3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11....
Transcript of 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11....
![Page 1: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/1.jpg)
3D Ricci flow since Perelman
John LottUC-Berkeley
http://math.berkeley.edu/˜lott
October 25, 2015
![Page 2: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/2.jpg)
Outline of the talk
1. Homogeneous spaces and the geometrization conjecture
2. The geometrization conjecture and Ricci flow
3. Finiteness of the number of surgeries
4. Long-time behavior
5. Flowing through singularities
![Page 3: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/3.jpg)
3D Ricci flow since Perelman
Homogeneous spaces and the geometrization conjecture
Geometrization conjecture and Ricci flow
Finiteness of the number of surgeries
Long-time behavior
Flowing through singularities
![Page 4: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/4.jpg)
A brief history of math
18th century : one dimensional spaces
19th century : two dimensional spaces
![Page 5: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/5.jpg)
A brief history of math
18th century : one dimensional spaces
19th century : two dimensional spaces
![Page 6: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/6.jpg)
A brief history of math
20th century : three dimensional spaces
21st century :
![Page 7: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/7.jpg)
A brief history of math
20th century : three dimensional spaces
21st century :
![Page 8: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/8.jpg)
Today’s talk
Remaining open questions from 20th century math.
First, how do we understand three dimensional spaces?
In terms of homogeneous spaces.
![Page 9: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/9.jpg)
Today’s talk
Remaining open questions from 20th century math.
First, how do we understand three dimensional spaces?
In terms of homogeneous spaces.
![Page 10: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/10.jpg)
Today’s talk
Remaining open questions from 20th century math.
First, how do we understand three dimensional spaces?
In terms of homogeneous spaces.
![Page 11: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/11.jpg)
Locally homogeneous metric spaces
A metric space X is locally homogeneous if all x , y ∈ X , thereare neighbourhoods U and V of x and y and an isometricisomorphism (U, x)→ (V , y).
The metric space X is globally homogeneous if for all x , y ∈ X ,there is an isometric isomorphism φ : X → X that φ(x) = y .
![Page 12: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/12.jpg)
Locally homogeneous metric spaces
A metric space X is locally homogeneous if all x , y ∈ X , thereare neighbourhoods U and V of x and y and an isometricisomorphism (U, x)→ (V , y).
The metric space X is globally homogeneous if for all x , y ∈ X ,there is an isometric isomorphism φ : X → X that φ(x) = y .
![Page 13: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/13.jpg)
Locally homogeneous Riemannian manifolds
Any Riemannian manifold M gets a metric space structure.
Theorem(Singer 1960) If M is a complete, simply connected Riemannianmanifold which is locally homogeneous, then M is globallyhomogeneous.
![Page 14: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/14.jpg)
Locally homogeneous Riemannian manifolds
Any Riemannian manifold M gets a metric space structure.
Theorem(Singer 1960) If M is a complete, simply connected Riemannianmanifold which is locally homogeneous, then M is globallyhomogeneous.
![Page 15: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/15.jpg)
From “The Geometries of 3-Manifolds” by Peter Scott
We will say that a smooth manifold M admits a geometricstructure if M admits a complete, locally homogeneousRiemannian metric.
It is a theorem of Singer that such a metric on a simplyconnected manifold X must be homogeneous, i.e. the isometrygroup of X must act transitively.
Thus we can regard the universal cover X of M, together withits isometry group, as a geometry in the sense of Klein, and wecan sensibly say that M admits a geometric structure modelledon X . Thurston has classified the 3-dimensional geometriesand there are eight of them.
![Page 16: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/16.jpg)
From “The Geometries of 3-Manifolds” by Peter Scott
We will say that a smooth manifold M admits a geometricstructure if M admits a complete, locally homogeneousRiemannian metric.
It is a theorem of Singer that such a metric on a simplyconnected manifold X must be homogeneous, i.e. the isometrygroup of X must act transitively.
Thus we can regard the universal cover X of M, together withits isometry group, as a geometry in the sense of Klein, and wecan sensibly say that M admits a geometric structure modelledon X . Thurston has classified the 3-dimensional geometriesand there are eight of them.
![Page 17: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/17.jpg)
Two-dimensional geometries
Globally homogeneous S2,
locally homogeneous
Globally homogeneous R2, locally homogeneous
Globally homogeneous H2, locally homogeneous
![Page 18: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/18.jpg)
Two-dimensional geometries
Globally homogeneous S2, locally homogeneous
Globally homogeneous R2, locally homogeneous
Globally homogeneous H2, locally homogeneous
![Page 19: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/19.jpg)
Two-dimensional geometries
Globally homogeneous S2, locally homogeneous
Globally homogeneous R2,
locally homogeneous
Globally homogeneous H2, locally homogeneous
![Page 20: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/20.jpg)
Two-dimensional geometries
Globally homogeneous S2, locally homogeneous
Globally homogeneous R2, locally homogeneous
Globally homogeneous H2, locally homogeneous
![Page 21: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/21.jpg)
Two-dimensional geometries
Globally homogeneous S2, locally homogeneous
Globally homogeneous R2, locally homogeneous
Globally homogeneous H2,
locally homogeneous
![Page 22: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/22.jpg)
Two-dimensional geometries
Globally homogeneous S2, locally homogeneous
Globally homogeneous R2, locally homogeneous
Globally homogeneous H2, locally homogeneous
![Page 23: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/23.jpg)
Three-dimensional Thurston geometries
S3, R3, H3
S2 × R, H2 × R
Nil, Sol, ˜SL(2,R)
These are all globally homogeneous.
