shift operator

The shift operator is the small, concrete example that demolishes finite-dimensional intuition in one stroke. On ell^2, the right shift takes a sequence (x_1, x_2, x_3, ...) and pushes everything one slot to the right, inserting a zero: (0, x_1, x_2, x_3, ...). It is as simple as an operator gets, and yet it does something no matrix can: it is injective and length-preserving but not surjective.

Precisely: the right shift S on ell^2 is S(x_1, x_2, ...) = (0, x_1, x_2, ...). It is an isometry, ||S x|| = ||x|| for all x, hence injective and bounded with ||S|| = 1. But its range is exactly the sequences starting with 0, a proper closed subspace — so S is not onto, and has no inverse. Its adjoint is the left shift S^*(x_1, x_2, ...) = (x_2, x_3, ...), which is onto but not injective.

Why it shatters intuition: in finite dimensions a linear map is injective if and only if it is surjective (rank-nullity forces it). The shift shows this equivalence is purely finite-dimensional. S is one-to-one yet misses a whole coordinate's worth of the space. Even more striking, S has no eigenvalues at all — S x = lambda x forces x = 0 — so its point spectrum is empty, while its spectrum is the entire closed unit disc.

Why it matters: the shift is the universal building block of operator theory. The Wold decomposition writes any isometry in terms of shifts; the shift's invariant subspaces are described by Beurling's theorem via inner functions in the Hardy space; and it is the standard counterexample whenever someone forgets that infinite dimensions are different. Keep it in your pocket as the antidote to over-trusting finite-dimensional reflexes.