)
Dirk Englund
5. Interaction between atoms and fields
Overview † Quantum theory for atomfield interaction † A twolevel system in a singlemode EM field: Rabi oscillations † Normal mode splitting † Multimode interaction † WignerWeisskopf theory of spontaneous emission References † S. Haroche and D. Kleppner, “Cavity quantum electrodynamics”, Physics Today, January 1989, pp. 2430 2. † K. Vahala, “Optical microcavities”, Nature, vol. 424, pp. 839846 (Aug. 14 2003) † H. J. Kimble, “Structure and dynamics in cavity quantum electrodynamics”, in Cavity Quantum Electrodynamics, edited by P. Berman, pp. 203267 (Academic Press 1994) † Y. Yamamoto and A. Imamoglu, Mesoscopic quantum optics, Wiley and sons, 1999 (chapter 6) † M. O. Scully and M. S. Zubairy, Quantum optics, Cambridge University Press 1997 (chapters 6, 9) † H. Mabuchi and A. C. Doherty, “Cavity quantum electrodynamics: coherence in context,” Science, vol. 298, pp. 13721377 (2002). Atomfield interaction An emitter and an EM field interact primarily through the EM's electric field and the electronic dipole of the emitter, which can be induced or always present, e.g., in the case of a polar molecule. It is almost always a good approximation to assume that the extent of the dipole is much smaller than the wavelength of the EM field. In this case, we can consider the electric field to be given at just one point, and we obtain the interaction Hamiltonian: Hint,2 =  m × E (0.1) E = E HrA L is the electric field at the location of the atom m = Ÿ rHrL r d3 r is the electronic dipole moment The minus sign means that the energy is lowest if the dipole is parallel to the electric field. Let's consider now a simple atom, which has just one electron, that interacts with the electromagnetic field in an optical cavity. For simlicity, let’s consider only a single mode of the cavity for now. The electron has a charge qe = 1.6 ÿ 1019 C and is at position r with respect to the nucleus (see figure (a) below). Hint,2 = Hqe rL × E
2
Chap_5.nb
qe
r
(a)
1 Ν 2
Ω
(b)
Figure. (a) The electron (charge qe ) at a position r with respect to the nucleus. (b) The electronic state is approximated as a twolevel system with excited states Ie\ and ground state g\, with transition energy —w. The single mode EM field has frequency n.
g
e
JaynesCummings Hamiltonian Let's consider a single, linearly polarized mode at a frequency n in the cavity that interacts with the atom. The total Hamiltonian is given by H = HA + HF + Hint (0.2) In addition to the interaction Hamiltonian Hint , we also have terms for the atom energy (HA ) and the field energy (HF L. These have the following form: HA = 1 2 ª — w e_ [e 1 2 — w sz 1 2 — w g_ [g (0.3)
where sz ª e\ Xe  g\ Xg is the population operator and we defined w ª weg , the transition frequency of the atom. The field energy is given by the familiar expression ` ` 1 1 HF = 2 — n Ia† a + 2 M
Finally, the interaction Hamiltonian is Hint = Hqe rL × E. For this term, we first need to derive the form in the quantum theory. We assume that although the atoms may have many states, only one transition between two levels is close to the EM frequency n. Let's also assume that the lower of these transitions is the electron's ground state in the atom. As a result, we can assume that only two states of the atom are involved: a ground state g\ and an excited state e\. The states e\ and g\ represent the complete basis set of the atom, so e\ Xe + g\ Xg = 1. We may now expand qe r = 1 ÿ qe r ÿ 1 = H e\ Xe + g\ Xg L qe rH e\ Xe qe r = qe H e\ Xe r e\ Xe + g\ Xg r e\ Xe + e\ Xe r g\ Xg + g\ Xg r g\ Xg L + g\ Xg L (0.4)
