Let $v$ be a positive integer. I have a representation $\rho_v$ of $USp(4) = \{g\in M_2(\mathbb{H})\,|\,g^{T}\bar{g}\}$, where $\mathbb{H}$ is Hamiltons quaternions. The representation $\rho_v$ has young diagram consisting of two rows of length $v$.

I gather that by converting quaternions to matrices with complex entries that $\rho_v$ is really the irreducible representation of $Sp_4(\mathbb{C})$ that has highest weight $v(L_1 + L_2)$.

I now have to find $tr(\rho_v(\gamma))$ for certain matrices $\gamma$.

I understand that I have to use the Weyl character formula but how do I use this to find the trace of the rep acting on a single element? I have barely used the character formula so I apologise if this is simple.


You are working in the compact group of unitary symplectic $4\times4$ matrices, in which every matrix is diagonalizable (because of the unitary condition); moreover there are only two independent eigenvalues, since for each eigenvalue the inverse is also an eigenvalue (because of the symplectic condition). Therefore each group element is conjugate to a diagonal matrix with diagonal entries $x,y,x^{-1},y^{-1}$ for some complex values $x,y$ of modulus $1$; it suffices to know the trace (which is invariant on conjugacy classes) on these matrices. The trace will be given by a Laurent polynomial in $x,y$ which in addition is symmetric in $x$ and $y$ and invariant under interchange of $x$ and $x^{-1}$.

The Weyl character formula gives you this Laurent polynomial as a quotient of two alternating polynomials (in the sense that they change sign under the symmetries indicated above). The result for the weight you indicated is not easy to express as a closed formula in $v$, but for instance for $v=3$ you get a Laurent polynomial with terms $x^3y^3+x^3y+x^2y^2+x^2+2xy+2$, to be completed with other terms like $x^{-3}y^3$ , $xy^3$ and $2x^{-1}y^{-1}$ to establish the required symmetry (there are $4+8+4+4+4+1=25$ terms in all; I'm too lazy to write them all down). That this is indeed the right formula can be checked by multiplying the $25$-term polynomial by the Weyl denominator $x^2y-xy^2+x^{-1}y^2-x^{-2}y+x^{-2}y^{-1}-x^{-1}y^{-2}+xy^{-2}-x^2y^{-1}$, which should give $$ x^5y^4-x^4y^5+x^{-4}y^5-x^{-5}y^4+x^{-5}y^{-4}-x^{-4}y^{-5}+x^4y^{-5}-x^5y^{-4}. $$ Thus that $25$-term polynomial gives you the trace of the action of any matrix in the group as function of its eigenvalues.

  • $\begingroup$ Ok so to begin with you end up with 8 terms on both the numerator and the denominator of the WCF, but after cancellation you end up being left with a polynomial with 25 terms? Is the formula in my latest post correct then? (math.stackexchange.com/q/169084/25381) $\endgroup$ – fretty Jul 11 '12 at 11:01
  • $\begingroup$ Is there even a need to cancel the denominator out or can I just sub in the values of $x$ and $y$ in the beginning? $\endgroup$ – fretty Jul 11 '12 at 11:03
  • $\begingroup$ It is not really cancellation that is going on, it is polynomial division, which can create lots of terms (think of dividing $x^n-y^n$ by $x-y$). The division has to be exact (without remainder) by a subtle symmetry argument. But actually performing a division (not obvious with multiple indeterminates) is a rather difficult way to obtain the quotient $\endgroup$ – Marc van Leeuwen Jul 11 '12 at 11:27
  • $\begingroup$ As for your question about "canceling" the denominator before substitution: this is not necessary for concrete values of $x,y$ for which the denominator is nonzero. But I cannot think of much use of actually performing this division for concrete numerical values of $x,y$ (unless you want to plot the character as a function); it is the formal quotient that seems to convey the information in a more useful form. $\endgroup$ – Marc van Leeuwen Jul 11 '12 at 11:33
  • $\begingroup$ One more remark; the term "cancelling" is OK if you use it in a multiplicative sense. I was thinking of cancelling in an additive sense, which of course is nonsense here. $\endgroup$ – Marc van Leeuwen Jul 11 '12 at 11:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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