--- title: Parity Condition in Graphs, Games & Quantum url: https://www.emergentmind.com/topics/parity-condition type: topic --- # Parity Condition in Graphs, Games & Quantum A parity condition is a requirement formulated modulo \(2\). Across the literature surveyed here, the term denotes several technically distinct kinds of constraint: prescribed degree congruences in spanning subgraphs, winning conditions on infinite plays, consistency equations in parity-based quantum encodings, symmetry under parity-time conjugation, and the odd-or-even behavior of arithmetic sequences such as partition values on quadratic progressions [1301.7565; 2307.05762; 2105.06233; 2509.09553]. What unifies these uses is not a single definition, but the role of mod-\(2\) structure as a compact way to encode feasibility, acceptance, symmetry, or information. ## 1. Degree congruence and the graph-theoretic parity condition In graph factor theory, the standard formalization is the \((g,f)\)-parity factor. For integer-valued functions \(g,f:V(G)\to \mathbb Z_{\ge 0}\) with \[ g(v)\le f(v), \qquad g(v)\equiv f(v)\pmod 2, \] a spanning subgraph \(F\) is a \((g,f)\)-parity factor if \[ g(v)\le d_F(v)\le f(v), \qquad d_F(v)\equiv f(v)\pmod 2 \] for every vertex \(v\). In the constant case \(g\equiv a\), \(f\equiv b\), this yields an \((a,b)\)-parity factor, meaning \[ a\le d_F(v)\le b, \qquad d_F(v)\equiv a\equiv b \pmod 2, \] so the allowed degrees are exactly \(\{a,a+2,\dots,b\}\) [1301.7565; 1606.04608; 2305.16108]. Lovász’s \((g,f)\)-parity theorem gives the fundamental obstruction criterion. With disjoint \(S,T\subseteq V(G)\), a parity factor exists exactly when a deficiency term of the form \[ \eta(S,T)=g(T)-\sum_{x\in T} d_{G-S}(x)-f(S)+\tau(S,T) \] is never positive, where \(\tau(S,T)\) counts components \(C\) of \(G-(S\cup T)\) satisfying a parity test \[ e_G(V(C),T)+f(V(C))\equiv 1\pmod 2 \] [1301.7565]. In the constant case, this becomes the standard way to convert a parity-factor problem into inequalities over vertex cuts and component counts. A particularly compact specialization is the intrinsic “restricted minimum degree” criterion. A graph \(G\) contains a factor \(F\) with \[ d_F(v)\ge g(v), \qquad d_F(v)\equiv g(v)\pmod 2 \quad (\forall v\in V(G)) \] if and only if, for every \(T\subseteq V(G)\), \[ g(T)-\sum_{x\in T} d_G(x)+\tau(T)\le 0, \] where \(\tau(T)\) counts components \(C\) of \(G-T\) such that \[ e_G(V(C),T)+g(V(C))\equiv 1\pmod 2 \] [1301.7565]. For constant \(g\equiv m\), this yields explicit even-factor and odd-factor criteria. When \(m\) is even, \[ m|T|-\sum_{x\in T} d_G(x)+\tau(T)\le 0 \] with \(\tau(T)\) counting components satisfying \(e_G(V(C),T)\equiv 1\pmod 2\); when \(m\) is odd, the same inequality holds but \(\tau(T)\) counts components satisfying \[ e_G(V(C),T)+|V(C)|\equiv 1\pmod 2 \] [1301.7565]. These formulations make the graph-theoretic parity condition exact rather than heuristic. The parity obstruction is local at the level of components of \(G-T\), but it is also global, because the inequalities compare total degree demand on \(T\) with total incident degree supply. ## 2. Connectivity, degree, and spectral conditions forcing parity factors A large part of the graph-theoretic literature asks when crude global hypotheses force the Lovász inequalities automatically. The resulting criteria are diverse, but all treat the parity condition as a structural consequence of density or connectivity rather than as an explicit constraint to be