QUANTUM OPTICS IN ELECTRIC CIRCUITS

763693S

Erkki Thuneberg

Department of physics
University of Oulu
2010
 Abstract ration of aluminium. The Josephson tunnel junctions are
formed at the overlap between the long thin island parallel
With present nanofabrication methods it is possible to
make such small electric circuits that quantum effects beco- to the centre conductor and the fingers extending from the
me essential. The circuits behave like artificial atoms and much larger reservoir coupled to the ground plane.
the methods to deal with them resemble those used in
quantum optics and NMR rather than traditionally used Practicalities
by electrical engineers. This course is an introduction to Write your name, department/group, year class, and
the physics of such circuits. One major topic is how to email address on the list.
include dissipation into quantum mechanics. This will be
answered by deriving a master equation, and applying it The lectures are given in English if there is sufficient
to a harmonic oscillator and to a two-level system. The demand for it, otherwise in Finnish.
realization of the two-level system requires a nonlinear ele- The web page of the course is
ment, for which superconducting Josephson junctions are https://wiki.oulu.fi/display/763693S/Etusivu
used. Another theme is different types of noise (thermal,
shot, quantum). These can be derived by applying scat- The web page contains the lecture material (these notes),
tering formalism which considers electrons in a conductor the exercises and later also the solutions to the exercises.
like waves in a transmission line. We try to answer, among See the web page also for possible changes in lecture and
other things, if noise is present at zero temperature, is su- exercise times.
percurrent noisy, and can zero-point fluctuations be mea-
sured. Time table 2010
Moderneilla valmistusmenetelmillä voidaan tehdä pieni- Luennot: ti 10-12, sali TE320 (7.9. - 30.11.)
kokoisia sähköisiä piirejä, joissa kvanttimekaaniset ilmiöt
Harjoitukset: ma 10-12, sali TE320 (20.9. - 29.11.)
ovat olennaisia. Nämä piirit toimivat kuin keinotekoiset
atomit ja niiden kuvaamiseen käytetään menetelmiä, jot- Examination: date to be fixed later
ka ovat tutumpia kvanttioptiikassa ja ydinmagneettises- Teaching assistant: Matti Silveri
sa resonanssissa kuin sähköopissa. Tämä kurssi on joh-
datus tällaisten piirien fysiikkaan. Yksi pääaihe on kuin- Doing exercises is essential for learning. In addition,
ka liittää häviölliset ilmiöt kvanttimekaniikkaan. Tämä showing completed exercised will affect your final evalua-
tehdään johtamalla master-yhtälö, ja sitä sovelletaan har- tion. (You can improve by one, for example, from 3 to
moniseen oskillaattoriin ja kaksitasosysteemiin. Kaksitaso- 4.) Start calculating (at home) before the exercise time, 2
systeemin toteuttaminen edellytää epälineaarista element- hours is too short time to start from scratch.
tiä, jona käytetään suprajohtavaa Josephson-liitosta. Toi-
nen pääteema on erityyppiset kohinat kuten lämpö-, isku-
ja kvanttikohina. Nämä voidaan johtaa käyttäen sironta-
formalismia, missä elektroneja johteessa kuvataan kuin aal-
toja aaltoputkessa. Tarkoitus on vastata mm. onko nol-
lalämpötilassa kohinaa, kohiseeko supravirta ja voidaanko
nollapistevärähtelyja mitata.

Figure: [A. Wallraff et al, Nature 431, 162 (2004)] Inte-

grated circuit for cavity QED. a, The superconducting nio-
bium coplanar waveguide. b, The capacitive coupling to the
input and output lines. c, False colour electron micrograph
of a Cooper pair box (blue) fabricated onto the silicon sub-
strate (green) into the gap between the centre conductor
(top) and the ground plane (bottom) of a resonator (beige)
using electron beam lithography and double angle evapo-

1
1. Introduction temporal modes (wavelets), connection to information
theory, Landauer formula
Content 2e2
MT G= (4)
The following is only an appetizer, and details will be h
explained during the course. G conductance, M number of transverse modes, T trans-
mission probability.
harmonic oscillator
introduction to noise, spectral density SII (ω),
electric elements, electric harmonic oscillator, stripline thermal noise SII (ω) = 2GkT ,
microwave oscillator, optical cavity, nanomechanical oscil-
lator shot noise SII (ω) = ehIi,
familiar classical forced and damped harmonic oscillator quantum noise SII (ω) = 2G~ωΘ(ω)
solutions, quadrature variables, rotating wave approxima- scattering states, derivation of thermal, shot, and quan-
tion tum noise in scattering formalism
γ
α̇ = −iω0 α − α + if. (1)
2
Prerequisites
quantum harmonic oscillator (no dissipation), energy ei-
genstates E = ~ω0 (n + 12 ) Prior knowledge on quantum mechanics is needed. The
more you know the easier it is, but I try to explain the
Pgeneral quantum state |ψi (t)i, density operator ρ = quantum mechanics that is used. Knowledge of classical
i Pi |ψi ihψi | Lagrangian and Hamiltonian mechanics is needed at some
general derivation of master equation with dissipative places.
terms, interaction picture, Markov approximation, rota-
ting wave approximation, the master equation for harmonic Books and articles
oscillator The first three are standard references in quantum op-
dρ i tics:
= − [H0 + Hd , ρ]
dt ~ D. F. Walls and G. J. Milburn, Quantum optics (Sprin-
γ
+ (N + 1)(2aρa† − a† aρ − ρa† a) ger, Berlin, 1994, second edition 2008). Although not very
2 pedagogical, I have been reading mainly this.
γ
+ N (2a† ρa − aa† ρ − ρaa† ) (2)
2 M. O. Scully and M. S. Zubairy, Quantum optics (Cam-
bridge, 1997).
thermal state solution, coherent states, decaying solution,
driving terms, general solutions using translation operator. C. Cohen-Tannoudji, J. Dupont-Roc and G. Grynberg,
Atom-photon interactions, basic processes and applications
for average values, the classical and quantum harmonic
(Wiley, New York, 1992).
oscillators are identical.
S. Haroche and J.-M. Raimond, Exploring the quantum
statistical aspect, special case of fluctuation-dissipation
- atoms, cavities and photons (Oxford 2006). Contains tho-
theorem
rough introductions to various topics, experiments with cat
two-state system states.

qubit = two-state system, how to construct using Jo- L. Allen and J. H. Eberly, Optical resonance and two-
sephson junctions, circuit analysis with Josephson junc- level atoms (Dover, New York, 1975). Good account of
tions. Bloch equations, but no master equation.

master equation for two level system, rotating wave ap- M. Devoret: Quantum fluctuations in electrical
proximation, Bloch equations. circuits, in Les Houches, Session LXIII, 1995, Quan-
tum Fluctuations, (Elsevier 1997). p 351-386. Circuit
u theory with Josephson junctions. (download at
u̇ = −∆v − ,
T20 http://www.eng.yale.edu/qlab/archives.htm)
v
v̇ = ∆u − 0 + κEw, S. Datta: Electronic transport in mesoscopic systems
T2
(Cambridge, 1995). Explains the scattering formulation
w − weq
ẇ = − − κEv. (3) and Landauer formula.
T1
Ya. M. Blanter and M. Büttiker: Shot noise in mesosco-
harmonic oscillator coupled to a two-state system, pic conductors, Phys. Rep. 336, 1-166 (2000). Scattering
Jaynes-Cummings model formulation of noise.
C. Beenakker and C. Schönenberger: Quantum Shot Noi-
Electric transport and noise
se, Physics Today, May 2003, p. 37-42. A shorter discussion
transmission lines, transverse modes of shot noise.

2
 Practical applications of quantum electric 2. Harmonic oscillator
circuits
Study oscillations around some equilibrium state. If the
No one knows yet amplitude of the oscillations is small, nonlinear terms are
- quantum computation not important, one is led to harmonic oscillator.
- sensitive measuring devices Examples of harmonic oscillator:
Here the main motivation is just to understand the phe- - mass in spring, mechanical pendulum (for small ampli-
nomena! tudes)
- waves in string: each oscillation mode is a harmonic oscil-
Notation lator
- sound waves: every mode is a harmonic oscillator
e is the charge of an electron. I have tried to formulate
electric analog - LC oscillator, modes in a transmission
everything so that it is irrelevant whether this quantity is
line, and modes of electromagnetic wave (any frequency:
negative or positive.
rf, microwave, infrared, light etc) are examples harmonic
~ = h/2π Planck’s constant. oscillators.
a  b means that |a| is much smaller than |b|, and cor-
respondingly for a  b.
a/bc = a/(bc).
The step function

1, if x > 0,
Θ(x) = (5)
0, if x < 0.

cosh x = 21 (ex + e−x ), sinh x = 1 x

2 (e − e−x ), tanh x =
1/ coth x = sinh x/ cosh x.

Figure: [D. Gunnarson et al,

Phys. Rev. Lett. 101, 256806 (2008)] LC oscillator coupled
to superconducting single-electron transistor. The area of
white rectangles in a and b is shown enlarged in b and c,
respectively.

Figure: [A. Naik et al, Nature 443, 193 (2006)] Scan-

ning electron microscope image of the device: a 21.9-MHz,
doubly clamped, SiN and Al nanomechanical resonator
(NR) coupled to a superconducting single-electron transis-
tor (SSET), with simplified measurement circuit diagram.
The SSET is read out by applying 1.17 GHz microwaves
through the directional coupler, and monitoring the signal
reflected off the LCR tank circuit formed by LTank , CTank

3
and the SSET. ∇ · B = 0,
∂E
∇ × B = 0 µ0 + µ0 j.
2.1 Classical harmonic oscillator ∂t
Mechanical harmonic oscillator has the equation of mo- The fields E(r, t) and B(r, t) can be represented as sums
tion of 3 vector components and spatial Fourier components.
Consider a single mode, for example
ṗ = mq̈ = −kq − mγ q̇ + F (6)
E = E(t)ŷeikx , B = iB(t)ẑeikx . (10)
with spatial coordinate q, momentum p = mq̇, mass m,
friction constant mγ, spring constant k, external driving Assume there is some source current j0 (r, t), for example,
force F and time derivative marked with a dot. This can an oscillating charge. It has one Fourier component
be written in the form j0 (t)ŷeikx that couples to the mode above. In addition we
assume some conductivity σ of the medium, j = j0 + σE.
mq̈ + mγ q̇ + kq = F. (7) (Alternatively, could consider dielectric loss.) Substituting
to the Maxwell equations we get
In the limiting case k = 0, the oscillator equation (6)
reduces to describe a free particle. Thus in this special limit 0 µ0 Ė = kB − µ0 σE − µ0 j0 ,
the harmonic oscillator also applies to the description the Ḃ = −kE. (11)
free-electron model of solids (giving the Drude conductivity
etc.). This again is of the same form as the other oscillator equa-
tions above. Thus a mode of electromagnetic radiation can
Consider an electric LC oscillator. Electric current I = be considered as a harmonic oscillator. The whole electro-
Q̇, where Q is the charge. The voltage V can be written magnetic field has a huge number of modes which all can
V = Φ̇ where Φ is the magnetic flux. (This form will be be considered as independent harmonic oscillators.
convenient later, but not necessary at this point.) For the
familiar electric elements we have the relations: capacitor We notice that all the equations of motion (7), (8), (9),
Q = CV , inductor Φ = LI, and resistor V = RI. We study and (11) are of the same form, so we need to study only
first the series oscillator shown in the figure. It is driven by one of them. Although we use the notation of mechanical
a voltage source (the voltage is independent of the current oscillator in the following, it should be remembered that
flowing through). the analysis is valid for all kinds of harmonic oscillators.
p
We define ω0 = k/m. Thus k = mω02 and the equations
L C R of motion can be written in the form

ṗ = −mω02 q − γp + F
V p
q̇ = . (12)
m
For this we deduce
A more symmetric form is
X Q
V = Vi = + RQ̇ + LQ̈. (8) ṗ √ p F
i
C √ = −ω0 mω0 q − γ √ +√
mω0 mω0 mω0
Its dual, the parallel oscillator, is shown in the following √ p
mω0 q̇ = ω0 √ . (13)
figure. It is driven by a current source (the current is inde- mω0
pendent of the voltage over the source).
We define the complex parameters
1 √ p
I L C R α = √ ( mω0 q + i √ )
2~ mω 0
1 √ p
α∗ = √ ( mω0 q − i √ ). (14)
2~ mω0
For this we find similarly
Think here ~ simply as some number, there is no quantum
X Φ̇ Φ mechanics involved yet. In these variables the equations of
I= Ii = C Φ̈ + + . (9)
R L motion (12) get an alternative form
i

(Work out the details as an exercise.) γ γ iF

α̇ = −iω0 α − α + α∗ + √ (15)
2 2 2~mω0
The Maxwell equations
γ γ iF
ρ α̇∗ = iω0 α∗ − α∗ + α − √ .
∇·E = , 2 2 2~mω0
0
∂B The latter equation is the complex conjugate of the former,
∇×E =− , and therefore one of them is sufficient. The solution in the
∂t

4
absence of driving and damping is α(t) = α(0)e−iω0 t , i.e. Exercise: plot the real and imaginary part of α(t) as a
a rotation in the complex α plane with angular frequency function of time, and depict the trajectory of α(t) in the
ω0 . complex plane for
In the following we assume the driving force is harmo- a) f0 = 0, α(0) = 1, ω0 = ωd = 1, γ = 0.1;
nic, F (t) ∝ cos(ωd t). Note that this is no real restriction, b) f0 = 1, α(0) = 0, ω0 = ωd = 1, γ = 0.1, and interpret
since any function F (t) can be represented as a Fourier se- the results.
ries, and because the equation of motion (15) is linear, the Let us calculate the energy stored in the oscillator as the
solutions corresponding to each Fourier component can be sum of kinetic and potential energies
solved separately, and finally summed together. For nota-
tional beauty we also write the prefactor in a convenient p2 1
E= + mω02 q 2 . (22)
form. So we study a driving force of the form 2m 2
p
F (t) = 2~mω0 f (t), Expressing the energy in terms of α (14) gives the very
f (t) = 2f0 cos(ωd t). (16) suggestive form
E = ~ω0 |α|2 . (23)
The equations of motion (7) or (15) have an exact so-
Note however that at this point ~ is just some arbitrary
lution (Exercise). However, later we often need a rotating
normalization factor that also appears in the definition of
wave approximation, and it is appropriate to introduce it
α (14) and thus it cancels out in the energy (23).
already here. The assumption is that the driving and dis-
sipative terms are small relative to the first terms on the Calculating the time derivative of E (23) and using the
right hand side of (15). For that we transform to a frame equation of motion (19) one gets
rotating at the drive frequency by writing
dE
= Pabs − Pdiss ,
α(t) = β(t)e−iωd t . (17) dt
Pabs = 2~ω0 =(f0 eiωd t α),
Substitution gives
γ γ 2iωd t ∗ Pdiss = ~ω0 γ|α|2 , (24)
β̇ = −i(ω0 − ωd )β − β + e β
2 2 where = denotes the imaginary part. Here the first term on
+if0 (1 + e2iωd t ). (18) the right hand side can be interpreted as absorption, the
If the driving frequency is near resonance, ωd ≈ ω0 , the power transferred from the driving force to the oscillator.
first term on the right hand side is small, and supposedly The latter term, which always is negative, is the dissipation,
β(t) changes slowly in time. In contrast, the two “counter the power lost in friction.
rotating” terms contain a rapidly oscillating factor e2iωd t . Let us calculate the absorption and dissipation in the
The claim is that the effect of these average out on the time steady state (20). We find that they are equal, and given
scale we are interested in, which is determined by “secular” by
non-oscillating terms on the right hand side. Therefore we
simply drop out the rapidly oscillating terms. (The accu- ~ω0 γf02
Pabs = Pdiss = . (25)
racy of the rotating wave approximation could be tested (ω0 − ωd )2 + 14 γ 2
by comparison with the exact solution, or by doing pertur-
bation theory with respect of the counter-rotating terms.) One often thinks the absorbed power as function of the
Transforming back to α we have the equation of motion drive frequency, Pabs (ωd ). Its maximum corresponds to
for a classical harmonic oscillator in the rotating wave ap- resonance, where the drive frequency equals the natural
proximation frequency of the oscillator, ωd = ω0 . The form of Pabs (ωd )
γ in Eq. (25) is called Lorentzian line shape. Its full width at
α̇ = −iω0 α − α + if0 e−iωd t . (19) half maximum (often abbreviated as FWHM) equals γ.
2
Pabs
The equation motion (19) can be solved without dif-
ficulties. The steady state solution is obtained by assuming Ñ Ω0 f02 Γ
4
α = constant × e−iωd t , and one finds
f0 e−iωd t 3
αss (t) = . (20)
ω0 − ωd − 2i γ
The general solution of an initial value problem is 2

