Abstract
We study interpretations of the Tutte and characteristic polynomials of matroids. If M is a matroid with rank function r whose ground set E is given with a linear ordering <, then \(X\subseteq E\) is called \((M,<)\)-compatible if \(X\cap C\ne \{\min (C)\}\) for each circuit C of M. We show that the Tutte polynomial of M equals \(\sum x^{r(M/X)}y^{r^*(M|X)}\) where X runs through the subsets of E such that X and \(E{{\setminus }}X\) are \((M^*,<)\)- and \((M,<)\)-compatible, respectively. Similarly, the characteristic polynomial of M equals \(\sum (-1)^{|X|}(k-1)^{r(M/X)}\) where X runs either through \((M^*,<)\)-compatible subsets of E, or through the independent sets of M such that X and \(E{{\setminus }}X\) are \((M^*,<)\)- and \((M,<)\)-compatible, respectively.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Let M be a matroid on a finite set E with rank function r. The Tutte polynomial of M is (see [4])
This invariant was introduced in [22] and encodes many properties of graphs and matroids. Applications in combinatorics, knot theory, statistical physics, and coding theory are surveyed in [3,4,5,6,7, 11, 24, 25]. Evaluating the Tutte polynomial at a point or finding its coefficients is in general a \(\sharp P\)-hard problem [2, 14], even for planar graphs [23], and evaluations are in general difficult even to approximate [10]. T(M; x, y) is fully characterized by the following rules
The duality formula is
and the convolution formula (see [8, 17, 18]) is
where \(M|Z=M-(E{{\setminus }}Z)\).
The characteristic polynomial of M is (see [4, 26])
This generalizes chromatic and flow polynomials of graphs (see [4, 25, 26]). Relations with some other combinatorial structures are studied in [1, 26]. Polynomial p(M; k) is fully characterized by the following rules
2 Interpretations
We recall some basic properties of matroids. If \(X\subseteq E\), we shall usually write \(r(M{-}X)\), r(M / X), \(r^*(M{-}X)\), \(r^*(M/X)\) for \(r_{M-X}(E{\setminus } X)\), \(r_{M/X}(E{\setminus } X)\), \(r_{(M-X)^*}(E{\setminus } X)\), \(r_{(M/X)^*}(E{\setminus } X)\), respectively. For any \(X,Y\subseteq E\) (see cf. [21]),
Denote by \(I_M\) the set of isthmuses of M. If \(X\subseteq E\) and \(e\in E\), \(e\notin I_M, X\), then \(r((M{-}e)/X)= r_{(M-e)}(E{{\setminus }}\{e\})-r_{(M-e)}(X) =r_{M}(E{{\setminus }}\{e\})-r_{M}(X)= r_{M}(E)-r_{M}(X)= r(M/X)\), i.e.,
Let \({\mathcal {C}}(M)\) denote the family of circuits of M. By [21, Proposition 3.1.1],
It is an easy exercise to prove that (see cf. Exercise 2 in [21, Section 3.1])
Let < be a linear ordering of E. For each nonempty \(X\subseteq E\), we denote by \(\min (X)\) and \(\max (X)\) the minimal and maximal element of X with respect to <, respectively.
We say that \(X\subseteq E\) is \((M,<)\)-compatible if for each \(C\in {\mathcal {C}}(M)\), \(C\cap X\ne \{\min (C)\}\). Clearly, no \((M,<)\)-compatible set can contain a loop of M.
Denote by \({\mathcal {E}}(M,<)\) the family of all \((M^*,<)\)-compatible subsets of E, by \({\mathcal {D}}(M,<)=\{X\in {\mathcal {E}}(M,<);E{{\setminus }}X\in {\mathcal {E}}(M^*,<)\}\), and by \({\mathcal {P}}(M,<)\) the set of couples (X, Y) such that \(X,Y\subseteq E\), \(X\cap Y=\emptyset \), X is \((M^*,<)\)-compatible, Y is \((M,<)\)-compatible, and X and Y are maximal with this property (i.e., for each \(e'\in E{\setminus }(X\cup Y)\), \(X\cup \{e'\}\) is not \((M^*,<)\)-compatible and \(Y\cup \{e'\}\) is not \((M,<)\)-compatible). Clearly, \(I_{M^*}\subseteq X\) and \(I_{M}\subseteq Y\) for each \((X,Y)\in {\mathcal {P}}(M,<)\). Furthermore, \(\{(X,E{{\setminus }}X);X\in {\mathcal {D}}(M,<)\}\subseteq {\mathcal {P}}(M,<)\), but the inclusion is an equality because in the following statement we prove that \({\mathcal {D}}(M,<)\) and \({\mathcal {P}}(M,<)\) have the same cardinality.
