= Input penalty of a history Hamiltonian
{title2=$H_{\rm in}=Q\otimes|0\rangle\langle0|$}
The <input penalty of a history Hamiltonian> tests prescribed <ancilla qubits> at clock time zero while leaving the <quantum witness> unrestricted. A zero <ancilla qubit> is tested by the bad-state projector $|1\rangle\langle1|$; a plus <ancilla qubit> is tested by $|-\rangle\langle-|$.
Let $Q=\sum_{j=1}^m\Pi_j$, where each $\Pi_j$ tests a different <ancilla qubit>. These <orthogonal projections> commute because they act on separate <tensor product> factors. Thus $Q$ is a <positive operator> with integer <eigenvalues> between zero and $m$. Zero is essential: its <kernel> contains exactly the valid ancillary input subspace, tensored with arbitrary <quantum witness> states. Only the nonzero <eigenvalues> are positive integers.
The clock-conditioned term is $H_{\rm in}=Q\otimes|c_0\rangle\langle c_0|$. For a normalized <computational history state> $|\eta\rangle=(T+1)^{-1/2}\sum_t|\psi_t\rangle|c_t\rangle$ with clock labels forming an <orthonormal basis>,
$$
\langle\eta|H_{\rm in}|\eta\rangle=\frac{\langle\psi_0|Q|\psi_0\rangle}{T+1}.
$$
Together with correct propagation, zero input penalty selects histories with valid initial data. The displayed expectation concerns this fixed history state; a bound on the full <Hamiltonian>'s lowest <eigenvalue> also has to account for propagation penalties.
Back to article page