a sum rule for thermal conductivity and dynamical thermal...
TRANSCRIPT
![Page 1: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/1.jpg)
A Sum Rule for Thermal Conductivity and Dynamical
Thermal Transport Coefficients in
Condensed Matter
Sriram ShastrySriram ShastryUCSC, Santa Cruz, CAUCSC, Santa Cruz, CA
Work supported by
DMR 0408247Seminar @ UC Davis, 9 Feb, 2006
![Page 2: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/2.jpg)
ABSTRACT
New ideas about reliable computations of thermoelectric transport coefficients of Strongly Correlated matter are discussed. The dynamical thermal conductivity κ(ω) and dynamical thermopower play a central role, together with the optical conductivity.
Quite surprisingly, a sum rule for thermal conductivity κ(ω) is found, in perfect analogy with the f-sum rule for σ(ω). The new formalism gives interesting insights into the triangular lattice thermopower, which is of current interest in the sodium cobaltates.
![Page 3: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/3.jpg)
Plan of the talk
•Motivation
•Hall Constant at High Frequency and the method of high frequencies
•Sum rule for thermal conductivity
•Thermopower and Lorentz number
•Cobaltates and some predictions beyond Heikes formula
•Numerical results for Hall and thermopower on small clusters using exact diagonalization.
•Prospects for further work
![Page 4: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/4.jpg)
Motivation:
Experiments:
Na x Co O2 :Happy symbiosis of basic theory and technology.
• Large Thermo power
•Huge B dependence of Thermopower
•Strange Hall constant
Skutterudites:
•Cages and Rattlers; or how to manipulate lattice thermal conductivity
Heavy Fermi Systems
Theory:
•Updating Boltzmann theory: long lived almost free quasi particles are not a good starting point for most of these materials.
•Effect of strong correlations on transport.
•Understanding Mott Hubbard physics: what are holes?
![Page 5: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/5.jpg)
Hall Effect in Strongly Correlated MatterHALL CONSTANT and what is a hole?
Question: What is “n” for a Mott Hubbard system? Electron number of hole number ( measured from half filling)?
Real space versus k space!!
Standard expression says that Hall constant is a measure of carrier concentration:
RH =1
−|e|ncNaïve expectation from non interacting picture
0.5 1 1.5 2
-75
-50
-25
25
50
75
![Page 6: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/6.jpg)
0 0.250.50.75 1 1.251.51.75
-4
-2
0
2
4 Correlated theory for Hall constant: SSS (1993)
Behaviour for square lattice Mott Hubbard system. Also expected for triangular lattice at low T ( work in progress). Notice there are THREE zero crossings
t<0 Triangular lattice at T> |t| is always hole like. No zero crossings in either case.
0.25 0.5 0.75 1 1.25 1.5 1.75
-80
-60
-40
-20
0.25 0.5 0.75 1 1.25 1.5 1.75
20
40
60
80
![Page 7: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/7.jpg)
Question: What about the Fermi surface curvature and all that!!
Theoretical Possibility that:
•Bare fermi surface curvature remains electron like whereas Hall constant changes sign.
•Not necessarily in conflict with ARPES, since that probe measures renormalized fermi surface shape. Luttinger volume theorem makes no guarantees on shape of fermi surface, and it is to be expected that for non spherical FS’s, there are large distortions.
•Important frontier theoretical questions: FS shape renormalization compared to Hall constant sign/magnitude.
![Page 8: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/8.jpg)
How do we calculate any of these things reliably? T-J type models are the correct framework but notoriously hard
•Auxiliary fields: slave bosons/fermions/… nice but hardly reliable in the phyisical cases
•Numerical techniques QMC, diagonalization, High Temperature expansions,,promising but defeated by Kubo formulas involving sums over states and a tricky DC limit in most cases.
