Affine geometry of curves explained

SL(n,R)\ltimesRn.

In the classical Euclidean geometry of curves, the fundamental tool is the Frenet - Serret frame. In affine geometry, the Frenet - Serret frame is no longer well-defined, but it is possible to define another canonical moving frame along a curve which plays a similar decisive role. The theory was developed in the early 20th century, largely from the efforts of Wilhelm Blaschke and Jean Favard.

The affine frame

Let x(t) be a curve in

Rn

. Assume, as one does in the Euclidean case, that the first n derivatives of x(t) are linearly independent so that, in particular, x(t) does not lie in any lower-dimensional affine subspace of

R2

. Then the curve parameter t can be normalized by setting determinant

\det\begin{bmatrix}

x

,&\ddot{x

}, &\dots, &^ \end = \pm 1.

Such a curve is said to be parametrized by its affine arclength. For such a parameterization,

t\mapsto[x(t),

x

(t),...,x(n)(t)]

determines a mapping into the special affine group, known as a special affine frame for the curve. That is, at each point of the quantities

x,x

,...,x(n)

define a special affine frame for the affine space

Rn

, consisting of a point x of the space and a special linear basis
x

,...,x(n)

attached to the point at x. The pullback of the Maurer - Cartan form along this map gives a complete set of affine structural invariants of the curve. In the plane, this gives a single scalar invariant, the affine curvature of the curve.

Discrete invariant

The normalization of the curve parameter s was selected above so that

\det\begin{bmatrix}

x

,&\ddot{x

}, &\dots, &^ \end = \pm 1.If n≡0 (mod 4) or n≡3 (mod 4) then the sign of this determinant is a discrete invariant of the curve. A curve is called dextrorse (right winding, frequently weinwendig in German) if it is +1, and sinistrorse (left winding, frequently hopfenwendig in German) if it is -1.

In three-dimensions, a right-handed helix is dextrorse, and a left-handed helix is sinistrorse.

Curvature

Suppose that the curve x in

Rn

is parameterized by affine arclength. Then the affine curvatures, k1, …, kn−1 of x are defined by

x(n+1)=

k
1x

+ … +kn-1x(n-1).

That such an expression is possible follows by computing the derivative of the determinant

0=\det\begin{bmatrix}

x

,&\ddot{x

}, &\dots, &^ \end\dot\, = \det \begin\dot, &\ddot, &\dots, &^ \end

so that x(n+1) is a linear combination of x′, …, x(n-1).

Consider the matrix

A=\begin{bmatrix}

x

,&\ddot{x

}, &\dots, &^ \end

whose columns are the first n derivatives of x (still parameterized by special affine arclength). Then,

A

=\begin{bmatrix}0&1&0&0&&0&0\\ 0&0&1&0&&0&0\\ \vdots&\vdots&\vdots&&&\vdots&\vdots\\ 0&0&0&0&&1&0\\ 0&0&0&0&&0&1\\ k1&k2&k3&k4&&kn-1&0 \end{bmatrix}A=CA.

In concrete terms, the matrix C is the pullback of the Maurer - Cartan form of the special linear group along the frame given by the first n derivatives of x.

See also

References

  • Book: Guggenheimer, Heinrich. Differential Geometry. 1977. Dover. 0-486-63433-7.
  • Book: Spivak, Michael. Michael Spivak

. Michael Spivak. A Comprehensive introduction to differential geometry (Volume 2). 1999. Publish or Perish. 0-914098-71-3.