I'm trying to get the following result confirmed in Mathematica: $$ \left.\frac{\partial\mathop{K_{\nu}}\nolimits\!\left(x\right)}{\partial\nu}% \right|_{\nu=\pm\frac{1}{2}}=\pm\sqrt{\frac{\pi}{2x}}\mathop{E_{1}}\nolimits% \!\left(2x\right)e^{x}, $$ cf. http://dlmf.nist.gov/10.38.E7. Here $K_{\nu}(z)$ denotes the MacDonald function and $E_1$ denotes the exponential integral.

However, Mathematica seems to be unaware of the above result. All I get for

D[BesselK[a, z], {a, 1}] /. a -> 1/2


$$ \text{BesselK}^{(1,0)}\left(\frac{1}{2},z\right). $$

Can Mathematica be made more specific here?

  • 2
    $\begingroup$ FullSimplify[FunctionExpand[Derivative[1, 0][BesselK][1/2, z]]] might be somewhat closer. $\endgroup$ Mar 21, 2016 at 20:47
  • $\begingroup$ ...and of course the result in b.gatessucks's comment can be massaged into the required form. $\endgroup$ Mar 21, 2016 at 22:10

1 Answer 1


As noted by b.gatessucks, a preliminary application of FunctionExpand[] allows the confirmation of this identity:

FullSimplify[FunctionExpand[Derivative[1, 0][BesselK][1/2, z]] ==
             Sqrt[π/2] ExpIntegralE[1, 2 z] Exp[z]/Sqrt[z]]

FullSimplify[FunctionExpand[Derivative[1, 0][BesselK][-1/2, z]] ==
             -Sqrt[π/2] ExpIntegralE[1, 2 z] Exp[z]/Sqrt[z]]

It should be noted that the particular form of the result in Mathematica uses $\frac1{\sqrt{z}}$ instead of $\sqrt{\frac1z}$, which corresponds to a different choice of branch cut on the negative real axis.

Only recently has a formula for $\frac{\partial}{\partial\nu}K_\nu(z)$ for general values of $\nu$ been found; see e.g. Brychkov's article and this accompanying blog post.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.