Basic idea
Think high frequency
![Page 9: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/9.jpg)
First serious effort to understand Hall constant in correlated matter:
S S, Boris Shraiman and Rajiv Singh, Phys Rev Letts ( 1993)
Introduced object
0* lim lim ( )/H xy
BR B
ωρ ω
→ →∞=
•Easier to calculate than transport Hall constant
•Captures Mott Hubbard physics to large extent
2( ) (0) /(1 )xy xy iσ ω σ ωτ= +Motivation: Drude theory has
( ) (0) /(1 )xx xx iσ ω σ ωτ= +
2( )( ) xy
xxxy
σ
σρ ω =Hence relaxation time cancels out in the Hall
resistivity
![Page 10: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/10.jpg)
* 22 v [ , ] /x yiH xxhBR N J Jπ τ−= < > < >
•Very useful formula since
•Captures Lower Hubbard Band physics. This is achieved by using the Gutzwiller projected fermi operators in defining J’s
•Exact in the limit of simple dynamics ( e.g few frequencies involved), as in the Boltzmann eqn approach.
•Can compute in various ways for all temperatures ( exact diagonalization, high T expansion etc…..)
•We have successfully removed the dissipational aspect of Hall constant from this object, and retained the correlations aspect.
•Very good description of t-J model, not too useful for Hubbard model.
•This asymptotic formula usually requires ω to be larger than J
![Page 11: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/11.jpg)
Comparison with Hidei Takagi and Bertram Batloggdata for LSCO showing change of sign of Hall constant at delta=.33 for square lattice
![Page 12: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/12.jpg)
Here δ = ρ -1.
Since Fermi temperature seems low, the large T limit may work, so we predict: RH will not saturate with T.
Predict linear T dependence and known slope.
![Page 13: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/13.jpg)
0 1
(2 ) (1 / 2)(1 3 / 2)| | 4 (1 ) (1 )
1 2
( ) (4 3 )| | 4 (2 )( 1) 4(2 )( 1)
[ ]
[ ]
n
n n nv kTe c t n n n n
n
n n nv kTe c t n n n n
R
R
< <
− − −− −
< <
−− − − −
= − +
= − +
Tr lattice only Tr and square lattice
Triangular lattice prediction at high T (2003) Kumar Shastry
To leading order in T/J we find
![Page 14: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/14.jpg)
2 4 6 8 10
5
10
15
20
25
T
RH
As a function of T, Hall constant is LINEAR for triangular lattice!!
We suggest that transport Hall = high frequency Hall constant!!
•Origin of T linear behaviour in triangular lattice has to do with frustration. Loop representation of Hall constant gives a unique contribution for triangular lattice with sign of hopping playing a non trivial role.
O(β t)4
B O(β t)3
Triangular lattice
square lattice
![Page 15: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/15.jpg)
Hall constant as a function of T for x=.68 ( CW metal ). T linear over large range 2000
to 4360 ( predicted by theory of triangular lattice transport KS)
T Linear resistivity
STRONG CORRELATIONS & Narrow Bands
![Page 16: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/16.jpg)
Triangular lattice: High Temperature expansion result is a successful prediction: namely non saturation of R_H and that the slope is a measure of bandwidth.
Many things to do further: currently under study
ω T dependence.
Experimentally: Determining the R(ω) curve from optics to dc? Interesting terahertz frequency range for NCO.
Lesson learnt from Hall study:
RECOGNIZE CORRECT COMBINATION OF TRANSPORT CONSTANTS
YES NO
σxy(ω)ρxy(ω)σxx(ω)
![Page 17: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/17.jpg)
A Sum Rule for Thermal Conductivity and Dynamical Thermal Transport Coefficients in Condensed Matter -I
S S Summer 05, August 29, 05
Cond-mat/0508711
New formalism with new results:->
•Thermal conductivity sum rule, analogous to plasma sum rule
•Thermo-power formula that is better than the Heikes- Mott -Zener formula. Transport contribution is evaluated and is correct at ALL temperatures for the free electron case, and presumably close to DC answer at all frequencies
•Thermoelectric figure of merit
•Lorentz ratio
•Nernst effect
Explicit results for NCO- a useful prediction regarding design of higher Thermopower materials.
