I have a confusion with regards to the principle of QM that states that time evolution must be unitary. In particular, given that states transform through time as $|\Psi(t)\rangle = U(t)|\Psi(0)\rangle$; does the condition: $$\langle\Phi(0)|\Psi(0)\rangle=0 \ \implies \ \langle\Phi(t)|\Psi(t)\rangle=0 $$

imply that $U$ must be unitary, or is it imposed on $U$?

I realize that with the before mentioned condition it is posible to prove that for two orthonormal memebers of the basis, $i,j$: $\langle i| U^{\dagger}U | j \rangle = 0$. Is it posible to prove that for the same member the product is 1?

  • $\begingroup$ If $U$ is unitary then $U^{\dagger}$ is the inverse of $U$ by definition. $\endgroup$ Sep 27, 2017 at 14:10
  • $\begingroup$ Yes, but the question is about how to show that, that is the case $\endgroup$
    – NBit
    Sep 27, 2017 at 14:18
  • $\begingroup$ Since $\vert \Psi(t)\rangle = U\vert\Psi(0)\rangle$ then clearly $$ \langle \Phi(t)\vert \Psi(t)\rangle = \langle \Phi(0)\vert U^\dagger\,U\vert \Psi(0)\rangle = \langle\Phi(0)\vert\Psi(0)\rangle $$ valid $\forall t$ implies $U^\dagger U=\hat 1\, \forall t$ as well, provided your kets are arbitrary. (Hopefully this should be enough.) $\endgroup$ Sep 27, 2017 at 14:32
  • $\begingroup$ The condition of orthogonality preservation you stated alone would also allow for operators that satisfy $U^\dagger U = c \cdot 1$ with any constant $c$. Usually one says that unitarity comes from norm preservation, i.e. $< \psi(t) ,\psi(t)> = <\psi(0),\psi(0)>$. $\endgroup$
    – Luke
    Sep 27, 2017 at 15:23

1 Answer 1


It is an interesting elementary problem. After I while I proved the following proposition where I use $A^*$ for the adjoint of $A$.

Proposition. Let $U: H \to H$ be a bounded operator over a Hilbert space $H$. The following conditions are equivalent.

(a) For $x,y \in H$,$\quad$ $\langle x|y\rangle =0$ implies $\langle Ux|Uy\rangle =0$

(b) $U^*U = cI$ for some real $c\geq 0$.

Before proving the statement I observe that even if $c=1$, $U$ is not necessarily unitary, because unitarity is $U^*U=UU^*=I$. And here $UU^*=I$ generally fails when $H$ is infinite dimensional (otherwise it is trivially true as a consequence of $U^*U=I$). For $c=1$, $U$ is an isometry not necessarily surjective.

Proof. It is obvious that (b) implies (a), so we prove that (a) implies (b). Condition (a) can be rephrased as $y \perp x$ implies $y \perp U^*Ux$. As a consequence $U^*Ux \in \{\{x\}^\perp\}^\perp$ which is the linear span of $x$. In other words $U^*U x = \lambda_x x$ for some $\lambda_x\in \mathbb C$. My goal is now proving that $\lambda_x$ does not depend on $x$.

To this end, consider a couple of vectors $x \perp y$ with $x, y \neq 0$. Using the argument above we have $$U^*U x = \lambda_x x\:,\quad U^*U y = \lambda_y y\:, \quad U^*U (x+y) = \lambda_{x+y} (x+y)\:.\tag{1}$$ Linearity of $U^*U$ applied to the last identity leads to $$U^*Ux + U^*Uy = \lambda_{x+y}x + \lambda_{x+y}y\:,$$ namely $$U^*Ux- \lambda_{x+y}x = -(U^*Uy- \lambda_{x+y}y)\:\:.$$ Exploiting the first two identities in (1) we get $$(\lambda_x- \lambda_{x+y})x = -(\lambda_y- \lambda_{x+y})y\:.$$ Since $x \perp y$ and $x,y \neq 0$, the only possibility is that $$\lambda_x = \lambda_{x+y} = \lambda_y\:.$$ So a couple of orthogonal non-vanishing vectors has the same $\lambda_x$.

To conclude consider a Hilbert basis $\{x_n\}$ of $H$ so that, if $z\in H$, $$z = \sum_n c_n x_n \tag{2}$$ for complex numbers $c_n$. Since $U^*U$ is continuous ($U$ is bounded), $$U^*Uz = \sum_n c_n U^*Ux_n = \sum_n c_n \lambda_{x_n}x_n\tag{3}$$ But we know from the previous argument that $\lambda_{x_n} = \lambda_{x_m}$ so that, indicating with $c$ the common value of the $\lambda_{x_n}$, (3) can be rewritten as $$U^*Uz = \sum_n c_n cx_n = c\sum_n c_n x_n = cz\:.$$ Since $z\in H$ was arbitrary, we have found that $$U^*U=c I\:.$$ Taking the adjoint of both sides we obtain $c=\overline{c}$ so that $c$ is real. Finally, $$0 \leq \langle Ux | Ux\rangle = \langle x| U^*U x\rangle = c \langle x| x \rangle$$ so that $c\geq 0$. QED

  • $\begingroup$ This is nice but does the part "Is it posible to prove that for the same member the product is 1?" not imply $c=\lambda_x=1$? $\endgroup$ Sep 27, 2017 at 19:32
  • $\begingroup$ I do not understand, we already know that $U$ such thst $U^*U= cI$ satisfies the initial hypotheses also for $c\neq 1$. If it were possible to prove that $c=1$ we would exclude this case that we know exists. $\endgroup$ Sep 27, 2017 at 19:40
  • $\begingroup$ Something's clearly throwing me off here. I'll come bac k to in a couple of days. $\endgroup$ Sep 27, 2017 at 19:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.