Warning : unlike in two dimensions, not every compactthree-dimensional manifold admits a geometric structure, i.e.admits a locally homogeneous Riemannian metric.
![Page 24: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/24.jpg)
Three-dimensional Thurston geometries
S3, R3, H3
S2 × R, H2 × R
Nil, Sol, ˜SL(2,R)
These are all globally homogeneous.
Warning : unlike in two dimensions, not every compactthree-dimensional manifold admits a geometric structure, i.e.admits a locally homogeneous Riemannian metric.
![Page 25: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/25.jpg)
Three-dimensional Thurston geometries
S3, R3, H3
S2 × R, H2 × R
Nil, Sol, ˜SL(2,R)
These are all globally homogeneous.
Warning : unlike in two dimensions, not every compactthree-dimensional manifold admits a geometric structure, i.e.admits a locally homogeneous Riemannian metric.
![Page 26: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/26.jpg)
Three-dimensional Thurston geometries
S3, R3, H3
S2 × R, H2 × R
Nil, Sol, ˜SL(2,R)
These are all globally homogeneous.
Warning : unlike in two dimensions, not every compactthree-dimensional manifold admits a geometric structure, i.e.admits a locally homogeneous Riemannian metric.
![Page 27: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/27.jpg)
Three-dimensional Thurston geometries
S3, R3, H3
S2 × R, H2 × R
Nil, Sol, ˜SL(2,R)
These are all globally homogeneous.
Warning : unlike in two dimensions, not every compactthree-dimensional manifold admits a geometric structure, i.e.admits a locally homogeneous Riemannian metric.
![Page 28: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/28.jpg)
Geometrization conjecture
If M is a compact orientable 3-manifold then there is acanonical way to split M into pieces, using certain embedded2-spheres and 2-tori. (The collection of 2-spheres and 2-toricould be empty.)
Conjecture (Thurston, 1982)
The ensuing pieces have geometric structures, i.e. admit locallyhomogeneous metrics
![Page 29: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/29.jpg)
Geometrization conjecture
If M is a compact orientable 3-manifold then there is acanonical way to split M into pieces, using certain embedded2-spheres and 2-tori. (The collection of 2-spheres and 2-toricould be empty.)
Conjecture (Thurston, 1982)
The ensuing pieces have geometric structures, i.e. admit locallyhomogeneous metrics
![Page 30: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/30.jpg)
Geometrization conjecture
If M is a compact orientable 3-manifold then there is acanonical way to split M into pieces, using certain embedded2-spheres and 2-tori. (The collection of 2-spheres and 2-toricould be empty.)
Conjecture (Thurston, 1982)
The ensuing pieces have geometric structures, i.e. admit locallyhomogeneous metrics
![Page 31: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/31.jpg)
Geometric decomposition
Cut along the 2-spheres and cap off the resulting pieces with3-balls.
Cut along the 2-tori. The interiors of the ensuing pieces shouldadmit complete locally homogeneous metrics.
![Page 32: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/32.jpg)
Geometric decomposition
Cut along the 2-spheres and cap off the resulting pieces with3-balls.
Cut along the 2-tori. The interiors of the ensuing pieces shouldadmit complete locally homogeneous metrics.
![Page 33: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/33.jpg)
Geometric decomposition
Cut along the 2-spheres and cap off the resulting pieces with3-balls.
Cut along the 2-tori. The interiors of the ensuing pieces shouldadmit complete locally homogeneous metrics.
![Page 34: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/34.jpg)
Geometric decomposition
Cut along the 2-spheres and cap off the resulting pieces with3-balls.
Cut along the 2-tori. The interiors of the ensuing pieces shouldadmit complete locally homogeneous metrics.
![Page 35: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/35.jpg)
Poincare conjecture
The geometrization conjecture implies the Poincare conjecture :
Suppose that the original 3-manifold M is simply connected.
By Van Kampen’s theorem, after we cut along 2-spheres andadd caps, the resulting components {Mi} are still simplyconnected. (π1(M) = ?iπ1(Mi).)
Any 2-torus T in the canonical decomposition is supposed tohave π1(T )→ π1(Mi) injective. Since π1(T ) = Z2 andπ1(Mi) = {e}, there can’t be such tori.
The geometrization conjecture now says that each Mi has alocally homogeneous metric. Since Mi is simply connected andcompact, it must be diffeomorphic to S3.
Then the original 3-manifold M is a “connected sum” of3-spheres, and is also diffeomorphic to S3.
![Page 36: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/36.jpg)
Poincare conjecture
The geometrization conjecture implies the Poincare conjecture :
Suppose that the original 3-manifold M is simply connected.
By Van Kampen’s theorem, after we cut along 2-spheres andadd caps, the resulting components {Mi} are still simplyconnected. (π1(M) = ?iπ1(Mi).)
Any 2-torus T in the canonical decomposition is supposed tohave π1(T )→ π1(Mi) injective. Since π1(T ) = Z2 andπ1(Mi) = {e}, there can’t be such tori.
