Ballentine-sections 3.2 and 3.3

[WannabeNewton]

This isn't a question about chapter 3 of Ballentine itself but is related to the notation of the chapter. In chapter 2 of Srednicki's QFT text, Srednicki goes about deriving the commutator relations for the generators of the proper orthochronous Lorentz group and consequently the structure coefficients of the Lorentz algebra similarly to Ballentine's derivation of these quantities for the Galilei group. However the notations differ slightly and I'm having trouble parsing Srednicki's notation having become comfortable with Ballentine's notation.

He starts with an infinitesimal Lorentz transformation $\Lambda^{\mu}{}{}_{\nu} = \delta^{\mu}{}{}_{\nu} + \delta \omega^{\mu}{}{}_{\nu}$ where $\delta \omega_{\mu\nu} = -\delta \omega_{\nu\mu}$. In expression 2.12 he writes, for a unitary transformation $U(\Lambda)$ of an infinitesimal Lorentz transformation, $U(1 + \delta \omega) = I + \frac{i}{2\hbar}\delta \omega_{\mu\nu}M^{\mu\nu}$.

I assume by $1 + \delta \omega$ he means the identity operator $1$ and the matrix $\delta \omega^{\mu}{}{}_{\nu}$ now construed as an operator $\delta \omega$, with both operators acting on Minkowski space-time. Now he's using abstract indices to represent the 2-form $\delta \omega_{\mu\nu}$ and then contracting it with the indices on $M^{\mu\nu}$ and based on his definition of the angular momentum operator $\mathbf{J}$ component-wise by $J_i = \frac{1}{2}\epsilon_{ijk}M^{jk}$ and the Lorentz boost operator $\mathbf{K}$ also component-wise by $K_{i} = M^{0i}$ it seems that the indices on $M^{\mu\nu}$ aren't abstract indices but rather just labels for Hermitian generators of the proper orthochronous Lorentz group along the different spatial axes in which case I can neither make sense of how he gets $U(1 + \delta \omega) = I + \frac{i}{2\hbar}\delta \omega_{\mu\nu}M^{\mu\nu}$ nor make sense of how to even parse it.

Thanks.

 

[George Jones]

Each each $\Lambda$ is Lorentz transformation matrix; consequently, each $\Lambda^{\mu}{}{}_{\nu}$ is a component of a matrix, i.e., a number; so, each $\delta \omega_{\mu \nu}$ is a number (infintesimal?); aach $M^{\mu \nu}$ is an operator (or matrix, if the space is finite-dimensional and a basis has been chosen); $\delta \omega_{\mu \nu} M^{\nu \mu}$ is a linear combination of operators.

 

[strangerep]

Adding a little to what George said...

I assume by $1 + \delta \omega$ he means the identity operator $1$ and the matrix $\delta \omega^{\mu}{}{}_{\nu}$ now construed as an operator $\delta \omega$, with both operators acting on Minkowski space-time.

An element of the abstract Lorentz group can be represented as 4x4 matrices acting as transformations on Minkowski spacetime, or it can be represented as a unitary operator on Hilbert space. Srednicki is obviously mapping between these 2 representations. In more general situations, however, you may need to think of $U$ as a mapping from abstract group elements to their concrete representations as unitary operators. But of course, when reading Srednicki's textbook, you got to keep these two distinctions straight, since he's doing a rep-to-rep mapping.

Hence: when he writes $U(1+\delta\omega)$, the argument is obviously a 4x4 matrix, denoting a transformation applicable on Minkowski spacetime. But the rhs of (2.12) is an operator on Hilbert space, so he uses $I$ instead of $1$.

(And now I wonder whether I've only confused the issue.)

 

[WannabeNewton]

Each each $\Lambda$ is Lorentz transformation matrix; consequently, each $\Lambda^{\mu}{}{}_{\nu}$ is a component of a matrix, i.e., a number; so, each $\delta \omega_{\mu \nu}$ is a number (infintesimal?); aach $M^{\mu \nu}$ is an operator (or matrix, if the space is finite-dimensional and a basis has been chosen); $\delta \omega_{\mu \nu} M^{\nu \mu}$ is a linear combination of operators.