` Now define the matrix element mij = qe Xi r j\, the atom transition operator sij = i] X j , the raising operator ` ` s+ = e\ Xg , and the lowering operator s = g\ Xe . Now we have ` ` ` ` (0.5) qe r = Imee see + mgg sgg + meg seg + mge sge M The electric field at the location of the atom is
Chap_5.nb
3
EHrA L =
—w 2V
` ` ` sin kz z Ia + a† M e,
(0.6)
` where e is the polarization direction of the field and z is the position of the atom. Let's write this for a general ` electric field, replacing sin kz z e by the normalize field EHrA L ê maxH EHrL L, where rA is the position of the atom inside the cavity. Then we have EHrA L = —w 2V 1 maxH EHrL L —w 2 V maxH EHrL L
2
` ` IEHrA L a + E* Hr A L a† M
(0.7)
EHrA L = i
` ` IEHr A L a + E* Hr A L a† M
(0.8)
Let's introduce the mode volume for a general dielectric cavity  one which may have a varying dielectric index. The mode volume is given by Vmode =
3 Ÿ d r eHrL EHrL 2
maxIeHrL EHrL 2 M
,
(0.9)
The electric field is then EHrA L = —w 2 Vmode maxHeHrL EHrL L
2
` ` IEHr A L a + E* Hr A L a† M
(0.10)
Now, because the position operator is an odd function, the parity of the states means that mii = qe Xi r i\ = 0. Also, meg = mge * . Thus, Hint = qe r × E = —w 2 Vmode maxIe0 EHrL M —w 2 Vmode maxIe0 EHrL M
2 2
(0.11) ` ` ` ` Imeg seg + mge sge M ÿ IEHrA L a + E* HrA L a† M (0.12)
=
` ` ` ` Imeg ÿ seg EHrA L a + meg * ÿ seg E* HrA L a†
(0.13)
` ` ` ` + mge sge ÿ EHrA L a + meg * sge ÿ E* HrA L a† M = —w 2 Vmode maxIe0 EHrL M
2
` ` ` ` Imeg ÿ s+ EHrA L a + meg * ÿ s+ E* HrA L a†
(0.14)
` ` ` ` + mge ÿ s E HrA L a + meg * ÿ s E* HrA L a† M ` ` ` ` The terms that go like s+ a† and s a do not conserve energy and are dropped. Then
4
Chap_5.nb
Hint =
—w 2 Vmode maxIe0 EHrL M
2
` ` ` ` Imeg ÿ s+ EHrA L a + meg * ÿ s E* HrA L a† M
(0.15)
Now introduce the coupling parameter gHrL : gHrL = 1 — —w 2 Vmode maxIe0 EHrL 2 M Imeg ÿ EHrA LM (0.16)
Now the interaction Hamiltonian is ` ` ` ` Hint = — Ig* HrL s a† + gHrL s+ a M Let's rewrite the atomfield coupling parameter as gHrL = where † g0 =  meg — EHrA L —w 2 Vmode e0 eM = Rabi frequency 1 — —w 2 Vmode maxIe0 EHrL 2 M Imeg ÿ EHrA LM = g0 yHrA L cos x
(0.17)
(0.18)
† yHrA L =
= atom position mode overlap EHrM L meg EHrA L ÿ = atom orientation in the field † cosHxL = EHrA L meg † rM = point where the field intensity eHrL EHrL 2 is maximized † eM = eHrM L 3 2 Ÿ d r eHrL EHrL † Vmode = eM EHrM L 2 We see that g0 is maximized for perfect field and polarization overlap, and that g0 increases as Summarizing, the total Hamiltonian is given by H = HA + HF + Hint Where: HA = 1 2 — w sz (0.20) (0.19)
1 V
.
` ` 1 HF = — n a† a + 2 ` `† ` ` * Hint = — Ig HrL s a + gHrL s+ a M Strong and weak coupling regimes of cavity QED:
Chap_5.nb
5
 strong coupling: g > k, g  weak coupling: g < k or g
Interaction Picture
[reference: Scully/Zubairy, Ch. 5]
Suppose the total Hamiltonian is given by H = H0 + Hint , (0.21) where we already know the eigensystem for the Hamiltonian H0 , and the term Hint represents an interaction Hamiltonian. The Schrödinger equation i — ¶∂t yHtL\ = H yHtL\ can be formally integrated for the state ket yHtL\ = UHtL yH0L\ where the time evolution operator obeys (0.24) H UHtL , with an initial value of UH0L = 1. — In general, the time evolution has rapidly varying terms due to the constant part of the Hamiltonian. We can cancel out these fast rotations by choosing a frame which rotates at an angular frequency H0 /—. In this interaction picture, the state vectors will then only rotate at a rate that depends on the interaction term Hint . To this end, we define ¶∂t UHtL = yI HtL\ = exp Then ¶∂t yI HtL\ = B¶∂t exp =B i — H0 exp i — i — i — i — H0 t F yHtL_ + exp i — i — i — H0 t ¶∂t yHtL_ i — i — HH0 + Hint L yHtL_ F HH0 + Hint L yHtL_ F (0.26) (0.27) (0.28) (0.29) i — H0 t yHtL_ (0.25) i (0.23) (0.22)