The geometrization conjecture now says that each Mi has alocally homogeneous metric. Since Mi is simply connected andcompact, it must be diffeomorphic to S3.
Then the original 3-manifold M is a “connected sum” of3-spheres, and is also diffeomorphic to S3.
![Page 37: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/37.jpg)
Poincare conjecture
The geometrization conjecture implies the Poincare conjecture :
Suppose that the original 3-manifold M is simply connected.
By Van Kampen’s theorem, after we cut along 2-spheres andadd caps, the resulting components {Mi} are still simplyconnected. (π1(M) = ?iπ1(Mi).)
Any 2-torus T in the canonical decomposition is supposed tohave π1(T )→ π1(Mi) injective. Since π1(T ) = Z2 andπ1(Mi) = {e}, there can’t be such tori.
The geometrization conjecture now says that each Mi has alocally homogeneous metric. Since Mi is simply connected andcompact, it must be diffeomorphic to S3.
Then the original 3-manifold M is a “connected sum” of3-spheres, and is also diffeomorphic to S3.
![Page 38: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/38.jpg)
Poincare conjecture
The geometrization conjecture implies the Poincare conjecture :
Suppose that the original 3-manifold M is simply connected.
By Van Kampen’s theorem, after we cut along 2-spheres andadd caps, the resulting components {Mi} are still simplyconnected. (π1(M) = ?iπ1(Mi).)
Any 2-torus T in the canonical decomposition is supposed tohave π1(T )→ π1(Mi) injective. Since π1(T ) = Z2 andπ1(Mi) = {e}, there can’t be such tori.
The geometrization conjecture now says that each Mi has alocally homogeneous metric. Since Mi is simply connected andcompact, it must be diffeomorphic to S3.
Then the original 3-manifold M is a “connected sum” of3-spheres, and is also diffeomorphic to S3.
![Page 39: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/39.jpg)
Poincare conjecture
The geometrization conjecture implies the Poincare conjecture :
Suppose that the original 3-manifold M is simply connected.
By Van Kampen’s theorem, after we cut along 2-spheres andadd caps, the resulting components {Mi} are still simplyconnected. (π1(M) = ?iπ1(Mi).)
Any 2-torus T in the canonical decomposition is supposed tohave π1(T )→ π1(Mi) injective. Since π1(T ) = Z2 andπ1(Mi) = {e}, there can’t be such tori.
The geometrization conjecture now says that each Mi has alocally homogeneous metric. Since Mi is simply connected andcompact, it must be diffeomorphic to S3.
Then the original 3-manifold M is a “connected sum” of3-spheres, and is also diffeomorphic to S3.
![Page 40: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/40.jpg)
Poincare conjecture
The geometrization conjecture implies the Poincare conjecture :
Suppose that the original 3-manifold M is simply connected.
By Van Kampen’s theorem, after we cut along 2-spheres andadd caps, the resulting components {Mi} are still simplyconnected. (π1(M) = ?iπ1(Mi).)
Any 2-torus T in the canonical decomposition is supposed tohave π1(T )→ π1(Mi) injective. Since π1(T ) = Z2 andπ1(Mi) = {e}, there can’t be such tori.
The geometrization conjecture now says that each Mi has alocally homogeneous metric. Since Mi is simply connected andcompact, it must be diffeomorphic to S3.
Then the original 3-manifold M is a “connected sum” of3-spheres, and is also diffeomorphic to S3.
![Page 41: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/41.jpg)
3D Ricci flow since Perelman
Homogeneous spaces and the geometrization conjecture
Geometrization conjecture and Ricci flow
Finiteness of the number of surgeries
Long-time behavior
Flowing through singularities
![Page 42: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/42.jpg)
Ricci flow approach to geometrization
Hamilton’s Ricci flow equation
dgdt
= − 2 Ricg .
This is like a nonlinear heat equation for a Riemannian metric g.
The ordinary heat equation
dfdt
= 4f
acts on functions f on a fixed (compact connected) Riemannianmanifold M. It takes an initial function f0 and evolves it intosomething homogeneous (i.e. constant).
Maybe the Ricci flow will evolve an initial Riemannian metricinto something homogeneous.
![Page 43: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/43.jpg)
Ricci flow approach to geometrization
Hamilton’s Ricci flow equation
dgdt
= − 2 Ricg .
This is like a nonlinear heat equation for a Riemannian metric g.
The ordinary heat equation
dfdt
= 4f
acts on functions f on a fixed (compact connected) Riemannianmanifold M. It takes an initial function f0 and evolves it intosomething homogeneous (i.e. constant).
Maybe the Ricci flow will evolve an initial Riemannian metricinto something homogeneous.
![Page 44: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/44.jpg)
Ricci flow approach to geometrization
Hamilton’s Ricci flow equation
dgdt
= − 2 Ricg .
This is like a nonlinear heat equation for a Riemannian metric g.
The ordinary heat equation
dfdt
= 4f
acts on functions f on a fixed (compact connected) Riemannianmanifold M. It takes an initial function f0 and evolves it intosomething homogeneous (i.e. constant).
Maybe the Ricci flow will evolve an initial Riemannian metricinto something homogeneous.
