---
title: Complex Circle Manifold (CCM)
url: https://www.emergentmind.com/topics/complex-circle-manifold-ccm
type: topic
---

# Complex Circle Manifold (CCM)

Complex Circle Manifold (CCM) is a context-dependent term rather than a single universally fixed mathematical object. In current arXiv usage it denotes at least three distinct constructions: a compact, connected almost complex manifold endowed with a \(J\)-preserving \(S^1\)-action with isolated fixed points; a compact complex manifold obtained from a product of principal circle bundles through a holomorphic \(\mathbb{C}\)-bundle quotient; and the product of unit circles \(\mathcal{M}=\{z\in\mathbb{C}^n: |z_i|=1\ \forall i\}\), used as a Riemannian manifold for constant-modulus optimization [1411.6458] [1012.0668] [2508.07396]. The common motif is circular symmetry, but the underlying geometry, topology, and analytic role differ substantially across these literatures.

## 1. Terminological scope and principal usages

The non-uniformity of the term is explicit in the literature. In almost complex geometry, CCM refers to a manifold carrying a circle action; in complex-analytic geometry, it refers to a manifold built from circle bundles; in optimization, it denotes the feasible set of unit-modulus complex variables.

| Usage | Defining object | Typical context |
|---|---|---|
| Almost complex CCM | Compact, connected almost complex \(2n\)-manifold with a \(J\)-preserving \(S^1\)-action and isolated fixed points | Localization, Chern numbers, Hamiltonianity |
| Complex-analytic CCM | \(S(L_1)\times S(L_2)\) endowed with a complex structure induced from a holomorphic \(\mathbb{C}\)-action on \(L_1\times L_2\) | Non-Kähler compact complex manifolds |
| Optimization CCM | \(\{z\in\mathbb{C}^n: |z_i|=1\ \forall i\}\), or matrix analogue \(\{Z\in\mathbb{C}^{n\times m}: |Z_{ij}|=1\}\) | Constant-modulus waveform, beamforming, and IRS design |

This terminological divergence is not merely cosmetic. In the optimization literature, CCM is explicitly contrasted with the complex sphere manifold \(\{s\in\mathbb{C}^L:\|s\|_2=1\}\): the CCM has real dimension \(L\), whereas the complex sphere has real dimension \(2L-1\), because CCM fixes the modulus of each entry individually rather than imposing a single global energy constraint [1904.07329]. This suggests that the acronym must always be interpreted relative to local definitions rather than assumed to identify a unique canonical manifold.

## 2. CCM in almost complex geometry with circle actions

In the differential-topological literature, a CCM is a compact, connected almost complex manifold \((M,J)\) of real dimension \(2n\) endowed with a \(J\)-preserving circle action with isolated fixed points. At each fixed point \(p\in M^{S^1}\), the isotropy representation on \(T_pM\cong \mathbb{C}^n\) splits into one-dimensional weight spaces with nonzero integer weights \(w_{p,1},\dots,w_{p,n}\). If \(\lambda(p)\) denotes the number of negative weights at \(p\), and \(N_j\) is the number of fixed points with exactly \(j\) negative weights, then \(N_j=N_{n-j}\). The Euler characteristic equals the number of fixed points, so \(c_n[M]=|M^{S^1}|=\sum_{j=0}^n N_j>0\). ABBV localization yields the standard fixed-point formula, and in particular shows that fixed points must carry weights of both signs [1411.6458].

A central invariant is the index \(k_0\) of \((M,J)\), defined as the largest integer such that \(c_1=k_0\eta\) modulo torsion for some \(\eta\in H^2(M;\mathbb{Z})\); if \(c_1\) is torsion, then \(k_0=0\). When \(M\) is simply connected, the minimal Chern number equals \(k_0\). In the Hamiltonian symplectic case with isolated fixed points, the manifold is simply connected, the minimal Chern number coincides with the index, and \(1\le k_0\le n+1\). These relations place CCMs inside the broader geography problem for almost complex and symplectic manifolds, where one seeks sharp constraints on Chern numbers, fixed-point data, and possible circle actions [1411.6458].

