Have another look at this loop:
for t=i*j
t(0,:)=0;
t(2.5,:)=0;
t(:,0)=0;
t(:,2.0)=0;
for t=~0
t(x,y)=(t(x+1,y)+t(x-1,y)+t(x,y+1)+t(x,y-1))/4;
end
end
You have not assigned anything to i or j, so both values will have their default meaning of sqrt(-1) . sqrt(-1)*sqrt(-1) is -1, so your outer loop is "for t=-1" which means that the loop is to be executed once with the value of t set to the scalar value -1 . Inside the loop, you attempt to access t(0,:) . As t is a scalar, it's last dimension is 1, so that will be the same as trying to access t(0,1) which will fail because arrays cannot be subscripted at 0.
If it did not fail there, it would fail at the next line, t(2.5,:) as it is not allowed to access arrays at fractional indices.
If you managed to get through the next several lines, you would hit your "for t=~0" . ~0 is the scalar containing the logical value "true", so you would be asking for the loop to be executed once with t temporarily assigned true . And then you would try to access that 1x1 scalar at all kinds of non-integral points defined by x and y...
Arrays must be indexed with non-negative integers. The array position might logically correspond to some x value such as 2.5 but you need to maintain that list of correspondences independently. If you have a point in logical coordinates that you want to convert to array coordinates, histc() or mod() or interp1() can be useful.
