Numerical symbolic differentiation and evaluate at 0.

I am trying to do numerical symbolic differentiation of the kappa function below after which I want to evaluate at 0 and simplify. This is my code:

clear
syms delta gamma rho lambda T sigma u r

f_1(u) = 1i*u*rho -(1-exp(-lambda*T))*(u^2+1i*u)/(2*lambda);

f_2(u) = 1i*u*rho-(u^2+2*1i*u)/(2*lambda);

I(u) = delta/lambda*(sqrt(gamma^2-2*f_1(u))-sqrt(gamma^2-2*1i*rho*u))...
         - 2*delta*f_2(u)/(lambda*sqrt(gamma^2 - 2*f_2(u))) ...
          *(atanh(sqrt((gamma^2-2*1i*rho*u)/(gamma^2 - 2*f_2(u)))) ...
          - atanh(sqrt( (gamma^2-2*f_1(u))/(gamma^2- 2*f_2(u)))));

kappa(u) = 1i*u*T*(r-lambda*(delta*rho)/(sqrt(gamma^2 -2*rho)))  ...
     -sigma*(u^2+1i*u)/(2*lambda)*(1-exp(-lambda*T)) ...
     + lambda*I(u);

c_1(u) = diff(kappa,u);

simplify(c_1(0))

But I get the error: Error in MuPAD command: Division by zero. [_power]

It might be a typo, but I am pretty sure there aren't any. In case I have not, and my friend has neither, it is solvable in Mathematica (I do not have access to or ever tried using the program my self). I've tried all sorts of stuff, but I cannot get Matlab to evaluate it.

I have 2014b of course with the symbolic toolbox (but it might be possible for me to update).

Any ideas?

[John D'Errico]

REALLY a bad idea to name a variable gamma. since there is a terribly useful function named gamma, that you might actually want to use one day. Get used to using a better name that does not cause a conflict, and your code will become more stable in the future.

[Henrik Dam]

Hi John. Thanks for the tip, I did not think of that. I will do so from now on.

[Walter Roberson]

Another potential difficulty occurs with using capital-I as a symbolic variable name: capital-I is how the symbolic toolbox represents sqrt(-1)

[Henrik Dam]

Thanks for the tip Walter :)