--- title: 'Good Cut Subspaces: Theory and Applications' url: https://www.emergentmind.com/topics/good-cut-subspaces type: topic --- # Good Cut Subspaces: Theory and Applications “Good Cut Subspaces” is not a single standardized term across the literature represented here. In Lorentzian geometry it denotes distinguished subsets of the four-complex-dimensional solution manifold of the good cut equation; in combinatorial geometry it refers to \(k\)-flats on which partition guarantees exceed the per-subspace Ham-Sandwich bound; in coding theory it is closely tied to families of subspaces with uniformly small intersections against prescribed cuts; in aperiodic order it can be read operationally as a choice of physical subspace yielding linear repetitivity; and in several other settings it labels subspaces singled out by extremal behavior across coordinate cuts, affine root-system cuts, or bipartite entanglement cuts [1007.4215] [1903.12516] [2204.13069] [2301.06997]. ## 1. Scope of the term across research areas The cited literature assigns “good cut” structure to subspaces by imposing a cut condition that is stronger than generic behavior in the ambient category. In each case, the subspace is distinguished not merely by inclusion but by the existence of a canonical family of cuts, slices, or projections with additional regularity, symmetry, or extremality. | Area | Ambient object | “Good cut” feature | |---|---|---| | Lorentzian geometry | \( \mathcal{I}^+ \), horizons, H-space | Shear-free or asymptotically shear-free NGCs from good cut functions [1007.4215] | | Discrete geometry | \( G_k(\mathbb{R}^d) \) | A \(k\)-flat with enhanced Ham-Sandwich or center-transversal guarantees [1903.12516] | | Finite-field/coding theory | \( V(k,q^m) \) | Small total intersections with low-dimensional \(F_{q^m}\)-subspaces [2204.13069] | | Aperiodic order | Cut-and-project schemes | A physical subspace yielding linear repetitivity [2301.06997] | | Quantum information | Multipartite Hilbert spaces | NPT or 1-distillability across every bipartition [2201.07918] | | Numerical linear algebra / graph theory | \(k\)-subspaces of \(\mathbb{R}^n\) | Large principal-angle distance from every coordinate subspace [2511.02387] | | Root systems | Affine subspaces in \(\mathbb{E}\) | Root slices with interval, orbit, and projection structure [2109.00944] | This suggests a unifying pattern: a good cut subspace is typically a subspace on which a cut operation becomes structurally rigid, extremal, or classification-friendly. ## 2. Good cut subspaces in Lorentzian geometry and H-space In the good-cut-equation literature, a good cut is a cross-section \(u = Z(\zeta,\bar{\zeta})\) of a null \(3\)-surface \(N\) such that the null directions constructed from the tangent to the cut determine an (asymptotically) shear-free null geodesic congruence at \(N\). The relevant setting takes \(N\) to have topology \(S^2 \times \mathbb{R}\), with coordinates \(u\) and complex stereographic coordinates \((\zeta,\bar{\zeta})\) on \(S^2\), and works on the complexification \(N_{\mathbb{C}}\), where \(u,\zeta,\bar{\zeta}\) are independent complex variables close to real and conjugate values [1007.4215]. The generalized good cut equation is posed on a sphere cross-section with conformal factor \(P = V P_0\), where \(P_0(\zeta,\bar{\zeta}) = 1 + \zeta \bar{\zeta}\). In the metric-sphere case \(V=1\), it reduces to the standard good cut equation \[ \overline{ð}_{0}^{2} G \equiv \partial_{\bar{\zeta}}\!\left(P_{0}^{2}\,\partial_{\bar{\zeta}} G\right) = \sigma\!\big(G,\zeta,\bar{\zeta}\big). \] The generalized equation is conformally equivalent to the standard equation after the change of stereographic coordinate \[ \bar{\zeta}^{\ast} = \frac{\bar{\zeta} + W}{1 - W\,\zeta}, \qquad \zeta^{\ast} = \zeta, \] with \(W\) determined by \[ V^{-2} = 1 + \overline{ð}_{0} W. \] Accordingly, all generalized versions share the same solution-space structure as the standard good cut equation. For sufficiently regular \(\sigma\), the solution space is a four-complex-dimensional manifold known as H-space, with solutions written as \[ u = G(z^{a}, \zeta, \bar{\zeta}), \qquad a=0,1,2,3. \] In the homogeneous case \(\sigma=0\), the regular solution space is spanned by the \(l=0,1\) modes, \[ G_{0}(z^{a}, \zeta, \bar{\zeta}) = z^{a} l_{a}(\zeta,\bar{\zeta}), \] so H-space reduces to complex Minkowski space. This homogeneous four-complex-parameter family is the “Minkowski good cut subspace.” Its associated congruences are shear-free everywhere, not merely asymptotically. More generally, the full H-space remains four-complex-dimensional for \(\sigma \neq 0\); the higher harmonics are fixed by \(\sigma\), while the \(l=0,1\) modes enter as the constants of integration in the Green’s-function iteration. The phrase “Good Cut Subspaces” in this setting refers to distinguished subsets of H-space obtained by imposing additional geometric or physical conditions. The explicit examples given are the complex Minkowski subspace \((\sigma=0)\), real slices selected by the requirement that \(u\) be real on the real sphere and that the associated \(L\)-field be regular, and horizon subspaces on a vacuum non-expanding horizon. On a horizon, the time-independent generalized equation \[ ð^{2} G = \sigma(\zeta,\bar{\zeta}) \] can be shifted by a complex supertranslation \(u \to u + f(\zeta,\bar{\zeta})\) so that \(\sigma=0\), and then mapped conformally to the homogeneous good cut equation; the resulting horizon good cut subspace is therefore equivalent to the Minkowski good cut subspace. Each point or curve \(z^{a} = \xi^{a}(\tau)\) in H-space yields a good cut function \(Z(\tau,\zeta,\bar{\zeta})\) and hence a shear-free or asymptotically shear-free null direction field through \[ L(u,\zeta,\bar{\zeta}) = ð_{0} Z(\tau,\zeta,\bar{\zeta}), \qquad u = Z(\tau,\zeta,\bar{\zeta}). \] ## 3. Geometric partition theory: Ham-Sandwich and center-transversal subspaces In geometric transversal theory, “good cut subspaces” arise from continuous assignments of mass distributions to subspaces of \(\mathbb{R}^d\). For \(1 \le k \le d\), a mass assignment is a continuous map \[ \mu: G_k(\mathbb{R}^d) \longrightarrow M_k, \] where \(G_k(\mathbb{R}^d)\) is the Grassmann manifold of \(k\)-dimensional linear subspaces and \(M_k\) is the space of \(k\)-dimensional mass distributions. The central phenomenon is that by varying the \(k\)-flat, one can sometimes obtain simultaneous partition guarantees for more masses than are available inside a fixed \(k\)-flat [1903.12516]. The paper’s main theorem states that if \(\mu_1,\dots,\mu_{n+d-k}\) are mass assignments on \(G_k(\mathbb{R}^d)\), with \(n \le k \le d\), then there exists a \(k\)-dimensional linear subspace \(h\) such that \(\mu_1^h,\dots,\mu_{n+d-k}^h\) have a common \((n-1,k)\)-center transversal. Equivalently, there is an affine \((n-1)\)-flat \(g \subset h\) such that every halfspace \(H \subset h\) containing \(g\) satisfies \[ \mu_i^h(H) \ge \frac{1}{k-n+2}\,\mu_i^h(h) \] for all \(i\). In the special case \(k=n\), one obtains a \(k\)-flat on which \(d\) assigned masses admit a simultaneous Ham-Sandwich cut. The paper therefore identifies a good cut subspace as a \(k\)-flat on which enhanced partition guarantees—beyond the per-flat bound \(k\)—hold for a larger family of assigned masses. A second theorem treats \((k-1)\)-horizontal subspaces, namely \(k\)-dimensional subspaces containing \(e_1,\dots,e_{k-1}\). If \(\mu_1,\dots,\mu_{d-k+2}\) are mass assignments on \(\mathrm{Hor}_k^{k-1}(\mathbb{R}^d)\), with \(2 \le k \le d\), then there exists a \(k\)-dimensional \((k-1)\)-horizontal subspace \(h\) such that the assigned masses admit a common Ham-Sandwich cut inside \(h\). In the language used in the paper, “choosing \(k-1\) vectors” means fixing those coordinate directions and retaining one free direction. The same work extends the perspective from single hyperplanes to families of \(n\) hyperplanes. For a family \(\mathcal{L}\) of oriented hyperplanes, the parity-colored regions \[ R^+ := \{p \in \mathbb{R}^d : \lambda(p)\ \text{is even}\}, \qquad R^- := \{p \in \mathbb{R}^d : \lambda(p)\ \text{is odd}\} \] are defined using the parity of the number of positive sides containing \(p\). The paper proves a relaxed version of Langerman’s conjecture: any \(dn\) mass distributions in \(\mathbb{R}^d\) can be almost simultaneously bisected by \(n\) hyperplanes. Throughout, the existence proofs use configuration-space/test-map constructions, a Borsuk–Ulam-type obstruction, Stiefel–Whitney classes, and quasi-sections rather than algorithms. ## 4. Finite-field and coding-theoretic formulations In the finite-field literature on subspace designs, “good cutting” is formulated through uniformly small intersections with low-dimensional subspaces. The ambient space is \(V = V(k,q^m)\), a \(k\)-dimensional vector space over \(F_{q^m}\), and the basic objects are ordered families of \(F_q\)-subspaces \(U_i \le V\). The paper’s central notion is an \((s,A)_q\)-subspace design: an ordered set \(\mathcal{U} = (U_1,\ldots,U_t)\) such that \[ \dim_{F_{q^m}}\!