γ
α(t) = αss (t) + [α(0) − αss (0)]e−iω0 t− 2 t
1
f0 γ
= (e−iωd t − e−iω0 t− 2 t )
ω0 − ωd − 2i γ Ωd - Ω0
−iω0 t− γ2 t -3 -2 -1 1 2 3
Γ
+α(0)e . (21)

5
 It is also instructive to look at We see that a† is the Hermitian conjugate of a as i† = i∗ =
−i. By direct calculation we can now verify that
f0
eiωd t αss (t) ≡ αss (0) =
ω0 − ωd − 2i γ [a, a† ] = 1. (31)
i
f0 (ω0 − ωd + 2 γ)
= 1 2 = Aeiδ (26) From this relation it follows that the Hilbert space where
(ω0 − ωd )2 + 4γ a and a† operate is spanned by states of the form
in the complex plane. We decompose this complex number |ni, n = 0, 1, 2, . . . , ∞, (32)
into real amplitude A and phase δ. Here δ is the phase shift
between the oscillator spatial coordinate q and the driving and the following relations determine how a and a† operate
force f . For small driving frequencies δ ≈ 0, i.e. the “par- on these,
ticle” just tilts in the direction of the force [as is evident
from (7) when the derivative terms can be neglected]. With a† a|ni = n|ni,
√
increasing drive frequency δ grows and gets the value π/2 a|ni = n |n − 1i,
at resonance. It approaches π for ωd → ∞, i.e. the particle †
√
a |ni = n + 1 |n + 1i. (33)
tilts in the direction opposite to the force. Apart from cons-
tant factors, the imaginary part of αss (0) is the absorption Interpretation: a† a gives the number of quanta, a annihila-
(25). The real part of αss (0) makes the phase shift to differ tes one quantum and a† creates one quantum. The states
from π/2 and is called dispersion. (32) are known as number states or Fock states.
Exercise: plot the absorption (simply =[αss (0)]), disper- For completeness, we give a proof of results (32) and
sion and the phase shift δ as a function of ωd , and depict (33).
the trajectory of αss (0) in the complex plane for f0 = 1, Proof. (read at home) We study the operator
ω0 = 1, γ = 0.1 and interpret the results.
ň = a† a. (34)
One often used quantity is the quality factor or Q. It is
a dimensionless number defined as the resonance angular It is Hermitian (verify). The eigenvalues of a Hermitian
frequency times the energy stored divided by the dissipated operator are real valued. Let us label the eigenstates of the
power. Using expressions above we calculate operator ň by using the eigenvalue n (a real number)
ω0 E ω0
Q= = . (27) ň|ni = n|ni. (35)
Pdiss γ
We assume the eigenstates are normalized, hn|ni = 1. We
The validity of the rotating wave approximation requires
see that the eigenvalue n cannot be negative,
Q  1.
X
n = hn|ň|ni = hn|a† a|ni = hn|a† |mihm|a|ni
2.2 Quantization X
m
2
= |hm|a|ni| ≥ 0. (36)
Simple harmonic oscillator m

Consider now the harmonic oscillator (12) but without We easily calculate from (31) that
damping and driving,
[ň, a] = −a. (37)
ṗ = −mω02 q
p This implies
q̇ = . (28)
m
ň(a|ni) = aň|ni − a|ni = (n − 1)(a|ni), (38)
Quantum mechanics makes p and q operators which are
Hermitian, p† = p, q † = q. They have to satisfy the com- and we see that a|ni either is a state corresponding to the
mutation relations eigenvalue n − 1, i.e. a|ni = c|n − 1i, or else a|ni = 0. The
latter alternative implies ň|ni = 0, so it is possible only if
[q, p] = i~, (29) n = 0. In the former case normalization gives

where [A, B] = AB − BA is a commutator. Notice that |c|2 = hn|a† a|ni = n, (39)

[A, A] ≡ 0 and therefore [q, q] = [p, p] = 0.
and therefore we fix
We make the same change of variables as above (14), but √
instead of α and α∗ we write a and a† , a|ni = n |n − 1i. (40)
1 √ p If one operates sufficiently many times with a, one should
a = √ ( mω0 q + i √ )
2~ mω 0 arrive at negative eigenvalues, which is in contradiction
1 √ p with equation (36). The way out of this is that n is an
a† = √ ( mω0 q − i √ ). (30)
2~ mω0 integer, so that a|0i = 0, and the process (40) ends. (Note

6
the essential difference between the n = 0 eigenstate |0i These consists of the exponential (49) multiplied by Her-
and the zero of the linear space 0.) Correspondingly one mite polynomials and normalization constants.
can deduce results for a† , and we get the results (33). E
The energy of the oscillator (22) is the same as the Ha- Ñ Ω0
miltonian of the simple harmonic oscillator
6
p2 1 2 2
H= + mω0 q . (41) 5
2m 2
Writing it in terms of a and a† (14) gives 4
  3
1
H = ~ω0 a† a + . (42) 2
2
This allows us to easily find the energy eigenvalues of the 1
m Ω0
oscillator using (33). The eigenstates are |ni (32) q
-4 -2 0 2 4 Ñ
H|ni = En |ni (43)
Figure: the parabolic potential energy, the 6 lowest ener-
and the eigenenergies are gy eigenvalues (horizontal lines) and the corresponding wa-
  ve functions (50).
1
En = ~ω0 n + . (44) Ψ20
2
We see the energy is quantized to multiples of ~ω0 . n is 0.4
interpreted as the number of quanta in the oscillator. The
lowest energy state |0i satisfies a|0i = 0. It corresponds to 0.2
no quanta, n = 0, and zero point energy E = ~ω0 /2. The
higher eigenstates can be generated from this by repeated m Ω0
q
application of a† , -10 -5 5 10 Ñ
(a† )n -0.2
|ni = √ |0i. (45)
n!
-0.4
Above we derived the energy eigenstates of the harmo- Figure: the wave function ψ20 (q).
nic oscillator in the abstract form (45). The more usual
Schrödinger approach also gives the explicit wave functions
Forced harmonic oscillator
ψn (q) = hq|ni. (46) We include the driving term by postulating a hamilto-
nian
These are hardly needed in this course, but for complete- p2 1
ness let us construct them as well. H= + mω02 q 2 − F (t)q. (51)
2m 2
In the Schrödinger approach one constructs explicit Here the first two terms are the same as the energy of
representation for the operators q and p so that the for- undriven oscillator (22), and the last one is a new term
mer is simply multiplication by q and the latter describing driving. Applying the classical Hamilton equa-
tions
d ∂H ∂H
p = −i~ . (47) ṗ = − , q̇ = (52)
dq ∂q ∂p
Substituting these to a (30) and applying to a|0i = 0 gives we get the same equations of motion as above (12) except
the damping term. This justifies the Hamiltonian (51).
√ ~ d
( mω0 q + √ )ψ0 (q) = 0. (48) Quantizing p and q, and making the change of variables
mω0 dq
(14) gives
The normalized solution of this is  
† 1
 mω  14  mω  H = ~ω0 a a + − ~(a + a† )f (t). (53)
0 0 2 2
ψ0 (q) = exp − q . (49)
π~ 2~
This form is exact. However, in the following we often make
All higher states can be obtained by using (45) again the rotating wave approximation. This is best jus-
r r !n tified going back to the classical Hamiltonian (since we
1 mω0 ~ d
ψn (q) = √ q− ψ0 (q). (50) have not yet discussed the time dependence in the quan-
n! 2~ 2mω0 dq tum case) by replacing a and a† by α and α∗ . Using again

7
α(t) = β(t) exp(−iωd t) with slow β, we see that the driving the oscillators is constructed in analogy with the driving
term in (53) is composed of two rapidly oscillating terms term so that instead of a fixed function f (t) the time de-
and two secular terms. Dropping the former and changing pendence comes from dynamical variables Γ + Γ† , which
back to quantized notation gives depend linearly on bk ’s and b†k ’s with coupling constants
  gk . Such a bilinear coupling keeps the quadratic form of
1
H = ~ω0 a† a + − ~f0 (a† e−iωd t + aeiωd t ). (54) the Hamiltonian, which makes it not too difficult to solve.
2
While (55) should be considered as a model of friction
We see that the forcing causes transitions between the in the case of a mechanical oscillator, it is more than that
eigenstates of the simple harmonic oscillator. for an electromagnetic resonator. The preferred oscillator
is a mode in the resonator, and it couples to a continuum
Although we could, we are not interested to work on this of electromagnetic modes outside of the resonator.
further before we also have the damping term.
The Hamiltonian looks complicated. There is an exact
solution, at least in the absence of forcing: one can trans-
Including damping form to normal coordinates (remember Analytic mechanics
The damping term −γp in the equations of motion (12) course). In these, the system just consists of oscillators that
cannot be reproduced by any simple term in a Hamilto- can be solved independently. This solution is unsatisfacto-
nian. Therefore simple minded quantization of the damped ry for several reasons. 1) the transformation to the normal
harmonic oscillator fails. coordinates and back is complicated, 2) the model contains
a lot of microscopic details (ωk , gk ) and we would like to
To have success, we have to think what friction real-
get rid of these as much as possible, 3) we want to apply
ly is. It means coupling of macroscopic variables to some
the damping also to other systems than the harmonic oscil-
microscopic degrees of freedom. Energy is then transferred
lator.
from the macroscopic variables to the microscopic ones. If
the number microscopic degrees of freedom is large and In order to make progress, we study more quantum mec-
they are sufficiently disordered, the energy will effective- hanics and introduce the density operator.
ly not return back to the macroscopic degrees of freedom.
Macroscopically this is seen as dissipation, loss of energy.
In order to describe dissipation in quantum mechanics,
the obvious way is to include the microscopic degrees of
freedom. In a strict sense this means that we have to know
what they are. However, we could guess that the gross
features of dissipation are independent of the details of
the microscopic world. Therefore we could use a model for
them. The simplest model we know is the harmonic oscil-
lator.
Thus we construct the Hamiltonian

H = H0 + Hd + Hbath + Hint , (55)

1
H0 = ~ω0 (a† a + ), (56)
2
Hd = −~(a + a† )f (t), (57)
~ωk b†k bk ,
X
Hbath = (58)
k
X
Hint = −~(a + a† )(Γ + Γ† ), Γ= gk bk . (59)
k
Here H0 is the Hamiltonian of our preferred harmonic
oscillator, Hd is the forcing term (53), Hbath describes a
set of microscopic harmonic oscillators, and Hint gives the
coupling between the macroscopic and microscopic oscilla-
tors (we neglect the zero point energy). The operators bk
and b†k are the annihilation and creation operators for the
microscopic oscillators with frequencies ωk and satisfy

[bk , b†l ] = δkl , [bk , bl ] = [b†k , b†l ] = 0, (60)

i.e. all commutators having two different oscillators (inclu-

ding also the preferred one) vanish. The coupling between

8
3. Quantum mechanics Let A be a linear operator in the ket space, A|bi = |ci.
The linearity means

3.1 Vector spaces A(|bi + |ci) = A|bi + A|ci,

Let us briefly remind about quantum mechanics using A(λ|bi) = λA|bi. (68)
the Dirac notation. We already used the notation |ni for the The same operator can be made to operate “backwards”
harmonic oscillator states. In general we assume a linear in the bra space, hc|A, by requiring
vector space formed by elements |ai called state vectors or
states. The assumption about linear vector space means (hc|A)|bi = hc|(A|bi). (69)
that addition and multiplication by a scalar are defined
The Hermitian conjugate of A is denoted by A† . It has the
|ai + |bi = |ci, λ|di = |f i, (61) property
hb|A|ci = hc|A† |bi∗ . (70)
with some elements |ci and |di also belonging to the vector
space, and λ is any complex number. With vector space we The operator A is Hermitian if
require the standard rules
hb|A|ci = hc|A|bi∗ . (71)
|ai + |bi = |bi + |ai, †
Exercise: check that the operators a and a (33) are Her-
(|ai + |bi) + |ci = |ai + (|bi + |ci), mitian conjugates and a† a is Hermitian.
λ(µ|di) = (λµ)|di, If
λ(|ai + |bi) = λ|bi + λ|ai. (62)
A|bi = λ|bi, (72)
In addition, we define another linear vector space formed it is said that |bi is an eigenstate of A with eigenvalue λ. It
by states ha| so that corresponding to any “ket” state |ai can be shown that for a Hermitian operator all eigenvalues
there is a “bra” state ha|, and vice versa, satisfying are real. Two states |bi and |ci having different eigenvalues
λ 6= µ with a Hermitian operator, are orthogonal, hc|bi = 0.
|ai + |bi = |ci ⇔ ha| + hb| = hc|,
Let |ek i be an orthonormal basis, i.e. any state vector
λ|di = |f i ⇔ λ∗ hd| = hf |. (63)
can be represented as
We see that the bra space is a copy of the ket space with
X
|ai = ck |ek i, hel |ek i = δl,k . (73)
the only difference that complex conjugation appears in the k
multiplication by a scalar. Using this, we further require
the vector space to form a Hilbert space, so that there exist The coefficients are then given by ck = hek |ai. In the fol-
inner product of two state vectors ha| and |bi and denoted lowing we also assume |ai is normalized, ha|ai = 1.
by Physical quantities correspond to operators in quantum
mechanics. Suppose a measurement corresponding to ope-
ha|bi ≡ ha||bi = λ (64) rator B. The possible values obtained in a measurement
are the eigenvalues λ of the operator B. Suppose we se-
with a complex number λ. With inner product one requires lect the basis |e i (73) from eigenstates of the operator B.
k
that Then the probability to find the eigenvalue λ is given by
the sum of the squares of the coefficients, |ck |2 of the ei-
ha|(|bi + |ci) = ha|bi + ha|ci,
genstates corresponding to the eigenvalue λ. The average
ha|(λ|bi) = λha|bi, value, denoted by hBi, in state |ai is given by
ha|bi = hb|ai∗ ,
hBi = ha|B|ai. (74)
ha|ai = 0 ⇔ |ai = 0. (65)
Suppose that all eigenvalues of a Hermitian operator are
The first two mean linearity with respect to the ket state.
different i.e. the spectrum is nondegenerate. This allows to
Combined with the third these imply linearity also with
label the eigenstates by the eigenvalue, |λi. It also allows to
respect to the bra state.
represent an arbitrary state |ai by the set of numbers hλ|ai.
Exercise: check that the states A special case of this is the wave function ψa (q) = hq|ai
∞
already used above (46).
X
λn |ni, (66)
n=0 3.2 Time depedence
where λn are complex and |ni (32), form a linear vector Based on direct correspondence with classical mechanics,
space. Show that this is Hilbert space if one defines the time dependence in quantum mechanics is given by
dA ∂A
hm|ni = δm,n . (67) i~ = [A, H] + i~ . (75)
dt ∂t