![Page 45: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/45.jpg)
Ricci flow approach to geometrization
Hamilton’s Ricci flow equation
dgdt
= − 2 Ricg .
This is like a nonlinear heat equation for a Riemannian metric g.
The ordinary heat equation
dfdt
= 4f
acts on functions f on a fixed (compact connected) Riemannianmanifold M. It takes an initial function f0 and evolves it intosomething homogeneous (i.e. constant).
Maybe the Ricci flow will evolve an initial Riemannian metricinto something homogeneous.
![Page 46: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/46.jpg)
Surfaces
For the Ricci flow on a compact surface, after rescaling themetric approaches a locally homogeneous metric.
![Page 47: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/47.jpg)
Surfaces
For the Ricci flow on a compact surface, after rescaling themetric approaches a locally homogeneous metric.
![Page 48: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/48.jpg)
Surfaces
For the Ricci flow on a compact surface, after rescaling themetric approaches a locally homogeneous metric.
![Page 49: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/49.jpg)
Surfaces
For the Ricci flow on a compact surface, after rescaling themetric approaches a locally homogeneous metric.
![Page 50: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/50.jpg)
Singularities in 3D Ricci flow
Some components may disappear, e.g. a round shrinking3-sphere.
![Page 51: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/51.jpg)
Neckpinch
A 2-sphere pinches off. (Drawn one dimension down.)
![Page 52: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/52.jpg)
Neckpinch
A 2-sphere pinches off. (Drawn one dimension down.)
![Page 53: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/53.jpg)
Neckpinch
A 2-sphere pinches off. (Drawn one dimension down.)
![Page 54: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/54.jpg)
Hamilton’s idea of surgery
![Page 55: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/55.jpg)
Hamilton’s idea of surgery
![Page 56: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/56.jpg)
Hamilton’s idea of surgery
![Page 57: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/57.jpg)
Role of singularities
Singularities are good because we know that in general, wehave to cut along some 2-spheres to see the geometric pieces.
They are also problematic because they may cause lots oftopologically trivial surgeries. (Spitting out 3-spheres.)
Remark : the surgeries are done on 2-spheres, not 2-tori.
![Page 58: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/58.jpg)
Role of singularities
Singularities are good because we know that in general, wehave to cut along some 2-spheres to see the geometric pieces.
They are also problematic because they may cause lots oftopologically trivial surgeries. (Spitting out 3-spheres.)
Remark : the surgeries are done on 2-spheres, not 2-tori.
![Page 59: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/59.jpg)
Role of singularities
Singularities are good because we know that in general, wehave to cut along some 2-spheres to see the geometric pieces.
They are also problematic because they may cause lots oftopologically trivial surgeries. (Spitting out 3-spheres.)
Remark : the surgeries are done on 2-spheres, not 2-tori.
![Page 60: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/60.jpg)
Role of singularities
Singularities are good because we know that in general, wehave to cut along some 2-spheres to see the geometric pieces.
They are also problematic because they may cause lots oftopologically trivial surgeries. (Spitting out 3-spheres.)
Remark : the surgeries are done on 2-spheres, not 2-tori.
![Page 61: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/61.jpg)
Intuitive way to prove the geometrization conjectureusing Ricci flow
Step 1 : Show that one can perform surgery.
a. Show that singularities are only caused by componentsdisappearing or by 2-spheres pinching down.
b. Show that the surgery times do not accumulate.
Step 2 : Show that only a finite number of surgeries occur.
Step 3 : Show that after the singularities are over, as timeevolves, the locally homogeneous pieces in the Thurstondecomposition asymptotically appear.
(Relevant geometries : R3, H3, H2 × R, ˜SL(2,R), Sol, Nil.)
![Page 62: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/62.jpg)
Intuitive way to prove the geometrization conjectureusing Ricci flow
Step 1 : Show that one can perform surgery.
a. Show that singularities are only caused by componentsdisappearing or by 2-spheres pinching down.
b. Show that the surgery times do not accumulate.
Step 2 : Show that only a finite number of surgeries occur.
Step 3 : Show that after the singularities are over, as timeevolves, the locally homogeneous pieces in the Thurstondecomposition asymptotically appear.
(Relevant geometries : R3, H3, H2 × R, ˜SL(2,R), Sol, Nil.)
![Page 63: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/63.jpg)
Intuitive way to prove the geometrization conjectureusing Ricci flow
Step 1 : Show that one can perform surgery.
a. Show that singularities are only caused by componentsdisappearing or by 2-spheres pinching down.
b. Show that the surgery times do not accumulate.
Step 2 : Show that only a finite number of surgeries occur.
Step 3 : Show that after the singularities are over, as timeevolves, the locally homogeneous pieces in the Thurstondecomposition asymptotically appear.
(Relevant geometries : R3, H3, H2 × R, ˜SL(2,R), Sol, Nil.)
![Page 64: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/64.jpg)
Intuitive way to prove the geometrization conjectureusing Ricci flow
Step 1 : Show that one can perform surgery.
a. Show that singularities are only caused by componentsdisappearing or by 2-spheres pinching down.
b. Show that the surgery times do not accumulate.
Step 2 : Show that only a finite number of surgeries occur.