H0 t F yHtL_ + exp i — H0 yHtL_F + exp i — Hint F yHtL_
H0 t BH0 t B
= exp = exp Finally,
H0 t B
H0 t B
¶∂t yI HtL\ = ¶∂t yI HtL\ = where
i — i —
exp
i —
H0 t Hint exp 
i —
H0 t
yI HtL_
(0.30) (0.31)
U0 † HtL Hint U0 HtL yI HtL_
6
Chap_5.nb
U0 HtL = exp 
i —
H0 t
(0.32)
We may define Hint in the interaction picture: HI = U0 † HtL Hint U0 HtL Then the state evolves in the interaction picture according to ¶∂t yI HtL\ = i — HI yI HtL_ (0.34) (0.33)
` An operator AS in the Schrödinger picture is transformed in the interaction picture as ` ` AI = U0 † HtL AS U0 HtL A twolevel atom interacting with a singlemode field Recall from above that the total Hamiltonian for this problem is H = H0 + Hint Where: H0 = HA + HF = Hint ` ` 1 — w sz + — n a† a + 2 2 ` `† ` ` = — g Is a + s+ a M 1
(0.35)
(0.36)
(0.37)
In the interaction picture, HI = U0 † HtL Hint U0 HtL ` ` ` ` = — U0 † HtL g Is a† + s+ a M U0 HtL ` ` ` ` = — g IU0 † HtL s U0 HtL U0 † HtL a† U0 HtL + U0 † HtL s+ a U0 HtLM ` ` ` ` = — g IIei w t s M Iei n t a† M + U0 † HtL s+ a U0 HtLM, where we used the identity @A, @A, BDD + ... 2! ` from Chapter 1 and transformed a† as ` U0 † HtL a† U0 HtL = ei w t Ia
†
(0.38) (0.39) (0.40) (0.41)
ea A B ea A = B  a@A, BD +
a2
(0.42)
aM
Hi w tL2 † ` ` a† ei w t Ia aM = a† + Ii w tAa† a, a† EM + IAa† a, Aa† a, a† EEM + ... 2! Hi w tL2 2! Ia† M + ...
(0.43)
` = a† + Ii w t a† M + ` = a† ei w t
` ` ` and similarly for U0 † HtL s U0 HtL and U0 † HtL s+ a U0 HtL. Then we arrive at the interaction Hamiltonian in the interaction frame:
Chap_5.nb
7
` ` ` and similarly for U0 † HtL s U0 HtL and U0 † HtL s+ a U0 HtL. Then we arrive at the interaction Hamiltonian in the interaction frame: ` ` ` ` (0.44) HI = — gIei D t s a† + ei D t s+ a M, where D = w  n. Solution by probability amplitude [reference: Scully/Zubairy] The state in the interaction picture is yI HtL\, which evolves according to the Schrödinger equation HI yI HtL_ — The state in the interaction frame is given by slowly varying amplitudes ce,n and cg,n : ¶∂t yI HtL\ = yI HtL\ = ‚ Acg,n g, n] + ce,n e, n]E
n
i
(0.45)
(0.46) (0.47) (0.48) (0.49) (0.50) (0.51) n e, n  1]E (0.52)
¶∂t yI HtL\ = 
i —
HI yI HtL_
i ` `† ` ` ° ° ‚n Acg,n g, n] + ce,n e, n]E =  I— g Iei D t s a + ei D t s+ a MM ‚n Acg,n g, n] + ce,n e, n_F — ` `† ` ` ° ° ‚n Acg,n g, n] + ce,n e, n]E =  i g Iei D t s a + ei D t s+ a M ‚n Acg,n g, n] + ce,n e, n]E ` ` ` ` ` ` Iei D t s a† + ei D t s+ a M ‚ Acg,n g, n] + ce,n e, n]E = ‚ Acg,n ei D t s a† g, n] + n n ` ` ` ` ` ` ce,n ei D t s a† e, n] + cg,n ei D t s+ a g, n] + ce,n ei D t s+ a e, n]E = ‚ A0 + ce,n ei D t n n + 1 g, n + 1] + cg,n ei D t n e, n  1] + 0E n + 1 g, n + 1] + cg,n ei D t
° ° ‚n Acg,n g, n] + ce,n e, n]E =  i g ‚n Ace,n ei D t Now premultiply by Xg, n + 1 ï ° cg,n+1 = i g ce,n ei D t n+1
(0.53)
Premultiply by Xe, n  1 ï ° c = i g c ei D t n
e,n1 g,n
(0.54)
or ° ce,n = i g cg,n+1 ei D t In summary, ° cg,n+1 = i g ce,n ei D t n + 1 ° ce,n = i g cg,n+1 ei D t n + 1 (0.56) n+1 (0.55)
We can solve this by the Laplace transform with initial conditions, let's say cg,n H0L = 1 and cg,n H0L = 0. Then we have
8
Chap_5.nb
cg,n+1 HtL = Bcos ce,n = where 2ig
Wn t 2 n+1

iD Wn
sin
Wn t 2
F ei D tê2 (0.57)
Wn
sin
Wn t 2
ei D tê2
Wn 2 = D2 + 4 g2 Hn + 1L. Suppose that D=0, so cg,n+1 HtL = cosIg ce,n =  i sinIg n+1 tM n + 1 tM
