---
title: Poincaré–Carrollian Intertwiner
url: https://www.emergentmind.com/topics/poincare-carrollian-intertwiner
type: topic
---

# Poincaré–Carrollian Intertwiner

The **Poincaré–Carrollian intertwiner** denotes a class of maps that send Poincaré-covariant data to Carrollian data while preserving the relevant algebraic action. In current usage, the term includes an Inönü–Wigner contraction of Lie algebras, a rigging-induced projection from ambient Lorentzian geometry to intrinsic Carrollian geometry on null hypersurfaces, Fourier or Penrose transforms from bulk amplitudes or twistor representatives to Carrollian boundary observables, BRST-induced maps from worldsheet cohomology to Carroll group representations, and, in string theory, explicit operators exchanging compactified Poincaré and Carrollian sectors [2510.21651] [2402.04120] [2501.11011] [2606.05401]. The shared criterion is an intertwining relation such as
\[
T\,\rho_{\mathrm{Poincar\acute e}}(X)=\rho_{\mathrm{Carroll}}(\phi(X))\,T,
\]
or, in superalgebraic form, preservation of brackets and anticommutators.

## 1. Algebraic foundation: contraction, conformal extension, and BMS relation

At the algebraic level, the standard starting point is the ultra-relativistic contraction of the Poincaré algebra. In the review on Carrollian geometry, the Carroll algebra in \(d+1\) dimensions is generated by rotations \(J_{ij}\), boosts \(K_i\), spatial translations \(P_i\), and time translations \(P_0\), with brackets
\[
[K_i,P_j]=i\,\delta_{ij}P_0,\qquad [K_i,P_0]=0,
\]
while the corresponding Poincaré bracket is
\[
[K_i,P_0]=i\,P_i.
\]
This is the characteristic ultra-local limit in which boosts cease to mix time and space translations. The same review gives the group-level contraction: from finite Poincaré transformations, the Carrollian limit rescales \(t=Cx_0\), \(\vec v=C\vec\beta\), \(b=Ca_0\) with \(C\to\infty\), producing
\[
t'=t+\vec v\cdot R\vec x+b,\qquad \vec x'=R\vec x+\vec a.
\]
It formulates the contraction as an algebraic intertwiner \(\phi\colon \mathfrak{iso}(1,d)\to \mathfrak{carr}(d+1)\) [2510.21651].

A parallel construction appears in conformal field theory. In the ultra-relativistic scaling \(x^i\to x^i\), \(t\to \epsilon t\), \(\epsilon\to0\), the relativistic conformal generators contract to the finite conformal Carrollian algebra generated by
\[
H=\partial_t,\qquad P_i=\partial_i,\qquad B_i=x_i\partial_t,
\]
together with \(D\), \(K\), and \(K_i\). In \(d=4\), this algebra admits an infinite enhancement by supertranslations \(M_f=f(x^i)\partial_t\); the same work emphasizes that the conformal Carrollian groups are isomorphic to Bondi–Metzner–Sachs groups and presents the contraction pair \((\phi,T_\epsilon)\), where \(T_\epsilon\) rescales coordinates and fields so that
\[
T_\epsilon \rho_P(X)=\rho_C(\phi(X)) T_\epsilon
\]
in the limit \(\epsilon\to0\) [1901.10147].

The conformal extension is also tied directly to BMS. For a flat Carrollian structure, the conformal Carroll generators include an infinite-dimensional supertranslation sector \(M_f=f(x^i)\partial_t\), and on a \(3\)-dimensional Carroll manifold with \(2\)-sphere spatial metric one has
\[
\mathfrak{ccarr}_{2}(3)=\mathfrak{sl}(2,\mathbb{C})\ltimes \mathbb{R}^S\cong \mathfrak{bms}_4.
\]
This algebraic identification explains why many intertwiners are simultaneously Poincaré–Carrollian and Poincaré–BMS maps [2510.21651].