Step 3 : Show that after the singularities are over, as timeevolves, the locally homogeneous pieces in the Thurstondecomposition asymptotically appear.
(Relevant geometries : R3, H3, H2 × R, ˜SL(2,R), Sol, Nil.)
![Page 65: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/65.jpg)
Intuitive way to prove the geometrization conjectureusing Ricci flow
Step 1 : Show that one can perform surgery.
a. Show that singularities are only caused by componentsdisappearing or by 2-spheres pinching down.
b. Show that the surgery times do not accumulate.
Step 2 : Show that only a finite number of surgeries occur.
Step 3 : Show that after the singularities are over, as timeevolves, the locally homogeneous pieces in the Thurstondecomposition asymptotically appear.
(Relevant geometries : R3, H3, H2 × R, ˜SL(2,R), Sol, Nil.)
![Page 66: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/66.jpg)
Intuitive way to prove the geometrization conjectureusing Ricci flow
Step 1 : Show that one can perform surgery.
a. Show that singularities are only caused by componentsdisappearing or by 2-spheres pinching down.
b. Show that the surgery times do not accumulate.
Step 2 : Show that only a finite number of surgeries occur.
Step 3 : Show that after the singularities are over, as timeevolves, the locally homogeneous pieces in the Thurstondecomposition asymptotically appear.
(Relevant geometries : R3, H3, H2 × R, ˜SL(2,R), Sol, Nil.)
![Page 67: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/67.jpg)
Perelman’s work
Step 1 : Show that one can perform surgery.
a. Show that singularities are only caused by componentsdisappearing or by 2-spheres pinching down.
b. Show that the surgery times do not accumulate.
Done by Perelman.
![Page 68: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/68.jpg)
Perelman’s work
Step 1 : Show that one can perform surgery.
a. Show that singularities are only caused by componentsdisappearing or by 2-spheres pinching down.
b. Show that the surgery times do not accumulate.
Done by Perelman.
![Page 69: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/69.jpg)
Perelman’s work
Step 2 : Show that only a finite number of surgeries occur.
From Perelman’s first Ricci flow paper : Moreover, it can beshown ... that the solution is smooth (if nonempty) from somefinite time on.
From Perelman’s second Ricci flow paper : This is a technicalpaper, which is a continuation of [I]. Here we verify most of theassertions, made in [I, §13]; the exceptions are ... thesmoothness of the solution from some time on, which turnedout to be unjustified, and, on the other hand, irrelevant for theother conclusions.
![Page 70: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/70.jpg)
Perelman’s work
Step 2 : Show that only a finite number of surgeries occur.
From Perelman’s first Ricci flow paper : Moreover, it can beshown ... that the solution is smooth (if nonempty) from somefinite time on.
From Perelman’s second Ricci flow paper : This is a technicalpaper, which is a continuation of [I]. Here we verify most of theassertions, made in [I, §13]; the exceptions are ... thesmoothness of the solution from some time on, which turnedout to be unjustified, and, on the other hand, irrelevant for theother conclusions.
![Page 71: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/71.jpg)
Perelman’s work
Step 2 : Show that only a finite number of surgeries occur.
From Perelman’s first Ricci flow paper : Moreover, it can beshown ... that the solution is smooth (if nonempty) from somefinite time on.
From Perelman’s second Ricci flow paper : This is a technicalpaper, which is a continuation of [I]. Here we verify most of theassertions, made in [I, §13]; the exceptions are ... thesmoothness of the solution from some time on, which turnedout to be unjustified, and, on the other hand, irrelevant for theother conclusions.
![Page 72: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/72.jpg)
What Perelman actually showed
For any t , one can define a “thick-thin” decomposition of thetime-t manifold (assuming that it’s nonsingular). Then for largebut finite t , the following properties hold.
1. The interior of the thick part carries a complete finite-volumehyperbolic metric. (This uses Ricci flow.)
2. The thin part is a “graph manifold”. (This doesn’t use Ricciflow. Stated by Perelman, proofs by Shioya-Yamaguchi,Morgan-Tian, Bessieres-Besson-Boileau-Maillot-Porti andKleiner-L.)
3. The interface between the thick and thin parts consists of“incompressible” 2-tori (Hamilton).
Graph manifolds were known to have a geometricdecomposition. Along with knowledge of the topological effectsof surgeries, this proved the geometrization conjecture.
![Page 73: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/73.jpg)
What Perelman actually showed
For any t , one can define a “thick-thin” decomposition of thetime-t manifold (assuming that it’s nonsingular). Then for largebut finite t , the following properties hold.
1. The interior of the thick part carries a complete finite-volumehyperbolic metric. (This uses Ricci flow.)
2. The thin part is a “graph manifold”. (This doesn’t use Ricciflow. Stated by Perelman, proofs by Shioya-Yamaguchi,Morgan-Tian, Bessieres-Besson-Boileau-Maillot-Porti andKleiner-L.)
3. The interface between the thick and thin parts consists of“incompressible” 2-tori (Hamilton).
Graph manifolds were known to have a geometricdecomposition. Along with knowledge of the topological effectsof surgeries, this proved the geometrization conjecture.
![Page 74: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/74.jpg)
What Perelman actually showed
For any t , one can define a “thick-thin” decomposition of thetime-t manifold (assuming that it’s nonsingular). Then for largebut finite t , the following properties hold.
