The Two Casimirs of the Lorentz Algebra
Introduction
A Casimir operator is an element of the universal enveloping algebra that commutes with every generator. Its value is therefore constant on an irreducible representation, and a complete set of Casimirs is a set of numbers that names the representation. For the Lorentz algebra there are exactly two, and this article fixes them, normalises them once, computes their eigenvalues on the $(m,n)$ carrier, and shows what the pair decides that either one alone does not.
The two numbers are $m(m+1)+n(n+1)$ and $m(m+1)-n(n+1)$. The first is symmetric in the two labels and therefore cannot tell $(m,n)$ from $(n,m)$. The second is antisymmetric, and its sign is the chirality. Together they name the representation, which is what makes the $(m,n)$ table a classification rather than a list.
The article assumes The Six Generators and Their Commutators for the table and the first Casimir's normalisation, The Two Commuting Complex Triples for the split into two summands, and The Isomorphisms and Their Dimensions for the rank. It contrasts the two labels with the Poincaré labels of Wigner's Classification: Two Casimirs, Three Orbits.
Where a Casimir Lives
The Universal Enveloping Algebra
The Casimirs are not elements of the Lie algebra. The algebra's product is the bracket, and the bracket of two generators is a generator; a product of two generators, in the sense of an associative product, exists only after the algebra has been embedded in an associative algebra generated by it. That algebra is the universal enveloping algebra $\mathcal{U}(\mathfrak{so}(1,3))$, and its elements are polynomials in the generators with the bracket imposed as a relation:
$$ \mathcal{U}(\mathfrak{so}(1,3))=\frac{\mathbb{C}\langle J_x,J_y,J_z,K_x,K_y,K_z\rangle}{\langle T_aT_b-T_bT_a-f_{ab}{}^{c}T_c\rangle}. $$
Inside the biquaternion algebra the enveloping algebra is realised concretely, because the algebra is associative and the generators are its elements. So the Casimirs of this block are products of biquaternions, and they are computed by ordinary multiplication.
The Definition of a Casimir
An element $C\in\mathcal{U}$ is a Casimir when
$$ [C,T_a]=0\quad\text{for every generator }T_a . $$
By Schur's lemma its image in an irreducible representation is a scalar multiple of the identity, and that scalar is the representation's label.
Why There Are Two
The complexified algebra is $\mathfrak{sl}(2,\mathbb{C})\oplus\mathfrak{sl}(2,\mathbb{C})$, of rank $2$. The rank of a semisimple algebra is the number of independent Casimirs, so the count is two. That is the structural reason, and it is why the label is a pair and not a single number.
The Two Invariants
The First
$$ \mathcal{C}_1=\tfrac12\bigl(\vec K^{2}-\vec J^{2}\bigr), \qquad \vec J^{2}=\sum_{\eta}J_\eta J_\eta,\quad \vec K^{2}=\sum_\eta K_\eta K_\eta , $$
with the corpus's generators $J_\eta=\tfrac12\vec e_\eta$ and $K_\eta=\tfrac12 i\vec e_\eta$ of The Six Generators and Their Commutators. The menu writes it as $\tfrac12(\vec J^{2}-\vec K^{2})$, and that is the same operator written with the Hermitian generators $\mathcal{J}=i\vec J$, $\mathcal{K}=i\vec K$; §The Normalisation, Fixed Once gives the dictionary and the reason the two spellings are not interchangeable letter for letter.
That it commutes with all six generators is checked directly from the table: it is invariant under the rotations because it is a rotational scalar built from two vectors, and invariant under the boosts because the boost variation of $\vec K^{2}-\vec J^{2}$ is the cross term, which cancels.
The Second
$$ \mathcal{C}_2=\tfrac{i}{2}\bigl\{\vec J,\vec K\bigr\}, \qquad \bigl\{\vec J,\vec K\bigr\}=\sum_\eta\bigl(J_\eta K_\eta+K_\eta J_\eta\bigr), $$
and the menu's spelling $-\tfrac{i}{2}\{\vec J,\vec K\}$ is again the Hermitian-generator form.
