura2 . Using the method we recently calculated the temperature dependence of the MA constants of SmFe12 and showed that K1>0K_{1}>0 and K2<0K_{2}<0 in consistence with experimental results Ogawa2 . The report of the work, however, contains only the final results and no details of computational procedure have been presented. As a results no explanations on the mechanism for the results that K1>0K_{1}>0 and K2<0K_{2}<0 have been given.

The purpose of the present study is thus to clarify the origin of the finite temperature magnetic properties of SmFe12 compound by statistical and analytical ways. To this end, we describe the details of the statistical method and develop a novel analytical method. The analytical procedure is able to derive simple relations between the temperature dependence of magnetic properties and parameters determined by first-principles electronic structure calculations. The treatment of the JJ-mixing effects adopted previously by the other groups Kuzminmix ; Magnani will be modified, and the results will be compared with the statistical results of the temperature dependence of magnetic properties of SmFe12. Good agreement between the analytical and statistical results guarantees the applicability of the modified analytical formula to Sm compounds.

In the following, we give the model Hamiltonian, the parameters of which are determined by the first-principles, and present the calculation procedure for finite temperatures, especially the statistical method to obtain the MA constants and magnetization curves, and explain the modified analytical method. The latter method may clarify the relations among the free energy of the system, the CF, and the exchange field. Using the analytical method, we will show that the mechanism of K1>0K_{1}>0 and K2<0K_{2}<0 in SmFe12 is attributed to the characteristic lattice structure around Sm ions, that is, crystallographic 2bb-sites on cc-axis adjacent to Sm are vacant. We also present results on the magnetization process and nucleation fields by calculating Gibbs free energy. As pointed out in Ref. Tsuchiura2 , this analytical spin model can be easily extended to Sm ions around the intergranular phases, which is crucially important in the coercivity mechanism Sepehri-Amin1 ; Sepehri-Amin2 ; Sepehri-Amin3 .

This paper is organized as follows. The model Hamiltonian is explained in Section II, and the procedure of the statistical and analytical method are explained in Section III. Section IV shows the results of temperature dependence of magnetic properties calculated in the statistical and analytical methods. A summary of our work is given in Section V.

II Model Hamiltonian

We adopt a following Hamiltonian to investigate the magnetic properties of RR transition metal (TMTM) compounds:

^=\displaystyle\hat{\cal H}= 1V0j=1nR^R,j\displaystyle\frac{1}{V_{0}}\sum_{j=1}^{n_{R}}\hat{\cal H}_{R,j}
+K1TM(T)sin2θTM𝑴TM(T)𝑩,\displaystyle+K_{1}^{TM}(T)\sin^{2}\theta^{TM}-{\bm{M}}^{TM}(T)\cdot{\bm{B}}, (1)

where ^R,j\hat{\cal H}_{R,j} is a Hamiltonian for RR ion on jj-th site and nRn_{R} is the number of RR ion in the unit cell volume V0V_{0}. Second and third term represent the phenomenological treatment of MA energy and Zeeman term on TMTM sublattice, where K1TM(T)K_{1}^{TM}(T) and 𝑴TM(T){\bm{M}}^{TM}(T) are the temperature dependent anisotropy constant and magnetization vector of TMTM sublattice, respectively, and θTM\theta^{TM} is the polar angle of 𝑴TM(T){\bm{M}}^{TM}(T) against the cc-axis. 𝑴TM(T){\bm{M}}^{TM}(T) is given as MTM(T)𝒆TMM^{TM}(T){\bm{e}}^{TM} by using the absolute value of the sublattice magnetization MTM(T)M^{TM}(T) and a directional vector 𝒆TM{\bm{e}}^{TM} of 𝑴TM(T){\bm{M}}^{TM}(T). MTM(T)M^{TM}(T) is defined by a part of magnetization subtracting the 4f4f electron contribution from the total magnetization. 𝑩{\bm{B}} is an applied field.

II.1 Hamiltonian of Single RR Ion

The Hamiltonian for 4f4f shell in jj-th RR ion in Eq. (1) is

^R,j\displaystyle\hat{\cal H}_{R,j} =i=1n4fh^j(i)+18πε0ii=1n4fe2|𝒓^i𝒓^i|,\displaystyle=\sum_{i=1}^{n_{4f}}\hat{h}_{j}(i)+\frac{1}{8\pi\varepsilon_{0}}\sum_{i\neq i^{\prime}=1}^{n_{4f}}\frac{e^{2}}{|\hat{\bm{r}}_{i}-\hat{\bm{r}}_{i^{\prime}}|}, (2)

with

h^j(i)=\displaystyle\hat{h}_{j}(i)= ξ𝒍^i𝒔^i+2μB𝒔^i𝑩ex,j(T)\displaystyle\xi\hat{\bm{l}}_{i}\cdot\hat{\bm{s}}_{i}+2\mu_{B}\hat{\bm{s}}_{i}\cdot{\bm{B}}_{{\rm ex},j}(T)
+ri2|R4f(ri)|2Vj(𝒓i)𝑑ri\displaystyle+\int r_{i}^{2}|R_{4f}(r_{i})|^{2}V_{j}({\bm{r}}_{i})dr_{i}
+μB(𝒍^i+2𝒔^i)𝑩.\displaystyle+\mu_{B}(\hat{\bm{l}}_{i}+2\hat{\bm{s}}_{i})\cdot{\bm{B}}. (3)

The first and second terms in Eq. (2) represent the single electron contribution and the electron-electron repulsion in 4f4f shell, respectively, where n4fn_{4f} is the number of 4f4f electrons, ε0\varepsilon_{0} and ee are the vacuum permittivity and the elementary charge, respectively. h^j(i)\hat{h}_{j}(i) in Eq. (3) is the Hamiltonian for ii-th 4f4f electron on jj-th RR site, where the first term in Eq. (3) is the spin-orbit interaction (SOI) between spin (𝒔^i\hat{\bm{s}}_{i}) and orbital (𝒍i^\hat{{\bm{l}}_{i}}) angular momenta, with a coupling constant ξ\xi. The second term represents the exchange interaction between spin moment and temperature dependent exchange field 𝑩ex,j(T)=𝒆TMBex,j(T){\bm{B}}_{{\rm ex},j}(T)=-{\bm{e}}^{TM}B_{{\rm ex},j}(T) on jj-th RR site, where μB\mu_{B} is the Bohr magneton. The third and fourth terms are the CF and Zeeman terms, respectively. In the expression of CF, Vj(𝒓i)V_{j}({\bm{r}}_{i}) and R4f(ri)R_{4f}(r_{i}) are Coulomb potential and radial parts of the 4f4f wave function on jj-th RR site, respectively. Note that the kinetic energy and screened central potential terms are effectively taken into account in the formation of 4f4f orbital.

To obtain the electronic properties at T=0T=0, we apply the first-principles and determine the parameters in the Hamiltonians in Eq. (3). We use the full-potential linearized augmented plane wave plus local orbitals (APW+lo) method implemented in the WIEN2k code wien2k . The Kohn-Sham equations are solved within the generalized-gradient approximation (GGA). To simulate localized 4f4f states, we treat 4f4f states as atomic-like core states, which is so called opencore method Richter1 ; Novak ; Hummler0 ; Richter2 ; Divis1 ; Divis2 .

We calculate the ground state properties of SmFe12 such as Coulomb potential, charge distribution, and sublattice magnetizations. In accord with the previous theoretical studies for SmFe12 Harashima ; Koerner ; Delange , we assume that Sm ion has trivalent-like electronic structure. The exchange fields Bex,j(0)B_{{\rm ex},j}(0) at T=0T=0 are determined from an energy increase caused by spin flip of 4f4f electrons Brooks ; Yoshioka , and CFs acting on ii-th 4f4f electron are directly estimated from Coulomb potential Vj(𝒓i)V_{j}({\bm{r}}_{i}) acting on jj-th RR site. It is noted that the single ion Hamiltonian ^R,j\hat{\cal H}_{R,j} thus determined for jj-th RR ions includes effects of TMTM atoms surrounding the RR ions as a mean field.

Practically, the CF term is rewritten as the following formula Novak ; Richter2 :

0rcri2|R4f(ri)|2\displaystyle\int_{0}^{r_{c}}r_{i}^{2}|R_{4f}(r_{i})|^{2} Vj(𝒓i)dri=l,mAl,jmrlal,mtlm(θ^i,φ^i),\displaystyle V_{j}({\bm{r}}_{i})dr_{i}=\sum_{l,m}\frac{A_{l,j}^{m}\langle r^{l}\rangle}{a_{l,m}}t_{l}^{m}(\hat{\theta}_{i},\hat{\varphi}_{i}), (4)
Al,jmrl=\displaystyle A_{l,j}^{m}\langle r^{l}\rangle= al,m0rc𝑑riri2|R4f(ri)|2\displaystyle a_{l,m}\int_{0}^{r_{\rm c}}dr_{i}r_{i}^{2}|R_{4f}(r_{i})|^{2}
×dΩiVj(𝒓i)tlm(θi,φi),\displaystyle\times\int d\Omega_{i}V_{j}({\bm{r}}_{i})t_{l}^{m}(\theta_{i},\varphi_{i}), (5)

where Al,jmrlA_{l,j}^{m}\langle r^{l}\rangle is CF parameter on jj-th RR site, al,ma_{l,m} is a numerical factor Hutchings , tlm(θ^i,φ^i)t_{l}^{m}(\hat{\theta}_{i},\hat{\varphi}_{i}) is tesseral harmonic function of a solid angle Ω=(θ^i,φ^i)\Omega=(\hat{\theta}_{i},\hat{\varphi}_{i}), and rcr_{\rm c} is a cut-off radius.

Values of CF parameters Al,jmrlA_{l,j}^{m}\langle r^{l}\rangle in Eq. (5), exchange field Bex,j(0)B_{{\rm ex},j}(0) in Eq. (3), TMTM-sublattice magnetization MTM(0)M^{TM}(0) in Eq. (1) in SmFe12 are shown in TABLE 1. The lattice constants used in this calculations are the experimental values a=b=8.35a=b=8.35 Å and c=4.8c=4.8 Å Hirayama2 . For Wycoff positions, we apply the theoretically optimized ones given in Ref. Harashima . The crystal structure of SmFe12 is shown in Fig. 1.

Table 1: Values of CF potentials Al,jmrlA_{l,j}^{m}\langle r^{l}\rangle [K], exchange field μBBex,j(0)/kB\mu_{\rm B}B_{{\rm ex},j}(0)/k_{\rm B} [K], and TMTM-sublattice magnetization V0MTM(0)V_{0}M^{TM}(0) [μB\mu_{\rm B}] in SmFe12 calculated by first-principles, where μB\mu_{\rm B} and kBk_{\rm B} are Bohr magneton and Boltzmann constant, respectively, and V0=a×b×cV_{0}=a\times b\times c. We note that Al,jmrlA_{l,j}^{m}\langle r^{l}\rangle and μBBex,j(0)/kB\mu_{\rm B}B_{{\rm ex},j}(0)/k_{\rm B} are independent of site index jj.
A2,j0r2A_{2,j}^{0}\langle r^{2}\rangle A4,j0r4A_{4,j}^{0}\langle r^{4}\rangle A4,j4r4A_{4,j}^{4}\langle r^{4}\rangle A6,j0r6A_{6,j}^{0}\langle r^{6}\rangle A6,j4r6A_{6,j}^{4}\langle r^{6}\rangle μBBex,j(0)/kB\mu_{\rm B}B_{{\rm ex},j}(0)/k_{\rm B} V0MTMV_{0}M^{TM}(0)
-71.4 -21.3 -49.3 5.9 3.0 296.1 51.6

II.2 Single RR Ion Hamiltonian
in LSLS Coupling Regime

We here apply the concept of LSLS coupling to the single electron Hamiltonian of Eq. (3) with Russell Saunders states |L,S;J,M|L,S;J,M\rangle, due to the strong Coulomb interaction between 4f4f electrons. According to the Hund’s rule, we specify the quantum number of total orbital (spin) moment L(S)L(S). Total angular momentum JJ is varied from |LS||L-S| to L+SL+S, and MM is the magnetic quantum number. Thus the single ion Hamiltonian in Eq. (2) can be reduced to:

^R\displaystyle\hat{\cal H}_{R} =^so+^ex+^CF+^Z,\displaystyle=\hat{\cal H}_{{\rm so}}+\hat{\cal H}_{\rm ex}+\hat{\cal H}_{\rm CF}+\hat{\cal H}_{\rm Z}, (6)
^so\displaystyle\hat{\cal H}_{\rm so} =λ𝑳^𝑺^,\displaystyle=\lambda\hat{\bm{L}}\cdot\hat{\bm{S}}, (7)
^ex\displaystyle\hat{\cal H}_{\rm ex} =2μB𝑺^𝑩ex(T),\displaystyle=2\mu_{\rm B}\hat{\bm{S}}\cdot{\bm{B}}_{\rm ex}(T), (8)
^CF\displaystyle\hat{\cal H}_{\rm CF} =l,m,mBlmΘlLCm(l)(𝑳^),\displaystyle=\sum_{l,m,m^{\prime}}B_{l}^{m}\Theta_{l}^{L}C_{m}^{(l)}(\hat{\bm{L}}), (9)
^Z\displaystyle\hat{\cal H}_{\rm Z} =μB(𝑳^+2𝑺^)𝑩.\displaystyle=\mu_{\rm B}(\hat{\bm{L}}+2\hat{\bm{S}})\cdot{\bm{B}}. (10)

Hereafter, the site index jj is omitted for single-ion quantity. 𝑳^\hat{\bm{L}} and 𝑺^\hat{\bm{S}} are total orbital and spin momenta of 4f4f electrons, respectively, Bl0=(2l+1)/4πAl0rl/al,0B_{l}^{0}=\sqrt{(2l+1)/4\pi}A_{l}^{0}\langle r^{l}\rangle/a_{l,0} and Bl±|m|=(1)m(2l+1)/8π[Al|m|rliAl|m|rl]/al,mB_{l}^{\pm|m|}=(\mp 1)^{m}\sqrt{(2l+1)/8\pi}\left[A_{l}^{|m|}\langle r^{l}\rangle\mp iA_{l}^{-|m|}\langle r^{l}\rangle\right]/a_{l,m} for m0m\neq 0, and ΘlL=LiCm(l)(θ^i,φ^i)L/LiCm(l)(𝑳^)L\Theta_{l}^{L}=\langle L\parallel\sum_{i}C^{(l)}_{m}(\hat{\theta}_{i},\hat{\varphi}_{i})\parallel L\rangle/\langle L\parallel\sum_{i}C^{(l)}_{m}(\hat{\bm{L}})\parallel L\rangle. In the treatment of SOI, we should note that the eigenstates of LSLS coupling are specified by the quantum number of JJ. In general, the term ^so\hat{\cal H}_{\rm so} is dominating in Eq. (6). Thus JJ is a good quantum number in most of the RR-4f4f systems. Because the LSLS coupling in Sm compounds is weak compared with other RR ones, it is necessary to include excited JJ-multiplets. Hereafter, we abbreviate the states |L,S;J,M|L,S;J,M\rangle as |J,M|J,M\rangle.