1. The interior of the thick part carries a complete finite-volumehyperbolic metric. (This uses Ricci flow.)
2. The thin part is a “graph manifold”. (This doesn’t use Ricciflow. Stated by Perelman, proofs by Shioya-Yamaguchi,Morgan-Tian, Bessieres-Besson-Boileau-Maillot-Porti andKleiner-L.)
3. The interface between the thick and thin parts consists of“incompressible” 2-tori (Hamilton).
Graph manifolds were known to have a geometricdecomposition. Along with knowledge of the topological effectsof surgeries, this proved the geometrization conjecture.
![Page 75: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/75.jpg)
What Perelman actually showed
For any t , one can define a “thick-thin” decomposition of thetime-t manifold (assuming that it’s nonsingular). Then for largebut finite t , the following properties hold.
1. The interior of the thick part carries a complete finite-volumehyperbolic metric. (This uses Ricci flow.)
2. The thin part is a “graph manifold”. (This doesn’t use Ricciflow. Stated by Perelman, proofs by Shioya-Yamaguchi,Morgan-Tian, Bessieres-Besson-Boileau-Maillot-Porti andKleiner-L.)
3. The interface between the thick and thin parts consists of“incompressible” 2-tori (Hamilton).
Graph manifolds were known to have a geometricdecomposition. Along with knowledge of the topological effectsof surgeries, this proved the geometrization conjecture.
![Page 76: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/76.jpg)
What Perelman actually showed
For any t , one can define a “thick-thin” decomposition of thetime-t manifold (assuming that it’s nonsingular). Then for largebut finite t , the following properties hold.
1. The interior of the thick part carries a complete finite-volumehyperbolic metric. (This uses Ricci flow.)
2. The thin part is a “graph manifold”. (This doesn’t use Ricciflow. Stated by Perelman, proofs by Shioya-Yamaguchi,Morgan-Tian, Bessieres-Besson-Boileau-Maillot-Porti andKleiner-L.)
3. The interface between the thick and thin parts consists of“incompressible” 2-tori (Hamilton).
Graph manifolds were known to have a geometricdecomposition. Along with knowledge of the topological effectsof surgeries, this proved the geometrization conjecture.
![Page 77: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/77.jpg)
What Perelman actually showed
For any t , one can define a “thick-thin” decomposition of thetime-t manifold (assuming that it’s nonsingular). Then for largebut finite t , the following properties hold.
1. The interior of the thick part carries a complete finite-volumehyperbolic metric. (This uses Ricci flow.)
2. The thin part is a “graph manifold”. (This doesn’t use Ricciflow. Stated by Perelman, proofs by Shioya-Yamaguchi,Morgan-Tian, Bessieres-Besson-Boileau-Maillot-Porti andKleiner-L.)
3. The interface between the thick and thin parts consists of“incompressible” 2-tori (Hamilton).
Graph manifolds were known to have a geometricdecomposition. Along with knowledge of the topological effectsof surgeries, this proved the geometrization conjecture.
![Page 78: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/78.jpg)
Is the intuitive picture correct?
Step 2, on the finiteness of the number of surgeries, was stillopen.
Step 3 : Show that after the singularities are over, as timeevolves, the locally homogeneous pieces in the Thurstondecomposition asymptotically appear.
Perelman showed that this is true for the “thick” part. Heshowed that its geometry is asymptotically hyperbolic. Whathappens on the “thin” part was still open.
Remark : Answering these questions has no topologicalimplication. We already know that the geometrizationconjecture holds. Rather, they are analytic questions about theRicci flow.
![Page 79: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/79.jpg)
Is the intuitive picture correct?
Step 2, on the finiteness of the number of surgeries, was stillopen.
Step 3 : Show that after the singularities are over, as timeevolves, the locally homogeneous pieces in the Thurstondecomposition asymptotically appear.
Perelman showed that this is true for the “thick” part. Heshowed that its geometry is asymptotically hyperbolic. Whathappens on the “thin” part was still open.
Remark : Answering these questions has no topologicalimplication. We already know that the geometrizationconjecture holds. Rather, they are analytic questions about theRicci flow.
![Page 80: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/80.jpg)
Is the intuitive picture correct?
Step 2, on the finiteness of the number of surgeries, was stillopen.
Step 3 : Show that after the singularities are over, as timeevolves, the locally homogeneous pieces in the Thurstondecomposition asymptotically appear.
Perelman showed that this is true for the “thick” part. Heshowed that its geometry is asymptotically hyperbolic. Whathappens on the “thin” part was still open.
Remark : Answering these questions has no topologicalimplication. We already know that the geometrizationconjecture holds. Rather, they are analytic questions about theRicci flow.
![Page 81: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/81.jpg)
Is the intuitive picture correct?
Step 2, on the finiteness of the number of surgeries, was stillopen.
Step 3 : Show that after the singularities are over, as timeevolves, the locally homogeneous pieces in the Thurstondecomposition asymptotically appear.
Perelman showed that this is true for the “thick” part. Heshowed that its geometry is asymptotically hyperbolic. Whathappens on the “thin” part was still open.
