Specific Heat in Thermodynamics

Heat capacity: Specific Heat: dq dq

cP = P cV =
C = lim ⎜
⎛ ΔQ ⎞ C dq dT dT V
ΔT →0 ΔT
⎟ c= =
⎝ ⎠ m dT Per mass or usually “per mole”: CP and CV

⎛ dU ⎞ ⎛ ∂S ⎞ ⎛ ∂S ⎞
dQ = dU + dW = dU + PdV ⇒ CV = ⎜ ⎟ dQ = TdS ⇒ C P = T ⎜ ⎟ ; CV = T ⎜ ⎟
⎝ dT ⎠V ⎝ ∂T ⎠ P ⎝ ∂T ⎠V
neglect the volume expansion when reversible system in equilibrium with its environment
solid is heated, dW ~ 0
sum over all T1

∫
Ei
− possible states (i) S (T1 ) = C
dT
Free energy: F = − k BT ln Z Z = ∑e k BT with energies Ei T
0
⎛ ∂S ⎞ ⎛∂ F ⎞ i 2
dF = d (U − TS ) = − SdT − PdV ⇒ CV = T ⎜ ⎟ = −T ⎜⎜ 2 ⎟⎟
⎝ ∂T ⎠V ⎝ ∂T ⎠V
Note lower boundary
On integration
• Powerful method in materials characterization
⎛ dQ ⎞ ⎛ ∂H ⎞
dH = TdS = dQ ⇒ C P = ⎜ ⎟ = ⎜ ⎟
• Modern time: high level of automatization (PPMS) ⎝ dT ⎠ P ⎝ ∂T ⎠ P

• Notice that what we calculate is CV and what we measure is CP. For isobaric conditions

Introduction to thermal and transport techniques

 Simplest Example: 1D Monoatomic Lattice
Chain of N identical atoms with M: xn (t ) = na + u n (t ), n ∈ N
..
1 M u = K (un +1 − un ) − K (un − un −1 ), n ∈ N
∇qω (q) = 0
i ( q j na −ω j t )
Search for harmonic solution: un (t ) = A j e ( cusp, max, min )
Periodic boundary condition un(t)=un+N(t) → exp(iqjNa)=1
2πj
qj = , j∈N
Na
iq j a − iq j a
2 Mω 2j A j = KA j (e +e − 2) ∇ qω (q ) = const.
6
2K K q ja
3 ω 2j = (1 − cos q j a ) ⇒ ω j = 2 sin velocity of excitations in the chain (group velocity):
M M 2
cg = ∂ω/∂q = cscos(qa/2)sgn(q)
π N cg=0 at the zone boundary |q| = π/a = 2π/λ (Bragg law)
Extremal values: qj = ± ;( j = )
a 2 → at the zone boundary lattice effects are strongest,
⎛ π π⎞ atoms oscillate out of phase for |q|=π/a
First Brilluen zone: q j = ⎜ − ...... ⎟
⎝ a a⎠ 5
4
Number of possible q For low q → ω(q) ~ |q|cs
What is phonon density
values = N (number of (cs ~ speed of sound)
of states in 1D chain?
unit cells in the system)
Introduction to thermal and transport techniques
 Simple example: Diatomic chain in 1D
n-1 n n+1
n

u1(na): displacement of atom n,1

Spring K Spring G
u2(na): displacement of atom n,2

Coupled equations of motion: See textbook, also Aschroft/Mermin “Solid State Physics”
..
M 1 u1 (na) = − K [u1(na) − u2 (na)] − G[u1(na) − u2 ((n − 1)a )] K +G 1
ω 2 (k ) = ± K 2 + G 2 + 2 KG cos qa
.. M eff M eff
M 2 u 2 (na) = − K [u2(na) − u1 (na )] − G[u 2(na) − u1 ((n − 1)a )]
Optical branch. Atoms vibrate
against each other with their
center of mass fixed
There are N 2π n Long wavelength modes can interact with
q= electromagnetic radiation (oppositely charged
values of q, a N ions can be excited by E of the light wave
N={1…N}:
For each q, there are 2 solutions, Acoustic branch. Atoms and their
total of 2N normal modes - phonons center of mass move together as in
acoustical vibration
ω~ck (sound waves)
Introduction to thermal and transport techniques
General Case and Phonon Density of States
If there is s ions per unit cell and N cells, there will be r
3Ns degrees of freedom and 3s normal modes for each q
phonon. The lowest 3 branches are acoustic. Remaining
3(s-1) branches are optical.
Each mode has its own polarization vector:
r
paralel to q - longitudinal mode mixed excitations are
r
perpendicular to q - transverse mode of course possible

Phonon spectrum of real materials (diamond, s=2, 6 normal modes)

include also interaction beyond nearest neighbors, electron – phonon
coupling, anharmonicity…

extremal points of phonon

Define phonon DOS: number of
dispersion curves in BZ states per energy interval:
r
Vdk
ρ (ω ) = ∑ ∫ δ (ω − ωμ (q))
μ (2π ) 3

ρ (ω )dω total number of modes in

infinitesimal range ω+dω
r r per total V
If ω μ (q0 ) = ω then expand: group velocity vμ ( q0 ) s
Vdq
r r r r ∂ω N =∫ ; ∑1 = 3s
ωμ (q ) = ωμ (q0 ) + (q − q0 ) rμ (2π )3 μ
∂q
van Hove singularity
r dSω dω Vdω dSω
3 ∑ ∫
Phonon dispersion relation by dω = ∇ qωμ (q )dq⊥ ⇒ dq = r ρ (ω )dω = r
∇ qω μ ( q ) (2π ) μ ω =ωμ ( qr ) ∇ qω μ (q )
INS:λ, E match
Introduction to thermal and transport techniques
 Longitudinal and transverse modes

dω
Group Velocity: vg =
dK

e
) Mod GaAs
(LA
Frequency, ω

c
s ti
LO LO

ou
8
TO TO
A c
al

Frequency (1012 Hz)

ud
i n
o de 6

i t
ng A )M LA LA

Lo (T 4

stic
u
A co 2 TA
TA

rse
s ve 0
an
0.4 0.2 0 0.2 0.4 0.6 0.8 1.0

Tr L (111) Direction Γ Ka/π (100) Direction X

0 Wave vector, K π/a

Introduction to thermal and transport techniques
 Linear oscillator in quantum representation
Consider harmonic oscillator: V=(1/2)kx = (1/2)mω x →
2 2 2 H =
p2 1
+
2m 2
mω 2 2
x =
1 2
2m
p + (mωx) 2 [ ]
1
Introduce raising and lowering operators: a † = (− ip + mωx ), a = 1 (ip + mωx )
2hmω 2hmω
So that [a,a†]=1 (easy to show since [xi,pj]=iħδij) → aa†=1+a†a