Refer to caption
Figure 2: Energy levels as a function of Bex(T)/Bex(0)B_{\rm ex}(T)/B_{\rm ex}(0) of the Sm-4f4f states in SmFe12 at 𝒆TM=𝒏c{\bm{e}}^{TM}={\bm{n}}_{c} and for 𝑩=𝟎{\bm{B}}={\bm{0}}. High energy levels originated from J=13/2J=13/2 and J=15/2J=15/2 multiplets are above 6000 K, which are not shown.

The energy levels of Sm-4f4f states in SmFe12 depend on 𝑩ex(T){\bm{B}}_{\rm ex}(T) and applied field 𝑩{\bm{B}}. Fig. 2 shows the Bex(T)/Bex(0)B_{\rm ex}(T)/B_{\rm ex}(0) dependence of the energy levels for 𝒆TM{\bm{e}}^{TM}=𝒏c{\bm{n}}_{c}, which is a unit vector parallel to cc-axis, and for 𝑩=𝟎{\bm{B}}={\bm{0}}. The data needed are given in TABLE 1. As for SOI constant, we use experimental value of λ/kB=ξ/5kB=411\lambda/k_{B}=\xi/5k_{B}=411 K Elliott . At Bex(T)=0B_{\rm ex}(T)=0 the SS is strongly coupled with LL to form a Kramers doublet with a total angular momentum JJ due to the large LSLS coupling with fine CF splitting. With increasing Bex(T)/Bex(0)B_{\rm ex}(T)/B_{\rm ex}(0), the exchange field breaks the time-reversal symmetry and lift the degeneracy.

II.3 Phenomenological Model for TMTM Sublattice

For finite temperature magnetic properties of TM, we apply a phenomenological formula assuming uniform MTM(T)M^{TM}(T) and K1TM(T)K_{1}^{TM}(T). For MTM(T)M^{TM}(T), we apply the Kuz’min formula Kuzminmag :

MTM(T)MTM(0)\displaystyle\frac{M^{TM}(T)}{M^{TM}(0)} =Bex(T)Bex(0)=α(T),\displaystyle=\frac{B_{\rm ex}(T)}{B_{\rm ex}(0)}=\alpha(T), (11)
α(T)\displaystyle\alpha(T) =[1s(TTC)3/2(1s)(TTC)5/2]1/3,\displaystyle=\left[1-s\left(\frac{T}{T_{\rm C}}\right)^{3/2}-(1-s)\left(\frac{T}{T_{\rm C}}\right)^{5/2}\right]^{1/3}, (12)

where, TCT_{\rm C} is Curie temperature and ss is a fitting parameter. The temperature dependence of K1TM(T)K_{1}^{TM}(T) has been expressed by an extended power law Miura-pl :

K1TM(T)K1TM(0)=\displaystyle\frac{K_{1}^{TM}(T)}{K_{1}^{TM}(0)}= α3(T)+87C1[α3(T)α10(T)]\displaystyle\alpha^{3}(T)+\frac{8}{7}C_{1}\left[\alpha^{3}(T)-\alpha^{10}(T)\right]
+87C2[α(T)31811α(T)10+711α(T)21],\displaystyle+\frac{8}{7}C_{2}\left[\alpha(T)^{3}-\frac{18}{11}\alpha(T)^{10}+\frac{7}{11}\alpha(T)^{21}\right], (13)

where C1C_{1} and C2C_{2} are fitting parameters.

In present study for SmFe12 compound, we use values of s=0.01s=0.01 and TC=555T_{\rm C}=555 K in Eq. (12), as used by Hirayama etet alal. Hirayama2 . They showed that the magnetization agrees well with experimental measurement for SmFe12. The values of C1C_{1}, C2C_{2} and V0K1TM(0)V_{0}K_{1}^{TM}(0) in Eq. (13) are determined as 0.263-0.263, 0.237-0.237 and 47.747.7 K, respectively, by fitting the expression to observed data for YFe11Ti in Ref. Nikitin .


III Method of Model Calculations

III.1 Statistical Method

To calculate the finite temperature magnetic properties, we use the model Hamiltonian and calculate MA and magnetic moment for Sm 4f4f electrons using the statistical method for the partial system. Using the eigenvalues of the Hamiltonian Eq. (6), we express the free energy density as,

G(𝒆TM,T,𝑩)=\displaystyle G({\bm{e}}^{TM},T,{\bm{B}})= 1V0j=1nRgj(𝒆TM,T,𝑩)\displaystyle\frac{1}{V_{0}}\sum_{j=1}^{n_{R}}g_{j}({\bm{e}}^{TM},T,{\bm{B}})
+K1TM(T)sin2θTM𝑩𝑴TM(T),\displaystyle+K_{1}^{TM}(T)\sin^{2}\theta^{TM}-{\bm{B}}\cdot{\bm{M}}^{TM}(T), (14)
gj(𝒆TM,T,𝑩)=\displaystyle g_{j}({\bm{e}}^{TM},T,{\bm{B}})= kBTlnZj(𝒆TM,T,𝑩),\displaystyle-k_{\rm B}T\ln Z_{j}({\bm{e}}^{TM},T,{\bm{B}}), (15)
Zj(𝒆TM,T,𝑩)=\displaystyle Z_{j}({\bm{e}}^{TM},T,{\bm{B}})= nexp[En,j(𝒆TM,T,𝑩)kBT],\displaystyle\sum_{n}\exp\left[-\frac{E_{n,j}({\bm{e}}^{TM},T,{\bm{B}})}{k_{\rm B}T}\right], (16)

where gj(𝒆TM,T,𝑩)g_{j}({\bm{e}}^{TM},T,{\bm{B}}) is Gibbs free energy for RR-4f4f partial system, En,j(𝒆TM,T,𝑩)E_{n,j}({\bm{e}}^{TM},T,{\bm{B}}) and Zj(𝒆TM,T,𝑩)Z_{j}({\bm{e}}^{TM},T,{\bm{B}}) are the eigenvalue and the partition function of jj-th RR Hamiltonian ^R,j\hat{\cal H}_{R,j} [Eq. (6)] for given 𝒆TM{\bm{e}}^{TM}, respectively. The direction of the TMTM magnetization 𝒆TM{\bm{e}}^{TM} is treated as an external parameter. The equilibrium condition of the system for given TT and 𝑩\bm{B} is:

G(𝒆0TM,T,𝑩)=min𝒆TMG(𝒆TM,T,𝑩),G({\bm{e}}_{0}^{TM},T,{\bm{B}})=\min_{{\bm{e}}^{TM}}G({\bm{e}}^{TM},T,{\bm{B}}), (17)

where 𝒆0TM{\bm{e}}_{0}^{TM} is the direction of TMTM sublattice magnetization in the equilibrium. In practice, we determine the minimal G(𝒆TM,T,𝑩)G({\bm{e}}^{TM},T,{\bm{B}}) numerically by changing 𝒆TM{\bm{e}}^{TM}.

The MA energy is given by the free energy G(𝒆TM,T,𝟎)G({\bm{e}}^{TM},T,{\bm{0}}) with different directional vector 𝒆TM{\bm{e}}^{TM}. In the tetragonal symmetry, gj(𝒆TM,T,𝟎)g_{j}({\bm{e}}^{TM},T,{\bm{0}}) in G(𝒆TM,T,𝟎)G({\bm{e}}^{TM},T,{\bm{0}}) is formally expressed as Kuzminlinear ; Miura-pl :

gj(𝒆TM,T,𝟎)=\displaystyle g_{j}({\bm{e}}^{TM},T,{\bm{0}})= p=1[kp,j(T)+q=1p/2kp,jq(T)cos(4qφTM)]\displaystyle\sum_{p=1}^{\infty}\left[k_{p,j}(T)+\sum_{q=1}^{\lfloor p/2\rfloor}k_{p,j}^{q}(T)\cos(4q\varphi^{TM})\right]
×sin2pθTM+C(T),\displaystyle\times\sin^{2p}\theta^{TM}+C(T), (18)

where θTM\theta^{TM} and φTM\varphi^{TM} are polar and azimuthal angle of 𝒆TM{\bm{e}}^{TM}, respectively, p/2\lfloor p/2\rfloor indicates the greatest integer of p/2p/2, and kp,j(T)k_{p,j}(T) and kp,jq(T)k_{p,j}^{q}(T) are out-of-plane and in-plane MA constant for jj-th RR ion. The C(T)C(T) is an angle independent constant. The series expansion does not guarantee the convergence Kuzmin2nd ; Kuzminlimit , however, for finite pp, kp,jq(T)k_{p,j}^{q}(T) can be obtained from the comparison between Taylor series of gj(𝒆TM,T,𝟎)g_{j}({\bm{e}}^{TM},T,{\bm{0}}) of Eqs. (15) and (18) with respect to θTM\theta^{TM} for a fixed φTM\varphi^{TM} Sasaki ; Miura-direct as:

gj(𝒆TM,T,𝟎)=\displaystyle g_{j}({\bm{e}}^{TM},T,{\bm{0}})= gj(0)(T)+gj(1)(T)θTM\displaystyle g_{j}^{(0)}(T)+g_{j}^{(1)}(T)\theta^{TM}
+12!gj(2)(T)(θTM)2+,\displaystyle+\frac{1}{2!}g_{j}^{(2)}(T)(\theta^{TM})^{2}+\cdots,
gj(n)(T)=\displaystyle g_{j}^{(n)}(T)= ngj(θTM,φTM,T,𝟎)(θTM)n|θTM=0,φTM=π/8,\displaystyle\left.\frac{\partial^{n}g_{j}(\theta^{TM},\varphi^{TM},T,{\bm{0}})}{\partial(\theta^{TM})^{n}}\right|_{\begin{subarray}{c}\theta^{TM}=0,\\ \varphi^{TM}=\pi/8\end{subarray}},

and

gj(𝒆TM,T,𝟎)=\displaystyle g_{j}({\bm{e}}^{TM},T,{\bm{0}})= k1,j(T)(θTM)2\displaystyle k_{1,j}(T)(\theta^{TM})^{2}
+[23!k1,j(T)+k2,j(T)](θTM)4+,\displaystyle+\left[-\frac{2}{3!}k_{1,j}(T)+k_{2,j}(T)\right](\theta^{TM})^{4}+\cdots,

respectively, which are resulting in

k1,j(T)\displaystyle k_{1,j}(T) =12gj(2)(T),\displaystyle=\frac{1}{2}g_{j}^{(2)}(T), (19)
k2,j(T)\displaystyle k_{2,j}(T) =13k1,j(T)+14!gj(4)(T),\displaystyle=\frac{1}{3}k_{1,j}(T)+\frac{1}{4!}g_{j}^{(4)}(T), (20)

etc. Using MA energy on the single RR ion in Eq. (18), the total MA constants are obtained as

K1(T)\displaystyle K_{1}(T) =1V0j=1nRk1,j(T)+K1TM(T),\displaystyle=\frac{1}{V_{0}}\sum_{j=1}^{n_{R}}k_{1,j}(T)+K_{1}^{TM}(T), (p=1),\displaystyle(p=1), (21)
Kp(q)(T)\displaystyle K^{(q)}_{p}(T) =1V0j=1nRkp,j(q)(T),\displaystyle=\frac{1}{V_{0}}\sum_{j=1}^{n_{R}}k_{p,j}^{(q)}(T), (p2),\displaystyle(p\geq 2), (22)

where Kp(T)K_{p}(T) and Kpq(T)K_{p}^{q}(T) are out-of-plane and in-plane MA constant in whole system.

The orbital and spin components of the magnetic moment of a single RR ion in the equilibrium, can be calculated by:

𝒎L,j(T,𝑩)\displaystyle{\bm{m}}_{L,j}(T,{\bm{B}}) =μBnρn,j(T,𝑩)n,j|𝑳^|n,j,\displaystyle=-\mu_{\rm B}\sum_{n}\rho_{n,j}(T,{\bm{B}})\langle n,j|{\hat{\bm{L}}}|n,j\rangle, (23)
𝒎S,j(T,𝑩)\displaystyle{\bm{m}}_{S,j}(T,{\bm{B}}) =2μBnρn,j(T,𝑩)n,j|𝑺^|n,j,\displaystyle=-2\mu_{\rm B}\sum_{n}\rho_{n,j}(T,{\bm{B}})\langle n,j|{\hat{\bm{S}}}|n,j\rangle, (24)

respectively, where ρn,j(T,𝑩)=exp[βEn,j(𝒆0TM,𝑩)]/Zj(𝒆0TM,T,𝑩)\rho_{n,j}(T,{\bm{B}})=\exp\left[-\beta E_{n,j}({\bm{e}}^{TM}_{0},{\bm{B}})\right]/Z_{j}({\bm{e}}^{TM}_{0},T,{\bm{B}}) and |n,j|n,j\rangle is the nn-th eigenstate for En,j(𝒆0TM,𝑩)E_{n,j}({\bm{e}}_{0}^{TM},{\bm{B}}), and the total magnetization Ms(T,𝑩)M_{\rm s}(T,{\bm{B}}) is given as,

𝑴s(T,𝑩)\displaystyle{\bm{M}}_{\rm s}(T,{\bm{B}}) =1V0j=1nR𝒎j(T,𝑩)+MTM(T)𝒆0TM,\displaystyle=\frac{1}{V_{0}}\sum_{j=1}^{n_{R}}{\bm{m}}_{j}(T,{\bm{B}})+M^{TM}(T){\bm{e}}_{0}^{TM}, (25)

with 𝒎j(T,𝑩)=𝒎L,j(T,𝑩)+𝒎S,j(T,𝑩){\bm{m}}_{j}(T,{\bm{B}})={\bm{m}}_{L,j}(T,{\bm{B}})+{\bm{m}}_{S,j}(T,{\bm{B}}).

Finally, to confirm the convergence of the probability weights for excited-JJ multiplet states at 𝑩=𝟎{\bm{B}}={\bm{0}}, we define a following weight function:

WJ(T)=n,Mρn,j(T,𝟎)|n,j|J,M|2.W_{J}(T)=\sum_{n,M}\rho_{n,j}(T,{\bm{0}})|\langle n,j|J,M\rangle|^{2}. (26)

In the case of SmFe12 crystal, the value of WJ(T)W_{J}(T) is independent of site index jj.

Table 2: Probability weight for each JJ-multiplet calculated by WJ(T)W_{J}(T) in Eq. (26). For J=13/2J=13/2 and 15/215/2, WJ(T)=0.0W_{J}(T)=0.0.
JJ 5/2 7/2 9/2 11/2
TT=0 0.93217 0.06548 0.00229 0.00005
T=TCT=T_{\rm C} 0.90536 0.09049 0.00406 0.00009

The results are shown in TABLE 2, which indicates good convergence of weight for the number of the excited JJ-multiplets even at T=TC=555T=T_{\rm C}=555 K. Thus in the calculation using statistical method for SmFe12, we take the excited JJ-multiplets up to J=9/2J=9/2. In the analytical calculation, the JJ-mixing effects are approximately treated only for the lowest-JJ multiplet by using unitary transformation.