Ok good that's what I figured the notation was representing. So just to tie it back to Ballentine's notation, if we take the definitions $J_{i} = \epsilon_{ijk}M^{jk}$ and $K_{i} = M^{i0}$ and note that $M^{ij} = \epsilon^{ijk}J_{k}$ as a result of the first definition, we have $\frac{1}{2}\delta \omega_{\mu\nu}M^{\mu\nu} = \delta \omega_{i0}M^{i0}+\frac{1}{2}\epsilon^{ijk}\omega_{ij}J_{k} = \delta \omega_{0x}K_{x}+\delta \omega_{0y}K_{y}+\delta \omega_{0z}K_{z} +\omega_{xy}J_{z} + \omega_{yz}J_{x} + \omega_{zx}J_{y}$
so if I wrote $U(1 + \delta \omega) = I + i\sum_{\mu = 1}^{6}s_{\mu}K_{\mu}$ as in Ballentine (units of $\hbar = 1$) and made the identifications $s_1 = \delta \omega_{0x}$, $s_{4} = \delta \omega_{xy}$ etc. and $K_{1} = K_{x}$, $K_{4} = J_{z}$ etc. would it be fair to say it's equivalent to what Srednicki writes?

Also, consider a killing field $\xi^{\mu}$ in Minkowski space-time and a constant 4-velocity field $u^{\mu}$. Letting $x^{\mu}$ be the position vector field relative to $u^{\mu}$, we can decompose $\xi^{\mu}$ as $\xi^{\mu} = 2E^{[\nu}u^{\mu]}x_{\nu} + \epsilon^{\mu\nu\alpha\beta}x_{\nu}u_{\alpha}B_{\beta}$ where $E^{\mu},B^{\mu}$ are space-like vectors. Pure Lorentz boosts correspond to the "electric" part $2E^{[\nu}u^{\mu]}x_{\nu}$ and pure rotations correspond to the "magnetic part" $\epsilon^{\mu\nu\alpha\beta}x_{\nu}u_{\alpha}B_{\beta}$ where the rotation axis is given by $B^{\mu}$.

If we go to the rest frame of $u^{\mu}$ and consider an elemental pure rotation given by $B^{\mu} = (\partial_z)^{\mu}$ for example then $\xi = x\partial_{y}-y\partial_{x}$ which is in obvious 1-1 correspondence with the generator of rotations about the $z$ axis $J_z = X P_{y} -Y P_{x}$. For an elemental pure Lorentz boost in the $x$ direction, for which $E^{\mu} = (\partial_{x})^{\mu}$, we get $\xi = x\partial_t+t\partial_x$. Can an analogous 1-1 correspondence between this and $K_z$, the generator of Lorentz boosts in the $x$ direction, be made in the same obvious manner? This is just so I can connect the Hermitian generators of the proper orthochronous Lorentz group back to the familiar notion of Minkowski space-time killing fields as the generators of isometric flows in Minkowski space-time.

Adding a little to what George said...
In more general situations, however, you may need to think of $U$ as a mapping from abstract group elements to their concrete representations as unitary operators.

Could you elucidate this point? I understand that when we write $U(1 + \delta \omega)$, we basically just saying that to each representation of the infinitesimal Lorentz transformation as an operator $1 + \delta \omega; \delta^{\mu}{}{}_{\nu} + \delta \omega^{\mu}{}{}_{\nu}$ on Minkowski space-time there is associated a representation of the same infinitesimal Lorentz transformation as a unitary operator $I + \frac{i}{2\hbar}\delta \omega_{\mu\nu} M^{\mu\nu}$ on Hilbert space, a rep to rep mapping as you put it, but I'm afraid I don't quite understand the quoted statement above.

Thanks guys!

 

[rubi]

Could you elucidate this point? I understand that when we write $U(1 + \delta \omega)$, we basically just saying that to each representation of the infinitesimal Lorentz transformation as an operator $1 + \delta \omega; \delta^{\mu}{}{}_{\nu} + \delta \omega^{\mu}{}{}_{\nu}$ on Minkowski space-time there is associated a representation of the same infinitesimal Lorentz transformation as a unitary operator $I + \frac{i}{2\hbar}\delta \omega_{\mu\nu} M^{\mu\nu}$ on Hilbert space, a rep to rep mapping as you put it, but I'm afraid I don't quite understand the quoted statement above.