† † 2 †
aψ n = cn ψ n +1 ⇒ ψ n aa ψ n = ψ n +1 c n∗ c n ψ n +1 = cn = c n2 = ψ n aa + 1ψ n = n + 1 → cn = n +1
†
aψ n = n +1ψ n +1
† 2
aψ n = dn ψ n −1 ⇒ ψ n aa ψ n =n= ψ n −1 d n∗ d n ψ n −1 = dn = dn ⇒ dn = n
aψ n = nψ n −1

Now we write Hamiltonian in terms of raising and lowering operators: H

( − ip + m ω x )( ip + m ω x ) 1 ⎛ p2 mω 2x2 mωh ⎞
†
aa =
2mωh
=
1
2mωh
2 2
[
p + ( m ω x ) + im ω ( xp − px ) = ⎜
ω h ⎜⎝ 2 m
+
2
−
2m
] ⎟⎟
⎠

⎛ † 1⎞ ⎛ † 1⎞
H = h ω ⎜ aa + ⎟ = h ω ⎜ aa − ⎟
⎝ 2⎠ ⎝ 2⎠ energy of linear harmonic oscillator
of frequency ω is quantized in ħω

Introduction to thermal and transport techniques

 Einstein Model of Lattice Specific Heat
Collection of uncoupled quantum oscillators, each vibrating with the same frequency ωE
Number of oscillators is equal to number of degrees of freedom in the system

Average energy:
<n> = average quantum number for oscillator = [exp(βħω)-1]-1 (Bose-Einstein distribution)
∞ ∞
⎛ 1⎞
∑ ⎜ n + ⎟hω e
⎝ 2 ⎠
− β nhω
hω ∑ nhω e − β nhω
hω ∂ ⎛ ∞ ⎞ hω ∂ 1 hω hω ⎛ 1⎞
E (T ) = n
= + n
= − ln ⎜ ∑ e − β n h ω ⎟ = − ln − β hω
= + − β hω
= ⎜ n + ⎟hω
∞
2 ∞
2 ∂β ⎝ n ⎠ 2 ∂β 1− e 2 1− e ⎝ 2⎠
∑e
n
− β nhω
∑e
n
− β nhω

Specific heat in Einstein approximation:

∑xn
n
= (1 − x) −1
ΘE – Einstein temperature
2 hω / k T
⎛ ∂U ⎞ ⎛ ∂ ⎞ ⎛ ∂ 1 ⎞ 3 N A k B (h ω E / k BT ) e E B
CV =⎜ ⎟ =⎜ 3N A E ⎟ = ⎜ 3 N A ( n + )hω E ⎟ =
⎝ ∂T ⎠V ⎝ ∂ T ⎠V ⎝ ∂T 2 ⎠ e hω E / k BT − 1
2
( )
High temperature limit (T>> ΘE) – exponents are replaced with expansions and we get Dulong – Petit result:
⎡ 1 ⎛ ΘE ⎞
2
⎤
CV = 3 R ⎢1 − ⎜ ⎟ + ..... ⎥ ≈ 3 R
⎣⎢ 12 ⎝ T ⎠ ⎦⎥
Low temperature limit (T<<ΘE) – predicts that specific heat should decrease to zero exponentially:

C V = 3 R (Θ E / T )2 e − ( Θ E /T )

Real materials: atoms are coupled, vibrate collectively at many different ω

Introduction to thermal and transport techniques
 Debye model of Lattice Specific Heat 1
For thermodynamic properties optical modes are irrelevant (low T).
We keep acoustic modes and replace them with purely linear mode with the same initial dispersion

Enter discrete nature of the solid: total number of vibrational modes is normalized to 3NA

∃maximum (cutoff) frequency ωD

Relation between ω and wave vector
The number 3NA is large (10-24), therefore we
consider vibrational levels as continuous and q is defined by Debuye approximation
for sound waves: ~sound velocity
write number of modes in ω, ω+dω:
ωD q c
ω =c =
∫ ρ (ω )d ω
0
= 3N A 2π λ
Periodic boundary conditions (L is the dimension of representative cube of continuum):
[i (xq x + yq y + zq z )] [i ( x + L ) q x + ( y + L ) q y + ( z + L ) q z ] 2π 4π
e =e ⇒ q x , q y , q z = 0,± ,± ,...
r r L L
Volume in reciprocal space for each wavevector is (2πL)3 → number of allowed values of q per unit volume of q space is (L/2π)3 = V/8π3
r
Number of allowed q values is large, so q ~ continuous variable → number of modes with q or less is V/8π3 · volume of sphere with R = q:

V 4 dN 4π V 2
N = π q 3 ⇒ ρ (ω ) = = ω
8π 3
3 dω c3

12 π V 12 π V 9N A
Real solid, elastic waves for each q have ρ (ω ) = ω2 Total number of oscillators ω 2dω = ω 2dω
longitudinal component and 2 transverse: c
3
in interval ω, ω+dω: c
3
ω 3
D

Cutoff frequency for vibrational spectrum 1/ 3

⎛ 3N A ⎞
ωD = c ⎜ ⎟
Introduction to thermal and transport techniques ⎝ 4π V ⎠
 Debye model of Lattice Specific Heat 2
Specific heat in Debuye model (x=ħω/kBT, ΘD=ħωD/kB): 4π4/15