III.2 Analytical Method

According to hierarchy of energy scale in RR intermetallic compounds: ^so^ex^CF^Z\hat{\cal H}_{\rm so}\gg\hat{\cal H}_{\rm ex}\gg\hat{\cal H}_{\rm CF}\sim\hat{\cal H}_{\rm Z}, we develop an analytical method for finite temperature magnetic properties, which enables us to connect the thermodynamic properties directly to our model parameters based on electronic states. Practically, we generalize the analytical expression of Gibbs free energy Yamada ; Kuzminlinear to include the effects of JJ-mixing using a first-order perturbation for the CF potential and Zeeman energy. We also derive an analytical expression for the magnetization curve, which enables us to estimate the CF potential using the observed results. The procedure of the formalism consists of (i) construction of starting Hamiltonian for single RR ion, (ii) approximation for diagonal matrix element of an effective Hamiltonian, (iii) finite temperature perturbation for single RR ion, and (iv) thermodynamic analysis.

III.2.1 Effective Lowest-JJ Multiplet Hamiltonian
for Single RR Ion

To restrict ^R\hat{\cal H}_{R} in low-energy subspace for ^so^ex\hat{\cal H}_{\rm so}\gg\hat{\cal H}_{\rm ex}, the effective lowest-JJ multiplet Hamiltonian ^ReffJ\hat{\cal H}_{R}^{{\rm eff}J} is obtained by unitary transformation and projection, where the off-diagonal matrix elements between inter-JJ multiplets become negligibly small, and compensating term ^mix\hat{\cal H}_{\rm mix} is added in diagonal element for lowest-JJ multiplet. We here introduce modified version of effective Hamiltonian as explained below.

First, we define a rotational operator 𝒟^(𝒆TM)\hat{\cal D}({\bm{e}}^{TM}) which transforms the quantization axis to 𝒆TM{\bm{e}}^{TM}. With this operator, the Hamiltonian ^R\hat{\cal H}_{R} and ^A\hat{\cal H}_{A} (AA=ex, CF and Z) is transformed to:

𝒟^(𝒆TM\displaystyle\hat{\cal D}^{\dagger}({\bm{e}}^{TM} )^R𝒟^(𝒆TM)\displaystyle)\hat{\cal H}_{R}\hat{\cal D}({\bm{e}}^{TM})
^R=^so+^ex+^CF+^Z,\displaystyle\equiv\hat{\cal H}_{R}^{\prime}=\hat{\cal H}_{\rm so}^{\prime}+\hat{\cal H}_{\rm ex}^{\prime}+\hat{\cal H}_{\rm CF}^{\prime}+\hat{\cal H}_{\rm Z}^{\prime}, (27)
^so=\displaystyle\hat{\cal H}_{\rm so}^{\prime}= λ2[𝑱^2L(L+1)S(S+1)],\displaystyle\frac{\lambda}{2}\left[\hat{\bm{J}}^{2}-L(L+1)-S(S+1)\right], (28)
^ex=\displaystyle\hat{\cal H}_{\rm ex}^{\prime}= 2Bex(T)C0(1)(𝑺^),\displaystyle-2B_{\rm ex}(T)C_{0}^{(1)}(\hat{\bm{S}}), (29)
^CF=\displaystyle\hat{\cal H}_{\rm CF}^{\prime}= l,m,mBlmΘlL[𝒟m,m(l)(𝒆TM)]Cm(l)(𝑳^),\displaystyle\sum_{l,m,m^{\prime}}B_{l}^{m}\Theta_{l}^{L}[{\cal D}_{m,m^{\prime}}^{(l)}({\bm{e}}^{TM})]^{\ast}C_{m^{\prime}}^{(l)}(\hat{\bm{L}}), (30)
^Z=\displaystyle\hat{\cal H}_{\rm Z}^{\prime}= μBm,mbm(1)[𝒟m,m(1)(𝒆TM)][Cm(1)(𝑳^)+2Cm(1)(𝑺^)],\displaystyle\mu_{\rm B}\sum_{m,m^{\prime}}b_{-m}^{(1)}[{\cal D}_{m,m^{\prime}}^{(1)}({\bm{e}}^{TM})]^{\ast}\left[C_{m^{\prime}}^{(1)}(\hat{\bm{L}})+2C_{m^{\prime}}^{(1)}(\hat{\bm{S}})\right], (31)

where 𝑱^=𝑳^+𝑺^\hat{\bm{J}}=\hat{\bm{L}}+\hat{\bm{S}}, Cq(k)(𝑨^)C_{q}^{(k)}(\hat{\bm{A}}) is the spherical tensor operator with rank kk for angular momentum 𝑨^\hat{\bm{A}} Edomonds , and bm(1)b_{m}^{(1)} is a magnetic field tensor: b0(1)=Bzb_{0}^{(1)}=B_{z} and b±1(1)=(±Bx+iBy)/2b_{\pm 1}^{(1)}=-(\pm B_{x}+iB_{y})/\sqrt{2}. 𝒟m,m(l)(𝒆TM)=𝒟m,m(l)(φTM,θTM,0){\cal D}_{m,m^{\prime}}^{(l)}({\bm{e}}^{TM})={\cal D}_{m,m^{\prime}}^{(l)}(\varphi^{TM},\theta^{TM},0) is the Wigner’s DD function. Now we apply a unitary transformation (Schrieffer-Wolf transformation Schrieffer ) to ^R\hat{\cal H}_{R}^{\prime},

eiΩ^^ReiΩ^=^R+i[Ω^,^R]+O(Ω^2),e^{i\hat{\Omega}}\hat{\cal H}_{R}^{\prime}e^{-i\hat{\Omega}}=\hat{\cal H}_{R}^{\prime}+i\left[\hat{\Omega},\hat{\cal H}_{R}^{\prime}\right]+O(\hat{\Omega}^{2}), (32)

and introduce a projection operator 𝒫^J=M=JJ|J,MJ,M|\hat{\cal P}_{J}=\sum_{M=-J}^{J}|J,M\rangle\langle J,M|, by which the space of the JJ-multiplet is restricted to the lowest one. The operator Ω^\hat{\Omega} is defined so as to remove the first-order off-diagonal matrix elements for JJ in ^R\hat{\cal H}_{R}^{\prime}:

iJ[Ω^,𝒫^J^R𝒫^J]=J𝒫^J^R𝒫^J^R.i\sum_{J^{\prime}}\left[\hat{\Omega},\hat{\cal P}_{J^{\prime}}\hat{\cal H}_{R}^{\prime}\hat{\cal P}_{J^{\prime}}\right]=\sum_{J^{\prime}}\hat{\cal P}_{J^{\prime}}\hat{\cal H}_{R}^{\prime}\hat{\cal P}_{J^{\prime}}-\hat{\cal H}_{R}^{\prime}. (33)

Apparently, J,M|Ω^|J,M=0\langle J,M|\hat{\Omega}|J,M^{\prime}\rangle=0. The second term of the right-hand-side of Eq. (32) has now a diagonal matrix with corrections to the diagonal elements in the original ^R\hat{\cal H}_{R}^{\prime}. The second and higher-order terms in Ω^\hat{\Omega} are neglected. By inserting Eq. (33) to Eq. (32), we obtain

^ReffJ=\displaystyle\hat{\cal H}_{R}^{{\rm eff}J}= 𝒫^JeiΩ^^ReiΩ^𝒫^J^RJ+^mix,\displaystyle\hat{\cal P}_{J}e^{i\hat{\Omega}}\hat{\cal H}_{R}^{\prime}e^{-i\hat{\Omega}}\hat{\cal P}_{J}\equiv\hat{\cal H}_{R}^{J}+\hat{\cal H}_{\rm mix}, (34)
^RJ=\displaystyle\hat{\cal H}_{R}^{J}= 𝒫^J^R𝒫^J=EJ+^exJ+^CFJ+^ZJ,\displaystyle\hat{\cal P}_{J}\hat{\cal H}_{R}^{\prime}\hat{\cal P}_{J}=E_{J}+\hat{\cal H}_{\rm ex}^{J}+\hat{\cal H}_{\rm CF}^{J}+\hat{\cal H}_{\rm Z}^{J}, (35)
^mix=\displaystyle\hat{\cal H}_{\rm mix}= i2𝒫^J[Ω^,^R]𝒫^J,\displaystyle\frac{i}{2}\hat{\cal P}_{J}\left[\hat{\Omega},\hat{\cal H}_{R}^{\prime}\right]\hat{\cal P}_{J}, (36)

where EJ=λ[J(J+1)L(L+1)S(S+1)]/2E_{J}=\lambda[J(J+1)-L(L+1)-S(S+1)]/2 and ^AJ=𝒫^J^A𝒫^J\hat{\cal H}_{A}^{J}=\hat{\cal P}_{J}\hat{\cal H}_{A}^{\prime}\hat{\cal P}_{J} (AA=ex, CF and Z). We here classify analytical models depending on the approximation to the matrix element of Ω^\hat{\Omega} for JJJ\neq J^{\prime} in Eq. (33) as follows:

  • model A: Lowest-JJ multiplet without mixing as:

    J,M|Ω^A|J,M=0,\langle J,M|\hat{\Omega}^{\rm A}|J^{\prime},M^{\prime}\rangle=0,
  • model B: Effective lowest-JJ multiplet with mixing as:

    J,M|Ω^B|J,M=iJ,M|^1|J,MEJEJ,\langle J,M|\hat{\Omega}^{\rm B}|J^{\prime},M^{\prime}\rangle=i\frac{\langle J,M|\hat{\cal H}_{1}|J^{\prime},M^{\prime}\rangle}{E_{J^{\prime}}-E_{J}},
  • model C: Modified effective lowest-JJ multiplet with mixing (present study) as:.

    J,M|\displaystyle\langle J,M| Ω^C|J,M\displaystyle\hat{\Omega}^{\rm C}|J^{\prime},M^{\prime}\rangle
    =\displaystyle= iJ,M|^1|J,MEJEJi(EJEJ)2\displaystyle i\frac{\langle J,M|\hat{\cal H}_{1}|J^{\prime},M^{\prime}\rangle}{E_{J^{\prime}}-E_{J}}-\frac{i}{(E_{J^{\prime}}-E_{J})^{2}}
    ×M′′[J,M|^1|J,M′′J,M′′|^1|J,M\displaystyle\times\sum_{M^{\prime\prime}}\left[\langle J,M|\hat{\cal H}_{1}|J^{\prime},M^{\prime\prime}\rangle\langle J^{\prime},M^{\prime\prime}|\hat{\cal H}_{1}|J^{\prime},M^{\prime}\rangle\right.
    J,M|^1|J,M′′J,M′′|^1|J,M],\displaystyle\left.\qquad\quad-\langle J,M|\hat{\cal H}_{1}|J,M^{\prime\prime}\rangle\langle J,M^{\prime\prime}|\hat{\cal H}_{1}|J^{\prime},M^{\prime}\rangle\right],

where ^1^R^so\hat{\cal H}_{1}\equiv\hat{\cal H}_{R}^{\prime}-\hat{\cal H}_{\rm so}^{\prime}. The approximations are referred to as model A, B and C, hereafter. By using Ω^B\hat{\Omega}^{\rm B}, Magnani etet alal. derived the effective lowest-JJ multiplet Hamiltonian Magnani and Kuz’min had also derived an equivalent approximation for anisotropy constants Kuzminmix . In the latter work, it was pointed out that the approximations of the models A and B are not applicable to the Sm compounds due to relatively small λ\lambda. In the present study, we have modified Ω^B\hat{\Omega}^{\rm B} to Ω^C\hat{\Omega}^{\rm C}.

III.2.2 Approximation for Diagonal Matrix Element of ^ReffJ\hat{\cal H}_{R}^{{\rm eff}J}

The energy levels for 4f4f electron system are obtained by the exact diagonalization of ^R\hat{\cal H}_{R} in Eq. (6), and the diagonal matrix elements of ^ReffJ\hat{\cal H}_{R}^{{\rm eff}J} can be expressed as:

J,M|\displaystyle\langle J,M| ^ReffJ|J,M\displaystyle\hat{\cal H}_{R}^{{\rm eff}J}|J,M\rangle
=J,M|^R|J,M+J,M|^mix|J,M,\displaystyle=\langle J,M|\hat{\cal H}_{R}^{\prime}|J,M\rangle+\langle J,M|\hat{\cal H}_{\rm mix}|J,M\rangle, (37)

through two unitary transformations by 𝒟^(𝒆TM)\hat{\cal D}({\bm{e}}^{TM}) and eΩ^e^{-\hat{\Omega}}. The first term in Eq. (37) can be obtained by using the relation 𝒟m,0(l)(φTM,θTM,0)=Ylm(θTM,φTM){\cal D}_{m,0}^{(l)}(\varphi^{TM},\theta^{TM},0)=Y_{l}^{m}(\theta^{TM},\varphi^{TM}) and Wigner Eckert theorem Edomonds ,

J,M|\displaystyle\langle J,M| ^R|J,M\displaystyle\hat{\cal H}_{R}^{\prime}|J,M\rangle
=\displaystyle= EJ2(gJ1)μBBex(T)J,M|C0(1)(𝑱^)|J,M\displaystyle E_{J}-2(g_{J}-1)\mu_{\rm B}B_{\rm ex}(T)\langle J,M|C_{0}^{(1)}(\hat{\bm{J}})|J,M\rangle
+l,mAlmrlΘlJtlm(𝒆TM)al,mJ,M|C0(l)(𝑱^)|J,M\displaystyle+\sum_{l,m}A_{l}^{m}\langle r^{l}\rangle\Theta_{l}^{J}\frac{t_{l}^{m}({\bm{e}}^{TM})}{a_{l,m}}\langle J,M|C_{0}^{(l)}(\hat{\bm{J}})|J,M\rangle
+μBgJ(𝒆TM𝑩)J,M|C0(1)(𝑱^)|J,M,\displaystyle+\mu_{\rm B}g_{J}\left({\bm{e}}^{TM}\cdot{\bm{B}}\right)\langle J,M|C_{0}^{(1)}(\hat{\bm{J}})|J,M\rangle, (38)

where ΘlJ\Theta_{l}^{J} is the Stevens factor Stevens ; Hutchings . By using the model C with Ω^C\hat{\Omega}^{\rm C}, the second term in Eq. (37) is approximated as,