The correct mathematical setting to talk about this stuff is the theory of unitary representations of Lie groups. In general, you are given a Lie group $G$, that is a group with an additional manifold structure, such that the group multiplication and the inversion are smooth maps. In your case, this group is the Poincare group (or rather it's connected component of the identity). A unitary representation of a group is a pair $(\mathcal H, \pi)$ with $\mathcal H$ being a Hilbert space and $\pi:G\rightarrow U(\mathcal H)$ is a (strongly continuous) map from the group into the unitary operators on $\mathcal H$ such that $\pi(e) = \mathrm{id}_\mathcal{H}$ and $\pi(g)\pi(h)=\pi(gh)$. The goal of representation theory is to classify all these representations (up to equivalence).

For the Poincare group, it turns out that you can classify all the (irreducible, roughly speaking "minimal") (projective) representations by two numbers, which we call mass and spin. This is due to Wigner ("On unitary representations of the inhomogeneous Lorentz group").

There is much more to be said about this. For example the role of the infinitesimal transformations is played by Lie algebras and their representations and there is a way to relate them to the representations of the Lie group (that's why we actually do it) and one has to deal with some complications. But I'm not sure how much you want to learn about it, so I stop here for now.

 

[strangerep]

For an elemental pure Lorentz boost in the $x$ direction, for which $E^{\mu} = (\partial_{x})^{\mu}$, we get $\xi = x\partial_t+t\partial_x$. Can an analogous 1-1 correspondence between this and $K_z$, the generator of Lorentz boosts in the $x$ direction, be made in the same obvious manner? This is just so I can connect the Hermitian generators of the proper orthochronous Lorentz group back to the familiar notion of Minkowski space-time killing fields as the generators of isometric flows in Minkowski space-time.

I guess you meant $K_x$ ?

It's interesting to me to see how you're approaching this in terms of more general techniques usually associated with GR and DG. When I learned this stuff, I never thought about Killing fields, etc, at all.

But anyway, the answer is essentially "yes". A more low-brow way to get the explicit generator is to consider an ordinary 1-parameter transformation of the form $x^\mu \to x'^{\,\mu}(x^\lambda, p)$, where $p$ is the parameter. The associated generator $P$ can be found via this formula:
$$
\def\Pdrv#1#2{\frac{\partial #1}{\partial #2}}
\def\pdrv#1{\frac{\partial}{\partial#1}}
P ~:=~ \left. \Pdrv{x'^{\,\mu}}{p} \right|_{p = p_0} \pdrv{x^\mu} ~,
$$
where $p_0$ denotes the value of $p$ corresponding to the identity transformation. I.e., the first derivative above is to be evaluated at the identity.

In the case of a coordinate transformation corresponding to a boost in velocity, one gets essentially what you wrote.

Could you elucidate this point? I understand that when we write $U(1 + \delta \omega)$, we basically just saying that to each representation of the infinitesimal Lorentz transformation as an operator $1 + \delta \omega; \delta^{\mu}{}{}_{\nu} + \delta \omega^{\mu}{}{}_{\nu}$ on Minkowski space-time there is associated a representation of the same infinitesimal Lorentz transformation as a unitary operator $I + \frac{i}{2\hbar}\delta \omega_{\mu\nu} M^{\mu\nu}$ on Hilbert space, a rep to rep mapping as you put it, but I'm afraid I don't quite understand the quoted statement above.

Urk, I knew I was just confusing matters by jumping ahead like that.

Rubi has already given a good answer, and I'd probably just add to the confusion by trying to embelish it. But feel free to ask further questions if you can't keep your curiosity under control.

 

[WannabeNewton]

I guess you meant $K_x$ ?

Ah yeah sorry typo.

In the case of a coordinate transformation corresponding to a boost in velocity, one gets essentially what you wrote.

