応用物理学会学術講演会 神奈川大学,2003年3月29日 12半導体B,12.1探索的材料物性 29p-ZD-1 近藤半導体とは何か 埼玉大学理学部物理学科 佐宗哲郎 近藤半導体は普通の半導体と何が違う か? 近藤半導体は役に立つか? 1. 近藤絶縁体とは 2. 近藤絶縁体のエネルギー・ギャップの形成機構 3. 近藤絶縁体における多体効果 4. 近藤絶縁体の強磁場誘起絶縁体-金属転移 5. 近藤絶縁体の熱電効果 1 4f and 5f electrons in rare earth and actinide ions under CEF and spin-orbit interaction L=3, S=1/2 J=5/2, 7/2 due to spin-orbit interaction Ce: J=7/2 4f ~7000K ~10~100K J=5/2 spinorbit 7 8 CEF Yb: J=7/2 is lower than J=5/2 8 states with J=7/2 under cubic CEF: 8 (1) 8 ( 2) 7 3 5 1 5 0 Y3 Y3 Y3 12 28 21 3 2 1 3 2 15 1 Y3 Y3 Y33 Y3 14 14 28 28 2 電子相関の役割 金属の自由電子模型 電子比熱 C/T Cel T , 2 3 T2 D( F )k B , D( F ) 2 Pauli B 2 D( F ) スピン帯磁率 T 1 2m 2 2 2 3/ 2 常磁性! 自由を奪うもの: 散乱 緩和時間 k 1 電気抵抗 0 2 | V (k k ' ) |2 D( F ) m ne 2 0 T 自由を奪うもの: 相互作用 遍歴性と局在性 s電子 遷移金属 d電子 アクチナイド金属 5f電子 希土類金属 4f電子 > アルカリ金属 > > 磁気モーメントの発生 (磁性は電子相関から生じる) U=0の時 EF U>0 Ed+U EF Ed Ed 3 4 近藤効果 希薄磁性合金における 電気抵抗極小の現象 R(T ) R0 AT 5 cJ log(T / TK ) TK:近藤温度, c:不純物濃度 T s-d模型 H s d k ck ck J k S s ' ck ck ' ' kk ' ' J. Kondo (1964) 3次摂動によるスピン反転 散乱過程 TK De1/ J 伝導電子 局在スピン Ed+U 近藤模型とAnderson模型の等価性 H A k c k ck Ed nd Und nd k (Vk ck d V * k d ck ) EF Ed 8V 2 H s d ( J ) U k 希土類不純物による近藤効果 Ce: 4f1, L=3, S=1/2, J=5/2, Jz=-5/2,・・・,5/2, N=6重縮退 Yb: 4f13, L=3, S=1/2, J=7/2, Jz=-7/2,・・・,7/2, N=8重縮退 5 Universal Behaviors in Kondo Effect 帯 磁 率 (T ) (T ) T(T) 電 気 抵 抗 C T TK TK R(T) C/T C(T)=T+AT3 比 熱 熱 電 能 Cel(T) T TK S(T) log T 0.01TK TK 100TK -log(T/TK) 6 T/TK 近藤絶縁体とは・・ 電子相関の強いバンド絶縁体である。 ・帯磁率 χ YbB12 高温でCurie-Weiss則 低温で非磁性 FeSi ・比熱 T C/T Egap<T<TKで C=T+AT3, 0 ・電気抵抗 ρ Egap T2 ~ e Egap/T 低温で絶縁体 T •低温で,狭いギャップ(数十~数百meV)の絶縁体。 •フェルミ準位に電子がないにもかかわらず近藤効果 的ふるまい(?)を示す。 •電子数偶数個で立方晶の4f ,5f化合物 (YbB12, Ce3Bi4Pt3など)に見られる。 •d電子系でも類似 (FeSiなど)。 ただし,χ(T=0)→0 •CeNiSnは斜方晶で擬ギャップ 7 •TmSeは電子数奇数個で,Mott絶縁体,TN=5K Heavy Fermions with Energy Gap (Kondo Insulators) high (mJ/K 2mole) C/(0) (K) SmS 総 電 子 数 78 SmB6 92 80 YbB12 130 80 TmSe 103 Ce3Bi4Pt3 740 Ce3Sb4Pt3 612 CeRhSb 154 CeRhAs 146 60 Tmax(K) TK(K) 81 320 80 46 320 70 small 23 320 35 113 TN~3K 4 144 CeFe4P12 CeNiSn Eg (K) 1500 136 190 92 12 2~5 U3Ni3Sb4 230 U3Pd3Sb4 260 U3Pt3Sb4 170 UFe4P12 pseudogap Tc=3.15K UNiSn 170 800 TN=45K ThNiSn 168 800 no order Sm3Se4 322 790 Sm3Te4 394 600 1600 TN=0.7K Heavy Fermions with low carrier density 電子数 high C/(0) Tmax TK nc/R E CeN CeP CeAs TN=8K 20 CeBi Yb4As3 205 92 no peak 0.05 TN=6K 0.05 TN=25K 0.001 T0~300K 8 近藤絶縁体を記述する模型 •近藤格子模型 (U=∞) •周期アンダーソン模型 (U=0 → ∞)と その一般形(2-バンド模型など) ○バンド計算でギャップが開いていて,そこに多体効 果が加わっている。 ○複雑な結晶構造や軌道縮退のために,ギャップが 開かないことも有る。 ○多体効果とスピンのゆらぎの効果は,連続したも のである。 ○近藤効果がきれいに見えるとは限らない。 k Ef k Ef 9 No gap! 近藤絶縁体を記述する模型 •近藤格子模型 (U=∞) •周期アンダーソン模型 (U=0 → ∞)と その一般形(2-バンド模型など) ○バンド計算でギャップが開いていて,そこに多体効 果が加わっている。 ○複雑な結晶構造や軌道縮退のために,ギャップが 開かないことも有る。 ○多体効果とスピンのゆらぎの効果は,連続したも のである。 ○近藤効果がきれいに見えるとは限らない。 k Ef k Ef 10 No gap! Electronic structure of the Kondo Insulator YbB12 Harima and Yanase (1992) 4f J=7/2 J=5/2 4f J=7/2 J=5/2 11 LDA+U → Eg~0.01eV 4f J=7/2 8 7,6 J=5/2 2002/03/15 15:07:02 fcc2.dat T. Saso and H. Harima(2003) Nearly-Free-Electron model? Empty Band for fcc lattice (a=5.65A) cannot reproduce the band. Energy (Ryd) 2 12 Saso (C)T. 0 K X W L X Tight-binding band for YbB12 Yb: 5d t2g band 5d (xy, yz, zx) Effective d-d hopping through B12 clusters + - + + - - 2px±2py B12 Yb 5dxy 13 Construction of d-f matrix elements CEF and s.o. 8 states with J=7/2 under cubic CEF: 8 (1) 8 ( 2) 7 3 5 1 5 0 Y3 Y3 Y3 28 21 12 3 2 1 3 2 15 1 Y3 Y3 Y33 Y3 14 14 28 28 Slater-Koster integrals by K. Takegahara, et al., J. Phys. C13(1980)583.---- CEF but no s.o. Conversion of Ylm into l=3 cubic harmonics yields: 1 5 x(5 x 2 3r 2 ) i y (5 y 2 3r 2 ) 2 21 8(1) [ 1 9 5 x ( y 2 z 2 ) i y ( z 2 x 2 ) ] z (5 z 2 3r 2 ) 2 7 21 8( 2 ) [ 1 5 x(5 x 2 3r 2 ) i y (5 y 2 3r 2 ) 2 7 1 3 3 x ( y 2 z 2 ) i y ( z 2 x 2 ) ] z( x 2 y 2 ) 2 7 7 Slater-Koster integrals d-f mixing matrix elements: 8(1) xy yz zx xy yz zx 8(1) 8(2) 8( 2) 0 0 it 2 (c x s y is x c y ) 5it1 (cx s y is x c y ) 4t1c y s z it1c y s z 3it 2 s y cz 2t 2 c y s z 4it1s z cx it1c z sx 3it 2 c z s x 2it 2 s z c x 0 5it1 (cx s y is x c y ) it2 (cx s y is x c y ) 0 it1s y c z 4t1c y s z 2t 2 c y s z 3it 2 s y c z it1cz s x 4it1s z cx 2it 2 s z c x 3it 2 cz s x t1 14 5 15 (df ), t2 (df ), s sin(k / 2), c cos(k / 2), x, y, z 56 56 0.9 d,d bands x2-y2, 3z2-r2 Energy (Ryd) (dds) only xy, yz, zx 0 K X W L X 2002/08/10 14:17:26 dxy-band2.dat d-only bands Ek ( xy ) xy, yz, zx Ek ( yz ) Ek ( zx ) •Three independent bands •Two-dimensional bands ky kx ) cos( ) 2 2 ky k 3(dd ) cos( ) cos( z ) 2 2 k k 3(dd ) cos( z ) cos( x ) 2 2 3(dd ) cos( E •Conduction band at X 0.3 point is doubly degenerate. 0 4f 8 15 -0.3 K X W L X (dd)-(df)-(ff) tight-binding band (dd)=0.06, (df)=0.01, (df)=-0.005, (ff)=-0.002, LDA+U shift DE=-0.005 (in Ryd) Energy(Ryd) 1.2 2002/09/14 16:39:45 YbB12-dos.dat Eg=0.003 Ryd 0.8 K X W L X 400 YbB12 DOS 非対称! 0 0.8 1 16 Mechanism of the gap opening Besides the Kramers degeneracy, No gap! Gap opens! or Semimetal! No gap! 17 Gap opens! Semimetal! d-d model: xy yz Ekxy 0 0 0 3t s x s y xy yz zx x2 y2 3z 2 r 2 0 V k V V Ef V 0 0 1.2 V V 0 E f Energy (Ryd) 0 3z 2 r 2 3t s x s y 3 3 Ekyz 0 t s x s y t s x s y 2 2 2002/10/29 12:28:02 dd-band05.dat 3 3 0 Ekzx t s x s y t s x s y 2 2 3 3 3 t s x s y t s x s y ts x s y 0 2 2 2 3 3 3 t s x s y t s x s