(0.58)
(0.59)
Note that even for n = 0, the system undergoes oscillations! Let's consider an InAs QD in PC nanocavity. g0 = meg — l n
3
—w 2 Vmode eM e0 ; w= 2pc l ;
(0.60)
Vmode º
(0.61) (0.62)
meg º  qe ÿ d, d = size of QD ~ 1 nm ï meg ~1.6 ÿ 1028 C ÿ m Then
ï
g0 2p
=
meg —
—w 2 Vmode e0
º 60 GHz
(0.63)
Chap_5.nb
9
n gê2pHGHzL Dê2pHGHzL
1 100.701 275
Pn,e ; Pn+1,g 1.0
0.8
0.6
0.4
0.2
0
2
4
6
8
10
tHpsL
Normal mode splitting Recall the Jaynes Cummings Hamiltonian: H = H0 + Hint Where: H0 = HA + HF = Hint 1 `` — w sz + — n a a† + 2 2 ` `† ` ` = — g Is a + s+ a M g, n + 1\< 1 (0.65) (0.64)
Let's consider the system states that are connected through the emission or absorption of a photon: 8 e, n\, (0.66) This system is said to be in the n+1excitation manifold. In the absence of interaction, the eigenstates are the states which diagonalize H0 : 8 e, n\, with energies Ee,n = Xe, n H0 e, n\ = 1 2 —w +—n n+ 1 2 1 2 3 2 (0.68) (0.69) g, n + 1\< (0.67)
Eg,n+1 = Xg, n H0 g, n\ = 
—w +—n n+
10
Chap_5.nb
If the interaction is present, we need to diagonalize the total Hamiltonian, H = H0 + Hint =
1 2
Xe, n H e, n\ Xe, n H g, n + 1\ Xg, n + 1 H e, n\ Xg, n + 1 H g, n + 1\
1
(0.70)
=
— w + — n In + 2 M  g* n+1
1
g
n+1
3
 2 — w + — n In + 2 M g
1
(0.71)
=
1 2
— d + — n Hn + 1L  g* n+1
n+1
 2 — d + — n Hn + 1L
(0.72)
where d = w  n. This problem is solved (HW) to given the eigenenergies E± = — w Hn + 1L ± —2 g
2
Hn + 1L +
—d 2
2
(0.73)
Suppose d = w  n = 0. Then E± = — w Hn + 1L ± — g n+1 (0.74)
and the corresponding eigenstates are E+ : n +\ = e, n\ + g, n + 1\ 2 E : n \ = e, n\  g, n + 1\ 2 (0.76) (0.75)
The states 8 e, n\, g, n + 1\< are called the bare states, and the states { n+\,  n\} are termed the dressed states of the coupled cavityatom system. If the system is prepared in an initial state yH0L\, then it will evolve in time according to (0.77) yHtL\ = H n +\ Yn + ei E+ t ê— + n ] Yn  ei E t ê— M yHtL] So suppose that the system is initially in the bare state
the dressed states is expressed with the phase factors proportional to the eigenenergies. For example, if the system is initially prepared in the state ex,n>, the initial state can be expressed also as an equal superposition of the dressed sates: 1n + 2n " (0) = 2 1n e ! iE+t / ! + 2n e !iE!t / ! The time evolution is then expressed as: " (t ) = 2 n +\ + n \ Therefore, ifyH0L\ = e, n\ = the atomfield interaction is turned on and detuning between the cavity field frequency and the atom transition frequency is not present, the energy level E0 splits into 2 two energy levels E+ and E! , corresponding to the even and odd modes, respectively.
Chap_5.nb
11
(0.78)
Then
yHtL\ =
n +\ ei E+ t ê— + n ] ei E t ê— 2
E+ = !# (n + 1) + ! g
(n + 1)
1n =
ex, n + g , n + 1 2
ex,n>, g,n+1>
EE340, Lecture notes, Part 4 (Spring 2007) Jelena Vuckovic Where DE = HE+  E L and E =# (n+ +) E Lgê 2.nThus, 2n = ex, n ! g, n + 1 HE + 1 ! ! ( + 1) E! = ! After —L + is loaded into the cavity 2 e, n\ cosHDE t êan atomg, n + 1\ sinHDE t ê —L and strongly coupled to the cavity field, the cavity This splitting is usuallyE t ê— as the indication that the atomcavity systemand the splitting equal to 2 g (n + 1) occurs. yHtL\ = ei used transmission spectrum changes, has reached the (0.80) strong coupling regime (for reference, see Kimble, e.g.) 2 For example, if we had an empty cavity (no atom or exciton inside it), it would feature the following transmission spectrum: 2 g (n + 1)
E0 = !# (n + 1)
= ei E t ê—
n +\ ei DE t ê2 — + n ! g ei DE1t)ê2 — 2 ] (n + 2
(0.79)
T(#)
T(!)