WORK in progress-
![Page 18: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/18.jpg)
R∗H = limB→0;ω→∞
1
Bρxy(ω)
Formalism gives results that are typically the analogs of the Hall constant R* and are equal time correlators of non trivial operators. Hard to evaluate, but not impossible, unlike the DC counterparts.
Need frequency dependence of various objects
κ(ω) ρ(ω) S(ω) L(ω) Z(ω)
![Page 19: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/19.jpg)
Thermal conductivity and thermoelectric power
Frequency dependent objects: Also have REAL experimental motivation:
Relevant frequency scales:100GHz clock speed CPU involves significant change in dissipative characteristics: for silicon.
τ ∼ 10−11sec
New class of experiments are now possible with femto second laser pulse heating, followed by detection of electric voltage or temperature.
τ ∼ 10−15sec
This gets us into electronic time scales
![Page 20: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/20.jpg)
K = K0 +K1e−iωct,
with adiabatic switching from the infinitely remote past t = −∞ as usual, andK0 =
PrK(r) =
Pr(H(r) − µ n(r)). Here H(r) is the energy density, and
since we are mainly dealing with lattice models, we sum over r. The operator
K1 =Xr
ψ(r)K(r),
where ψ(r) is a small (pseudo) gravitational field with some spatial variationsuch that its average is zero.
Physical idea of Luttinger is that ψ(r) is a pseudo temperature, it is amechanical analog of temperature. The action β(r)K(r) tries to be locallyconstant, and so if ψ(r) increases locally then T increases locally too. Finallyuse Einstein relation to relate the response to ∇ψ to that to ∇T , but basicallythink of
∇ψ(r) = ∇T (r)T
![Page 21: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/21.jpg)
Θxx = − limkx→0
d
dkx[JQx (kx), K(−kx)]
Here we commute the Heat current with the energy density to get the thermal operator
Re κ(ω) =π
~Tδ(ω)DQ +Re κreg(ω) with
Re κreg(ω) =π
TLµ1− e−βω
ω
¶ X²n 6=²m
pn|hn|JQx |mi|2δ(²m − ²n − ~ω), (
DQ =1
L
⎡⎣hΘxxi− ~ X²n 6=²m
pn − pm²m − ²n |hn|J
Qx |mi|2
⎤⎦ . (
The sum rule for the real part of the thermal conductivity (an even functiof ω) follows Z ∞
0
Re κ(ω)dω =π
2~TLhΘxxi.
Comment: New sum rule.
completely parallel to f-sum rule for electrical conductivity on a lattice
![Page 22: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/22.jpg)
κ(ωc) =i
T~ωcDQ +
1
TLZ ∞0
dteiωctZ β
0
dτhJQx (−t− iτ)JQx (0)i.
Where
DQ =1
L
"hΘxxi− ~
Xn,m
pn − pm²m − ²n |hn|J
Qx |mi|2
#.
In normal dissipative systems, the correction to Kubo’s formula is zero.
So, what does Θ look like and what is its value? Answer is model dependent,and in brief, Θ is the specific heat times a velocity
Θxx
~T=1
dCµv
2eff
![Page 23: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/23.jpg)
Thermo-power follows similar logic:
< Jx >= σ(ω)Ex + γ(ω)(−∇T )then the thermopower is
S(ω) =γ(ω)
σ(ω).
Φxx = − limk→0
d
dkx[Jx(kx),K(−kx)].
This is the thermo electric operator
γ(ωc) =i
~ωcTL
"< Φxx > −~
Xn,m
pn − pm²n − ²m + ~ωc hn|Jx|mihn|J
Qx |mi
#.
Dγ =1
L
"< Φxx > −~
Xn,m
pn − pm²m − ²n hn|Jx|mihn|J
Qx |mi
#.
γ(ωc) =i
~ωcTDγ +
1
TLZ ∞0
dteiωctZ β
0
dτhJx(−t− iτ)JQx (0)i,
![Page 24: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/24.jpg)
High frequency limits that are feasible and sensible similar to R*
L∗ =hΘxxiT 2hτxxi (1)
Z∗T =hΦxxi2T 2hΘxxihτxxi . (2)
S∗ =hΦxxiT hτxxi . (3)
Hence for any model system, armed with these three operators, we can compute the Lorentz ratio, the thermopower and the thermoelectric figure of merit!
