--- title: Mean Subtree Order in Graphs url: https://www.emergentmind.com/topics/mean-subtree-order type: topic --- # Mean Subtree Order in Graphs Mean subtree order is the average number of vertices in a subtree, and it sits at the intersection of enumerative graph theory, extremal graph theory, and generating-function methods. In the classical tree setting, a subtree is a connected induced subgraph, so the invariant measures the first moment of the subtree-size distribution. Two later extensions coexist in the literature. One extends the induced-subgraph viewpoint from trees to all connected graphs by averaging over connected induced vertex sets; the other, used for general graphs, counts all tree subgraphs, not necessarily induced. This terminological bifurcation is central: many structural theorems are parallel across the two settings, but edge-addition and extremal phenomena can differ substantially [2103.15174] [2010.15522] [2508.20593]. ## 1. Definitions and competing conventions For a tree \(T\), the standard definition counts induced connected subgraphs. If \(S_k(T)\) denotes the number of subtrees of order \(k\), then the mean subtree order is \[ \mathrm{M}(T)=\frac{\sum_{k\ge 1} k\,S_k(T)}{\sum_{k\ge 1} S_k(T)}. \] The associated generating polynomial is written as \(F_T(x)=\sum_{k\ge 1} S_k(T)x^k\), giving \[ \mathrm{M}(T)=\frac{F_T'(1)}{F_T(1)}. \] This convention underlies the classical work of Jamison and most of the extremal theory for trees [2010.15522]. For connected graphs, one important extension replaces subtrees by connected induced subgraphs. If \(c_k(G)\) is the number of connected induced subgraphs of order \(k\), then \[ M(G)=\frac{\sum_{k=1}^n k\,c_k(G)}{\sum_{k=1}^n c_k(G)}. \] Writing \(C_G(z)=\sum_{k=1}^n c_k(G)z^k\), one has \(M(G)=C_G'(1)/C_G(1)\). In this setting the quantity is also called the mean connected induced subgraph order or mean CIS order [2103.15174] [1811.05430]. A different general-graph convention counts all tree subgraphs. If \(s_k(G)\) is the number of \(k\)-vertex subtrees of a connected graph \(G\), then \[ \mu(G)=\frac{\sum_{k=1}^n k\,s_k(G)}{\sum_{k=1}^n s_k(G)}=\frac{S_G'(1)}{S_G(1)}, \] where \(S_G(x)=\sum_{k=1}^n s_k(G)x^k\). Distinct edge sets on the same vertex set are counted separately, and subtrees are not required to be induced [2508.20593] [2308.12808] [1911.05794]. | Setting | Objects counted | Mean formula | |---|---|---| | Tree \(T\) | induced connected subgraphs of \(T\) | \(\mathrm{M}(T)=F_T'(1)/F_T(1)\) | | Connected graph \(G\), CIS convention | connected induced subgraphs | \(M(G)=C_G'(1)/C_G(1)\) | | Connected graph \(G\), subtree-subgraph convention | connected acyclic subgraphs | \(\mu(G)=S_G'(1)/S_G(1)\) | Local variants are equally important. For trees, \(\mathrm{M}_v(T)\) denotes the average order of all subtrees containing a fixed vertex \(v\), and more generally \(\mu_T(S)\) denotes the average order of all subtrees containing a fixed subtree \(S\). Normalized versions are called densities: \(D_T=\mu_T/n\) for the global tree density, \(D(G)=M(G)/n\) for connected induced subgraphs of a graph, and \(D_T(S)=(\mu_T(S)-|S|)/(n-|S|)\) for local density at a subtree \(S\) [2010.15522] [2103.15174] [2306.13422]. ## 2. Classical tree theory and the path minimum The foundational extremal theorem is due to Jamison: among all trees of order \(n\), the path \(P_n\) uniquely minimizes the mean subtree order. Explicitly, \[ M(T)\ge \frac{n+2}{3}, \] with equality if and only if \(T\cong P_n\). Since subtrees of \(P_n\) are exactly contiguous vertex intervals and there are \(n-k+1\) of order \(k\), one obtains \[ M(P_n)=\frac{n+2}{3}. \] This is the baseline lower bound for the subject [2103.15174] [1811.05430]. A later contraction theorem gives a second route to the same bound. If \(e\) is any edge of a tree \(T\), then contracting \(e\) produces a tree \(T/e\) satisfying \[ \mu_T-\mu_{T/e}\ge \frac13, \] with equality if and only if \(T\) is a path. The inequality is strict whenever \(e\) does not lie on a pendant path. Repeated contraction to a single vertex immediately yields \[ \mu(T)\ge 1+\frac{n-1}{3}=\frac{n+2}{3}, \] again with equality only for \(P_n\) [2310.17757]. The same minimization result admits a structural proof via local transformations. The Gluing Lemma shows that when a rooted tree is attached to a path, the mean subtree order is maximized when the attachment point is central; the Strong Gluing Lemma strengthens this to strict monotonicity as the gluing point moves inward. These tools yield a conceptually simple proof that every non-path tree has a standard \(1\)-associate with strictly smaller mean subtree order, and iteration leads to the path [1707.01874]. The path theorem is sometimes treated as merely a tree result. That is misleading. It is the prototype for several later generalizations, and the sharp constant \((n+2)/3\) persists in broader settings [2103.15174] [2508.20593]. ## 3. Maximal mean subtree order in trees The opposite extremal direction is structurally subtler. Jamison conjectured that among all trees on \(n\) vertices, a caterpillar attains the maximum mean subtree order. This Caterpillar Conjecture remains open, but the asymptotic order of magnitude is known: \[ \max_{|T|=n}\mu_T = n-2\log_2 n + O(1). \] More precisely, if \(f(x)=x-2^x\) on \([0,1]\), extended \(1\)-periodically, then \[ n-2\log_2 n + f(2\log_2 n)+o(1)\le \max_{|T|=n}\mu_T \] and \[ \max_{|T|=n}\mu_T\le n-2\log_2 n + 2 + f(2\log_2 n)+o(1). \] Thus the global maximum is pinned down to within \(2+o(1)\) [2010.15522]. The local extremum is sharper. If \(\mu_T(r)\) is maximal over all trees \(T\) of order \(n\) and vertices \(r\in V(T)\), then \(T\) must be a broom and \(r\) is the end of the handle opposite the leaves. Consequently, \[ \max_{|T|=n,\;v\in V(T)} \mu_v(T) = n-\log_2 n+\tfrac12 f(2\log_2 n)+o(1). \] This local result explains why broom-like and double-broom-like structures recur in extremal constructions [2010.15522]. Optimal or near-optimal trees are highly elongated. There exists a constant \(C\) such that if a tree \(T\) of order \(n\) has diameter \(d\), then \[ \mu_T \le n-\log_2 n-2\log_2(n-d)+C. \] As a corollary, if \(\hat T_n\) is optimal, then \(n-\operatorname{diam}(\hat T_n)\) is \(o(n^{1/2+\delta})\) for every \(\delta>0\). The same paper shows that an optimal tree has \(\Theta(n^4)\) subtrees, and its central part \(C(T)\) induces a connected subtree with at least \(n-n^{1/3}\) vertices and at most \(16\) leaves [2010.15522]. Several structural refinements are known. If \(T\) is optimal in \(\mathcal{T}_n\) or in the caterpillar family \(\mathcal{C}_n\) and \(n\ge 4\), then every leaf is adjacent to a vertex of degree at least \(3\); equivalently, every limb has order \(1\). In \(\mathcal{C}_n\), the number of leaves in an optimal caterpillar is \(\Theta(\log_2 n)\), while in \(\mathcal{T}_n\) the best general bound is \(O(\log_2 n)\) [1707.01874]. Degree constraints force a different asymptotic regime. For trees whose internal vertices all have degree at least \(3\), the density satisfies \[ \frac12 \le D(T) < \frac34, \] and both bounds are tight. For series-reduced trees, a sequence has density tending to \(1/2\) if and only if the proportion of leaves tends to \(1\), and tends to \(3/4\) if and only if the leaf proportion tends to \(1/2\) while the twig proportion tends to \(0\) [1312.7233] [2103.15174]. ## 4. Local mean, local density, and anchored subtree structure Local mean order often reveals more structure than the global average. Jamison’s local monotonicity states that for vertex sets \(A\subseteq B\), \[ \mu_T(A)\le \mu_T(B), \] with equality if and only if the smallest subtree containing \(A\) equals that containing \(B\). In particular, \[ \mu_T(v)\ge \mu_T \] for every vertex \(v\), with strict inequality when \(|T|>1\). A further lower bound states that for any subtree \(S\subseteq T\), \[ \mu_T(S)\ge \frac{|S|+|T|}{2}. \] These inequalities make local means a natural vehicle for proving global bounds [2010.15522] [2306.13422]. A central technical tool is the index. For a rooted tree \(T_v\), \[ i(T_v)=\frac{R_{T_v}(v)}{N_{T_v}(v)\bigl(1+N_{T_v}(v)\bigr)}, \] and more generally \(i(v;S)\) measures the effect of adding or deleting a leaf relative to a fixed subtree \(S\). The Index Lemma gives exact increment formulas: \[ \mu_T(S+w)=\mu_T(S)+i(w;S) \] for a neighbor \(w\) of \(S\), and \[ \mu_T(S-v)=\mu_T(S)-i(v;S) \] for a leaf \(v\) of \(S\). The bound \(0 \frac13 \] for every connected graph. The proof proceeds by induction on \(n\) and combines several ingredients: the conditioned lower bound \[ A(G,H)\ge \frac{n+h}{2} \] for a connected subset \(H\) of size \(h\); the inequality \[ av(G,x)\bigl(N(G,x)-1\bigr)\ge N(G-x); \] upper bounds \(av(G,x)\le (n-1)/2\) in \(2\)-connected regions; a cut-vertex inequality controlling the product \(\prod_i N(G_i',x)\); and the identity \[ S(G)=\sum_{x\in V(G)} N(G,x). \] Together these show that the path is the unique minimizer not only among trees or block graphs, but among all connected graphs under the connected-induced-set convention [2103.15174]. Canonical examples illustrate the scale of the invariant. Under the connected-induced-set convention, \(M(P_n)=(n+2)/3\); for the cycle, \[ M(C_n)=\frac{\frac{n^2(n-1)}2+n}{n(n-1)+1}\sim \frac n2; \] for the star \(K_{1,n-1}\), \[ M(K_{1,n-1})\sim \frac{n+1}{2}; \] and for the complete graph, \[ M(K_n)=\frac{n\,2^{n-1}}{2^n-1}\sim \frac n2. \] High connectivity and large blocks therefore push the average well above the path bound [2103.15174]. Under the non-induced subtree-subgraph convention, the same path minimum has now also been confirmed for connected graphs: \[ \mu(G)\ge \frac{n+2}{3}, \] with equality only for \(P_n\). In this convention the complete graph is conjectured to maximize the mean. Cayley’s formula yields \[ s_k(K_n)=\binom{n}{k}k^{k-2}, \] hence \[ \mu(K_n)= \frac{\sum_{k=1}^n \binom{n}{k}k^{k-1}} {\sum_{k=1}^n \binom{n}{k}k^{k-2}} = n-\frac1e+o(1). \] A uniformly random subtree of \(K_n\) therefore has size \(n-\Theta(1)\), with expected deficit tending to \(1/e\) [2508.20593]. Several graph classes admit stronger lower bounds than the universal \((n+2)/3\). For connected cographs of order \(n\), \[ \frac n2 < A(G)\le \frac{n+1}{2}, \] with equality on the right only for \(n=1\). This suggests that forbidding long induced paths or forcing dense module structure raises the mean substantially above the path threshold [2103.15174]. ## 6. Local mean order beyond trees: \(k\)-trees and related structures The theory extends naturally to \(k\)-trees. A sub-\(k\)-tree of a \(k\)-tree \(T\) is a subgraph that is itself a \(k\)-tree, and the global mean order is \[ \mu(T)=\frac{R(T)}{N(T)}. \] For a fixed \(k\)-clique \(C\), the local mean \(\mu(T;C)\) averages over all sub-\(k\)-trees containing \(C\). The shifted local mean \(\mu^\bullet(T;C)=\mu(T;C)-k\) is often used because the clique \(C\) is mandatory in every counted object [2309.16545]. The principal local extremal theorem is that for \(k\ge 2\), if a \(k\)-clique \(C\) maximizes \(\mu(T;C)\) in a \(k\)-tree \(T\) of order \(n\ne k+2\), then \(C\) has degree \(1\). A related formulation states that the maximum local mean order occurs at a \(k\)-clique that is not a major \(k\)-clique, so a maximizer must have degree \(1\) or \(2\). The proof passes through the characteristic \(1\)-tree \(T'_C\), for which \[ \mu(T;C)=\mu(T'_C;C)+k-1, \] and then uses tree-level Kelmans and partial Kelmans operations to move away from major \(k\)-cliques while increasing local mean [2309.16545] [2309.11885]. A second global-local comparison generalizes Jamison’s tree inequality. For any \(k\)-tree \(T\) and any \(k\)-clique \(C\), \[ \mu(T;C)<2\,\mu(T). \] This bound is sharp, with \(k\)-brooms providing asymptotically extremal examples. In the same framework, if \(T\) has no \(k\)-clique of degree \(2\), then \[ \frac{n+k}{2}-o_n(1)<\mu(T)<\frac{3n+k-3}{4}, \] and for sufficiently large \(n\) the \(k\)-star is the unique minimizer of \(\mu(T)\) in this class [2309.16545]. These \(k\)-tree results are not merely analogical. They show that the mean subtree order belongs to a broader family of clique-anchored mean-order invariants controlled by recursive decompositions, characteristic trees, and local-global inequalities. A plausible implication is that many structural phenomena first observed for trees are manifestations of a more general elimination-order geometry, although a full transfer to arbitrary chordal graphs is not stated in the cited work. ## 7. Edge operations, nonmonotonicity, and open problems A persistent misconception is that adding edges should always increase mean subtree order. Under the non-induced subtree-subgraph convention this is false. There are connected graphs \(G\subset H\) on the same vertex set with \(\mu(H)<\mu(G)\), and the decrease from adding a single edge can be as large as \[ \left(\frac13-o(1)\right)n \] asymptotically. The construction uses a long path with large leaf-stars at its ends; adding one closing edge creates a family in which almost all subtrees contain the new edge and have mean order asymptotic to \((2/3)n\), whereas the original tree has mean close to \(n\) [1911.05794]. The nonmonotonicity is robust. For every positive integer \(k\), there exist infinitely many pairs of connected graphs \(G\subset H\) with \(|E(H)\setminus E(G)|=k\) such that \(\mu(H)<\mu(G)\). The same work confirms a conjectural comparison \(\mu(K_m+nK_1)<\mu(K_{m,n})\) for fixed \(m\) and sufficiently large \(n\), again illustrating that adding edges can lower the mean rather than raise it [2308.12808]. Even the weaker statement that some edge addition must help is delicate. It is proved for trees: every tree \(T\) of order at least \(3\) has nonadjacent vertices \(v,w\) such that \[ M(T+vw)>M(T). \] But universal edge-addition monotonicity fails, and edge deletion can also move the mean in either direction in general graph families [1911.05794] [2508.20593]. Several open problems remain central. Under the non-induced subtree-subgraph convention, the complete graph \(K_n\) is conjectured to maximize \(\mu(G)\) among connected graphs of order \(n\). A reduction shows that it would suffice to prove \[ s_{n-1}(G)\ge \frac{s_{n-1}(K_n)}{s_n(K_n)}\,s_n(G), \] which would imply both \(\mu(G)\le \mu(K_n)\) and an upper bound on the probability that a random subtree is spanning [2508.20593]. Under the connected-induced-set convention, two problems highlighted after the path-minimum theorem ask whether minimum degree at least \(3\) forces \(D(G)>1/2\), and whether there is an absolute upper bound less than \(1\) on \(D(G)\) for graphs with minimum degree at least \(3\) [2103.15174]. Taken together, these results position mean subtree order as a sharply structured, but convention-sensitive, graph invariant. In trees it is controlled by path-minimization, broom-like local maximizers, and near-caterpillar global maximizers. In connected graphs it admits both induced and non-induced extensions, each with a unique path minimizer but with different monotonicity behavior under edge operations. The invariant therefore serves simultaneously as a measure of combinatorial cohesion and as a test case for how local substructure counts propagate through global graph architecture. Source: https://www.emergentmind.com/topics/mean-subtree-order