#0
#
!0 !
11 In the weakcoupling case, the irreversible decay rates dominate over the coherent coupling rate; in other words, the gas phase atoms (incl. Rydberg enough and with This mode splitting has been observed experimentally withatomcavity field system does not have atoms) time to couple coherently before dissipation occurs. The atom attempts to initiate a number Rabi single quantum dots. In the limit of a low field, the atom and cavity may be viewed as destructively, and the two oscillators which oscillations with different frequencies; these oscillations interfere are coupled through a weak spring: irreversible decay (spontaneous emission) occurs. We now derive the spontaneous emission rates for an atom in free space, uniform dielectric, and finally an atom in a cavity.
Spontaneous emission rate in free space: Einstein’s A coefficient In free space, an atom interacts with a continuum of radiation modes. The atom tries to initiate Rabi oscillations with each one of these modes and with different Rabi frequencies, but all of these oscillations interfere destructively and irreversible spontaneous emission occurs. The system is again described by the JaynesCummings Hamiltonian, but with the multimode field: H=HA+HF+Hint , ! 1$ ' † * ! † H F = ! "!k! % ak! ak! + " , and H int = ! i" g k! (rA )ak! " $ $ g k! (rA )" + ak! ! ! 2# & k k ! ! where in free space case g k (rA ) = g 0,k! cos(% k! ), ! ˆ µ "!k! µ ' e! g 0,k! = eg , and cos(% k! ) = eg k . " 2& 0V µeg
"# HA = "z, 2
no coupling Weak coupling regime coupling 4.3.2.
(
)
ˆ V is the quantization volume, and ek! is the orientation of the electric field for the mode ! k . In a box with dimensions V=Lx ( Ly ( Lz and by applying periodic boundary conditions (as we have seen in Chapter 3) the resonant modes are defined as ! 2) 2) 2) ˆ ˆ ˆ k = nx x + ny y + n z z , where nx, ny, and nz are integers. The volume in kspace Lx Ly Lz
12
12
Chap_5.nb
w1 w2 g time to watch motion release system
10 14. 0.6 20
WignerWeisskopf theory of spontaneous emission
References: M. O. Scully and M. S. Zubairy, Quantum optics, Cambridge University Press 1997, Sect. 6.3. Y. Yamamoto and A. Imamoglu, Mesoscopic quantum optics, Wiley and sons, 1999, Sect. 6.6.
An twolevel atom in a single mode vacuum will, as we saw above, oscillate back and forth between the two states at the vacuum Rabi frequency. This clearly does not happen for an atom in free space: the atom will simply decay through a process we call spontaneous emission. To explain spontaneous emission, we need to take into account not one, but a continuum of optical modes. The emitter couples to all of these simultaneously, leading to a dephasing that results in an exponential decay of the excited state probability.
Pe HtL, wê2p=1 GHz 1.0
0.8
0.6
0.4
0.2
1
2
3
4
5
tHnsL
In the interaction picture, the Hamiltonian is
Chap_5.nb
13
` ` HI = — ‚ Agk * HrA L s+ ak ei Hwnk L t + H.cE
k
(0.81)
where gk * HrA L = gk expHi k ÿ rA L. Assume the initial state is is yHtL\ = e, 0\ and the state evolution is yHtL\ = ce HtL e, 0\ + ‚ cg,k g, 1k ] k The state evolution is described by the Schrödinger equation, ¶∂t yI HtL\ = i — HI yI HtL_ (0.84) (0.85) (0.86) (0.83) (0.82)
° ° ce HtL e, 0\ + ‚ cg,k g, 1k ] k ` ` ` ` = iI ‚ Agk * HrA L s+ ak ei Hwnk L t + gk HrA L s a† k ei Hwnk L t EM yI HtL]
k
= i ‚ Agk * HrA L cg,k e, 0k ] ei Hwnk L t + gk HrA L ce HtL ei Hwnk L t g, 1k ]E
k
Multiplying from the left by Xe, 0 gives ° ce HtL = i ‚ Agk * HrA L cg,k ei Hwnk L t E k Multiplying from the left by Xg, 1k gives ° c = i g Hr L c HtL ei Hwnk L t
g,k k A e
(0.87)
(0.88)
We'd like to eliminate the coefficients cg,k . So let's formally integrate the above equation to obtain cg,k HtL = i gk HrA L ‡ ce Ht 'L ei Hwnk L t' dt'
0 t
(0.89)
Then we have the following differential equation for ce : ° ce HtL =  ‚ gk * HrA L 2 ‡ ce Ht 'L ei Hwnk L Htt'L dt' k
0 t
(0.90)
Now make the following approximations:  infinite space ö closely spaced modes in frequency ö replace sum by integral ‚k ö ‡ DHkL d3 k where DHkL is the density of electromagnetic states. In a homogenous medium, DHkL = 2 polarizations volume in k  space per mode V H2 pL3
3 ‡ d k=2
(0.91)
=
¶
V H2 pL3
, so
2p 0 p
(0.92)
‚k ö 2
V H2 pL3
2 ‡ k dk ‡ 0
0
df ‡ sin q dq
(0.93)
where V is the quantization volume.