y 0 ts x s y 2 2 2 0 Constant mixing model: k 0 V V x2 y 2 zx 0.8 K X W L X Separate mixing model: k 0 V 0 0 V k 0 0 Ef V 0 0 k V V 0 0 E f 0 V 0 Ef 0 0 k 0 V 0 0 V E f 18 Summary •New LDA+U band calculation was performed for the most typical Kondo insulator YbB12. The gap (Eg=0.0013Ryd) opened after additional shift of 4f level by 0.3Ryd. •The conduction bands can be expressed very well by the simple tightbinding model (t2g band with effective d-d hopping matrix (dd) through B12 cluster). It is impossible to express it by the nearly free-electron model. •Introduction of the d-f hybridization (df) opened up the energy gap if the filled band is pulled down slightly. •Relationship of the gap-opening and the degeneracy was discussed. •The CEF ground state must be 8, since otherwise a gap does not open!! Ref. T. Saso and H. Harima, to appear in J. Phys. Soc. Jpn. 71 (2003) No.5 Future problems •Explicit inclusion of B12 clusters may improve the dispersion curves. •Explicit separation of the Kramers degeneracy may provide alternative representation. (see Maehira, et al.) •Based on the present model, we have to take into account the correlation effect ( via e.g. DMFT?), and calculate the physical 19 quantities. FeSi B-20 structure •Magnetic susceptibility (Jaccarino, et al., 1967) vanishes at low T due to the gap, and exhibits Curie law at high T with a peak at 500K •Magnetic specific heat (Kanomata, et al., 2000) has a peak at 220K. •Optical conductivity (Damascelli, et al., 1997) has a gap of 570 cm-1 (810K) at low T, but it is filled up at 300K 20 •Band calculation (Kulatov, et al.) gives an indirect gap of 0. 04eV and direct gap of 0.16eV. •Band calculation (Ohta, et al., 1994) gives an indirect gap of 0. 02eV. ( ) d ( ) ( ) f ( )[1 f ( )] +Drude term Correct formula: ( ) d ( ) ( ) f ( ) f (21 ) Spin-Fluctuation (SCR) theory by Takahashi and Moriya (1979) Sum rule for the dynamical susceptibility: 3 0 0 (q, ) 2 d coth( ) Im Si 2T q 1 U 0 (q, ) 0 (0,0) 1 U 0 (0,0) (T ) is treated as a parameter to be determined at each T. Takahashi (1997) 3 2 0 d[1 2n( )] Im q 2 0 (q, ) 3 (n 2 n n ) 1 U 0 (q, ) 4 For correlated metal: Si (U ) For an insulator (Eg>0): 0(q,) is calculated from the T-independent band model with a constant gap. •This theory can explain (T) and C(T) qualitatively, but not (,T). 2 3 n (T E g , n 1) 4 22 FLEX (Fluctuation Exchange) •single-particle self-consistent theory •weak coupling theory 1 G(k , i ) i k (k , i ) 1 U (k , i ) Un T G (k q, i i ) 2 2 q 3U (q, i ) U (q, i ) 2 (q, i ) 1 U (q, i ) 1 U (q, i ) (q, i ) T G(k , i )G(k q, i i ) k U→0 1 (k , i ) Un U 2T (q, i )G(k q, i i ) 2 q This is SC-SOPT self-energy. The lower and upper Hubbard bands are not reproduced! Thus, FLEX cannot be used for strong correlation! 