Abstract Nonsense

Crushing one theorem at a time

Alternating Forms

Point of post: In this post we discuss the concept of alternating forms as is discussed in section 30 of Halmos.


So in our last post we restricted our study of multilinear algebra to that of spaces of the form \text{Mult}_n\left(\mathscr{V}\right) in the attempt to better approximate multilinear forms which are near and dear to us, (e.g. the determinant). Thus, we continue on this path to study alternating forms, the abstractification of the observation that if two columns (or rows) in a matrix are the same, then the determinant is zero. This is really where things start to get interesting.

Alternating Forms

Let \mathscr{V} be a finite dimensional F-space and K\in\text{Mult}_n\left(\mathscr{V}\right) we call K alternating if


for any \ell,k\in[n]. In other words K is zero whenever any two of it’s arguments are equal. The first result that we can see from such n-forms is that they’re alternating. Indeed

Theorem: Let \mathscr{V} be a finite dimensional F-space and let K\in\text{Mult}_n\left(\mathscr{V}\right)  be alternating, then K is skew-symmetric.

Proof: We first note for any two \ell,k\in[n] we have by multilinearity that

\begin{aligned}K(x_1,\cdots,\underbrace{x+y}_{k^{\text{th}}},\cdots,\underbrace{x+y}_{\ell^{\text{th}}},\cdots,x_n)=& K\left(x_1,\cdots,x,\cdots,x,\cdots,x_n\right)\\ &+K\left(x,\cdots,y,\cdots,x,\cdots,x_n\right)\\ &+K\left(x_1,\cdots,y,\cdots,y,\cdots,x_n\right)\\ &+K\left(x_1,\cdots,y,\cdots,x,\cdots,x_n\right)\end{aligned}

But, since K is alternating we may conclude that the left hand side and the first and third terms of the right hand side vanish to give us that

\displaystyle K(x_1,\cdots,x,\cdots,y,\cdots,x_n)=-K(x_1,\cdots,y,\cdots,x,\cdots,x_n)

from where it follows by the arbitrariness of \ell,k that for any transposition \tau\in S_n we have that

\tau K=-K

and thus if \pi=\tau_1\cdots\tau_m we may recall by the first theorem in the last post that

\pi K=(\tau_1\cdots\tau_m)K=\tau_1(\cdots(\tau_m K)\cdot)=\tau_1(\cdots(-\tau_2 K)\cdots)=\cdots=(-1)^m K

(more properly by induction) from where it clearly follows that \pi K=\text{sgn}(\pi)K and thus K is skew-symmetric. \blacksquare

Now, intuitively it seems as though the converse of the above is true since, for example taking bilinear forms we see that if B is skew-symmetric then

B(x,x)=-\underbrace{B(x,x)}_{\text{switching }x\text{s}}

from where we would like to conclude that B(x,x)=0. Unfortunately though there are fields for which x=-x and x\ne 0, take \mathbb{Z}_2 for example. In particular, this is true precisely when the character of the field is equal to 2. Thus, with this in mind we are able to make sense of the restrictions in the following theorem:

Theorem: Let \mathscr{V} be a finite dimensional F-space with \text{char}\left(F\right)>2, then if K\in\text{Mult}_n\left(\mathscr{V}\right) is skew-symmetric then K is alternating.

Proof: We merely note that for any \ell,k\in[n] we have that


and thus adding to both sides we see that


and since 1+1 is a unit we may conclude that


and since \ell,k\in[n] and x\in\mathscr{V} were arbitrary it follows that K is alternating. \blacksquare

We will see later (as one of the homework problems) an example where the above can fail for fields of characteristic two. Next is a theorem which is patently obvious, yet very important. It basically says that if \{x_1,\cdots,x_n\} is a linearly dependent set of vectors in \mathscr{V} and K\in\text{Mult}_n\left(\mathscr{V}\right) is alternating then K(x_1,\cdots,x_n)=0. This is clear, as I said, because one of the vectors in \{x_1,\cdots,x_n\} is a linear combination of the others. Thus, when one uses multilinearity to expand it one has a sum of terms each term of which has two identical entries in the K part (see below).

