I am trying to understand Bell's famous paper from 1964 on the EPR paradox.

I stumbled over equation $(9)$ where it says

$$\left.\begin{align}A(\vec a,\vec\lambda)&=\operatorname{sign}\vec a\cdot\vec\lambda\\B(a,b)&=-\operatorname{sign}\vec b\cdot\vec\lambda \end{align}\right\}\tag9$$

Apart from the missing vector arrows I don't understand why it's $B(a, b)$ there. Could Bell actually have meant $B(b, \lambda)$ instead? I wondered about this because I also don't understand how he concludes $(10)$ from this, but this might be another question.

  • 4
    $\begingroup$ It's somehow comforting to see even very famous papers have mistakes :) $\endgroup$
    – Andrew
    Commented Aug 11, 2021 at 19:39

1 Answer 1


Yes, this is a typo on the left-hand side of the second equation; it is correct that $$B(\vec{b},\vec{\lambda})=-\mathrm{sign}\,\vec{b}\cdot\vec{\lambda}.$$

To see how he concludes (10), let's rewrite and put the normalization in explicitly for this uniform distribution: $$P(\vec{a},\vec{b})=\frac{\int d\lambda A(\vec{a},\vec{\lambda})B(\vec{b},\vec{\lambda})}{\int d\lambda}.$$ The integral is over the surface of a sphere with unit radius, so the denominator will be $4\pi$. The product $A(\vec{a},\vec{\lambda})B(\vec{b},\vec{\lambda})=-\mathrm{sign}\,(\vec{a}\cdot\vec{\lambda})\mathrm{sign}\,(\vec{b}\cdot\vec{\lambda})$ will either be $+1$ or $-1$, and we need to determine for what fraction of the values of $\vec{\lambda}$ each of the $\pm1$ values occur.

The rest is geometry. Set the coordinate system, without loss of generality, to have $\vec{a}$ point to the top of the sphere and $\vec{b}$ be at polar angle $\theta$ and azimuthal angle $0$. This is always allowed because the integral over $\lambda$ is spherically symmetric. The region of values of $\vec{\lambda}$ for which $\mathrm{sign}\vec{a}\cdot\vec{\lambda}$ is $+1$ is the top half of the sphere; the region for which $\mathrm{sign}\vec{b}\cdot\vec{\lambda}$ is $+1$ is the hemisphere whose pole points in the direction of $\vec{b}$. The overlap of the two $+1$ regions forms a spherical wedge with angular width $\pi-\theta$. By symmetry, the spherical wedge on the opposite side of the sphere also has angular width $\pi-\theta$, and there both values are $-1$. (You can check the limits: when $\theta=0$, they fully overlap, and when $\theta=\pi$ they do not.)

From this picture, we see that $\mathrm{sign}\,(\vec{a}\cdot\vec{\lambda})\mathrm{sign}\,(\vec{b}\cdot\vec{\lambda})=1$ for $2\pi-2\theta$ out of the possible $2\pi$ angular width and $\mathrm{sign}\,(\vec{a}\cdot\vec{\lambda})\mathrm{sign}\,(\vec{b}\cdot\vec{\lambda})=-1$ for the other $2\theta$. The total probability is thus (remembering the extra $-$ sign) $$P(\vec{a},\vec{b})=-\frac{2\pi-2\theta}{2\pi}+\frac{2\theta}{2\pi}=-1+\frac{2\theta}{\pi}, $$ as per Bell.

  • $\begingroup$ Thanks a lot for this detailed explanation! I could follow this, except for the point of the normalization. Where is the $4\pi$ used? $\endgroup$ Commented Aug 11, 2021 at 21:26
  • 1
    $\begingroup$ Yeah, I didn't really use it directly. Instead, I concluded that we only have to care about one angular coordinate, so that was normalized by $2\pi$. A more complete mathematical proof would do the integral over all of the surface area and show there to be symmetry about one axis $\endgroup$ Commented Aug 11, 2021 at 21:30

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.