Okay, good so far
Each time you iterate the for loop, t will be a scalar containing one of those values. You declared T to be tmax rows, same as the maximum value for t, so you are on track.
Each iteration of the for loop, you are overwriting all of T. So after the for loop finishes, T is going to contain a single value that is the one calculated in the last iteration of the loop.
That is setting up to write into 20 different rows for Y, which is fine so far.
X=(-1).^n.*exp(-alpha.*n.^2.*pi^2.*t/R^2);
Each iteration of the n loop, you overwrite all of X, and you do not write into Y at all.
but X is a scalar... and what is the point of having initialized Y if you are going to overwrite it with sum(X) ?
As described above, at each iteration of the for t loop, t will be a scalar. At the end of the loop, t will have the last value assigned to it, which would be tmax, and will be a scalar.
As explored above, you are overwriting T with a scalar for each iteration of the loop, so T is going to end up a scalar.
Hint:
Also
X=(-1).^n.*exp(-alpha.*n.^2.*pi^2.*t/R^2);
with n = 20 and t = 5000 the exp() term will be exp(-35479) approximately, which is far far too small to bother with.
You can demonstrate that with those coefficients and the maximum possible n, that there is no point doing t > 100 -- that beyond that you would underflow to 0.
Are you certain you want to proceed at integer t up to 5000? It would be more common to have a factor to convert that to a shorter time scale, such as t/1000 for bound of 5 seconds. Time 5000 seconds for an exponential process is not going to have anything useful left of the equation by 5000.
