fplot(@(x) x.^2, [eps(realmin), 10])
eps(realmin) is the smallest representable positive floating point number. That is as close as MATLAB numeric processing can get to not including 0 but including limit(epsilon, epsilon > 0), but it is only about 10^(-330) which is not even close to being satisfactory as sufficiently "infinitely small". You can switch to the Symbolic toolbox, but you with it you cannot do much better than 10^(-50000000), maybe 10^(-99999999) but no better, which is still infinitely too large to be a satisfactory approximation of infinitely small.
