For linear maps and with finite-dimensional , the rank inequality for a composition isRestrict to . The rank-nullity theorem gives , and the intersection has dimension at most .
For example, the product of two rank-four endomorphisms of a five-dimensional vector space has rank at least three. The bound controls how much rank can be lost when the image of one map meets the kernel of the next.
For compatible matrices , the inequality is . One proof uses the block matrix . Subtract times its lower block row from its upper block row and then subtract its first block column times from its second. The result is , with rank . On the other hand, choose independent columns of from the first block column and independent columns of from the second. The corresponding columns of are independent: project a relation onto the upper block first, and then onto the lower block. Thus .
Articles by others on the same topic
There are currently no matching articles.