14
Chap_5.nb
° ce HtL =  ‚ gk * HrA L 2 ‡ ce Ht 'L ei Hwnk L Htt'L dt' k
0
t
(0.94) (0.95)
ö2
V H2 pL3
3 * 2 i Hwnk L Htt'L dt' ‡ d k gk HrA L ‡ ce Ht 'L e 0
t
The vacuum Rabi frequency term is gk * HrA L where yHrA L = 1, g0,k =  meg — — nk 2 Vmode,k e0 eM,k (0.97)
2
= Hg0,k L2 cos2 x
(0.96)
Let's assume that the emitter is in a uniform medium, so Vmode,k and eM,k = n2 are the same for all modes. Then, g0,k =  meg —n — nk 2 V e0 (0.98)
Assume that the dipole is in the zdirection, so that x=q. Then the integral becomes ° ce HtL = 2
2p
V H2 pL3
2 ‡ k dk ‡ 0
¶
2p
0
df ‡ sin q dq Hg0,k L2 cos2 q2 ‡ ce Ht 'L ei Hwnk L Htt'L dt'
0 0
p
t
(0.99) (0.100)
Using ‡
0
p 2 df ‡ sin q cos2 q2 dq = 2 p ÿ , we have 3 0 2 ¶
V ° ce HtL = ‡ 3 p2 0
 meg —n
— nk 2 V e0
k2 dk ‡ ce Ht 'L ei Hwnk L Htt'L dt'
0
t
(0.101)
¶ meg 2 t nk ° HtL =  1 ce k2 dk ‡ ce Ht 'L ei Hwnk L Htt'L dt' ‡ 0 6 p2 0 — n2 2 e0
(0.102)
Replace k =
n nk c
, where c is the speed of light in vacuum, so (0.103)
¶ n3 t meg 2 1 ° HtL =  1 ce nk 3 dnk ‡ ce Ht 'L ei Hwnk L Htt'L dt' ‡ 0 6 p2 — n2 2 e0 0 c3
=
meg 2 n 12 p2 — e0 c3
3 i Hwnk L Htt'L dt' ‡ nk dnk ‡ ce Ht 'L e 0 0
¶
t
(0.104)
Consider the time integral
i Hwnk L Htt'L dt'. ‡ cg,k Ht 'L e 0 t
Chap_5.nb
15
The exponent varies rapidly so that the integral averages to zero. Only when w is near nk will the integral be non  zero. Therefore we assume that in the integral, we can approximate nk ~ w. For the same reason, we can also replace the lower integration limit by ¶, so ° ce HtL º meg 2 n 12 p — e0 c
2 3
w3 ‡ dt' ce Ht 'L ‡
0
t
¶
¶
„ nk ei Hwnk L Htt'L
(0.105)
t meg 2 n w3 ° ce HtL º ‡ dt' ce Ht 'L H2 p dHt  t 'LL 12 p2 — e0 c3 0 2 3 ° HtL º  meg n w c HtL ce e 6 p — e0 c3
(0.106)
(0.107)
This has the solution ce HtL = exp or ce HtL = exp where G = n G0 , with G0 = meg 2 w3 3 p — e0 c3
2
meg 2 n w3 6 p — e0 c3 G 2
t
(0.108)
t ,
(0.109) (0.110) (0.111)
=
1 4 p e0
4 w3 meg 2 3 — c3
(0.112)
The excited state population therefore decays as ce HtL = expHG tL (0.113) with a spontaneous emission lifetime t = 1 ê G. Note that the spontaneous emission rate is proportional to the refractive index n.
