In mathematics, the Menger curvature of a triple of points in n-dimensional Euclidean space Rn is the reciprocal of the radius of the circle that passes through the three points. It is named after the Austrian-American mathematician Karl Menger.
Let x, y and z be three points in Rn; for simplicity, assume for the moment that all three points are distinct and do not lie on a single straight line. Let Π ⊆ Rn be the Euclidean plane spanned by x, y and z and let C ⊆ Π be the unique Euclidean circle in Π that passes through x, y and z (the circumcircle of x, y and z). Let R be the radius of C. Then the Menger curvature c(x, y, z) of x, y and z is defined by
c(x,y,z)=
1{R}. | |
If the three points are collinear, R can be informally considered to be +∞, and it makes rigorous sense to define c(x, y, z) = 0. If any of the points x, y and z are coincident, again define c(x, y, z) = 0.
Using the well-known formula relating the side lengths of a triangle to its area, it follows that
c(x,y,z)=
1{R} | |
= |
4A | |
|x-y||y-z||z-x| |
,
where A denotes the area of the triangle spanned by x, y and z.
Another way of computing Menger curvature is the identity
c(x,y,z)= | 2\sin\anglexyz |
|x-z| |
\anglexyz
Menger curvature may also be defined on a general metric space. If X is a metric space and x,y, and z are distinct points, let f be an isometry from
\{x,y,z\}
R2
cX(x,y,z)=c(f(x),f(y),f(z)).
Note that f need not be defined on all of X, just on , and the value cX (x,y,z) is independent of the choice of f.
Menger curvature can be used to give quantitative conditions for when sets in
Rn
\mu
Rn
cp(\mu)=\int\int\intc(x,y,z)pd\mu(x)d\mu(y)d\mu(z).
E\subseteqRn
c2(H1|E)<infty
H1|E
E
The basic intuition behind the result is that Menger curvature measures how straight a given triple of points are (the smaller
c(x,y,z)max\{|x-y|,|y-z|,|z-y|\}
p>3
f:S1 → Rn
\Gamma=f(S1)
f\in
| ||||
C |
(S1)
cp(H1|\Gamma)<infty
0<Hs(E)<infty
0<s\leq | 1 |
2 |
c2s(Hs|E)<infty
E
C1
\Gammai
Hs(E\backslashcup\Gammai)=0
1 | |
2 |
<s<1
c2s(Hs|E)=infty
1<s\leqn
In the opposite direction, there is a result of Peter Jones:[4]
E\subseteq\Gamma\subseteqR2
H1(E)>0
\Gamma
\mu
E
\muB(x,r)\leqr
x\inE
r>0
c2(\mu)<infty
H1(B(x,r)\cap\Gamma)\leqCr
x\in\Gamma
c2(H1|E)<infty
Analogous results hold in general metric spaces:[5]