Theorem 1
Let < be a linear ordering of elements of a matroid M. Then \(|{\mathcal {D}}(M,<)|=|{\mathcal {P}}(M,<)|=T(M;1,1)\) and
Proof
In order to unify and simplify notation, denote by \({\mathcal {P}}_1(M,<)={\mathcal {P}}(M,<)\), \({\mathcal {P}}_2(M,<)=\{(X,E{{\setminus }}X);X\in {\mathcal {D}}(M,<)\}\), and by \(g(M,<;x,y)\) the right hand side of (12), i.e.,
Clearly, \({\mathcal {P}}_2(M,<)\subseteq {\mathcal {P}}_1(M,<)\) and the equality occurs if and only if \(|{\mathcal {P}}_1(M,<)|=|{\mathcal {P}}_2(M,<)|\). We use induction on \(|E{\setminus }(I_{M^*}\cup I_{M})|\) to prove that \(T(M;x,y)=g(M,<;x,y)\) and \(|{\mathcal {P}}_i(M,<)|=T(M;1,1)\) for \(i=1,2\).
If \(E=I_{M^*}\cup I_{M}\), then \({\mathcal {P}}_i(M,<)=\{(I_{M^*},I_{M})\}\), \(|{\mathcal {P}}_i(M,<)|=1=T(M;1,1)\) (\(i=1,2\)), and \(T(M;x,y)=x^{|I_{M}|}y^{|I_{M^*}|}=g(M,<;x,y)\).
If \(E\ne I_{M^*}\cup I_{M}\), choose \(e=\max \left( E{\setminus }(I_{M^*}\cup I_{M})\right) \) and define
Suppose that \((X,Y)\in {\mathcal {P}}_1(M{-}e,<)\) and \(e'\in E{\setminus }(X\cup Y\cup \{e\})\).
-
(i)
Let \(C\in {\mathcal {C}}(M^*)\). If \(C{\setminus }\{e\}\in {\mathcal {C}}(M^*/e)\), then \(e\notin X\), \(\min (C)=\min (C{{\setminus }}\{e\})\), and \(X\cap C=X\cap (C{{\setminus }}\{e\})\ne \{\min (C{{\setminus }}\{e\})\}\). If \(C=\bigcup _{i=1}^nC_i\), \(C_i\in {\mathcal {C}}(M^*/e)\), then \(X\cap C_i\ne \{\min (C_i)\}\), \(i=1,\dots ,n\), \(\min (C)\in \{\min (C_i);i=1,\dots ,n\}\), whence \(X\cap C\ne \{\min (C)\}\). Thus by (11), \(X\cap C\ne \{\min (C)\}\). Since this holds for each \(C\in {\mathcal {C}}(M^*)\), X is \((M^*,<)\)-compatible.
-
(ii)
If \(C\in {\mathcal {C}}(M)\) and \(e\in C\) (resp. \(e\notin C\)), then \(e\in (Y{\cup }\{e\})\cap C\ne \{\min (C)\}\) because \(e\ne \min (C)\) (resp. \((Y{\cup }\{e\})\cap C=Y\cap C\ne \{\min (C)\}\)), i.e., \(Y\cup \{e\}\) is \((M,<)\)-compatible.
