Recurrence Relations in The Rayleigh-Ritz Method
Recurrence Relations in The Rayleigh-Ritz Method
Recurrence Relations in The Rayleigh-Ritz Method
138, 2000
Calvin Stubbins
Department of Physics & Astronomy, Franklin & Marshall College
Lancaster PA, USA
§1. Introduction
As can be seen in Eq. (1.5), the Rayleigh-Ritz approach assumes that the linear
coefficients are independent. However, it is easy to show that this is not the case.
If we substitute Eq. (1.1) into the Schrödinger equation HΨ = EΨ, a recurrence
relation for the coefficients is formed. In particular we get
∞
∞
an Hmn = E an Fmn . (2.1)
n=1 n=1
Here E is the exact energy, which is an unknown. For each value of m we obtain
one equation which may be used to solve for one of the a in terms of the other
coefficients and the energy. In general, the number of unknowns will be greater than
the number of equations. For a basis with N terms, we will let R represent the
number of independent linear coefficients. In the traditional variational method R
is taken to be N. In the new approach, in general, R is less than N and hence any
of the N coefficents can be written as a linear combination of the R independent
coefficients. We write the recurrence relationship as
am = gmj aj , (2.2)
j
where the sum has R terms. The coefficients gmj may be found directly from
Eq. (2.1). So in the new approach, Eq. (1.5) is replaced by
∂am
= gmi . (2.3)
∂ai
Minimization of the Rayleigh quotient leads to the following matrix equation
Aφ = ΛBφ. (2.4)
For a basis set with N terms, these matrices have dimensions R×R. The eigenvectors
contain the values of the independent a. The eigenvalues will be approximations to
the exact energies. The elements of the matrices A and B are
N
N
Aij = gmi gnj Hmn and Bij = gmi gnj Fmn . (2.5)
mn=1 mn=1
752 C. Stubbins
§3. Results
We apply this method to two models, the quartic anharmonic oscillator and the
Zeeman effect in Hydrogen. For the anharmonic oscillator, R = 2 for all values of N .
Therefore, the N × N matrix diagonalization of the traditional variational method
is replaced by a 2 × 2 matrix diagonalization in this new approach. For an N -basis,
the R × R numerical results are essentially the same as the traditional N × N results.
For the Zeeman problem N, depends on the size of the basis. As the basis
gets bigger, the number of independent coefficients, also gets bigger. However, R