![Page 25: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/25.jpg)
So we naturally ask
•what do these operators look like
•how can we compute them
• how good an approximation is this?
In the preprint: several models worked out in detail
•Lattice dynamics with non linear disordered lattice
•Hubbard model
•Inhomogenous electron gas
•Disordered electron systems
•Infinite U Hubbard bands
•Lots of detailed formulas: we will see a small sample for Hubbard model and see some tests…
![Page 26: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/26.jpg)
Thermo power operator for Hubbard model
Φxx = −qe2
X~η,~η0,~r
(ηx + η0x)2t(~η)t(~η0)c†
~r+~η+~η0,σc~r,σ − qeµ
X~η
η2xt(~η)c†~r+~η,σc~r,σ +
qeU
4
X~r,~η
t(~η)(ηx)2(n~r,σ + n~r+~η,σ)(c
†~r+~η,σc~r,σ + c
†~r,σc~r+~η,σ). (1)
This object can be expressed completely in Fourier space as
Φxx = qeX~p
∂
∂px
©vxp (ε~p − µ)
ªc†~p,σc~p,σ (2)
+qeU
2LX
~l,~p,~q,σ,σ0
∂2
∂l2x
nε~l + ε~l+~q
oc†~l+~q,σc~l,σc
†~p−~q,σ0c~p,σ0 . (3)
τxx =q2e~X
η2x t(~η) c†~r+~η,σc~r,σ or (1)
=q2e~X~k
d2ε~kdk2x
c†~k,σc~k,σ (2)
![Page 27: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/27.jpg)
Θxx =Xp,σ
∂
∂px
©vx~p (ε~p − µ)2
ªc†~p,σc~p,σ +
U2
4
Xη,σ
t(~η)η2x(n~r,σ + n~r+~η,σ)2c†~r+~η,σc~r,σ
−µUX~η,σ
t(~η)η2x(n~r,σ + n~r+~η,σ)c†~r+~η,σc~r,σ
−U8
X~η,~η0,σ
t(~η)t(~η0)(ηx + η0x)2 {3n~r,σ + n~r+~η,σ + n~r+~η0,σ + 3n~r+~η+~η0,σ} c†~r+~η+~η0,σc~r,σ
+U
4
X~η,~η0,σ
t(~η)t(~η0)(ηx + η0x)η0xc†~r+~η,σc~r,σ
nc†~r+~η,σc~r+~η+~η0,σ + c
†~r−~η0,σc~r,σ − h.c.
o. (1)
Interesting by product of these formulas: at T=0, <Φ > must vanish being entropy current, and hence the chemical potential can be expressed as a ratio of two operators. This is pretty surprising, and can be verified in some cases: half filled Hubbard model in any dimension for bipartite lattices µ= U/2
![Page 28: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/28.jpg)
Free Electron Limit and Comparison with the Boltzmann Theory
It is easy to evaluate the various operators in the limit of U→0, and thisexercise enables us to get a feel for the meaning of these various somewhatformal objects. We note that
hτxxi = 2q2eLXp
n~pddpx
£vx~p¤
hΘxxi = 2LXp
n~pddpx
£vx~p(ε~p−µ)2
¤hΦxxi = 2qe
LXp
n~pddpx
£vx~p(ε~p−µ)
¤. (1)
At low temperatures, we use the Sommerfield formula after integrating byparts, and thus obtain the leading low T behaviour:
hτxxi = 2q2eρ0(µ)h(vx~p )2iµhΘxxi = T 2
2π2k2B3
ρ0(µ)h(vx~p )2iµ (1)
hΦxxi = T 22qeπ
2k2B3
∙ρ00(µ)h(vx~p )2iµ + ρ0(µ)
d
dµh(vx~p )2iµ
¸, (2)
![Page 29: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/29.jpg)
We may form the high frequency ratios
S∗ = Tπ2k2B3qe
d
dµlog£ρ0(µ)h(vx~p )2iµ
¤L∗ =
π2k2B3q2e
. (1)
It is therefore clear that the high frequency result gives the same Lorentz numberas well as the thermopower that the Boltzmann theory gives in its simplest form.
