Solve function not giving complex solutions

Hello. For some reason solve does not seem to give all the existing solutions to simple equations. For instance, entering:

syms x
s=solve(x^2+1,x)

I get:

Warning: Explicit solution could not be found. > In solve at 179

s =

[ empty sym ]

The two solution I expected to see are of course +i and -i. I use Matlab R2014A. Any idea what might be causing this? Thanks in advance.

Answer 1

[Yoav B]

I contacted Mathworks support. It turns out that all is needed here is to clear the default assumptions for symbolic variables. The following code would do the job.

syms x clear

[Star Strider]

Interesting. Appreciate your following this up.

Answer 2

[Star Strider]

I got the same result. The Symbolic Math Toolbox very much seems to need tweaking.

Using vpasolve returns the correct result:

syms x 
s = vpasolve(x^2 + 1,x)

s =
  1.0i
 -1.0i

[Yoav B]

Thank you. Well, I gues I'll use vpasolve instead. The funny thing is that even the code given in the documentation

syms x
solve(x^4 + 1 == 2*x^2 - 1)

doesn't work. I wonder how this is possible.

[Star Strider]

My pleasure.

Beats me. I get the same result (in R2015b).

You discovered this, so I’ll let you report it to Contact Support as a quite definite bug. Include the URL of this thread in your message to them so you don’t have to repeat all of it.

Before you contact them, have MATLAB open, and type ver in the Command Window. Copy-paste the output to the appropriate window in the web page.

Best I can do!

Meanwhile, if you desperately need to do symbolic calculations, the Wolfram Alpha website will let you do an incredible number of things for the cost of the electricity and communications time to type them in and copy the result. I use it frequently when the Symbolic Math Toolbox (that is not free) leaves me in the lurch.

[John D'Errico]

Not a bug as far as I can see. Works fine. R2015b.

syms x
s=solve(x^2+1==0,x)
s =
 -1i
  1i

s=solve(x^2+1,x)
s =
 -1i
  1i

[Star Strider]

I got the same

s =
Empty sym: 0-by-1

result OP got in R2015b. That’s the reason I suggested the vpasolve approach. I don’t know what the problem is.

[Yoav B]

Thank you both. I'll report it.