The anticommutator is needed because $J_\eta$ and $K_\eta$ do not commute. The factor $i$ makes the operator Hermitian, since $J_\eta$ is anti-Hermitian, $K_\eta$ is Hermitian and $i$ compensates; the factor $\tfrac12$ makes its eigenvalues the integers and half-integers of the table below.
Neither Is the Other
$\mathcal{C}_1$ and $\mathcal{C}_2$ are algebraically independent: no polynomial in one equals the other. That independence is the rank statement of the previous section, and it is why the pair, and not either member, is the label.
The Normalisation, Fixed Once
The Two Generator Conventions
The corpus's generators are anti-Hermitian vectors, $J_\eta=\tfrac12\vec e_\eta$, and Hermitian ones, $K_\eta=\tfrac12 i\vec e_\eta$; the physicist's are all Hermitian, $\mathcal{J}_\eta=iJ_\eta$ and $\mathcal{K}_\eta=iK_\eta$. Because $\mathcal{J}^{2}=-J^{2}$ and $\mathcal{K}^{2}=-K^{2}$, each Casimir has two spellings and they differ by one sign:
$$ \mathcal{C}_1=\tfrac12\bigl(\vec K^{2}-\vec J^{2}\bigr)\ \text{(corpus)} =\tfrac12\bigl(\mathcal{J}^{2}-\mathcal{K}^{2}\bigr)\ \text{(physicist)}, $$
$$ \mathcal{C}_2=\tfrac{i}{2}\bigl\{\vec J,\vec K\bigr\}\ \text{(corpus)} =-\tfrac{i}{2}\bigl\{\mathcal{J},\mathcal{K}\bigr\}\ \text{(physicist)}, $$
and the two spellings of each line give the same numbers. The menu writes the physicist's spellings; this article writes the corpus's, because the numerical checks of the corpus are performed with the biquaternion generators.
The error to avoid is a mixed spelling. Substituting $\vec J\to i\mathcal{J}$ into $\tfrac12(\vec K^{2}-\vec J^{2})$ without also substituting $\vec K$ gives $-\tfrac12(\mathcal{J}^{2}-\mathcal{K}^{2})$, which is the negative of the Casimir: the defining representation would come out at $-\tfrac34$ instead of $+\tfrac34$, and the sign would look like a result rather than a mistake.
| convention | $\mathcal{C}_1$ | $\mathcal{C}_2$ | value on $(\tfrac12,\tfrac12)$ |
|---|---|---|---|
| corpus, $J$ anti-Hermitian, $K$ Hermitian | $\tfrac12(\vec K^{2}-\vec J^{2})$ | $\tfrac{i}{2}\{\vec J,\vec K\}$ | $(\tfrac32,\,0)$ |
| physicist, all Hermitian | $\tfrac12(\mathcal{J}^{2}-\mathcal{K}^{2})$ | $-\tfrac{i}{2}\{\mathcal{J},\mathcal{K}\}$ | $(\tfrac32,\,0)$ |
The last column is one pair of numbers and it is the same in both rows, which is the whole point of the table.
Why a Normalisation Must Be Fixed Once
A Casimir is defined up to an additive constant and a multiplicative factor, and both are usually chosen so that the eigenvalues come out as integers. This block fixes
$$ \mathcal{C}_1\ \text{with eigenvalue}\ m(m+1)+n(n+1),\qquad \mathcal{C}_2\ \text{with eigenvalue}\ m(m+1)-n(n+1), $$
and every other article quotes these integers. A different book will have $\mathcal{C}_1$ differing by a factor of $2$ or a constant, and no physics is lost, but a numerical comparison between books requires the dictionary. The dictionary is the table above and the sentence after it.
