Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

The following is a historical question, but first some background:

Let $\psi$ be a linear operator from a vector space to itself. The following two expressions, viewed as formal power series, can be shown to be equal (even for an infinite dimensional vector space):

$$ \det(1 - \psi T) = \exp \left(-\sum_{s=1}^{\infty} \text{Tr}(\psi^{s})T^{s}/s\right) $$

Here is a simple example using the associated matrix in the finite dimensional case:

enter image description here

For more on this topic (and how to ensure the definitions make sense for infinite dimensional vector spaces) see: N. Koblitz, p-adic Numbers, p-adic Analysis, and Zeta-Functions, Springer-Verlag, New York, 1984.

In particular, see Koblitz's book for how ($p$-adic versions of) this identity can be used in Bernard Dwork's proof of the rationality of the zeta function (resolving the first of the Weil Conjectures).

I would welcome any insight as to how one thinks up, observes, or conjectures such an identity in the first place, but my question is: what is the origin of this identity and where else has it appeared (perhaps in a somewhat modified form)?

share|cite|improve this question
Its discussed on the wikipedia page for the determinant. It can be viewed as a form of the identity $\exp(\mathrm{tr}(M))=\det\exp(M)$. – Chris Godsil Jan 13 '13 at 20:10
@ChrisGodsil If you can write up something formal showing why this is the case, I will accept it as an answer. (Unless others can shed more light on where else this identity shows up and what its history is.) – Benjamin Dickman Jan 13 '13 at 21:20
Since you ask the question in relation to its use in Dwork's proof of the first Weil conjectures... I think this kind of identity was known to Weil, who first defined the zeta-function of a variety in fairly general terms. You might look at his 1949 paper where he stated his conjectures, just to see if it provides some inspiration for you. – KCd Jan 14 '13 at 4:23
up vote 3 down vote accepted

The development of this operator identity can be traced back till the very beginning of the 20th century.

I think the paper On the numerical evaluation of Fredholm determinants from Folkmar Bornemann (2008) could be useful for you. Bornemann presents in Section 3: Definition and Properties of Fredholm and Operator Determinants of this paper four different representations of your stated operator identity. He writes:

For a trace class operator $A \in \mathcal{J}_1(\mathcal{H})$ there are several equivalent constructions that all define one and the same entire function \begin{align*} d(z)=\text{det}(I+zA)\qquad(z\in\mathbb{R}) \end{align*} in fact, each construction has been chosen at least once, in different places of the literature, as the basic definition of the operator determinant:

$1.)$ Gohberg and Krein ($1969$, p. $157$) define the determinant by the locally uniformly convergent (infinite) product \begin{align*} \text{det}(I+zA)=\prod_{n=1}^{N(A)}(1+z\lambda_n(A)) \end{align*} which possesses zeros exactly at $z_n=-1/\lambda_n(A)$, counting multiplicity.

$2.)$ Gohberg et al. ($1990$, p. $115$) define the determinant as follows. Given any sequence of finite rank operators $A_n$ with $A_n\rightarrow A$ converging in trace class norm, the sequence of finite dimensional determinants \begin{align*} \text{det}\left(I+zA_n\upharpoonright_{\text{ran}(A_n)}\right) \end{align*} (which are polynomials in z) converges locally uniform to $\text{det}(I+zA)$, independently of the choice of the sequence $A_n$. The existence of at least one such sequence follows from the singular value representation at $(2.1)$.

$3.)$ Dunford and Schwartz ($1963$, p. $1029$) define the determinant by what is often called Plemelj’s formula \begin{align*} \text{det}(I+zA)=\exp(\text{tr}\ \text{log}(I+zA))=\exp\left(-\sum_{n=1}^{\infty}\frac{(-z)^n}{n}\text{tr}A^n\right) \end{align*} which converges for $|z|<1/|\lambda_1(A)|$ and can analytically be continued as an entire function to all $z\in\mathbb{C}$.

$4.)$ Grothendieck ($1956$, p. $347$) and Simon ($1977$, p. $254$) define the determinant most elegantly with a little exterior algebra (Greub $1967$). With $\bigwedge^n(A)\in\mathcal(J)_1\left(\bigwedge^n(\mathcal{H})\right)$ being the $n^\text{th}$ exterior product of $A$, the power series \begin{align*} \text{det}(I+zA)=\sum_{n=0}^{\infty}z^n\text{tr}\bigwedge^n(A) \end{align*} converges for all $z\in\mathbb{C}$. Note that $\text{tr}\bigwedge^{n}(A)=\sum_{i_1<\cdots<i_n}\lambda_{i_1}(A)\cdots\lambda_{i_n}(A)$ is just the $n^\text{th}$ symmetric function of the eigenvalues of $A$.

Proofs of the equivalence can be found in (Gohberg et al. $2000$, Chap. $2$) and (Simon $2005$, Chap. $3$).

In a footnote for $3.)$ above, Bornemann provides additional information to the Plemelj Formula, namely:

Plemelj ($1904$, Eq. $(62)$) had given a corresponding form of the Fredholm determinant for integral operators. However, it can already be found in Fredholm ($1903$, p. $384$).

So, the formula seems to go back till Fredholm $1903$. I suppose according to some references I've found, it's the following paper: I. Fredholm, Sur une classe d`équation fonctionelle, Acta Math. $27, 365-390 (1903)$.

share|cite|improve this answer
@Benjamin Dickman: Thanks for the bounty. Since you put the question long time ago, its really good to see the answer pleases you! :-) – Markus Scheuer May 3 '14 at 17:11

Using $\det\exp=\exp\mathrm{tr}$, we have $$ \det(I-tA) = \det \exp\log(I-tA) = \exp\mathrm{tr}\log(I-tA) $$ and as $$ \log(I-tA) = -\sum_{k\ge1} \frac1k t^k A^k, $$ we have $$ \exp\mathrm{tr}\log(I-tA) = \exp\left( -\sum_{k\ge1} \frac1k t^k \mathrm{tr}(A^k)\right) $$ This shows that your identity is a form of the $\det\exp$ identity.

The question about the origins is interesting and I do not have much to say (which was why I restricted myself to a comment in the first place). For complex matrices it is very easy to prove that $\det\exp(A)=\exp\mathrm{tr}(A)$, which might explain why I have never seen a name attached to it.

I have seen it used in connection with Lie groups (to prove that sl is the Lie algebra of SL), and (in forms like what you stated) in work on enumeration.

share|cite|improve this answer

Your Answer


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.