Skip to content

Perturbation Theory up to Second Order

CSIR NET chemistry: Rayleigh-Schrödinger perturbation to second order in energy, first-order wavefunction, box and oscillator examples, degeneracy.

  • 20 min read
  • Updated
On this page · 18 sections
  1. Prerequisites
  2. The Hook: A Box with a Bump
  3. Setting Up the Problem
  4. First Order
  5. Second Order
  6. Exact Check: The Two-Level System
  7. Toolkit: Matrix Elements You Will Need
  8. Worked Examples: Particle in a Box
  9. Worked Examples: Harmonic Oscillator
  10. Degenerate Levels: First Order
  11. Perturbation vs Variation
  12. Memory Tricks
  13. Don’t Confuse
  14. Common Mistakes
  15. Practice
  16. Quick Revision Table
  17. Key Takeaways
  18. What’s Next

Prerequisites

The Hook: A Box with a Bump

Setting Up the Problem

Split the Hamiltonian into a solvable part and a small remainder:

$$\hat H = \hat H^{(0)} + \lambda\hat H'$$

The unperturbed problem \( \hat H^{(0)}\psi_n^{(0)} = E_n^{(0)}\psi_n^{(0)} \) is solved exactly, with orthonormal \( \psi_n^{(0)} \). The bookkeeping parameter \( \lambda \) (set to 1 at the end) tracks the order. Expand the exact eigenvalue and eigenfunction of level \( n \):

$$E_n = E_n^{(0)} + \lambda E_n^{(1)} + \lambda^2 E_n^{(2)} + \ldots, \qquad \psi_n = \psi_n^{(0)} + \lambda\psi_n^{(1)} + \lambda^2\psi_n^{(2)} + \ldots$$

This is Rayleigh–Schrödinger perturbation theory. In this section the level \( n \) is assumed non-degenerate.

Substituting into \( \hat H\psi_n = E_n\psi_n \) and collecting powers of \( \lambda \):

OrderEquation
\( \lambda^0 \)\( (\hat H^{(0)} - E_n^{(0)})\psi_n^{(0)} = 0 \) (the unperturbed problem)
\( \lambda^1 \)\( (\hat H^{(0)} - E_n^{(0)})\psi_n^{(1)} = (E_n^{(1)} - \hat H')\psi_n^{(0)} \)
\( \lambda^2 \)\( (\hat H^{(0)} - E_n^{(0)})\psi_n^{(2)} = (E_n^{(1)} - \hat H')\psi_n^{(1)} + E_n^{(2)}\psi_n^{(0)} \)

Choose intermediate normalisation, \( \langle\psi_n^{(0)}|\psi_n^{(k)}\rangle = 0 \) for \( k \geq 1 \): corrections contain no component along the unperturbed state.

First Order

Energy. Multiply the \( \lambda^1 \) equation on the left by \( \psi_n^{(0)*} \) and integrate. Because \( \hat H^{(0)} \) is Hermitian, the left side is \( \langle\psi_n^{(0)}|(\hat H^{(0)} - E_n^{(0)})|\psi_n^{(1)}\rangle = 0 \). So

First-order energy
$$\boxed{E_n^{(1)} = \langle\psi_n^{(0)}|\hat H'|\psi_n^{(0)}\rangle = H'_{nn}}$$

The first-order correction is the average of the perturbation over the unperturbed state.

Wavefunction. Expand \( \psi_n^{(1)} = \sum_{m\neq n} a_m\psi_m^{(0)} \) and project the \( \lambda^1 \) equation onto \( \psi_m^{(0)} \) with \( m \neq n \): \( (E_m^{(0)} - E_n^{(0)})a_m = -H'_{mn} \). Hence

First-order wavefunction
$$\psi_n^{(1)} = \sum_{m\neq n}\frac{H'_{mn}}{E_n^{(0)} - E_m^{(0)}}\,\psi_m^{(0)}, \qquad H'_{mn} = \langle\psi_m^{(0)}|\hat H'|\psi_n^{(0)}\rangle$$

The perturbation mixes in other states, each in proportion to its coupling and inversely to its energy gap.

