---
title: P-Divisibility in Quantum Systems and Arithmetic
url: https://www.emergentmind.com/topics/p-divisibility
type: topic
---

# P-Divisibility in Quantum Systems and Arithmetic

P-divisibility is not a uniform term across mathematics and mathematical physics. In open quantum systems it denotes a positivity-based divisibility property of a dynamical map: for every \(t\ge s\ge 0\), there exists an intermediate propagator \(V_{t,s}\) such that \(\Lambda_t=V_{t,s}\Lambda_s\) and \(V_{t,s}\) is positive and trace-preserving. In arithmetic, algebraic geometry, modular forms, coding theory, and related areas, the corresponding expression is usually \(p\)-divisibility for a fixed prime \(p\), meaning divisibility by \(p\) or \(p^n\), often formulated through local-global principles, \(p\)-adic valuations, or congruence laws. The common motif is factorization or divisibility constrained by an ambient structure—positivity in dynamics, Galois cohomology in arithmetic, alterations in geometry, or \(p\)-adic integrality in generating functions [2502.15498][1204.5831].

## 1. Quantum-dynamical P-divisibility

In the open-systems literature, P-divisibility is weaker than CP-divisibility and stronger than mere positivity of the one-parameter family itself. If \(\Lambda_t\) is the reduced dynamics, P-divisibility requires positive trace-preserving intermediate maps \(V_{t,s}\) for all \(t\ge s\), whereas CP-divisibility requires each \(V_{t,s}\) to be completely positive. Operationally, P-divisibility preserves states of the system alone, while CP-divisibility preserves states even after embedding into an arbitrary ancilla extension [2502.15498].

For two-level systems governed by a time-local master equation with rates \(\Gamma(t)\), \(\gamma_+(t)\), and \(\gamma_-(t)\), the qubit characterization is explicit. The dynamical map is P-divisible at time \(t\) if and only if
\[
|\gamma_-(t)|\le \gamma_+(t),
\]
and, whenever
\[
2\Gamma(t)\le \gamma_+(t),
\]
one also has
\[
\gamma_-^2(t)\le 4\Gamma(t)\big(\gamma_+(t)-\Gamma(t)\big).
\]
The same paper proves equivalence, for qubits, between several criteria that had appeared separately in the literature: positivity of intermediate propagators, trace-norm contraction on Hermitian operators, a Bloch-ball geometric criterion, and Kossakowski’s positivity condition. In the same model class, the hierarchy is
\[
\text{CP-divisible} \Rightarrow \text{P-divisible} \Rightarrow \text{BLP monotone},
\]
with reverse implications failing in general. Two special regimes collapse the hierarchy: if the instantaneous fixed point lies on the Bloch sphere, P-divisibility coincides with CP-divisibility; if the dynamics is unital, P-divisibility coincides with the BLP no-information-backflow condition [2502.15498].

Geometrically, the qubit criterion is encoded by the inward-pointing condition of the infinitesimal Bloch vector field on the sphere. Writing
\[
R(z)=z^2(\Gamma-\gamma_+)+z\gamma_- - \Gamma,
\]
P-divisibility is equivalent to \(R(z)\le 0\) for all \(z\in[-1,1]\). This reduction to a one-variable quadratic is one of the reasons the two-level case admits a complete classification [2502.15498].

## 2. Tensor powers, stability, and finite-step analogues

A central structural theorem concerns the second tensor power. For a time-local family \(\{\Lambda_t\}_{t\ge 0}\), the map \(\Lambda_t\) is CP-divisible if and only if \(\Lambda_t\otimes\Lambda_t\) is P-divisible; equivalently, \(\Lambda_t\otimes\Lambda_t\) is P-divisible if and only if it is CP-divisible. Thus CP-divisibility of the single-copy dynamics is stable under passage to two identical noninteracting copies, and P-divisibility of the doubled dynamics is already strong enough to recover complete positivity of the original intermediate propagators [1610.04634].

The same work isolates an important limitation. For time-dependent generators, positivity of \(\Lambda_t\otimes\Lambda_t\) at each fixed time does not imply complete positivity of \(\Lambda_t\). The distinction is between pointwise positivity of the tensor square and P-divisibility of the tensor-square propagators \(\Lambda_{t,s}\otimes\Lambda_{t,s}\). The former is too weak, whereas the latter is equivalent to CP-divisibility of the original family [1610.04634].