checked subset by subset. | Setting | Hypothesis | Conclusion | |---|---|---| | General graphs | \(m\)-edge-connected with \(\delta(G)\ge m+1\) for even \(m\), or \((m+1)\)-edge-connected for odd \(m\) | even or odd factor \(F\) with \(d_F(v)\ge m\) [1301.7565] | | \(r\)-regular graphs | parity-dependent conditions on \(r,a,b,m\), including \(r/m\le b\) and \(a\le r(1-1/m)\) in the case \(r\) even, \(a,b\) odd | existence of an \((a,b)\)-parity factor [1110.3198] | | Degree/Ore-type condition | \(n\ge \frac{b(a+b)(a+b+2)}{2a}\), \(\delta(G)\ge a+\frac{b}{a}-1\), and \(\max\{d_G(u),d_G(v)\}\ge \frac{an}{a+b}\) for nonadjacent \(u,v\) | existence of an \((a,b)\)-parity factor [1606.04608] | | Spectral condition | \(n\ge \max\{2a(a+1)+b-3,\ (2a+2)(a+1)\}\) and \(\rho(G)\ge \rho(H_{n,a})\) with \(H_{n,a}=K_{a-1}\vee (K_1\cup K_{n-a})\) | existence of an \((a,b)\)-parity factor unless \(G=H_{n,a}\) [2305.16108] | For edge connectivity, the parity gain can be explicit. If \(m\) is even and \(G\) is \(m\)-edge-connected with minimum degree \(m+1\), then \(G\) contains an even factor \(F\) with \(d_F(v)\ge m\). If \(m\) is odd and \(G\) is \((m+1)\)-edge-connected, then \(G\) contains an odd factor \(F\) with \(d_F(v)\ge m\) [1301.7565]. The proof mechanism is characteristic: edge connectivity bounds the cut size of each parity-obstructing component, and when the required cut parity disagrees with the connectivity threshold, the lower bound increases by \(1\), producing the contradiction. Regular graphs admit a more delicate parity-sensitive picture. For an \(m\)-edge-connected \(r\)-regular graph, sufficient conditions depend on the parities of \(r,a,b\). In the case \(r\) even and \(a,b\) odd, \(|G|\) even, the conditions \[ \frac{r}{m}\le b, \qquad a\le r\left(1-\frac{1}{m}\right) \] force an \((a,b)\)-parity factor; the other two cases use the odd integer \(m^*\in\{m,m+1\}\) with \(m^*\equiv 1\pmod 2\), and require either \[ a\le r\left(1-\frac{1}{m^*}\right) \] or \[ \frac{r}{m^*}\le b \] according to the parity pattern of \(r,a,b\) [1110.3198]. The paper explicitly states that these edge-connectivity conditions are sharp. The Ore-type degree condition of Nishimura-style factor theory also extends to parity factors. If \[ n \ge \frac{b(a+b)(a+b+2)}{2a}, \qquad \delta(G)\ge a+\frac{b}{a}-1, \] and every nonadjacent pair \(u,v\) satisfies \[ \max\{d_G(u),d_G(v)\}\ge \frac{an}{a+b}, \] then \(G\) has an \((a,b)\)-parity factor [1606.04608]. A plausible implication is that the parity obstruction can be overcome whenever local degree deficits are ruled out strongly enough on nonedges. Spectral forcing gives a different extremal language. Wang, Yu, Hu, and Wen show that if \[ \rho(G)\ge \rho(H_{n,a}), \qquad H_{n,a}=K_{a-1}\vee (K_1\cup K_{n-a}), \] under the stated order condition, then \(G\) has an \((a,b)\)-parity factor unless \(G=H_{n,a}\) [2305.16108]. Here the parity condition is controlled indirectly through adjacency spectral radius, with the exceptional join graph serving as the sharp obstruction. ## 3. Strong parity properties and universal parity prescriptions in graphs A stronger usage of “parity condition” requires a graph to realize not one prescribed parity pattern, but every admissible one. This leads to the strong parity property, all \((a,b)\)-parity factors, and the \(k\)-strong parity property. A graph has the strong parity property if for every even set \(X\subseteq V(G)\), there exists a spanning subgraph \(F\) such that \[ \delta(F)\ge 