Let $T$ be an endomorphism of a finite dimensional vector space $V$. Suppose that $(v_1,\ldots v_n)$ is an ordered basis of $V$. And let $[T]$ be the matrix of $T$ with respect to this basis.

Is there a way to compute the matrix of $\Lambda^k(T)$ with respect to the obvious basis given $[T]$? The 'obvious' basis is the $\binom{n}{k}$ $k$-wedge tuples from $\{v_1,\ldots,v_n\}$ with increasing indices.

There have been many questions on here about how to compute the characteristic polynomial and the coefficients $(-1)^k\text{tr}(\Lambda^k(T))$, but I haven't seen any interest in the matrix of $\Lambda^k(T)$ itself. I suspect it can be built from the minors of $[T]$, but I have no idea how to proceed.

up vote 4 down vote accepted

Let $T(v_i) = \sum_j a_{ji} \cdot v_j$, so that $(a_{ji})$ is the matrix of $T$ w.r.t. to $(v_1,\dotsc,v_n)$. The wedges $v_{i_1} \wedge \dotsc \wedge v_{i_k}$ with $i_1<\dotsc<i_k$ form a basis of $\Lambda^k(V)$. We have:

$$\Lambda^k(T)(v_{i_1} \wedge \dotsc \wedge v_{i_k}) = T(v_{i_1}) \wedge \dotsc \wedge T(v_{i_k})=\sum_{j_1,\dotsc,j_k} a_{j_1,i_1} \dotsc a_{j_k,i_j} \cdot v_{j_1} \wedge \dotsc \wedge v_{j_k}$$

If the indices $j_\ell$ are not pairwise distinct, the summand is zero. If not, we may choose a permutation to order the indices. The result is: $$\sum_{j_1<\dotsc<j_k} \sum_{\pi \in \Sigma_k} \mathrm{sgn}(\sigma) a_{j_{\pi(1)},i_1} \dotsc a_{j_{\pi(k)},i_j} \cdot v_{j_1} \wedge \dotsc \wedge v_{j_k}$$ Thus, the entry of the matrix of $\Lambda^k(T)$ in the row $(j_1<\dotsc<j_k)$ and the column $(i_1<\dotsc<i_k)$ is $$ \sum_{\pi \in \Sigma_k} \mathrm{sgn}(\sigma) a_{j_{\pi(1)},i_1} \dotsc a_{j_{\pi(k)},i_j} .$$ This is nothing else than the determinant of the submatrix of $(a_{ji})$ corresponding to the rows $(j_1<\dotsc<j_k)$ and the columns $(i_1<\dotsc<i_k)$.

Conclusion: The entries of the matrix of $\Lambda^k(T)$ are the $k$-minors of the matrix of $T$.

I thought it might be worthwhile to add an explicit example. Consider $T:\mathbb{R}^3 \rightarrow \mathbb{R}^3$ with $$ [T] = \left[ \begin{array}{ccc} a & d & g \\ b & e & h \\ c & f & i \end{array}\right]$$ In particular, using the usual $e_1 = [ 1,0,0 ]^T=(1,0,0)$, $e_2 = (0,1,0)$, $e_3=(0,0,1)$ we have $$ T(e_1) = (a,b,c), \ \ \ \ T(e_2) = (d,e,f),\ \ \ \ T(e_3) = (g,h,i). $$ Calculate the wedge products of these images \begin{align} T(e_1) \wedge T(e_2) &= (a,b,c) \wedge (d,e,f) \\ &= (ae-bd)e_1 \wedge e_2+(af-cd)e_1 \wedge e_3+(bf-ce)e_2 \wedge e_3 \end{align} Next, \begin{align} T(e_1) \wedge T(e_3) &= (a,b,c) \wedge (g,h,i) \\ &= (ah-gb)e_1 \wedge e_2+ (ai-cg)e_1 \wedge e_3+(bi-ch)e_2 \wedge e_3 \end{align} and \begin{align} T(e_2) \wedge T(e_3) &= (d,e,f) \wedge (g,h,i) \\ &= (dh-eg)e_1 \wedge e_2+(di-fg)e_1 \wedge e_3+(ei-fh)e_2 \wedge e_3 \end{align} The basis for $\Lambda^2 \mathbb{R}^3$ with increasing indices is $\{ e_1 \wedge e_2, e_1 \wedge e_3, e_2 \wedge e_3 \}$. Considering this the standard basis, we find in view of the calculations above and the definition $(\Lambda^2 T)(v,w) = T(v) \wedge T(w)$, $$ [\Lambda^2 T] = \left[ \begin{array}{c|c|c} ae-bd & ah-gb & dh-eg \\ \hline af-cd & ai-cg & di-fg \\ \hline bf-ce & bi-ch & ei-fh \end{array}\right]$$ This is a matrix of minors of $[T]$. However, given our choice of basis, the minors are out of place relative to their usual assignment in the Laplace expansion. If we instead order the basis $\{ e_2 \wedge e_3, e_1 \wedge e_3, e_1 \wedge e_2 \}'$ then $$ [\Lambda^2 T]' = \left[ \begin{array}{c|c|c} ei-fh & bi-ch & bf-ce \\ \hline di-fg & ai-cg & af-cd \\ \hline dh-eg & ah-gb & ae-bd \end{array}\right]$$ That's better. Now the minors are where they ought to be in my humble opinion. I suppose this is not terribly surprising, the strictly increasing basis is not natural for making the correspondence for vectors and one and two forms in $\mathbb{R}^3$.

  • That was very illuminating. Thank you – Robert Wolfe Jul 25 '14 at 4:18
  • 1
    glad to help, I once made a sign-error in the 12-slot of the above. All the sudden I had students taking cell-phone pictures... – James S. Cook Jul 25 '14 at 4:43

Your Answer

 

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.