23 バンド計算の状態密度を用いた 2-band Hubbard 模型 H (tij1 ci1 c j1 tij2 ci2 c j 2 ) ij U (ni1 ni1 ni 2 ni 2 ) i U 2 (ni1 ni 2 ni 2 ni1 ) バンド内 バンド間 (反平行スピン) i U 3 (ni1 ni 2 ni1 ni 2 ) i J (ci1 ci1 ci2 ci 2 1 2) バンド間 (平行スピン) 交換相互作用 i (1,2 はband index) Parameters: U=0.5eV,J=0.35U, U2=U-J, U3=U-2J Band calculation at T=0 by Yamada et al., J. Phys.: Condens. Matter 11 (1999) L309, but the gap is 24 enlarged by 16%. (Self-Consistent Second-Order Perturbation Theory in d→∞) SC-SOPT 1 11 (i ) Un U 2T 2 G22 (i ' )G22 (i 'i )G11 (i i ) 2 ', G11 (i ) k 1 i k 11 (i ) 2 ( 2) Σ11 ( ) U U 2 2 2 + 2 + U3 U2 U2 2 2 2 U3 + J J Density of states: ( ) 1 ImG11 ( i ) G22 ( i ) Optical conductivity: 2 e 2 2 ( ) (Assume breaking of momentum conservation) d ( ) ( ) f ( ) f ( ) (Good for interband transition) Density of states and (,T) become strongly T-dependent. 25 Optical conductivity of FeSi by SCSOPT and comparison with experiment K.Urasaki and T.S.,JPSJ 68(1999)3477. Exp. data taken from A. Damascelli, et al., Physica B 230-232 (1997) 787. •The gap (~700K) is filled up at 300K. •The gap is filled rather slowly at T<100K, but is filled faster at T>150K. •The gap edge shifts to low when T rises. •Rigid band model cannot reproduce the correct temperaturedependence. 26 電気伝導度の公式(d→∞) d→∞では,バーテックス 補正は効かない。 2e 2 f (T ) d L ( )( ) 3 厳密に言うと,d→∞では, vk2~O(1/d)→0 2 1 2 L( ) k ImG(k , ) N k G (k , ) 1 1 k ( ) k i 2 ( ) d→∞では, (k,) はkによらない。 1 2 ( ) 2 L( ) d ' ( ' ) ( ' ) 1 2 2 ( ' ) ( ) 2 ( ) ( ) 2 ( ) ( ) よって,Boltzmann方程式と同じ形に書ける: 2e 2 f 2 (T ) d ( ) ( ) ( )( ) 3 2 注意: 周期系では Umklappが必要 光学伝導度の公式(d→∞) 2e 2 f ( ) f ( ) ( ) d L( , ) 3 1 2 L( , ) k Im G (k , ) Im G (k , ) N k ( ) 2 ( ) d k Im G (k , ) Im G (k , ) 1 ( ) 2 ( ) Im R A ( ) ( ) 0 ( ) 2 ( ) ( ) 27 Specific heat of FeSi E 1 d f ( ) Im ( ) G( ) 2 2 E CV (T ) T V K. Urasaki and T. Saso, Correlation Effects in Multi-Band Hubbard Model and Anomalous Properties of FeSi, ``New Properties of Matter due to Ordering and Fluctuation of Electron Orbitals--Comprehensive Study of f- and dElectrons--'' News Letter Vol.1 No.2 (2000) p.83. •Peak at T=200~300K agrees with experiments. Kanomata, et al. (2000) 28 Magnetic susceptibility of FeSi The susceptibility can be obtained by 1. Solving the Bethe-Salpeter equation 2. Numerical derivative of the electron number. n n h 0 h lim •Peak position is too low compared to the experiments. •Curie behavior at high T is not well reproduced. •Inclusion of spin fluctuation is necessary. 