## 2. Geometric realizations on null hypersurfaces and timelike infinity

A second major usage is geometric. In the null-hypersurface framework, a Carrollian manifold is a \((d+1)\)-dimensional manifold equipped with a degenerate metric \(q_{ab}\) and a nowhere-vanishing kernel vector field \(\ell^a\), satisfying
\[
q_{ab}\ell^b=0,\qquad \ell^a\neq0.
\]
The review on Carrollian geometry defines a geometric intertwiner \(T\) as restriction followed by projection with the rigged projector
\[
\Pi_\mu{}^\nu=\delta_\mu{}^\nu-n_\mu k^\nu,
\]
where \(n_\mu\) is the null normal and \(k^\mu\) a null rigging vector. The rigging-induced connection
\[
D_a T_b{}^c=\Pi_a{}^\mu \Pi_b{}^\nu (\nabla_\mu T_\nu{}^\rho)\Pi_\rho{}^c
\]
is shown to coincide with the preferred intrinsic Carrollian connection. In this setting the intertwining statement is
\[
T\big(\rho_P(X)\cdot \text{bulk data}\big)=\rho_C\big(\phi(X)\big)\cdot T(\text{bulk data}),
\]
for ambient symmetries preserving the hypersurface. The same framework projects Gauss and Codazzi–Mainardi relations and rewrites Einstein’s equations as conservation of the null Brown–York stress tensor
\[
T_a{}^b=\frac{1}{8\pi G}(W_a{}^b-W\,\delta_a{}^b),\qquad D_b T_a{}^b=\frac{1}{8\pi G}\Pi_a{}^\alpha G_{\alpha\beta}\ell^\beta
\]
[2510.21651].

A related, but distinct, geometric intertwiner appears at timelike infinity \(\mathsf{Ti}\). There, \(\mathsf{Ti}_{d+1}\) is realized as \(\mathbb{R}\times H^d\) with Carrollian structure
\[
n=\partial_\tau,\qquad q=0\cdot d\tau^2+h_{ab}(x)\,dx^a dx^b.
\]
Its isometries form the Poincaré group, and massive spin-\(s\) Carrollian fields are constructed by induced representations from the isotropy subgroup \(\mathrm{ISO}(d)\). The field is defined by
\[
\phi(\tau,x)=U(\tau,x)\phi(0)U(\tau,x)^{-1},\qquad U(\tau,x)=e^{x^a B_a}e^{\tau H},
\]
and the quadratic Casimir acts as
\[
[\mathcal C_2,\phi(x)]=-\partial_\tau^2\phi(x).
\]
Imposing the massive irreducibility condition yields
\[
(\partial_\tau^2+m^2)\phi_{a_1\ldots a_s}(\tau,x)=0.
\]
The corresponding Poincaré–Carrollian intertwiner \(I_s\) maps bulk massive spin-\(s\) fields in Minkowski space to Carrollian spin-\(s\) fields on \(\mathsf{Ti}\), both asymptotically and through an explicit momentum-space kernel, and satisfies
\[
I_s\,\rho_{\mathrm{Poincar\acute e}}(g)=\rho_{\mathrm{Carroll}}(g)\,I_s
\]
[2402.05190].

These two geometric constructions differ in causal setting—null hypersurfaces versus timelike infinity—but both realize the same idea: bulk Lorentzian data are projected to Carrollian data without losing the relevant symmetry action. This suggests that the intertwiner is best regarded as a structural bridge rather than a single canonical operator.

## 3. Boundary transforms from scattering amplitudes

In amplitude theory, the intertwiner is an explicit integral transform from momentum-space scattering data to Carrollian correlators at null infinity. One formulation considers a massless scalar in \(4\)D Minkowski space with asymptotic field \(\Sigma(u,\Omega)\). If \(\mathcal M\) denotes the amputated connected momentum-space scattering matrix element, the Carrollian amplitude is
\[
\Big\langle \prod_{j=1}^{m+n}\Sigma_j(u_j,\Omega_j;\sigma_j)\Big\rangle
=\Big(\frac{1}{8\pi^2 i}\Big)^{m+n}\prod_{j=1}^{m+n}\int_0^\infty d\omega_j\,e^{-i\sigma_j\omega_j u_j}\,
(2\pi)^4\delta^{(4)}\!\Big(\sum_j p_j\Big)\,i\mathcal M.
\]
The same construction introduces bulk-to-boundary kernels
\[
D_\sigma(u,\Omega;x)= -\frac{\sigma}{8\pi^2\,(u+n\!\cdot\! x-i\sigma\epsilon)},\qquad
K_\sigma(u,\Omega;x)=\frac{i\sigma}{4\pi^2\,(u+n\!\cdot\! x-i\sigma\epsilon)^2},
\]
which implement the Fourier map on each external leg. In this setting the Fourier transform operator \(\mathcal F\) intertwines the Poincaré representation on \(\mathcal M\) with the Carrollian diffeomorphism representation on the boundary amplitude \(A\) [2402.04120].