Thanks! So it seems then, after rewriting the pure boost killing fields along the $x$-axis as $t\partial_{x} + x\partial_{t}\rightarrow x^{0}\partial^{1} - x^{1}\partial^{0}$ (where the negative sign comes from raising time indices using the Minkowski metric) and similarly for pure boost killing fields along the other axes, as well as rewriting the rotational killing fields about the $z$-axis as $x\partial_{y} - y\partial_{x}\rightarrow x^{1}\partial^{2} - x^{2}\partial^{1}$ and similarly for the pure rotational killing fields about the other axes, that $M^{\mu\nu} \propto (x^{\mu}\partial^{\nu} - x^{\nu}\partial^{\mu})$ with the proportionality being given by some appropriate quantization factor. If so, why didn't Srednicki just write down an explicit form for $M^{\mu\nu}$ in the above manner and how is $M^{\mu\nu}$ different from the $\mathcal{L}^{\mu\nu} = \frac{\hbar}{i}(x^{\mu}\partial^{\nu} - x^{\nu}\partial^{\mu})$ that Srednicki defines in problem 2.8? For example $\mathcal{L}^{01} = \frac{\hbar}{i}(x^{1}\partial^{0} - x^{0}\partial^{1}) = i\hbar(x^{0}\partial^{1}-x^{1}\partial^{0})$ and this seems to correspond to the boost operator $K_{x}$ based on the form of the boost killing field in the $x$ direction.

 

[WannabeNewton]

There is much more to be said about this. For example the role of the infinitesimal transformations is played by Lie algebras and their representations and there is a way to relate them to the representations of the Lie group (that's why we actually do it) and one has to deal with some complications. But I'm not sure how much you want to learn about it, so I stop here for now.

Thanks for the awesome primer rubi! I'm familiar with this from classical mechanics as well as general smooth manifold theory but I haven't personally had a chance to find a QM/QM for mathematicians book that developed all of the above in a detailed manner in the context of QM (Hall contains it but the text does it very briefly in a single chapter so I found it hard to absorb anything effectively from it). Do you have any particular reference(s) in mind?

 

[marmoset]

This is beyond me but I think the book Peter Woit is preparing (current version available here http://www.math.columbia.edu/~woit/qmbook.pdf) might interest you. From the preface:

These are the course notes prepared for a class taught at Columbia during the
2012-13 academic year. The intent was to cover the basics of quantum mechanics,
from a point of view emphasizing the role of unitary representations of
Lie groups in the foundations of the subject. The approach to this material is
simultaneously rather advanced, using crucially some fundamental mathematical
structures normally only discussed in graduate mathematics courses, while
at the same time trying to do this in as elementary terms as possible.

 

[strangerep]

[...] So it seems then, after rewriting the pure boost killing fields along the $x$-axis as $t\partial_{x} + x\partial_{t}\rightarrow x^{0}\partial^{1} - x^{1}\partial^{0}$ (where the negative sign comes from raising time indices using the Minkowski metric)

Yes.

why didn't Srednicki just write down an explicit form for $M^{\mu\nu}$ in the above manner and [...]

I learned my basic QFT from Peskin & Schroeder. Later, I bought a copy of Srednicki when several other people said nice things about it, but I was a bit disappointed, tbh.

But rest assured, you'll find plenty more (bigger) issues to complain about as you wend your way through multiple QFT texts at various levels.


BTW, regarding Peter Woit's draft textbook,... keep in mind that it's aimed at mathematicians. Therefore, it inevitably omits many of the important physical insights that are contained in Ballentine (yet are hard to find elsewhere). This is not to criticize Peter's book -- but only as a warning that a mathematical emphasis tends to gloss over important physical insights (just as a physics emphasis tends to gloss over some of the subtler maths). One needs to study the subject well from both perspectives.

You might also want to take a look at this book:

Michele Maggiore, "A Modern Introduction to QFT",
Oxford Univ. Press, 2005, ISBN 0-19-852074-3

He places less emphasis than other QFT books on practical stuff like calculating cross-sections, etc, in particle physics, and more emphasis on the underlying methods and ideas. I found that side of the subject easier to absorb from Maggiore than from Weinberg's lengthy tomes. If you look through ch2 of Maggiore, you should be able to decide whether it's suitable for you at this time.