9
Here H is the Hamiltonian operator and A = These are precisely the same as the Heisenberg picture
A(q1 , q2 , . . . , qn , p1 , p2 , . . . pn , t) is an arbitrary operator, equations discussed above (75). Thus this transformation
which depends on the canonical operators qi and pi (i = also justifies the Schrödinger picture postulated above.
1, . . . , n), and, in some cases, also can depend explicitly on One more picture is useful in the following. We decom-
time, giving rise to the second term on the right hand side pose the Hamiltonian
of (75). An example of such an operator is the driving term
in (51), A = −F (t)q. H = H0 + H1 , (85)
As discussed above, the operators are not directly obser-
where H0 usually is some simple, time-independent part
vable quantities, only their eigenvalues. This has the con-
and H1 is more complicated “interaction” part of H. Let
sequence that time dependence in quantum mechanics can
us fix U (t) by the simple part only
be formulated in different ways called pictures so that all
observable quantities are still the same. The equation of dU i
motion above (75) is valid in Heisenberg picture. In this i~ = H0 U ⇔ U = exp(− H0 t). (86)
dt ~
picture the state vectors have no time dependence [as all is
expected to be in (75)]. In alternative pictures part of the Then (81) reduce to equations in the interaction picture
time dependence is transferred to the state vectors. For d|a0 i
that purpose assume a unitary time-dependent operator i~ = H10 |a0 i (87)
dt
U (t). The unitarity means dA0 ∂A 0
† † i~ = [A0 , H0 ] + i~ , (88)
UU = 1 = U U (76) dt ∂t
with a unity operator 1. If we now define new operators where H00 = H0 .
and state vectors by As an example consider a and a† taking as H0 the simple
A0 = U † AU, |a0 i = U † |ai, (77) harmonic oscillator Hamiltonian (42). Using (88) we find
we see that all expectation values are unchanged, a(t) = ae−iω0 t , a† (t) = a† eiω0 t , (89)

hb0 |A0 |c0 i = hb|A|ci. (78) where the time dependence in a(t) and a† (t) denotes
interaction picture. a and a† are the time-independent
The same applies to the eigenvalues, i.e. there is no physical
Schrödinger picture operators. (Exercise)
change. However, the equations for the new state vectors
and operators are changed. Now it is most convenient to
first postulate the following equations in the Schrödinger 3.3 Density operator
picture In order to simplify the description of damping, we ha-
d|ai ve to average over the environmental degrees of freedom.
i~= H|ai (79) However, in quantum mechanics we cannot simply take an
dt
dA ∂A average over a wave function (needs justification). Instead,
i~ = i~ . (80) we first form a density operator, and this allows taking
dt ∂t
averages.
Here the state vectors have time dependence determined
by the Hamiltonian H, but the operators only have the The density operator is the operator
explicit time dependence. Making the transformation (77) X
we have ρ= pi |ai ihai |. (90)
i
d|a0 i dU †
i~ = (U † HU + i~ U )|a0 i Here pi is the probability that the system is in state
dt dt
dA0

dU †

∂A |ai i. The state vectors |ai i are assumed to be normalized,
0
i~ = A , −i~ U + U † i~ U. (81) hai |ai i = 1. Also the probabilities have to be normalized,
dt dt ∂t P
i pi = 1. The case that only one state is possible is called
(Exercise) pure state and the density operator reduces to
Let us fix U (t) by the condition
ρ = |aiha|. (91)
dU
i~ = HU, (82) The nice thing with the density operator is that the expec-
dt
tation value of any operator B can be calculated as
and we define ∂A0 /∂t = U † (∂A/∂t)U . [In order to see that
this is a reasonable definition consider the example case hBi = Tr(Bρ), (92)
A = −F (t)q]. Then (81) reduce to
where the trace Tr means the sum over the basis vectors
d|a0 i
i~ = 0 (83) of an arbitrary orthonormal basis |ek i,
dt X
dA0 0 0 ∂A0 Tr(. . .) = hek | . . . |ek i. (93)
i~ = [A , H ] + i~ . (84)
dt ∂t k

10
In order to show (92) we calculate Here β = 1/k
P B T and Z is determined by the normalization
XX condition i pi = 1 or (96). Applying this to the simple
Tr(Bρ) = hek |ai ipi hai |B|ek i harmonic oscillator (42) gives that the average number of
k i
X quanta is given by the Bose distribution
= pi hai |B|ai i, (94)
1
i hni = ≡N (102)
eβ~ω0 − 1
which is the expectation value (74) each state being weigh-
ted by its probability pi . and the density operator is
∞
A useful property of the trace is that for any operators X
A and B ρ = (1 − e−β~ω0 )e−β~ω0 n |nihn|
n=0
∞ n
Tr(AB) = Tr(BA). (95)

X 1 N
= |nihn|. (103)
n=0
1 + N 1+N
Exercise: show that
(Exercise)
Tr ρ = 1. (96)

In order to get some insight, let us calculate the density 3.5 Correlation functions
operator corresponding to the state |ai = a1 |e1 i + a2 |e2 i Sometimes important information can be expressed by a
in the basis (|e1 i, |e2 i). We get the density matrix correlation function
|a1 |2 a1 a∗2
   
ρ11 ρ12 hA(t2 )B(t1 )i, (104)
ρ= = . (97)
ρ21 ρ22 a∗1 a2 |a2 |2
where two different times appear. This notation implicitly
This pure state should be contrasted with a “mixture” of
says that the operators should be expressed in the Heisen-
states |e1 i and |e2 i, where the two states appear with clas-
berg picture.
sical probabilities p1 and p2 . This gives
    Exercise: Calculate ha(t)i, ha† (t)i, and the correlation
ρ11 ρ12 p1 0 † †
ρ= = . (98) functions ha(t)a (0)i and ha (t)a(0)i for simple harmonic
ρ21 ρ22 0 p2 oscillator in equilibrium at temperature T and try to in-
The essential difference is the presence of off-diagonal com- terpret the result (for T = 0 and T > 0).
ponents ρ12 and ρ21 in the “coherent” superposition (97).
(Exercise)
We wish to have dynamical equation for the density
operator. Taking the derivative of ρ (90) and using the
Schrödinger equation (79) we get
dρ
i~ = [H, ρ] . (99)
dt
This is known as von Neumann equation. Note that the
density operator is exceptional that it does not follow the
general rule for operators in the Schrödinger picture (80).
Rather the equation (99) resembles the Heisenberg-picture
equation (84), but has the opposite sign. In the interaction
picture (87) one gets
dρ0
i~ = [H10 , ρ0 ] (100)
dt
i i
H10 = exp( H0 t)H1 exp(− H0 t).
~ ~
(Exercise)

3.4 Equilibrium statistical mechanics

In equilibrium at temperature T the probability of each
state is given by the Gibbs distribution
1 −βEi 1
pi = e ⇔ ρ = e−βH . (101)
Z Z

11
4. Master equation where the time development operator

l(t) = (U̇1 (t) + U̇2 (t) + . . .)U −1 (t). (113)

4.1 Derivation
We are now ready to go back to the model of damping We assume that the interaction term is such that
(55). Our program is to formulate a dynamical equation for
the density operator of the whole system, solve it assuming TrR [V (t)ρR ] = 0. (114)
weak coupling between the macroscopic and microscopic
degrees of freedom, and then average over the microscopic This implies that U1 = 0. It follows that up to second order
ones. This should lead to a master equation for the density
operator of the macroscopic system only. (WM, section 6.) l(t) = U̇2 (t)
Z t
For generality, write 1
= − 2 dt1 TrR [V (t), [V (t1 ), ρR (0)(·)]],
~ 0
H = HS + HR + V (105)
and thus
with HS denoting system, HR the reservoir and V their
coupling. Let w(t) denote the density operator of the dρ(t) 1
Z t
complete system. In the interaction picture it satisfies = − dt1 TrR [V (t), [V (t1 ), ρR (0)ρ(t)]].
dt ~2 0
dw(t) (115)
i~ = [V (t), w(t)] . (106)
dt
The reduced density opeator of the system is We now specify to the system-reservoir coupling (59).
First we change to the interaction picture. Secondly, we
ρ(t) = TrR [w(t)], (107) make again the rotating wave approximation. It will be
where TrR marks the trace over reservoir degrees of free- seen that the most important case is that all frequencies
dom only. We assume that at initial time t = 0 the system are of similar magnitude †ωd ∼ ω0 ∼ ωk . Thus we can drop
and the reservoir are uncorrelated, terms of type abk and a† bk . We are left with the interaction

w(0) = ρ(0)ρR (0). (108) V (t) = −~(a† Γ(t)eiω0 t + aΓ† (t)e−iω0 t ),

X
We integrate (106) Γ(t) = gk bk e−iωk t . (116)
Z t k
i
w(t) = w(0) − dt1 [V (t1 ), w(t1 )] . (109)
~ 0 Substituting into (115) we get
Iterating this gives Z t
dρ(t)
∞ t t1 tn−1
= − dt1 TrR [(a† Γ(t)eiω0 t + aΓ† (t)e−iω0 t ),
dt
Z Z Z
i X 0
w(t) = w(0) + (− )n dt1 dt2 . . . dtn
n=1
~ 0 0 0 [(a† Γ(t1 )eiω0 t1 + aΓ† (t1 )e−iω0 t1 ),
×[V (t1 ), [V (t2 ), . . . [V (tn ), w(0)] . . .]]. (110) ρR (0)ρ(t)]]. (117)

Taking the trace over the reservoir gives Writing out also the double commutator, there will be 16
∞ different terms in total. Let us look at one of them,
i n t
X Z Z t1 Z tn−1
ρ(t) = ρ(0) + (− ) dt1 dt2 . . . dtn
~ 0 0 0
Z t
n=1
(−1) 2
dt1 TrR [aΓ† (t)e−iω0 t ρR (0)ρ(t)a† Γ(t1 )eiω0 t1 ]
×TrR [V (t1 ), [V (t2 ), . . . [V (tn ), ρ(0)ρR (0)] . . .]] 0
Z t
= (1 + U1 (t) + U2 (t) + . . .)ρ(0) †
= aρ(t)a dt1 TrR [Γ† (t)ρR (0)Γ(t1 )]e−iω0 t eiω0 t1
= U (t)ρ(0), (111) 0
Z t
where Un (t) is an operator acting on ρ, = aρ(t)a† dt1 TrR [ρR (0)Γ(t1 )Γ† (t)]e−iω0 (t−t1 )
Z t Z t1 Z tn−1 0
i = aρ(t)a† I, (118)
Un (t) = (− )n dt1 dt2 . . . dtn
~ 0 0 0
where we have used the cyclic property of trace (95). We see
×TrR [V (t1 ), [V (t2 ), . . . [V (tn ), ρR (0)(·)] . . .]].
that the trace is the expectation value (92) in the reservoir
(Exercise: commutator as a linear operator) We now take and we write
the time derivative
Z t
dρ(t) I = dt1 hΓ(t1 )Γ† (t)ie−iω0 (t−t1 )
= (U̇1 (t) + U̇2 (t) + . . .)ρ(0)
dt 0
Z t
= (U̇1 (t) + U̇2 (t) + . . .)U −1 (t)ρ(t)
gk gl∗ hbk b†l iei(−ωk t1 +ωl t) e−iω0 (t−t1 ) .
X
= dt1
= l(t)ρ(t), (112) 0 k,l

12
Thus we have to evaluate the expectation value hbk b†l i for where
the reservoir oscillators. Assuming they initially are in ther-
mal equilibrium, we can use (103) for each oscillator. Also γ = 2πρ(ω0 )|g(ω0 )|2 (127)
using (60) we get and
hbk b†l i = δk,l [1 + N (ωk )] (119) ∞
Z
1
∆=P d ρ(ω0 + )|g(ω0 + )|2
−∞ 
t
×[1 + N (ω0 + )].
Z X (128)
I = dt1 |gk |2 [1 + N (ωk )]ei(ωk −ω0 )(t−t1 ) .
0 k
Physically ∆ is the frequency shift of the system caused
Assuming the reservoir oscillators have continuous spect- by coupling to the reservoir. Usually this is a small effect,
rum with density ρ(ω) and we neglect it here. Later in studying a two-level system
Z t Z ∞ we see that this shift is similar to the Lamb shift observed
I = dt1 dω1 ρ(ω1 )|g(ω1 )|2 [1 + N (ω1 )] in atoms.
0 0
One term in the master equation (117) is thus given by
×ei(ω1 −ω0 )(t−t1 ) . (120)
(118) and (126). The task that remains is to calculate the
Changing the integration variables we have rest 15 of the 16 terms. Eight of the terms vanish because
Z t Z ∞ averages hbk bl i and hb†k b†l i vanish in a thermal state (103).
I = dτ dω1 ρ(ω1 )|g(ω1 )|2 The 7 remaining terms can be calculated with slight modi-
0 0 fications of the above. After that we finally get the master
×[1 + N (ω1 )]ei(ω1 −ω0 )τ equation for damped harmonic oscillator in the interaction
Z t Z ∞ picture
= dτ dρ(ω0 + )|g(ω0 + )|2
dρ γ
0 −ω0 = (N + 1)(2aρa† − a† aρ − ρa† a)
×[1 + N (ω0 + )]eiτ dt 2
γ
Z t Z ∞ + N (2a† ρa − aa† ρ − ρaa† ). (129)
= dτ df ()eiτ (121) 2
0 −ω0 Recall that N is the Bose occupation at the oscillator
with f () = ρ(ω0 + )|g(ω0 + )| [1 + N (ω0 + )]. The in- frequency (102), and γ is a new parameter describing dam-
2

tegrand oscillates which leads to cancellations. The main ping, as soon will be seen. If the reservoir is at zero tem-
contribution comes from ||τ . 1. Assuming tω0  1, we perature, T = 0, we have N = 0 and (129) reduces to
can then extend the integrations to infinity, but simulta- dρ γ
neously we add a convergence parameter η > 0 to guaran- = (2aρa† − a† aρ − ρa† a). (130)
dt 2
tee that there is no contribution from τ = ∞,
Z ∞ Z ∞ If there are other “interaction” terms in the Hamilto-
I = dτ df ()e(i−η)τ (122) nian of the system, they can be included according to the
0 −∞
general rules of density operator in the interaction picture
and in the end we take the limit η → 0. Making the time (100). In particular, the forcing term (57), after a rota-
integration gives ting wave approximation, can be added to give the master
Z ∞ equation for forced and damped harmonic oscillator in the
f ()
I = i d . (123) interaction picture
−∞  + iη
dρ i
We use the symbolic identity = − [HdI , ρ]
dt ~
1 1 γ
= P ∓ iπδ(), (124) + (N + 1)(2aρa† − a† aρ − ρa† a)
 ± iη  2
γ
+ N (2a† ρa − aa† ρ − ρaa† ), (131)
where P denotes the principal value defined by 2
Z ∞ HdI = −~f0 [a† (t)e−iωd t + a(t)eiωd t ], (132)
f ()
P d
−∞  with a(t) and a† (t) in the interaction picture (89).
Z −a Z ∞ 
f () f () Finally, we can also write the master equation in the
= lim d + d (125)
a→0 −∞  a  Schrödinger picture
(Exercise). We get dρ i
= − [H0 + Hd , ρ]
Z ∞ dt ~
f () γ
I = iP d + πf (0) + (N + 1)(2aρa† − a† aρ − ρa† a)
−∞  2
γ γ
= i∆ + [N (ω0 ) + 1], (126) + N (2a† ρa − aa† ρ − ρaa† ) (133)
2 2

13
with (56) and (57). More explicity (Exercise). In the absence of forcing the solution is
dρ
= −iω0 [a† a, ρ] + if0 [a† e−iωd t + aeiωd t , ρ] hn(t)i = N + (hn(0)i − N )e−γt . (140)
dt
γ
+ (N + 1)(2aρa† − a† aρ − ρa† a) Even if initially hni = 0, the occupation will grow
2
γ and approaches exponentially the Bose occupation N =
+ N (2a† ρa − aa† ρ − ρaa† ). (134) (e~ω0 /kB T − 1)−1 . This can be understood that the reser-
2
voir at temperature T > 0 has excitations, and these will
Let us comment the derivation of the master equation. excite the system as well, until a thermal equilibrium is
We assumed the interaction between system and reservoir reached so all oscillators are at the same temperature.
as weak. The coupling was taken into account to second or-
We see that hai vanishes in the thermal equilibrium. This
der in the interaction, which is known as Born approxima-
can also be seen directly from the equilibrium ρ (103),
tion. Note that the second order approximation was made
which only has diagonal elements. This can be unders-
in expressing ρ̇ against ρ at the same instant (112). Compa-
tood so that there are quanta present in the oscillator, but
red to the system at the initial time t = 0 terms of all orders
their phases are random, leading to vanishing hai. Clas-
are included. (Even though the interaction is weak at each
sically we can think an ensemble of different αk = Ak eiφk .
instant, it can have a huge effect over long time.) Another
If their phases φk are random we have hαi = 0 but nonzero
less explicit assumption is called Markovian, the time deri-
h|α|2 i = hA2 i, which is the classical quantity corresponding
vative ρ̇(t) only depends on ρ(t), not any longer history of
to hni.
ρ. Thus the reservoir is assumed to forget any effect caused
upon it more rapidly than the system can change its state. It is also possible to calculate correlation functions (104).
We also used the rotating wave approximation. The form of The trick used to calculate correlations like
the damping term in the master equations above is known
as Lindblad form. hA(t)B(0)i, (141)