Second Order

Project the \( \lambda^2 \) equation onto \( \psi_n^{(0)} \). The left side vanishes again; with intermediate normalisation \( \langle\psi_n^{(0)}|\psi_n^{(1)}\rangle = 0 \), so the \( E_n^{(1)} \) term drops out too, leaving \( E_n^{(2)} = \langle\psi_n^{(0)}|\hat H'|\psi_n^{(1)}\rangle \). Inserting \( \psi_n^{(1)} \):

Second-order energy
$$\boxed{E_n^{(2)} = \sum_{m\neq n}\frac{|H'_{mn}|^2}{E_n^{(0)} - E_m^{(0)}}}$$

So the energy through second order is

$$E_n \approx E_n^{(0)} + H'_{nn} + \sum_{m\neq n}\frac{|H'_{mn}|^2}{E_n^{(0)} - E_m^{(0)}}$$

Note what each order needs: \( E^{(1)} \) needs only \( \psi^{(0)} \); \( E^{(2)} \) needs \( \psi^{(1)} \) (equivalently, all the off-diagonal matrix elements).

What the second-order formula tells you without calculation

  1. The ground-state \( E^{(2)} \) is never positive. For the lowest level every denominator \( E_0^{(0)} - E_m^{(0)} \) is negative and every numerator is \( \geq 0 \).
  2. Level repulsion. A state above level \( n \) contributes a negative term (pushes \( n \) down); a state below contributes a positive term (pushes \( n \) up). Coupled levels move apart, exactly as in the linear variation method.
  3. Near states dominate. The \( 1/\Delta E \) weighting means the nearest coupled states usually give most of \( E^{(2)} \).
  4. Symmetry kills terms. If \( H'_{mn} = 0 \) by parity or angular momentum, that state contributes nothing. If \( H'_{nn} = 0 \) by symmetry (an odd perturbation in a state of definite parity), the first non-zero correction is second order.
graph TD
    A["Perturbation H prime"] --> B{"Is H prime nn zero
by symmetry?"} B -->|No| C["First-order shift
E1 = H prime nn"] B -->|Yes: odd H prime and a parity state| D["First non-zero correction
is second order"] C --> E["Second order: sum of
coupling squared over gap"] D --> E E --> F{"Is coupling much smaller
than every gap?"} F -->|Yes| G["Series useful"] F -->|No: degenerate or near-degenerate| H["Use degenerate PT or
diagonalise exactly"]

Validity

The expansion is useful only if

$$\left|H'_{mn}\right| \ll \left|E_n^{(0)} - E_m^{(0)}\right| \quad \text{for every } m \text{ that couples to } n$$

When a coupled state is degenerate with \( n \), the formula divides by zero. That is not a failure of the physics but a sign that the unperturbed functions were the wrong starting point (see Degenerate Levels below).

Exact Check: The Two-Level System

Take two unperturbed levels \( E_1^{(0)} \) and \( E_2^{(0)} = E_1^{(0)} + \Delta \), with no diagonal perturbation and an off-diagonal coupling \( V \). The exact energies come from a 2 × 2 determinant:

$$E_\pm = E_1^{(0)} + \frac{\Delta}{2} \pm \sqrt{\frac{\Delta^2}{4} + V^2}$$

For \( |V| \ll \Delta \), expand the square root: \( \sqrt{\Delta^2/4 + V^2} \approx \frac{\Delta}{2} + \frac{V^2}{\Delta} \). The lower level becomes \( E_1^{(0)} - V^2/\Delta \): exactly the second-order formula, since \( E_1^{(1)} = 0 \) and the only term is \( |V|^2/(E_1^{(0)} - E_2^{(0)}) \).

Toolkit: Matrix Elements You Will Need

Particle in a box (\( 0 \leq x \leq L \)): \( \psi_n = \sqrt{2/L}\sin(n\pi x/L) \), \( E_n^{(0)} = n^2h^2/8mL^2 \).

  • \( \langle x\rangle_n = L/2 \) for every \( n \) (symmetric probability density).
  • Probability in either half of the box is exactly \( \frac{1}{2} \) for every \( n \).
  • \( |\psi_n(L/2)|^2 = \frac{2}{L} \) for odd \( n \) and \( 0 \) for even \( n \) (node at the centre).