The Casimirs in Terms of the Two Triples
The Expansion
With $\vec J_\pm=\tfrac12(\vec J\pm i\vec K)$ of The Two Commuting Complex Triples,
$$ \vec J_+^{2}=\tfrac14\bigl(\vec J^{2}-\vec K^{2}\bigr)+\tfrac{i}{4}\bigl\{\vec J,\vec K\bigr\}, \qquad \vec J_-^{2}=\tfrac14\bigl(\vec J^{2}-\vec K^{2}\bigr)-\tfrac{i}{4}\bigl\{\vec J,\vec K\bigr\}. $$
Comparing with the two Casimirs in the corpus's spelling,
$$ \mathcal{C}_1=-\bigl(\vec J_+^{2}+\vec J_-^{2}\bigr),\qquad \mathcal{C}_2=\vec J_+^{2}-\vec J_-^{2}. $$
The minus sign in the first line is not a mistake and it is worth stating in words: with the corpus's anti-Hermitian generators the quadratic Casimir of a triple is negative, $\vec J_+^{2}=-m(m+1)$, because a product of two anti-Hermitian generators is negative semidefinite. With the physicist's generators it is positive, and then $\mathcal{C}_1=\vec J_+^{2}+\vec J_-^{2}$ as the plan states. The two statements are one statement and the sign travels with the generator.
Why the Form Matters
Any statement about $\mathcal{C}_1$ and $\mathcal{C}_2$ can be read as a statement about the two summands. "The pair labels the representation" becomes "one label for each summand". "The second Casimir is the chirality" becomes "the difference of the two spins". The rewriting is not decoration: it is what makes the next article's table a table of two independent half-integers.
The Eigenvalues on $(m,n)$
The Table
On the carrier $(m,n)$, where the first summand acts with spin $m$ and the second with spin $n$,
$$ \mathcal{C}_1=m(m+1)+n(n+1),\qquad \mathcal{C}_2=m(m+1)-n(n+1). $$
The first few carriers:
| $(m,n)$ | dimension | $\mathcal{C}_1$ | $\mathcal{C}_2$ |
|---|---|---|---|
| $(0,0)$ | $1$ | $0$ | $0$ |
| $(\tfrac12,0)$ | $2$ | $\tfrac34$ | $\tfrac34$ |
| $(0,\tfrac12)$ | $2$ | $\tfrac34$ | $-\tfrac34$ |
| $(\tfrac12,\tfrac12)$ | $4$ | $\tfrac32$ | $0$ |
| $(1,0)$ | $3$ | $2$ | $2$ |
| $(0,1)$ | $3$ | $2$ | $-2$ |
| $(1,\tfrac12)$ | $6$ | $\tfrac{11}{4}$ | $\tfrac54$ |
| $(1,1)$ | $9$ | $4$ | $0$ |
Why the Eigenvalues Are Real
Both Casimirs are Hermitian operators in the corpus's convention, so their eigenvalues are real and their eigenvectors can be chosen orthogonal.
- $\vec J^{2}$ is Hermitian because $J_\eta$ is anti-Hermitian, so $J_\eta^{2}$ is Hermitian and the sum of Hermitian operators is Hermitian.
- $\vec K^{2}$ is Hermitian for the same reason, since $K_\eta$ is Hermitian.
- $\{\vec J,\vec K\}$ is Hermitian because $J_iK_i+K_iJ_i$ is a sum of products whose adjoint is itself, and the factor $i$ in $\mathcal{C}_2$ makes the combination Hermitian.
So the two eigenvalues are real numbers, and the table's entries are those numbers. This is worth stating because a complex algebra's Casimirs could in principle take complex values on a complex carrier; they do not, and the reason is the real form's star.
The Check
The values are checked numerically on three carriers computed from the representation matrices: the defining $2\times2$ representation, where $\mathcal{C}_1=\mathcal{C}_2=\tfrac34$; the four-vector congruence representation, where $\mathcal{C}_1=\tfrac32$ and $\mathcal{C}_2=0$; and the eight-dimensional adjoint, where $\mathcal{C}_1=0$ on the two central directions and $\mathcal{C}_1=\pm2$ on the traceless part with $\mathcal{C}_2=\pm2$, the signs running together. All three agree with the table to machine precision.
What the Pair Decides
The First Casimir Alone Does Not Separate
$m(m+1)+n(n+1)$ is symmetric under $m\leftrightarrow n$. So
$$ \mathcal{C}_1(\tfrac12,0)=\mathcal{C}_1(0,\tfrac12)=\tfrac34,\qquad \mathcal{C}_1(1,0)=\mathcal{C}_1(0,1)=2,\qquad \mathcal{C}_1(\tfrac32,\tfrac12)=\mathcal{C}_1(\tfrac12,\tfrac32), $$
and in general every carrier is degenerate with its mirror image. A classification by $\mathcal{C}_1$ alone therefore names representations only up to parity of the labels, and that parity is exactly the chirality a physicist wants to know.