ωD ωD 3 Θ /T
⎛ ∂U ⎞ ∂ 1 ⎛ hω ⎞ e hω / k BT 9N ⎛ T ⎞ D
x 4 e x dx
CV =⎜ ⎟ = ∫ ρ (ω )( n + )h ω d ω = ∫0 B ⎜⎝ k B T ⎟⎠ ( e h ω / k B T − 1) 2 ω D3 A ω d ω = 9 R ⎜⎜⎝ Θ D
k ⎜ ⎟ 2
⎟⎟ ∫ (e
⎝ ∂T ⎠V ∂T 0
2 ⎠ 0
x
−1 )
2

High temperatures: T>>ΘD we replace exponential with Low temperatures T<<ΘD , upper limit of integral is ∞ and we get
series function and obtain Dulong Petit result
3
⎡ 1 ⎛ΘD ⎞
2
⎤ 12 4 ⎛ T ⎞ N 234 R
CV = 3 R ⎢1 − ⎜ ⎟ + ... ⎥ CV = π R ⎜⎜ ⎟⎟ = β T 3 = 234 R θ −D3T 3 θ D ( N atoms) = 3
⎣⎢ 20 ⎝ T ⎠ ⎦⎥ 5 ⎝ ΘD ⎠ β

Note that it holds for 1 atom in unit cell = 3 phonon branches

Usually good for T < (ΘD/10 - ΘD/50). For T > Debuye region:
ρ (ω ) = α 1ω 2 + α 2 ω 4 + α 3ω 6 + ... ⇒ C V = β 1T 3 + β 2 T 5 + β 3T 7 + ....

Physical significance: Debuye temperature is a measure of the stiffnes of the crystal: above ΘD all modes are getting
excited, and below ΘD modes begin to be “frozen out”, marking rapid reduction in CV with decreasing temperature.

Shortcoming of the model: θD = θD(T)

but at high T all vibrational modes are
excited so θD=const. (classical result)

Introduction to thermal and transport techniques

 Application of Debye Model
Debye model can be applied succesfully to many materials with more than one atom in
the unit cell.
Sometimes high temperature
range can also have CP ~ T3 0.4
which is unphysical.
UCoGa5

Cp/T (J/mol.K2)
At T > θD the model is not valid. 0.3
Optical phonon mode contribution
is not negligible. Thermal
expansion may not be negligible 0.2
Debye temperatures of solids:

0.1
2 2
T (K )
0.0
0 100 200 300 400 500
Introduction to thermal and transport techniques
Specific Heat Dominated by ~ T3 Phonon Contribution

10 7
C = 3 ηkB = 4.7 ×10 6 J
m3 −K

6
10

Diamond
3

5
Specific Heat, C (J/m -K)

10
Each atom has
a thermal energy
10 4
of 3kBT
C ∝ T3
10 3 Classical
Regime
2
10
θ D = 1860 K

10 1 1
10 10 2 10 3 10 4
Temperature, T (K)

Introduction to thermal and transport techniques

 Electronic specific heat
1
Gas obeying Fermi-Dirac statistics. Fermi Dirac distribution function: f (ε ) =
e ( ε −ε F ) / k B T + 1
At T=0 all energy levels with ε<εF (~ TF=εF/kB) are occupied, rest are vacant. At T>0 electrons with ε~kBT of εF have
sufficient thermal energy to become excited to vacant levels.

At T<TF, fraction of electrons with ε~kBT is of the order of T/TF, contributing to U ~ NA(T/TF)kBT and to C ~ 2RT/TF

Since TF ~ (104 – 105)K → Ce ~ 10-2R which is ~ 1% of lattice contribution (think Dulong Petit).
∞ ∞
Internal energy of a system of N electrons is: U = ∫ εf (ε )n(ε )dε where N = ∫ f (ε )n(ε )dε
0 0
Total density of states for both spin directions is n(ε), so that n(ε)dε is the number of energy levels in interval ε, ε+dε
∞
∂N ∂f
At very low temperatures T<<TF, we can write 0 = ε F = ∫εF n(ε )dε
∂T 0 ∂T
Nonzero only in the vicinity of εF.

∞ ∞ ∞
⎛ ∂U ⎞ ∂ ∂f x 2 e x dx
CV = ⎜ ⎟ = ∫ εf (ε )n(ε )dε = n(ε F ) ∫ (ε − ε F ) dε = k BT ∫ x
2 For x0 = -∞
~ π2/3
⎝ ∂T ⎠V ∂T 0 0
∂T x0
(e + 1) 2

We introduce x = (ε-εF)/kBT and x0 = -TF/T

π 2
Ce = n(ε F )k B2T = γT
3 Linear T dependence of electronic specific heat
Introduction to thermal and transport techniques
 Free electron model
Simplest but rather useful approximation. Electrons move as free particles with dispersion:
ε=
h2 2
2m
k =
h 2
2m
k x + k y2 + k z2 [ ]
Periodic boundary conditions: ki = 2πn/L (i=x,y,z), (n = 0, ±1, ±2….) → there is one allowed
r of k for each cell of (2π/L)3 = 8π3/V → number of allowed wave vectors per unit V of
value
k space is V/8π3 r
The lowest energy state is obtained by placing pairs of electrons in states with smallest k ,
occupying all states within sphere of radius kF:
1/ 3 2/3
⎛ 4πk F3 ⎞ V ⎛ 3π 2 N ⎞ h 2 ⎛ 3π 2 N ⎞
⎜⎜ ⎟⎟ 3 • 2 = N ⇒ k F = ⎜⎜ ⎟⎟ ⇒ εF = ⎜ ⎟
⎝ 3 ⎠ 8π ⎝ V ⎠ 2m ⎜⎝ V ⎟⎠
Fermi energy and radius of Fermi sphere depend only on electron density N/V
3/ 2
V ⎛ 2mε ⎞
3/ 2
dN V ⎛ 2m ⎞
Number of states N with ε < εF is: N = 2 ⎜ 2 ⎟ ,density of states: n (ε ) = = ⎜ 2 ⎟ ε
1/ 2