Remark : Answering these questions has no topologicalimplication. We already know that the geometrizationconjecture holds. Rather, they are analytic questions about theRicci flow.
![Page 82: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/82.jpg)
3D Ricci flow since Perelman
Homogeneous spaces and the geometrization conjecture
Geometrization conjecture and Ricci flow
Finiteness of the number of surgeries
Long-time behavior
Flowing through singularities
![Page 83: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/83.jpg)
Finiteness of the number of surgeries
Theorem(Bamler 2013) Starting from any compact Riemannian3-manifold, Perelman’s Ricci-flow-with-surgery only encountersa finite number of surgeries.
Furthermore, for large time t, if what’s left is nonempty then thesectional curvatures decay like O(t−1).
To be more precise, there is a parameter in Perelman’sRicci-flow-with-surgery that determines the scale at whichsurgery is performed.
The statement is that if this parameter is small enough (whichcan always be achieved) then there is a finite number ofsurgeries.
![Page 84: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/84.jpg)
Finiteness of the number of surgeries
Theorem(Bamler 2013) Starting from any compact Riemannian3-manifold, Perelman’s Ricci-flow-with-surgery only encountersa finite number of surgeries.
Furthermore, for large time t, if what’s left is nonempty then thesectional curvatures decay like O(t−1).
To be more precise, there is a parameter in Perelman’sRicci-flow-with-surgery that determines the scale at whichsurgery is performed.
The statement is that if this parameter is small enough (whichcan always be achieved) then there is a finite number ofsurgeries.
![Page 85: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/85.jpg)
Finiteness of the number of surgeries
Theorem(Bamler 2013) Starting from any compact Riemannian3-manifold, Perelman’s Ricci-flow-with-surgery only encountersa finite number of surgeries.
Furthermore, for large time t, if what’s left is nonempty then thesectional curvatures decay like O(t−1).
To be more precise, there is a parameter in Perelman’sRicci-flow-with-surgery that determines the scale at whichsurgery is performed.
The statement is that if this parameter is small enough (whichcan always be achieved) then there is a finite number ofsurgeries.
![Page 86: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/86.jpg)
Finiteness of the number of surgeries
Theorem(Bamler 2013) Starting from any compact Riemannian3-manifold, Perelman’s Ricci-flow-with-surgery only encountersa finite number of surgeries.
Furthermore, for large time t, if what’s left is nonempty then thesectional curvatures decay like O(t−1).
To be more precise, there is a parameter in Perelman’sRicci-flow-with-surgery that determines the scale at whichsurgery is performed.
The statement is that if this parameter is small enough (whichcan always be achieved) then there is a finite number ofsurgeries.
![Page 87: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/87.jpg)
Smooth flows
CorollaryA compact 3-manifold M admits a smooth Ricci flow that existsfor all positive time if and only if π2(M) = π3(M) = 0, i.e. if theuniversal cover of M is diffeomorphic to R3.
![Page 88: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/88.jpg)
Bounded curvature
Theorem(Bamler 2013) Starting from any compact Riemannian3-manifold, Perelman’s Ricci-flow-with-surgery only encountersa finite number of surgeries. Furthermore, for large time t, ifwhat’s left is nonempty then the sectional curvatures decay likeO(t−1).
Relevance of the second statement :
In Ricci flow, the Riemannian metric has engineering dimensionlength2 and time has engineering dimension length2.
So the scale-invariant time-t metric is g(t) = g(t)t .
The statement is that for large time, the rescaled metrics {g(t)}have uniformly bounded sectional curvatures.
This is good because we know lots about metrics with boundedsectional curvature (Cheeger-Fukaya-Gromov).
![Page 89: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/89.jpg)
Bounded curvature
Theorem(Bamler 2013) Starting from any compact Riemannian3-manifold, Perelman’s Ricci-flow-with-surgery only encountersa finite number of surgeries. Furthermore, for large time t, ifwhat’s left is nonempty then the sectional curvatures decay likeO(t−1).Relevance of the second statement :
In Ricci flow, the Riemannian metric has engineering dimensionlength2 and time has engineering dimension length2.
So the scale-invariant time-t metric is g(t) = g(t)t .
The statement is that for large time, the rescaled metrics {g(t)}have uniformly bounded sectional curvatures.
This is good because we know lots about metrics with boundedsectional curvature (Cheeger-Fukaya-Gromov).
![Page 90: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/90.jpg)
Bounded curvature
Theorem(Bamler 2013) Starting from any compact Riemannian3-manifold, Perelman’s Ricci-flow-with-surgery only encountersa finite number of surgeries. Furthermore, for large time t, ifwhat’s left is nonempty then the sectional curvatures decay likeO(t−1).Relevance of the second statement :
In Ricci flow, the Riemannian metric has engineering dimensionlength2 and time has engineering dimension length2.
So the scale-invariant time-t metric is g(t) = g(t)t .
The statement is that for large time, the rescaled metrics {g(t)}have uniformly bounded sectional curvatures.
This is good because we know lots about metrics with boundedsectional curvature (Cheeger-Fukaya-Gromov).