1,\qquad d_F(u)\equiv 1 \pmod 2 \text{ for }u\in X,\qquad d_F(v)\equiv 0 \pmod 2 \text{ for }v\notin X. \] Lu, Yang, and Zhang’s characterization states that this holds if and only if \[ c(G-S)\le \sum_{v\in S} d_G(v)-2|S|+1 \] for every \(S\subset V(G)\). On the spectral side, if \(G\) is connected of order \(n\), minimum degree \(\delta\ge 3\), \(n\ge 282\), and \[ \rho(G)\ge \rho(H(n,\delta)), \] then \(G\) has the strong parity property unless \[ G=H(n,\delta)=K_\delta \vee \Bigl(K_{n-\delta(\delta-1)-1}\cup (\delta(\delta-2)+1)K_1\Bigr) \] [2402.13601]. The paper on all \((a,b)\)-parity factors imposes an even stronger universality requirement. A graph has all \((a,b)\)-parity factors if it has an \(h\)-factor for every function \[ h:V(G)\to \{a,a+2,\ldots,b\} \] with \(h(v)\equiv b\pmod 2\) and the necessary global parity compatibility. Under \[ n\ge 3(b+1)(a+b),\qquad \delta(G)\ge \frac{b^2-b}{a}, \] and \[ \max\{d_G(u),d_G(v)\}\ge \frac{bn}{a+b} \] for every pair of nonadjacent vertices \(u,v\), a connected graph has all \((a,b)\)-parity factors [2009.03032]. Here the parity condition is not a single congruence class to be achieved once; it is a uniform realizability statement over all allowable vertexwise parity-bounded degree prescriptions. Kano and Matsumura’s \(k\)-strong parity property adds a lower bound outside the odd-degree set. For even \(k\ge 2\), a graph has the \(k\)-strong parity property if for every even \(X\subseteq V(G)\) there is a spanning subgraph \(F\) with \[ d_F(u)\equiv 1 \pmod 2 \quad (u\in X), \] and \[ d_F(v)\in \{k,k+2,k+4,\ldots\} \quad (v\in V(G)\setminus X). \] Their characterization is \[ c(G-S)\le \sum_{v\in S} d_G(v)-k|S|+1 \] for all \(S\subseteq V(G)\). Subsequent work gives both a size condition and a signless Laplacian spectral radius condition forcing this property [2505.20352]. This suggests that “strong” parity conditions are best viewed as deletion-component inequalities rather than merely as degree congruences. ## 4. Parity objectives in infinite games and verification In game theory and formal verification, a parity condition is an \(\omega\)-regular acceptance criterion defined by priorities assigned to states. In the most common convention represented here, a play \(\pi\) satisfies even parity iff the minimum priority seen infinitely often is even: \[ \pi\in EPAR \iff \min(\mathrm{Inf}(\pi)) \text{ is even}. \] This is the convention used for simple stochastic games with energy-parity objectives, energy parity games, and mean-payoff parity games [2307.05762; 1001.5183; 1706.06139]. A different convention also appears: one paper defines the winner by the parity of the highest priority seen infinitely often [1807.10210]. The presence of both conventions is a standard source of confusion; the surveyed papers simply adopt different priority orderings. Once combined with quantitative objectives, the parity condition becomes the qualitative component of a mixed specification. In simple stochastic games, the central objective is \[ EN(k)\cap EPAR, \] where \[ \pi\in EN(k)\iff k+\sum_{i=0}^{n-1} c_i>0 \text{ for all } n\ge 0. \] The value of this objective can be approximated in \(2\)-NEXPTIME, and \(\varepsilon\)-optimal strategies for both players require \[ O\!