J,\displaystyle\langle J, M|^mix|J,M\displaystyle M|\hat{\cal H}_{\rm mix}|J,M\rangle
\displaystyle\sim 1ΔsoJ,M|^ex|J+1,M\displaystyle-\frac{1}{\Delta_{\rm so}}\langle J,M|\hat{\cal H}_{\rm ex}^{\prime}|J+1,M\rangle
×J+1,M|^ex+2^CF+2^Z|J,M\displaystyle\times\langle J+1,M|\hat{\cal H}_{\rm ex}^{\prime}+2\hat{\cal H}_{\rm CF}^{\prime}+2\hat{\cal H}_{\rm Z}^{\prime}|J,M\rangle
×[1J+1,M|H^ex|J+1,MJ,M|H^ex|J,MΔso],\displaystyle\times\left[1-\frac{\langle J+1,M|\hat{H}_{\rm ex}^{\prime}|J+1,M\rangle-\langle J,M|\hat{H}_{\rm ex}^{\prime}|J,M\rangle}{\Delta_{\rm so}}\right], (39)

where Δso=λ(J+1)\Delta_{\rm so}=\lambda(J+1). Contributions from ^CF\hat{\cal H}_{\rm CF}^{\prime} and ^Z\hat{\cal H}_{\rm Z}^{\prime} are neglected in the second term of the square bracket. By using Wigner-Eckert theorem Edomonds and the relation for products of the matrix elements of the spherical tensor operators given by Eq. (5) in chapter 12. of Ref. Varshalovich , the diagonal matrix element is expressed as follows:

J,M\displaystyle\langle J,M |^mix|J,M\displaystyle|\hat{\cal H}_{\rm mix}|J,M\rangle
=\displaystyle= Δex(T)L+13SJ,M|𝒯1(𝑱^)|J,M\displaystyle-\Delta_{\rm ex}(T)\frac{L+1}{3S}\langle J,M|{\cal T}_{1}(\hat{\bm{J}})|J,M\rangle
l,mAlmrlΞlJtlm(𝒆TM)al,ml(l+1)2l+1J,M|𝒯l(𝑱^)|J,M\displaystyle-\sum_{l,m}A_{l}^{m}\langle r^{l}\rangle\Xi_{l}^{J}\frac{t_{l}^{m}({\bm{e}}^{TM})}{a_{l,m}}\frac{l(l+1)}{2l+1}\langle J,M|{\cal T}_{l}(\hat{\bm{J}})|J,M\rangle
+(𝒆TM𝑩)2(L+1)3(J+1)J,M|𝒯1(𝑱^)|J,M,\displaystyle+\left({\bm{e}}^{TM}\cdot{\bm{B}}\right)\frac{2(L+1)}{3(J+1)}\langle J,M|{\cal T}_{1}(\hat{\bm{J}})|J,M\rangle, (40)

where Δex(T)=2(gJ1)μBBex(T)\Delta_{\rm ex}(T)=-2(g_{J}-1)\mu_{\rm B}B_{\rm ex}(T). We here use the relation J=LSJ=L-S assuming RR as light rare-earth and Ξ6J=22/(33×7×11)\Xi_{6}^{J}=-2^{2}/(3^{3}\times 7\times 11) and 22×17/(35×7×112)-2^{2}\times 17/(3^{5}\times 7\times 11^{2}) for Ce3+ and Sm3+, respectively, and ΞlJ=ΘlJ\Xi_{l}^{J}=\Theta_{l}^{J} in the other cases.

More explicit expression of 𝒯1(𝑱^){\cal T}_{1}(\hat{\bm{J}}) depends on further approximations. So far two approximations have been adopted; one completely neglect the term J,M|^mix|J,M\langle J,M|\hat{\cal H}_{\rm mix}|J,M\rangle, that is, ^mix=0\hat{\cal H}_{\rm mix}=0 Kuzminlinear , and the other is an approximation to neglect the second term  in the square bracket in Eq. (39) which was adopted by Kuz’min Kuzminmix and Magnani etet alal Magnani . According to the model approximations of Ω^X\hat{\Omega}^{X} with X=X=A, B, and C, the quantities 𝒯l(𝑱^){\cal T}_{l}(\hat{\bm{J}}) are denoted as 𝒯lX(𝑱^){\cal T}_{l}^{X}(\hat{\bm{J}}) with XX=A, B, and C. Clearly 𝒯^lA=0\hat{\cal T}_{l}^{\rm A}=0, and for XX=B and C,

𝒯lB(C)(𝑱^)=\displaystyle{\cal T}_{l}^{\rm B(C)}(\hat{\bm{J}})= Δex(T)Δso[2J+l+12𝒱l1B(C)(𝑱^)\displaystyle\frac{\Delta_{\rm ex}(T)}{\Delta_{\rm so}}\left[\frac{2J+l+1}{2}{\cal V}_{l-1}^{\rm B(C)}(\hat{\bm{J}})\right.
22J+l+2𝒱l+1B(C)(𝑱^)],\displaystyle\left.-\frac{2}{2J+l+2}{\cal V}_{l+1}^{\rm B(C)}(\hat{\bm{J}})\right], (41)

with

𝒱lB(𝑱^)=\displaystyle{\cal V}_{l}^{\rm B}(\hat{\bm{J}})= C0(l)(𝑱^),\displaystyle C_{0}^{(l)}(\hat{\bm{J}}),
𝒱lC(𝑱^)=\displaystyle{\cal V}_{l}^{\rm C}(\hat{\bm{J}})= C0(l)(𝑱^)+Δex(T)ΔsoL+S+1S(J+2)\displaystyle C_{0}^{(l)}(\hat{\bm{J}})+\frac{\Delta_{\rm ex}(T)}{\Delta_{\rm so}}\frac{L+S+1}{S(J+2)}
×[l(2Jl+1)(2J+l+1)4(2l+1)C0(l1)(𝑱^)\displaystyle\times\left[\frac{l(2J-l+1)(2J+l+1)}{4(2l+1)}C_{0}^{(l-1)}(\hat{\bm{J}})\right.
+l+12l+1C0(l+1)(𝑱^)],\displaystyle\qquad\left.+\frac{l+1}{2l+1}C_{0}^{(l+1)}(\hat{\bm{J}})\right], (42)

where we formally set C0(1)(𝑱^)=0C_{0}^{(-1)}(\hat{\bm{J}})=0.

The energy levels EnE_{n} for 4f4f electron system, which consist of the lowest energy E1E_{1} to the 2J2J-th excited energy E2J+1E_{2J+1}, are now expressed as,

EMX=J,M|^R|J,M+J,M|^mixX|J,M,E_{M}^{X}=\langle J,M|\hat{\cal H}_{R}^{\prime}|J,M\rangle+\langle J,M|\hat{\cal H}_{\rm mix}^{X}|J,M\rangle, (43)

(XX=A, B, and C) with M=JM=-J to JJ for the model A, B, and C.

Refer to caption
Figure 3: (color online) Calculated energy levels of the Sm-4f4f states in SmFe12 at 𝑩=𝟎{\bm{B}}={\bm{0}}. Analytical results EMXE^{X}_{M} with XX=A, B, and C for corresponding model approximations in Eq. (43) are given by thick green (A), blue (B) and red lines (C), respectively. To clarify the contributions from ^so,^ex,\hat{\cal{\cal H}}_{\rm so},\hat{\cal H}_{\rm ex}, and ^CF\hat{\cal H}_{\rm CF}, we take original Hamiltonian ^R\hat{\cal H}_{R} as ^R=^so\hat{\cal H}_{R}=\hat{\cal H}_{\rm so}, ^so+^ex\hat{\cal H}_{\rm so}+\hat{\cal H}_{\rm ex}, and ^so+^ex+^CF\hat{\cal H}_{\rm so}+\hat{\cal H}_{\rm ex}+\hat{\cal H}_{\rm CF}. The numerically exact results are also shown by thin black lines.

Fig. 3 shows the diagonal matrix element EMXE_{M}^{X} (XX=A, B, and C) of the effective lowest-JJ multiplet Hamiltonian ^ReffJ\hat{\cal H}_{R}^{{\rm eff}J} at T=0T=0 in Eq. (43). Note that CF coefficients and exchange fields are determined by the first principles, and the same values are used for models A, B and C. The results are compared with the exact results. To distinguish the contribution from each ^so,\hat{\cal H}_{\rm so}, ^ex\hat{\cal H}_{\rm ex}, and ^CF\hat{\cal H}_{\rm CF} in ^R\hat{\cal H}_{R} of Eq. (6), the original Hamiltonian ^R\hat{\cal H}_{R} is taken as ^so\hat{\cal H}_{\rm so}, ^so+^ex\hat{\cal H}_{\rm so}+\hat{\cal H}_{\rm ex}, or ^so+^ex+^CF\hat{\cal H}_{\rm so}+\hat{\cal H}_{\rm ex}+\hat{\cal H}_{\rm CF}.

Let us first describe the characteristics for the result ^R=^so+^ex\hat{\cal H}_{R}=\hat{\cal H}_{\rm so}+\hat{\cal H}_{\rm ex}. In model A, the sixfold degeneracy of energy levels given by ^so\hat{\cal H}_{\rm so} splits into equi-energy levels as EMA=EJ+Δex(0)ME_{M}^{\rm A}=E_{J}+\Delta_{\rm ex}(0)M. In model B, the equi-energy levels shift to lower energy states by JJ-mixing term, EMB=EJ+Δex(0)M|J,M|^ex|J+1,M2/Δso.E_{M}^{\rm B}=E_{J}+\Delta_{\rm ex}(0)M-|\langle J,M|\hat{\cal H}_{\rm ex}|J+1,M\rangle^{2}/\Delta_{\rm so}. In model C, the energy shifts, which was over-estimate by the JJ-mixing term, are corrected.

The results obtained by ^R=^so+^ex+^CF\hat{\cal H}_{R}=\hat{\cal H}_{\rm so}+\hat{\cal H}_{\rm ex}+\hat{\cal H}_{\rm CF} show that the effect of CF potentials on the energy levels is weak, as expected, and they reproduce the results obtained by the numerical exact diagonalization method as shown in Fig. 3.

III.2.3 Finite Temperature Perturbation for Single RR Ion

We apply the first-order perturbation at finite temperature assuming ^exJ^CFJ+^ZJ+^mixX\hat{\cal H}_{\rm ex}^{J}\gg\hat{\cal H}_{\rm CF}^{J}+\hat{\cal H}_{\rm Z}^{J}+\hat{\cal H}_{\rm mix}^{X}. The unperturbed and perturbed Hamiltonians are ^exJ=Δex(T)C0(1)(𝑱^)^(0)\hat{\cal H}_{\rm ex}^{J}=\Delta_{\rm ex}(T)C_{0}^{(1)}(\hat{\bm{J}})\equiv\hat{\cal H}^{(0)} and ^CFJ+^ZJ+^mix^\hat{\cal H}_{\rm CF}^{J}+\hat{\cal H}_{\rm Z}^{J}+\hat{\cal H}_{\rm mix}\equiv\hat{\cal H}^{\prime}, respectively. Note that ^so\hat{\cal H}_{\rm so} is effectively taken into account in the JJ-multiplet formation of the RR ion. The approximated Gibbs free energy for RR-4f4f partial system on jj-th RR site up to first-order perturbation is formally expressed as gj(𝒆TM,T,𝑩)=kBlnZ0(T)+MρM(0)(T)J,M|^|J,Mg_{j}({\bm{e}}^{TM},T,{\bm{B}})=-k_{\rm B}\ln Z_{0}(T)+\sum_{M}\rho_{M}^{(0)}(T)\langle J,M|\hat{\cal H}^{\prime}|J,M\rangle, where EM(0)(T)=Δex(T)ME_{M}^{(0)}(T)=\Delta_{\rm ex}(T)M, Z0(T)=Mexp[βEM(0)(T)]Z_{0}(T)=\sum_{M}\exp[-\beta E_{M}^{(0)}(T)], and ρM(0)(T)=exp[βEM(0)(T)]/Z0(T)\rho_{M}^{(0)}(T)=\exp[-\beta E_{M}^{(0)}(T)]/Z_{0}(T). More explicitly, it is given as,

g(𝒆TM,T,𝑩)=\displaystyle g({\bm{e}}^{TM},T,{\bm{B}})= kBTMρM(0)(T)lnρM(0)(T)\displaystyle k_{\rm B}T\sum_{M}\rho_{M}^{(0)}(T)\ln\rho_{M}^{(0)}(T)
+MρM(0)(T)EM,\displaystyle+\sum_{M}\rho_{M}^{(0)}(T)E_{M}, (44)

by using EME_{M} in Eq. (43). It is noted that g(𝒆TM,T,𝑩)g({\bm{e}}^{TM},T,{\bm{B}}) is model dependent because EME_{M} equals to EMAE_{M}^{\rm A}, EMBE_{M}^{\rm B} or EMCE_{M}^{\rm C}, corresponding to the model adopted.

By using Helmholtz free energy f(𝒆TM,T)f({\bm{e}}^{TM},T) for RR-4f4f partial system, the Gibbs free energy in the modified effective lowest-JJ model is given as,

g(𝒆TM,T,𝑩)=\displaystyle g({\bm{e}}^{TM},T,{\bm{B}})= f(𝒆TM,T)m(T)𝒆TM𝑩,\displaystyle f({\bm{e}}^{TM},T)-m(T){\bm{e}}^{TM}\cdot{\bm{B}}, (45)
m(T)=\displaystyle m(T)= μB[gJJBJ1(x)2(L+1)3(J+1)TJ1(x)],\displaystyle\mu_{\rm B}\left[g_{J}JB_{J}^{1}(x)-\frac{2(L+1)}{3(J+1)}T_{J}^{1}(x)\right], (46)

with

f(𝒆TM,T)=\displaystyle f({\bm{e}}^{TM},T)= kBTMρM(0)(T)lnρM(0)(T)\displaystyle k_{\rm B}T\sum_{M}\rho_{M}^{(0)}(T)\ln\rho_{M}^{(0)}(T)
+fex(T)+fCF(𝒆TM,T),\displaystyle+f_{\rm ex}(T)+f_{\rm CF}({\bm{e}}^{TM},T), (47)
fex(T)=\displaystyle f_{\rm ex}(T)= Δex(T)[JBJ1(x)+L+13STJ1(x)],\displaystyle-\Delta_{\rm ex}(T)\left[JB_{J}^{1}(x)+\frac{L+1}{3S}T_{J}^{1}(x)\right], (48)
fCF(𝒆TM,T)=\displaystyle f_{\rm CF}({\bm{e}}^{TM},T)= l,mAlmrlΞlJtlm(𝒆TM)al,m\displaystyle\sum_{l,m}A_{l}^{m}\langle r^{l}\rangle\Xi_{l}^{J}\frac{t_{l}^{m}({\bm{e}}^{TM})}{a_{l,m}}
×[JlBJl(x)+l(l+1)2l+1TJl(x)].\displaystyle\times\left[J^{l}B_{J}^{l}(x)+\frac{l(l+1)}{2l+1}T_{J}^{l}(x)\right]. (49)

Here xJΔex(T)/kBTx\equiv J\Delta_{\rm ex}(T)/k_{\rm B}T, and the model dependence appears in TJl(x)T_{J}^{l}(x), which is denoted as TJl,X(x)T_{J}^{l,X}(x) with XX=A, B or C. For XX=A, TJl,A(x)=0T_{J}^{l,\rm{A}}(x)=0 and for XX=B and C,