Harmonic oscillator: \( E_v^{(0)} = (v + \frac{1}{2})\hbar\omega \), and with the ladder operators

$$\hat x = \sqrt{\frac{\hbar}{2m\omega}}\left(\hat a + \hat a^\dagger\right), \qquad \langle v+1|\hat x|v\rangle = \sqrt{\frac{\hbar}{2m\omega}}\sqrt{v + 1}$$
  • \( \hat x \) connects \( v \) to \( v \pm 1 \) only; \( \hat x^2 \) to \( v, v \pm 2 \); \( \hat x^3 \) to \( v \pm 1, v \pm 3 \); \( \hat x^4 \) to \( v, v \pm 2, v \pm 4 \).
  • Odd powers have zero diagonal elements (parity), so they shift the energy only at second order.
  • \( \langle v|x^2|v\rangle = \frac{\hbar}{m\omega}\left(v + \frac{1}{2}\right) \); \( \langle v|x^4|v\rangle = \left(\frac{\hbar}{2m\omega}\right)^2(6v^2 + 6v + 3) \).

Worked Examples: Particle in a Box

(a) Constant shift everywhere, \( H' = V_0 \). \( E_n^{(1)} = V_0 \); all off-diagonal elements vanish (orthogonality), so \( E^{(2)} = 0 \) and the result is exact. A trivial case, but options often hide it.

(b) Step over half the box, \( H' = V_0 \) for \( 0 < x < L/2 \), zero elsewhere. \( E_n^{(1)} = V_0 \times (\text{probability in the left half}) = V_0/2 \) for every \( n \). The second-order term is non-zero and negative for the ground state.

(c) Linear slope, \( H' = \varepsilon x \). \( E_n^{(1)} = \varepsilon\langle x\rangle = \varepsilon L/2 \) for every \( n \): the first-order shift equals the potential at the centre.

(d) Delta function at the centre, \( H' = \gamma\,\delta(x - L/2) \). \( E_n^{(1)} = \gamma|\psi_n(L/2)|^2 = 2\gamma/L \) for odd \( n \) and \( 0 \) for even \( n \). Levels with a node at the centre do not feel the bump: this is the box-with-a-bump answer from the hook.

(e) Half-sine bump, \( H' = V_0\sin(\pi x/L) \). For \( n = 1 \):

$$E_1^{(1)} = \frac{2V_0}{L}\int_0^L\sin^3\frac{\pi x}{L}\,dx = \frac{2V_0}{L}\cdot\frac{4L}{3\pi} = \frac{8V_0}{3\pi} \approx 0.849\,V_0$$

For \( n = 2 \) the same integral gives \( \frac{32V_0}{15\pi} \approx 0.679\,V_0 \): the \( n = 2 \) density has a node exactly where the bump peaks.

Worked Examples: Harmonic Oscillator

These four perturbations cover most oscillator questions. All shifts below were checked against exact diagonalisation.

Perturbation \( H' \)\( E_v^{(1)} \)\( E_v^{(2)} \)Exact?
\( bx \) (linear)0\( -\dfrac{b^2}{2m\omega^2} \) for every \( v \)exact: complete the square, the parabola shifts sideways and down
\( \frac{1}{2}\delta k\,x^2 \) (stiffer spring)\( \dfrac{\delta k}{2k}\left(v + \frac{1}{2}\right)\hbar\omega \)\( -\dfrac{1}{8}\left(\dfrac{\delta k}{k}\right)^2\left(v + \frac{1}{2}\right)\hbar\omega \)first two terms of \( \sqrt{1 + \delta k/k} \)
\( cx^3 \) (cubic)0 (odd)\( -\dfrac{c^2}{\hbar\omega}\left(\dfrac{\hbar}{2m\omega}\right)^3(30v^2 + 30v + 11) \)approximate
\( dx^4 \) (quartic)\( d\left(\dfrac{\hbar}{2m\omega}\right)^2(6v^2 + 6v + 3) \)negative, order \( d^2 \)approximate

Degenerate Levels: First Order

If \( \psi_n^{(0)} \) belongs to a set of \( g \) degenerate functions \( \phi_1, \ldots, \phi_g \), the second-order formula fails for couplings inside the set. The cure is to choose the right zeroth-order combinations first: diagonalise \( \hat H' \) inside the degenerate set.

$$\det\left|H'_{ij} - E^{(1)}\delta_{ij}\right| = 0, \qquad H'_{ij} = \langle\phi_i|\hat H'|\phi_j\rangle, \quad i, j = 1, \ldots, g$$

The \( g \) roots are the first-order corrections; the perturbation may lift the degeneracy fully, partly or not at all. The eigenvectors are the “correct” zeroth-order functions, after which non-degenerate formulas apply to couplings outside the set.