4.2 Applications is called quantum regression theorem. We write the full

trace of the Heisenberg operators
Averages
hA(t)B(0)i = Tr[U † (t)A(0)U (t)B(0)w(0)]
As first step learning the master equation, let us calcula- = Tr[A(0)U (t)B(0)w(0)U † (t)]. (142)
te some expectation values. Consider the annihilation ope-
rator, The last form has Schrödinger picture time-independent
dhai d operator A(0) and the time-dependent density matrix with
= Tr(aρ). (135) the only difference that B(0) appears at initial time t = 0.
dt dt
In the derivation of the master equation we did not use
This can be calculated in any picture and the result is the
any special property of the density operator. Thus we can
same. In the Schrödinger picture a is time independent and
calculate (142) as above but using the initial value B(0)ρ(0)
dhai dρ for the density operator instead of ρ(0). As an example,
= Tr(a ) (136)
dt dt calculate
(Exercise). Substituting from (134) and using (31) and (95) dha† (t)a(0)i
one finds = iω0 ha† (t)a(0)i − if0 ha(0)ieiωd t
dt
dhai γ γ
= −iω0 hai − hai + if0 e−iωd t . (137) − ha† (t)a(0)i. (143)
dt 2 2
This equation is precisely the same as the classical equation The solution in thermal equilibrium, f = 0 is
0
(19) for α. This allows to identify the coefficient γ in the
master equation as the same damping coefficient as in the γ
ha† (t)a(0)i = N eiω0 t− 2 |t| , (144)
classical theory.
The results for α can now be translated to hai. For i.e. thermal equilibrium shows nonvanishing correlations
example, in the absence of forcing, f0 = 0, we have ex- that vanish in time exponentially.
ponential decay
γ
ha(t)i = ha(0)ie−iω0 t− 2 t . (138) Number representation
Let us now study the full density matrix, not only the
Let us apply the same prescription to the number ope-
averages. One obvious possibility is to represent the density
rator n = a† a. We get
matrix in the number basis (32)
dha† ai
= if0 (ha† ie−iωd t − haieiωd t ) ∞ X ∞
dt X
ρ = ρmn |mihn|. (145)
−γha† ai + γN (139) m=0 n=0

14
This is an infinite matrix, but often the matrix can be where α is a complex-valued parameter. This is called
truncated to a finite size with some maximum m and n. displacement operator, as will be evident from relation
The master equation in the number basis is (156) below. Using the relation eA+B = eA eB e−[A,B]/2 ,
which is valid if [A, [A, B]] = [B, [A, B]] = 0, the displace-
dρmn ment operator has the alternative forms
= −iω0 (m − n)ρmn
dt √
√
+if0 e−iωd t ( mρm−1,n − n + 1ρm,n+1 ) D(α) = exp(αa† − α∗ a)
√ √
+if0 eiωd t ( m + 1ρm+1,n − nρm,n−1 ) = exp(−|α|2 /2) exp(αa† ) exp(−α∗ a)
+γ(N + 1) = exp(|α|2 /2) exp(−α∗ a) exp(αa† ). (153)
p 1
×[ (m + 1)(n + 1)ρm+1,n+1 − (m + n)ρmn ] The exponentials (or any function) of an operator should
2
√ 1 be understood as Taylor expansions
+γN [ mnρm−1,n−1 − (m + n + 2)ρmn ]. (146)
2 ∞
X β k Ak
We notice, that only the driving terms contain nondiagonal exp(βA) = . (154)
k!
k=0
elements of ρmn . Assuming f0 = 0 we can thus write the
equation for P (n) = ρnn as We verify the properties
dP (n)
= t+ (n − 1)P (n − 1) + t− (n + 1)P (n + 1) D† (α) = D−1 (α) = D(−α) (155)
dt
−[t+ (n) + t− (n)]P (n), (147)
D† (α)aD(α) = a+α
where † †
D (α)a D(α) = a† + α ∗ (156)
t+ (n) = γN (n + 1)
because [a, f (a† )] = f 0 (a† ), as follows from the commutator
t− (n) = γ(N + 1)n (148)
(31).
can be interpreted as transition probabilities from state n Using the translation operator we construct the following
to states n + 1 and n − 1. In the steady state they satisfy states
detailed balance conditions |αi = D(α)|0i. (157)
t+ (n)P (n) = t− (n + 1)P (n + 1). (149) These are called Glauber states or coherent states. (The
latter may be confusing since “coherent” can mean also
Verify as an exercise that the Bose distribution (103) sa- other things.) In order to understand the meaning of these
tisfies the detailed balance condition and thus is a steady states, let us consider the following.
state solution of the master equation (147). Im α
Im β |α>=D(α)|0>
The physical interpretation of the transition probabili-
ties will be discussed more later in the connection of a Im α
Re β
two-state system.
θ Re α φ
As one particular solution we can consider the tem-
|0> Re α
perature relaxation (140) starting from the ground state
P (n) = δn,0 . Substituting the Bose distribution (103) into The Hamiltonian of the simple harmonic oscillator has
the master equation (147) we see that the solution has the circular symmetry in the q − p variables when expressed in
Bose distribution at all times with the average number of the α variables (23). Since the number states are nondege-
quanta satisfying nerate, the wave functions (50) are the same when looked
−γt in any coordinates <β (or =β) rotated with respect to α by
hn(t)i = N (1 − e ) (150)
arbitrary angle θ, (except a possible phase shift). Thus the
and thus corresponding to temperature ground state |0i can be represented by a circularly sym-
metric dot centered at α = 0 of width 1. Now the Glauber
~ω0 1 state |αi is represented by the same dot but displaced by
T (t) = (151)
kB ln[1 + N −1 (1 − e−γt )−1 ] α = |α|eiφ . This state is described by a wave function that
is identical to |0i except a coordinate shift, which for the
with N evaluated at the reservoir temperature. <β axis would be |α| cos(φ − θ).
Here are some more useful properties of Glauber states.
Glauber states The representation with number states
The following operator is of particular interest ∞
2 X αn
|αi = e−|α| /2
√ |ni (158)
D(α) = exp(αa† − α∗ a), (152) n=0 n!

15
and the property that they are eigenstates of the annihila- Now as long as α satisfies the classical equation (19), the
tion operator last two terms vanish. This means that the drive completely
drops out of the transformed master equation (165). Based
a|αi = α|αi. (159) on this we can now construct several possible solutions for
the driven and damped harmonic oscillator.
The distribution of the photon number in Glauber states
follows the Poissonian distribution 1) At T = 0, the equilibrium solution for the transformed
2n
master equation (165) is the ground state |0i. Transforming
2 −|α|2 |α| back to the the original representation, the full solution is
Pn = |hn|αi| = e (160)
n! a Glauber state |αi (157) with α satisfying the classical
with the average and the variance equation (19).
2) Starting from an arbitrary initial state but reservoir
hňi = |α|2 , σn2 ≡ hň2 i − hňi2 = hňi, (161)
at T = 0, the transformed master equation (165) gives
and standard deviation σn = hňi (Exercise). The Glau- relaxation to the ground state, and the solution after initial
p

ber states are not orthogonal, relaxation is the Glauber state solution above.
2 3) At general T , the equilibrium solution for the transfor-
|hα|βi|2 = e−|α−β| (162)
med master equation (165) is the thermal equilibrium state
(103). The full solution is the thermal state with “center
but are approximately so if |α − β|  1. They satisfy the
of mass” α satisfying the classical equation (19).
completeness relation
Z 4) Any thermal relaxation in the transformed master
1 2 equation (165) takes place independently of the classical
|αihα|d α = 1. (163)
π motion (19) of “center of mass” α. Thus one can have, for
example, initially a coherent state experiencing the thermal
relaxation (150) simultaneously as α changes according to
Elimination of driving terms the classical equation (19).
Let us try the transformation (77) with U = D(α) on on
the full master equation with the drive (134). We get that Cat states
the effective Hamiltonian corresponding to H = H0 + Hd
Consider a coherent superposition of two Glauber states,
is [neglecting constants and using (156)]
|ψi = N (|αi + |βi). (166)
∂D† (α)
H̃ = D† (α)HD(α) + i~ D(α)
∂t N is a normalization constant, which is very close
Here √
= ~ω0 (a† + α∗ )(a + α) − ~f0 [(a† + α∗ )e−iωd t to 1/ 2 when |α − β|  1, because of (162). The density
α∗ α operator corresponding to such a state is
+(a + α)eiωd t ] + i~[(− − a† )α̇ + ( + a)α̇∗ ]
2 2
ρ = N 2 (|αihα| + |βihβ| + |αihβ| + |βihα|) . (167)
= ~ω0 a† a + ~a† [ω0 α − f0 e−iωd t − iα̇]
+~a[ω0 α∗ − f0 eiωd t + iα̇∗ ] + constant. (164) It is of fundamental interest to study how such an initial
state evolves with time. In particular, the last two terms
The transformed density matrix obeys (we drop tilde) are off-diagonal and contain information about quantum
dρ i γ coherence between states |αi and |βi. That such an inter-
= − [H̃, ρ] + (N + 1)[2(a + α)ρ(a† + α∗ ) ference exists on microscopic states is an essential part of
dt ~ 2
the success of the quantum mechanics in explaining phy-
−(a + α )(a + α)ρ − ρ(a† + α∗ )(a + α)]
† ∗
sics. However, when applied to macroscopic systems, such
γ
+ N [2(a† + α∗ )ρ(a + α) − (a + α)(a† + α∗ )ρ coherence has not been observed. A famous discussion of
2 this was given by Schrödinger, who considered a cat being
−ρ(a + α)(a† + α∗ )] either alive or dead as the two states, and according to nai-
i γ ve extrapolation of quantum mechanics, one could expect
= − [H̃, ρ] + (N + 1)(2aρa† − a† aρ
~ 2 to see oscillations between the alive and dead states of the
−ρa† a − α[a† , ρ] + α∗ [a, ρ]) cat. With the Glauber states we can now test this expec-
γ tation since the Glauber states can be macroscopic, there
+ N (2a† ρa − aa† ρ − ρaa† + α[a† , ρ] − α∗ [a, ρ])
2 is no upper limit on the average number of quanta (161).
γ
= −iω0 [a† a, ρ] + (N + 1)(2aρa† − a† aρ − ρa† a) In order to study this, we first want to verify that
2
γ ∗
+ N (2a ρa − aa† ρ − ρaa† )
†
ρa = ||αihβ||e−αβ (168)
2
γ
−i[a† , ρ](ω0 α − f0 e−iωd t − iα̇ − i α) is a solution of the master equation (130) if α and β satisfy
2 the corresponding classical equation. (For simplicity assu-
∗ iωd t ∗ γ ∗
−i[a, ρ](ω0 α − f0 e + iα̇ + i α ). (165) me T = 0 and f0 = 0.) Here we have used unnormalized
2

16
Glauber states defined by 5. Two-state system
∞ In the following we wish to consider the case that the
2 X αn
||αi ≡ e|α| /2
|αi = √ |ni (169) energy levels are not equally spaced. In this case, one can
n=0 n! often approximate the system by two levels only. A two-
state system means that one can excite the system from the
and we need the following properties lower to the higher level, but any third level is sufficiently
off resonance that its presence can be neglected. A two-
a||αi = α||αi, hα||a† = hα||α∗ ,
state system is often called a qubit, a short for quantum
d||αi dhα||
a† ||αi = , hα||a = . (170) bit.
dα dα∗
In atomic and molecular spectra non-equally spaced le-
After some calculation one can now see that (168) indeed vels occur as a rule. However, the electric elements we have
satisfies the master equation (Exercise). All that remains is considered above (capacitors, inductors and resistors) are
to multiply ρa (168) with an appropriate constant in order all linear. Any circuit made of such elements is linear (equi-
to produce any of the four terms in (167). For the third valently the Hamiltonian is quadratic, at most), and has
term we find the time evolution equally spaced levels. In order to make a two-state system,
1 2 2 2 2 ∗ ∗ one needs a nonlinear element. In the following we study a
ρ3 = N 2 e 2 (|α| +|β| −|α0 | −|β0 | ) eα0 β0 −αβ |αihβ| Josephson junction. This is a tunnel junction between two
2 − 21 (|α0 |2 −2α0 β0∗ +|β0 |2 )(1−e−γt ) superconducting metals. It is an ideal nonlinear element.
= N e |αihβ|
2 iφ −γt The practical drawback is that it works only at low tempe-
= N 2 e−|α0 | (1−e )(1−e ) |αihβ|, (171)
ratures, where superconductivity appears. In practice this
where α0 and β0 refer to the initial values of α and β, and in limitation is not serious because in order to see quantum
the last line we have assumed β0 = e−iφ α0 . For short times effects in electric circuits, one has to cool to low tempera-
and φ = π the exponential factor is e−2γhnit . For large hni tures anyway to get rid of noise.
this decays very rapidly. The time constant (2γhni)−1 for
the decay is half of the time constant it takes to emit one 5.1 Circuit analysis
quantum to the reservoir. Therefore coherent oscillations
between macroscopic states can only take place during a Before going to Josephson junctions, let us exercise a bit
very short time, if at all. circuit analysis using fluxes and Lagrangian approach. The
general rules are given by Devoret. Here is an attempt to
The result above can also be interpreted that the reser- reproduce them.
voir is continuously watching the system. Rather than “re-
servoir”, we should speak about the “environment” of the A circuit consists of branches. To each branch b we as-
system, which contains everything except the system itself. sociate a potetial difference Vb and a current Ib . To be
Once the system emits one quantum to the environment, precise, these could be defined
the phase of that photon contains information about the Z end of b
system. Thus it acts like a measurement, independently of Vb (t) = E · ds,
whether any human watches it or not. This kills the cohe- beginning of b
rence between the states, and leaves only the classical pro-
I
1
bability [the first two terms in (167)]. Which of these two I b (t) = B · ds. (172)
µ0 around b
is realized in any particular experiment, can be determined
if the information contained in the emitted photon can be About directions we follow the standard but confusing way
revealed. If not, one has to keep both classical alternatives. of counting Vb positive in the opposite direction as the po-
The Glauber states are robust against the effect of the sitive Ib direction. (A more systematic procedure would be
environment. Such states are called pointer states, i.e. they to write Ohm’s law as I = −V /R, the potential decreases
correspond to some value of a pointer in some classical in the direction of the current.)
measuring device. Any coherent superposition of Glauber We define branch fluxes and charges by
states is fragile, i.e. it rapidly decays to a classical proba-
Z t
bility distribution of different Glauber states. The same in
Φb (t) = Vb (t0 )dt0 ,
other words, the pointer in the measuring device can have −∞
only classical probability, it cannot exist in any coherent Z t
superposition (except of very short times). Qb (t) = Ib (t0 )dt0 . (173)
−∞
Note that also the number states |ni are fragile. In the
presence of dissipation a high number state n  1 rapid- The integration is started at some time past where no vol-
ly decomposes to an incoherent probability distribution of tages and charges were present. The energy stored in a
Glauber states
√ |αi with different phases φ, where α = Aeiφ branch is obtained by calculating
and A ≈ n.
Z t
There are real experiments with the cat states, see HR. Eb (t) = Vb (t0 )Ib (t0 )dt0 . (174)
−∞

