Spin and magnetic field of a photon

In summary, the spin of a photon does not have a direction in the same sense as the magnetic field of the photon. The electromagnetic field of a single photon is transversal, with both the magnetic and electric fields perpendicular to the direction of propagation. In the Zeeman effect, the sigma and pi photons emitted have different polarizations, with the sigma photons having m=+-1 and the pi photons having m=0. This is related to the change in the state of the atom when the photon is emitted and the angular momentum of the photon about an axis parallel to the magnetic field. Overall, the polarization of the light is connected to the average angular momentum of the photons emitted.
  • #36
Now everything is a big mess again! Photons are not as simple as other quanta since they are massless and have spin 1. The first subtlety is that they have only 2 not three spin states. That's because there's no restframe for massless quanta. The full group theoretical foundation can be found in my lecture notes on QFT in appendix B:

http://th.physik.uni-frankfurt.de/~hees/publ/lect.pdf

The 2nd subtlety is that there's no position operator for photons. You can't treat photons in a first-quantization way to begin with. It's the very point where you really need QFT, if you want to get things right (although there are claims in the literature to the contrary).

The 3rd subtlety is that the naive canonical quantization doesn't work, because the classical Maxwell theory is a gauge theory. As is also shown in my QFT notes, that must be so from the group-theoretical point of view, if you don't want to introduce continuous spin-like degrees of freedom, which are not observed.

A quick hand-wavy way out (for the free field) is to completely fix the gauge. For the free em. field the most convenient choice is the radiation gauge, which consists of two gauge conditions for the vector field
$$A^0=0, \quad \vec{\nabla} \cdot \vec{A}=0.$$
The equation of motion is
$$\Box \vec{A}=(\partial_t^2-\Delta)\vec{A}=0.$$
Now we look for plane-wave modes and get
$$\vec{A}(t,\vec{x})=a(\vec{k},\lambda) \vec{\epsilon}_{\lambda} \exp(-\mathrm{i} \omega t + \mathrm{i} \vec{k} \cdot \vec{x})+c.c.$$
##\lambda## labels two polarization vectors, which can be chosen arbitrarily, but they must fulfill ##\vec{\epsilon}_{\lambda} \cdot \vec{k}=0## (i.e., you have two linearly independent choices). The most natural choice are the two circular polarized modes, which corresponding to helicity eigenmodes to the eigenvectors ##h \in \{-1,1 \}##. Then there's the dispersion relation ##\omega=|\vec{k}|##.

Now we skip all the subtleties of canonical quantization by observing that we have a set of harmonic oscillators there and thus we can make the ansatz
$$\hat{\vec{A}}(t,\vec{x}) = \sum_{\vec{k},\lambda} \frac{1}{\sqrt{2 \omega}} \hat{a}(\vec{k},\lambda) \vec{\epsilon}(\vec{k},\lambda) \exp(-mathrm{i} k \cdot x)|_{k^0=\omega}+\text{h.c.}$$
Here, I've assumed a large quantization volume as a cube of length ##L## with periodic boundary conditions for the field operator, so that the momenta are discrete, ##\vec{k} \in \frac{2 \pi}{L} \mathbb{Z}^3##.

The creation and annhiliation operators fulfill the Bose commutator relations
$$[\hat{a}(\vec{k},\lambda) ,\hat{a}(\vec{k}',\lambda')]=0, \quad [\hat{a}(\vec{k},\lambda),\hat{a}^{\dagger}(\vec{k}',\lambda')]=\delta_{\vec{k},\vec{k}'} \delta_{\lambda,\lambda'}.$$
The total four-momentum is given by
$$\hat{P}^{\mu} = \sum_{\vec{k},\lambda} \hat{N}(\vec{k},\lambda) k^{\mu}, \quad \hat{N}(\vec{k},\lambda)=\hat{a}^{\dagger}(\vec{k},\lambda) \hat{a}(\vec{k},\lambda).$$
The ground state is the vacuum ##|\Omega \rangle## defined by
$$\hat{a}(\vec{k},\lambda) |\Omega \rangle=0.$$
A single-mode single-photon state is given by
$$|\vec{k},\lambda \rangle=\hat{a}^{\dagger}(\vec{k},\lambda) |\Omega \rangle.$$
A single-mode coherent state that is most closely related to a classical em. field is given by the eigenvector of ##\hat{a}(\vec{k},\lambda)##,
$$\hat{a}(\vec{k},\lambda) |\alpha \rangle = \alpha |\alpha \rangle.$$
Since ##\hat{a}## is NOT self-adjoint ##\alpha \in \mathbb{C}##. It's not a state of definite photon number. The probability to find ##n## photons (of momentum ##\vec{k}## and with polarization ##\lambda## of course) is given by a Poisson distribution
$$P(n)=\exp(-n|\alpha|^2) \frac{|\alpha|^{2n}}{n!}$$
The mean photon number and four-momentum is
$$\langle N(\vec{k},\lambda) \rangle=|\alpha|^2, \quad \langle \vec{P} \rangle=k^{\mu} |\alpha|^2.$$