## 3. Hilbert polynomial, rigidity, and Chern-number constraints

For an almost complex CCM with \(k_0\neq 0\), one chooses a line bundle \(L_0\to M\) with \(c_1(L_0)=\eta_0:=c_1/k_0\) modulo torsion and defines the Hilbert polynomial \(H(z)\in \mathbb{Z}[z]\) by
\[
H(k)=\operatorname{Ind}(L_0^k),\qquad k\in \mathbb{Z}.
\]
Riemann–Roch gives
\[
H(k)=\int_M e^{k\eta_0}\operatorname{Td}(M)
=\sum_{h=0}^n \frac{k^h}{h!}\int_M \eta_0^h\,\operatorname{Td}_{n-h}(M),
\]
so the coefficients of \(H\) are explicit linear combinations of the Chern numbers \(c_1^h\operatorname{Td}_{n-h}[M]\). In particular, \(H(0)=\operatorname{Td}(M)[M]=N_0\), and the leading coefficients include
\[
a_n=\frac{c_1^n[M]}{k_0^n n!},\qquad
a_{n-1}=\frac{c_1^n[M]}{2k_0^{n-1}(n-1)!},
\]
together with a corresponding formula for \(a_{n-2}\) involving \(c_1^{n-2}c_2[M]\) [1411.6458].

The Hilbert polynomial satisfies a reciprocity symmetry,
\[
H(z)=(-1)^n H(-z-k_0),
\]
and if \(k_0\ge 2\) then
\[
H(-1)=H(-2)=\cdots=H(-k_0+1)=0.
\]
If \(n\equiv k_0\pmod 2\), there is an additional zero at \(-k_0/2\). These zeros generate linear equations among the Chern numbers. The generating function \(G(t)=\sum_{k\ge 0}H(k)t^k\) has the form
\[
G(t)=\frac{U(t)}{(1-t)^{m+1}},
\]
with \(m=\deg(H)\), \(U(0)=N_0\), and a palindromicity relation for \(U\). In the symplectic case, \(N_0\in\{0,1\}\), with \(N_0=1\) if and only if the action is Hamiltonian and \(N_0=0\) exactly if it is non-Hamiltonian. For large index, the rigidity becomes explicit: if \(k_0=n+1\), then
\[
H(z)=\frac{N_0}{n!}\prod_{j=1}^n (z+j),
\]
while if \(k_0=n\), then
\[
H(z)=\frac{2N_0}{n!}\left(z+\frac{n}{2}\right)\prod_{j=1}^{n-1}(z+j).
\]
For \(k_0=n-1\) and \(k_0=n-2\), the theory yields strong linear relations between \(c_1^n[M]\) and \(c_1^{n-2}c_2[M]\), and these relations become practical criteria distinguishing Hamiltonian from non-Hamiltonian circle actions in the symplectic category [1411.6458].

## 4. Fixed-point classification and low-dimensional structure

The fixed-point theory of almost complex circle actions is exceptionally rigid in low cardinalities. If a compact, connected almost complex manifold has exactly one fixed point, then the manifold is a point. If it has exactly two fixed points, then either \(\dim M=2\), with weights \(\{a\}\) and \(\{-a\}\), or \(\dim M=6\), with weights
\[
\{-a-b,a,b\}\quad\text{and}\quad \{-a,-b,a+b\}.
\]
If there are exactly three fixed points, then \(\dim M=4\), and the weights are
\[
\{a+b,a\},\qquad \{-a,b\},\qquad \{-b,-a-b\},
\]
matching the weight data of a standard circle action on \(\mathbb{CP}^2\) [1510.00952].

Four fixed points already display a richer landscape. In real dimension \(8\), every compact almost complex manifold with a circle action and exactly four fixed points has the same Hirzebruch \(\chi_y\)-genus and the same Chern numbers as \(S^2\times S^6\):
\[
\chi_y(M)=-y+2y^2-y^3,
\]
\[
\int_M c_1^4=0,\quad
\int_M c_1^2c_2=0,\quad
\int_M c_2^2=0,\quad
\int_M c_1c_3=4,\quad
\int_M c_4=4.
\]
In particular, such a manifold is unitary cobordant to \(S^2\times S^6\) [2001.10699].