The Pair Does
$\mathcal{C}_2=m(m+1)-n(n+1)$ is antisymmetric, so the pair separates each carrier from its mirror. Conversely, given the pair $(c_1,c_2)$ one recovers
$$ m(m+1)=\tfrac12(c_1+c_2),\qquad n(n+1)=\tfrac12(c_1-c_2), $$
and each quadratic determines its non-negative half-integer. So the map from carriers to pairs of numbers is injective, and the Casimirs are a complete set of labels.
The Sign of the Second Casimir Is the Chirality
The two spinors are told apart by one sign:
$$ (\tfrac12,0):\ (\mathcal{C}_1,\mathcal{C}_2)=(\tfrac34,\tfrac34),\qquad (0,\tfrac12):\ (\mathcal{C}_1,\mathcal{C}_2)=(\tfrac34,-\tfrac34). $$
The first is the left-handed Weyl spinor and the second the right-handed one, in the convention where $(\tfrac12,0)$ is carried by left multiplication on the spinor module. The Spinor Representation of the Lorentz Group in Biquaternionic Form fixes that convention and The Spinor Module in Biquaternionic Form and Its Lorentz Action proves which side is which; the point here is only that the distinction is a sign and that a single Casimir cannot carry it.
A Worked Check
On $(\tfrac12,0)$ the direct computation with the $2\times2$ images gives $\mathcal{C}_1=\tfrac34$ and $\mathcal{C}_2=\tfrac34$, and on $(0,\tfrac12)$, which is the conjugate representation, $\mathcal{C}_1=\tfrac34$ and $\mathcal{C}_2=-\tfrac34$. The two computations differ in the sign of $K$'s action, which is what conjugating a representation does, and the sign of $\mathcal{C}_2$ follows it. Verified numerically on both carriers.
The Difference from the Poincaré Casimirs
The Poincaré Pair
The Poincaré algebra has two Casimirs as well,
$$ P^{2}=\sum_\mu\varepsilon_\mu P_\mu P_\mu,\qquad W^{2}=\sum_\mu\varepsilon_\mu W_\mu W_\mu, $$
with $\tilde W_\mu=\tfrac12\sum_{\nu,\sigma}\varepsilon_\nu\varepsilon_\sigma\,\epsilon_{\mu\nu\rho\sigma}\tilde P_\nu\tilde M_{\rho\sigma}$ the Pauli–Lubanski vector. On a massive one-particle space $\tilde P^{2}=m^{2}c^{2}$ and $\tilde W^{2}=-\hbar^{2}m^{2}c^{2}s(s+1)$, so the pair is the mass and the spin.
Why They Are Different Objects
The Poincaré Casimirs cannot be built from the Lorentz generators alone, because $P_\mu$ is not in the Lorentz algebra. They are Casimirs of a larger algebra, and their eigenvalues label orbits of a larger group. So:
| Lorentz Casimirs | Poincaré Casimirs | |
|---|---|---|
| built from | $J,K$ only | $P$ and $M$, with $W$ |
| eigenvalues | a pair of half-integers $(m,n)$ | a mass and a spin $(m,s)$ |
| carrier | finite-dimensional | infinite-dimensional, unitary |
| names | the field's chirality content | the particle |
What Each Label Means
The Lorentz pair names the field; the Poincaré pair names the particle. A field of type $(m,n)$ can be used to build a particle of some mass and spin, and which field builds which particle is a statement about the field equations and not about the algebra. Wigner's Classification: Two Casimirs, Three Orbits proves the particle side and states the honest boundary: the finite carriers $(1,0)$, $(0,1)$ and $(\tfrac12,\tfrac12)$ are not Wigner modules at all.
