Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

If I understand correctly, the analytic multiplicity of a linear operator say $T:V\to V$ is the amount of times $\lambda$ shows up as a root in the characteristic polynomial (Assuming you have a matrix $A$ representing $T$ with respect to some basis of $V$, then the characteristic equation is $\det(A-\lambda I)$).

I understand how to find this, but what exactly does this mean for the linear operator?

Also, how do we find the geometric multiplicity of $T$? What is it's significance? I tried looking it up on Wikipedia but they are using some notation and words that I am not familiar with; the definitions there are usually quite formal. Am I correct in thinking that it is the greatest amount of linearly independent vectors of the eigenvalues of $T$? Do all the eigenvalues of a linear operator have the same amount of linearly independent vectors?

Edit: Here is a specific example. We have the matrix, say $A$, with $1$s all along the diagonal and above, and $0$s below. Since this matrix is upper triangular, the eigenvalues of $A$ are $1$. Looking at the equations we have to satisfy to find the eigenvector(s) of an eigenvalue $1$, we get: \begin{align*} x_1 + x_2 + \cdots + x_n &= \lambda x_1\\ x_2 + \cdots + x_n &= \lambda x_2\\ &\vdots\\ x_n &=\lambda x_n \end{align*}

It seems like we need to do some inspection here, it would be hard (or at least I can't see how) to show this formally. It seems to me like there are two possible linear independant eigenvectors here,
$$\displaystyle \vec 0$$ and $$\displaystyle (1,0,...,0)$$
So the geometric multiplicity is $1$?

share|improve this question
1  
@Fdart: The algebraic multiplicity (usual name, rather than "analytic", but okay) is not of the operator, but of the eigenvalue $\lambda$; each eigenvalue has $\lambda$ has an associated algebraic multiplicity. Same for the geometric multiplicity: not of $T$, but of $\lambda$ as an eigenvalue of $T$. –  Arturo Magidin Feb 17 '11 at 3:54
    
Ok, that makes a lot more sense :) –  fdart17 Feb 17 '11 at 3:55
2  
The geometric multiplicity of $\lambda$ is the dimension of its corresponding eigenspace. –  PEV Feb 17 '11 at 3:56
    
@Fdart17: $\mathbf{0}$ never counts, because no collection of vectors that contains $\mathbf{0}$ can be linearly independent. –  Arturo Magidin Feb 17 '11 at 4:39
    
@Fdart17: I've added the analysis of this particular matrix. –  Arturo Magidin Feb 17 '11 at 4:47

1 Answer 1

up vote 8 down vote accepted

The geometric multiplicity of $\lambda$ tells you how big a subspace of $V$ you can find where $T$ acts simply as "multiplication by $\lambda$" (that is, how big, dimensionally speaking, the subspace spanned by the eigenvectors of $\lambda$ is).

The analytic/algebraic multiplicity of $\lambda$ tells you how big that space "should" be for $V$ and $T$ to have a "nice" decomposition of the following kind: you can express $V$ as a direct sum of subspaces $E_{\lambda_1}$, $E_{\lambda_2},\ldots,E_{\lambda_k}$ (where $\lambda_1,\ldots,\lambda_k$ are the distinct roots of the characteristic polynomial) so that on each $E_{\lambda_i}$, $T$ acts just by "multiplication by $\lambda_i$". For $V$ to really be equal to the sum of these spaces, you need $\dim(E_{\lambda_i})$, which is the geometric multiplicity of $\lambda_i$, to equal the algebraic multiplicity of $\lambda_i$.

It has other properties, but I think that's a good place to start.

Yes, the geometric multiplicity is the largest possible number of linearly independent eigenvectors of $T$ associated to $\lambda$ (vectors $\mathbf{v}$, $\mathbf{v}\neq\mathbf{0}$, such that $T(\mathbf{v}) = \lambda\mathbf{v}$; that is, vectors on which $T$ acts just by "multiplication by $\lambda$).