17
A capacitor is described by the relation Vb = Qb /C. Using We can now take the limit C3 → ∞, fixing Φ̇b = U . Neglec-
Ib = Q̇b we get the energy ting constants we have
Q2b C Φ̇2b C(Φ̇a − U )2 Φ2
Eb = = . (175) L(Φa , Φ̇a ) = − a. (181)
2C 2 2 2L
where we have used Vb = Φ̇b in the latter form. An inductor Using the standard procedure of analytical mechanics, one
is described by the relation Φb = LIb . Using the same defines the canonical momentum corresponding to the node
relations we get the energy flux Φa as
Φ2b LQ̇2b
Eb = = . (176) ∂L
2L 2 Qa = = C(Φ̇a − U ). (182)
∂ Φ̇a
In the following we wish to construct the Lagrangian,
In general this can be identified as the node charge, the
which is defined as kinetic energy minus potential energy.
sum of the charges on the capacitor plates attached to the
In using the flux variables, we can associate capacitive ener-
node. Now there is only one such capacitor, which contri-
gies as kinetic (because they depend on Φ̇) and inductive
butes the whole Qa . The Hamiltonian is calculated using
energies as potential (because they depend on Φ).
the standard rule
A dc voltage source can be represented by adding a fake X
capacitor to the circuit, and then taking the limit C → ∞ H= Φ̇i Qi − L. (183)
and Q → ∞ while V = Q/C remains constant. Corres- nodes i
pondingly, a dc current source can be represented by an
inductor, and an ac source by an LC circuit. In the present case this gives (neglecting constant again)

Not all branch variables are independent variables, they C Φ̇2a Φ2a
are constrained by Kirchhoff rules. They say that the vol- H = Φ̇ a Qa − L = + . (184)
2 2L
tages around each closed loop must sum to zero, and the
the currents arriving to each node must also sum to zero. Writing it as a function of the canonical variables gives
In terms of flux and charge variables these are finally
X
Φb = Φ̃l (177) (Qa + CU )2 Φ2
H(Φa , Qa ) = + a. (185)
all b around l 2C 2L
X
Qb = Q̃n , (178) Except the shift CU caused by the battery, the result is
all b arriving at n
the same as obtained above (41), taking into account the
where Φ̃ and Q̃ are constants. [The node rule (178) is not correspondence of the constants in (7) and (9).
l n
explicitly used in the following. It is automatically satisfied
by the procedure that is used below.] 5.2 Josephson junctions
Φ1 Φ2 In the following we use a simple model of the Joseph-
Φb son junction. In superconductivity the electrons form pairs,
Φa
L C U which are called Cooper pairs. Here we consider only the
C3 Φ3
Cooper pairs, and neglect any unpaired electrons. This is a
reasonable approximation as long as the voltage V is small
compared to the energy gap ∆ of the superconductor
Let us consider an example, the circuit in the figure on
the left. The right hand figure describes how we can assign ∆
V  . (186)
branch (Φ1 , Φ2 and Φ3 ) and node fluxes (Φground = 0, Φa e
and Φb ) and change the battery to a capacitor. Often one uses aluminum, which becomes superconducting
The loop rule (177) gives Φ1 + Φ2 + Φ3 = Φ̃. We express below Tc = 1.2 K. The energy gap at T  Tc is approxima-
the branch fluxes in terms of the node ones, tely 0.2 meV (corresponding to 2 K in temperature units,
i.e. ∆/kB ). Thus we can neglect unpaired electrons for
Φ1 = Φ a T  1 K and V  0.2 mV. For Nb the conditions are
Φ2 = Φ b − Φa somewhat more relaxed.
Φ3 = Φ̃ − Φb . (179) There is another limitation to be encountered soon. We
The next step is to write the Lagrangian as a function of need a metallic island that is coupled to the environment
the node fluxes. We get so that it matters if there is one more or less electron on
the island. This means that the capacitive energy of one
L(Φa , Φb , Φ̇a , Φ̇b ) electron Ec = e2 /2C should exceed the thermal energy
C(Φ̇b − Φ̇a )2 C3 Φ̇2b Φ2 kB T . With nanofabrication methods one can make such
= + − a. (180) small structures that C ∼ 10−15 F, meaning that T < 1 K.
2 2 2L

18
 We now define a Josephson element by the energy parallel to the Josephson element (figure). Do the analysis
for this circuit diagram (exercise) to find the result
2πΦ
E = −EJ cos . (187)
Φ0 (Q + Q0 )2 2πΦ
H(Φ, Q) = − EJ cos . (193)
2CΣ Φ0
Here Φ0 = h/2e is the flux quantum and EJ is a cons-
tant (coupling energy). Although the cosine form (187) This is the Hamiltonian of a Cooper pair box. We have defi-
may look strange, a very simple interpretation of this is gi- ned Q0 = Cg U , CΣ = Cg + CJ . The lead from the battery
ven below in connection of equation (199). Using the fact to the capacitance Cg is called “gate” and the electrode
that Ė = IV = I Φ̇ we can find the current between the Josephson junction and the gate capacitance
2πΦ is called “island” or “box”.
I = Ic sin = Ic sin φ. (188)
Φ0 Until now the circuit analysis has been classical. We can
shift to quantum theory using the standard canonical quan-
The latter form follows because we define phase tization procedure. For mechanical system q and p are ca-
2π 2e nonical conjugate variables, and in quantum theory they
φ= Φ = Φ. (189) become operators satisfying the commutation relation (29).
Φ0 ~
For the electric circuit considered, the canonical variables
The relation (188) is known as dc Josephson effect saying are Φ and Q, and thus analogously they become operators
that at V = 0 there is a constant supercurrent that depends satisfying
sinusoidally on Φ (or φ). The maximum supercurrent Ic is
given by [Φ, Q] = i~. (194)

2πEJ 2eEJ In Schrödinger picture Φ corresponds to the spatial coor-

Ic = = . (190) dinate q (or x) and Q is then the analog of p (47) operator
Φ0 ~
d
We also see the ac Josephson effect saying that at cons- Q = −i~ . (195)
dΦ
tant V 6= 0 there is alternating current as Φ(t) = V t +
constant. The frequency of the alternating current is The Hamiltonian (193) corresponding to the Schrödinger
eigenvalue equation for the wave function u(Φ) is
2e
ω = V. (191)  2
~ 1 d
−i~ + Q0 u(Φ)
At voltage 0.1 mV this gives the frequency ν = ω/2π = 48 2CΣ dΦ
GHz. 2πΦ
− EJ cos u(Φ) = Eu(Φ). (196)
Φ0
5.3 Cooper pair box Alternatively, this can be written using φ (189)
 2
d Q0
EJ C U 4Ec −i + u(φ)−EJ cos φ u(φ) = Eu(φ), (197)
dφ 2e

where Ec = e2 /2CΣ is the single electron charging energy.

We see that this Schrödinger equation has potential V =
Let us now repeat the same circuit calculation as abo- −EJ cos φ which is periodic.
ve but replacing the inductor with a Josephson junction. Before analyzing the solutions of the Schrödinger equa-
Going through the calculation one notices that there is not tion (197), it is useful (if not necessary) to discuss the repre-
much change, and the result is sentation of the same equation in the charge basis. The ei-
(Q + CU )2 2πΦ genstates of the charge operator (195) are plane waves
H(Φ, Q) = − EJ cos . (192)
2C Φ0 ψQ (Φ) = eiQΦ/~ . (198)
(We have dropped the subindex a as unnecessary.)
Operating with the Cooper pair box Hamiltonian on this
EJ gives

Cg U d
= H(Φ, −i~ )ψQ (Φ) (199)
dΦ
2
 
(Q + Q0 ) EJ i2eΦ/~
CJ = eiQΦ/~ − (e + e−i2eΦ/~ ) .
2CΣ 2
Let us add one more realistic feature. A real Josephson The first term is the capacitive energy times the original
junction always can store charge on the metal electrodes. charge eigenstate, as one should expect. The second, Jo-
This can be represented as an additional capacitance CJ sephson term gives two different charge eigenstates. Their

19
charges differ from Q by ±2e. Thus we see a simple in- 1) Ec  EJ , the kinetic energy (of lowest states) is small
terpretation of the Josephson energy term (187): it trans- compared to the potential. Close to the minimum at φ = 0
fers one Cooper pair across the junction, from side 1 to side the cosine potential is approximately harmonic. Thus the
2, or vice versa. lowest energy eigenvalues are harmonic oscillator like, with
4
Take a look on the circuit diagram of the Cooper pair a small correction coming from the φ term in the Taylor
box above. In the absence of the Josephson element, the series of − cos(φ). Q0 has negligible effect on the states
since ψ(±π) ≈ 0.
island is coupled to the rest of the circuit by capacitors only.
Thus the charge on the island cannot change. Introducing 2) Ec  EJ , called charge limit. Consider first EJ = 0.
the Josephson element changes this situation so that the As discussed above the solutions in the phase representa-
charge can be an integral multiple n of 2e, i.e. the possible tion are plane waves. The energies are 4Ec (n + Q0 /2e)2 .
values of charge E

Q = Q1 + 2en (200)
n =1 n =0 n =-1

with a constant Q1 . Let us denote a charge eigenstate ha- EJ=/ 0

ving n Cooper pairs on the island by |ni. Thus the Hamil-
tonian (193) can be written in the charge representation EJ=0
as
Q0
" 2 -e 0 e 2e 3e 4e

X Q0
H = 4Ec n + |nihn|
n
2e
# For any given Q0 , the capacitive energy is minimized by
EJ the integer n that is closest to Q0 /2e. This leads to the
− (|n + 1ihn| + |n − 1ihn|) . (201)
2 set of parabolas as a function of Q0 shown in the figure.
The lowest state is nondegenerate except at −e, e, 3e etc,
(Here we have neglected Q1 since it can be eliminated by where the levels cross. If one now turns on a small EJ ,
redefinition of Q0 .) Note that in this respect the Joseph- the degeneracy is lifted: the system is in a superposition of
son element is simpler than an inductor. An inductor can states where there are either n or n + 1 Cooper pairs on
pass an arbitrary charge and thus Qa in (185) can vary the island. As depicted in the figure, there are two closely
continuously and a countable charge basis as in (201) is spaced energy levels, and all others are at much higher
insufficient. energies. The levels depend essentially on Q0 as there is
What is the meaning of the discrete charge (200) in the not much effect of EJ except near the charge degeneracy
flux/phase space? The most general allowed wave function points.
is a discrete superposition of the plane waves (198), As a summary, we see that a two-state system is for-
X med in circuits containing a Josephson junction. We have
u(φ) = un einφ , (202) discussed here so-called charge qubit. Also two other types
n of qubits (“fulx” and “phase”) can be constructed using
which is now simpler in terms of phase φ = 2eΦ/~ (189) Josephson junctions.
and we neglect Q1 . But this condition is the same as
u(φ + 2π) = u(φ), (203) 5.4 Rabi oscillations
i.e. the wave function u in (197) has to be periodic and thus We now consider a general two-state system. We label
carries independent information only in range −π < φ ≤ π, the two levels as a and b, where b is the lower energy state.
say. Note that u and the potential V = −EJ cos φ have the E
same period. |a>

It is now convenient to define a new wave function by

the transformation
|b>
iQ0 φ/2e iπQ0 /e
ψ(φ) = e u(φ), ψ(φ + 2π) = e ψ(φ). (204)
The energy separation of the levels is denoted by ~Ω0 ,
Substituting into (197) gives where Ω0 ≥ 0. The general wave function can be written

d2 
ψa

− 4Ec ψ(φ) − EJ cos φ ψ(φ) = Eψ(φ). (205) or |ψi = ψa |ai + ψb |bi. (206)
dφ2 ψb
Thus the dependence on the “off-set charge” Q0 has moved
from the Schrödinger equation to the boundary conditions. In the matrix representation the unperturbed Hamiltonian
can be written
The equation (205) has exact solution in terms of Mat-
hieu functions. We learn more by considering two limiting ~Ω0
cases. H0 = σz , (207)
2

20
where we use the Pauli matrices Thus the system oscillates between the ground and excited
      states at the angular frequency ΩR0 = 2f0 , known as the
0 1 0 −i 1 0
σx = , σy = , σz = . Rabi frequency. In the general (nonresonant) case there
1 0 i 0 0 −1
is
p oscillation at the (generalized) Rabi frequency ΩR =
(208)
4f02 + (Ω0 − ωd )2 . Rabi oscillations were first observed
It is also convenient to define
by Rabi in atoms (1937). In a Josephson-junction qubit
   
0 0 0 1 they were seen by Nakamura et al (1999).
σ− = , σ+ = . (209)
1 0 0 0 In the following it is useful to introduce a representation
of the density matrix in terms a fictitious spin vector S.
These are the same as a and a† operating on |1i and |0i
Any 2 × 2 matrix can be written in terms of unit matrix
(33) with the exception that all higher states are cut off,
I and the three Pauli matrices (208). Taking into account
σ− |ai = |bi, σ+ |ai = 0, that Trρ = 1, we can write it as
 1 
σ− |bi = 0, σ+ |bi = |ai. (210) 1 2 + Sz Sx − iSy
ρ= I +S·σ = 1 . (219)
2 Sx + iSy 2 − Sz
Let us consider a driven (not damped) qubit. With the The inverse of this is
analogy to the harmonic oscillator (53) the Hamiltonian is
1
Sx = (ρab + ρba )
~Ω0 2
H = H0 + Hd = σz − 2~f0 σx cos ωd t. (211) i
2 Sy = (ρab − ρba )
2
The Schrödinger equation 1
Sz = (ρaa − ρbb ). (220)

ψ̇a

i

Ω0 −4f0 cos(ωd t)

ψa
 2
=− .
ψ̇b 2 −4f0 cos(ωd t) −Ω0 ψb Note that S is a classical vector, not an operator. It can be
(212) represented by a point in the Bloch sphere depicted below.
We can try to remove the time depedence of H by making
transformation (77) with Sz
 −iω t/2 
e d 0
U = exp(−iωd tσz /2) = (213)
0 eiωd t/2

and get
~

Ω 0 − ωd −4f0 cos(ωd t)eiωd t
 Sy
H̃ = .
2 −4f0 cos(ωd t)e−iωd t −(Ω0 − ωd ) Sx
(214)
If we are interested only in a slowly moving solution in this Note that the representation (219) is valid for arbitrary
frame, we may neglect the fast oscillating terms, and get a statistical mixtures. For a pure state the fictitious spin is
time independent Hamiltonian normalized to 1/2 corresponding to points on the surface
  of the Bloch sphere. For incoherent mixtures Sx2 + Sy2 +
~ Ω 0 − ωd −2f0 Sz2 < 1/4 and they correspond to states inside of the Bloch
H̃rwa = . (215)
2 −2f0 −(Ω0 − ωd ) sphere.
The eigenvalues of this are given by Consider now an arbitrary 2×2 matrix Hamiltonian H =
q cI + (~/2)Ω · σ. Substituting this and ρ (219) into the
~ von Neumann equation (99) gives the equation of motion
E± = ± 4f02 + (Ω0 − ωd )2 . (216)
2 (exercise)
At resonance (ωd = Ω0 ) the eigenvalues are ±~f0 , and the dS
= Ω × S. (221)
solution is dt
      Thus the vector Ω is the angular velocity of the fictitious
ψa −1 1 spin S. We can think Ω as a dimensionless fictitious mag-
=A e−if0 t + B eif0 t . (217)
ψb 1 1 netic field around which the spin rotates.
Assuming b as the initial state at t = 0 gives A = B = It is of interest to interpret the previous Rabi oscillation
1/2 and thus ψa = i sin f0 t and ψb = cos f0 t. We find the solution in this picture. First, the static magnetic field ma-
density matrix kes the spin to rotate around z by Ω0 . The transformation
(213) corresponds to changing to a frame rotating around z
|ψa |2 ψa ψb∗
 
ρ = by ωd . If ωd = Ω0 , the only motion in this frame is around
ψb ψa∗ |ψb |2 the driving field Ω = −2f0 x̂ (in RWA), and thus causes
the Rabi oscillation. If ωd 6= Ω0 there is rotation around
 