The operators for the electromagnetic field components are given by
$$\hat{\vec{E}}(t,\vec{x})=-\partial_t \hat{\vec{A}}(t,\vec{x}), \quad \hat{\vec{B}}(t,\vec{x})=\vec{\nabla} \times \hat{\vec{A}}(t,\vec{x}).$$
For a given state you can calculate expectation values as for any observable according to Born's rule.
 
Last edited:
  • Like
Likes PeterDonis
Physics news on Phys.org
  • #37
PeterDonis said:
And what is the preparation procedure? How are we preparing a "plane-wave one-photon state"?
There is no plane-wave one-photon state (in the limit ##L \rightarrow \infty## in the formalism of my previous posting) since in this limit you have the usual ##\delta##-distribution as in ordinary QT,
$$\langle \vec{k},\lambda|\langle \vec{k}',\lambda' \rangle=\delta^{(3)}(\vec{k}-\vec{k}') \delta_{\lambda \lambda '}.$$
So for a true one-photon state, you must use a wave packet.

The preparation procedure for a single photon was very complicated not too long ago. Nowadays it's quite simple, because of the discovery of parametric down-conversion. There a laser shines into certain types of crystals (BBO=beta barium borat) and due to non-linear optics effects there's a process where one photon out of the laser field (emitting light close to a coherent state as described above) is absorbed in the crystal and two phase-matched polarization entangled photons emitted, obeying energy-momentum conservation. You use one photon (the "signal") to be sure to have created a photon pair (only about 1 out of 10^{12} laser photons are converted according to Wikipedia), using a polarizer. Then you are sure to have the other photon in a given (quite well-determined) momentum and definite polarization state (the "idler"). This is also known as a "heralded photon". This prepares true single-photon states!

https://en.wikipedia.org/wiki/Spontaneous_parametric_down-conversion

Sometimes you read that you simply might dim down a laser very much so that the average number of photons ##|\alpha|^2 \simeq 1##, but that's far from being a single-photon source. The corresponding coherent state is a superposition of mostly the vacuum state and the single-photon state but also contains small mixtures of all multi-photon states.
 
  • Like
Likes PeterDonis
  • #38
vanhees71 said:
There is no plane-wave one-photon state

That's what I thought.
 
  • #39
PeterDonis said:
That's what I thought.

In the real world there are no plane waves, period. Statements about plane waves are still useful, however, both in the limiting sense and in the Fourier decomposition sense.
 
  • #40
LeandroMdO said:
Statements about plane waves are still useful

For some purposes, yes. But I'm still not sure that justifying claims like "the magnetic field of a photon is perpendicular to its direction of propagation" (or, for that matter, "the spin of a photon is parallel with its direction of propagation") is one of them. If there is no "one photon state" that we can actually prepare and measure that has this property, then I don't think the statement is justified. If there is, it isn't a plane wave state, so what state is it?
 
  • #41
vanhees71 said:
Then you are sure to have the other photon in a given (quite well-determined) momentum and definite polarization state

How would this state be described mathematically?
 
  • #42
It's something like
$$|\Psi \rangle=\int_{\mathbb{R}^3} \mathrm{d}^3 \vec{k} \frac{1}{\sqrt{(2 \pi)^3 2 \omega(\vec{k})}} A(\vec{k}) \hat{a}^{\dagger}(\vec{k},\lambda)|\Omega \rangle,$$
where ##A(\vec{k}6)## fullfills
$$\int_{\mathbb{R}^3} \frac{\mathrm{d}^3 \vec{k}}{(2 \pi)^3 2 \omega(\vec{k})} |A(\vec{k})|^2=1.$$
 
  • #43
vanhees71 said:
It's something like

I see how this state has a definite polarization, since there is no integration over ##\lambda##; but since there is integration over ##k##, isn't this state a superposition of different momentum states, and therefore not a state with a definite value of momentum?
 
  • #44
I thought we were talking about single photon states not single photon states being exact momentum eigenstates? Why are you constantly and artificially complicating things?
 
  • #45
DrDu said:
I thought we were talking about single photon states not single photon states being exact momentum eigenstates?