3π ⎝ h ⎠ dε 2π 2 ⎝h ⎠

1/ 3
m ⎛⎜ 3 zN AV m ⎞⎟
2
Where z=N/NA is the conduction electron/atom
Density of states at εF: n(ε F ) = 2
πh ⎜⎝ π ⎟
⎠
ratio and Vm is the molar volume
2/3
π2 m 1 / 3 ⎛ πVm ⎞
Ce = n(ε F ) k T = γ 0T ⇒ γ 0 = 2 ( zN A ) ⎜
2
⎟ kB
2
Ce/T in free electron model
3 B
h ⎝ 3 ⎠
Introduction to thermal and transport techniques
 Thermal Relaxation Calorimetry 1
Sample of unknown heat capacity Cx is attached to
Based on a measurement of thermal response
a sample platform with a thermal grease (e.g.,
of a sample calorimeter assembly to a change
apiezone N grease).
in heating conditins.
The platform consists of a thin sapphire or silicon
disc, which has high thermal conductivity. A thin-
film heater is evaporated onto the bottom of the
platform, and the platform temperature Tp is
determined from a bare temperature sensor attached
to it.

Wires thermally link the platform to a copper heat

sink held at a constant temperature T0. They create
a thermal link between the bath and platform with a
thermal conductance K1. They also provide
electrical connections to the temperature sensor and
heater.

Power P is applied to the platform via the thin-film

Heat flow diagram for a standard relaxation calorimeter heater, and a system of differential equations is
Cryogenics 43, 369 (2003) solved
Introduction to thermal and transport techniques
 Thermal Relaxation Calorimetry: Single τ
Combined addenda HC dTP
(platform, T sensor, heater, P = Ca + K 2 (Tp − Tx ) + K1 (Tp − T0 )
and grease) dt
dT
0 = C x x + K 2 (Tx − Tp ) Thermal conductance of
sample – platform thermal link
dt
Power P is applied to the heater, the platform sample assembly warms to a temperature T0 + ΔT =
T0 +P/K1. If the thermal connection sample - platform is very strong (K2 >> K1, Tx ≈ Tp), we get:
dTP
P = (Ca + C x ) + K1 (Tp − T0 )
dt
Power P is discontinued and the platform/sample assembly will cool to the bath temperature T0:

Tp (t ) = T0 + ΔTe −t /τ ;τ = (Ca + C x ) / K1
For small ΔT (ΔT/T << 1), we can ignore T dependence of Ca, Cx and K1 and get Cx via τ
In this method (using single τ) a steady state at constant P and Tp > T0 followed by relaxation
to T0 can be used to determine both K1 and Cx. K1 is determined by measuring the temperature
change ΔT that results when power P is applied. , Addenda heat capacity Ca can be determined
from a decay measurement with no sample attached to the platform.
Cryogenics 43, 369 (2003) Introduction to thermal and transport techniques
 Thermal Relaxation Calorimetry: Double τ
This method is applied when K2 >> K1 due to poor sample - platform thermal link. Therefore
Tx ≠ Tp. Thermal decay of Tp is described using two exponentials:

Tp (t ) = T0 + Ae − t /τ1 + Be − t /τ 2
Time constant τ2 is usually much shorter than the other. Thus, there are two relaxation times:
1. Shorter relaxation process (τ2) between the sample and platform
2. Longer gradual process (τ1) due to thermal relaxation between the platform/sample and the
heat-sink temperature bath.

By measuring decay curves it is possible to determine τ1, τ2, K2, and Cx given known values for
Ca and K1. Series of decays cycles (10–100) can be averaged at each temperature to obtain data
scatter of less than 1%.

Drawbacks of the method:

1. Measurements can be time consuming and are becoming impractical for τ1 ~ 100s
2. It is assumed that sample Cx does not vary much during T0 + ΔT, which may not be satisfied
near the phase transition. Better results are obtained using single τ method near the phase
transition since near the phase transition Cx >> Ca and we get:
T − T0
C x (T ) = − K1
Introduction to thermal and transport techniques dT / dt
Quantum Design PPMS

Introduction to thermal and transport techniques

PPMS Heat Capacity Option Hardware

Introduction to thermal and transport techniques

PPMS Heat Capacity Puck

sample grease
vacuum
platform

T base

T base
(3x3 mm)
wires
heater thermometer

After applying heat pulse P(t), thermal response of a system is:

dT (t ) Base temperature
Ctotal = P(t ) − K w (T (t ) − Tb )
dt
Heat coupling through wires, thermal conductance
Ctotal = (sample+platform+grease). Heater power P(t) is varied, and
T vs time t is measured, minimizing the difference between
measured temperatures and the model. Here, Tb is the base
temperature, Kw is the heat coupling through the wires.

We choose heat ⎧P (0 ≤ t ≤ t 0 )
P(t ) = ⎨ 0
pulse as: ⎩0 (t > t0 )
Introduction to thermal and transport techniques
 Quantum Design Two τ Method

Heat pulse applied

34.57 Sample is not 100% thermally attached
34.43
to the platform, so we have to solve two
Temperature (K)

coupled differential equations. The T(t)

34.29 then usually looks like the blue curve.
34.14

34.00

33.86

33.71

33.57
0 2 4 6 8 10 12 14 16 18 20

Time (sec) Sample, platform T

dTp (t )
Cplatform = P(t ) − K w (T p (t ) − Tb ) + K g (Ts (t ) − Tp (t ) )
dt
= − K g (Ts (t ) − Tp (t ) )
dTs (t ) Thermal coupling sample – platform through grease
Csample
dt
Introduction to thermal and transport techniques
 Data Fitting Procedure
dT (t )
Let’s consider first ideal case of 100% coupling: Ctotal = P(t ) − Kw (T (t ) − Tb )
dt
⎧P0 (0 ≤ t ≤ t0 ) Ton (0) = Tb
For applied heat pulse: P(t ) = ⎨ and boundary conditions:
⎩ 0 (t > t 0 ) Ton (t0 ) = Toff (t0 ),

We obtain T(t): T (t ) = ⎨
( )
⎧Ton (t ) = P0τ 1 − e−t /τ / Ctotal + Tb (0 ≤ t ≤ t0 ) Where:
(
⎩Toff (t ) = P0τ 1 − e )
−t0 /τ −(t −t0 ) /τ
e / Ctotal + Tb (t > t0 ) τ = Ctotal / Kw

P0, t0 are known.