-
(iii)
If \(X\cup \{e'\}\) is not \((M^*/e,<)\)-compatible, there exists \(C'\in {\mathcal {C}}(M^*/e)\) such that \(C'\cap (X{\cup }\{e'\})=\{\min (C')\}\), thus \(e'=\min (C')\ne e\). If \(C'{\cup }\{e\}\in {\mathcal {C}}(M^*)\) (resp. \(C'\in {\mathcal {C}}(M^*)\)) then \(e\notin X\), \(\min (C')=\min (C'{\cup }\{e\})\), and \((C'{\cup }\{e\})\cap (X{\cup }\{e'\})=C'\cap (X{\cup }\{e'\})=\{\min (C'{\cup }\{e\})\}\) (resp. \(C'\cap (X{\cup }\{e'\})=\{\min (C')\}\)), whence by (10), \(X\cup \{e'\}\) is not \((M^*,<)\)-compatible.
-
(iv)
If \(Y\cup \{e'\}\) is not \((M{-}e,<)\)-compatible, there exists \(C'\in {\mathcal {C}}(M{-}e)\) such that \(C'\cap (Y{\cup }\{e,e'\})=C'\cap (Y{\cup }\{e'\})=\{\min (C')\}\), i.e., \(Y\cup \{e,e'\}\) is not \((M,<)\)-compatible.
By (i)–(iv), \((X,Y{\cup }\{e\}))\in {\mathcal {P}}'_1(M,<)\). From (i) and (ii), it also follows that if \((X,Y)\in {\mathcal {P}}_2(M{-}e,<)\) then \((X,Y{\cup }\{e\})\in {\mathcal {P}}'_2(M,<)\).
Suppose that \((X,Y')\in {\mathcal {P}}'_1(M,<)\) and \(e'\in E{\setminus }(X\cup Y')\).
-
(v)
If \(C\in {\mathcal {C}}(M^*/e)\) and \(C\cup \{e\}\in {\mathcal {C}}(M^*)\) (resp. \(C\in {\mathcal {C}}(M^*)\)), then \(\min (C)=\min (C{\cup }\{e\})\), and \(X\cap (C{\cup }\{e\})=X\cap C\ne \{\min (C{\cup }\{e\})\}\) (resp. \(X\cap C\ne \{\min (C)\}\)), whence by (10), X is \((M^*/e,<)\)-compatible.
-
(vi)
If \(C\in {\mathcal {C}}(M{-}e)\), then \((Y'{{\setminus }}\{e\})\cap C=Y'\cap C\ne \{\min (C)\}\), i.e., \(Y'{{\setminus }}\{e\}\) is \((M{-}e,<)\)-compatible.
-
(vii)
If \(X\cup \{e'\}\) is not \((M^*,<)\)-compatible, there exists \(C'\in {\mathcal {C}}(M^*)\) such that \(C'\cap (X{\cup }\{e'\})=\{\min (C')\}\), whence \(e'=\min (C')\ne e\). If \(C'{{\setminus }}\{e\}\in {\mathcal {C}}(M^*/e)\), then \((C'{{\setminus }}\{e\})\cap (X{\cup }\{e'\})=\{\min (C')\}\). If \(C'=\bigcup _{i=1}^nC'_i\), \(C'_i\in {\mathcal {C}}(M^*/e)\), there exists j such that \(\min (C')\in C'_j\), whence \(\min (C'_j)=\min (C')\) and \(C'_j\cap (X{\cup }\{e'\})=\{\min (C'_j)\}\). Thus by (11), \(X\cup \{e'\}\) is not \((M^*/e,<)\)-compatible.
-
(viii)
If \(Y'\cup \{e'\}\) is not \((M,<)\)-compatible, there exists \(C'\in {\mathcal {C}}(M)\) such that \(C'\cap (Y'{\cup }\{e'\})=\{\min (C')\}\), thus \(e'=\min (C')\ne e\), \(e\notin C'\) (because \(e\in Y'\)), and \(C'\cap ((Y'{{\setminus }}\{e\})\cup \{e'\})=\{\min (C')\}\), i.e., \((Y'{{\setminus }}\{e\})\cup \{e'\}\) is not \((M{-}e,<)\)-compatible.
By (v)–(viii), \((X,Y'{\setminus }\{e\}))\in {\mathcal {P}}_1(M{-}e,<)\). From (v) and (vi), it also follows that if \((X,Y')\in {\mathcal {P}}'_2(M,<)\) then \((X,Y'{\setminus }\{e\})\in {\mathcal {P}}_2(M{-}e,<)\).
