This is probably a well-known issue, but I could not find a clear discussion in the literature, and I think others could find it useful.

Consider a real-analytic function germ $f:(\mathbb R^2,0) \rightarrow \mathbb R$, it is represented by a convergent power series $\sum_{i,j}a_{ij}x^iy^j \in \mathbb R\{x,y\}$. Suppose $f$ has a singularity at $0$, i.e. $df(0)=0$. Define its Jacobian ideal as the ideal generated by its partial derivatives: $I_{df} = \mathbb R\{x,y\}\langle f_x,f_y\rangle$, and its real local algebra as: $$Q_f := \frac{\mathbb R\{x,y\}}{I_{df}}.$$

Let $\mu_{\mathbb R} := \dim_{\mathbb R} Q_f$. The singularity is called algebraically isolated if $\mu_{\mathbb R}< \infty$. It is called isolated if, in some small neighborhood of $0$, $df(x,y)=0$ only at $(x,y)=0$ (i.e., the usual meaning).

Now take the complexification of $f$ by simply considering the series as a complex one: $\sum_{i,j}a_{ij}z^iw^j$. It is convergent in some polydisc and defines an holomorphic germ $f^{\mathbb C}:(\mathbb C^2,0) \rightarrow \mathbb C$. Take its (complex) Jacobian ideal, that is: $I_{df^{\mathbb C}} = \mathbb C\{x,y\}\langle f^{\mathbb C}_z,f^{\mathbb C}_w\rangle$, and the complex local algebra $$Q^{\mathbb C}_f := \frac{\mathbb C\{x,y\}}{I_{df^{\mathbb C}}}.$$

The Milnor number of the complexified singularity is defined by: $\mu_{\mathbb C} := \dim_{\mathbb C} Q_f^{\mathbb C}.$ A complex singularity is called isolated if $\mu_{\mathbb C} < \infty$. (In the holomorphic case it can be proven that a singularity is isolated if and only if, in some small neighborhood of $0$, $df(z,w)=0$ only at $(z,w)=0$, so we don't need to specify ''algebraically'' here.)


  1. What is the relationship between the real-analytic singularity and its complexification? In particular: between the two local algebras, and between $\mu_{\mathbb R}$ and $\mu_{\mathbb C}$? Is there a class of functions where things work better?
  2. Is true (or false) that a real-analytic singularity is algebraically isolated if and only if its complexification is isolated?
  3. In the real setting: is ''algebraically isolated'' stronger then simply ''isolated''?
  4. Any reference!

Thank you.

  • 3
    $\begingroup$ It seems to me that $Q_f^{\mathbb{C}} \cong Q_f \otimes \mathbb{C}$, so $\mu_{\mathbb{R}} = \mu_{\mathbb{C}}$. Am I missing something? $\endgroup$ – David E Speyer Feb 26 '16 at 13:51

As David Speyer has already suggested, we have $Q_f^\mathbb{C}\simeq Q_f\otimes\mathbb{C}$, so that answers the first question (and the second question).

For the third question, consider $f = (x^2+y^2)^2$. The singularity at $(x,y)=0$ is clearly isolated, but it is not algebraically isolated, since $\mu_\mathbb{R} = \infty$. Of course, over the complex numbers, $(z^2+w^2)^2$ has a non-isolated singularity at $(z,w)= (0,0)$; the two lines of singularities $z = \pm i\,w$ intersect at $(0,0)$.

The various confusions will be cleared up by noting that the ring of germs of analytic functions in two variables is a UFD, and, the singularity will be algebraically isolated as long as $f$ does not have any repeated irreducible factors. 'Isolated' is weaker over the reals simply because all the factors of $f$ might have no real zeros other than at the origin.

  • $\begingroup$ Thanks to both of you. I missed a clear understanding of $Q_f \otimes \mathbb C\simeq Q_f^{\mathbb C}$. Indeed one should verify a few, trivial here but necessary, things: the map $\mathbb R\{x,y\}\otimes \mathbb C \rightarrow \mathbb C\{z,w\}$, $f\otimes \alpha \mapsto \alpha f^C$, 1) respects operations and convergence, 2) has a well defined inverse (since the ''formal real and imaginary part'' of a convergent complex series, put in $\mathbb R\{x,y\}$, converge), 3) commutes with the operation of "taking $i$-th derivative", so maps $I_{df} \otimes\mathbb C$ onto $I_{df^{\mathbb C}}$. $\endgroup$ – peter Feb 26 '16 at 18:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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