getting Complex double in matlab while excel solves it

I had this formula in excel few years ago with less variation of WL, but now want to have a bigger range and put it into matlab and it gives me Complex double and the curve is totally different from what I expect or that came out in excel, can anyone help me here

WL=0:1E-6:4e-3;

SS=0.06977+7.0625*(1-exp(-(WL-0.26053)./0.15994)).^2.28411.*exp(-(WL-0.26053)./0.2285);

Answer 1

[Stephen23]

(1-exp(-(WL-0.26053)./0.15994)).^2.28411
%
%  ^^^^                      ^             positive values (around 5)
%^^^                          ^            negative values (around -4)
%                              ^^^^^^^^^   fractional power of negative values.

What do you expect a fractional power of negative values to give you?

>> (-4).^2.28411
ans =  14.884 + 18.473i

[Asliddin Komilov]

I know why its happening, the trouble is I am getting Complex double for the whole range of the outputs even for positive values.

[Stephen23]

" the trouble is I am getting Complex double for the whole range of the outputs even for positive values"

Please show an example of this: I would like to see a positive value raised to a positive fractional power that returns a complex result.

[Stephen23]

Asliddin Komilov's "Answer" moved here:

if you copy the following to the command window you will see that data will come out in complex, resulting in a plot as in 'screen2', at the same time similar calculated data gives a different curve ('screen1') in excel.

WL=1E-3:1E-3:4;
SS=0.06977+7.0625*(1-exp(-(WL-0.26053)./0.15994)).^2.28411.*exp(-(WL-0.26053)./0.2285)

[Stephen23]

As I explained in my answer, you are calculating the exponent of a negative number using a fractional power, which produces a complex result. Taking a look at your example:

>> WL=1E-3:1E-3:2;
>> numel(WL)
ans = 2000
>> X = 1-exp(-(WL-0.26053)./0.15994);
>> Y = Y.^2.28411;
>> find(X(:)<0) % indices of the negative values!
1
2
3
4
...  lots of sequential indices here
259
260
>> find(imag(Y(:))>0) % indices of where negative values -> complex result.
1
2
3
4
...  lots of sequential indices here
259
260

You can easily avoid taking the exponent of negative values using max:

>> Z = max(0,1-exp(-(WL-0.26053)./0.15994)).^2.28411
>> find(imag(Z)>0)
ans = []

Or set those values to NaN, or whatever suits your algorithm.

"getting Complex double in matlab while excel solves it"

In fact your screenshots make it clear that MATALB solves it (correctly, giving complex results for those negative values), whereas Excel just returns a an error message.

[Asliddin Komilov]

it worked, thanks