In dimension \(6\), Jang’s four-fixed-point classification splits into six cases, including \(\mathbb{CP}^3\)-type, \(Q^3\)-type, and several Todd-genus-zero types. One of the previously unknown Todd-genus-zero cases can be realized by Kustarev’s surgery on two copies of \(S^6\), producing a connected almost complex \(6\)-manifold diffeomorphic to \(S^4\times S^2\). The resulting action is not equivariantly diffeomorphic to a linear action, yielding an exotic \(S^1\)-action that preserves an almost complex structure [2308.06832].

Higher-dimensional lower bounds are also sharp. In real dimension \(10\), any circle action on a compact almost complex manifold with a fixed point has at least six fixed points. This minimum is attained by \(\mathbb{CP}^5\) and by \(S^6\times \mathbb{CP}^2\). The proof excludes the possibility of four fixed points by combining \(\chi_y\)-rigidity, ABBV localization, and integrality of Chern numbers [2411.03242].

## 5. CCM as compact complex manifolds built from circle bundles

A distinct usage, due to Sankaran and Thakur, starts from compact complex manifolds \(X_i\) and holomorphic line bundles \(\bar{L}_i\to X_i\). Writing \(S(L_i)\) for the associated principal circle bundle and \(L_i=\bar{L}_i\setminus X_i\) for the associated principal \(\mathbb{C}^*\)-bundle, one forms \(S(L)=S(L_1)\times S(L_2)\) and \(L=L_1\times L_2\). A CCM is then \(S(L)\) endowed with a complex structure induced from a holomorphic principal \(\mathbb{C}\)-bundle structure on \(L\to L/\mathbb{C}\), where the \(\mathbb{C}\)-action is holomorphic, free, and has closed, properly embedded leaves, with each orbit intersecting \(S(L)\) transversely in exactly one point. This identifies \(L/\mathbb{C}\) diffeomorphically with \(S(L)\) and makes \(S(L)\) into a compact complex manifold [1012.0668].

Three classes of complex structures are constructed. Scalar type always exists and arises from the proper holomorphic embedding
\[
\alpha_\tau:\mathbb{C}\to \mathbb{C}^*\times \mathbb{C}^*,\qquad z\mapsto (e^z,e^{\tau z}),\qquad \operatorname{Im}(\tau)>0.
\]
The quotient \(S_\tau(L)=L/\mathbb{C}_\tau\) is a complex manifold diffeomorphic to \(S(L_1)\times S(L_2)\), and the induced fibration over \(X_1\times X_2\) has elliptic-curve fiber \(\mathbb{C}/(\mathbb{Z}+\tau\mathbb{Z})\). Diagonal type generalizes this by using admissible \(\mathbb{C}\)-actions inside a torus \(T_1\times T_2\) satisfying a weak hyperbolicity condition. Linear type specializes to the homogeneous setting \(X_i=G_i/P_i\) with \(\bar{L}_i\) negative ample and constructs the action from \(\Lambda\in \operatorname{Lie}(\mathfrak{B}_1\times \mathfrak{B}_2)\), with \(\Lambda_s\) weakly hyperbolic and \(\Lambda_u\) sufficiently small [1012.0668].

These manifolds generalize classical Calabi–Eckmann manifolds: taking \(X_i=\mathbb{P}^{n_i}\) and \(\bar{L}_i=\mathcal{O}_{\mathbb{P}^{n_i}}(-1)\) gives \(S(L_i)\cong S^{2n_i+1}\), and the scalar-type construction recovers complex structures on \(S^{2n_1+1}\times S^{2n_2+1}\). The class is intrinsically non-Kähler in broad generality: if \(H^1(X_1;\mathbb{R})=0\) and \(c_1(\bar{L}_1)\in H^2(X_1;\mathbb{R})\) is nonzero, then \(S(L)\) admits no symplectic structure and hence is non-Kähler for any complex structure. Under projectivity and Cohen–Macaulay assumptions on the affine cones, there is a vanishing theorem for \(H^q(S(L);\mathcal{O}_{S(L)})\), and in homogeneous cases with \(\Lambda_u=0\) and maximal parabolics, the meromorphic function field \(K(S(L))\) is purely transcendental over \(\mathbb{C}\) with transcendence degree at most \(\dim S(L)-1\) [1012.0668].