For two degenerate functions with \( H'_{11} = H'_{22} = 0 \) and \( H'_{12} = V \), the roots are \( \pm|V| \) and the correct functions are \( (\phi_1 \pm \phi_2)/\sqrt{2} \). The linear Stark effect in the \( n = 2 \) level of hydrogen is exactly this case, worked on the applications page.

Perturbation vs Variation

FeaturePerturbation theoryVariation method
Needssolvable \( \hat H^{(0)} \) close to \( \hat H \)a sensible trial function
Givessystematic series \( E^{(0)} + E^{(1)} + E^{(2)} + \ldots \)single number, an upper bound
Bound?only through first order for the ground statealways an upper bound to the ground state
Excited statesyes, level by levelvia orthogonality or higher secular roots
Fails whencoupling comparable to gaps; degeneracytrial function has the wrong shape or symmetry
Chemistry exampleMP2 (second-order Møller–Plesset correction)Hartree–Fock, configuration interaction

Memory Tricks

Don’t Confuse

ThisvsThat
\( E^{(1)} = H'_{nn} \) (diagonal element)\( E^{(2)} \) uses off-diagonal \( H'_{mn} \) only
Denominator \( E_n^{(0)} - E_m^{(0)} \) (state of interest first)\( E_m^{(0)} - E_n^{(0)} \): flips the sign of every term
Non-degenerate PT: one level, sum over othersDegenerate PT: diagonalise \( H' \) inside the degenerate block
\( E^{(0)} + E^{(1)} \): variational upper bound for the ground state\( E^{(0)} + E^{(1)} + E^{(2)} \): no bound guaranteed
Linear \( bx \) on an oscillator: shift \( -b^2/2m\omega^2 \), spacing unchangedQuartic \( dx^4 \): shift grows with \( v \), spacing changes
Box step over half: \( V_0/2 \) for all \( n \)Box central delta: \( 2\gamma/L \) odd \( n \), 0 even \( n \)

Common Mistakes

Practice

All questions below are original practice questions written for this page. They are not from any CSIR NET paper.

Level 1: Foundation

  1. Write the first-order energy correction in words.
  2. What is the sign of the second-order correction to a non-degenerate ground state?
  3. A particle in a box feels \( H' = V_0 \) everywhere inside. What are \( E^{(1)} \) and \( E^{(2)} \)?
Answers
  1. The expectation value of the perturbation over the unperturbed wavefunction, \( \langle\psi_n^{(0)}|\hat H'|\psi_n^{(0)}\rangle \).
  2. Negative or zero. Every denominator \( E_0^{(0)} - E_m^{(0)} \) is negative and every numerator is non-negative.
  3. \( E^{(1)} = V_0 \), \( E^{(2)} = 0 \). A constant has no off-diagonal elements between orthogonal states; the first-order result is exact.

Level 2: Part B Style

Practice Q1. A particle in a one-dimensional box (\( 0 \leq x \leq L \)) is perturbed by \( H' = V_0 \) for \( 0 \leq x \leq L/2 \) and \( H' = 0 \) for \( L/2 < x \leq L \). The first-order corrections to the \( n = 1 \) and \( n = 2 \) levels are:

(a) \( V_0/2 \) and \( V_0/2 \) (b) \( V_0/2 \) and 0 (c) \( V_0 \) and \( V_0/2 \) (d) \( V_0/4 \) and \( V_0/2 \)

Answer: (a)

\( E_n^{(1)} = V_0\int_0^{L/2}\frac{2}{L}\sin^2\frac{n\pi x}{L}dx = \frac{V_0}{2} \) for every \( n \), because each \( |\psi_n|^2 \) is symmetric about the centre and so puts exactly half the probability in each half.