 

[dextercioby]

Ah yeah sorry typo. Thanks! So it seems then, after rewriting the pure boost killing fields along the $x$-axis as $t\partial_{x} + x\partial_{t}\rightarrow x^{0}\partial^{1} - x^{1}\partial^{0}$ (where the negative sign comes from raising time indices using the Minkowski metric) and similarly for pure boost killing fields along the other axes, as well as rewriting the rotational killing fields about the $z$-axis as $x\partial_{y} - y\partial_{x}\rightarrow x^{1}\partial^{2} - x^{2}\partial^{1}$ and similarly for the pure rotational killing fields about the other axes, that $M^{\mu\nu} \propto (x^{\mu}\partial^{\nu} - x^{\nu}\partial^{\mu})$ with the proportionality being given by some appropriate quantization factor. If so, why didn't Srednicki just write down an explicit form for $M^{\mu\nu}$ in the above manner and how is $M^{\mu\nu}$ different from the $\mathcal{L}^{\mu\nu} = \frac{\hbar}{i}(x^{\mu}\partial^{\nu} - x^{\nu}\partial^{\mu})$ that Srednicki defines in problem 2.8? For example $\mathcal{L}^{01} = \frac{\hbar}{i}(x^{1}\partial^{0} - x^{0}\partial^{1}) = i\hbar(x^{0}\partial^{1}-x^{1}\partial^{0})$ and this seems to correspond to the boost operator $K_{x}$ based on the form of the boost killing field in the $x$ direction.

Coming from a thorough study of differential geometry into quantum mechanics/field theory, I assume you wish to stick as closely as possible to the differential geometric concepts proved useful to you in learning relativity. There is only one source that I know which attacks QFT coming from the diff. geometry used in relativity: the notes of R. Geroch at the Univ. of Chicago which are now being published as a book and offered for sale. You should venture yourself in reading those.

 

[vanhees71]

I'd rather recommend Weinberg's marvelous books on QFT. They are not very much based on differential geometry, but have a very clear and complete exposition of the symmetry of the Poincare group, which you do not find so easily in other textbooks.

Some aspects you also find in the very nice book

Sexl, Roman U., Urbandtke, Helmuth K.: Relativity, Groups, Particles, Springer, 2001

but this deals only with the representation theory not with the further development of QFT and S-Matrix theory.

Peskin and Schroeder is a standard text, but has to read with some caveats. There are a lot of topics not very carefully treated. The most shocking moment was when I found logarithms with dimensionful arguments in the chapter about the renormalization group (sic!). It's not so much the many typos in this book which I find disturbing but such conceptional weaknesses which are the more serious in view of the fact that relativistic interacting QFT is anyway mathematically not strictly formulated and maybe can never be. The more importance should be given to the parts which can be made right.

Srednicky is not so bad, but I don't understand, why he deals with \phi^3 theory, which is a priori unstable. Formally you can demonstrate some formal issues in perturbative renormalization, but otherwise it has not much value. On the other hand, it gives a careful exposition of the important and not so trivial topics like the LSZ reduction formalism and S-matrix theory.

To begin with, I'd recommend also Ryder. A more path-integral based approach you find in the book by Bailin and Love, Gauge Field Theories, which I liked very much too when I learned QFT. The best choice, in my opinion, still is Weinberg's Quantum Theory of Fields (3 vols., with vol. 3 about SUSY QFT), as mentioned above. Another somewhat more advanced book is Boehm, Dittmaier, Joos, Gauge theories of the strong and electroweak interaction (Teubner).

If you are interested in the special topic of QFT in curved spacetime, then have a look at deWitts famous review:

B. deWitt, Quantum Field Theory in Curved Space Time, Phys. Rept. 19, 295 (1975)
http://dx.doi.org/10.1016/0370-1573(75)90051-4

 

[andresB]

The thing I recall is it took me a little while to fully understand - its a bit dense.

The other thing I recall is there is a bit of thought required about exactly why a free particle forms an irreducible set.

The other thing is the exact key assumption being made in quantitization - it took me a while to nut that one out.

I'm having problems with those two , any help would be appreciated.

 

[bhobba]

I'm having problems with those two , any help would be appreciated.

This thread is over 18 months old.

I think it would be wise to start a new one detailing exactly what your problems are.

It been a long while since I have gone through it myself.

Thanks
Bill

 

[strangerep]

I'm having problems with those two , any help would be appreciated.

I agree with Bhobba -- start a new thread. Also, about irreducibility, be sure to read the relevant appendices in Ballentine's textbook which cover this in a bit more detail.