All uknowns are obtained by minimizing the difference

∑(T (t ) − T )
2
i i
between measured temperatures Ti and those obtained i
from the model at the same time ti.
model measured

In the real case of non-ideal coupling, more unknowns are fitted at the same time. The
process is more extensive numerically. We fit only to the expression for Tp since the
thermometer in the system is on the platform, no thermometer is attached to the sample
and we assume T(sample) = T (platform) = T(wires).

Introduction to thermal and transport techniques

 Schottky Anomaly 1
Consider system with discrete energy levels. When temperature is comparable to
level separation Δ, specific heat has a broad peak (Schottky anomaly).

Consider a general case of a system with multiple non-degenerate levels ε1…..εn..

Average thermal energy is:
n ⎧ n2
⎛ − Δi ⎞ ⎛ n ⎛ − Δi ⎞ ⎞ ⎫
∑Δ e − ( Δ i / k BT ) ⎪ ∑ Δ2i exp⎜⎜ ⎟⎟ ⎜ ∑ Δ i exp⎜⎜ ⎟⎟ ⎟ ⎪
⎜ ⎝ B ⎠⎟
i
dU R ⎪ i =0 k T
⎝ B ⎠ − i =0 k T ⎪
U=N i =0
⇒ Cschottky = = 2⎨ n ⎬
⎛ − Δ i ⎞ ⎜⎜ n ⎛ − Δ i ⎞ ⎟⎟
A n
dT T ⎪
∑e − ( Δ i / k BT )

⎪ ∑
exp⎜⎜
k T
⎟⎟
⎜ ∑ exp⎜⎜
k T
⎟⎟ ⎟ ⎪
⎪
i =0
⎩ i =0 ⎝ B ⎠ ⎝ i = 0 ⎝ B ⎠ ⎠ ⎭ TΔ=Δ/kB
Now reduce this to a two level system with energies ε0=0 and ε1=Δ Level
with degeneracies g0 and g1: separation
2 in K
N A g1Δe − ( Δ / k BT )
dU g0 ⎛ Δ ⎞ e ( Δ / k BT )
U= ⇒ Cschottky = = R ⎜⎜ ⎟⎟
g 0 + g1e −( Δ / k BT ) dT [
g1 ⎝ k BT ⎠ 1 + ( g 0 / g1 )e ( Δ / k BT ) ]
2

Low temperature limit (T<<TΔ): High temperature limit (T>>TΔ): Entropy:

∞
Cschottky
S schottky = ∫
2 2
g 0 ⎛ TΔ ⎞ −(TΔ / T ) C g 0 g1 ⎛ TΔ ⎞ dT
Cschottky =R ⎜ ⎟ e schottky = R 2 ⎜ ⎟ T
g1 ⎝ T ⎠ ( g 0 + g1 ) ⎝ T ⎠ 0

Introduction to thermal and transport techniques

S schottky = R ln(1 + ( g1 / g 0 ) )
 Schottky Anomaly 2
Peak magnitude depends on ratio of g’s. It increases with degeneracy difference. Usually large
when compared with other contributions, and dominant if occurs at low temperatures:

Cschottky ~ R, other contributions ~ 10-2R

In everyday life presents a problem since separation of other contributions is non - trivial

ΔE In the simplest case of two-level system,

Cschottky shows as an anomaly with
maximum at ~ 0.4·ΔE. For a more
complex level system, the Schottky
specific heat is smooth function without
any clear anomalies, but it can be fit with
exponential and ~ T-2 terms

Nuclear Schottky anomaly – when interaction removes degeneracies of nuclear levels – could be
produced by external magnetic field, by hyperfine magnetic field from conduction electrons or by
CEF gradients. Introduction to thermal and transport techniques
 Schottky Anomaly Example
Given the Hund’s rule, groundstate multiplet J we would expect Rln(2J + 1) entropy
associated with the magnetic state. This is what we find for Gd: (S = 7/2, L = 0 and J = 7/2)
therefore S ~ R ln(8). For other rare earths the spin-orbit coupling gives rise to crystalline
electric field (CEF) splitting.

e.g.: Ce (J =5/2) in cubic point symmetry Δ

Δ = 18 K
This can be seen in the Cp as a
Schottky anomaly. This is
clearly shown at the right in
PrAgSb2. CP is modeled as a
two level system.

Introduction to thermal and transport techniques

Heavy fermion superconductor PrOs4Sb12

Phys. Rev. B 73, 104503 (2006)

Introduction to thermal and transport techniques
 Specific Heat in Disordered Solids

Think of Schottky for two single level system with energy configurations ±ε:
2
⎛T ⎞ ⎛T ⎞
Cschottky = k B ⎜ Δ ⎟ sec h 2 ⎜ Δ ⎟
⎝T ⎠ ⎝T ⎠

Disordered solids ~ collection of two level systems characterized by ±ε:

Low T upper limit ∞

2 ∞
⎛ ε ⎞ ⎛ ε ⎞ π2/12
CV = 2k B ∫ P(ε )⎜⎜ ⎟⎟ sec h 2 ⎜⎜ ⎟⎟dε = 2k B2 P0T ∫ x 2 sec hxdx
⎝ k BT ⎠ ⎝ k BT ⎠ 0
Assume P(ε)=const=P0 and x=ε/kBT

Specific heat of disordered two level system has linear T contribution to heat capacity:

π2
CV = P0 k B2T
6

Introduction to thermal and transport techniques

 Paramagnetic Metals at Low Temperatures
At low temperatures both electronic and vibrational excitations contribute:

Cp = aT + bT3
Clean separation of a and b is usually done by plotting C/T versus T2.
As we have seen a is proportional to N(EF) and b is proportional to the Debye temperature.