1 1 − cos(2f0 t) i sin(2f0 t)
= . (218) Ω = −2f0 x̂ + (Ω0 − ωd )ẑ.
2 −i sin(2f0 t) 1 + cos(2f0 t)

21
 We see that with proper driving fields, the qubit can be Similarly to the case of harmonic oscillator, the equi-
brought to any orientation. This is the purpose in quan- librium state of the qubit is determined by the detailed
tum engineering. The problem is that there are unwanted balance condition
couplings to environment, whose effect we will study next.
ρ̇aa = 0 ⇔ γ(N + 1)ρaa = γN ρbb . (226)

5.5 Bloch equations This leads to the Gibbs distribution of the qubit

In order to get the damping terms we resort to the mas- ρaa = e−~Ω0 /kB T ρbb . (227)
ter equation. The master equation for the two-level sys-
tem density matrix is analogous to the harmonic oscillator In deriving the master equation we assumed the coupling
(133) with substitutions a → σ − , a† → σ + and thus (59) or V (t) = −~[σ − (t) + σ + (t)][Γ(t) + Γ† (t)] between
the system and the reservoir. In addition, we could also
dρ i consider a coupling of the form V (t) = ~Γ̃(t)σz , where Γ̃(t)
= − [H0 + Hd , ρ]
dt ~ is some interaction-picture Hermitian operator acting in
γ the reservoir. (Note that the interaction picture operator
+ (N + 1)(2σ − ρσ + − σ + σ − ρ − ρσ + σ − )
2 σz (t) ≡ σz .) The derivation of the master equation in this
γ
+ N (2σ + ρσ − − σ − σ + ρ − ρσ − σ + ). (222) case is much simpler and one finds on the right hand side
2 of the master equation (222) an additional term
Here H0 + Hd is two-state Hamiltonian (211). The boson γφ
occupation (102) has to be evaluated at the level separation − [σz , [σz , ρ]], (228)
4
~Ω0 , i.e.
where
  Z ∞
1 1 β~Ω0
N= = coth −1 , (223) γφ = 4 dthΓ̃(t)Γ̃(0)i. (229)
exp(β~Ω0 ) − 1 2 2 0

where the latter form is sometimes more convenient. Adding this to (224) gives

Let us stydy the master equation first without driving. ρ̇aa = −γ(N + 1)ρaa + γN ρbb
Calculating (222) for the 2 × 2 matrix componentes we get ρ̇ab = −iΩ0 ρab − γ(N + 21 )ρab − γφ ρab
ρ̇ba = iΩ0 ρba − γ(N + 12 )ρba − γφ ρba
ρ̇aa = −γ(N + 1)ρaa + γN ρbb
ρ̇bb = γ(N + 1)ρaa − γN ρbb . (230)
ρ̇ab = −iΩ0 ρab − γ(N + 12 )ρab
ρ̇ba = iΩ0 ρba − γ(N + 12 )ρba We see that there is no effect on the diagonal terms, but
ρ̇bb = γ(N + 1)ρaa − γN ρbb . (224) the off-diagonal ones acquire additional damping. This can
be understood that the longitudinal coupling induces fluc-
We notice that the diagonal and off-diagonal elements of tuations into the level separation Ω0 . Thus different rea-
ρ develop independently. The diagonal elements conserve lizations rotate at different speeds, and lead to diminishing
the normalization ρaa + ρbb = 1 because ρ̇bb = −ρ̇aa . We of the average coherence. Note that the essential quantity
also can give physical interpretation to the three terms to this dephasing is fluctuations at ω ≈ 0 (229), whereas
contributing to for the mixing terms (∝ γ) the fluctuations at ω ≈ Ω0
are dominant. To avoid such fluctuations is one of the
ρ̇aa = −γN ρaa − γρaa + γN ρbb . (225) main problems in realizing quantum operations in elect-
ric circuits.
The term γN ρbb corresponds to absorption of a quantum
Let us now represent the damping terms using the spin
from the environment. The term −γN ρaa corresponds to
representation (219). Combining everything gives
stimulated emission of a quantum to the environment. Fi-
nally, the term −γρaa corresponds to spontaneous emis- dSx 1
= (Ω × S)x − Sx
sion of a quantum to the environment. The interpretation dt T2
of absorption and emission comes because these processes dSy 1
increase or decrease the energy of the qubit, and from ener- = (Ω × S)y − Sy
dt T2
gy conservation. We notice that absorption and stimulated dSz 1
emission are possible only if N 6= 0, i.e. the environment is = (Ω × S)z − (Sz − Sz0 ). (231)
dt T1
at finite temperature, T > 0.
Here we assume the main static part of Ω equals Ω0 ẑ, and
One may ask how can the spontaneous emission take
place by coupling the qubit to the vacuum only (thinking 1 1 ~Ω0
Sz0 = − = − tanh
the reservoir oscillators as electromagnetic modes). One 2(2N + 1) 2 2kT
could interpret that even vacuum is not completely empty 1 ~Ω0
since there are zero-point oscillations, and these could kick = γ(2N + 1) = γ coth
T1 2kT
the qubit from the upper to the lower state. (We return to 1 1
this soon.) = + γφ , (232)
T2 2T1

22
where Sz0 is the equilibrium value of the fictitious spin and f (t−τ ) is important. The spectral density of f (t) is defined
T1 and T2 are two relaxation times. Equations (231) are as
Bloch equations that have extensively been used in nuclear Z Tw /2
magnetic resonance and in optics. Sf (ω) = dτ eiωτ Rf (τ ), (237)
w −T /2
In the literature the Bloch equations appear in various
notations. Allen&Eberly write the Bloch equations in the where Tw gives the width of the time window used in the
rotating frame in the form Fourier transform. (More on Fourier transform on page 27.)
u Returning to fr = Γ + Γ† we have
u̇ = −∆v −
T2 Z ∞
v̇ = ∆u −
v
+ κEw Sfr (ω) = dτ eiωτ hδfr (t)δfr (t − τ )i
T2 −∞
Z ∞
w − weq
ẇ = − − κEv. (233) = dτ eiωτ hfr (τ )fr (0)i
T1 −∞
Z ∞
Here u, v and w are the three components of S, weq = Sz0 , = dτ eiωτ hΓ(τ )Γ† (0) + Γ† (τ )Γ(0)i
−∞
∆ = Ω0 − ωd , and κE = 2f0 is the driving field. The steady Z ∞
state solution for constant κE is gi gj (hbi b†j ie−iωi τ + hb†i bj ieiωi τ )
X
= dτ eiωτ
−∞ i,j
(∆T2 )(κET2 )
u = −weq Z ∞
1 + (∆T2 )2 + T1 T2 (κE)2
X
= dτ eiωτ gi2 [(N + 1)e−iωi τ + N eiωi τ ]
κET2 −∞ i
v = weq ∞ ∞
1 + (∆T2 )2 + T1 T2 (κE)2
Z Z
= dτ eiωτ dω1 ρ(ω1 )g 2 (ω1 )
1 + (∆T2 )2 −∞ 0
w = weq . (234)
1 + (∆T2 )2 + T1 T2 (κE)2 ×[(N (ω1 ) + 1)e−iω1 τ + N (ω1 )eiω1 τ ]
= 2πρ(|ω|)g 2 (|ω|) (238)
The last one can also be written as
×[(N (ω) + 1)Θ(ω) + N (−ω)Θ(−ω)]
T1 T2 (κE)2
w − weq = −weq . (235) because
1 + (∆T2 )2 + T1 T2 (κE)2
Z ∞
Note that the last of equations (233) is an energy equation dτ eiωτ = 2πδ(ω) (239)
−∞
(apart from constant factor) similar to (24) for harmonic
oscillator. Thus the absorption is proportional to v, the dis- and we have used the step function Θ(x) (5). We see that
sipation is proportional to w − weq , and in the steady state the frequency ω in Sfr (ω) can be either negative or positive.
they are equal. For small driving these reduce to the same Using (102) we can write both parts in the same expression
as for a harmonic oscillator. At higher drives one gets “sa- as
turation”, where the absorption approaches a maximum.
Another effect is “resonance fluorescence” that the emit- Sgn(ω)γ(ω)
Sfr (ω) = , (240)
ted radiation is not at the same frequency as the drive, 1 − e−~ω/kB T
but has three different peaks centered at ωd and ωd ± ΩR0 . where Sgn(ω) = ω/|ω| and γ was defined above (127) for
This spectrum can be determined by studying a Fourier positive frequencies and we extend it symmetrically for ne-
transform of the correlation function (104). gative ones,

γ(ω) = 2πρ(|ω|)g 2 (|ω|). (241)

5.6 Fluctuation-dissipation theorem
Let us have an interlude to study in more detail the Using the form (16) we can write the spectral density for
force fr exerted by the reservoir on the system. By the the unreduced force F in (7), where ω0 = ω is the angular
analogy of (57) and (59), this force is the operator fr = frequency at which we are interested to analyze F (t). We
Γ+Γ† . This force cannot be characterized by its mean value get
since it vanishes. Instead, we can use spectral density. For 2~ωmγ(ω)
later use we give here full definition of this quantity for an SF (ω) = . (242)
1 − e−~ω/kB T
arbitrary quantity f (t). We first define δf = f − hf i, the
This result is known as fluctuation-dissipation theorem. On
deviation of f from its mean value. Secondly we define the
autocorrelation function Rf of f by the left hand side SF represents fluctuations or noise, the
random force exerted by the reservoir on the system. On
Rf (τ ) = hδf (t)δf (t − τ )i. (236) the right hand side γ represent dissipation, the frictional
force caused by the reservoir on the system. This theorem
(In some cases an additional averge over time t may be says that these are related, one cannot have dissipation
needed.) As f may be an operator, the order of f (t) and without fluctuations, if T > 0, and vice versa.

23
 Of special interest is the classical limit of (242), meaning to cause absorption, which means energy flow from the en-
kB T  ~ω. Expanding the exponential factor we get vironment to the system. This process is possible only if
the environment is at a finite temperature, T > 0.
SF (ω) = 2kB T mγ(ω). (243)
@KD
SV
R kB
Although we have been using the notation of mechanical 800
oscillator, we must remember that all results are valid for
600
all harmonic oscillators, only the constants have different
names. In particular, comparing (7) with the electrical 400
oscillators equations (8) or (9), we see that voltage and
200
current fluctuations are related to the resistance
@HzD
Ω
-4. ´ 1012-2. ´ 1012 2. ´ 1012 4. ´ 1012 2Π
SV (ω) = 2kB T R,
SI (ω) = 2kB T /R. (244) Figure: Noise in a resistor as a function of frequency at
temperatures T = 293 K, T = 100 K and T = 0 (from top
The physics is that a resistor not only dissipates energy, but to down). We see that the classical result (244) is an ap-
also acts as a source of current and voltage fluctuations. proximation that works reasonably in a limited frequency
range around ω = 0. (The range is quite large, ω/2π . 1012
Hz, compared to typical frequencies in standard electro-
SV
true
nics.) For large negative ω the noise goes small as the en-
V=RI

= SI = vironment becomes insufficient to supply such high energy

resistor
V=RI

and zero point fluctuations cannot give off any energy. For
large positive ω the noise increases meaning that at large
energies zero point fluctuations, causing spontaneous emis-
The famous relation (244) was derived by Nyquist (1928) sion, become the dominant cause of emission compared to
and was motivated by the experimental results by John- thermal excitations.
son (1928). [Electrical engineers usually define S different-
ly from the present physicist’s way so that they have a 5.7 Coupled qubit and harmonic oscilla-
factor 4 instead of 2 in (244).] tor
As an ideal circuit element, the resistance R of a resistor Consider a two-state system coupled to a single harmo-
is independent of frequency. This certainly is an idealiza- nic oscillator. Two realizations of such systems were shown
tion, but is approximately true for real resistors in a rat- in the beginning of this course (A. Wallraff et al and D.
her wide range of frequencies. (At least this is limited by Gunnarson et al). The hamiltonian for this system is again
ω  1/τ where τ ∼ 10−14 s is the Drude relaxation time of obtained by adapting our previous model of one plus a re-
electrons in a metal). Thus the noise spectral density (244) servoir of oscillators (55)-(59). The simplest way is to cut
is independent of frequency. Such noise is called “white”. In the reservoir to a single two-state system, and we get
particular, it is also symmetric, SV (−ω) = SV (ω). This is
~Ω0
natural for classical noise since any asymmetry must arise H= σz + ~ω0 (a† a + 21 ) − ~gσx (a + a† ). (246)
from non-commutation of f (t) and f (t − τ ) in (236). 2
In the following we are interested in the case that the level
Let us now return back to the general expression (242), separations are not much different, Ω − ω  Ω + ω [the
0 0 0 0
or equivalently reason also being that there is no exact solution of (246)].
This allows to use the rotating wave approximation, and
2~ωR
SV (ω) = . (245) we get the Jaynes-Cummings (JC) model
1 − e−~ω/kB T
~Ω0
HJC = σz + ~ω0 (a† a + 12 ) − ~g(σ + a + σ − a† ), (247)
What is the meaning of the frequency asymmetry ω ↔ −ω 2
in this expression? For that it is better to go back to the last which we are now going to study.
form of equation (238). We see that the positive frequency
noise is proportional to N + 1 and the negative one to N . The space where the JC Hamiltonian (247) operates is a
Since there is no other sources of these factors in the deri- direct product of the qubit space and the harmonic oscilla-
vation of the master equation, we conclude that the terms tor space. The former has basis |ai and |bi, and the latter
proportional to N + 1 and N in the master equation are |ni with n = 0, 1, 2, . . .. Thus the JC space has basis formed
caused by positive and negative frequency fluctuations. Re- by |a, ni and |b, ni. One can now notice that HJC couples
ferring to our discussion of the physical processes caused by the state |a, ni only to itself and to the state |b, n + 1i. To
these terms [in connections of (148) and (225)], we can thus help to visualize that it may be useful to write Hamiltonian
identify positive frequency fluctuations to cause emission, in the form
ω0 (a† a + 12 ) + Ω20
 
i.e. energy flow from the system to the environment. This −ga
HJC = ~ .
is effective only if the system initially is in an excited sta- −ga† ω0 (a† a + 21 ) − Ω20
te. Conversely we identify negative frequency fluctuations (248)

24
We can now search the eigenstates in the form S2 S2
= (1 + S + + . . .)H(1 − S + + . . .)
  2 2
ψa |ni 1
. (249) = H + [S, H] + [S, [S, H]] + . . . . (251)
ψb |n + 1i 2

This reduces the problem to diagonalizing the matrix Choosing [S, H0 ] = −H1 cancels the leading order term
√ giving
1 Ω0
 
