LECTURE 2.

Prof.R.Parthasarathy

Time-dependent Perturbation Theory Consider a system with a hamiltonian H = H0 + H with the solvable eigenvalue problem: H0 |un > = En |un >, describing the stationary states in the absense of H . Let H depend on time. Then, transitions between the stationary states of H0 can be eected. We assume that the time dependence is slowly varying nature - adiabatic case. The Schrdinger equation o i h | > = H| >, t
En t

(1)

can be analysed by expanding | > in terms of ei h |un > - eigenfunctions of the unperturbed H0 with exponential time dependence.

|(t) > =

n a (t)|u > ei Eh t , n n n

(2)

where stands for summation over discrete n and integration over continuos n. Note the expansion coecents an (t) depend on time. Substituting in (1),
En t i i { n (an (t) an (t)En )ei h |un >} = h h + n a (t)E ei Eh t |u > n n n n n a (t)H |u > ei Eh t . (3) n n n

Taking the scalar product with |uk > and using < uk |un >= k,n (appropriately for continuous case), we get, ak (t) = where kn = Ek En h (5) 1 an (t) < uk |H |un > eikn t , i n h (4)

Now writing an (t) as an (t) = a(0) (t) + a(1) (t) + 2 a(2) (t) + , n n n 1 (6)

and replacing H (t) by H (t) in (7), we have {ak (t) + ak (t) + 2 ak (t) + } =
(0) (1) (2)

1 < uk |H (t)|un > i n h {a(0) + 2 a(1) + }eikn t . (7) n n

Equating the like powers of , we get a series of expressions: = 0, 1 (1) < uk |H (t)|un > a(0) eikn t , ak = n i n h 1 (2) ak = < uk |H (t)|un > a(1) eikn t , n i n h , In general, ak
(s+1)

ak

(0)

(8)

1 i h
(0)

< uk |H (t)|un > a(s) eikn t . n

(9)

Consider the equation ak = 0. This tells that ak is constant in time. This stands for the initial condition before the perturbation is applied. Let (0) us assume that all except one of the ak are zero, so that the system initially (0) is in one denite state. As an example, ak =< uk |um >= km or (k m), (1) m the initial state index. Then the equation for ak in (10), becomes simply 1 < uk |H (t)|um > eikm t , giving i h 1 t (1) ak (t) = < uk |H (t)|um > eikm t dt . i h ak
(1)

(10)

This is the rst order result. Knowing this one can proceed to the second order and so on. If H (t) is independent of time except for being turned on at one time and o at a later time, then the limits of the integration can be taken as 0 and t. Then, we get ak (t) =
(1)

< uk |H |um > eikm t 1 , h km 2

(11)

and the probability of nding the system in the state |uk > at t is, |ak (t)|2 =
(1)

sin2 (km t/2) 4 | < uk |H |um > |2 . 2 km h2

(12)

2 The quantity sin2 (km t/2)/(km ) can be plotted against km . It is easy to see that the peak occurs when km = 0 and the peak value is t2 /4. The function will vanish whenever km = 2/t, 4/t, etc. This gives the area under the curve will be proportional to t. Hence, the probability will be proprtional to t and therefore we consider the probability per unit time. We consider a particular group of nal states k. The nal state being a continuum may have nite width of energy levels so that in the vicinity of |uk > there could be many stationary states having energies close to Ek . So we sum over them, (1) 2 k |ak (t0 )| and replace the sum by an integral over the number of states which are closely located. Introduce density of nal states as (E) = dN dE so that dN = (E)dE and then the above sum for the transition probability per unit time becomes,

1 t

|ak (t)|2 (Ek ) dEk .

(1)

(13)

The matrix element and the density of nal states will be slowly varying with Ek and so can be taken out of the integral. Then, we have = 1 4| < uk |H |um > |2 (Ek ) t h

sin2 (km t/2) dkm . 2 km

(14)

The limits of the integral above extending from to is justied by observing the behaviour of the integrand which peaks near zero and subsequent peaks are considerably small. Verify this by plotting. The value of the integral is t (Show this). Hence, 2 = 2 (Ek ) | < uk |H |um > |2 , h (15)

which is an important formula, Fermis Golden Rule. Harmonic perturbation

Let H (t) depend on time harmonically and be switched on at t = 0 and o at t = t0 ; t0 > 0. Say, < uk |H (t)|um > = 2 < uk |H |um > sint, > 0. Then, substituting in (12), ak (t0 ) =
(1) t0 2 < uk |H |um > sint eikm t dt , i h 0 < uk |H |um > ei(+km )t0 1 ei(km )t0 1 = , (17) h i( + km ) i(km )

(16)

giving the rst order amplitude at the end of the harmonic perturbation, for (0) the system initially in one denite state ak . When km = , the amplitude (1) aK (t0 ) becomes appreciably large. When the external perturbation has an angular frequency nearly equal to km , the rst term in (19) contributes signicantly. This requirement km = = Ek Em , h Ek = Em h,