The Comparison That Is Wrong
A sentence such as "the Lorentz Casimirs are mass and spin" is wrong on both counts. Mass is not a Lorentz label, because the Lorentz group acts transitively on the mass hyperboloid's directions and not on its mass; and spin is not one of the two Lorentz labels, because the two labels are a pair of independent half-integers and the spin $s$ is a diagonal combination of them.
Why This Makes the Table a Classification
Three properties together.
- Completeness. The two Casimirs separate every carrier, as the invertibility above shows.
- Invariance. They are constant on each irreducible representation, so no two carriers can share a pair unless they are the same carrier.
- Constructivity. Each carrier can be built, and its dimension is $(2m+1)(2n+1)$, so the classification is effective and not merely a labelling.
A list of representations would be a list of matrices someone wrote down. A classification is a set of invariants that decides membership and equality. The two Casimirs are what move the $(m,n)$ table from the first to the second, and that is why the table of The Finite-Dimensional Representations and Their Unitarity Defect is a theorem rather than an inventory.
Summary
The Lorentz algebra has exactly two independent Casimirs, $\mathcal{C}_1=\tfrac12(\vec J^{2}-\vec K^{2})$ and $\mathcal{C}_2=-\tfrac{i}{2}\{\vec J,\vec K\}$, written here in the Hermitian convention and equal in the corpus's anti-Hermitian one, where they read $\tfrac12(\vec K^{2}-\vec J^{2})$ and $\tfrac{i}{2}\{\vec J,\vec K\}$ and give the same numbers. They are the sum and the difference of the quadratic Casimirs of the two commuting triples, $\mathcal{C}_1=-(\vec J_+^{2}+\vec J_-^{2})$ and $\mathcal{C}_2=\vec J_+^{2}-\vec J_-^{2}$. On the carrier $(m,n)$ their eigenvalues are $m(m+1)+n(n+1)$ and $m(m+1)-n(n+1)$, both real because both operators are Hermitian. The first is symmetric and cannot separate a carrier from its mirror; the second is antisymmetric and its sign is the chirality, so $(\tfrac12,0)$ carries $(\tfrac34,\tfrac34)$ and $(0,\tfrac12)$ carries $(\tfrac34,-\tfrac34)$. These are not the Poincaré Casimirs of mass and spin, which live in a larger algebra and label particles rather than fields. Together the pair makes the $(m,n)$ table a classification.
Summary of Notation
| symbol | meaning |
|---|---|
| $\mathcal{C}_1=\tfrac12(\vec K^{2}-\vec J^{2})$ | the first Casimir, corpus spelling |
| $\mathcal{C}_2=\tfrac{i}{2}\{\vec J,\vec K\}$ | the second Casimir, corpus spelling |
| $\vec J_\pm=\tfrac12(\vec J\pm i\vec K)$ | the two complex triples |
| $(m,n)$ | the pair of half-integers labelling a carrier |
| $C_1=P^{2}$, $C_2=-W^{2}$ | the Poincaré Casimirs, written with a plain $C$ to keep them apart from $\mathcal{C}_{1,2}$ |
| $\mathcal{J}=i\vec J$, $\mathcal{K}=i\vec K$ | the Hermitian generators |
Further Reading
- B. C. Hall, Lie Groups, Lie Algebras, and Representations, 2nd ed., Springer, 2015 — the universal enveloping algebra, Casimir elements and their role in the classification.
- W. Fulton and J. Harris, Representation Theory: A First Course, Springer, 1991 — the Casimirs of $\mathfrak{sl}_2$ and of a direct sum, and highest-weight labelling.
- S. Weinberg, The Quantum Theory of Fields, Vol. 1, Cambridge University Press, 1995, §5.6 — the two invariants $\vec J^{2}-\vec K^{2}$ and $\vec J\cdot\vec K$, the $(A,B)$ labels, and the eigenvalues $A(A+1)+B(B+1)$.
- E. P. Wigner, "On Unitary Representations of the Inhomogeneous Lorentz Group", Annals of Mathematics 40 (1939) 149–204 — the Poincaré Casimirs and the classification by mass and spin.
- Wigner's Classification: Two Casimirs, Three Orbits — the particle side, and the boundary between the two classifications.
- The Finite-Dimensional Representations and Their Unitarity Defect — the table this article makes into a classification.