Practice Q2. A harmonic oscillator of mass \( m \) and frequency \( \omega \) is perturbed by \( H' = bx \). Through second order, the ground-state energy is:

(a) \( \frac{1}{2}\hbar\omega \) (b) \( \frac{1}{2}\hbar\omega + \frac{b^2}{2m\omega^2} \) (c) \( \frac{1}{2}\hbar\omega - \frac{b^2}{2m\omega^2} \) (d) \( \frac{1}{2}\hbar\omega - \frac{b^2}{m\omega^2} \)

Answer: (c)

\( E^{(1)} = b\langle x\rangle_0 = 0 \). \( E^{(2)} = |H'_{10}|^2/(-\hbar\omega) = -\frac{b^2\hbar}{2m\omega}\cdot\frac{1}{\hbar\omega} = -\frac{b^2}{2m\omega^2} \). It is also the exact answer (complete the square). (b) has the wrong sign; (d) drops the \( \frac{1}{2} \) from the matrix element.

Practice Q3. A perturbation \( H' = dx^4 \) acts on a harmonic oscillator. The first-order correction to the ground state is:

(a) \( \frac{3d\hbar^2}{4m^2\omega^2} \) (b) \( \frac{3d\hbar^2}{m^2\omega^2} \) (c) \( \frac{d\hbar^2}{4m^2\omega^2} \) (d) 0

Answer: (a)

\( \langle 0|x^4|0\rangle = 3\left(\frac{\hbar}{2m\omega}\right)^2 = \frac{3\hbar^2}{4m^2\omega^2} \). (d) would be right for \( x^3 \) (odd), not \( x^4 \). (c) uses \( \langle x^2\rangle^2 \) without the factor 3.

Practice Q4. A particle in a box of length \( L \) is perturbed by \( H' = \gamma\,\delta(x - L/2) \). Which levels have zero first-order shift?

(a) all odd \( n \) (b) all even \( n \) (c) only \( n = 1 \) (d) none

Answer: (b)

\( E_n^{(1)} = \gamma|\psi_n(L/2)|^2 = \frac{2\gamma}{L}\sin^2\frac{n\pi}{2} \): zero for even \( n \) (node at the centre), \( 2\gamma/L \) for odd \( n \).

Level 3: Part C Style

Practice Q5. Two states of a molecule have unperturbed energies 0 and 2.0 eV. A perturbation has zero diagonal elements and an off-diagonal element of 0.20 eV between them. Which statements are correct?

  1. The second-order shift of the lower state is \( -0.020 \) eV.
  2. The exact shift of the lower state is about \( -0.0198 \) eV.
  3. The upper state moves up by the same amount as the lower state moves down.
  4. If the gap were reduced to 0.40 eV with the same coupling, second order would still be accurate to within 1%.

(a) 1, 2 and 3 only (b) 1 and 2 only (c) 1, 2 and 4 only (d) 1, 2, 3 and 4

Answer: (a) 1, 2 and 3 only
  1. Correct. \( -V^2/\Delta = -0.04/2.0 = -0.020 \) eV.
  2. Correct. \( 1.0 - \sqrt{1.0 + 0.04} \approx -0.0198 \) eV.
  3. Correct. For an isolated pair with zero diagonal perturbation the exact roots are symmetric about the midpoint, and the second-order shifts are \( \mp V^2/\Delta \).
  4. Wrong. With \( \Delta = 0.40 \) eV, \( V/\Delta = 0.5 \). Second order gives \( -0.100 \) eV; exact gives \( 0.20 - \sqrt{0.04 + 0.04} \approx -0.083 \) eV, about 20% off.

Practice Q6. A particle in a box (\( 0 \leq x \leq L \)) is perturbed by \( H' = V_0\sin(\pi x/L) \). The first-order corrections \( E_1^{(1)} \) and \( E_2^{(1)} \) are:

(a) \( \frac{8V_0}{3\pi} \) and \( \frac{32V_0}{15\pi} \) (b) \( \frac{8V_0}{3\pi} \) and \( \frac{8V_0}{3\pi} \) (c) \( \frac{4V_0}{3\pi} \) and \( \frac{16V_0}{15\pi} \) (d) \( \frac{2V_0}{\pi} \) and 0

Answer: (a)

\( E_1^{(1)} = \frac{2V_0}{L}\int_0^L\sin^3\frac{\pi x}{L}dx = \frac{2V_0}{L}\cdot\frac{4L}{3\pi} = \frac{8V_0}{3\pi} \approx 0.849V_0 \).