(18)

with |um > being the initial state, the nal state has energy depleted by h and so corresponds to emission process. Similarly when = km , the second term in (19) dominates, and now Ek = Em + h, corresponding to absorption process. The transfer or reception occurs signicantly when is nearly equal to the system frequency. Consider a case in which the initial state |um > is a discrete bound state and the nal state |uk > is one of the continuous set of dissociated states. Then Ek > Em . The second term in (19) needs to be considered. The rst order transition probability is then, |ak (t t0 )|
(1) 2

| < uk |H |um > |2 ei(km )t0 1 2 | |, (km ) h2 (km )t0 4| < uk |H |um > |2 = sin2 . 2 2 2 h (km ) =

(19)

The dependence on km can be easily studied. It is like sin2 (xt0 /2)/x2 . When x = km = 2/t0 , 4/t0 , the function vanishes. When x = 4

3/t0 , 5/t0 ,, the function peaks but progressively decreasing at the peak values. It acquires a maximum when x = 0. Transition Probability per unit time The nal state being a continuum may have nite width of energy levels so that in the vicinity of |uk > there could be many stationary states having (1) energies close to Ek . So we sum over them, k |ak (t0 )|2 and replace the sum by an integral over the number of states which are closely located. Introduce density of nal states as (E) = dN so that dN = (E)dE and then the dE above sum becomes |ak (t0 )|2 (E)dE =
(1)

| < uk |H |um > |2 4 sin2 (km )t0 /2 (km )2 h2 (Ek )dEk . (20)

We assume that (Ek ) is a very slowly varying function of Ek and taken as a constant. Next, we denote | < uk |H |um > |2 as averaged over the nal states. Then, the transition probability per unit time is = 1 4 | < uk |H |um > |2 (Ek ) t0 h2 sin2 (km )/2 t0 dEk . (km )2
t0 dEk , 2 h

(21)

By letting x = simplies to

1 (km 2

)t0 and so dx =

the above expression

2 | < uk |H |um > |2 (Ek ) h

sin2 x dx. x2

(22)

We have encountered the function sin earlier and its main peak contributing x2 2 to the integral most. So we replace the above integral by sin2 x whose x value is . Hence, the transition probability per unit time becomes, = 2 (Ek ) | < uk |H |um > |2 , h (23)

which is the celebrated Fermis Golden Rule of Quantum Mechanics. 5

Non-harmonic case and second order Let us consider H (t) = e t H , where is vanishingly small. Then (12) gives ak (t) =
(1)

(24)

< uk |H |um > e t+ikm t , h km i

(25)

and so the probability of transition to |uk > Pk (t) =

e2 t | < uk |H |um > |2 = 2 km + h2

(26)

The rate of change of the probability (probability per unit time) is then, k = 2
2 (km + 2)

e2 t

| < uk |H |um > |2 . h2

(27)

Now taking the limit 0, using the formula (x) = limit 0

1 x2 +

(28)

k =

2 | < uk |H |um > |2 (Ek Em ). h

(29)

Now introducing the density of nal states and paralleling the steps in (22) to (24), we have k = 2 | < uk |H |um > |2 (Ek ). h (30)

In (31) and (32), we have matrix elements connecting states of equal energy. It may happen that the matrix element may vanish. Then the rst 6

order transition probability is zero. So we examine the second order case now. From the master equation (11), we have (2) (31) i ak (t) = n < uk |H (t)|un > a(1) (t) eikn t , h n where for a(1) (t) we will substitute (22) and use H (t) = e t H . Then, we get n
(2) i 2 ak (t) = h n

< uk |H |un >< un |H |um >

e2 t+ikm t , nm i

(32)

which upon integrating over t from to t, gives,

e2 t+ikm t < uk |H |un >< un |H |um > . (nm i )(ikm + 2 ) (33)

The transition probability is obtained by modulus squaring the above. That is, Pk (t) = |ak (t)|2 =
(2) (2)

1 e4 t | < uk |H |un >< un |H |um > |2 2 . 2 n (nm + 2 )(km + 4 2 ) h4 (34)

Hence, the transition probability (in second order) per unit time is (upon dierentiating above with respect to t), 1 4 e4 t n | < uk |H |un >< un |H |um > |2 2 (35) . 2 km + 4 2 (nm + 2 ) h4 Now taking the limit 0, we have 2 < uk |H |un >< un |H |um > 2 (2) km = | n | (Ek Em ). (36) h hnm Once again we introduce the density of nal states to arrive at 2 < uk |H |un >< un |H |um > 2 (2) km = | n | (Ek ). (37) h hnm The above result means that if the system initially in |um > cannot make a transition to the nal state |uk > in the rst order by H due to the vanishing of the matrix element < uk |H |um >, connecting states of same energy for some reasons, then in the second order the system can make a transition to intermediate state |un > and then to |uk >. Such intermediate transitions do not conserve energy and are not actually seen. These are called virtual transitions. k
(2)