A different but related use of divisibility appears in stochastic maps. There the question is finite-step factorization: given a stochastic matrix \(P\), does there exist a stochastic matrix \(Q\) such that \(P=Q^2\)? This is a positive, probability-preserving midpoint factorization, but not a continuous-time P-divisibility condition. The distinction is explicit: finite divisibility asks for a prescribed stochastic root, whereas embeddability asks for a full continuous-time semigroup representation \(P=\exp(Qt)\). The decision problem for stochastic square roots is NP-complete, and the same hardness extends to nonnegative matrices and CPTP maps [1411.7380].

This separation matters conceptually. In the quantum Markovianity literature, P-divisibility is a two-time property of all propagators \(V_{t,s}\). In the complexity-theoretic literature on stochastic matrices, divisibility is a single finite factorization problem. The two notions are adjacent but not interchangeable [1411.7380].

## 3. Local-global \(p\)-divisibility in arithmetic geometry

In arithmetic, \(p\)-divisibility usually concerns whether divisibility by a prime power can be detected locally. For an elliptic curve \(\mathcal E/k\), the local-global divisibility problem for \(p^n\) asks whether a point \(P\in \mathcal E(k)\) satisfying
\[
P=p^nD_v \quad \text{in } \mathcal E(k_v)
\]
for all but finitely many places \(v\) must already satisfy
\[
P=p^nD \quad \text{in } \mathcal E(k).
\]
The cohomological obstruction is the first local cohomology group
\[
H^1_{\mathrm{loc}}(G_n,\mathcal E[p^n]),
\qquad
G_n=\operatorname{Gal}(k(\mathcal E[p^n])/k),
\]
and vanishing of this group rules out counterexamples [1104.4762].

For elliptic curves over number fields, the decisive obstruction identified in the 2011 refinement is \(k\)-rational \(p\)-torsion. If \(k\) does not contain \(\mathbb Q(\zeta_p+\bar\zeta_p)\) and \(\mathcal E(k)\) has no torsion point of exact order \(p\), then local divisibility by \(p^n\) for all but finitely many places is equivalent to global divisibility by \(p^n\), for every \(n\ge 1\). The proof proceeds through the structure of the mod-\(p\) Galois image \(G_1\), diagonal and triangular decompositions of the higher image \(G_n\), and vanishing results for \(H^1_{\mathrm{loc}}(G_n,\mathcal E[p^n])\) using Sah’s theorem and inflation-restriction [1104.4762].

The precursor for \(p^2\) had already shown a sharper obstruction than rational \(p\)-isogenies. If \(p>3\), \(k\) does not contain the degree-\(p\) subfield of \(\mathbb Q(\zeta_{p^2})\), and local-global divisibility by \(p^2\) fails, then \(\mathcal E\) must have a \(k\)-rational point of exact order \(p\). Over \(\mathbb Q\), this reduces the possible exceptional primes for divisibility by \(p^2\) to \(\{2,3,5,7\}\) [1103.4963].

The framework extends beyond elliptic curves. For a commutative algebraic group \(\mathcal A/k\) with \(\mathcal A[p]\simeq (\mathbf Z/p\mathbf Z)^n\), sufficient conditions for local-global divisibility by \(p\) and for \(\Sha(k,\mathcal A[p])=0\) are given in terms of the Galois module \(\mathcal A[p]\). If \(\mathcal A[p]\) is very strongly irreducible, or a direct sum of very strongly irreducible modules, and
\[
p>\frac n2+1,
\]
then local-global divisibility by \(p\) holds in \(\mathcal A\) and \(\Sha(k,\mathcal A[p])=0\). For principally polarized abelian varieties, this implies divisibility by \(p\) in the Weil–Châtelet group and local-global divisibility by \(p\) in \(H^r(k,\mathcal A)\) for all \(r\ge 0\) [1603.05857].