The thermal conductivity cannot be found from this approach, but basically the formula is the same as the Drude theory with i/ω -> τ.
Some new results for strong correlations and triangular lattice:
Thermopower formula to replace the Heikes-Mott-Zener formula
![Page 30: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/30.jpg)
Leading High temperature term for the Triangular lattice and application to Sodium Cobalt Oxide
S∗ = − µ
qeT+
qe∆
T hτxxiwhere
∆ = −12
X~η,~η0,~r
(ηx + η0x)2 t(~η) t(~η0) hYσ0,σ(~r + ~η) c†~r+~η+~η0,σ0c~r,σi
This is a very useful alternate formula to the Heikes-Mott-Zener formula wherethe second term in Eq above is thrown out. It interpolates very usefully be-tween the standard formulas for low temperature as well as at high temperature.The second term represents the “transport” contribution to the thermopower,whereas the first term is the thermodynamic or entropic part, which dominatesat high temperaturefor S∗ we can actually make a systematic expansion in pow-ers of βt, unlike the dc counterpart.
![Page 31: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/31.jpg)
Leading high temp expansion:
hτxxi = 6Lq2ethc†1c0i = 3Lq2eβt2n(1− n).nc~r,σ, c
†~r0,σ0
o= δ~r,~r0
nδσ,σ0(1−n~r,σ)+(1−δσ,σ0)c†~r,σc~r,σ
o≡ Yσ,σ0 δ~r,~r0 (1)
∆∼−32Lt2Xσ,σ0hYσ0,σ(~η)c†~η+~η0,σ0c~0,σi.
The spins must be the same to the leading order in βt where we generate ahopping term c†~0,σc~η+~η0,σ froman expansion of exp(−βK), and hence a simpleestimation yields
∆=−34Lt3βn(1−n)(2−n)+O(β3).
This together with µ/kBT =log(n/2(1−n))+O(β2t2) gives us the result for0≤n≤1
S∗= kBqe
½log[2(1−n)/n]−βt2−n
4+O(β2t2)
¾,
![Page 32: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/32.jpg)
S∗ =kBqe
nlog[2(n− 1)/(2− n)] + βt
n
4+O(β2t2)
ofor 1 ≤ n ≤ 2 using particle hole symmetry.From the temperature dependence of the data of Terasaki et aland assuming
S ∼ S∗one finds that t = −1100K, and with this, S∗ ∼ 120µV/K, fairly closeto the observed value.
Note that in these high Temp expansions:
1. Correction is O(β) for triangular lattice but O(β^2) for square lattice. Hence larger transport correction for triangular lattice.
2. Prediction. If sign of t is +ve then ( unfrustrated case) S will reach its asymptotic Heikes-Mott-Zener value FROM ABOVE, hence a peak in S must exist. Such a case must be the largest S for any metallic system.
3. Go FOR IT GUYS!! Let us find a realization of this.
![Page 33: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/33.jpg)
![Page 34: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/34.jpg)
![Page 35: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/35.jpg)
20 40 60 80 100
0.92
0.94
0.96
0.98
Linear versus quadratic behavior of thermopwer as a fn of inverse T
![Page 36: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/36.jpg)
![Page 37: A Sum Rule for Thermal Conductivity and Dynamical Thermal ...physics.ucsc.edu/~sriram/talks_by_sriram/2005_2006/UCDavis.pdf · Sriram Shastry UCSC, Santa Cruz, CA Work supported by](https://reader033.vdocuments.mx/reader033/viewer/2022042712/5f9fa254d44c7d49f44286a8/html5/thumbnails/37.jpg)
Conclusions:
•New and promising formalism.
•Lots of computations are being carried out or planned at Santa Cruz.
•Thanks to my colleagues at SC: Jan Haerter and Mike Peterson