TJl,B(C)(x)=\displaystyle T_{J}^{l,{\rm B(C)}}(x)= Δex(T)Δso[2J+l+12VJl1,B(C)(x)\displaystyle\frac{\Delta_{\rm ex}(T)}{\Delta_{\rm so}}\left[\frac{2J+l+1}{2}V_{J}^{l-1,{\rm B(C)}}(x)\right.
22J+l+2VJl+1,B(C)(x)],\displaystyle\qquad\qquad\left.-\frac{2}{2J+l+2}V_{J}^{l+1,{\rm B(C)}}(x)\right], (50)

with

VJl,B(x)=\displaystyle V_{J}^{l,{\rm B}}(x)= JlBJl(x),\displaystyle J^{l}B_{J}^{l}(x), (51)
VJl,C(x)=\displaystyle V_{J}^{l,{\rm C}}(x)= JlBJl(x)Δex(T)ΔsoL+S+1S(J+2)\displaystyle J^{l}B_{J}^{l}(x)-\frac{\Delta_{\rm ex}(T)}{\Delta_{\rm so}}\frac{L+S+1}{S(J+2)}
×[l(2Jl+1)(2J+l+1)4(2l+1)Jl1BJl1(x)\displaystyle\times\left[\frac{l(2J-l+1)(2J+l+1)}{4(2l+1)}J^{l-1}B_{J}^{l-1}(x)\right.
+l+12l+1Jl+1BJl+1(x)],\displaystyle\qquad\left.+\frac{l+1}{2l+1}J^{l+1}B_{J}^{l+1}(x)\right], (52)

where BJl(x)B_{J}^{l}(x) is the generalized Brillouin function Kuzminlinear defined by (1)lJlBJl(x)=C0(l)(𝑱^)0(-1)^{l}J^{l}B_{J}^{l}(x)=\langle C_{0}^{(l)}(\hat{\bm{J}})\rangle_{0} with x=JΔex(T)/kBTx=J\Delta_{\rm ex}(T)/k_{\rm B}T for l0l\geq 0, where 𝑨^0=MρM(0)(T)J,M|𝑨^|J,M\langle\hat{\bm{A}}\rangle_{0}=\sum_{M}\rho_{M}^{(0)}(T)\langle J,M|\hat{\bm{A}}|J,M\rangle. The analytical expression of BJl(x)B_{J}^{l}(x) is given in Ref. Magnani and TJl,A(x)=0T_{J}^{l,\rm A}(x)=0, TJl,B(x)T_{J}^{l,\rm B}(x) and TJl,C(x)T_{J}^{l,\rm C}(x) are linear combination of BJl1(x)B_{J}^{l-1}(x) and BJl+1(x)B_{J}^{l+1}(x), and BJl2(x)B_{J}^{l-2}(x), BJl(x)B_{J}^{l}(x), and BJl+2(x)B_{J}^{l+2}(x), respectively, as shown in Eq. (50).

Because of the first-order perturbation for ^Z\hat{\cal H}_{\rm Z}^{\prime}, an analytical expression of the magnetic moment m(T)m(T) is obtained as m(T)=mL(T)+mS(T)m(T)=m_{L}(T)+m_{S}(T) with:

mL(T)\displaystyle m_{L}(T) =μB[L+1J+1JBJ1(x)+2(L+1)3(J+1)TJ1(x)],\displaystyle=\mu_{\rm B}\left[\frac{L+1}{J+1}JB_{J}^{1}(x)+\frac{2(L+1)}{3(J+1)}T_{J}^{1}(x)\right], (53)
mS(T)\displaystyle m_{S}(T) =2μB[SJ+1JBJ1(x)+2(L+1)3(J+1)TJ1(x)],\displaystyle=-2\mu_{\rm B}\left[\frac{S}{J+1}JB_{J}^{1}(x)+\frac{2(L+1)}{3(J+1)}T_{J}^{1}(x)\right], (54)

where mL(T)m_{L}(T) and mS(T)m_{S}(T) are orbital and spin component of magnetic moment on the RR ion. It is noted that mL(T)m_{L}(T) and mS(T)m_{S}(T) are model dependent because of the model dependence of TJl(x)T_{J}^{l}(x) as shown above.

Within the finite temperature perturbation theory, the angular 𝒆TM{\bm{e}}^{TM} dependent part of single RR ion free energy f(𝒆TM,T)f({\bm{e}}^{TM},T) in Eq. (47) with the tetragonal symmetry can be written by:

f(𝒆TM,T)=\displaystyle f({\bm{e}}^{TM},T)= k1(T)sin2θ\displaystyle k_{1}(T)\sin^{2}\theta
+[k2(T)+k21(T)cos4φTM]sin4θTM\displaystyle+\left[k_{2}(T)+k_{2}^{1}(T)\cos 4\varphi^{TM}\right]\sin^{4}\theta^{TM}
+[k3(T)+k31(T)cos4φTM]sin6θTM\displaystyle+\left[k_{3}(T)+k_{3}^{1}(T)\cos 4\varphi^{TM}\right]\sin^{6}\theta^{TM}
+C(T),\displaystyle+C(T), (55)

which is a truncated form of g(𝐞TM,T,𝟎)g({\bf e}^{TM},T,{\bm{0}}) in Eq. (18). The C(T)C(T) is an angle independent constant. For example, the leading anisotropy constants for a trivalent magnetic light RR ion (Ce3+, Pr3+, Nd3+, Pm3+, and Sm3+) can be written as follows:

k1(T)\displaystyle k_{1}(T) =3[J2BJ2(x)+65TJ2(x)]A20r2Ξ2J\displaystyle=-3\left[J^{2}B_{J}^{2}(x)+\frac{6}{5}T_{J}^{2}(x)\right]A_{2}^{0}\langle r^{2}\rangle\Xi_{2}^{J}
40[J4BJ4(x)+209TJ4(x)]A40r4Ξ4J\displaystyle-40\left[J^{4}B_{J}^{4}(x)+\frac{20}{9}T_{J}^{4}(x)\right]A_{4}^{0}\langle r^{4}\rangle\Xi_{4}^{J}
168[J6BJ6(x)+4213TJ6(x)]A60r6Ξ6J,\displaystyle-168\left[J^{6}B_{J}^{6}(x)+\frac{42}{13}T_{J}^{6}(x)\right]A_{6}^{0}\langle r^{6}\rangle\Xi_{6}^{J}, (56)
k2(T)\displaystyle k_{2}(T) =35[J4BJ4(x)+209TJ4(x)]A40r4Ξ4J\displaystyle=35\left[J^{4}B_{J}^{4}(x)+\frac{20}{9}T_{J}^{4}(x)\right]A_{4}^{0}\langle r^{4}\rangle\Xi_{4}^{J}
+378[J6BJ6(x)+4213TJ6(x)]A60r6Ξ6J.\displaystyle+378\left[J^{6}B_{J}^{6}(x)+\frac{42}{13}T_{J}^{6}(x)\right]A_{6}^{0}\langle r^{6}\rangle\Xi_{6}^{J}. (57)

All terms of MA constants kp(q)(T)k^{(q)}_{p}(T) in model A,B, and C are given by linear terms with respect to AlmrlA_{l}^{m}\langle r^{l}\rangle.

Refer to caption
Figure 4: (color online) Temperature dependence of (a) TJl,B(C)(x)T_{J}^{l,\rm B(C)}(x) scaled by TJl,B()>0T_{J}^{l,{\rm B}}(\infty)>0 in Eq. (50) for model B(C) with broken(solid) curves and (b) generalized Brillouin function BJl(x)/BJl()B_{J}^{l}(x)/B_{J}^{l}(\infty) Kuzminlinear with J=5/2J=5/2 and x=JΔex(T)/kBx=J\Delta_{\rm ex}(T)/k_{\rm B}, where the temperature is scaled by Curie temperature TCT_{\rm C}. The dashed-dotted line represent the value of RJR_{J} (see text).

We may rewrite the approximations used and adopted in the present formalism by using TJl,X(x)T_{J}^{l,X}(x) in Eq. (50) as follows:

  • model A: Lowest-JJ multiplet without mixing as Δex(T)/Δso=0\Delta_{\rm ex}(T)/\Delta_{\rm so}=0 or TJl,A(x)=0T_{J}^{l,{\rm A}}(x)=0 Kuzminlinear

  • model B: Effective lowest-JJ multiplet with mixing as [Δex(T)/Δso]2=0\left[\Delta_{\rm ex}(T)/\Delta_{\rm so}\right]^{2}=0 or TJl,B(x)T_{J}^{l,{\rm B}}(x) Kuzminmix ; Magnani

  • model C: Modified effective lowest-JJ multiplet with mixing as TJl,C(x)T_{J}^{l,{\rm C}}(x) (present study).

At T=0T=0, we have found a following simple relation holds between TJl,C()T_{J}^{l,{\rm C}}(\infty) and TJl,B()T_{J}^{l,{\rm B}}(\infty) as:

RJ=TJl,C()TJl,B()=1Δex(0)Δso(L+S+1)JS(J+2).\displaystyle R_{J}=\displaystyle\frac{T_{J}^{l,{\rm C}}(\infty)}{T_{J}^{l,{\rm B}}(\infty)}=1-\frac{\Delta_{\rm ex}(0)}{\Delta_{\rm so}}\frac{(L+S+1)J}{S(J+2)}. (58)

Because RJR_{J} is independent of ll, relations among the models XX=A, B, and C on mL,SX(0)m_{L,S}^{X}(0) and kp(q,)X(0)k_{p}^{(q,)X}(0) can be generally expressed as follows:

mL,SC(0)\displaystyle m_{L,S}^{\rm C}(0) =mL,SA(0)+RJ[mL,SB(0)mL,SA(0)],\displaystyle=m_{L,S}^{\rm A}(0)+R_{J}\left[m_{L,S}^{\rm B}(0)-m_{L,S}^{\rm A}(0)\right], (59)
kp(q,)C(0)\displaystyle k_{p}^{(q,)\rm C}(0) =kp(q,)A(0)+RJ[kp(q,)B(0)kp(q,)A(0)].\displaystyle=k_{p}^{(q,)\rm A}(0)+R_{J}\left[k_{p}^{(q,)\rm B}(0)-k_{p}^{(q,)\rm A}(0)\right]. (60)

At finite temperatures, TJl,B(C)(x)T_{J}^{l,{\rm B(C)}}(x) for J=5/2J=5/2 scaled by TJl,B()>0T_{J}^{l,{\rm B}}(\infty)>0 are shown in Fig. 4(a) for the SmFe12 compound. Here, Δex(T)/Δso\Delta_{\rm ex}(T)/\Delta_{\rm so} is taken to be 0.206α(T)0.206\alpha(T). For comparison purpose, we also show the BJl(x)/BJl()B_{J}^{l}(x)/B_{J}^{l}(\infty) in Fig. 4(b). BJl(x)B_{J}^{l}(x) decays faster than TJl,B(C)(x)T_{J}^{l,{\rm B(C)}}(x) with increasing temperature. Thus the JJ-mixing effects included in TJl,B(C)(x)T_{J}^{l,{\rm B(C)}}(x) remain even at high temperatures.

III.2.4 Thermodynamic Analysis

Finally we investigate the thermodynamical instability by using the thermodynamic relation between Gibbs and Helmholtz free energy, which explicitly contains the CF potentials and the exchange field determined by first principles. We have to note that above the room temperature the exchange contribution ^ex\hat{\cal H}_{\rm ex} decreases with increasing temperature at a rate proportional to α(T)\alpha(T), so the energy hierarchy is changed and thermal fluctuation effects have to be considered as kBT^CF^exk_{\rm B}T\gg\hat{\cal H}_{\rm CF}\sim\hat{\cal H}_{\rm ex}. Even in this case, the formulation derived here based on generalized Brillouin function holds as shown by Kuz’min in Refs. Kuzminmix ; Kuzminlimit . In this thermodynamic analysis, we use the model C.

By applying the finite temperature perturbation theory to the lowest-JJ multiplet Hamiltonian, the approximated Gibbs free energy density for whole system can be expressed as:

G(𝒆TM,T,𝑩)\displaystyle G({\bm{e}}^{TM},T,{\bm{B}}) =F(𝒆TM,T)𝑴s(T)𝑩,\displaystyle=F({\bm{e}}^{TM},T)-{\bm{M}}_{\rm s}(T)\cdot{\bm{B}}, (61)
F(𝒆TM,T)\displaystyle F({\bm{e}}^{TM},T) =1V0j=1nRfj(𝒆TM,T)+K1TM(T)sin2θTM,\displaystyle=\frac{1}{V_{0}}\sum_{j=1}^{n_{R}}f_{j}({\bm{e}}^{TM},T)+K_{1}^{TM}(T)\sin^{2}\theta^{TM}, (62)
𝑴s(T)\displaystyle{\bm{M}}_{\rm s}(T) =[1V0j=1nRmj(T)+MTM(T)]𝒆TM,\displaystyle=\left[\frac{1}{V_{0}}\sum_{j=1}^{n_{R}}{m}_{j}(T)+M^{TM}(T)\right]{\bm{e}}^{TM}, (63)

where F(𝒆TM,T)F({\bm{e}}^{TM},T) is Helmholtz free energy density for whole system with model C and fj(𝒆TM,T)f_{j}({\bm{e}}^{TM},T) and mj(T)𝒆TMm_{j}(T){\bm{e}}^{TM} are corresponding energy for 4f4f-shell and expectation value of magnetic moment on jj-th RR ion given in Eq. (47) and Eq. (46), respectively. The temperature dependence of G(𝒆TM,T,𝑩)G({\bm{e}}^{TM},T,{\bm{B}}) can be expressed as the linear combination of the generalized Brillouin functions for RR ion BJl(JΔex/kBT)B_{J}^{l}(J\Delta_{\rm ex}/k_{\rm B}T) and the temperature coefficient for TMTM ion α(T)\alpha(T) in Eq. (12). The equilibrium condition is the same as Eq. (17), where 𝒆0TM{\bm{e}}_{0}^{TM} becomes the direction of total magnetization in the equilibrium. We can also analyze the instability of magnetic metastable states, which are crucially important in permanent magnetic materials. The metastable condition is δG(T,𝒆TM,𝑩)>0\delta G(T,{\bm{e}}^{TM},{\bm{B}})>0 for given TT and 𝑩{\bm{B}} with |eTM|=1|e^{TM}|=1.

The MA constants in whole system are obtained by combining the contribution from RR sublattice in Eq. (55) with Fe sublattice same as Eqs. (21) and (22). K1(T)K_{1}(T) can be substituted into the so called Krönmuller equation Kronmuller1 ; Kronmuller-text to obtain the coercive field

Bc(T)\displaystyle B_{\rm c}(T) =αBN(T)NeffMs(T),\displaystyle=\alpha B_{\rm N}(T)-N_{\rm eff}M_{\rm s}(T), (64)
BN(T)\displaystyle B_{\rm N}(T) =2K1(T)Ms(T),\displaystyle=\frac{2K_{1}(T)}{M_{\rm s}(T)}, (65)

where Bc(T)B_{\rm c}(T) and BN(T)B_{\rm N}(T) are coercive and nucleation field, respectively. α(<1)\alpha(<1) is microstructural parameter and NeffN_{\rm eff} is local effective demagnetization factor Kronmuller-text . The BN(T)B_{\rm N}(T) gives upper limit of Bc(T)B_{\rm c}(T).