Therefore \({\mathcal {P}}'_i(M,<)=\{(X,Y\cup \{e\});(X,Y)\in {\mathcal {P}}_i(M{-}e,<)\}\) (\(i=1,2\)). For any \((X,Y')\in {\mathcal {P}}'_2(M,<)\), we have \(e\notin I_M, X\) whence by (9), \(r((M{-}e)/X)= r(M/X)\) and clearly, \(M-Y'=(M{-}e)-(Y'{{\setminus }}\{e\})\). Thus by induction hypothesis, \(|{\mathcal {P}}'_i(M,<)|=|{\mathcal {P}}_i(M{-}e,<)|=T(M{-}e;1,1)\) (\(i=1,2\)) and
The latter equation holds true also for \(M^*\), whence by (3),
Since \({\mathcal {P}}_i(M^*,<)=\{(Y,X);(X,Y)\in {\mathcal {P}}_i(M,<)\}\) and for each \((X,Y)\in {\mathcal {P}}_i(M,<)\), \(r(M/X)=r^*(M^*{-}X)\), \(r^*(M{-}Y)=r(M^*/Y)\), and \(e\in X\cup Y\), we have
Thus by (2), \(|{\mathcal {P}}_i(M,<)|=T(M;1,1)\) (\(i=1,2\)) and \(T(M;x,y)=g(M,<;x,y)\). \(\square \)
Denote by \({\mathcal {S}}(M,<)=\{X\in {\mathcal {D}}(M,<);r(X)=|X|\}\).
Theorem 2
Let < be a linear ordering of elements of a matroid M. Then \(|{\mathcal {E}}(M,<)|=T(M;1,2)\), \(|{\mathcal {S}}(M,<)|=T(M;1,0)\), and
Proof
We use induction on \(|E{\setminus } I_M|\) to prove the first equation from (13) and that \(|{\mathcal {E}}(M,<)|=T(M;1,2)\). If \(E=I_M\), then \({\mathcal {E}}(M,<)=\{\emptyset \}\), \(|{\mathcal {E}}(M,<)|=1=T(M;1,2)\), \(r(M)=|I_M|\), and \(p(M;k)=(k-1)^{r(M)}\) as claimed.
If \(E\ne I_M\), choose \(e=\max (E{\setminus } I_M)\) and denote by \({\mathcal {E}}^+=\{X\in {\mathcal {E}}(M,<);\, e\notin X\}\), \({\mathcal {E}}^-=\{X\in {\mathcal {E}}(M,<);\, e\in X\}\).
If \(X\in {\mathcal {E}}(M{-}e,<)\), then by item (i) from the proof of Theorem 1, X is \((M^*,<)\)-compatible, i.e., \({\mathcal {E}}(M{-}e,<)\subseteq {\mathcal {E}}^+\). Similarly, the reverse implication follows from item (v). Thus \({\mathcal {E}}(M{-}e,<)={\mathcal {E}}^+\).
If \(X\in {\mathcal {E}}(M/e,<)\), then for each \(C\in {\mathcal {C}}(M^*)\) satisfying \(e\in C\) (resp. \(e\notin C\)), we have \(e\in (X{\cup }\{e\})\cap C\ne \{\min (C)\}\), because \(e\ne \min (C)\) (resp. \((X{\cup }\{e\})\cap C=X\cap C\ne \{\min (C)\}\)), whence \(X\cup \{e\}\) is \((M^*,<)\)-compatible. If \(X'\in {\mathcal {E}}^-\), then for each \(C\in {\mathcal {C}}(M^*{-}e)\), \((X'{{\setminus }}\{e\})\cap C=X'\cap C\ne \{\min (C)\}\), whence \(X'{\setminus }\{e\}\) is \((M^*{-}e,<)\)-compatible. Thus \({\mathcal {E}}(M/e,<)=\{X'{{\setminus }}\{e\};X'\in {\mathcal {E}}^-\}\).