Theorem: Let \mathscr{V} be a m-dimensional F-space and \{x_1,\cdots,x_n\}\subseteq\mathscr{V} a linearly dependent set of vectors. Then, for any alternating K\in\text{Mult}_n\left(\mathscr{V}\right) it’s true that


Proof: Since \{x_1,\cdots,x_n\} is linearly dependent then we may assume without loss of generality that there exists scalars \alpha_2,\cdots,\alpha_n\in F such that

\displaystyle x_1=\sum_{j=2}^{n}\alpha_j x_j

(we may assume this since one of the vectors must be a linear combination of the others, and thus with the possibility of reordering we may assume that vector is the “first” one). Thus

\displaystyle K\left(\sum_{j=2}^{n}\alpha_j x_j,\cdots,x_n\right)=\sum_{j=2}^{n}\alpha_j K\left(x_j,\cdots,x_n\right)=\sum_{j=2}^{n}\alpha_j 0=0

From where the conclusion follows. \blacksquare

The obvious question is, does the converse hold. Namely, if one has a n-linear alternating form K and does K(x_1,\cdots,x_n)=0 imply that \{x_1,\cdots,x_n\} is linearly dependent? The answer is yes in the case when n=\dim_F\mathscr{V}=m.

Theorem: Let \mathscr{V} be an m-dimensional F-space and K\in\text{Mult}_m\left(\mathscr{V}\right) be non-zero and alternating. Then, if \{x_1,\cdots,x_m\} are linearly independent then K(x_1,\cdots,x_m)\ne0.

Proof: Clearly since the dimension coincides with the number of vectors we know that \{x_1,\cdots,x_m\} forms a basis for \mathscr{V}. So, let y_1,\cdots,y_m\in\mathscr{V} be arbitrary. Then, we know that

\displaystyle y_1=\sum_{j_1=1}^{m}\alpha_{1,j}x_j,\cdots,y_m=\sum_{j_m=1}^{n}\alpha_{m,j}x_j\quad (1)

so that

\displaystyle K(y_1,\cdots,y_m)=\sum_{j_1=1}^{m}\cdots\sum_{j_m=1}^{m}\alpha_{1,j}\cdots\alpha_{m,j}K\left(x_{j_1},\cdots,x_{j_m}\right)

now, if at any point x_{j_k}=x_{j_\ell} then K(x_{j_1},\cdots,x_{j_m})=0 and if not then K(x_{j_1},\cdots,x_{j_m})=\pi K(x_1,\cdots,x_m) for some \pi, and by our first theorem this implies that K(x_{j_1},\cdots,x_{j_m})=\text{sgn}(\pi)K\left(x_1,\cdots,x_m\right). Thus, if we assume that K(x_1,\cdots,x_m)=0 then the two cases above imply that every term in (1) is zero, and thus the sum is zero. But, since y_1,\cdots,y_m were arbitrary it follows that K(y_1,\cdots,y_m)=0 for all y_1,\cdots,y_m\in\mathscr{V} contradictory to our assumption that K\ne \bold{0}. It follows that K(x_1,\cdots,x_m)\ne 0. \blacksquare


1. Halmos, Paul R. ” Finite-dimensional Vector Spaces,. New York: Springer-Verlag, 1974. Print


November 14, 2010 - Posted by | Algebra, Halmos, Linear Algebra | , ,


  1. […] that we have dealt enough with alternating -forms to get a “feel” for them, we can safely state some obvious theorems, which trim away […]

    Pingback by Characterization of Alternating Multilinear Forms, and the Determinant « Abstract Nonsense | November 15, 2010 | Reply

  2. […] Recall that an alternating map is a -linear map such that if any . The classic example of an alternating map is the -linear map given by the determinant (where we are obviously identifying an -tuple of elements with their associated -matrix). […]

    Pingback by The Tensor Algebra and Exterior Algebra (Pt. III) « Abstract Nonsense | May 10, 2012 | Reply

  3. […] back into the space. The embedding should look familiar either if you are familiar with the more classic approach to multilinear algebra (via the alternating map) or if you know Leibniz expansion of the determinant. Namely, we are […]

    Pingback by The Tensor Algebra and Exterior Algebra (Pt. IV) « Abstract Nonsense | May 10, 2012 | Reply

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: