Skip to content

Why does the XXX chain possess a whole family of commuting operators? The starting point is a local identity on just three two-dimensional spaces. You will prove that identity for the rational R-matrix, identify the spectral differences that make it work, and test its exceptional parameters. This opens the algebraic sequence leading to commuting charges, the periodic-chain Hamiltonian and regular Bethe states; it needs tensor products, not prior representation theory or a Bethe completeness theorem.

Required background. Act with matrices on tensor products and compose operators from right to left; the linear algebra bridge develops these operations. Construct a small spin-chain Hamiltonian applies them to physical spin operators and the spin permutation.

Helpful background. The XXX course gives the coordinate-wave viewpoint. The R-matrix convention reference distinguishes the related operator normalizations.

Let V=C2V=\mathbb C^2 with basis ∣↑⟩,∣↓⟩|\uparrow\rangle,|\downarrow\rangle. All tensor products here are ordinary, with no signs introduced by exchanging factors. On V⊗VV\otimes V, the permutation PP is defined by

P(v⊗w)=w⊗v,P2=I.P(v\otimes w)=w\otimes v, \qquad P^2=I.

On V1⊗V2⊗V3V_1\otimes V_2\otimes V_3, the subscript tells us which two factors are exchanged. Thus

P13(v1⊗v2⊗v3)=v3⊗v2⊗v1.P_{13}(v_1\otimes v_2\otimes v_3) =v_3\otimes v_2\otimes v_1.

The labels name tensor factors, not matrix entries. Put A=P12A=P_{12}, B=P13B=P_{13}, C=P23C=P_{23}. Acting on an arbitrary elementary tensor proves

AB=BC=CA,BA=CB=AC,ABC=CBA=B.AB=BC=CA,\qquad BA=CB=AC,\qquad ABC=CBA=B.

For example, ABAB first exchanges factors 1,31,3, then 1,21,2, sending the triple (v1,v2,v3)(v_1,v_2,v_3) to (v2,v3,v1)(v_2,v_3,v_1). The products BCBC and CACA do the same. Equality on elementary tensors implies equality on their linear span, hence on the full tensor product. These are the permutation relations used by Faddeev 1996, § 3, equations (33)–(39), PDF.

Do P12P_{12} and P23P_{23} commute? Apply both orders to ∣↓↑↑⟩|\downarrow\uparrow\uparrow\rangle.

Repair. P12P23P_{12}P_{23} gives ∣↑↓↑⟩|\uparrow\downarrow\uparrow\rangle, whereas P23P12P_{23}P_{12} gives ∣↑↑↓⟩|\uparrow\uparrow\downarrow\rangle. They share a tensor factor and do not commute. Operators on disjoint pairs, such as P12P_{12} and P34P_{34} on four factors, do commute. A drawing of crossing lines does not replace this ordering check.

For a dimensionless complex spectral parameter uu, define

R(u)=uI+iP.R(u)=uI+iP.

In the ordered basis ↑↑,↑↓,↓↑,↓↓\uparrow\uparrow,\uparrow\downarrow,\downarrow\uparrow,\downarrow\downarrow,

R(u)=(u+i0000ui00iu0000u+i).R(u)= \begin{pmatrix} u+i&0&0&0\\ 0&u&i&0\\ 0&i&u&0\\ 0&0&0&u+i \end{pmatrix}.

The relation to verify is

R12(a)R13(a+b)R23(b)=R23(b)R13(a+b)R12(a).\begin{aligned} &R_{12}(a)R_{13}(a+b)R_{23}(b)\\ &\hspace{1.5em}=R_{23}(b)R_{13}(a+b)R_{12}(a). \end{aligned}

Here R12R_{12}, for example, acts as the identity on factor 33. The parameters can also be written a=λ−μa=\lambda-\mu, b=μ−νb=\mu-\nu, and a+b=λ−νa+b=\lambda-\nu. The middle argument is constrained by these differences.

To see exactly where the constraint matters, keep the middle argument independent and call it zz. Expand

D(a,z,b)=(aI+iA)(zI+iB)(bI+iC)−(bI+iC)(zI+iB)(aI+iA).\begin{aligned} \mathcal D(a,z,b)={}&(aI+iA)(zI+iB)(bI+iC)\\ &-(bI+iC)(zI+iB)(aI+iA). \end{aligned}

The terms with zero or one permutation cancel immediately. The three-permutation terms cancel because ABC=CBAABC=CBA. The two-permutation terms are

−bAB−zAC−aBC+aCB+zCA+bBA.-bAB-zAC-aBC+aCB+zCA+bBA.

Use AB=BC=CAAB=BC=CA and BA=CB=ACBA=CB=AC to obtain

D(a,z,b)=(z−a−b)(AB−BA).\mathcal D(a,z,b)=(z-a-b)(AB-BA).

Setting z=a+bz=a+b proves the Yang–Baxter identity for every complex a,ba,b, including values at which an individual R-matrix is singular. No inverse was used. The Library proof connects this local identity to the chain-wide exchange relation.

The symmetric and antisymmetric subspaces of V⊗VV\otimes V have projectors

Π+=I+P2,Π−=I−P2.\Pi_+=\frac{I+P}{2},\qquad \Pi_-=\frac{I-P}{2}.

