MathLabs
TheoremProved

The Hilbert distance is a genuine metric, with straight segments as geodesics

Statement

For Ω\Omega as above, dΩd_\Omega is a metric on Ω\Omega: it is symmetric (dΩ(x,y)=dΩ(y,x)d_\Omega(x,y)=d_\Omega(y,x)), non-negative, and dΩ(x,y)=0d_\Omega(x,y)=0 if and only if x=yx=y. Moreover, for any x,z∈Ωx,z\in\Omega and any point yy on the straight segment [x,z][x,z], dΩ(x,y)+dΩ(y,z)=dΩ(x,z)d_\Omega(x,y)+d_\Omega(y,z)=d_\Omega(x,z); consequently the ordinary Euclidean straight-line segment [x,z][x,z] (traversed at the appropriate non-uniform speed) is a geodesic of (Ω,dΩ)(\Omega,d_\Omega), and dΩd_\Omega satisfies the triangle inequality with equality exactly along such segments.

Why is it true?

The formula dΩ(x,y)=12ln⁡[a,x,y,b]d_\Omega(x,y) = \frac12\ln[a,x,y,b] is symmetric under swapping the roles of xx and yy together with aa and bb, because relabelling turns ∣ay∣∣bx∣∣ax∣∣by∣\frac{|ay||bx|}{|ax||by|} into itself; and because a<x<y<ba<x<y<b forces ∣ay∣>∣ax∣|ay|>|ax| and ∣bx∣>∣by∣|bx|>|by|, the cross-ratio is a product of two numbers each greater than 11, so it exceeds 11 exactly when x≠yx\neq y, making dΩd_\Omega strictly positive off the diagonal. The additivity along a fixed line is really a statement about ordinary numbers on the real line: cross-ratios computed from the same pair of boundary points a,ba,b telescope multiplicatively as a third collinear point is inserted, so a logarithm turns that telescoping product into a sum — the same phenomenon you can check numerically on any three points of a segment.

Proof sketch

Symmetry and positivity. Write [a,x,y,b]=∣ay∣∣ax∣⋅∣bx∣∣by∣[a,x,y,b]=\frac{|ay|}{|ax|}\cdot\frac{|bx|}{|by|}. Since a<x<y<ba<x<y<b on the line, ∣ay∣=∣ax∣+∣xy∣>∣ax∣|ay|=|ax|+|xy|>|ax| and ∣bx∣=∣by∣+∣xy∣>∣by∣|bx|=|by|+|xy|>|by|, so both factors exceed 11 when x≠yx\ne y, hence [a,x,y,b]>1[a,x,y,b]>1 and dΩ(x,y)>0d_\Omega(x,y)>0; when x=yx=y both factors equal 11. Relabelling (x,y,a,b)↦(y,x,b,a)(x,y,a,b)\mapsto(y,x,b,a) sends ∣ay∣∣bx∣∣ax∣∣by∣\frac{|ay||bx|}{|ax||by|} to ∣bx∣∣ay∣∣by∣∣ax∣\frac{|bx||ay|}{|by||ax|}, the identical number, so dΩ(x,y)=dΩ(y,x)d_\Omega(x,y)=d_\Omega(y,x).

Additivity on a line. Let x,y,zx,y,z be collinear with yy between xx and zz, and let a,ba,b be the (common) boundary intersections of that line. Writing all four one-dimensional distances along the line, [a,x,y,b]⋅[a,y,z,b]=∣ay∣∣ax∣∣bx∣∣by∣⋅∣az∣∣ay∣∣by∣∣bz∣=∣az∣∣ax∣∣bx∣∣bz∣=[a,x,z,b],[a,x,y,b]\cdot[a,y,z,b]=\frac{|ay|}{|ax|}\frac{|bx|}{|by|}\cdot\frac{|az|}{|ay|}\frac{|by|}{|bz|}=\frac{|az|}{|ax|}\frac{|bx|}{|bz|}=[a,x,z,b], since the factors ∣ay∣|ay| and ∣by∣|by| cancel. Taking 12ln⁡\frac12\ln of both sides gives dΩ(x,y)+dΩ(y,z)=dΩ(x,z)d_\Omega(x,y)+d_\Omega(y,z)=d_\Omega(x,z), exactly the equality case of the triangle inequality, so the straight segment [x,z][x,z] realizes the Hilbert distance and is a geodesic.

Triangle inequality off a line. For x,y,zx,y,z not collinear, convexity of Ω\Omega is what is needed: projecting the pair (x,z)(x,z) through yy only ever enlarges the relevant cross-ratio compared with going straight from xx to zz, because the boundary points seen from yy lie no closer to xx or zz than the boundary points seen directly along xzxz — this monotonicity, proved carefully using the convexity of Ω\Omega, gives dΩ(x,z)≤dΩ(x,y)+dΩ(y,z)d_\Omega(x,z)\le d_\Omega(x,y)+d_\Omega(y,z) in general (see Hilbert 1895; a full derivation is given in the Handbook of Hilbert Geometry, Ch. 1).

Topics that use this theorem

Step-by-step proofs

No step-by-step proof yet for this theorem.

References

  1. Athanase Papadopoulos, Marc Troyanov (eds.) (2014). Handbook of Hilbert Geometry · DOI:10.4171/147
  2. Athanase Papadopoulos, Marc Troyanov (2014). From Funk to Hilbert Geometry · arXiv:1406.6983
  3. David Hilbert (1895). Über die gerade Linie als kürzeste Verbindung zweier Punkte · DOI:10.1007/BF02096204