IV Calculated Results for SmFe12

IV.1 Valence Mechanism of Magnetic Anisotropy

We first calculate the charge density distribution and Coulomb potential at 0 K on constituent atoms of SmFe12 lattice (Fig. 1) using the first principles. The calculated results determine the values of CF acting on 4f4f electrons, the magnitude of the exchange field Bex(0)B_{\rm ex}(0) acting on the JJ, and the magnitude of TMTM sublattice magnetization. These values are used for parameter values in the model Hamiltonian. The contribution to the CF from the charge density distribution inside (outside) the muffin-tine sphere radius is called ”valence (lattice) contribution” Hummler0 . If the CF is dominated by the former contribution, we call the mechanism of the MA ”valence mechanism” Tsuchiura1 .

The charge density distributions of single RR ion are approximately replaced with charge density on atomic orbitals of 6p6p and 5d5d states. To evaluated the valence contribution to CF parameters Al0rl(val)A_{l}^{0}\langle r^{l}\rangle(val), we introduce distribution parameters Δn6p(2),Δn5d(2)\Delta n_{6p}^{(2)},\Delta n_{5d}^{(2)} Coehoorn ; Sakuma and Δn5d(4)\Delta n_{5d}^{(4)} defined as,

Δnnl(l)=4π2l+1al,0m𝑑Ωtl0(θ,φ)|tlm(θ,φ)|2nnl,m,\Delta n_{n^{\prime}l^{\prime}}^{(l)}=\frac{4\pi}{2l+1}a_{l,0}\sum_{m^{\prime}}\int d\Omega\ t_{l}^{0}(\theta,\varphi)|t_{l^{\prime}}^{m^{\prime}}(\theta,\varphi)|^{2}n_{n^{\prime}l^{\prime},m^{\prime}}, (66)

where Ω\Omega is the solid angle and mm^{\prime} indicates the multiple orbitals for the quantum number (nln^{\prime}l^{\prime}). The shape of the function tl0(θ,φ)t_{l}^{0}(\theta,\varphi) in Eq. (66) is given in Fig. 5(c).

The particular cases are as follows:

Δn6p(2)=\displaystyle\Delta n_{6p}^{(2)}= 15[n6p,z12(n6p,x+n6p,y)],\displaystyle\frac{1}{5}\left[n_{6p,z}-\frac{1}{2}(n_{6p,x}+n_{6p,y})\right], (67)
Δn5d(2)=\displaystyle\Delta n_{5d}^{(2)}= 17[n5d,z2+12(n5d,xz+n5d,yz)\displaystyle\frac{1}{7}\left[n_{5d,z^{2}}+\frac{1}{2}(n_{5d,xz}+n_{5d,yz})\right.
(n5d,x2y2+n5d,xy)],\displaystyle\quad\left.\vphantom{\frac{1}{7}}-(n_{5d,x^{2}-y^{2}}+n_{5d,xy})\right], (68)
Δn5d(4)=\displaystyle\Delta n_{5d}^{(4)}= 128[n5d,z223(n5d,xz+n5d,yz)\displaystyle\frac{1}{28}\left[n_{5d,z^{2}}-\frac{2}{3}(n_{5d,xz}+n_{5d,yz})\right.
+16(n5d,x2y2+n5d,xy)],\displaystyle\qquad\left.+\frac{1}{6}(n_{5d,x^{2}-y^{2}}+n_{5d,xy})\right], (69)

where nnl,mn_{n^{\prime}l^{\prime},m^{\prime}} is the occupation number of the (nln^{\prime}l^{\prime}, mm^{\prime}) orbital. We note that Δn6p(4)=0\Delta n_{6p}^{(4)}=0. Valence contribution of A20r2A_{2}^{0}\langle r^{2}\rangle and A40r4A_{4}^{0}\langle r^{4}\rangle are determined as Richter1 ; Hummler0 ,

A20r2(val)\displaystyle A_{2}^{0}\langle r^{2}\rangle(val) =F(2)(4f,6p)Δn6p(2)+F(2)(4f,5d)Δn5d(2),\displaystyle=F^{(2)}(4f,6p)\Delta n_{6p}^{(2)}+F^{(2)}(4f,5d)\Delta n_{5d}^{(2)}, (70)
A40r4(val)\displaystyle A_{4}^{0}\langle r^{4}\rangle(val) =F(4)(4f,5d)Δn5d(4),\displaystyle=F^{(4)}(4f,5d)\Delta n_{5d}^{(4)}, (71)

with the Slater-Condon parameters:

F\displaystyle F (4f,nl)(l){}^{(l)}(4f,n^{\prime}l^{\prime})
=e24πε00rcr<lr>l+1r2|R4f(r)|2r2|Rnl(r)|2𝑑r𝑑r>0,\displaystyle=\frac{e^{2}}{4\pi\varepsilon_{0}}\iint_{0}^{r_{c}}\frac{r_{<}^{l}}{r_{>}^{l+1}}r^{2}|R_{4f}(r)|^{2}r^{\prime 2}|R_{n^{\prime}l^{\prime}}(r^{\prime})|^{2}dr^{\prime}dr>0, (72)

where r<=min(r,r)r_{<}=\min(r,r^{\prime}) and r>=max(r,r)r_{>}=\max(r,r^{\prime}). Via Eqs. (70) and (71), the distribution parameters Δnnl(l)\Delta n_{n^{\prime}l^{\prime}}^{(l)} determine Al0rl(val)A_{l}^{0}\langle r^{l}\rangle(val). It may be noted that no 6p6p and 5d5d orbitals exist for A60r6(val)A_{6}^{0}\langle r^{6}\rangle(val).

Refer to caption
Figure 5: (color online) (a) Atomic position of first (8i8i), second (8j8j), and third (8f8f) neighbor Fe atoms of Sm ion in SmFe12, (b) illustration of valence mechanism Tsuchiura1 in SmFe12, and (c) typical tesseral harmonic functions as basis of CF Hamiltonian, where signs represent the phase.

A simple explanation for the appearance of the uniaxial MA in a Sm ion surrounded by Fe atoms is given as follows. Fig. 5(a) shows the lattice strucutre of SmFe12 Hirayama2 ; Harashima . Left panel of Fig. 5(b) shows the location of Sm and Fe on (010) plane of the lattice. Because of the short atomic distance between Sm and the first nearest neighbor (n.n.) Fe(8i8i) sites, the distribution of valence electrons on Sm extends within the aba-b plane as shown in Fig. 5(c). According to the negative sign of t20(θ,φ)t_{2}^{0}(\theta,\varphi) in Fig. 5(d), the distribution parameters defined by Eq. (66) in terms of electron numbers of 6p6p and 5d5d-orbitals are negative; Δn6p(2)=0.0012\Delta n_{6p}^{(2)}=-0.0012, Δn5d(2)=0.0011\Delta n_{5d}^{(2)}=-0.0011. Therefore, we obtain A20r2(val)<0A_{2}^{0}\langle r^{2}\rangle(val)<0 by Eq. (70) in agreement with the numerical value of A20r2A_{2}^{0}\langle r^{2}\rangle shown in TABLE 1. As shown by Eq. (56) the main contribution of the MA constant k1(T)k_{1}(T) is given by a product of A20A_{2}^{0} and the positive value of Stevens factor Θ20\Theta_{2}^{0}, and k1(T)k_{1}(T) becomes positive. This means that the K1(T)>0K_{1}(T)>0 because K1TM(T)>0K_{1}^{TM}(T)>0.

On the other hand, second neighbor Fe(8jj) and third neighbor Fe(8ff) atoms of Sm atom are situated obliquely upward as shown in Fig. 5(b). According to the negative sign of t40(θ,φ)t_{4}^{0}(\theta,\varphi) shown in Fig. 5(d), we obtained Δn5d(4)=0.0013\Delta n_{5d}^{(4)}=-0.0013 using Eq. (69), and A40r4(val)<0A_{4}^{0}\langle r^{4}\rangle(val)<0 from Eq. (71). Again the negative value is consistent with the numerical values of A40r4A_{4}^{0}\langle r^{4}\rangle. The main contribution of MA constant k2(T)k_{2}(T) comes from a product of A40r4A_{4}^{0}\langle r^{4}\rangle and the positive value of Θ40\Theta_{4}^{0}, and results in K2(T)<0K_{2}(T)<0.

Thus, the sign of MA constants K1(T)K_{1}(T) and K2(T)K_{2}(T) are determined by the configuration of Sm and Fe atoms in the lattice. In the following, we investigate the JJ-mixing effect on single Sm magnetic properties at T=0T=0 K.

IV.2 JJ-Mixing Effect and Zero-Temperature Magnetic Properties of SmFe12 Compound

To clarify the JJ-mixing effect on single-ion magnetic properties, we show the calculated results of the magnetic moments mL,S(0)m_{L,S}(0) and the MA constants k1,2(0)k_{1,2}(0) for model A, B, and C in TABLE 3. We used Eqs. (53) and (54) for mL,S(0)m_{L,S}(0) and Eqs. (56) and (57) for k1,2(0)k_{1,2}(0), and the values of AlmrlA_{l}^{m}\langle r^{l}\rangle, Bex(0)B_{\rm ex}(0), and MTM(0)M^{TM}(0) in TABLE 1. As a reference, we also show the results obtained by the statistical method: mL,S(0,𝟎)m_{L,S}(0,{\bm{0}}) in Eqs. (23) and (24) and k1,2(0)k_{1,2}(0) defined in Eqs. (19) and (20). Both the analytical and statistical results give k1(0)>0k_{1}(0)>0 and k2(0)<0k_{2}(0)<0 for three models A, B, and C. The calculated results in model C (present model) agree best with the statistical ones.

We find that the absolute values of mL,S(0)m_{L,S}(0) and k1,2(0)k_{1,2}(0) in model B and C are larger than those in model A, which is attributed to inclusion of the JJ-mixing effects. The model B proposed in the previous studies Kuzminmix ; Magnani over-estimated the JJ-mixing effects by 1/RJ1/R_{J} compared with model C, where RJ=0.44R_{J}=0.44 in Eq. (58) for SmFe12 compound. Actually, values of mL,SCm_{L,S}^{\rm C} and k1,2Ck_{1,2}^{\rm C} in TABLE 3 satisfy the relation in Eq. (59) and (60). The results in present study (XX=C) quantitatively agree well with statistical ones except for k2C(0)k^{\rm C}_{2}(0). The discrepancy in k2C(0)k^{\rm C}_{2}(0) may be due to omitting the 2nd order terms of A20r2A_{2}^{0}\langle r^{2}\rangle in Eq. (57), which have a positive contribution independent of the sign of A20r2A_{2}^{0}\langle r^{2}\rangle Kuzmin2nd .

Table 3: Magnetic moments mL,S(0)m_{L,S}(0) [μB\mu_{\rm B}] in Eqs. (53) and (54) and MA constants k1,2(0)k_{1,2}(0) [K] in Eqs. (56) and (57) for model A, B, and C at 0 K for single Sm ion. Results obtained by Boltzamann statistics of mL,S(0,𝟎)m_{L,S}(0,{\bm{0}}) defined by Eqs. (23) and (24) and k1,2(0)k_{1,2}(0) defined by Eqs. (19) and (20) are also shown in the fifth column.
model A B C statistics
mLm_{L} 4.29 5.04 4.62 4.70
mSm_{S} -3.57 -5.08 -4.24 -4.39
k1k_{1} 60.2 144.5 97.7 101.1
k2k_{2} -14.0 -74.6 -40.9 -23.5

IV.3 Finite Temperature Magnetic Properties
of SmFe12 Compound

Refer to caption
Figure 6: (color online) Temperature dependence of (a) magnetic moments of Sm ion mL,S(T)m_{L,S}(T) and m(T)m(T) at 𝑩=𝟎{\bm{B}}={\bm{0}} and (b) MA constants per single Sm ion k1,2(T)k_{1,2}(T) calculated by using models A, B and C. Results obtained by Boltzmann statistics are shown by broken curves. (c) Temperature dependent MA constants K1,2(T)K_{1,2}(T) in SmFe12 compound by using statistical method for Sm sublattice contribution k1,2(T)k_{1,2}(T), which are compared with experimental ones by the Sucksmith-Thompson (circles) Hirayama2 and the anomalous Hall effect (triangles) Ogawa2 . For both calculated and experimental results in Fig. (c), K1(T)K_{1}(T) and K2(T)K_{2}(T) are shown by solid and broken curves, respectively.

Calculated results of finite temperature magnetic properties for a single Sm ion in equilibrium at 𝒆0TM=𝒏c{\bm{e}}^{TM}_{0}={\bm{n}}_{c}: the magnetic moment mL,S(T)m_{L,S}(T) in Eqs. (53) and (54) and the MA constants k1,2(T)k_{1,2}(T) in Eqs. (56) and (57) are shown in Fig. 6(a) and (b), respectively. The results show that the JJ-mixing effect in model B increases the absolute values of both mL,S(T)m_{L,S}(T) and k1,2(T)k_{1,2}(T). The over-estimation in model B is modified by the present model C in the whole temperature range. Obtained results of model C reproduce well the statistical results for mL,S(T,𝟎)m_{L,S}(T,{\bm{0}}) in Eqs. (23) and (24) and for k1,2(T)k_{1,2}(T) in Eqs. (19) and (20) as shown by broken lines in Fig. 6(a) and (b).

The physical meaning of the increment of the absolute value of mL,S(T)m_{L,S}(T) and k1,2(T)k_{1,2}(T) by JJ-mixing may be given as follows. The expression of the free energy given by Eq. (48) includes the JJ-mixing effect in the second term of the square bracket. The term decrease fex(x)f_{\rm ex}(x) by μBBex(T)δS(T)-\mu_{\rm B}B_{\rm ex}(T)\delta S(T), where δS(T)=2(L+1)3(J+1)TJ1(T)>0\displaystyle\delta S(T)=\frac{2(L+1)}{3(J+1)}T_{J}^{1}(T)>0. Because of the decrease in fex(x)f_{\rm ex}(x), the absolute value of the spin C0(1)(𝑺^)0\langle C^{(1)}_{0}(\hat{\bm{S}})\rangle_{0} and orbital moments C0(1)(𝑳^)0\langle C^{(1)}_{0}(\hat{\bm{L}})\rangle_{0} along 𝒆0TM{\bm{e}}_{0}^{TM} are increased by δS(T)\delta S(T). The tensor operators C0(l)(𝑳^)0\langle C^{(l)}_{0}(\hat{\bm{L}})\rangle_{0} for even ll are also increased by l(l+1)2l+1TJl(x)\displaystyle\frac{l(l+1)}{2l+1}T_{J}^{l}(x), which contribute to increase in the absolute value of the MA constants kp(q)(T)k_{p}^{(q)}(T).