By (6) and induction hypothesis,
because by (9), \(r((M{-}e)/X)=r(M/X)\) for each \(X\in {\mathcal {E}}(M{-}e,<)={\mathcal {E}}^+\). Furthermore, \(|{\mathcal {E}}(M,<)|= |{\mathcal {E}}^+|+|{\mathcal {E}}^-|= |{\mathcal {E}}(M{-}e,<)|+|{\mathcal {E}}(M/e,<)|\) and \(M{-}e=M/e\) if e is a loop of M, thus by the last two rows of (2), \(|{\mathcal {E}}(M,<)|=T(M;1,2)\).
We prove the second part of (13). By (7), \(r^*(M|X)=|X|-r(X)\) whence \({\mathcal {S}}(M,<)=\{X\in {\mathcal {D}}(M,<);\,r^*(M|X)=0\}\). Moreover by (8), \(r(M/X)=r(M)-r(X)=r(M){-}|X|\) for each \(X\in {\mathcal {S}}(M,<)\). From (5) and (12), we have
Hence \(p(M;0)=|{\mathcal {S}}(M,<)|({-}1)^{r(M)}\) and by (5), \(|{\mathcal {S}}(M,<)|=T(M;1,0)\). \(\square \)
If \(I_{M^*}\ne \emptyset \) then \({\mathcal {S}}(M,<)=\emptyset \) (because \(I_{M^*}\subseteq X\) for each \(X\in {\mathcal {S}}(M,<)\) and \(r(I_{M^*})=0\)) and by (13), \(p(M;k)=0\).
Clearly, \({\mathcal {S}}(M,<)\subseteq {\mathcal {E}}(M,<)\), whence by (13),
Corollary 1
Let < be a linear ordering of elements of a matroid M. Then
Proof
By (7) and (8), \(r(M/Z)+r^*(M|Z)=r_{M/Z}(E{\setminus } Z)+r_{(M|Z)^*}(Z)= r_M(E)-r_M(Z)+|Z|+r_{M|Z}(\emptyset )-r_{M|Z}(Z)= r(M)+|Z|-2r(Z)+0\), whence
and applying the first equation from (13) for \(p(M/Z;1{-}x)\) and \(p((M|Z)^*;1{-}y)\) we get the first part of (15). By the second equation from (13),
and the second part of (15) follows from the fact that \(r(M)+|Z|+r(M/Z)+r^*(M|Z)=r(M)+|Z|+r(M)-r(Z)+|Z|+r(\emptyset )-r(Z)\) is even. \(\square \)
For example, suppose that \(M=U_{1,n}\), \(M^*=U_{n-1,n}\), and \(E=\{e_1,\dots ,e_n\}\) has a linear ordering < such that \(e_1<\dots <e_n\). Denote by \(2^{E}\) the powerset of E, \(S_i=\{e_1,\dots ,e_i\}\), and \(S^*_i=E{{\setminus }}S_i\) (\(i=0,\dots ,n\), \(S^*_n=S_0=\emptyset \)). We can check that
In accordance with Theorems 1, 2, and (3),
and in accordance with (14),
It would be interesting to study relations between \({\mathcal {E}}(M,<)\), \({\mathcal {D}}(M,<)\), \({\mathcal {P}}(M,<)\), and \({\mathcal {S}}(M,<)\) and other structures, that have the same cardinalities or are known to give expressions to compute the Tutte polynomial or the characteristic polynomial. Obvious examples are the bases, independent sets, internal and external activities (see [5, 22]), broken circuits (sets of the form \(C{\setminus }\{\min (C)\}\), \(C\in {\mathcal {C}}(M)\)), and the partitioning of \(2^E\) into T(M; 1, 1) intervals of type \([B{\setminus }\mathrm {Int}(B),B\cup \mathrm {Ext}(B)]\) (where B ranges over all bases of M and \(\mathrm{Int}(B)\) and \(\mathrm{Ext}(B)\) denote the usual sets of active elements with respect to B, see [9, 12, 19, 20]). Finally, notice a classical interpretation of \(p(M;0)=T(M;1,0)\) introduced in [13] in terms of totally cyclic and acyclic orientations of graphs satisfying additional conditions. These orientations represent certain equivalence classes of totally cyclic and acyclic orientations of graphs (see [15, 16]).