## 6. CCM as a Riemannian manifold for unit-modulus optimization

In signal processing, radar, communications, and control, CCM denotes the constant-modulus feasible set
\[
\mathcal{M}=\{z\in\mathbb{C}^n: |z_i|=1\ \forall i\},
\]
or, in matrix form,
\[
\mathcal{M}=\{Z\in\mathbb{C}^{n\times m}: |Z_{ij}|=1\ \forall i,j\}.
\]
It is a product of circles, \(\mathbb{T}^n\) or \((S^1)^{nm}\), equipped with the Euclidean metric restricted from the ambient complex space:
\[
g_z(\eta,\xi)=\operatorname{Re}(\eta^H\xi),
\qquad
g_Z(U,V)=\operatorname{Re}(\operatorname{trace}(U^H V)).
\]
The tangent and normal spaces are elementwise:
\[
T_z\mathcal{M}=\{v\in\mathbb{C}^n:\operatorname{Re}(v\odot \overline{z})=0\}
=\{\mathrm{i}z\odot \alpha:\alpha\in\mathbb{R}^n\},
\]
\[
N_z\mathcal{M}=\{\beta\odot z:\beta\in\mathbb{R}^n\}.
\]
The orthogonal projection is
\[
\operatorname{Proj}_z(u)=u-\big(\operatorname{Re}(u\odot \overline{z})\big)\odot z,
\]
the Riemannian gradient is \(\operatorname{grad}f(z)=\operatorname{Proj}_z(\nabla f(z))\), and the Hessian admits the explicit embedded-manifold form
\[
\operatorname{Hess}f(z)[\eta]
=\operatorname{Proj}_z(\nabla^2f(z)[\eta])
-\big(\operatorname{Re}(\nabla f(z)\odot \overline{z})\big)\odot \eta.
\]
The exact exponential map is
\[
\exp_z(\eta)=z\odot e^{\mathrm{i}\alpha},
\]
for \(\eta=\mathrm{i}z\odot \alpha\), and a practical retraction is the elementwise normalization
\[
R_z(\eta)=\operatorname{normalize}(z+\eta).
\]
The KKT formulation is equivalent to Riemannian stationarity: projecting the Euclidean gradient removes the Lagrange multipliers associated with the unit-modulus constraints [2508.07396].

For MIMO radar beampattern design, the manifold structure leads to the projection–descent–retraction (PDR) iteration
\[
\mathbf{s}_{k+1}
=
\mathbf{R}\Big(
\mathbf{s}_k-\beta\,\Pi_{T_{\mathbf{s}_k}\mathcal{S}^L}\big(\nabla \bar f(\mathbf{s}_k)\big)
\Big),
\]
where the cost is quadratic in the waveform vector and the constraints are entrywise constant modulus. For quadratic objectives, the paper proves monotonic cost decrease under explicit step-size and regularization conditions and states convergence to a local minima. The same framework accommodates an orthogonality penalty \(\alpha\|\mathbf{X}^H\mathbf{X}-N\mathbf{I}_M\|_F^2\), yielding near-orthogonal waveforms while preserving constant modulus [1904.07329].

The same geometry appears in wireless beamforming. In IRS-assisted uplink NOMA, the IRS phase vector \(\mathbf{w}\in\mathbb{C}^L\) satisfies \(|w_i|=1\), so the feasible set is again a CCM. The optimization is reformulated as a max–min SINR feasibility-expansion problem on the manifold, smoothed, and combined with an exact penalty for inequality constraints. A Riemannian trust-region method, implemented in Manopt, is then used to optimize the projected gradient over the manifold. In the reported experiments, the manifold algorithm outperforms SDR, SDP-DC, and SROCR benchmarks while running at much lower complexity [2206.01312].

Across these literatures, CCM is best understood as a family of circle-based geometric frameworks rather than a single object. In almost complex geometry, it encodes local weight data, localization identities, and rigidity of Chern numbers. In complex-analytic geometry, it produces compact non-Kähler manifolds from circle bundles and holomorphic \(\mathbb{C}\)-actions. In optimization, it is the embedded torus of unit-modulus variables supporting explicit projection, retraction, gradient, and Hessian formulas. The shared terminology reflects the centrality of \(S^1\), but the mathematical content is decisively discipline-specific.

Source: https://www.emergentmind.com/topics/complex-circle-manifold-ccm