They are respectively the three-dimensional triplet and one-dimensional singlet spaces. Therefore

R(u)=(u+i)Π++(u−i)Π−.R(u)=(u+i)\Pi_++(u-i)\Pi_-.

At u=0u=0, R(0)=iPR(0)=iP is an invertible permutation times a scalar. This property is called regularity and will turn a transfer matrix into translation. At u=iu=i, the singlet eigenvalue vanishes; at u=−iu=-i, the triplet eigenvalue vanishes. Away from those two points,

R(u)−1=uI−iPu2+1.R(u)^{-1}=\frac{uI-iP}{u^2+1}.

For a quick worked check, at u=2u=2 the two eigenvalues are 2+i2+i and 2−i2-i. The inverse acts with their reciprocals. The matrix itself is not unitary: R(2)†R(2)=5IR(2)^\dagger R(2)=5I. Singularity, regularity and unitarity are distinct statements.

This R-matrix is an auxiliary algebraic object. It is not the two-magnon amplitude ratio S12S_{12} from the coordinate lesson, and its complex spectral parameter is not automatically a physical lattice momentum.

Guided practice: complete a permutation proof

Section titled “Guided practice: complete a permutation proof”

Apply BCBC and CACA to (v1,v2,v3)(v_1,v_2,v_3) and compare with the stated action of ABAB. Then use B2=IB^2=I or a direct action to establish ABC=BABC=B.

Hint

Keep the rightmost operation first. For BCBC, exchange 2,32,3 before exchanging 1,31,3.

Solution

The products act as

BC:(v1,v2,v3)↦(v1,v3,v2)↦(v2,v3,v1),CA:(v1,v2,v3)↦(v2,v1,v3)↦(v2,v3,v1).\begin{aligned} BC:(v_1,v_2,v_3)&\mapsto(v_1,v_3,v_2)\mapsto(v_2,v_3,v_1),\\ CA:(v_1,v_2,v_3)&\mapsto(v_2,v_1,v_3)\mapsto(v_2,v_3,v_1). \end{aligned}

Thus both equal ABAB. For ABCABC, first CC sends the triple to (v1,v3,v2)(v_1,v_3,v_2), then ABAB sends it to (v3,v2,v1)(v_3,v_2,v_1), which is exactly BB. Reversing every factor order similarly gives CBA=BCBA=B. These identities hold for arbitrary vectors, not only spin-basis examples.

Independent practice: reject a false spectral assignment

Section titled “Independent practice: reject a false spectral assignment”

Choose a=1a=1, b=2b=2, but use middle argument z=0z=0. Compute D(1,0,2)∣↑↓↑⟩\mathcal D(1,0,2)|\uparrow\downarrow\uparrow\rangle without multiplying three 8×88\times8 matrices. Does checking the Yang–Baxter equation only on ∣↑↑↑⟩|\uparrow\uparrow\uparrow\rangle detect this error?

Hint

Use the derived factor (z−a−b)(AB−BA)(z-a-b)(AB-BA). Every permutation acts identically on the all-up vector.

Solution

Here z−a−b=−3z-a-b=-3, while

(AB−BA)∣↑↓↑⟩=∣↓↑↑⟩−∣↑↑↓⟩.(AB-BA)|\uparrow\downarrow\uparrow\rangle =|\downarrow\uparrow\uparrow\rangle-|\uparrow\uparrow\downarrow\rangle.

The residual is therefore −3-3 times this nonzero vector, with norm 323\sqrt2. On the all-up vector, AB−BAAB-BA vanishes, so that single test misses the incorrect parameter relation. An identity must hold on the whole space; a highly symmetric vector is a weak negative control.

Transfer: normalize a real-parameter R-matrix

Section titled “Transfer: normalize a real-parameter R-matrix”

For real uu, show that R^(u)=R(u)/u2+1\widehat R(u)=R(u)/\sqrt{u^2+1} is unitary. Does multiplying R(u)R(u) by a scalar function preserve the Yang–Baxter identity? Distinguish this statement from invertibility at every complex parameter.

Hint

Use P†=PP^\dagger=P and P2=IP^2=I. Count the three scalar factors on each side of the identity.

Solution

For real uu, R(u)†=uI−iPR(u)^\dagger=uI-iP, so R(u)†R(u)=(u2+1)IR(u)^\dagger R(u)=(u^2+1)I and R^†R^=I\widehat R^\dagger\widehat R=I. Replacing R(u)R(u) by f(u)R(u)f(u)R(u) multiplies both sides of Yang–Baxter by the same scalar f(a)f(a+b)f(b)f(a)f(a+b)f(b), wherever these functions are defined. It preserves the identity without requiring cancellation of that factor.

This does not guarantee an invertible matrix: zeros of ff can introduce new singularities, and the unscaled RR is already singular at u=±iu=\pm i. The real positive square root used above also does not specify a globally single-valued complex normalization. When constructing a Hamiltonian, a parameter-dependent normalization changes its logarithmic derivative and must be tracked.

You have proved the rational Yang–Baxter identity, tested a nontrivial failure and identified its singular parameters. Continue to Build monodromy and transfer matrices to combine one auxiliary space with every spin in a finite ring.

  • Faddeev, L. D. How Algebraic Bethe Ansatz works for integrable model. Les Houches lectures, arXiv:hep-th/9605187v1, 1996, 59 pp. Version record. Open PDF. Section and equation numbers identify the cited locations in this version.