False under the hypotheses as printed. Nonzero coefficients do not ensure the two-column dominant-subspace condition.
For a counterexample, take , eigenvalues
and an orthonormal eigenbasis for which
The two displayed coefficient vectors are orthonormal and can be completed to an orthogonal matrix, so such a real symmetric matrix exists. Every and is nonzero, but
Indeed, with ,
is a nonzero multiple of , while the remaining direction in
is
Thus
By the simultaneous iteration interpretation of the QR algorithm, the leading two columns of the accumulated factor span this same space. The leading block is therefore an orthogonal-coordinate representation of the compression of to this space. Since
we have
for every relevant , rather than . Hence the requested Hausdorff convergence does not hold.
For completeness, the intended statement becomes true if one adds
Let consist of the first two columns of . The Accumulated QR factorization identity gives
After division by , every component along
tends to zero because of the strict spectral gap, while makes the two surviving dominant components independent. Therefore these two-dimensional subspaces converge to the dominant invariant subspace
There are two-by-two orthogonal matrices such that
Because
up to the harmless index convention at , we obtain
in the matrix 2-norm. Orthogonal similarity preserves the spectrum, and the stated spectral perturbation bound now gives
Unshifted QR algorithm Created 2026-09-24 Updated 2026-11-05
The unshifted QR algorithm starts at , factors , and forms
for real data with orthogonal . Each step is an orthogonal similarity, so it preserves eigenvalues. It is the zero-shift case of the QR algorithm; complex data use a unitary factor and conjugate transpose.
Convergence needs hypotheses. For a real symmetric matrix with suitably separated eigenvalue moduli and nondegenerate starting projections, the simultaneous iteration interpretation of the QR algorithm explains convergence toward diagonal form. Equal moduli can prevent that behavior: for , the valid factorization , leaves every iterate equal to . Its eigenvalues remain coupled.
For nonsymmetric matrices, the useful target is generally a triangular form or a real Schur form, rather than a diagonal matrix. Shifts and block deflation in the QR algorithm improve practical convergence. The similarity identity by itself is not a convergence theorem.