16
Chap_5.nb
7.98953 µ 108
meg lHnmL nHrefr. indexL 550 2.5
Spontaneous emission rate in a nonhomogeneous medium The density of states in kspace is given by DHkL = 2 polarizations volume in k  space per mode = 2V H2 pL3 (0.114)
where the factor of 2 arises because each plane wave can have two polarization directions. In free space, we have the dispersion k2 = kx 2 + ky 2 + kz 2 = n2 ë c2 . Thus, we can solve for the energy density of states: DHnL dn = 4 p k2 dk DHkL = 4 p k2 dk DHnL = 4 p k2 V p2 2 V dk H2 pL3 dn e c (0.118) = k2 V 1 p2 c 2V H2 pL3 (0.115)
(0.116)
=
In2 ë c2 M n2 V p c
2 3
(0.117)
DHnL =
Chap_5.nb
17
Therefore we can rewrite the spontaneous emission rate (noting n~w) G0 = where meg — —w 2 e0 V 2p 3 g0
2
DHwL ,
(0.119) (0.120) (0.121)
g0 =
The expression for G0 agrees with Fermi's golden rule, which gives the transition probability per unit time from state i\ to state f \ : TiØ f = 2p — X f Hint i\ 2 DHwL, (0.122)
where X f Hint i\ is the matrix element on the interaction term Hint . Our expression for G0 differs from Fermi's Golden Rule by a factor of 3 because we assumed that the atom couples only to modes that are polarized in the direction of the dipole. In other words, it comes from the average of the term cos x. Alternatively, we could write our result as G0 = 2 p Y gHrA L 2 ] DHwL (0.123) which now has the same form as Fermi's Golden Rule. It is possible to change D (w) by modifying the electromagnetic environment of the emitter. For instance, we saw before that in a photonic crystal, certain modes within the bandgap are not allowed. In this case, D (w) would be lowered compared to free space. This has recently been demonstrated in 3D and 2D photonic crystals. In the latter, the emission rate of a quantum dot was shown to be reduced by 510 times, a phenomenon that can be used to direct emission only into desired modes and thus produce more efficient light emitters or solar cells. Note: if we solved for the coefficient ce HtL exactly, we would actually find an expression of the form (0.124) t + i Dw t , 2 where the frequency shift Dw is the Lamb shift which is due to a vacuum field induced AC Stark shift, as we discussed before. ce HtL = exp A single emitter in a strong single mode field G
1
Mollow's Triplet Consider the eigenenergies for the singleemitter  single mode system and for the sake of simplicity assume again that d = 0. We have the following energy for the n + 1 excitation manifold: EHn + 1L = — w Hn + 1L ± — g DE = EHn + 1L± ± EHnL± n+1 (0.129)
and so the energy difference when a photon is lost by the system is given by the four possible transitions (0.130)
18
Chap_5.nb
n+1 n transition
= I— w Hn + 1L ± — g = —w±— g
n + 1 M ± I— w n ± — g n M Mollow’s triplet n MM n, n=
(0.131) (0.132) (0.133)
n + 1 M ± I±— g
DE = — w + — g 9 n + 1 n+1 
n ,  n+1 + n, n+1 +
Thus, four transitions are possible. This is illustrated in the figure below for the general case of a transition from n + 1ön:
g, n 1 e, n
g, n e, n 1
n
For example, let's consider the following transitions: † manifold 1ö 0: DEH1 Ø 0L=— w ± — g † manifold 2ö 1: DEH2 Ø 1L=— w ±— g 9 2 ,  2 + 1, 2  1, 2 + 1= † manifold N ö N  1: DEHN Ø N  1L=— w ± — g 9 N  N  1 ,  N + N  1 , N  N  1 , N + N  1 = If N is large, we can approximate N º N  1 , and so N , 0, 0, 2 N =
† manifold for large N ö N  1: DEHN Ø N  1L º — w ± — g 92
The two central transitions therefore become degenerate as N becomes large. This degeneracy causes one central peak which is twice as large asTΩ side peaks. the
2.0
2g
1.5
n
1.0
12
0.5
600
400
200
200
400
600
Ω2ΠGHz
Suppose now that the atom is in free space and is illuminated by a strong coherent field a\. If Xn\ = a 2 p 1, then the variance of a around the mean photon number a 2 is relatively small. Then we can approximate n º Xn\ = a 2 , and so we should obtain a threepeak spectrum with the satellite peaks separated by 2 g N º 2g a .
Chap_5.nb
19
Suppose now that the atom is in free space and is illuminated by a strong coherent field a\. If Xn\ = a 2 p 1, then the variance of a around the mean photon number a 2 is relatively small. Then we can approximate n º Xn\ = a 2 , and so we should obtain a threepeak spectrum with the satellite peaks separated by 2 g N º 2g a .
