PHY 313: Quantum Mechanics 2 Lecture 1

Anil Shaji
School of Physics, IISER Thiruvananthapuram (Dated: August 4, 2011)

Quantum mechanics happens in Hilbert space; however, quantum particles - of which quantum mechanics provides a description - live in ordinary space. To paraphrase Asher Peres, there are no Hilbert spaces in the laboratory. In this course, we will learn a whole lot more about connecting the quantum kinematics and dynamics that happens in Hilbert space with quantities that one can measure in the lab. You already had quite a bit of exposure to quantum dynamics per say. How to solve the Schrodinger equation or the Heisenberg equation in a variety of situations including, the innite square well, nite square well, harmoic oscillator, Hydrogen atom and so on. We have made some connections with the measurable quantities including nding the spectra, transmission and reection coecients and so on. We are now going to go on a slightly dierent path with the understanding that to a great extent we solved the problems mentioned above mostly in the spirit of the drunk who looked for a lost coin under the lamp. These problems were picked mainly because they were simple enough to be solvable. This exhausts only a tiny fraction of the interesting problems one needs to solve. In most cases where one wants to nd the quantum dynamics of an experimental system, the equations of motion are not solvable. So we go looking for approximations or approximate solutions. This leads us to perturbation theory in quantum mechanics. As an aside, I should point out that there is a rather very well developed literature on perturbation theory in classical physics except that you dont normally bother learning it. This is because when it comes to so called real world problems in classical mechanics rather than going for sophisticated analytical and perturbative methods, one goes for numerical solutions in a computer. Unfortunately this is a lost cause when it comes to dealing with analogous real world quantum systems because the nature of the problem makes it uncomputable on computers working according to the laws of classical physics. This problem has given rise to the entire eld of quantum information processing and quantum computation but that is a story that we are not going to go into in this course. Let us start by building a heuristic picture of what one is trying to do in perturbation theory. Let us start with a normal vector a along the x-axis on a two dimensional plane. Now suppose some kind of disturbance rotates the vector through a small angle towards the y-axis. Now we will write the vector as a = cos() + sin(). x y What we are trying to do in perturbation theory is very much analogous to this. What we are after are eigenstates of some Hamiltonian. We might know the eigenstates of a very similar or close

2 by hamiltonian as
(0) H0 |n(0) = En |n(0) .

We are required to nd the the eigenkets and eigenvalues or else good approximations of the eigenkets and eigenvalues of the full Hamiltonian problem, (H0 + V )|n = En |n . Here V is known as the perturbation, usually taken to be small. What we are attempting here is to try and expand the vectors |n in terms of the basis furnished by the set of known vectors |n(0) . In fact here we try something more than just expressing the unknown eigenkets in terms of the know ones, we also attempt to directly try and write down the eigenvalues of the full Hamiltonian in terms of the eigenvalues of H0 . Coming back to our example of writing a slightly rotated vector in terms of the basis given by the unit vectors along the x and y directions it is clear that this is a lost cause if the rotation takes the original vector right out of the x y plane. The meaning of the word perturbation is really that it does not take the system out of the Hilbert space spanned by the eigenkets of the unperturbed Hamiltonian. It is in this sense that the perturbation is stated to be small. Usually one chooses to write the problem to be solved as (H0 + V )|n = En |n . The parameter is introduced to keep track of how many times the perturbation enters the picture. It is a real parameter and we can get back the original problem by setting 1. Physically one (0) expects a smooth transition from |n0 to |n and En to En as goes from 0 to 1. Mathematically the method rests on the expansion of the energy eigenvalues and energy eigenkets in powers of . This means that implicitly we assume the analyticity of the energy eigenvalues and eigenkets in the complex plane. In plain terms, we do not want the system to do something dramatically dierent - or for than matter jump out of the original Hilbert space - when is changed continuously within some reasonable range. Operationally, if the perturbative approach is to be useful, keeping only two or three terms in the expansion of the energy eigenvalues and eigenkets should give a very good approximation to the true solution of the full problem.

I.

TIME INDEPENDENT PERTRUBATION THEORY: NON-DEGERATE CASE

We start with the simplest case where the perturbation V is time independent. We also assume that the eigenvalues of the unperturbed Hamiltonian are not degenerate. The problem with the degenerate case is that then rather than getting eigenstates or eigenkets we get eigenspaces. Any vector in such a degenerate eigenspace satises the eigenvalue equation corresponding to the degenerate eigenvalues. However when we are trying to expand the eigenstate of the full Hamiltonian in terms of the eigenstates of the unperturbed one, we have to make a choice and pick as many orthogonal vectors from each degenerate eigenspace as its dimension and use them as the basis for

3 our expansion. This picking has to be done wisely depending on the nature of the perturbation in order to get a simple perturbative expansion. We will study how this choice has to be made in the next step and for the moment make the simplifying assumtion that the eigenvalues of H0 are non-degenerate. Before doing the formal development let us do a simple example. Consider a two state system with the Hamiltonian H = H0 + V = E1 |1 1| + E2 |2 2| + V12 |1 2| + V21 |2 1|. Let us further assume that the matrix elements of the perturbation V12 and V21 are real so that V12 = V21 = V . So we have H = H0 + V = E1 |1 1| + E2 |2 2| + V |1 2| + V |2 1|. We can write the Hamiltonian in matrix form as H= The eigenvalues of this matrix are E1 = and E1 = If we have
0 0 |V | |E1 E2 |, (0) (0) (0) (0)

E1

(0)

V
(0)

V E2

E1 + E2 2

(0)

(0)

(E1 E2 )2 + 2 V 2 . 4

(0)

(0)

E1 + E2 2

(0)

(0)

(E1 E2 )2 + 2 V 2 . 4

(0)

(0)

we can expand the term under the square root in powers of and we get
0 E1 = E1 +

2 V 2 (E1 E2 ) 2 V 2 (E2 E1 )
(0) (0) (0) (0)

+ ...

and
0 E2 = E2 +

+ ...

These sort of expansions are our aim in the general case without, of course, having to diagonalize the entire Hamiltonian in the rst place as we have done here. Going back to the full expressions for E1 and E2 that we got, let us treat E1 and E2 as functions of in the complex plane. Because of the square root term, we know that there is branch cut wherever the term under the square root is negative. The branch points are at |V | = i E1 E2 . 2
(0) (0)

4 So the condition for the convergence of the series expansion for the = 1 case is |V | |E1 E2 | . 2
(0) (0)

If this condition is not satised then the perturbative expansion is meaningless. Keeping this example as a simple reference point now we can go ahead with the formal development of nondegerate perturbation theory.