Introduction to thermal and transport techniques

 Effective mass

60

50
β
Cp/T (mJ/mol.K2)

40
LuNiAl Electronic specific heat coefficient γ can
30 be estimated from the low-temperature
Au specific-heat data, where the lattice part
20
reduces to ~T3 dependence.
10 Cu
γ In real metals, the γ value is often different
0 from that obtained by free electron model.
0 20 40 60 80 100
2 2
2 T (K ) We can introduce an effective mass m* to
Cp / T = γ + βT account for this difference, since n(EF) is
γ (mJmol-1K-2): proportional to the carrier mass me.

Li Na K Fe Mn Cu Zn Ag Au Al Ga m * γexp
Free e 0.8 1.1 1.7 0.6 0.6 0.5 0.8 0.6 0.6 0.9 1.0 =
me γf.e.

exp. 1.6 1.4 2.1 4.6 15.2 0.7 0.6 0.6 0.7 1.3 0.6
Introduction to thermal and transport techniques
 Heavy Fermions
The electronic specific heat can be used as a measure of the electron effective mass, based on
the free electron result of:
γ = kB2T ·(mkF/3ħ2)

Heavy Fermions are compounds with

exceptionally high values of γ for
T < TK, the Kondo temperature.

They are defined as compounds with

γ > 100 mJ/mol-K2, about 100 times
the value of g for Cu. Large γ means
large electron mass, ergo, heavy
fermion.

Introduction to thermal and transport techniques

 Phase Transitions
Specific heat can also be used to locate and characterize phase transitions. We can suppress
superconducting Tc with an applied magnetic field, so the Cp feature can be more clearly seen
via comparison to the non-superconducting (in high applied magnetic field) compound.

Local moment ordering can be seen even more

clearly (larger entropy). Shown here are a series
of transitions in antiferromagnetic DyAgGe.

Introduction to thermal and transport techniques

 First and Second Order Phase Transitions

2.5
8

2.0
HoNiAl
UO2
6

Cp/T (J/mol.K2)
Cp /T (J/mol.K )
2

1.5
4
1.0

2
0.5 T2
T1
0 0.0
20.0 25.0 30.0 35.0 40.0 0 4 8 12 16 20
T (K) T (K)
Not ideal δ function but sharp anomaly. TC can be determined e.g. by idealization of the
Clear identification of TC. specific heat jump under the constraint of entropy
⎛ ∂S ⎞ conservation. Circles are real experimental points,
Think that C P = T ⎜ ⎟
⎝ ∂T ⎠ P red line is the idealization drawn in the way that
So if S has discontinuity, Cp will have sharper the yellow and green areas are equal to satisfy the
(like ∂ρ/∂T at magnetic transition) entropy conservation.
Introduction to thermal and transport techniques
 Specific Heat and Critical Region
Basic scaling in the critical region (near
critical temperature that corresponds to
order – disorder transition):
−α −α '
C p,V ≈ t for T > TC and C p,V ≈ t T < TC
T − Tc
α is the critical exponent and t= )
Tc
Magnetic specific heat in Heisenberg
magnetic systems scales with α = α’ ≈
0.01 (Fe, Ni, EuO…)

Sometimes nature of transition may change

in magnetic field: Magnetic field induced
first order transition in heavy fermion
superconductor CeCoIn5 detected by
specific heat measurement

Phys. Rev. Lett. 89, 137002 (2002) Introduction to thermal and transport techniques
 Specific Heat of Spin Waves 1

For T > 0 thermal excitations create spin waves in magnetically ordered materials that propagate due to
Gyromagnetic ratio
exchange coupling between neighboring spins. Heisenberg hamiltonian:
rr Sum over all spins i and all j of their NN’s
H = −2J ∑ Si S j
ˆ Each spin has associated magnetic moment
r r
i, j μl = γhSl
Consider spin oscillations along linear ferromagnet. Ground state: all spins are aligned along z with excitation:
r r r 2J r r r rr Effective magnetic
− 2JSl (Sl −1 + Sl +1 ) = − μl (Sl −1 + Sl +1 ) = −μl Be (exchange) field
γh r r
r d (hS ) r r dSl 2J r r r
Exchange field Be will induce a torque on spin Sl : l
= γhSl × Be ⇒ = Sl × (Sl −1 + Sl +1 )
dt dt h
With approximation of small spin deviation (good at low T) : Slz ≈ S , Slx , Sly << S
d (Slx ) 2JS d (Sly ) 2JS dSlz
= (2Sl − Sl −1 − Sl +1 ),
y y y
=− (2Sl − Sl −1 − Sl +1 ),
x x x
=0
dt h dt h dt
We look for wave solutions (a = lattice parameter): Slx = ξ xei (ωt +lqa) , Sly = ξ y ei (ωt +lqa)
4JS
And get new sets of equations: − iωξx + (1− cosqa)ξ y = 0, 4JS (1− cosqa)ξx + iωξy = 0
h h
Introduction to thermal and transport techniques
 Specific Heat of Spin Waves 2

Nontrivial solutions if determinant is zero, we get spin wave dispersion relation:

4JS (qa) 2JSa2q 2

2
4JS
ω= (1− cosqa) ≈ =
At low T where low frequencies (long
wavelength dominate so (1-cosqa)=(1/2)(qa)2
h h 2 h
αf -constant that depends on crystal structure
2JSa2q2
Now in 3D in the small q limit this becomes: ω = α f • Works well in long wavelength limit for
h metallic ferromagnets too.