\left(2\textrm{-}\mathsf{EXP}(\|\mathcal G\|^{O(1)})\cdot \log\frac{1}{\varepsilon}\right) \] memory modes for unary rewards [2307.05762]. In two-player energy parity games, exponential memory is necessary and sufficient for winning strategies, the finite initial credit problem is in \(\mathrm{NP}\cap \mathrm{coNP}\), and player 2 has memoryless winning strategies [1001.5183]. Mean-payoff parity games conjoin the parity condition with a threshold or optimality condition on the mean payoff. The threshold problem can be solved in \[ O(n^{d-1}\cdot m\cdot W), \] and the value problem in \[ O(n^{d}\cdot m\cdot W\cdot \log(nW)) \] for \(n\) vertices, \(m\) edges, \(d\) priorities, and maximum absolute reward \(W\) [1706.06139]. Parity is again the hard qualitative constraint: if it fails, the payoff is set to \(-\infty\). The parity condition also supports mixed qualitative semantics. For Markov decision processes and turn-based stochastic games with two parity objectives, one of which must hold surely and the other almost-surely or limit-surely, the MDP problems lie in \(\mathbf{NP}\cap \mathbf{coNP}\) for finite-memory strategies, whereas the corresponding turn-based stochastic game problems are \(\mathbf{coNP}\)-complete [1804.03453]. By contrast, the algorithmic study of parity solvers itself shows that parity conditions can induce difficult recursive behavior: the Two Counters construction gives exponential lower bounds for a wide range of attractor-based parity game algorithms [1807.10210]. ## 5. Quantum, coding, and symmetry interpretations In parity quantum optimization, the parity condition is a consistency relation on encoded variables. A parity qubit represents a product of logical spins, and any collection of such products in which every logical index appears an even number of times must multiply to the identity. In operator form, \[ \tilde \sigma_z^{(a_1)}\tilde \sigma_z^{(a_2)}\cdots \tilde \sigma_z^{(a_n)}\ket{\tilde\psi}=\ket{\tilde\psi}. \] In bit language, if \(\mathbf w=\mathbf v\mathbf G\), then valid parity strings satisfy \[ \mathbf w\mathbf P^\top=0,\qquad \mathbf G\mathbf P^\top=0. \] Thus the parity condition is literally a parity-check code constraint derived from closed cycles of a graph or hypergraph [2105.06233]. In minimally universal parity quantum computing, the term refers to a sufficient condition on the chosen family of parity subsets \(\mathbb P=\{S_j\}\). Universality follows if every \(|S_j|\) is even, \(\bigcup_j S_j=\{1,\dots,n\}\), nonadjacent sets are disjoint, and consecutive sets intersect in exactly one element, with all overlaps distinct. This leads to the sharp minimality result that one auxiliary parity qubit suffices when the number of logical qubits is even, and two are necessary and sufficient when it is odd [2504.03556]. Here parity is no longer merely a constraint on outputs; it is the organizing principle of the computational model. A different quantum use appears in simulated \(\mathcal{PT}\)-symmetric evolution. With parity operator \(\mathcal P=\sigma_x\) and time reversal given by complex conjugation, the relevant condition is \[ [H,\mathcal{PT}]=0, \] equivalently \[ \mathcal P H\mathcal P = H^\dagger. \] The paper shows that local \(\mathcal{PT}\)-symmetric evolution alone is not sufficient for information transfer in the simulated setting: the density matrix and the corresponding measurements must contain complex numbers [2102.13630]. In this context, the parity condition is a symmetry condition rather than a feasibility or acceptance condition. ## 6. Arithmetic parity and parity as computational output In arithmetic, parity conditions can concern the values of number-theoretic functions along thin sequences. A recent result proves that if \(1