Singularities are important in differential geometryand algebraic geometry. Roughly speaking they are points at which a geometric object fails to be smooth.

I'll give a precise definition and then a discussion with some examples. Suppose that we have a subset X of Cn that is defined by the simultaneous vanishing+ of m polynomials f1(x1,...,xn),..., fm(x1,...,xn). Note that in one favourite model of the universe (as a Calabi-Yau manifold) space locally has this form.

Now take a point x on X. Form the Jacobian matrix J which is the mxn matrix with i,j entry dfi / dxj and evaluate at the point x, to obtain J(x). Then the point x is a singular point iff J(x) has rank strictly less than n - the dimension* of X. A point which is not singular is smooth.

OK that's the definition, what does it mean? Consider the special case of a hypersurface, that is we just have one defining equation f=f1. In this case the Jacobian is just a row (df/dx1 ... df/dxn) and X has dimension n-1. Thus for a singularity at x we are asking that the Jacobian matrix should vanish identically at x, that is all the partial derivatives of f have to vanish at that point.

For example, consider a quadratic cone with defining polynomial x2 + y2 + z2. This has a sharp looking point at the vertex of the cone at the origin but looks smooth everywhere else. This intuition fits with the definition because the Jacobian (2x 2y 2z) clearly vanishes at the vertex and nowhere else.

What we are seeing here illustrates a general principle, at most points of X the Jacobian matrix will have the correct rank n - the dimension of X and be smooth, and only at the smaller dimensional set defined by the vanishing of the appropriate minors of the Jacobian can it be singular.

Here's another example, consider the plane algebraic curve defined by y2-x3, (the cuspidal cubic curve). If you sketch this curve then you'll see it is symmetric about the x-axis. In the first quadrant as it moves away from the origin y is growing much faster than x. Thus the curve has a pronounced sharp point at the origin, which is indeed a singularity. At all other points of the curve you can draw a well-defined tangent line but at the origin this tangent line is not well-defined, there is a two-dimensional space in which we could draw tangents. In general, a singularity occurs where the tangent space is bigger than the dimension of the space under consideration.


+ Strictly speaking when I talk about X being defined by the simultaneous vanishing of f1,...,fm what I meam is that when I consider the ideal in the polynomial ring C[x1,...,xn] consisting of all polynomials which vanish on X then this ideal is generated by f1,...,fm.
* The dimension of such a set X given by the simultaneous vanishing of polynomials is defined as the maximum length of a chain of similarly defined subsets.