\big(\langle U_1,\ldots,U_t\rangle_{F_{q^m}}\big)\ge s \] and for every \(s\)-dimensional \(F_{q^m}\)-subspace \(W \le V\), \[ \sum_{i=1}^t \dim_{F_q}(U_i \cap W) \le A. \] An \(s\)-design is an \((s,s)_q\)-subspace design. In this setting, good cutting means that every low-dimensional \(W\) intersects the family \(U_i\) in small total \(F_q\)-dimension; for hyperplanes, this is the case \(s=k-1\) [2204.13069]. The theory establishes several sharp parameter bounds. Proposition 2.1 gives \(s \le A\). Theorem 2.8 shows that if \((U_1,\dots,U_t)\) is an \(s\)-design in \(V(k,q^m)\), then \[ \dim_{F_q}(U_i) \le \begin{cases} k, & \text{if } m < s+1,\\[4pt] \frac{mk}{s+1}, & \text{if } m \ge s+1. \end{cases} \] Maximum \(s\)-designs attain the upper dimension bound when \(m \ge s+1\), and Theorem 3.5 constructs them when \(k = (s+1)\ell\), \(m \ge s+1\), and \(t \le q-1\), by direct-sum glueing of linearized-polynomial constructions. The good-cut interpretation becomes especially concrete in the code correspondence. For a non-degenerate sum-rank code with generator blocks \(G=(G_1|\cdots|G_t)\), letting \(U_i\) be the \(F_q\)-span of the columns of \(G_i\), the paper shows that \[ w(xG)=N-\sum_{i=1}^t \dim_{F_q}(U_i\cap x^\perp), \] and hence \[ d=N-\max\left\{\sum_{i=1}^t \dim_{F_q}(U_i\cap H) : H \text{ hyperplane in } V(k,q^m)\right\}. \] Thus small sums of intersections are exactly the structural input producing large sum-rank distance. The same framework yields optimal subspace designs, MSRD codes, two-intersection sets with respect to hyperplanes, cutting blocking sets, minimal codes, and explicit lossless dimension expanders. A further refinement is the notion of a cutting design. An ordered set \((U_1,\dots,U_t)\) is a cutting design if for any pair of hyperplanes \(H,H'\), \[ U_i\cap H \subseteq U_i\cap H' \ \forall i \quad\Rightarrow\quad H=H'. \] If there exists \(c>0\) such that \[ \sum_{i=1}^t \dim_{F_q}(U_i\cap H)=c \] for every hyperplane \(H\), then the family is a non-degenerate cutting design. This is the design-theoretic analogue of a uniformly good cut profile. ## 5. Cut-and-project sets and linear repetitivity In aperiodic order, the phrase can be used operationally for physical subspaces in cut-and-project schemes that yield linear repetitivity. A \(k\)-to-\(d\) cut-and-project scheme is \[ \mathcal{S}=(\mathrm{tot},\mathrm{phy},\mathrm{intl},\Gamma,W), \] where \(\mathrm{tot}=\mathrm{phy}\oplus\mathrm{intl}\), \(\Gamma\) is a discrete cocompact lattice of rank \(k\), and \(W\subset\mathrm{intl}\) is a compact window equal to the closure of its interior. The associated cut-and-project set is \[ \Lambda:=\{x\in\Gamma_\vee\mid x^\star\in W\}, \] in non-singular position. The paper studies polytopal windows, including non-convex, disconnected, and labelled cases, and frames the quality of the cut through the dynamical property of linear repetitivity (\(\mathrm{LR}\)) [2301.06997]. Low complexity is defined by \[ p(r)\ll r^d, \] where \(p(r)\) counts distinct \(r\)-patches. The complexity exponent is \[ p(r)\asymp r^\alpha, \qquad \alpha=\max_{f\in F_0}\sum_{V\in f}\bigl(d-\mathrm{rk}(V)+\beta_V\bigr), \] with \(F_0\) the set of flags in the supporting subspaces of the window. Low complexity \(C\) is the sharp case \(\alpha=d\), equivalently the rank-sum condition \[ \sum_{V\in f}\mathrm{rk}(V)=k(n-1)\quad\text{for every flag }f\in F_0, \] together with hyperplane spanning \(\beta_V=n-1\). Under weak homogeneity, the main theorem states: - \(\mathrm{LR}\) holds if and only if \(C\) and \(D\) hold. Here \(D\) is the Diophantine property of the projected lattice \(\Gamma_<\subset \mathrm{intl}\): \[ \|g\|\ge c\,\eta(g)^{-\delta}, \qquad \delta=\frac{d}{n}, \] for all nonzero \(g\in \Gamma_<\). In the fully general case weak homogeneity is replaced by an inhomogeneous analogue \(DF\), formulated on flag-group projections and generalized vertex displacements. The paper’s Main Theorem B gives the two-sided criterion: if \(\mathrm{LR}\) holds then \(C\) and \(DF\) hold, and conversely there exists \(N\in\mathbb{N}\) such that if \(C\) and \(DF\) hold after replacing \(\Gamma\) by \(\frac1N\Gamma\), then \(\mathrm{LR}\) holds. Within this framework, a good cut subspace is the choice of physical space \(\mathrm{phy}\) for which the associated cut-and-project set is linearly repetitive. In weakly homogeneous settings, this means precisely minimal complexity plus a homogeneous Diophantine condition; in the general setting, minimal complexity plus inhomogeneous Diophantine avoidance of shifted vertex displacements. The paper applies this criterion to Ammann–Beenker, Penrose, generalized Penrose, and labelled-window variants, after MLD reduction when the internal space has the form \(\mathbb{R}^n\oplus G\oplus\mathbb{Z}^r\). ## 6. Other specialized uses: multipartite cuts, coordinate extremality, and affine root slices In quantum information, a “Good Cut Subspace” is defined directly in terms of bipartitions. For a multipartite Hilbert space \(H = H_{A_1}\otimes\cdots\otimes H_{A_n}\), a subspace \(S\subset H\) is a Good Cut Subspace if every density operator supported on \(S\) is NPT across every bipartition \(A|A^c\); a stronger form requires 1-distillability across every cut. The construction in “Construction of genuinely entangled multipartite subspaces from bipartite ones by reducing the total number of separated parties” takes bipartite NPT or 1-distillable subspaces arranged along a chain, tensors them, and joins adjacent subsystems. In the tripartite case, \[ \mathcal W_{A_1 A'_2 A_4} := J_{A_2 A_3 \to A'_2}\big(S^{(1)}_{A_1 A_2} \otimes S^{(2)}_{A_3 A_4}\big) \] has the property that any density operator supported on \(\mathcal W\) is NPT across every bipartition, and 1-distillable across every bipartition if each input subspace is 1-distillable. The same paper also proves that joining adjacent bipartite CESs yields a genuinely entangled subspace, with \[ G_{GME}(S_{ABC}) = \min\big(G(S_{AB_1}),\,G(G_{B_2C})\big) \] for the resulting tripartite joined product [2201.07918]. In numerical linear algebra and graph theory, the phrase is used informally for \(k\)-dimensional subspaces of \(\mathbb{R}^n\) that are maximally far from all coordinate \(k\)-subspaces in largest-principal-angle distance. For \(U,V\subset\mathbb{R}^n\) of dimension \(k\), \[ d(U,V)=\theta_k, \qquad \cos\theta_i=\sigma_i(Q_U^TQ_V). \] For a coordinate subspace \(C_S\), \[ d(U,C_S)=\arccos\big(\sigma_{\min}(Q_U[S,:])\big). \] The 2025 paper proves that if \(U\) is realized as the star space \(U=\mathrm{col}(W^{1/2}B^T)\) of a nontrivial \(2\)-connected series-parallel graph with graph-induced weights, then \[ \theta_k(U,C_S)\ge \arccos(1/\sqrt{n}) \] for every coordinate \(k\)-subspace \(C_S\). These star spaces therefore supply explicit extremal subspaces for the GTZ1997 coordinate-subspace hypothesis [2511.02387]. In the root-system literature, the paper “Root systems, affine subspaces, and projections” does not introduce the term verbatim, but it studies a natural class of affine cuts with the same flavor. For \(S\subseteq\Delta\) and \(\beta\in\Phi^+\), the slice \[ \Phi_{S,\beta}=\{\gamma\in\Phi:c_\alpha(\gamma)=c_\alpha(\beta)\ \text{for all }\alpha\in S\} \] is the intersection of \(\Phi\) with an affine subspace \(A=\beta+\mathrm{Span}(\Delta\setminus S)\). When \(\mathrm{Supp}(\beta)\cap S\neq\varnothing\), \(\Phi_{S,\mathbb{Z}\beta}\) is a root subsystem, \(\Phi_{S,\beta}\) is an interval in the root poset with explicit minimum and maximum, and \(W(\Delta\setminus S)\) acts transitively on the long roots in \(\Phi_{S,\beta}\) and on the short roots in \(\Phi_{S,\beta}\), so the slice breaks into at most two parabolic orbits. In codimension \(1\), the projections of these slices are weight polytopes, and the orthogonal projection of the root polytope onto a standard parabolic subspace satisfies a containment \[ \pi_U(P_\Phi)\subset r\cdot P_{\Phi\cap U} \] with \(1