No, not all eigenvalues have the same geometric multiplicity; for example, in the matrix $$\left(\begin{array}{cccc} 2 & 1 & 0 & 0\\\ 0 & 2 & 0 & 0\\\ 0 & 0 & 2 & 0\\\ 0 & 0 & 0 & 1 \end{array}\right),$$ the characteristic polynomial is $(2-\lambda)^3(1-\lambda)$, so the two eigenvalues are $\lambda_1=1$ and $\lambda_2=2$. The eigenvalue $\lambda_1=1$ has algebraic and geometric multiplicities both equal to $1$; $\lambda_2=2$ has algebraic multiplicity $3$, and geometric multiplicity $2$. (You can check the geometric multiplicity by finding the nullity of $A-2I$).

Added. Since the geometric multiplicity of an eigenvalue $\lambda_i$ is the dimension of the subspace $E_{\lambda_i}$, your first task in finding that dimension is to identify the vectors $\mathbf{v}$ for which $T(\mathbf{v})=\lambda_i\mathbf{v}$. This is equivalent to finding the vectors for which $(T-\lambda_i I)(\mathbf{v})=\mathbf{0}$. The reason this is a better problem to tackle is that it is easier to solve a system that looks like $B\mathbf{v}=\mathbf{0}$, than one that looks like $A\mathbf{v}=\lambda\mathbf{v}$.

So, you find the nullspace of $T-\lambda_iI$, that is, the collection of all vectors $\mathbf{v}$ for which $(T-\lambda_iI)(\mathbf{v})=\mathbf{0}$. Its dimension is precisely the geometric multiplicity of $\lambda_i$, so the geometric multiplicity of $\lambda_i$ is found by computing $\mathrm{nullity}(T-\lambda_iI)$.

For your specific example: we begin with the matrix corresponding to the standard basis: $$A =\left(\begin{array}{cccc} 1 & 1 & \cdots & 1\\\ 0 & 1 & \cdots & 1\\\ \vdots & \vdots & \ddots & \vdots \\\ 0 & 0 & \cdots & 1 \end{array}\right).$$ The characteristic polynomial is $\det(A-tI) = (1-t)^n$, so the only eigenvalue is $\lambda=1$, with algebraic multiplicity $n$.

To find the geometric multiplicity, take $A-1I$ ("$1I$" because we are taking $\lambda = 1$), and find its nullspace. Since $$A - I = \left(\begin{array}{ccccc} 0 & 1 & 1 &\cdots & 1\\\ 0 & 0 & 1 & \cdots & 1\\\ \vdots & \vdots & \vdots & \ddots & \vdots \\\ 0 & 0 & 0 & \cdots & 0 \end{array}\right),$$ finding the reduced row-echelon form will give you the solutions to $(A-I)\mathbf{x}=\mathbf{0}$. The reduced row-echelon form of $A-I$ is $$\left(\begin{array}{ccccc} 0 & 1 & 0 & \cdots & 0\\\ 0 & 0 & 1 & \cdots & 0\\\ \vdots & \vdots & \vdots & \ddots & \vdots\\\ 0 & 0 & 0 & \cdots & 1\\\ 0 & 0 & 0 & \cdots & 0 \end{array}\right),$$ so $\mathbf{x}=(x_1,x_2,\ldots,x_n)$ is in the nullspace if and only if $x_2=x_3=\cdots=x_n=0$. So the eigenvectors of $\lambda=1$ are all vectors of the form $(a,0,0,\ldots,0)$ for arbitrary $a$; however, because $\mathbf{0}$ is always a solution to $A\mathbf{x}=\lambda \mathbf{x}$ for any $\lambda$, we declare by fiat that an eigenvector has to be nonzero (this has no bearing on the geometric multiplicity of $\lambda$, because $\mathbf{0}$ can never be in a linearly independent set). A basis for this nullspace is given by $(1,0,\ldots,0)$, so the nullspace has dimension $1$. This dimension is the geometric multiplicity of $\lambda=1$.

So, in summary: $\lambda=1$ is the only eigenvalue; it has algebraic multiplicity $n$, and geometric multiplicity $1$. The eigenvectors are all nonzero multiples of $(1,0,0,\ldots,0)$.

share|improve this answer
    
I wish I count vote up more than once.. Thanks a lot! –  fdart17 Feb 17 '11 at 4:07

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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