29 YbB12 Sugiyama(1988) Yb2.9+ 30 Yanase-Harima (1992) YbB12 Optical conductivity Okamura (1998) Anisotropy of magnetization curves in YbB12 F. Iga, et al. (1998) •YbB12 has a cubic structure, so that the linear susceptibility must be isotropic. •Anisotropy may appear in non-linear regions. 31 動的分子場理論(d=∞理論) 自己エネルギーが局所的: ij () d ()ij ~ G ( ) 1 G ( ) 1 ( ) を無摂動Green関数とする 1不純物問題と同じ ~ G( ) : 中心サイト以外の効果を取り入れ た(cavity) Green関数 U •局所電子相関がfullに考慮されている。 •電子系に対する最良の1サイト理論。 有効不純物問題の解法 厳密対角化法 数値繰り込み群 量子モンテカルロ法 NCA, slave boson 改良反復摂動論 (mIPT) ~ ()を G ( ) を用いて,U2まで計算して繰り返す。 U→∞にも適用できる。(Mott転移) 32 7 300K Model 6 150K Yb3+ (J=7/2) 8 0K Periodic Anderson model (N=2) H tij ci c j E f f i f i ij i Vij (ci f j H .c.) U n fi n fi ij i E f E f h, h g f B H | | J z | | g f gc → (T=0) > 0 1 (We neglect gc since gf>gc.) Methods Dynamical Mean Field Theory (exact in d→∞) Solution of the impurity problem •Exact diagonalization of small clusters •Iterative perturbation theory (IPT) Exact in zero conduction band width E. Lange, cond-mat/9810208 Self-Consistent Second-Order Perturbation Theory (SCSOPT) 33 Field-induced insulator-to-metal transition (FIIMT) in Kondo insulators Two effects of magnetic field on Kondo insulators (1) Zeeman effect→ ↑ and ↓ bands are shifted Gap decreases. (2) Suppression of the Kondo effect → renormalization disappears. Gap increases. Which dominates? Answer Renormalization factor scarecely changes when there is a gap. Gap is closed mainly by the Zeeman effect. (Note that the DOS shape changes.) 34 M(H)/M() Magnetization curves 1 0.8 SCSOPT DMFT 0.6 EG(H).SMP 3-17-1999 19:54 0.4 U/D=2 U/D=4 0.2 0 0 0.2 0.4 0.6 0.8 1 H/TK Gap closing 0.3 0.2 Eg U=0 (Scaled at h=0) Rigid Band Eg(0)-2h 0.1 SCSOPT (U=2) 0 0 0.1 h 0.2 35 H. Yamada, Phys. Rev. B47(1993)11211. DF ( M ) 1 1 1 aM 2 bM 4 cM 6 2 4 6 Stoner theory +spin fluctuation (T>0) (T=0) E (m) d ( ) Un n n d ( ), E(h) E(m) mh () 36 Seebeck Coefficient S Metals: hole-like 2 k B2 L' ( ) S (T ) T 3 | e | L( ) T L( ) vc ( )2 c ( ) c ( ) electron-like S Semiconductors: S (T ) hole-like k B Ec, | e | k BT T electron-like As a portable refrigerators: Figure of Merit Z= S2/k k: thermal conductivity Power factor P=S2 Ordinary semiconductors: (T)=n(T)e n(T)~exp[-(Ec-)/kBT] P=S2 e.g. Bi2Te3 S2 Eg~0.15eV T ZT~1 at 300K 37 |S|max=2kB/e=173V/K at kBT=(Ec-)/2 Kelvin’s relations W. Thomson (Lord Kelvin), Proc. Roy. Soc. Edinburgh (1851) 91. Thermodynamic arguments : a unit charge is adiabatically moved through ABCDA. T+DT T0 T T0 metal b A metal a C B metal a D S ab DT ab (T DT ) ab (T ) ( b a )DT 0, ab (T DT ) ab (T ) b a DT 0 T DT T T Seebeck coefficient S(T), Peltier coefficient (T), Thomson coefficient (T) T (T ' ) 0 T' S (T ) T (T ' ) 0 T' dT ' (T ) T : Kelvin’s relations Contradictions: S (T ) dT ' , S (T ) S(T) should →0 at T=0, but S(T)→1/T in semiconductors?? (T ) TS (T ) ( Ec,v ) / e Π=finite even at T=0 !? One can remove finite heat at T=0?? Contradictory to 3rd law of thermodynamics !? 38 Modification of Kelvin’s argument at T→0 for insulatos: T+DT T0 |e| T T0 C Semicond. a D Semicond. b A Semicond. a B E e- Conduction band Ecb T+DT Eca T S ab DT ab (T DT ) ab (T ) ( b a )DT W 0, ab (T DT ) ab (T ) b a DT 0 T DT T T Q+W W A work necessary to push up a charge from Eca to Ecb. W (Ec Ec ) / | e | b a Modified Kelvin’s relations: T2=T Q T1=0 (valid only at T→0) ab (T ) Ec Ec S ab (T ) , T | e|T b a T a (T ' ) b (T ' ) 0 T' S ab (T ) Ec Ec dT ' | e|T b a Thus, E Ec Sab (T ) c | e|T b a but Πab(T)→0, T a (T ' ) b (T ' ) 0 T' dT ' 0 Contradiction to the 3rd law of the thermodynamics is resolved! 39 •At low but finite T, nonequilibrium effect must be included! Correct behaviors due to Modified Kelvin Relations: S S T T T T Many-Body Effect or Non-Stoichiometry S T 40 Heat current carried by the strongly correlated electrons G. D. Mahan, Solid State Physics 51 (1998) 81. Hubbard: jQ ( k ) c c k jQ PAM: k k k S k i ( titj )R ij ci c j Vt ij R ij (ci f j H .c.) 2 ij ij M. Jonson and G. D. Mahan, PRB42(1990)9350. Linear Response Theory e2 ni n j Uk ci c j 2 k d f [ F ( i , i ) Re F ( i , i )], e d f ( )[F ( i , i ) Re F ( i , i )], T F ( , ' ) k ck ( )ck ( ' ) j (0) k For detailed analysis and application to High Tc’s, see: H. Kontani, J. Phys. Soc. Jpn. 70 (2001) 2840; cond-mat/0206501, cond-mat/0204193 Many-Body Effect in terms of DMFT for PAM For Kondo insulators: •Self-energy () ≠0 even in the gap when T>0. •Quasi-particle DOS () ≠0 even in the gap at T>0 and is Tdependent due to (). •Thus, Seebeck coefficient S(T) ∝T, ~1, at low T. 41 Kondo Insulators Periodic Anderson Model (N=2) +DMFT 500 U=2 D=0.5 (T) [arb. unit] 100 50 Ef=-1 Ef=-0.8 Ef=-0.7 Ef=-0.6 Ef=-0.5 Ef=-0.3 Ef=0 10 5 1 0 0.2 0.1 T [K] S(T) [V/K] 600 Ef=-1 Ef=-0.8 Ef=-0.7 Ef=-0.6 Ef=-0.5 Ef=-0.3 Ef=0 400 200 0 0 0.1 T 0.2 42 Kondo insulators: S(T) ∝T, ~1, because ()≠0 at T>0. Yb1-xLuxB12 phonon drag? 43 F. Iga, et al. 2-Band model Using total DOS from the band calculation 1000 Density of States H. Harima 500 YbB12 Total f-part 4f 8 EF 0 0.86 0.87 0.88 0.89 (Ryd) 0.9 0.91 Small carrier (electron) doping: 200 S(T) (V/K) 100 YbB12 U=0 Eg=63K 0 Exp. n=0.01 n=0.005 n=0 -100 -200 100 200 T [K] 300 T. Saso (2001) 44 Temperature-dependence of the Kondo peak may be important. Temperature-dependence of DOS of PAM with N=2 1.5 U=2, Ef=-1.2 f () 1 c 0.5 0 T=0 T=1 -4 -2 0 2 Kondo peak disappears. 4 U=2, Ef=-1.5 1.5 () f 1 T=0 c 0.5 0 -4 -2 T=1 0 2 Kondo peak disappears. 4 45 Comparison with the rigid-band model (DOS is fixed to DOS at T=0) 100 S(T) [V/K] 0 -100 rigid band -200 -300 Due to temperature-dependence of DOS PAM, U=2, Ef=-1.2 -400 -500 0 0.1 X 0.2 S(T) [V/K] 0 rigid band -100 Due to temperature-dependence of DOS PAM, U=2, Ef=-1.5 -200 0 0.1 0.2 0.3 T 0.4 0.546 FeSi Sales, et al. PRB50(1994) 8207 Jarlborg (1999) Jarlborg (1995) 47 Band DOS (H. Yamada) + SCSOPT(d=∞) FeSi FeSi EF 4000 DOS (T) (Wcm) 6000 2000 -1000 0 0 100 0 1000 200 300 T 600 -4 1x10 holes (U=0) -4 1x10 holes (U=0.5 eV) Exp. S [V/K] 400 200 FeSi 0 -200 0 100 200 T [K] 300 48 Summary •Kelvin’s relations must be modified at low temperature limit of insulators and semiconductors. The contradiction to the 3rd law of the thermodynamics is resolved by the new formula. •In reality, the effect will be smeared by nonstoichiometry and manybody interaction. •The energy bands of the most typical Kondo insulator YbB12 can be expressed very well by the simple tight-binding model (t2g band with effective d-d hopping matrix (dd) through B12 cluster, and (df) mixing). It is impossible to express it by the nearly free-electron model. •The gap was reproduced by a realistic band model for the first time (after LDA+U correction). Relationship of the gapopening and the degeneracy was discussed. •The CEF ground state must be 8, since otherwise a gap does not open!! •The temperature-dependence of the quasi-particle density of states is indispensable to understanding the thermopower. It cannot be understood by the rigid band model. Future problems •Based on the present model, we have to take into account the correlation effect, and recalculate the thermopower, etc.. •Band anisotropy may be important for a comparison with experiment 49 on the thermopower. Summary •近藤絶縁体は,相関の強いバンド絶縁体である。 •電子相関とスピンのゆらぎの効果とは,連続的につな がったものである。 dとfも連続している。 •状態密度とギャップは,多体効果のため,強く温度依存 する。 •相互作用のバンド間成分の大きさで,温度依存性が異 なり得る。 •光学伝導度のギャップは,さらに強く温度依存する。 (K.Urasaki & TS, JPSJ 68(1999)3477) •σ(,T)の記述には,自己無撞着な計算が必要。(SCSOPTなど。) •バンド計算の状態密度を反映した,正しいσ(,T)が必 要。 •χ(T)にはスピンの揺らぎが大事。 •S(T)にはdopingが大事。 •動的分子場理論とスピンの揺らぎの理論の統合理論 が可能。(TS,JPSJ 68 (1999) 3941; 69 (2000) No.12; J.Phys. (2001)) •問題点:バーテックス補正 50 まとめ •近藤半導体は,多体効果の強いバンド絶縁体(半導 体)である。 •例:YbB12,FeSi,CeRhSn,etc. •多体効果が強いために,バンドギャップが近藤温度 程度の小さな値になる場合がある。 •多体効果が強いために,バンドギャップ(と状態密度 の形)が温度変化する。 •ギャップの成因は,基本的には半導体と同じである ので,バンド計算で決まる。ただし,多体効果が強い 場合,バンド計算に問題が生じる。また,希土類化合 物では,スピン-軌道相互作用と結晶場分裂を正しく 扱うことが必要である。 •近藤半導体は役に立つか? 低温で性能のよい熱電 素子となる可能性は十分にある。 51
© Copyright 2024 ExpyDoc