← all systems · see it in The System
Gromov hyperbolic space
'A hyperbolic metric space is a metric space satisfying certain metric relations (depending quantitatively on a nonnegative real number delta) between points'; 'introduced by Mikhael Gromov, generalizes the metric properties of classical hyperbolic geometry and of trees. Hyperbolicity is a large-scale property'. X is delta-hyperbolic iff all x, y, z, w satisfy the four-point condition (x,z)_w >= min((x,y)_w, (y,z)_w) - delta.
Properties published scienceour proposal
Sourced fields, and our own readings.
| boundary | Inside: a metric space whose four-point relations differ from those of a tree by at most a fixed delta, with its slim geodesic triangles, its tree-like finite subsets, exponential circle growth, linear isoperimetric inequality, its boundary at infinity and the classification of its isometries. Outside: local metric detail (bounded regions are invisible to hyperbolicity), the groups that act on it, and spaces that are not hyperbolic (the Euclidean plane, the square grid). |
| nested level | Field scale, rung 1, below the first fold. A kind of metric space; a large-scale generalisation of hyperbolic space and of trees; hyperbolic n-space, the hyperbolic plane and negatively curved CAT spaces are examples. |
| scale | field |
Measures (9) published science
| hyperbolicity constant of the hyperbolic plane and Hadamard manifolds of curvature <= -1 | 2 delta |
| hyperbolicity constant of metric trees | 0 delta |
| incircle diameter of an ideal triangle in the hyperbolic plane | 2 log 3 length (curvature -1) |
| minimum vertex degree for a hyperbolic 1-skeleton of a plane triangulation | 7 degree |
| tree-approximation constant for n points | C = delta h(n), h(n) = O(log n), optimal length |
| base-point change cost | 2 delta |
| Gromov, 'Hyperbolic groups' (Essays in group theory) | 1987 year |
| Krioukov et al., Hyperbolic geometry of complex networks | 2010 year |
| characteristic clock: none of its own | not applicable (a static metric condition) none |
Cross-scale patterns seen here (2) our proposal
Shapes that recur when the same kind of problem is solved at different scales, and how each shows in Gromov hyperbolic space. Each links to every other system that shows it.
- same role, different builds (partial)
One large-scale shape built from smooth curvature (hyperbolic space, 2-hyperbolic), from branching (trees, 0-hyperbolic), from groups (Cayley graphs of surface groups) and from human networks (effective hyperbolic geometry of the Internet). - same shape, opposite sign (partial)
The degree-7 rule: a plane triangulation with every vertex of degree 6 is Euclidean (quasi-isometric to the plane), with every degree 7 or more it is hyperbolic; the same threshold that separates Euclidean and hyperbolic tilings.
Valence: what it trades, and with whom (6) published science
Each exchange with a neighbour, read from both sides: who gains, who pays, and what carries it.
- Internet via sys:spacetime · Gromov hyperbolic space even
- Euclidean space via sys:spacetime · Gromov hyperbolic space even
- Hyperbolic space via sys:spacetime · Gromov hyperbolic space gain
- Hyperbolic plane via sys:spacetime · Gromov hyperbolic space gain
- Metric space via sys:spacetime · Gromov hyperbolic space gain
- Metric tree (real tree, R-tree; walked as an idea) via sys:spacetime · Gromov hyperbolic space gain
Partners and parts published science
Built from / builds (2)
Sits inside / holds (3)
floor filled by (10)
- Internet
- Hyperbolic space
- Riemannian metric of constant negative curvature
- Hyperbolic plane
- Metric space
- Metric tree (real tree, R-tree; walked as an idea)
- Gromov-hyperbolic group (Cayley graph)
- CAT(k) space of negative curvature
- Curve complex
- Cantor set
represents (1)
Kinds (6)
- Hyperbolic space
- Metric space
- Metric tree (real tree, R-tree; walked as an idea)
- Gromov-hyperbolic group (Cayley graph)
- CAT(k) space of negative curvature
- Curve complex
wired (1)
Notes published science
- Seen from far enough away, every hyperbolic space is a tree: any n points sit within delta times O(log n) of a finite tree, and its asymptotic cones are real trees.
- A triangulated floor turns hyperbolic exactly when every corner has 7 or more triangles; with 6 it stays flat.
- Krioukov et al. argue the Internet's skewed degrees and clustering are what an underlying hyperbolic geometry looks like, and that routing by that hidden geometry stays efficient under severe damage.
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
Gromov hyperbolic space (own: large-scale negative curvature of a metric space)
- L1: A metric space (X, d); the definition via the Gromov product 'does not require the space to be geodesic'
- L2: Gromov product (y,z)_x = (1/2)(d(x,y) + d(x,z) - d(y,z)) and the four-point condition with constant delta; checking one base point suffices at constant 2 delta
- L3: Slim triangles (credited to Rips): every geodesic triangle is delta-slim, each side within delta of the other two; equivalent to the delta-centre definition up to a constant k
- L4: Large-scale invariants: invariant under quasi-isometry; any n points lie near a finite tree within C = delta h(n), h(n) = O(log n), optimal; circles grow exponentially; a linear isoperimetric inequality characterises it; asymptotic cones are real trees; local (bounded) detail is invisible
- L5: Boundary at infinity: classes of sequences converging to infinity (Gromov product to infinity); for a regular tree the space of ends, a Cantor set; for H^n the (n - 1)-sphere; homeomorphic to Busemann functions modulo translations for proper X
- L6: Isometries classified by their action on the boundary: elliptic (bounded orbit), hyperbolic (exactly two boundary fixed points), parabolic (exactly one); groups acting on hyperbolic spaces (hyperbolic groups, mapping class group via the curve complex, Out(Fn))
velocity space as a 2-hyperbolic space (rapidities at large scale)
- L1: The velocity hyperboloid is a copy of hyperbolic space, 'the prototype of a Gromov hyperbolic space'
- L2: 2-hyperbolic at curvature -1; scaling the metric by lambda gives lambda delta, so 'the more (negatively) curved the space is, the lower the hyperbolicity constant'
Who fills each layer
- L2: Metric tree (real tree, R-tree; walked as an idea)
- L4: Metric tree (real tree, R-tree; walked as an idea)