## 4. Cohomological, geometric, and automorphic \(p\)-divisibility

A geometric formulation appears in coherent cohomology. For a proper morphism \(f:X\to S\) of noetherian schemes with \(S\) affine, there exists an alteration \(\pi:Y\to X\) such that
\[
\pi^*(H^i(X,\mathcal O_X))\subset p(H^i(Y,\mathcal O_Y))
\qquad (i>0).
\]
By iteration, higher coherent cohomology classes become arbitrarily \(p\)-divisible after passing to successive proper covers. When the base has dimension at most \(1\), the result strengthens to killing \(p\)-torsion after proper surjective cover. Properness is essential: the paper gives nonproper counterexamples where neither killing nor \(p\)-divisibility can occur by proper covers [1204.5831].

In modular-form theory, \(p\)-divisibility is tied to Fourier coefficients and degree-lowering via the Siegel operator. For Siegel modular forms, the paper on \(p\)-divisibility transposition shows that complete mod-\(p\) vanishing of positive-definite Fourier coefficients in degree \(n+1\) descends under \(\Phi\) to determinant-selective vanishing in degree \(n\). More precisely, the constructed forms satisfy
\[
a(F_k^{(n+1)},T)\equiv 0 \pmod p
\quad \text{for all } T\in \Lambda_{n+1}^+,
\]
while
\[
a(F_k^{(n)},T)\equiv 0 \pmod p
\quad \text{for } T\in \Lambda_n^+ \text{ with } \det(T)\not\equiv 0 \pmod p.
\]
The degree-1 endpoint recovers Wilton-type congruence phenomena for \(\tau(t)\) mod \(23\) as the shadow of a higher-degree pattern [2212.00917].

The Hermitian analogue has the same structure. For \(m\equiv 2\pmod 4\), \(p>m+3\), and \(D_K\not\equiv 0\pmod p\), there exist Hermitian modular forms \(G^{(m+1)}\) and \(G^{(m)}\) with
\[
\Phi(G^{(m+1)})=G^{(m)},
\]
such that all positive definite Fourier coefficients of \(G^{(m+1)}\) vanish mod \(p\), while in degree \(m\) the coefficients vanish mod \(p\) whenever \(\det(H^{(m)})\not\equiv 0\pmod p\). The proof uses explicit Hermitian Eisenstein coefficients, generalized Bernoulli numbers, Ikeda’s local polynomials, and a functional equation forcing local vanishing at a suitable prime \(q\) [2312.06318].

## 5. \(p\)-adic valuation frameworks: exponentials, codes, and representation counts

A large body of \(p\)-divisibility results is formulated as lower bounds on \(p\)-adic valuations. One abstract source is Dwork’s lemma for exponentials of power series. If
\[
H(z)=\exp(S(z)),
\]
then full integrality follows from the condition
\[
S(z^p)-pS(z)\in p\mathbf Z_p[[z]].
\]
The truncated versions show that weaker, finite-order control on \(S(z^p)-pS(z)\) still implies quantitative lower bounds on the valuations of the coefficients of \(H(z)\). In particular, if the Dwork condition is imposed only up to degree \(p^l\), the coefficients \(h_n\) of \(H(z)\) still satisfy explicit bounds of the form
\[
v_p(h_n)\ge \sum_{s=1}^{l-1}\left\lfloor \frac{n}{p^s}\right\rfloor -(l-m-1)\left\lfloor \frac{n}{p^l}\right\rfloor,
\]
under suitable hypotheses on the critical and higher coefficients. These bounds are then applied to subgroup-counting and permutation-representation problems [1412.7014].

In coding theory, divisibility is attached directly to Hamming weights. A linear code \(C\subseteq \mathbb F_q^n\) is \(\Delta\)-divisible if every codeword has weight divisible by \(\Delta\), and its \(p\)-adic valuation is
\[
\nu_p(C)=\max\{t\ge 0:\text{ every codeword of }C\text{ has weight divisible by }p^t\}.
\]
For trace codes \(\mathrm{Tr}_{q^m/q}(C)\), the exact value of \(\nu_p\) is given by a minimization formula involving base-\(p\) digit sums and Teichmüller power sums built from a generalized generator matrix over \(\mathbb F_{q^m}\). This extends Ward’s divisibility criterion from ordinary generator matrices over \(\mathbb F_q\) to trace constructions over extension fields and yields applications to abelian codes and Artin–Schreier equations [2605.19857].