References
Aigner, M.: Whitney numbers. In: White, N. (ed.) Combinatorial Geometries, pp. 139–160. Cambridge University Press, Cambridge (1987)
Annan, J.D.: The complexities of the coefficients of the Tutte polynomial. Discrete Appl. Math. 57, 93–103 (1995)
Brylawski, T.: The Tutte polynomial, part 1: general Theory. In: Barlotti, A. (ed.) Matroid Theory and Its Applications, pp. 126–275. Springer, Berlin (2010)
Brylawski, T., Oxley, J.: The Tutte polynomial and its applications. In: White, N. (ed.) Matroid Applications, pp. 123–225. Cambridge University Press, Cambridge (1992)
Crapo, H.H.: The Tutte polynomial. Aequationes Math. 3, 211–229 (1969)
Ellis-Monaghan, J., Merino, C.: Graph polynomials and their applications I: the Tutte polynomial. In: Dehmer, M. (ed.) Structural Analysis of Complex Networks, pp. 219–255. Birkhauser, Dordrecht (2011)
Ellis-Monaghan, J., Moffat, I. (eds.): Handbook of the Tutte Polynomial. CRC Press, Boca Raton (2020)
Etienne, G., Las Vergnas, M.: External and internal elements of a matroid basis. Discrete Math. 179, 111–119 (1998)
Gioan, E.: On Tutte polynomial expansion formulas in perspectives of matroids and oriented matroids. Preprint, arXiv:1807.06559
Goldberg, L.A., Jerrum, M.: Inapproximability of the Tutte polynomial. Inform. and Comput. 206, 908–929 (2008)
Goodall, A., Krajewski, T., Regts, G., Vena, L.: A Tutte polynomial for maps. Combin. Probab. Comput. 27, 913–945 (2018)
Gordon, G., Traldi, L.: Generalized activities and the Tutte polynomial. Discrete Math. 85, 167–176 (1990)
Greene, C., Zaslavsky, T.: On the interpretation of Whitney numbers through arrangements of hyperplanes, zonotopes, non-Radon partitions, and orientations of graphs. Trans. Amer. Math. Soc. 280, 97–126 (1983)
Jaeger, F., Vertigan, D.L., Welsh, D.J.A.: On the computational complexity of the Jones and Tutte polynomials. Math. Proc. Cambridge Philos. Soc. 108, 35–53 (1990)
Kochol, M.: Polynomial algorithms for canonical forms of orientations. J. Comb. Optim. 31, 218–222 (2016)
Kochol, M.: Bounds of characteristic polynomials of regular matroids. Manuscript (2018)
Kook, W., Reiner, V., Stanton, D.: A convolution formula for the Tutte polynomial. J. Combin. Theory Ser. B 76, 297–300 (1999)
Kung, J.P.S.: Convolution-multiplication identities for Tutte polynomials of graphs and matroids. J. Combin. Theory Ser. B 100, 617–624 (2010)
Las Vergnas, M.: Active orders for matroid bases. European J. Combin. 22, 709–711 (2001)
Las Vergnas, M.: The Tutte polynomial of a morphism of matroids 5. Derivatives as generating functions of Tutte activities. European J. Combin. 34, 1390–1405 (2013)
Oxley, J.G.: Matroid Theory. Oxford University Press, Oxford (1992)
Tutte, W.T.: A contribution to the theory of chromatic polynomials. Canad. J. Math. 6, 80–91 (1954)
Vertigan, D.: The computational complexity of Tutte invariants for planar graphs. SIAM J. Comput. 135, 690–712 (2005)
Welsh, D.J.A.: Complexity: Knots, Colourings and Counting. Cambridge University Press, Cambridge (1993)
Welsh, D.J.A.: The Tutte polynomial. Random Structures Algorithms 15, 210–228 (1999)
Zaslavsky, T.: The Möbius function and the characteristic polynomial. In: White, N. (ed.) Combinatorial Geometries, pp. 114–138. Cambridge University Press, Cambridge (1987)
Acknowledgements
Author thanks unknown referees for comments.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Partially supported by VEGA 2/0024/18.
Rights and permissions
About this article
Cite this article
Kochol, M. Interpretations of the Tutte and characteristic polynomials of matroids. J Algebr Comb 53, 1–9 (2021). https://doi.org/10.1007/s10801-019-00914-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10801-019-00914-6