A closely related construction defines Carrollian amplitudes as
\[
C_n=\prod_{i=1}^n\int_0^\infty d\omega_i\,2\pi\,e^{i\epsilon_i\omega_i u_i}\,A_n(\{\omega_i,z_i,\bar z_i\}),
\]
with first descendants
\[
\widetilde C_n=\Big(\prod_i\partial_{u_i}\Big)C_n.
\]
In that framework the Fourier kernel is the “basic intertwining” map between bulk momentum-space amplitudes and position-space insertions at \(\mathscr I\). It yields the global conformal Carrollian Ward identities
\[
\sum_i\Big[\Big(T+\frac{u_i}{2}(\partial_{z_i}Y+\partial_{\bar z_i}\bar Y)\Big)\partial_{u_i}
+Y\partial_{z_i}+\bar Y\partial_{\bar z_i}
+\frac{1+\epsilon_iJ_i}{2}\partial_{z_i}Y+\frac{1-\epsilon_iJ_i}{2}\partial_{\bar z_i}\bar Y\Big]C_n=0.
\]
The same work shows that the B-transform maps Carrollian correlators to celestial amplitudes, thereby placing the Carrollian transform between the usual momentum basis and the celestial Mellin basis [2312.10138].

In a more recent \(D=3\) scalar construction, the intertwiner is defined directly as a map from bulk on-shell operators or amplitudes to boundary Carrollian primaries \(O_{h,\xi}(z,u)\). For the massless case, the Mellin–Laplace kernel is
\[
K^s_{h,0}(z,u;\omega',z')=\omega'^{\,h-1}\,e^{-2is\omega' u}\,\delta(z'-z),
\]
and for amplitudes one obtains
\[
\Big\langle\prod_{i=1}^n O_{h_i,\xi_i}(z_i,u_i)\Big\rangle
=\prod_{i=1}^n\int d\mu(\lambda'_i)\,K_{h_i,\xi_i}(z_i,u_i;\lambda'_i)\,\mathcal A_n(\{\lambda'_i\}).
\]
For massive particles, the kernel requires a complex-support delta distribution,
\[
K^{M,s}_{1,\eta}(z,u;y',z')=\frac{m}{2}\,\delta_{\mathbb C}\!\big(z'-z-\eta i y'\big)\,e^{-ismy'^{-1}u},
\]
whereas for tachyonic unitary representations one finds
\[
K^{T}_{1,\eta i m}(z,u;y',z')=\frac{m}{2}\,\delta\!\big(z'-z+\eta y'\big)\,e^{-imy'^{-1}u}.
\]
The paper formulates this analytically as “real mass is imaginary”: massive kernels require a complex momentum shift, while tachyonic kernels have real support [2606.05401].

## 4. Twistor and Penrose formulations

Twistor theory provides another exact realization. In the twistor-space construction of the loop algebra \(Lw_{1+\infty}\), generators are holomorphic Hamiltonians \(g(Z,\bar Z)\) of homogeneous degree \(2\) in twistor coordinates, acting by the holomorphic Poisson bracket
\[
\delta_g\mathbf f=\{g,\mathbf f\}.
\]
The preferred twistor lift \(\mathbf T^s\) sends a Carrollian field \(\phi\) at null infinity to a twistor representative \(\mathbf f\), and the Penrose transform \(\mathcal P\), followed by the large-\(r\) limit \(L\), returns the corresponding boundary field. The boundary intertwiner is
\[
\mathcal I(g)\cdot\phi:=L\circ\mathcal P\circ \mathrm{ad}_g\circ \mathbf T^s(\phi),
\]
and satisfies
\[
[\mathcal I(g_1),\mathcal I(g_2)]\,\phi=\mathcal I(\{g_1,g_2\})\,\phi.
\]
For \(n=1\) it yields supertranslations; restricting to \(l\le1\) recovers ordinary Poincaré translations. For \(n=2\), globally holomorphic generators reproduce the Lorentz \(\mathrm{SL}(2,\mathbb C)\) action on the sphere, while Laurent coefficients generate superrotations [2402.00688].