If they're not exact momentum eigenstates, then they don't have a definite direction of the momentum, so it makes no sense to say that the magnetic field is transverse to the direction of the momentum--at least not unless you allow that statement to be made using expectation values, which you said you objected to (and which I don't think it's reasonable to use for one photon states anyway, since expectation values imply averaging over many observations).

Similar remarks apply to the direction of the magnetic field itself--is the "single photon state" in question an eigenstate of the magnetic field operator? I haven't worked that out for the state that @vanhees71 wrote down. But if it isn't, then the magnetic field doesn't have a definite direction either.
 
  • #46
PeterDonis said:
I see how this state has a definite polarization, since there is no integration over ##\lambda##; but since there is integration over ##k##, isn't this state a superposition of different momentum states, and therefore not a state with a definite value of momentum?
You can make ##A(\vec{k})## very sharply peakded around an arbitrary momentum ##\vec{k}=\vec{k}_0##. Then you get a true one-photon state that has nearly definite momentum and a precise polarization state.
 
  • #47
vanhees71 said:
a true one-photon state that has nearly definite momentum and a precise polarization state

Yes, "nearly definite". But that still means that when you talk about the "direction" of the momentum you're talking about an expectation value.

Also, what about the magnetic field operator? How would it act on this state?
 
  • #48
You can derive the action of ##\hat{\vec{B}}(x)## on my example state by inserting the mode decomposition given above, using the commutation relations for creation and annihilation operators.

I'm also not understanding, why you insist on plane-wave modes. They are generalized eigenfunctions of momentum. Also in classical electrodynamics plane waves are not realizable in nature exactly. They have diverging total energy and momentum.
 
  • #49
vanhees71 said:
You can derive the action of ##\hat{\vec{B}}(x)## on my example state by inserting the mode decomposition given above, using the commutation relations for creation and annihilation operators.

Ok, that's what I thought, and when I do that it does not look like your example state is an eigenstate of ##\hat{\vec{B}}(x)##.

vanhees71 said:
I'm also not understanding, why you insist on plane-wave modes

I don't "insist" on them as a description of the actual states involved; I understand perfectly well that those states are not plane waves. But if we are going to talk about the "direction of propagation" of a photon (for example if we want to say that the magnetic field is perpendicular to the direction of propagation), and the state is not a plane wave, then the only way of making sense of that direction is as an expectation value. And as has been discussed earlier in this thread, for a "single photon state" I don't think an expectation value can be given that interpretation, since taking an expectation value implies averaging over many measurements. (Similar remarks apply to the "direction" of the magnetic field as well, if I'm correct that the state you wrote down is not an eigenstate of the B operator.)
 
  • #50
It's not an eigenstate of ##\vec{B}##. Why should it be? It's an eigenstate of ##\hat{N}##!
 
  • #51
vanhees71 said:
It's not an eigenstate of ##\vec{B}##. Why should it be? It's an eigenstate of ##\hat{N}##!

That was my intuition, yes, but I wanted to check how the math worked out.
 
  • #52
PeterDonis said:
There's another caution here as well, in addition to the one I gave in my last post. If we are going to say that the magnetic field is orthogonal to the photon's momentum, we have to be able to measure each one without disturbing the other, i.e., their measurements must commute. Do they?

(Similar remarks would apply to statements about the relative directions of the magnetic field and the spin.)
No, definitely not. Let's take another example, namely the angular momentum of an electron. It is defined as ##L=r \times p##.
L is only non-zero if r isn't parallel to p.
Would you also insist that for the angular momentum to be well defined, you must be able to measure simultaneously r and p, which is impossible because they are non-commuting observables.
 
  • #53
The total angular momentum of an electron (in the non-relativistic approximation) is represented by self-adjoint operators,
$$\hat{\vec{J}}=\hat{\vec{L}}+\hat{\vec{S}}=\hat{\vec{x}} \times \hat{\vec{p}} + \hat{\vec{s}}.$$
Since it's impossible to determine (not measuring!) position and momentum of the electron simultaneously, your above description of determining orbital angular momentum of course doesn't make any sense. Nevertheless the electron can be in a definite orbital-angular momenum eigenstate, i.e., a simultaneous eigenstate of ##\hat{\vec{L}}^2 ## and ##\hat{L}_z##, which are commuting self-adjoint operators, thus representing compatible observables.

For photon states, of course the electric and magnetic components are not compatible. Nevertheless you can have definite photon-number eigenstates (as demonstrated above for a single-photon state).
 
  • #54
DrDu said:
Would you also insist that for the angular momentum to be well defined, you must be able to measure simultaneously r and p, which is impossible because they are non-commuting observables.

