Put
The Rayleigh quotient shift is exactly
Expanding the denominator gives
The question prints this as , so its sign convention requires
unless every vanishes. Equivalently, is the positive quadratic Rayleigh-quotient error.
The shifted inverse solve has the spectral decomposition
Only relative coefficients matter after normalization, so factor out :
From the shift expansion,
whereas, for ,
Consequently
After the final normalization,
Thus an eigenvector-direction error becomes in one step, proving the Local cubic convergence of Rayleigh quotient iteration for this symmetric matrix.