A further valuation-theoretic instance is the number of linear representations of an Abelian \(p\)-group. For \(\#\operatorname{Hom}(G,GL_n(\mathbb F_q))\), the non-modular case \(p\nmid q\) exhibits essentially sharp linear lower bounds in \(n\), expressed in terms of the invariant factors of \(G\) and the \(p\)-adic behavior of \(q^{dp^i}-1\). In the modular cyclic case \(q=p^v\), the valuation becomes quadratic:
\[
v_p\bigl(\#\operatorname{Hom}(C_{p^u},GL_n(\mathbb F_{p^v}))\bigr)\ge
v\,\frac{p^u-1}{p^u+1}\binom n2,
\]
with equality when \(n\equiv 0,1\pmod{p^u+1}\). Here \(p\)-divisibility is not a factorization property but a growth law for the \(p\)-adic order of representation counts [1709.04829].

## 6. Related notions and terminological neighbors

Several neighboring literatures use “divisibility” in ways that are adjacent to, but distinct from, either quantum P-divisibility or arithmetic \(p\)-divisibility. On Markoff-like cubic surfaces over \(\mathbb F_p\), the divisibility statement concerns orbit sizes under Vieta involutions: for generic parameters with \(s=3+a_1+a_2+a_3\neq 0\) and \(a_i^2\neq 4\), every nontrivial orbit has size divisible by \(p\). The proof is an orbit-averaging argument based on functions \(\Delta_i\) satisfying
\[
\Delta_1+\Delta_2+\Delta_3=s,
\qquad
\Delta_i(x)+\Delta_i(m_i x)=s,
\]
which forces \(s|\mathcal O|=0\) in \(\mathbb F_p\) [2509.02187].

In combinatorial \(q\)-series, the phrase may indicate congruence rather than factorization. For the type \(2\) \((p,q)\)-analogue of the \(r\)-Whitney numbers of the second kind,
\[
W^*_{m,r}[n,k;t]_{p,q}\equiv \binom{n}{k}p^{(n-k)(mt+r-1)} \pmod{pq},
\]
so the two-parameter structure collapses modulo \(pq\) to a binomial coefficient times a pure power of \(p\). The paper explicitly treats this as polynomial congruence modulo the ideal generated by \(pq\), not as ordinary divisibility by a fixed integer \(p\) [2208.04527].

Graph theory uses yet another vocabulary. A graph is 2-divisible if every induced subgraph can be partitioned into two parts with strictly smaller clique number, and perfectly divisible if every induced subgraph can be partitioned into a perfect induced subgraph and a remainder of smaller clique number. Every \((P_5,C_5)\)-free graph is 2-divisible, while every bull-free graph that is either odd-hole-free or \(P_5\)-free is perfectly divisible. These are structurally different notions, but they illustrate how “divisibility” often denotes recursive decomposability rather than arithmetic divisibility [1704.06667].

The same caution applies to partition congruences. The function \(PD(n)\), the number of partitions with designated summands, satisfies
\[
PD(3n+2)\equiv 0\pmod 3,
\]
and the paper proves the stronger identity
\[
\sum_{n=0}^\infty PD(3n+2)q^n
=
3\,\frac{(q^3;q^6)_\infty^3 (q^6;q^6)_\infty^6}{(q;q^2)_\infty^5 (q^2;q^2)_\infty^8}.
\]
Here the divisibility phenomenon is a Ramanujan-type congruence for coefficients of a generating function, not a dynamical or local-global divisibility principle [1208.2210].

Taken together, these usages suggest that “P-divisibility” is best treated as a context-sensitive term. In open quantum systems it is a precise positivity condition on intermediate propagators. In arithmetic and adjacent areas, the lowercase form \(p\)-divisibility usually refers instead to prime-power divisibility, \(p\)-adic valuation, or congruence modulo \(p\). The unifying theme is not a single formal definition, but the presence of a divisibility constraint controlled by a deeper ambient structure—positivity, Galois action, alterations, Fourier expansion, or valuation theory.

Source: https://www.emergentmind.com/topics/p-divisibility