Look at the X and Y limits on your MATLAB plot, then look at the X and Y limits on the plot from Wolfram Alpha.
MATLAB is giving you (a small section of) the same plot as Wolfram Alpha. Adjust your meshgrid call.
And FYI, instead of computing 2 * (denominator).^(-1) you could just use 2./denominator. That looks a little more like the original equation, doesn't it?
Once you have the adjusted plot, to prove to yourself that MATLAB got it right originally set the axes limits for the adjusted plot to the values they had in the plot in your screenshot above and compare with that screenshot.
axis([-2 2 -2 2 0 0.2])
They look pretty similar. Then restore the limits to their automatically chosen values to see the whole surface.
axis auto