Well, if ##L = r \times p##, and we can't measure ##r## and ##p## simultaneously, then how do we measure the observable ##L##? I understand that we can compute its expectation value, but we're not talking about expectation values (at least I thought we weren't).
 
  • #55
You can simultaneously determine ##\vec{L}^2## and one component in an arbitrary direction (usually one takes the ##z## direction) but never all three components since ##\hat{\vec{L}}^2## and ##\hat{L}_z##, because these two operators commute, but
$$[\hat{L}_j,\hat{L}_k]=\mathrm{i} \hbar \epsilon_{ijk} \hat{L}_i.$$
 
  • #56
PeterDonis said:
But if we are going to talk about the "direction of propagation" of a photon (for example if we want to say that the magnetic field is perpendicular to the direction of propagation), and the state is not a plane wave, then the only way of making sense of that direction is as an expectation value.

Even in classical mechanics the idea that E and B are orthogonal to the direction of propagation is not realized exactly. For example, if you have a circularly polarized wave packet of finite size, the Poynting vector on the edges of the packet points in an azimuthal direction, and carries the http://people.westminstercollege.edu/faculty/ccline/courses/phys425/AJP_54(6)_p500.pdf angular momentum of the wave. Something similar happens in a waveguide (see fig 2).

So clearly the idea that electric and magnetic fields are orthogonal to the direction of propagation must be handled with care in realistic cases. That is fine, but I prefer handling conceptual issues in easy ones.

PeterDonis said:
(Similar remarks apply to the "direction" of the magnetic field as well, if I'm correct that the state you wrote down is not an eigenstate of the B operator.)

Let me try a different argument. We've been considering circularly polarized states thus far because they're the "natural" ones, but we don't have to. We can just as well quantize photons in a linearly polarized basis. You can read out easily from the expressions for the quantized electromagnetic field what this would look like, simply by replacing the set of e^\mu vectors by e^x and e^y. It is then clear that while the value of the electric field does not commute with photon number, the direction of it does (it is the definition of polarization after all).

I still maintain that this is inessential, because even for circularly polarized light the electric and magnetic fields must lie on a plane perpendicular to the momentum vector (near the center of a wavepacket of sharply peaked momentum), but it might help.
 
  • #57
PeterDonis said:
Well, if ##L = r \times p##, and we can't measure ##r## and ##p## simultaneously, then how do we measure the observable ##L##? I understand that we can compute its expectation value, but we're not talking about expectation values (at least I thought we weren't).
You could use e.g. some Stern Gerlach type experiment, i.e. measure the deflection when passing an atom through a magnetic field.
But let's look at even a simpler property, velocity: You can define it as ## v=(x(t)-x(0))/t## (I consider it to be constant for simplicity of argument). Given this expression I think you would demand that we measure x at two times, so that the particle has to be in a position eigenstate. However, it is impossible for a particle to have both a precisely known position and velocity. What comes to our rescue is that v is invariant under a shift of x, i.e., ##v=(x(t)-x(0))/t =x(t)+\Delta X -(x(0)+\Delta X))/t##, that is, we don't need absolute information about the position of the particle. In fact, velocity can be measured with high accuracy e.g. from Doppler effect measurements, e.g. on planets orbiting distant stars whose distance is only very vaguely known.
The same holds in QM. Usually, we can measure any (hermitian) operator of r and p, although it does not commute with neither r nor p. The point is that we are looking only at some specific correlation between r and p and don't need absolute information on these quantities. Especially the relative direction of B and p or S is such a correlation.
 
  • #58
I don't know about a concrete procedure to prepare a precise angular-momentum state. The SG apparatus of course comes immediately to mind. I also guess you can nowadays prepare atoms of ultracold gases in traps in all kinds of quantum states using various techniques, particularly laser excitations with taylored pulsed em. fields and the like. Maybe, we have an expert in atomic physics around, who can tell more.
 
  • #59
vanhees71 said:
I also guess you can nowadays prepare atoms of ultracold gases in traps in all kinds of quantum states using various techniques, particularly laser excitations with taylored pulsed em. fields and the like. Maybe, we have an expert in atomic physics around, who can tell more.
Polarized atomic samples can be obtained by optical pumping. This, in combination with magnetic trapping, is commonly used to get clouds of atoms in a single hyperfine state ##|F, M_F \rangle##.
 
  • Like
Likes vanhees71

Similar threads

Replies
19
Views
1K
Replies
21
Views
1K
Replies
29
Views
2K
Replies
10
Views
3K
Back
Top