Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Source. Lemma 1, printed p. 150 (published PDF).

Statement. If M=(E,F)M=(E,\mathcal F) is a finite matroid and n∈Z≥0n\in\mathbb Z_{\ge0}, then

F(n)={I∈F:∣I∣≤n}\mathcal F_{(n)}=\{I\in\mathcal F:|I|\le n\}

is a matroid on EE, with rank r(n)(A)=min⁡(n,r(A))r_{(n)}(A)=\min(n,r(A)).

Proof. The family contains the empty set and is hereditary. If I,JI,J belong to it and ∣I∣<∣J∣|I|<|J|, ordinary matroid augmentation supplies e∈J∖Ie\in J\setminus I for which I∪{e}I\cup\{e\} is independent in MM. Its size is at most ∣J∣≤n|J|\le n, so it remains in F(n)\mathcal F_{(n)}. The augmentation characterization therefore proves the matroid property.

Every independent subset of AA in the truncation has size at most min⁡(n,r(A))\min(n,r(A)). A base of AA in MM has r(A)r(A) elements, and selecting min⁡(n,r(A))\min(n,r(A)) of them attains that bound. This also covers n=0n=0 and A=∅A=\varnothing. □\square

The source calls the proof of Lemma 1 obvious. The argument above expands it. The source asserts, without giving the argument, that a truncation of a graphic or transversal matroid need not stay in that particular class; later uses concern general matroids.