2J ' Sa2q Solutions doubly degenerate, Two

Antiferromagnets: ω = α a • (see Rev. Mod. Phys. 30, 1 (1950))
h spin wave modes with ω for each q
(long wavelength, small frequencies)
hωq V hωq q dq
∞ 2
V
∞
x 4
dx (x2 = q2b/kBT
We can write: U = ∑
2 ∫ ( hωq / k BT ) ∫
−
( hω / kBT )
= = b 3 / 2
(kBT) 5 / 2
and b=2αfJSa2)
q e − 1 2π 0 e − 1 2π 2
0 e x2
−1
Sum over allowed values of q (1st BZ) Number of modes with q,q+dq is n(q)dq = (dN/dq)dq=(V/2π2)q2dq
(Recall Debuye model)
3/ 2
⎛ dU ⎞ ⎛ kBT ⎞ For antiferromagnets:
Specific heat is then: CM = ⎜ ⎟ = c f N A B⎜
k ⎟
⎝ dT ⎠V ⎝ 2 JS ⎠ ⎛ dU ⎞ ⎛kT ⎞
3

CM = ⎜ ⎟ = ca N AkB ⎜ B ⎟
In insulators: In metals: ⎝ dT ⎠V ⎝ 2J ' S ⎠
(see Rev. Mod. Phys. 30, 1 (1950))
CP = βT + δT
3 3/ 2
CP = γT + βT + δT
3 3/ 2
Hard to resolve lattice vs magnons,
Introduction to thermal and transport techniques in metallic AFM linear term as well
 Entropy Associated with Magnetic Order
For these local moment
systems S ~ R ln D
S = ∫(Cp/T)dT
We can determine how much entropy (change) is
associated with a given state. For magnetic systems we
need to use the magnetic Cp. In practice, this is done by
subtracting off the Cp(T) data from a non-magnetic
analogue (e.g. LuAgGe from TmAgGe).

Introduction to thermal and transport techniques

 Specific Heat of BCS Superconductors

Attraction of electron pairs by virtual phonon exchange, leading to T – dependent gap 2Δ at the
Fermi level. At T = 0: 2Δ(0)=3.52kBTC

Number of broken pairs as the temperature is increased is proportional to exp[-2Δ(0)/kBT], so

contribution to electronic specific heat in superconducting state is:

Cels a = 8.5, b = 1.44 for 2.5 < TC/T < 6

= ae(−bTC / T ) a = 26, b = 1.62 for 7 < TC/T < 12
γTC Schrieffer, J. R. Theory of Superconductivity, Benjamin, New York 1965
Near TC there is abrupt discontinuity since gap vanishes at TC. No latent heat is released (second
order phase transition):

Cels (TC) − γTC

= 1.43
γTC

Other complications may involve : strong coupling, gap anisotropy, presence of two distinct
energy scales (two gaps).

Introduction to thermal and transport techniques

 Some Other Contributions to Specific Heat

Spin fluctuations (critically – damped spin waves) in nearly ferromagnetic systems when
exchange interaction is not strong enough to produce ordered state, for T << Tsf:
3
⎛T ⎞ ⎛T ⎞
Ce = γT + α ⎜ ⎟ ln⎜ ⎟ (Phys. Rev. Lett 17, 750 (1966))
⎜T ⎟ ⎜T ⎟
⎝ sf ⎠ ⎝ sf ⎠

BCS superconductors with

anisotropic gap have reduced
value of specific heat anomaly at
superconducting transition (see
Ann. Phys. 40, 268 (1966)):

ΔC
γTC
(
= 1.426 1 − 4 a 2 )

Introduction to thermal and transport techniques Phys. Rev. B 75, 064517 (2007)
 Superconducting and Magnetic Entropy
Nd1-xCexCoIn5

4 1.2
TN(C) TN(χ) TCOH (ρ)
x=1 TC(ρ) TC(C) TM (C)
1.0
x = 0.9
0.8 x = 0.7
SMAG (Rln2)

x = 0.6 10 10
3
0.6 x = 0.4 HEAVY FERMION
x = 0.2
0.4

T (K)
X=
x=0
[C-Clatt]/T (J/mol K )
2

0.2
LMM
0.0 1 1
2 0 2 4 6 8 10
T(K)
MAGNETIC ORDER SUPER
CONDUCTOR

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
1 X

Heat capacity and magnetic entropy of electronic

system that evolves from local moment magnetism
to heavy fermion supeconductivity via disordered
0
1 T (K) 10 heavy fermion magnetic states
Phys. Rev. B 77, 165129 (2008)
Introduction to thermal and transport techniques
 Thermal Expansion in Thermodynamics

⎛ dQ ⎞ ⎛ ∂H ⎞
At T = 0 CP = CV (entropy S = 0): C P = ⎜ ⎟ = ⎜ ⎟
⎝ dT ⎠P ⎝ ∂T ⎠P
At T ≠ 0 CP is always greater than CV (heating with P = const for dT cost energy to do
work expanding against external pressure; if V = const, there is no work done):

β 2VmT
General relation between CP and CV at T ≠ 0 is C p − CV =
B
β (volumetric expansion) 1 ⎛ ∂V ⎞
1 ⎛ ∂L 3
⎞ 3 ⎛ ∂L ⎞ α (linear expansion) B=− ⎜ ⎟
β = 3 ⎜⎜ ⎟⎟ = ⎜ ⎟ = 3α Vm (molar volume) V ⎝ ∂P ⎠T
L ⎝ ∂T ⎠P L ⎝ ∂T ⎠P B (isothermal compressibility)

In practice, CP – CV ~ 1% of CP at θD/3 and 0.1% of CP at θD/6 and is maximum (~ 10%) at

Tm (melting temperature).