A related twistor-to-Carrollian bridge appears in the “third-Fourier transform” of asymptotic radiation data. For negative helicity,
\[
\Psi_n^0(u,\lambda,\tilde\lambda)= i\int ds\,\hat\phi(\lambda,s\tilde\lambda)\,e^{-isu},
\]
and for positive helicity,
\[
\widetilde\Psi_n^0(u,\lambda,\tilde\lambda)= i\int ds\,\hat\phi(s\lambda,\tilde\lambda)\,e^{-isu}.
\]
Combined with the half-Fourier transform to twistor space, this yields a Lorentz-covariant map between momentum-space wavefunctions, twistor data, and Carrollian operators at null infinity. The same work verifies the intertwining property explicitly for tree-level MHV sectors and shows that the induced Carrollian operators carry the global conformal Carrollian action on \(\mathscr I\) [2312.10138].

The twistor and amplitude constructions should not be conflated. The twistor-side map is cohomological and Hamiltonian, whereas the amplitude-side map is integral-transform based. What they share is preservation of the Poincaré action after passing to Carrollian boundary data.

## 5. String-theoretic realizations

String theory supplies two especially concrete realizations. In the homogeneous RNS Carrollian superstring studied by Chen and Hu, the target spacetime is split into Poincaré directions \(X^a\) and Carrollian spatial directions \(X^i\), with generalized Carrollian boost
\[
\delta X^a=\Lambda^a{}_i\,X^i,\qquad \delta X^i=0.
\]
For two compactifications—one along a Poincaré direction and one along a Carrollian direction—the mode algebra is the same super-\(\mathrm{BMS}_3\) algebra. The paper defines an operator \(U\) on the flipped-vacuum Hilbert space by
\[
U\,A_0\,U^{-1}=B_0,\qquad U\,B_0\,U^{-1}=A_0,
\]
together with
\[
(R_T,k_T,w_T)\mapsto \Big(R_C=\frac{1}{2R_T},\,k_C=w_T,\,w_C=k_T\Big),
\]
while leaving nonzero oscillators and homogeneous fermions invariant. Because \(L_n\), \(M_n\), \(H_r\), and \(\tilde H_r\) are preserved,
\[
U\,\mathfrak{bms}_3^{(P)}\,U^{-1}=\mathfrak{bms}_3^{(C)}.
\]
This is an explicit Poincaré–Carrollian intertwiner implementing T-duality between the compactified Poincaré-sector and Carrollian-sector homogeneous superstrings [2501.11011].

Quantum Carrollian bosonic strings furnish a different mechanism. After fixing the Carrollian analogue of conformal gauge, the residual symmetry is the three-dimensional extended \(\mathrm{BMS}_3\) algebra. The matter realization uses \(D+1\) bosonic \(BC\) systems \((X^\mu,\Pi_\mu)\) of weights \((0,1)\), with
\[
T(z)=:X^\mu\Pi_\mu:(z),\qquad
M(z)=\frac12:\Pi_0\Pi_0:(z)-\frac{\tau^2}{2}\delta_{ij}:X^iX^j:(z),
\]
yielding \(c_L=2(D+1)\) and \(c_M=0\). Nilpotent BRST quantization requires \(c_L=52\), hence \(D=25\). The BRST cohomology is finite-dimensional at fixed momentum, vanishes unless \(p_0=0\), and obeys Poincaré duality. For \(p=(0,\mathbf p)\neq0\), the absolute cohomology transforms under the little group \(\mathrm{SO}(24)\); for \(p=0\), under \(\mathrm{SO}(25)\). The paper then defines an intertwiner
\[
\Phi_\tau\colon H^n(p)\to L^2\text{-sections over }O_\tau
\]
by sending a cohomology class \([\Psi]\) to the induced section
\[
s_\Psi(R\cdot\tau):=U(R)[\Psi].
\]
Translations act by phases \(e^{i\langle \tau,a\rangle}\), rotations act on the fiber, and boosts act trivially in the induced unitary irreducible representations considered. The intertwiner property
\[
\Phi_\tau(X\cdot[\Psi])=\rho(X)\cdot\Phi_\tau([\Psi])
\]
identifies the BRST/BMS module with Carroll group representation theory [2509.04397].