For \( n = 2 \): \( \frac{2V_0}{L}\int_0^L\sin^2\frac{2\pi x}{L}\sin\frac{\pi x}{L}dx \). With \( u = \pi x/L \), \( \sin^2 2u\sin u = 4\sin^3u\cos^2u \), and \( \int_0^\pi 4\sin^3u\cos^2u\,du = \frac{16}{15} \). So \( E_2^{(1)} = \frac{2V_0}{\pi}\cdot\frac{16}{15} = \frac{32V_0}{15\pi} \approx 0.679V_0 \).

The \( n = 2 \) shift is smaller because its density vanishes at the centre, where the bump is highest. (c) uses \( 1/L \) instead of \( 2/L \) for the normalisation, which halves both values.

Practice Q7. A harmonic oscillator is perturbed by \( H' = cx^3 \) (small \( c \)). Which statements are correct?

  1. \( E_v^{(1)} = 0 \) for every \( v \).
  2. In the ground state, \( x^3 \) couples \( v = 0 \) only to \( v = 1 \) and \( v = 3 \).
  3. \( E_0^{(2)} = -\frac{11c^2\hbar^2}{8m^3\omega^4} \).
  4. The magnitude of \( E_v^{(2)} \) decreases as \( v \) increases.

(a) 1, 2 and 3 only (b) 1 and 4 only (c) 2 and 3 only (d) 1, 2, 3 and 4

Answer: (a) 1, 2 and 3 only
  1. Correct. \( x^3 \) is odd and every \( |\psi_v|^2 \) is even.
  2. Correct. \( (a + a^\dagger)^3 \) acting on \( |0\rangle \) gives only \( |1\rangle \) and \( |3\rangle \). With \( x_0^2 = \hbar/2m\omega \): \( \langle 1|x^3|0\rangle = 3x_0^3 \), \( \langle 3|x^3|0\rangle = \sqrt{6}\,x_0^3 \).
  3. Correct. \( E_0^{(2)} = -\frac{c^2x_0^6}{\hbar\omega}\left(\frac{9}{1} + \frac{6}{3}\right) = -\frac{11c^2x_0^6}{\hbar\omega} = -\frac{11c^2\hbar^2}{8m^3\omega^4} \).
  4. Wrong. \( |E_v^{(2)}| \propto 30v^2 + 30v + 11 \), which grows with \( v \). Higher levels sample larger \( x \), where \( x^3 \) is bigger.

Quick Revision Table

IdeaFact
Split\( \hat H = \hat H^{(0)} + \hat H' \), \( \hat H^{(0)} \) solvable
First-order energy\( E_n^{(1)} = H'_{nn} \)
First-order function\( \psi_n^{(1)} = \sum_{m\neq n}\frac{H'_{mn}}{E_n^{(0)} - E_m^{(0)}}\psi_m^{(0)} \)
Second-order energy\( E_n^{(2)} = \sum_{m\neq n}\frac{\vert H'_{mn}\vert ^2}{E_n^{(0)} - E_m^{(0)}} \)
Ground state\( E_0^{(2)} \leq 0 \); \( E^{(0)} + E^{(1)} \) is a variational upper bound
Validity\( \vert H'_{mn}\vert \ll \vert E_n^{(0)} - E_m^{(0)}\vert \)
Two-level, exact\( \bar E \pm\sqrt{(\Delta/2)^2 + V^2} \); PT: \( \mp V^2/\Delta \)
Box: \( \varepsilon x \) / half step / central delta\( \varepsilon L/2 \) / \( V_0/2 \) / \( 2\gamma/L \) odd, 0 even
Oscillator \( bx \)\( E^{(1)} = 0 \), \( E^{(2)} = -b^2/2m\omega^2 \) (exact)
Oscillator \( dx^4 \)\( E^{(1)} = d(\hbar/2m\omega)^2(6v^2 + 6v + 3) \)
Oscillator \( cx^3 \)\( E^{(1)} = 0 \); \( E^{(2)} = -\frac{c^2}{\hbar\omega}(\frac{\hbar}{2m\omega})^3(30v^2 + 30v + 11) \)
Degenerate levelsolve \( \det\vert H'_{ij} - E^{(1)}\delta_{ij}\vert = 0 \) in the degenerate block

Key Takeaways

What’s Next