Introduction to thermal and transport techniques

 Thermal Expansion of Anharmonic Crystals
Harmonic crystals do not expand when heated and do not shrink when cooled since average
interatomic spacing does not increase with T.
Real materials have non-parabolic U(r) so there is thermal expansion and CV increases above
Dulong – Petit at T>θD

Cu

1 ⎛ ∂V ⎞
βT = ⎜ ⎟
V ⎝ ∂T ⎠ P

Grüneisen approximation: Δω/ω ~ γΔV/V

Grüneisen parameter is a measure of anharmonicity

1 ⎛ ∂V ⎞ ⎛ ∂P ⎞ 1 ⎛ ∂P ⎞ ⎛ ∂S ⎞ ⎛ ∂U ⎞ ⎛ 1⎞ hω hω
βT = − ⎜ ⎟ ⎜ ⎟ = ⎜ ⎟ T⎜ ⎟ =⎜ ⎟ P = −⎛⎜ ∂F ⎞⎟ U = ⎜ n + ⎟hω + U eq = +
V ⎝ ∂P ⎠T ⎝ ∂T ⎠V B ⎝ ∂T ⎠V ⎝ ∂T ⎠V ⎝ ∂T ⎠V ⎝ ∂V ⎠T ⎝ 2⎠ 2 1 − e − β hω
this
Harmonic = 0 We get γs,q ~ 2 γ s ,q = ∂ (ln ωs (q))
∂ ⎡ 1 ⎤ ⎡ ∂ (hωs (q ) ) ⎤ 1 ⎡ ∂ ⎤ ∂ r ∂ (ln V )
P=− ⎢ eq ∑
U + h ω s ( q ) ⎥ + ∑ ⎢− β hωs ( q ) ⎥ ⇒ β T = ∑ ⎢ − h ω s ( q ) ⎥ n s ( q )
∂V ⎣ q,s 2 ⎦ q , s ⎣ ∂V (e − 1) ⎦ B q , s ⎣ ∂V ⎦ ∂T
∂ r
CV = ∑ hω s (q ) ns ( q ) γCV
s ,q ∂T
Grüneisen
βT =
Introduction to thermal and transport techniques BV
Thermal Expansion by Capacitive Dilatometry 1
Rev. Sci. Instr. 77, 123907 (2006)
In a capacitive dilatometer the dilation L of a sample of length L manifests as a change in the
gap D between a pair of capacitor plates. For an ideal parallel-plate capacitive dilatometer in
vacuum the relationship between the measured capacitance C and D is
[ L(T ) − L(0)]
ε 0 A ε = 8.854·1019 pF/m Thermal expansivity ε =
L(0)
C= 0

D A = area of capacitor plates Thermal expansion α=

1 dL d (ln L)
=
L dT dT

Introduction to thermal and transport techniques

Thermal Expansion by Capacitive Dilatometry 2
Rev. Sci. Instr. 77, 123907 (2006)

Commercial capacitance bridges have ~ 10-7 pF resolution at 1kHz → 3·10-3Å for dilatometer
operation at 18 pF.
Part of calibration processs is to find an appropriate functional relationship between the
capacitor gap D and the measured capacitance C. This is done using a sample platform e long
enough to adjust the capacitor gap from its largest zero-force to its smallest shorted value.
A protractor (with an appropriately sized hole in its center)
is attached to the main flange o. The dilatometer is inverted
and the sample platform e is then rotated and tightened in
small steps; after each step the angular position of the sample
platform (read off the protractor) and the capacitance C
are measured. The capacitor gap is: Angle of dilatometer shorting
1 c
D = c1 (θ m − θ ) ⇒ = 1 (θ m − θ )
C ε0 A
Thread pitch, # threads/mm → #μm/mm
Distance from straight to tilted plate
ε 0 A ⎛⎜ ⎛ C ⎞
2
Deviation from linear behavior ⎞ ⎟ C when they short
since plates are not perfectly D= 1 + ⎜⎜ ⎟⎟
c1 ⎜ C ⎟
parallel. Can be estimated as: ⎝ ⎝ max ⎠ ⎠
Introduction to thermal and transport techniques
Thermal Expansion by Capacitive Dilatometry 3
Rev. Sci. Instr. 77, 123907 (2006)

Data acquisition process in temperature or magnetic field consists of two steps:

Since distance between plates
1. Measurement of reference sample (Cu) is T,H dependent but is independent
2. Measurement of unknown sample. of length or nature of the sample
in dilatometer

α=
1 dL 1 d
= [Dc − Ds ] + α Cu ⎡⎢1 + Ds − Dc ⎤⎥ When the reference Cu or Sample are mounted
L dT L dT ⎣ L ⎦
For long samples with thermal expansion close to Cu [Ds-Dc]/L ≈ 0, therefore (since dL = -dD):
Cell effect, measurement with Cu standard installed
1 dL 1 ⎛ dL ⎞ 1 ⎛ dL ⎞
α= = ⎜ ⎟ − ⎜ ⎟ + α Cu
L dT L ⎝ dT ⎠ cell + sample L ⎝ dT ⎠ cell +Cu Resolution ~ 0.03 – 0.11Å
Depends on cryogenic
⎛ dD ⎞ 1 ⎡ Di +1 − Di Di − Di −1 ⎤ system used and cell
Derivatives are evaluated as: ⎜ ⎟ = ⎢ + ⎥
⎝ dT ⎠i 2 ⎣ Ti +1 − Ti Ti − Ti −1 ⎦
Dilatometer is surrounded by Cu can to provide electrical and thermal insulation.
Introduction to thermal and transport techniques
 Thermal Expansion by Diffraction
Based on the measurement of unit cell by X-ray, neutron methods (will be discussed later in the course).

Unit cell volume

25

20

50

-6
15

α(1/K)*10
40

30 c-axis
10 b-axis
a-axis

20

-6
α (1/K)*10
5 10

0
0 100 200 300 400 500
T(K)
0
0 100 200 300 400 500
T(K)
High resolution data are required. Resolution comparable or smaller
than for capacitive dilatometry. Advantage: bond length, angles, atomic
Phys. Rev. B 72, 045103 (2005)
position information is available
Introduction to thermal and transport techniques
Thermal Expansion in CeCoIn5

Introduction to thermal and transport techniques Phys. Rev. Lett. 100, 136401 (2008)