ω0 (n +
√2) + 2 −g n + 1
H=~ 1
−g n + 1 ω0 (n + 32 ) − Ω20 H̃ = H0 + [S, H1 ] + . . . . (252)
 √  2
~ ∆
√ −2g n + 1
= (n + 1)~ω0 + , We see the second term is second order in H1 . We wish to
2 −2g n + 1 −∆
use this to the interaction term in the JC model (247). We
where ∆ = Ω0 − ω0 . The energies of the eigenstates are (± guess S = α(σ + a − a† σ − ), and find
correspond to higher and lower energy states)
[α(σ + a − a† σ − ), ~ω0 a† a] = ~ω0 α(σ + a + a† σ − )
~p
E±,n = (n + 1)~ω0 ± 4g 2 (n + 1) + ∆2 , (250)
2
[α(σ + a − a† σ − ), 21 ~Ω0 σz ] = −~Ω0 α(σ + a + a† σ − ),
where n = 0, 1, . . .. These states appear as pairs, but the
ground state (n = −1) is a single state (0, |0i)T with energy which for α = g/∆ eliminates the coupling term in first
E = −~∆/2. order. The new second order term is
E 1
ÑΩ0 [α(σ + a − a† σ − ), ~g(a† σ − + σ + a)]
2
= ~gα[σ + a, a† σ − ]
3.0
~g 2 † 1 1
= (a aσz + σz + ). (253)
2.5
∆ 2 2
The transformed JC Hamiltonian, neglecting constants and
2.0 corrections of order g 3 /∆2 , is

g2 g2
   
† ~
1.5 H̃ = ~ ω0 + σz a a + Ω0 + σz . (254)
∆ 2 ∆
1.0 We see that this is diagonal in σz and in the number ope-
rator a† a, and the energy eigenvalues can directly be read
0.5 from the Hamiltonian. But one should note that the ope-
rators a, a† and σz are not the same as above, because of
D
the transformation (251). Thus the eigenstates of H̃ (254)
-0.4 -0.2 0.2 0.4 Ω0
are called dressed states, i.e. they are linear combinations
of the original “bare” states |a, ni and |b, ni.
The physical content is that the energy eigenstates are
We note the first term of (254) is the harmonic oscillator,
superpositions of states with qubit in the upper state and
but the frequency is shifted depending on the state of the
qubit in the lower state but one more quantum in the re-
qubit. Alternatively, collecting together the terms propor-
sonator (249). If we have an initial state |b, ni, this is not
tional to σz one can interpret that the qubit level separa-
an energy eigenstate, and there will be Rabi oscillations
tion has two shifts. The first shift 2g 2 n/∆ is called ac Stark
between states |b, ni and |a, n − 1i. This is quite similar
shift because it is analogous to the shift of atomic energy
than we had for a driven two-state system above (217).
levels in alternating electric field (which has the same na-
The difference is that the driving field is now produced
me). The second shift g 2 /∆ is independent on the photon
by quanta in the resonator, and we can see explicitly that
number n and is called the Lamb shift, again with analogy
when the qubit flips up, one quantum is removed from the
to atoms. It is a shift of the qubit or the atom caused by
resonator.
coupling to the vacuum of electromagnetic field. A similar
Another noteworthy point is the dispersive limit where shift was encountered above (128) but was neglected there.
∆  g. One way to study this is to expand the exact solu-
The JC model is a good starting point for the analy-
tion above to leading powers in g/∆. Here we present the
sis of the experiment by Wallraff et al mentioned in the
same result using a transformation directly on the JC Ha-
introduction of this course.
miltonian (247). Quite generally assume H = H0 + H1 and
try to eliminate H1 in first order. We make a transforma-
tion

H̃ = eS He−S

25
6. Electron transport and noise in Assume T = 0 and suppose the Fermi energy corres-
mesoscopic structures ponds to the horizontal line in the previous figure. Then it
is said that there are three conducting modes (or transverse
modes) in the wire. The modes whose lowest energy is abo-
6.1 Transverse modes in a wire ve the Fermi energy are empty, and thus do not contribute
Consider electrons in an ideal conducting wire (Datta). to electric conduction.
For simplicity we consider electrons in two dimensions. The Next we want to determine the conductivity of the wire.
effective Schrödinger equation is We assume that the left end is connected to a reservoir of
chemical potential µL and the right end to a reservoir at
~2 2
− ∇ ψ(x, y) + V (x, y)ψ(x, y) = Eψ(x, y), (255) chemical potential µR and we assume they are in equili-
2m brium at zero temperature, T = 0. Thus the potential dif-
where m is the effective mass (as we do not include the ference V = (µL − µR )/e. We assume that the wire is fully
periodic lattice potential in V ). The wire is in the x direc- coherent (no decoherence) and ballistic (no scattering) so
tion and has finite width W in the “transverse” y direc- that we can discuss the system in terms of freely propaga-
tion. Again, for simplicity assume the potential for elect- ting waves (256). Then the occupation of the states in the
rons outside the wire is very high. Then we know that the wire is given in the following figure
possible wave functions in the y direction are sin(πny/W ) 2 E mW2

with n = 1, 2, 3, . . . for a wire between y = 0 and y = W . Π2 Ñ2

30
For a homogeneous wire [V (x, y) = V (y)] the energy ei-
25
genstates are
20
r
2 ikx πny 15
ψk,n (x, y) = e sin , (256)
LW W 10

5
where k is the wave number in the x direction and L is kW
the quantization length where we could require periodic -4 -2 0 2 4 Π
boundary conditions. The energies of the states (256) are
Here the states with k > 0 propagate from left to right
~2 and they are filled up to the energy µL . Conversely, the
  πn 2 
2
Ek,n = k + . (257) states with k < 0 propagate from right to left and they are
2m W
filled up to the energy µR . It is then clear that the net cur-
These are plotted in the figure. rent is carried by the states with k > 0 between the ener-
2 E mW2 gies µR and µL , i.e. the states between the two horizontal
Π2 Ñ2 lines in the following figure. At lower energies states propa-
30
gating at opposite velocities are equally occupied and thus
25 cannot carry any net current.
20 2 E mW2
15 Π2 Ñ2
30
10
25
5
kW 20

-4 -2 0 2 4 Π 15

10
With each wave function (256) we associate a velocity in
the x direction (effectively a Hamilton’s equation) 5
kW

1 dEk,n -4 -2 0 2 4 Π
v= . (258)
~ dk The current I and the conductance
In our simple model (257) this agrees with the classical re- 1 I eI
G= = = (260)
sult v = px /m = ~k/m. In thermal equilibrium at tempe- R V µL − µR
rature T the occupation of the states is given by the Fermi can be calculated using standard practices of condensed
distribution matter physics, see Datta. Instead, here we derive the same
1 result using an approach that is close to information theory.
f (E) = β(E−µ) , (259) In information theory one can ask what is the maximum
e +1
amount of information that can be transmitted in a channel
where µ is the chemical potential and β = 1/kB T . At zero within a given frequency interval (between frequencies ν2
temperature µ is the same as Fermi energy, and all states and ν1 ) and time interval ∆t. Our related question is, how
at lower energies are occupied, f = 1, and the states at many electrons can be transmitted in a wire within a given
higher energies are empty, f = 0. energy interval µL − µR and time interval ∆t. It turns out

26
that the result is different only in the respect that electrons result (266) is independent of the location of the frequency
are fermions whereas in information transfer one mostly interval, only its width enters the expression.
uses fields that are bosons, like the electromagnetic field. Instead of a full proof, we make the preceding statement
plausible by the following calculation. We define the func-
6.2 Wavelets tion
Consider functions of time, f (t), and their Fourier trans- Z ω2
forms φ(t) = dω e−iωt (267)
ω1
Z Tw /2
fˆ(ω) = dt eiωt f (t), (261) known as Shannon wavelet. We notice that in the ω space
−Tw /2 it is non-zero only within the range ω1 < ω < ω2 . We
calculate its overlap with the same function but translated
where Tw is the width of the time window used in the Fou-
in time by u,
rier transform. Independent information of f (t) is contai-
ned only in f (ω) at frequencies ω = ωn = 2πn/Tw with Z ∞
n = 0, ±1, ±2, . . .. The inverse Fourier transform returns dt φ∗ (t)φ(t − u)
−∞
f (t) within the interval Tw , Z ∞ Z ω2 Z ω2
0

∞ = dt dω dω 0 eiωt e−iω (t−u)

1 X Tw Tw −∞ ω ω1
f (t) = e−iωn t fˆ(ωn ) for − <t< . (262) Z ω2 1
Tw n=−∞ 2 2 iωu
= 2π dωe
ω
If the function f (t) is sufficiently well behaved, one can ,1ω2
2π
take the limit Tw → ∞, in which case the transform and = eiωu
its inverse relation take the forms iu
ω1
Z ∞
4π i(ω1 +ω2 )u/2 (ω2 − ω1 )u
fˆ(ω) = dt eiωt f (t), (263) = e sin . (268)
−∞ u 2
Z ∞
dω −iωt ˆ
f (t) = e f (ω). (264) This vanishes for u = 2πn/(ω2 −ω1 ) for integers n 6= 0. This
−∞ 2π shows that the next orthogonal function can be placed at a
Whenever there is doubt about using these, one can go distance u satisfying u(ω2 − ω1 ) = 2π. We can now tile the
back to the case of finite Tw , where possible problems are area in t − ν plane in different ways, either selecting ω2 and
easier to solve. Instead of the angular frequency ω one can ω1 as the end points of the range ∆ω, or by dividing this
also use the frequency ν = ω/2π, which sometimes is a in several frequency intervals, and define the orthogonal
more convenient variable. wavelets for each subinterval as above. All choices lead to
the same result (266) asymptotically for ∆t∆ω  1.
Consider now a f (t) which is essentially different from
zero only during a finite time, a pulse. From the formulas
above we see that its Fourier transform is nonzero in a 6.3 Landauer formula
finite interval of frequencies, i.e. the pulse is formed by a We can now apply the result (266) to calculate the con-
wave packet, which contains waves of different frequencies. ductivity of the wire. Consider first one transverse mode
In more detail, one can prove the uncertainty relation only. The electric current I = ene /∆t is the electric charge
1 1 e times the number of transmitted electrons ne per time ∆t.
σt σω ≥ ⇔ σt σν ≥ . (265) The wave functions that carry the current are in the energy
2 4π
range ∆E = eV and thus frequency range ∆ν = eV /h. The
Here σt is the standard deviation of t in the probability number of transmitted orthogonal wave functions is given
distribution P (t) ∝ |f (t)|2 and σν is the same for ν in by n = ∆t∆ν (266). Since electrons are fermions, there
P (ν) ∝ |fˆ(ν)|2 . are two electrons (because of spin) per each wave function.
Thus the current is
More important than the uncertainty relation (265) for
us is the following question. Given a time window ∆t and a 2e2
frequency window ∆ν = ∆ω/2π, what is the number n of I= V (269)
h
mutually orthogonal functions it can support. To be preci-
se, we expect the answer to be valid only asymptotically or the conductance
as ∆t∆ω  1. The answer, as we discuss below, is
2e2
G= . (270)
1 h
n= ∆t∆ω = ∆t∆ν. (266)
2π
This is the Landauer formula in its simplest form (one per-
We can interpret the result saying that each function takes fectly transmitting mode at T = 0). This is an amazingly
a unit area in t − ν plane. (This area is by factor 4π larger simple result since it does not depend on any materials pa-
than the minimum value of σt σν .) Notice also that the rameters, it contains only the fundamental constants e and

27
h. (We return to the discussion soon.) The inverse of (270) This is the same result as obtained in the t − ω space.
is known as “quantum resistance” Looking at the k − E pictures presented above, we see that
different modes have different ∆p but this effect is just
h
RQ = 2 = 12.9 kΩ. (271) cancelled by the differences in the velocity v (258) so that
2e each mode contributes just proportional to v∆p = ∆E.
The simple form of the Landauer formula allows seve- (Here the formulas assumed a small ∆p. If that is not the
ral generalizations. First, we can generalize to arbitrary case one should divide ∆p into thinner R slices and integrate,
number M of modes simply by multiplying the current but the result is unchanged because v dp = ∆E.)
by M . Second, we can generalize to imperfect wire. There Let us discuss the relation to information theory. For
the waves are transmitted with probability T and reflec- fermions each state can be either occupied or not. Thus
ted back by probability R = 1 − T . Since only transmitted the information that can be coded is one bit (or two bits
waves contribute to the current, the current is given by including the spin) per one orthogonal wave function. For
I = (2e2 /h)M T V . Third, we can generalize to arbitrary bosons there are more possibilities since one can have dif-
temperature by applying the argument leading to Landau- ferent numbers of bosons in one state. For example, with
er formula to arbitrary thin energy slices separately. The 16 different amplitude levels, one can code 4 bits of data.
result is The Shannon formula for channel capacity is
2e2
Z
I= dEM (E)T (E)[fL (E) − fR (E)] (272) C = M ∆ν log2 (1 + S/N ), (276)
h
with integration over the energy band. Here fL (E) and where C is the maximum number of bits transferred per
fR (E) are the Fermi distribution functions on the left and time and S/N is the “signal to noise ratio” that determines
right reservoirs. We have also included the possibility that how many amplitude levels can be resolved.
both the transmission probability and the number of mo- The Landauer formula is best demonstrated experimen-
des are functions of energy. Fourth, one can consider a ge- tally, if one has a method to change the mode number or
neral scattering from any incoming mode to any reflected the transmission. One system where this can be done is a
or transmitted mode. Without going into details (see Dat- semiconductor heterostructure, where one can adjust the
ta) we claim that by making a unitary transformation on width of the conducting region by gate electrodes.
the modes, one can make the transmission probability dia-
gonal, and the only
P effect in formula (272) is to replace
M (E)T (E) → i Ti (E), where Ti are the elements of
the diagonalized transmission probability matrix. Also, we
can argue that the Landauer formula is independent of the
specific assumptions of effective mass approximation or the
form of the potential in the effective Scrhödinger equation
(255). Thus we have essentially removed all assumptions
made at the start except the one of the coherent trans-
port.
Figure: B. J. van Wees et al, Phys. Rev. Lett. 60, 848
Let us work out the current (272) in the case there is no
(1988). The conductance in a semiconductor heterostruc-
energy dependence on M and T . At any temperature the
ture as a function of gate voltage. Quantized values of G
Fermi distribution gives
Z are seen when there is one or more modes whose Ti ≈ 1
and for other modes Ti ≈ 0.
dE[fL (E) − fR (E)] = µL − µR = eV (273)
Now that we roughly understand the plateaus in the
and thus plot, let us next concentrate on what happens at the steps
between the plateaus.
2e2
G= MT , (274)
h
which is an often appearing form of the Landauer formula. 6.4 Classical shot noise
To confirm that all has been done right, let us still con- Let us consider a mode with a transmission probability
sider deriving the Landauer formula in x − p variables. The T ∼ 12 . This means that some of the incident electrons
formula (266) in t − ω space can be directly applied in the are transmitted and some are reflected back. Compared to
spatial coordinate vs. wave vector space, x − k, and num- the case of T = 1, where a uniform flow of electrons takes
ber of orthogonal wave functions is thus n = ∆x∆k/2π. place (as will be shown later), there are now fluctuations
Remebering that p = ~k we can extent this to x − p phase in the current. Such a noise is called shot noise. It was first
space, implying that one wave function takes an area h in analyzed by W. Schottky in electron tubes (1918).
the phase space. The current is thus Let us first study the shot noise using classical physics.
ene 2e∆x∆p 2e 2e We assume that electrons arrive randomly at a rate λ and
I= = = v∆p = ∆E. (275) the arrivals are instantaneous. Thus the current is a sum
∆t h∆t h h

