I am working through Leonard Susskind's The Theoretical Minimum: Quantum Mechanics. In this book a statement called the "spin-polarization principle" is introduced, which essentially states that:

For any state $|A \rangle$, there exists a direction vector $\hat{n}$ such that $\vec{\sigma} \cdot \hat{n} \, |A \rangle = |A\rangle$.

(Here $\sigma$ is used for spin operators - I've seen $S_n, S_x, S_y, S_z$ used elsewhere).

I understand this to mean that for any state there always exists a direction that a spin-measuring apparatus can be oriented in such that it will measure the spin as $+1$ with $100\%$ certainty. We can therefore write that the expectation value of this observable is $1$:

$$\langle \vec{\sigma} \cdot \hat{n} \rangle = 1.$$

We have that $\vec{\sigma} \cdot \hat{n} = n_x \sigma_x + n_y \sigma_y + n_z \sigma_z$ where $n_x, n_y, n_z $ are the components of $\hat{n}$.

The book then states that "the expectation value of the perpendicular components of $\sigma$ are zero in the state $|A\rangle$" and then states that it "follows" from this that

$$\langle \sigma_x \rangle ^2 + \langle \sigma_y \rangle ^2 + \langle \sigma_z \rangle ^2 = 1.$$

I don't understand what the book means by this though, or how you deduce that the sum of the squares of the expected values of the spin components is 1.

I think it might be that the "perpendicular components" being $0$ refers to the spin component measured in a direction perpendicular to $\hat{n}$ in $3$D space (because if a $+1$ spin is prepared along $\hat{n}$ then the expected value of the spin measurement perpendicular to $\hat{n}$ (along a vector $\hat{m}$) is $\hat{n} \cdot \hat{m} = 0$.

We also can show that

$$\langle \vec{\sigma} \cdot \hat{n} \rangle = \langle A | \vec{\sigma} \cdot \hat{n} | A \rangle = n_x \langle \sigma_x \rangle + n_y \langle \sigma_y \rangle + n_z \langle \sigma_z \rangle,$$

which is the closest that I've got to showing that the sum of the squares of the expected values of the spin components is $1$.

What does the book mean by this statement, and how do we deduce that

$$\langle \sigma_x \rangle ^2 + \langle \sigma_y \rangle ^2 + \langle \sigma_z \rangle ^2 = 1?$$

  • $\begingroup$ Are you sure it isn't $\langle \sigma_x^2\rangle + \langle \sigma_y^2\rangle + \langle \sigma_z^2 \rangle = 1$ ? $\endgroup$
    – Andrew
    Commented Jul 23, 2020 at 11:39
  • 1
    $\begingroup$ @Andrew No. The sum of the expectation values of the squares is 3/4. $\endgroup$
    – mike stone
    Commented Jul 23, 2020 at 12:51

2 Answers 2


Start with any normalized state $\vert\psi\rangle=\alpha\vert{+}\rangle +\beta\vert{-}\rangle $. Indeed this most general $\vert\psi\rangle$ is of the form $\cos\beta/2 \vert +\rangle +e^{i\varphi}\sin\beta/2\vert - \rangle$ and the angles $\beta$ and $\varphi$ are related to the average values of the Pauli matrices, v.g $\langle \sigma_z\rangle=\cos\beta=n_z$.

Thus you immediately get \begin{align} \sum_i \langle \sigma_i\rangle^2 =\sum_i n_i^2=1 \end{align}

Note that this $\vert\psi\rangle$ is also an eigenstate of $\hat n\cdot\vec\sigma$.


Since state $A$ is polarized by the direction of $\vec n$, the expected value of observable $\sigma_x$ is $n_x$. Or $\langle \sigma_x \rangle = n_x$. $n_y$ and $n_z$ are not participating in the expected value of $\sigma_x$ because the component $\sigma_y$ and $\sigma_z$ are perpendicular to $\sigma_x$.

By symmetric arguments, $\langle \sigma_y \rangle = n_y$ and $\langle \sigma_z \rangle = n_z$.

Follow the derivation:

$$\langle \vec\sigma \cdot \vec n \rangle = n_x\langle\sigma_x\rangle + n_y\langle\sigma_y\rangle + n_z\langle\sigma_z\rangle = {\langle \sigma_x \rangle}^2 + {\langle \sigma_y \rangle}^2 + {\langle \sigma_z \rangle}^2 = 1$$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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