The magnetic moment of Sm ion m(T)m(T) is reversed at around Tcomp=350T_{\rm comp}=350 K in model C and calculation by Boltzmann statistics. The temperature is called compensation temperature. This phenomenon is observed also in other Sm compounds Adachi1 ; Adachi2 . Zhao etet alal pointed out that this phenomenon also appears at T=337T=337 K in Sm2Fe17Nx using statistical method including similar parameter values with ours such as μBBex(0)/kB=300\mu_{B}B_{\rm ex}(0)/k_{\rm B}=300 K and λ/kB=411\lambda/k_{\rm B}=411 K Zhao . Their results are comparable with ours, however, the mechanism has not been surveyed. In the present model C, the magnetic moment of Sm ion can be written as m(T)=gJμBJBJ1(x)μBδS(T)m(T)=g_{J}\mu_{\rm B}JB_{J}^{1}(x)-\mu_{\rm B}\delta S(T). Because μBδS(T)\mu_{\rm B}\delta S(T) is proportional to TJ1,C(x)T_{J}^{1,{\rm C}}(x) and monotonically increasing with temperature below T/TC=0.8T/T_{\rm C}=0.8 as shown in Fig. 4(a), the term compensates the gJμBJBJ1(x)g_{J}\mu_{\rm B}JB_{J}^{1}(x) at TcompT_{\rm comp}.

Fig. 6(c) shows the results of K1(T)K_{1}(T) and K2(T)K_{2}(T) obtained by statistical method in SmFe12 compound, which are compared with experimental ones denoted by exp 1 and exp 2 measured by the Sucksmith-Thompson method Hirayama2 and anomalous Hall effect Ogawa2 , respectively. At the whole temperature region the results of K1(T)K_{1}(T) agree well with the experiments. Our statistical results qualitatively reproduce the experimental results below 200 K. The negative K2(T)K_{2}(T) at low temperatures is origin of first-order magnetization process (FOMP) as discussed below.

IV.4 Thermodynamic Properties
of SmFe12 Compound

Refer to caption
Figure 7: (color online) Dependence of Helmholtz free energy density on 𝒆TM𝒏a{\bm{e}}^{TM}\cdot{\bm{n}}_{a} with 𝒆TM𝒏b=0{\bm{e}}^{TM}\cdot{\bm{n}}_{b}=0 in SmFe12 at (a) T=0T=0 and (b) 400 K. Analytical results F(𝒆TM,T)F({\bm{e}}^{TM},T) in Eq. (62) and results of G(𝒆TM,T,𝟎)G({\bm{e}}^{TM},T,{\bm{0}}) obtained by Boltzmann statistics in Eq. (14) are shown by solid and broken curves, respectively, in which the contribution from Fe sublattice is included. The dashed-dotted curves represent the Fe sublattice MA energy: K1TM(T)(𝒆TM𝒏a)2K_{1}^{TM}(T)({\bm{e}}^{TM}\cdot{\bm{n}}_{a})^{2}

Fig. 7 shows calculated results of the Helmholtz free energy density F(𝒆TM,T)F({\bm{e}}^{TM},T) given in Eq. (62) for model C as a function of 𝒆TM𝒏a{\bm{e}}^{TM}\cdot{\bm{n}}_{a} with 𝒆TM𝒏b=0{\bm{e}}^{TM}\cdot{\bm{n}}_{b}=0 at T=0T=0 K and 400400 K, where 𝒏a(b){\bm{n}}_{a(b)} is unit vector parallel to a(b)a(b)-axis. The results are compared with statistical ones of G(𝒆TM,T,𝟎)G({\bm{e}}^{TM},T,{\bm{0}}) in Eq. (14). When the direction of 𝒆TM{\bm{e}}^{TM} is changed, the free energy density on both Sm and Fe sublattice are increased. For the Sm sublattice, the energy increase originates from the CF, which can be expressed by the jfCF,j(𝒆TM,T)\sum_{j}f_{{\rm CF},j}({\bm{e}}^{TM},T) in Eq. (55), and for Fe sublattice, the energy increase can be written by: K1TM(T)sin2θTMK_{1}^{TM}(T)\sin^{2}\theta^{TM} with K1TM(T)=1.966K_{1}^{TM}(T)=1.966 and 0.3870.387 MJ/m3 at 0 and 400 K, respectively, which are much smaller than those of Sm sublattice jk1,j(T)/V0=8.059\sum_{j}k_{1,j}(T)/V_{0}=8.059 and 2.3102.310 MJ/m3. The analytical results agree well with statistical ones.

Refer to caption
Figure 8: (color online) Magnetization curves of SmFe12 at T=0T=0 and 400 K with applied field BB parallel to aa-axis in the equilibrium calculated by analytical (solid curves) and statistical (broken curves) methods in Eqs. (63) and (25), respectively, where μ0\mu_{0} is the magnetic constant. Dashed-dotted lines show tangent lines of magnetization curves at B=0B=0: y=[μ0Ms(T)2/2K1]By=[\mu_{0}M_{\rm s}(T)^{2}/2K_{1}]B. Values of BB at the circles correspond to the nucleation field BN(T)B_{\rm N}(T) obtained by using the free energy density of model C (see text).

Fig. 8 shows calculated results of magnetization curves in the equilibrium states of SmFe12 at T=0T=0 and 400 K, where the magnetic field 𝑩{\bm{B}} is applied along aa-axis. Analytical results of the magnetization along the aa-axis are compared with statistical ones. We have confirmed that the results in model C well reproduce the statistical ones. At T=0T=0, we find characteristic behavior of an abrupt change in the magnetization 𝑴s(T)𝒏a{\bm{M}}_{\rm s}(T)\cdot{\bm{n}}_{a} at B=BFPB=B_{\rm FP}. The change is called first-order magnetization process (FOMP) and the BFPB_{\rm FP} is called as FOMP field. At TT=400 K, no FOMP appears in both analytical and statistical results and the magnetization saturates at the MA field BAB_{\rm A}. In SmFe12, the magnetization curve at low temperatures were not reported, however, in SmFe11Ti compound, FOMP observed at T=5T=5 K and BFP=10B_{\rm FP}=10 T Hu , which is qualitatively consistent with our results.

Refer to caption
Figure 9: (color online) Temperature dependence of the nucleation field BN(T)B_{\rm N}(T), the MA field BA(T)B_{\rm A}(T), and the first-order magnetization process (FOMP) field BFP(T)B_{\rm FP}(T) obtained by using the approximated free energy density of model C neglecting K3(T)K_{3}(T), K21(T)K_{2}^{1}(T), and K31(T)K_{3}^{1}(T), the details of which are shown in appendix A. BFP(T)B_{\rm FP}(T) is the point of discontinuity in the FOMP realized below FOMP temperature TFPT_{\rm FP} (see text).

Let us consider the magnetization process along cc-axis and estimate nucleation field BNB_{\rm N} in the model C. The magnetization is first saturated as Ms(T)𝒏cM_{\rm s}(T){\bm{n}}_{c} along cc-axis by an infinitesimal field. Then the direction of the magnetic field is reversed and the magnitude is increased as B𝒏c-B{\bm{n}}_{c}. The original state continues to exist as a quasi-stable state as far as the condition for a first-order variation δG(𝒏c,T,B𝒏c)>0\delta G({\bm{n}}_{c},T,-B{\bm{n}}_{c})>0 is satisfied. The magnetization tends to decline when δG(nc,T,B𝒏c)=0\delta G(nc,T,-B{\bm{n}}_{c})=0. The applied magnetic field at which the latter condition is satisfied is the nucleation field, which has been given as BN=2K1(T)/Ms(T)B_{\rm N}=2K_{1}(T)/M_{\rm s}(T) Kronmuller1 .

Because BNB_{\rm N} corresponds to the field at which the magnetization begins to decline with infinitesimally angle θ\theta against cc-axis, the magnitude BNB_{\rm N} in realistic system can be estimated once the magnetization curve is obtained along a hard axis. Fig. 8 shows the magnetization curve along aa-axis calculated in the model C. BNB_{\rm N} is given by a crossing point of the magnetization curve in the saturated state and the tangential line of the magnetization curve at zero field y=[μ0Ms(T)2/2K1(T)]y=\left[\mu_{0}M_{\rm s}(T)^{2}/2K_{1}(T)\right]. When the value of y=μ0Msy=\mu_{0}M_{\rm s}, the magnetic field coincides with BNB_{\rm N} defined in Eq. (65).

The magnetization curves along hard and easy-axis in the case of K1(T)>0K_{1}(T)>0 and K2(T)<0K_{2}(T)<0 can be characterized by the nucleation field BN(T)B_{\rm N}(T), the FOMP field BFP(T)B_{\rm FP}(T), and the MA field BA(T)B_{\rm A}(T). These values are analytically expressed by using the ratio γ(T)=K1(T)/K2(T)\gamma(T)=K_{1}(T)/K_{2}(T) as,

BN(T)\displaystyle B_{\rm N}(T) =2K1(T)Ms(T),\displaystyle=\frac{2K_{1}(T)}{M_{\rm s}(T)}, (73)
BFP(T)\displaystyle B_{\rm FP}(T) =BN[xFP(T)+2γ(T)xFP(T)3]\displaystyle=B_{\rm N}[x_{\rm FP}(T)+2\gamma(T)x_{\rm FP}(T)^{3}] (74)
(0<xFP(T)<1),\displaystyle\qquad\qquad\qquad\qquad(0<x_{\rm FP}(T)<1),
BA(T)\displaystyle B_{\rm A}(T) =BN[1+2γ(T)]\displaystyle=B_{\rm N}[1+2\gamma(T)] (75)
(xFP(T)>1),\displaystyle\qquad\qquad\qquad\qquad(x_{\rm FP}(T)>1),

with

xFP(T)\displaystyle x_{\rm FP}(T) =13(1+3γ(T)2),\displaystyle=\frac{1}{3}\left(-1+\sqrt{-\frac{3}{\gamma(T)}-2}\right), (76)

where we use the approximate free energy density: F(𝒆TM,T)=K1(T)sin2θTM+K2(T)sin4θTMF({\bm{e}}^{TM},T)=K_{1}(T)\sin^{2}\theta^{TM}+K_{2}(T)\sin^{4}\theta^{TM}, in which the small contributions K3(T)K_{3}(T), K2,31(T)K_{2,3}^{1}(T) are neglected. Details are shown in Appendix A. Calculated results of BN(T)B_{\rm N}(T), BFP(T)B_{\rm FP}(T), and BA(T)B_{\rm A}(T) are shown in Fig. 9. The condition of FOMP appearance in model C is given by K2(T)<K1(T)<6K2(T)-K_{2}(T)<K_{1}(T)<-6K_{2}(T) between 0<xFP(T)<10<x_{\rm FP}(T)<1 in Eq. (74). As for SmFe12 compound, the FOMP is realized below T=281T=281 K \equiv TFPT_{\rm FP}, which is analytically obtained from the condition: K1(T)=6K2(T)K_{1}(T)=-6K_{2}(T). The curves of BFP(T)B_{\rm FP}(T) and BA(T)B_{\rm A}(T) are continuously connected at TFPT_{\rm FP}, which is called as FOMP temperature. When K1(T)<K2(T)K_{1}(T)<-K_{2}(T) the magnetization direction is in-plane at 𝑩=𝟎{\bm{B}}={\bm{0}}.

V Summary

The temperature dependence of magnetic anisotropy (MA) constants and magnetization of SmFe12 were investigated by using two methods for the model Hamiltonian which combines quantum and phenomenological ones for rare-earth (RR) and Fe subsystem, respectively. Parameter values of RR Hamiltonian were determined by the first-principles. First method adopts a numerical procedure with Boltzmann statistics for the Sm 4f4f electrons. The other one is an analytical method which deals with the magnetic states of RR ions with strong mixing of states with different quantum number of angular momentum JJ (JJ-mixing). We have modified the previous analytical methods for Sm ions which have relatively small spin-orbit interaction, and clarified that they over-estimate the JJ-mixing effects for Sm-transition metal compounds. It has been shown that the results of our analytical method agree with those obtained by statistical method. Our analytical method revealed that the increasing spin angular momentum with JJ-mixing caused by strong exchange field, enhances the absolute value of orbital angular momentum and MA constants via spin-orbit interaction, and that these JJ-mixing effects remain even above room temperature. The calculated results of MA constants show that K1(T)>0K_{1}(T)>0 and K2(T)<0K_{2}(T)<0 in SmFe12 in consistent with experiment.

The peculiar temperature dependence known as first-order magnetization process (FOMP) in SmFe12 has been attributed to the negative K2K_{2}. It was also verified that the requirement for the appearance of FOMP is given as K2<K1<6K2-K_{2}<K_{1}<-6K_{2}. The positive (negative) K1(2)K_{1(2)} appears due to an increase in the crystal field parameter A20r2A_{2}^{0}\langle r^{2}\rangle (A40r4A_{4}^{0}\langle r^{4}\rangle) caused by hybridization between 3d3d-electrons of Fe on 8i8i (8j8j) site and 5d5d and 6p6p valence electrons on Sm. The mechanism of K1>0K_{1}>0 and K2<0K_{2}<0 in SmFe12 has been thus clarified by using the expressions of K1K_{1} and K2K_{2} obtained in the analytical method. Shortly, the sign of K1K_{1} and K2K_{2} in SmFe12 is attributed to the characteristic lattice structure around Sm ions, that is, crystallographic 2bb-sites on cc-axis adjacent to Sm are vacant. We also present results on the magnetization process and nucleation fields by calculating Gibbs free energy.

The present method will be applied to derive a general expression of the free energy to analyze MA of non-uniform systems such as disordered compounds, surfaces, and interfaces. The results will be reported in a forthcoming paper.

Acknowledgements.
This work is supported by ESICMM Grant Number 12016013 and ESICMM is funded by Ministry of Education, Culture, Sports, Science and Technology (MEXT). T. Y. was supported by JSPS KAKENHI Grant Numbers JP18K04678. P. N. was supported by the project Solid21.