28
of delta functions This is the Schottky result that the shot noise spectral
X density is directly proportional to the current. This formula
I(t) = e δ(t − tn ), (277)
is also interesting because the constant of proportionality
n
is the electron charge. Even though the effect of one single
and the arrival times tn have the probability distribution electron may be too small to resolve, the discreteness of the
P (tn − tn−1 ) = λe−λ(tn −tn−1 ) Θ(tn − tn−1 ), (278) charge still could be seen in the fluctuations of the current.
Electron shot noise could be compared to rain falling on
where Θ is the step function (5), implying that tn ≥ tn−1 .
a roof. This makes noise because rain is not a continuos
We see that the average interval between two arrivals is
flow of water but consists of droplets. We hear the noise
1/λ and therefore the average current
although it may be difficult to resolve the sound of the
hIi = eλ. (279) individual droplets.
Also one can show that charge Q = en collected in time We see that similarly as the thermal noise (244), the
∆t obeys Poisson distribution shot noise is white, i.e. independent of frequency. This re-
n
sult is supported by experiments in a range of frequencies,
−θ θ but there are deviations for both at very high frequencies
Pn = e (280)
n! and at low frequencies. At very high frequencies one should
with the mean value θ = λ∆t (exercise). start to reveal the time structure of a single-electron trans-
1.0
mission event, which here was idealized to be a delta func-
tion. At low frequencies the experiments reveal a noise that
0.8
is larger than the Schottky result. This is called 1/f noise,
P5 P6
0.6 P3 P4 the name indicating that the spectral density increases at
P2 low frequencies approximately as 1/frequency. 1/f -noise is
0.4 P1
interpreted so that systems are not constant in time. For
0.2 P0 example, an impurity atom in a resistor changes state and
results in a change of the resistance. This noise occurs in
θ
0 1 2 3 4 5 all systems but its magnitude varies. It was first analyzed
Pn(θ=5) in electron tubes, see J. B. Johnson, Phys. Rev. 26, 71-85
0.20 (1925) for a nice historic study.
0.15
6.5 Transmission and noise in scattering
0.10
formalism
0.05 We return to consider the conducting wire. Now we ma-
ke a quantum mechanical calculation using the field opera-
n
0 2 4 6 8 10 12 14 tors. The calculation is rather formal and therefore not ve-
ry transparent. However, as the end result it gives several
Next we want to calculate the autocorrelation function already familiar results, like the Landauer formula, ther-
(236) for the electric current, mal and shot noise and the fluctuation-dissipation theory
RI (τ ) = hδI(t)δI(t − τ )i, (281) in addition to new results like combinations of thermal,
shot and quantum noise. The shot noise by this method
where δI = I − eλ. We argue physically that since the was first calculated by G. Lesovik (1989). For further de-
arrival times are random, the main contribution to RI co- tails see Blanter&Büttiker.
mes from the correlation of an electron with itself. Thus
we claim that the main part comes from terms We think of dividing the wire into a scattering region
and two “leads” on both sides of it. We label the leads by
RI (τ ) = hδI(t)δI(t − τ )i α that is either left or right. Both leads are assumed to ha-
ve a single mode only, for simplicity. In both leads we have
X
2
≈ e h δ(t − tn )δ(t − τ − tn )i
n
a coordinate xα that is positive towards the scattering re-
X gion. We consider the annihilation and creation operators
= e2 h δ(t − tn )iδ(−τ ) for electron waves propagating in the leads. Here aαk are
n
the annihilation operators for the waves moving in the po-
= e2 λδ(−τ ). (282) sitive x direction (incoming waves) and b the annihila-
α αk
In fact this result is exact, since the dropped terms can be tion operators for the waves moving in the negative xα di-
shown to cancel each other. The spectral density (237) is rection (reflected and transmitted waves), both with wave
then number k. These are required to obey the anticommutator
Z T /2 rules
w

SI (ω) = dτ eiωτ RI (τ ) = e2 λ
−Tw /2 {aαk , a†αk0 } = δk,k0 , {bαk , b†αk0 } = δk,k0 ,
= ehIi. (283) {aαk , aαk0 } = {a†αk , a†αk0 } = 0,

29
 = {bαk , bαk0 } = {b†αk , b†αk0 } = 0, (284)
We calculate its expectation value assuming the particles
arriving from the reservoirs are in thermal equilibrium, and
where {A, B} = AB + BA. These rules imply the Fermi
get
statistics, where the only occupations allowed in one state
are 0 and 1. hX̂kk0 i = δk,k0 |t|2 [fα (Ek ) − fβ (Ek )]. (291)
We define the field operator
Therefore
1 X ikxα
ψ̂α (xα , t) = √ (e aαk + e−ikxα bαk )e−iEαk t/~ , ge X ~k 2
L k>0 hIˆα (t)i = |t| [fα (Ek ) − fβ (Ek )]
L m
k>0
(285)
ge ∞ ge2 2
Z
where Eαk = Eα0 + ~2 k 2 /2m. The current operator in lead = dE|t|2 [fα (E) − fβ (E)] = |t| V (292)
α is defined as if ψ̂α were a usual wave function, h 0 h
"
ge~ † d assuming t independent of energy (or a small V and T ) in
ˆ
Iα (t) = ψ̂α (xα , t) ψ̂α (xα , t)
2mi dxα the last equality. Thus we have Ohm’s law I = GV with
ge2 2
  #
d † G = |t| . (293)
− ψ̂ (xα , t) ψ̂α (xα , t) . (286) h
dxα α
We have rederived the Landauer equation with the field-
Here g = 2 in order to include the electron spin. The idea theory formalism.
here is that applied to usual wave function ψ(xα ), Eq. (286)
gives the usual expression for current, and the only gene- Now consider the fluctuations. We are looking at a statio-
ralization here is that a and b operators indicate which of nary process, and therefore can add to the autocorrelation
these states are occupied. We substitute the field opera- function (236) an extra integration over time,
tors. Using that k varies only in a small region compared Z Tw /2
1
to its value, we take k outside of the summation (but still Rf (τ ) = dthδf (t)δf (t − τ )i (294)
preserve the difference in energies) and get Tw −Tw /2

ge~ 1 X X and we would like to take the limit Tw → ∞. For the power
Iˆα (t) = 2ikei(Eαk −Eαk0 )t/~
2mi L 0
spectrum we get
k>0 k >0

×(a†αk aαk0 − b†αk bαk0 )

Z Tw /2
Sf (ω) = dτ eiωτ Rf (τ )
ge X X ~k i(Eαk −Eαk0 )t/~ −Tw /2
= e
L m
Z Z
1
k>0 k0 >0 = dτ dteiωτ hδf (t)δf (t − τ )i
† † Tw
×(aαk aαk0 − bαk bαk0 ). (287) Z Z
1
= h dtδf (t)eiωt dτ δf (t − τ )e−iω(t−τ ) i
Using Tw
Z 1
dtei(~ω+Ek −Ek0 )t/~ = Tw δ~ω+Ek ,Ek0 → hδ(~ω+Ek −Ek0 ) = hδf (ω)δf (−ω)i, (295)
Tw
(288)
which is known as Wiener-Khinchin theorem. [In fact, one
we get
Z could define the spectral density Sf with the last form of
(295), and then the Wiener-Khinchin theorem would be
Iˆα (ω) = dte Iˆα (t)
iωt
the same as our definition of Sf (237). This way of defi-
geh X X ~k ning Sf is not favored here since δf (ω) diverges in the limit
= δ(~ω + Ek − Ek0 ) Tw → ∞, but Sf (ω) is supposed to stay finite.] [The treat-
L m
k>0 k0 >0 ment of the integration limits in (295) is not completely
×(a†αk aαk0 − b†αk bαk0 ). (289) satisfactory, but apparently the error is small in the limit
Tw → ∞.]
We now take the scattering into account by expressing the
outgoing waves in terms of the incoming, bαE = raαE + We calculate
taβE . Here r and t are the reflection and transmission ge2 h2 X X X X ~2 kq
amplitudes. In order to conserve unitarity |t|2 + |r|2 = 1. SI (ω) =
∆tL2 m2
This allows us to determine the reflection and transmission k>0 k0 >0 q>0 q 0 >0
probabilities R = |r|2 and T = |t|2 . We define a short hand ×δ(~ω + Ek − Ek0 )δ(−~ω + Eq − Eq0 )hδ X̂kk0 δ X̂qq0 i.
X̂kk0 = a†αk aαk0 − b†αk bαk0 (296)
= a†αk aαk0 − (r∗ a†αk + t∗ a†βk )(raαk0 + taβk0 ) For that we need the expectation value of for annihila-
= |t|2 a†αk aαk0 − |t|2 a†βk aβk0 tion/creation operators, and we calculate

−r∗ ta†αk aβk0 − rt∗ a†βk aαk0 . (290) ha†1 a2 a†3 a4 i = ha†1 a2 iha†3 a4 i + ha†1 a4 iha2 a†3 i, (297)

30
 hδXkk0 δXqq0 i = hXkk0 Xqq0 i − hXkk0 ihXqq0 i coupling the system to a set of harmonic oscillators, where
= − |t|2 a†βk aβk0 − r∗ ta†αk aβk0 − rt∗ a†βk aαk0 )the oscillation quanta are bosons. Here the coupling is to
h(|t|2 a†αk aαk0
electrons, which are fermions, but the thermal noise is the
×(|t|2 a†αq aαq0 − |t|2 a†βq aβq0 − r∗ ta†αq aβq0 − rt∗ a†βq aαq0 )i same. We can interpret the result that at T > 0 there
−hXkk0 ihXqq0 i are electron and hole excitations that can be transmitted
4
= δkq0 δk0 q {|t| fα (Ek )[1 − fα (Ek0 )] through the junction causing the current that has nonzero
fluctuations (302) but whose average vanishes (since V =
+|t|4 fβ (Ek )[1 − fβ (Ek0 )] 0).
+|rt|2 fα (Ek )[1 − fβ (Ek0 )]
2) Shot noise. We assume ω = 0 and T = 0 and get from
+|rt|2 fβ (Ek )[1 − fα (Ek0 )]}i (299)
= δkq0 δk0 q |t|2 {|t|2 fα (Ek )[1 − fα (Ek0 )]
ge2 2 2
Z
2
+|t| fβ (Ek )[1 − fβ (Ek0 )] SI (0) = |t| |r| dE{fα (E)[1 − fβ (E)]
h
2
+|r| fα (Ek )[1 − fβ (Ek0 )] +fβ (E)[1 − fα (E)]}
+|r|2 fβ (Ek )[1 − fα (Ek0 )]}, (298) ge2 2
= |t| (1 − |t|2 )|eV |. (303)
h
ge2 h2 X X ~2 k 2 In the limit |t|2  1 this reduces to the Schottky result
SI (ω) = SI = |eI| (283). In the opposite limit |t|2 = 1 one gets the
∆tL2 m2
k>0 k0 >0 interesting result that the shot noise vanishes for perfect
δ(~ω + Ek − Ek0 )δ(−~ω + Ek0 − Ek )|t|2 transmission. Thus the current flow at temperature T =
×{|t|2 fα (Ek )[1 − fα (Ek0 )] + |t|2 fβ (Ek )[1 − fβ (Ek0 )] 0 is “quiet” in a perfect conductor. This applies at the
2 2 plateaus in the conductance figure above [after eq. (276)],
+|r| fα (Ek )[1 − fβ (Ek0 )] + |r| fβ (Ek )[1 − fα (Ek0 )]}
whereas at the steps the current is noisy.
ge2 h X X ~2 k 2 2
= δ(~ω + E k − E k 0 )|t| Shot noise is often quantified by “Fano factor”, which
L2 m2
k>0 k0 >0 is the ratio F = SI /ehIi. Here we get F = 1 − |t|2 . For
×{|t|2 fα (Ek )[1 − fα (Ek0 )] + |t|2 fβ (Ek )[1 − fβ (Ek0 )] classical shot noise this is unity but it vanishes for perfect
+|r| fα (Ek )[1 − fβ (Ek )] + |r| fβ (Ek )[1 − fα (Ek )]}. transmission.
2
0
2
0

3) General ω = 0 gives from (300)

Going to the continuum limit we get
ge2
Z
SI (0) = dE|t|2 {fα (E)[1 − fα (E)]
ge2
Z Z
h
SI (ω) = dE dE 0 δ(~ω + E − E 0 )|t|2
h +fβ (E)[1 − fβ (E)] + |r|2 [fα (E) − fβ (E)]2 }. (304)
×{|t|2 fα (E)[1 − fα (E 0 )] + |t|2 fβ (E)[1 − fβ (E 0 )] We calculate, aided by Mathematica,
+|r|2 fα (E)[1 − fβ (E 0 )] + |r|2 fβ (E)[1 − fα (E 0 )]}. Z
(299) dE[f (E + eV /2) − f (E − eV /2)]2
 
eV eV
An alternative form of the same result is = 2kB T coth −1 , (305)
2kB T 2kB T
ge2
Z Z
SI (ω) = dE dE 0 δ(~ω + E − E 0 )|t|2 and therfore
h ge2 2

eV

×{fα (E)[1 − fα (E 0 )] + fβ (E)[1 − fβ (E 0 )] SI (0) = |t| 2kB T |t|2 + eV |r|2 coth . (306)
h 2kB T
+|r|2 [fα (E)fα (E 0 ) − fα (E)fβ (E 0 ) For a tunneling barrier, where the transmission probability
+fβ (E)fβ (E 0 ) − fβ (E)fα (E 0 )]}. (300) is small (|t|2  1), only the second term is important and

Below we consider special cases of this formula. ge2 2 eV

SI (0) = |t| eV coth . (307)
h 2kB T
1) Thermal equilibrium. We have fα (E) = fβ (E) and we
are interested in ω = 0. We get from (299) We see that the total noise is not a simple sum of thermal
and shot noise.
2ge2 2 SI
Z
SI (0) = |t| dEf (E)[1 − f (E)]. (301) G T kB
h 8

Using f (E)[1 − f (E)] = −kB T df (E)/dE we get 6

2ge2 2 4
SI (0) = |t| kB T = 2GkB T, (302)
h 2

which is the Johnson-Nyquist noise (244). It is interesting eV

to note that previously we derived the same formula by 0 2 4 6 8 T kB

31
 SI
4) General |t| independent of energy. We calculate, aided
eGV
by Mathematica, 4

eβ∆E
Z 3
dEf (E)[1 − f (E + ∆E)] = ∆E β∆E
e −1 2
 
1 ∆E
= ∆E 1 + coth , (308)
2 2kB T 1

ΩÑ
-2 -1 0 1 2 eV
Z
dE{f (E)[1 − f (E + ~ω + eV )] Figure: noise as a function of ω at T = 0 and T =
eV /4kB for small transmission.
+f (E + eV )[1 − f (E + ~ω)]}
1

~ω + eV
 5) Long conductor: a long conductor can be considered
= (~ω + eV ) 1 + coth as consisting of several phase coherent systems in series.
2 2kB T
  This leads to cancellation of shot noise, so that the thermal
1 ~ω − eV noise dominates in long conductors. More precisely, from
+ (~ω − eV ) 1 + coth
2 2kB T (302) we get for thermal noise
1 ~ω + eV
= ~ω + (~ω + eV ) coth SV (0) = R2 SI (0) = 2RkB T, (312)
2 2kB T
1 ~ω − eV
+ (~ω − eV ) coth , (309) and for shot noise (303)
2 2kB T
SV (0) = R2 SI (0) = eR2 I = eRV. (313)
and we get from (299)
Here the first is additive in resistance but the latter is not.
ge2
Z
SI (ω) = dE|t|2 {|t|2 fα (E)[1 − fα (E + ~ω)] The same conclusion holds also at finite frequencies, i.e. in
h an long conductor
+|t|2 fβ (E)[1 − fβ (E + ~ω)]
~ω 2R~ω
+|r|2 fα (E)[1 − fβ (E + ~ω)] SV (ω) = R~ω(1 + coth )= . (314)
2kB T 1 − e−β~ω
+|r|2 fβ (E)[1 − fα (E + ~ω)]}
ge2 2 2 ~ω
= |t| {|t| ~ω(1 + coth )
h 2kB T
1 ~ω + eV
+|r|2 [~ω + (~ω + eV ) coth
2 2kB T
1 ~ω − eV
+ (~ω − eV ) coth ]}
2 2kB T
ge2 2 ~ω
= |t| {~ω + |t|2 ~ω coth
h 2kB T
1 ~ω + eV
+|r|2 [ (~ω + eV ) coth
2 2kB T
1 ~ω − eV
+ (~ω − eV ) coth ]}. (310)
2 2kB T

Note that the first term is odd in ω and the rest is even.
At V = 0 this reduces to

~ω 2G~ω
SI (ω) = G~ω(1 + coth )= , (311)
2kB T 1 − e−β~ω

which agrees with the fluctuation-dissipation theorem

(245).
At finite V but temperature T = 0, SI (310) at a given
ω consists of three linear segments as a function of V . The
same holds as a function of ω at a fixed V . At finite T
there the two corners are rounded. The shot noise (at T =
0) is limited to frequencies less than ν = eV /h. This is
the rate of electrons (in one transverse mode) attempting
transmission according to (266) for ∆ν = eV /h.

32
7. Conclusion
As a summary, we started by studying the driven and
damped harmonic oscillator. In the quantized version, the
damping has to be represented by a coupling to a bath,
which was chosen to consist of harmonic oscillators. In
addition to damping, the bath also produces fluctuations.
The theory presented for harmonic oscillator can be simply
generalized to two-state systems, and to more complica-
ted systems. Finally, we studied the fluctuations caused by
coupling to a bath of electrons. The hope is that this cour-
se forms a good starting point for understanding current
research on topics that bridge quantum optics and electron
transport in mesoscopic condensed matter structures.

33