N kê2pHGHzL gê2pHGHzL
25 21. 21.401
Interaction of a single atom with a singlemode coherent state field Consider a 2level atom interacting with a coherent state, giving the joint state
¶ ¶
yHtL\ = ‚ ce,n HtL PHnL H e, n\ + ‚ cg,n+1 HtL PHnL H g, n + 1\
n=0 n=0
(0.134)
where the first term corresponds to the excited atom and the coherent field, and the second atom corresponds to the ground state atom and the coherent field + 1 photon. We know the photon statistics for the coherent state, so
¶
yHtL\ = ‚ ce,n HtL e a
n=0
2
ê2
an n!
¶
H e, n\ + ‚ cg,n+1 HtL e a
n=0
2
ê2
an n!
H g, n + 1\
(0.135)
Suppose that the atom begins in the excited state, so ce,n = 1, cg,n+1 = 0 at t = 0. We know from above that the probabilities evolve as cg,n+1 HtL = Bcos ce,n = 2ig Wn t 2 n+1 Wn iD Wn sin Wn t 2 F ei D tê2 (0.136)
sin
Wn t 2
ei D tê2
20
Chap_5.nb
where Wn 2 = D2 + 4 g2 Hn + 1L. But for now, let's suppose that D=0, so cg,n+1 HtL = cosIg ce,n =  i sinIg
¶
(0.137)
n+1 tM n + 1 tM
¶ 2
(0.138)
The probability of finding the atom in the excited state is Pe HtL = ‚ Xe, m yHtL\
m=0
= e a ‚
n=0
2
a
2n
n!
cos2 Ig
n + 1 tM
(0.139)
We see that each photon number component in the sum oscillates with a different Rabi frequency. After some time, these components will interfere and cancel. How long will this take? Well, the interference will occur first for the terms that are furthest separated in frequency. Let's suppose now that the coherent state has a relatively high photon number n p 1 so that we can approximat the photon number as a 2 ± a , noting again that the standard deviation of the Poissondistributed photon number is s = Xn\ = that the phase difference Df ~ 1 = tc Ig = tc g n + 1 + Dn  g a
2
. So we can then approximate the collapse time tc from the condition
n + 1  Dn M Dn n+1 Dn n+1 (0.140)
Hn + 1L
1+

1
= tc g = tc g
Hn + 1L 1 + Dn n+1
1 Dn 2 n+1
 1
1 Dn 2 n+1
º tc g ÿ 1 Df ~ 1 º tc g Thus, the collapse time is tc º g1 Note that this is independent of the field's photon (assuming it is large enough for our approximations)! For offresonant driving, this time is longer because we need to replace g Wn = D2 + 4 g2 Hn + 1L n + 1 by (0.143) (0.142) (0.141)
Then we derive Df ~ 1 = tc K D2 + 4 g2 Hn + Dn + 1L D2 + 4 g2 Hn  Dn + 1L O (0.144)
Chap_5.nb
21
ötc
g 1+
D2 4 a 2 g2
Thus, the collapse time is now approximately tc º g The Pe HtL = a
2 1
1+ is
D2 4 a 2 g2 from the following plot, which shows the
(0.145) probability
collapse
2
apparent cos2 Ig
a 2n e a ⁄¶ n! n=0
n + 1 tM (red) as well as terms cos2 Ig
n + 1 tM (blue) from nDn to n+Dn, for
= Xn\ =20.
Pe HtL, wê2p=1 GHz 1.0
0.8
0.6
0.4
0.2
1 Pe HtL (red) and the terms cos2 Jg
2 n + 1 tN (blue), for a
2
3 = Xn\ =20.
4
5
tHnsL
Now let's look at longer times, as shown in the plot below.
Pe HtL, wê2p=1 GHz 1.0
0.8
0.6
0.4
10 0.0
20
30
40
50
tHnsL
The oscillations show revivals! These occur at times tr when the photon number states rephase. This happens (0.144) when
22
Chap_5.nb
tr K2
D2 + 4 g2 Hn + Dn + 1L  2
D2 + 4 g2 Hn  DnL O º 2 p
(0.146)
where the factors of 2 come from the fact that the amplitudes cg,n , etc., are squared inside the sum. The first term is for the state n\ and the second term is for n  1\. This gives tr º 2 p g
1
n
1+
D2 4 n g2
(0.147)
Now we see that different photon number states do take somewhat different times to rephase. However, if
Dn n
` 1, then the variation in n is small.
The revival of the excited state is a purely quantum mechanical phenomenon since it results from the discrete number of excitations of the modes of the EM field.
¶
Pe HtL = e
a2
‚
m=0
a n!
2
cos2 Ig
n + 1 tM
(0.148)