The string-theoretic literature therefore uses the same term for two rather different objects: a T-duality-like operator \(U\) preserving super-\(\mathrm{BMS}_3\) in the superstring, and a BRST-to-UIR induction map \(\Phi_\tau\) in the bosonic theory. Both are exact in the sense that they preserve the relevant algebraic structure.

## 6. Supergeometric generalizations, obstructions, and conceptual limits

The supergeometric extension makes clear that an intertwiner need not exist globally. On the Carrollian superplane \(\Pi\mathbb S\simeq \mathbb R^{2|4}\), the even base is \(M\simeq\mathbb R^2\) with coordinates \((t,x)\), and the superplane is a principal \(\mathbb R^{1|2}\)-bundle with fiber coordinates \((t,\zeta^i)\). Its degenerate metric is
\[
g:=\delta x\otimes \delta x \pm 2\,\delta\eta^1\otimes \delta\eta^2,
\]
with kernel
\[
\ker(g)=\mathrm{Span}(\partial_t,\partial_{\zeta^1},\partial_{\zeta^2}).
\]
Carroll spinors are constructed as sections of a degenerate Clifford module, with the nilpotent Carrollian boost generator
\[
S_{tx}=\frac14[\theta,e_x]=\frac12\,\theta e_x,\qquad (S_{tx})^2=0.
\]
Once a principal Carrollian connection and a basic odd one-form \(\Psi\) are chosen, the odd vector fields
\[
Q_i=\nabla_{\eta^i}+\Psi_i\partial_t
\]
define an \(N=2\) Carrollian supersymmetry [2603.21677].

In that setting the paper defines an intertwiner abstractly as a linear map \(I\) between superalgebras preserving brackets. It then shows that when clocks are closed and the basic one-form is constant, the resulting Carrollian \(N=2\) superalgebra coincides with the Inönü–Wigner contraction of the Poincaré superalgebra; the contraction map acts as an intertwiner on the constant-coefficient subalgebra. But for the full geometric supersymmetry one has
\[
\{Q_i,Q_j\}=2\Psi_{(ij)}(x)\,\partial_t
\]
and, if the connection has odd dependence, additional vertical odd terms proportional to \(\partial_{\zeta^k}\). Because the Poincaré superalgebra has constant structure constants and no coordinate-dependent brackets, “there exists no linear intertwiner” from the rigid Poincaré superalgebra to the full Lie–Rinehart superpair. The paper therefore identifies a precise obstruction: the intertwiner exists only on the constant-coefficient subfamily [2603.21677].

Other limitations appear in non-supersymmetric contexts as well. The ultra-relativistic contraction of relativistic field theories is sector-dependent: distinct field rescalings lead to different Carrollian sectors, and some sectors are explicitly described as “non-nice” because the resulting equations of motion lose kinetic terms [1901.10147]. In the \(D=3\) scalar conformal-basis construction, bulk unitarity gives
\[
L_n^\dagger=-L_n,\qquad P_n^\dagger=-P_n,
\]
whereas boundary radial quantization gives
\[
L_n^\dagger=L_{-n},\qquad P_n^\dagger=P_{-n}.
\]
That construction therefore states explicitly that the intertwiner matches algebra actions, not Hilbert-space adjoint structures [2606.05401].

These caveats rule out a common misconception: the Poincaré–Carrollian intertwiner is not, in general, a unique or universal isometric equivalence. Depending on context, it may be a contraction, a projection, a Fourier kernel, a Penrose transform, a BRST induction map, or a T-duality operator; it may preserve full representations, only a subalgebra, or only classical equations of motion. What unifies these constructions is the controlled transport of symmetry from a Poincaré description to a Carrollian one.

Source: https://www.emergentmind.com/topics/poincare-carrollian-intertwiner