Appendix A Magnetization Process in Condition
of K1(T)>0K_{1}(T)>0 and K2(T)<0K_{2}(T)<0

To investigate the magnetization process in equilibrium along the cc-plane (e.g. aa-axis), we introduce the simplified model with magnetic anisotropy constants K1(T)>0K_{1}(T)>0 and K2(T)<0K_{2}(T)<0, which can be expressed by the Gibbs free energy as:

G(x,T,B)=K1(T)x2+K2(T)x4\displaystyle G(x,T,B)=K_{1}(T)x^{2}+K_{2}(T)x^{4}- BMs(T)x\displaystyle BM_{\rm s}(T)x (77)
(|x|1),\displaystyle(|x|\leq 1),

where x=𝑴s(T)𝒏a/Ms(T)x={\bm{M}}_{\rm s}(T)\cdot{\bm{n}}_{a}/M_{\rm s}(T) with total magnetization 𝑴s{\bm{M}}_{\rm s} and unit vector parallel to aa-axis 𝒏a{\bm{n}}_{a}. TT and 𝑩=B𝒏a{\bm{B}}=B{\bm{n}}_{a} (B>0B>0) are temperature and applied magnetic field, respectively. The equilibrium condition is:

G(x0,T,B)=min|x|1G(x,T,B),G(x_{0},T,B)=\min_{|x|\leq 1}G(x,T,B),

where x=x0(T,B)x=x_{0}(T,B) gives minimum of G(x,T,B)G(x,T,B). For K1(T)K2(T)K_{1}(T)\leq-K_{2}(T), the magnetization is always tilted to the aa-axis direction due to x0(T,B)=1x_{0}(T,B)=1. Otherwise, the magnetization curve is given by:

𝑴s(T,B)𝒏a=Ms(T)x0(T,B).{\bm{M}}_{\rm s}(T,B)\cdot{\bm{n}}_{a}=M_{\rm s}(T)x_{0}(T,B). (78)

The first-order magnetization process (FOMP) appears, when x0(T,B)x_{0}(T,B) has two values at certain BB, which is called as FOMP field BFPB_{\rm FP}.

To determine the x0(T,B)x_{0}(T,B) for K1(T)>K2(T)K_{1}(T)>-K_{2}(T), we show the first and second derivative of G(x,T,B)G(x,T,B) with respect to xx as:

G(x,T,B)x\displaystyle\frac{\partial G(x,T,B)}{\partial x} =2K1(T)x+4K2(T)x3BMs(T),\displaystyle=2K_{1}(T)x+4K_{2}(T)x^{3}-BM_{\rm s}(T), (79)
2G(x,T,B)x2\displaystyle\frac{\partial^{2}G(x,T,B)}{\partial x^{2}} =2K1(T)+12K2(T)x2.\displaystyle=2K_{1}(T)+12K_{2}(T)x^{2}. (80)

A inflection point of G(x,T,B)G(x,T,B) for x>0x>0 at fixed TT and BB is given by xc(T)=K1(T)/6K2(T)x_{c}(T)=\sqrt{-K_{1}(T)/6K_{2}(T)}. Hereafter, we consider following two cases: xc(T)1x_{c}(T)\geq 1 and xc(T)<1x_{c}(T)<1.

(i) The case of xc(T)1x_{c}(T)\geq 1

x0(T,B)x_{0}(T,B) is obtained from the condition G(x,T,B)/x=0\partial G(x,T,B)/\partial x=0 for 0<x10<x\leq 1, because 2G(x,T,B)/x2>0\partial^{2}G(x,T,B)/\partial x^{2}>0 is always satisfied. The saturating point of magnetization x0(T,B)=1x_{0}(T,B)=1 is obtained from the condition G(x,T,B)/x|x=1=0\partial G(x,T,B)/\partial x|_{x=1}=0 as:

B=2K1(T)Ms(T)[1+2γ(T)]BA(T),\displaystyle B=\frac{2K_{1}(T)}{M_{\rm s}(T)}[1+2\gamma(T)]\equiv B_{\rm A}(T), (81)

where γ(T)=K2(T)/K1(T)\gamma(T)=K_{2}(T)/K_{1}(T). The BAB_{\rm A} is so-called anisotropy field.

(ii) The case of xc(T)<1x_{c}(T)<1

x0(T,B)x_{0}(T,B) is obtained from the condition

G(x0,T,B)=min[G(xe,T,B),G(1,T,B)],\displaystyle G(x_{0},T,B)=\min\left[G(x_{e},T,B),G(1,T,B)\right], (82)

where xe(T,B)x_{e}(T,B) is determined by the condition of local minium as: G(x,T,B)/x=0\partial G(x,T,B)/\partial x=0 and xe(T,B)<xc(T)x_{e}(T,B)<x_{c}(T). In the magnetization process, x0(T,B)x_{0}(T,B) is continuously increased from zero with increasing BB according to x0(T,B)=xe(T,B)x_{0}(T,B)=x_{e}(T,B) for G(xe,T,B)<G(1,T,B)G(x_{e},T,B)<G(1,T,B). At B=BFPB=B_{\rm FP} such that G(xe,T,B)=G(1,T,B)G(x_{e},T,B)=G(1,T,B) is satisfied, x0(T,B)x_{0}(T,B) shows the abrupt jump and becomes saturated value of xe(T,B)=1x_{e}(T,B)=1. The condition is rewritten as:

(x01)[3K2(T)x02+2K2x0+K1(T)+K2(T)]=0.\displaystyle(x_{0}-1)\left[3K_{2}(T)x_{0}^{2}+2K_{2}x_{0}+K_{1}(T)+K_{2}(T)\right]=0. (83)

By solving the Eq. (83) for 0<x010<x_{0}\leq 1, two minimum points of G(x,T,B)G(x,T,B) with respect to xx are obtained at x0(T,B)=1x_{0}(T,B)=1 and

x0(T,B)=13(1+3γ(T)2)xFP(T).x_{0}(T,B)=\frac{1}{3}\left(-1+\sqrt{-\frac{3}{\gamma(T)}-2}\right)\equiv x_{\rm FP}(T). (84)

By using xFP(T)x_{\rm FP}(T), the field at which the FOMP occurs is determined by:

B=2K1(T)Ms(T)[xFP(T)+2γ(T)xFP(T)3]BFP(T).B=\frac{2K_{1}(T)}{M_{\rm s}(T)}[x_{\rm FP}(T)+2\gamma(T)x_{\rm FP}(T)^{3}]\equiv B_{\rm FP}(T). (85)

As a result, for 1<γ(T)<1/6-1<\gamma(T)<-1/6, the FOMP occurs between 𝑴s(T)𝒏a=Ms(T)xFP(T){\bm{M}}_{\rm s}(T)\cdot{\bm{n}}_{a}=M_{\rm s}(T)x_{\rm FP}(T) and Ms(T)M_{\rm s}(T).

References

  • (1) T. Miyake, K. Terakura, Y. Harashima, H. Kino, and S. Ishibashi, J. Phys. Soc. Jpn. 83, 043702 (2014).
  • (2) Y. Hirayama, Y. K. Takahashi, S. Hirosawa, K. Hono, Scr. Mater. 95, 70 (2015).
  • (3) G. C. Hadjipanayis, A. M. Gabay, A. M. Schönhöbel, A. Martín-Cid,J. M. Barandiaran, and D. Niarchos, Engineering 6, 141 (2020).
  • (4) K. Ohashi, Y. Yokoyama, R. Osugi, Y. Tawara, IEEE Trans. Magn. MAG-23, 3101 (1987).
  • (5) K. Ohashi, Y. Tawara, R. Osugi, and M. Shimao, J. Appl. Phys. 64, 5714 (1988).
  • (6) Bo-Ping Hu, Hong-Shuo Li, J. P. Gavigan, and J. M. D. Coey, J. Phys.: Condens. Matter 1, 755 (1989).
  • (7) T. Kuno, S. Suzuki, K. Urushibara, K. Kobayashi, N. Sakuma, M. Yano, A. Kato, and A. Manabe, AIP Advances 6, 025221 (2016).
  • (8) A. M. Schönhöbel, R. Madugundo, O. Yu. Vekilova, O. Eriksson, H. C. Herper, J. M. Barandiarán, G. C. Hadjipanayis, J. Alloys Compd. 786, 969 (2019).
  • (9) H. Kato, T. Nomura, M. Ishizone, H. Kubota, and T. Miyazaki, J. Appl. Phys. 87, 6125 (2000).
  • (10) Y. Hirayama, Y. K. Takahashi, S. Hirosawa, and K. Hono, Scr. Mater. 138, 62 (2017).
  • (11) D. Ogawa, X. D. Xu, Y. K. Takahashi, T. Ohkubo, S. Hirosawa, and K. Hono, Scr. Mater. 164, 140 (2019).
  • (12) D. Ogawa, T. Yoshioka, X. D. Xu, Y. K. Takahashi, H. Tsuchiura, T. Ohkubo, S. Hirosawa, and K. Hono, J. Magn. Magn. Mater. 497, 165965 (2020).
  • (13) H. Sepehri-Amin, Y. Tamazawa, M. Kambayashi, G. Saito, Y. K. Takahashi, D. Ogawa, T. Ohkubo, S. Hirosawa, M. Doi, T. Shima, and K. Hono, Acta Materialia 194, 337-342 (2020).
  • (14) R. F. L. Evans, W. J. Fan, P. Chureemart, and T. A. Ostler, and M. O. A. Ellis, and R. W. Chantrell, J. Phys.: Condens. Matter 26, 103202 (2014).
  • (15) Y. Toga, M. Matsumoto, S. Miyashita, H. Akai, S. Doi, T. Miyake, and A. Sakuma, Phys. Rev. B 94, 174433 (2016).
  • (16) M. Nishino, Y. Toga, S. Miyashita, H. Akai, A. Sakuma, and S. Hirosawa, Phys. Rev. B 95, 094429 (2017).
  • (17) T. Miyake and H. Akai J. Phys. Soc. Jpn. 87, 041009 (2018).
  • (18) T. Yoshioka and H. Tsuchiura, Appl. Phys. Lett. 112, 162405 (2018).
  • (19) H. Tsuchiura, T. Yoshioka, and P. Novák, Scr. Mater. 154, 248 (2018).
  • (20) M. Yamada, H. Kato, H. Yamamoto, and Y. Nakagawa, Phys. Rev. B 38, 620 (1988).
  • (21) R. Sasaki, D. Miura, and A. Sakuma, Appl. Phys. Express 8, 043004 (2015).
  • (22) D. Miura, R. Sasaki, and A. Sakuma, Appl. Phys. Express 8, 113003 (2015).
  • (23) D. Miura and A. Sakuma, AIP Advances 8, 075114 (2018).
  • (24) M. D. Kuz’min, Phys. Rev. B 46, 8219 (1992).
  • (25) M. D. Kuz’min and J. M. D. Coey, Phys. Rev. B 50, 12533 (1994).
  • (26) M. D. Kuz’min, Phys. Rev. B 51, 8904 (1995).
  • (27) Y. Millev and M. Fähnle, Phys. Rev. B 52, 4336 (1995).
  • (28) M. D. Kuz’min, J. Appl. Phys. 92, 6693 (2002).
  • (29) N. Magnani, S. Carretta, E. Liviotti, and G. Amoretti, Phys. Rev. B 67, 144411 (2003).
  • (30) Y. Harashima, K. Terakura, H. Kino, S. Ishibashi, and T. Miyake, JPS Conf. Proc. 5, 011021 (2015).
  • (31) W. Körner, G. Krugel, and C. Elsasser, Sci. Rep. 6, 24686 (2016).
  • (32) P. Delange, S. Biermann, T. Miyake, and L. Pourovskii, Phys. Rev. B 96, 155132 (2017).
  • (33) J. H. Van Vleck, The Theory of Electric and Magnetic Susceptibilities (Oxford University Press, Oxford, 1932).
  • (34) S. G. Sankar, V. U. S. Rao, E. Segal, W. E. Wallace, W. G. D. Frederick, and H. J. Garrett, Phys. Rev. B 11, 435 (1975).
  • (35) H. W. de Wijn, A. M. van Diepen, and K. H. J. Buschow. phys. stat. sol. (b) 76, 11 (1976).
  • (36) H. Sepehri-Amin, T. Ohkubo, T. Nishiguchi, S. Hirosawa, K. Hono, Scr. Matter 63, 1124 (2010).
  • (37) H. Sepehri-Amin, T. Ohkubo, S. Nagashima, M. Yano, T. Shoji, A. Kato, T. Schrefl, and K. Hono, Acta Materialia 61, 6622-6634 (2013).
  • (38) P. Blaha, K.Schwarz, G. Madsen, D. Kvasnicka, and J. Luitz, WIEN2k, An Augmented Pla ne Wave + Local Orbitals Program for Calculating Crystal Properties, Karlheinz Schwarz, TU Wien, Austria, 2001, ISBN 3-9501031-1-2.
  • (39) M. Richter, P. M. Oppeneer, H. Eschrig, and B. Johansson Phys. Rev. B 46, 13919 (1992).
  • (40) P. Novák, Phys. Stat. Sol. B 198, 729 (1996).
  • (41) K. Hummler and M. Fähnle, Phys. Rev. B 53, 3272 (1996).
  • (42) M. Richter, J. Phys. D 31, 1017 (1998).
  • (43) M. Diviš, K. Schwarz, P. Blaha, G. Hilscher, H. Michor, and S. Khmelevskyi, Phys. Rev. B 62, 6774 (2000).
  • (44) M. Diviš, J. Rusz, H. Michor, G. Hilscher, P. Blaha, and K. Schwarz, J. Alloys Compd. 403, 29 (2005).
  • (45) M. Brooks, L. Nordström, and B. Johansson, J. Phys. Condens. Mater 3, 3393 (1991).
  • (46) M. T. Hutchings, Solid State Phys. 16, 227 (1964).
  • (47) R. J. Elliott, Magnetic Properties of Rare Earth Metals (Plenum, New York, 1972), Chap. 1.
  • (48) M. D. Kuz’min, Phys. Rev. Lett. 94, 107204 (2005).
  • (49) S. A. Nikitin, I. S. Tereshina, V. N. Verbetskiǐ, and A. A. Salamova, Phys. Solid State 40, 258 (1998).
  • (50) A. R. Edomonds, Angular Momentum in Quantum Mechanics (Princeton University Press, Princeton, N. J., 1960)
  • (51) D. A. Varshalovich, A. N. Moskalev, and V. K. Khersonskii, Quantum Theory of Angular Momentum (Wold Scientific, Singapore, 1988).
  • (52) J. R. Schrieffer and P. A. Wolff, Phys. Rev. 149, 491 (1966).
  • (53) K. Stevens, Proc. Phys. Soc. A65 209 (1952).
  • (54) H. Kronmüller, K. D. Drust, and M. Sagawa, J. Magn. Magn. Mater. 69, 149 (1987).
  • (55) H. Kronmüller and M. Fähnle, Micromagnetism and the Microstructure of Ferromagnetic Solids (Cambridge University Press, Cambridge, 2003)
  • (56) H. Tsuchiura,T. Yoshioka, and P. Novák, IEEE Trans. Magn. 50, 2105004 (2014).
  • (57) R. Coehoorn, K. H. J. Buschow, and M. W. Dirken, and R. C. Thiel, Phys. Rev. B 42, 4645 (1990).
  • (58) A. Sakuma, J. Phys. Soc. Jpn. 61, 4119 (1992).
  • (59) H. Adachi, H. Ino, and H. Miwa, Phys. Rev. B 56, 349 (1997).
  • (60) H. Adachi, and H. Ino, Nature 401, 148 (1999).
  • (61) T. S. Zhao, X. C. Kou, R. Grössinger, and H. R. Kirchmayr, Phys. Rev. B 44, 2846 (1991).