![Page 91: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/91.jpg)
Bounded curvature
Theorem(Bamler 2013) Starting from any compact Riemannian3-manifold, Perelman’s Ricci-flow-with-surgery only encountersa finite number of surgeries. Furthermore, for large time t, ifwhat’s left is nonempty then the sectional curvatures decay likeO(t−1).Relevance of the second statement :
In Ricci flow, the Riemannian metric has engineering dimensionlength2 and time has engineering dimension length2.
So the scale-invariant time-t metric is g(t) = g(t)t .
The statement is that for large time, the rescaled metrics {g(t)}have uniformly bounded sectional curvatures.
This is good because we know lots about metrics with boundedsectional curvature (Cheeger-Fukaya-Gromov).
![Page 92: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/92.jpg)
Bounded curvature
Theorem(Bamler 2013) Starting from any compact Riemannian3-manifold, Perelman’s Ricci-flow-with-surgery only encountersa finite number of surgeries. Furthermore, for large time t, ifwhat’s left is nonempty then the sectional curvatures decay likeO(t−1).Relevance of the second statement :
In Ricci flow, the Riemannian metric has engineering dimensionlength2 and time has engineering dimension length2.
So the scale-invariant time-t metric is g(t) = g(t)t .
The statement is that for large time, the rescaled metrics {g(t)}have uniformly bounded sectional curvatures.
This is good because we know lots about metrics with boundedsectional curvature (Cheeger-Fukaya-Gromov).
![Page 93: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/93.jpg)
Bounded curvature
Theorem(Bamler 2013) Starting from any compact Riemannian3-manifold, Perelman’s Ricci-flow-with-surgery only encountersa finite number of surgeries. Furthermore, for large time t, ifwhat’s left is nonempty then the sectional curvatures decay likeO(t−1).Relevance of the second statement :
In Ricci flow, the Riemannian metric has engineering dimensionlength2 and time has engineering dimension length2.
So the scale-invariant time-t metric is g(t) = g(t)t .
The statement is that for large time, the rescaled metrics {g(t)}have uniformly bounded sectional curvatures.
This is good because we know lots about metrics with boundedsectional curvature (Cheeger-Fukaya-Gromov).
![Page 94: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/94.jpg)
Ingredients of the proof
Bamler’s proof uses all of Perelman’s work, and more. Some ofthe new ingredients :
1. Localizing Perelman’s estimates and applying them to localcovers of the manifold.
2. Use of minimal surfaces to control the geometry of the thinpart.
3. Use of minimal embedded 2-complexes.
![Page 95: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/95.jpg)
Ingredients of the proof
Bamler’s proof uses all of Perelman’s work, and more. Some ofthe new ingredients :
1. Localizing Perelman’s estimates and applying them to localcovers of the manifold.
2. Use of minimal surfaces to control the geometry of the thinpart.
3. Use of minimal embedded 2-complexes.
![Page 96: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/96.jpg)
Ingredients of the proof
Bamler’s proof uses all of Perelman’s work, and more. Some ofthe new ingredients :
1. Localizing Perelman’s estimates and applying them to localcovers of the manifold.
2. Use of minimal surfaces to control the geometry of the thinpart.
3. Use of minimal embedded 2-complexes.
![Page 97: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/97.jpg)
Ingredients of the proof
Bamler’s proof uses all of Perelman’s work, and more. Some ofthe new ingredients :
1. Localizing Perelman’s estimates and applying them to localcovers of the manifold.
2. Use of minimal surfaces to control the geometry of the thinpart.
3. Use of minimal embedded 2-complexes.
![Page 98: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/98.jpg)
3D Ricci flow since Perelman
Homogeneous spaces and the geometrization conjecture
Geometrization conjecture and Ricci flow
Finiteness of the number of surgeries
Long-time behavior
Flowing through singularities
![Page 99: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/99.jpg)
Long-time behavior
From Bamler’s result, to understand the long-time behavior ofthe Ricci flow, it is enough to restrict to smooth Ricci flows.
The only case that we completely understand is when M admitssome hyperbolic metric. Then from Perelman’s work, for anyinitial metric on M, as t →∞ the rescaled Riemannian metricg(t) approaches the metric on M of constant sectionalcurvature − 1
4 .
Question : if M doesn’t admit a hyperbolic metric, what are thecandidate geometries for the long-time behavior?
![Page 100: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/100.jpg)
Long-time behavior
From Bamler’s result, to understand the long-time behavior ofthe Ricci flow, it is enough to restrict to smooth Ricci flows.
The only case that we completely understand is when M admitssome hyperbolic metric. Then from Perelman’s work, for anyinitial metric on M, as t →∞ the rescaled Riemannian metricg(t) approaches the metric on M of constant sectionalcurvature − 1
4 .
Question : if M doesn’t admit a hyperbolic metric, what are thecandidate geometries for the long-time behavior?
![Page 101: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/101.jpg)
Long-time behavior
From Bamler’s result, to understand the long-time behavior ofthe Ricci flow, it is enough to restrict to smooth Ricci flows.
The only case that we completely understand is when M admitssome hyperbolic metric. Then from Perelman’s work, for anyinitial metric on M, as t →∞ the rescaled Riemannian metricg(t) approaches the metric on M of constant sectionalcurvature − 1
4 .
Question : if M doesn’t admit a hyperbolic metric, what are thecandidate geometries for the long-time behavior?
![Page 102: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/102.jpg)
Quasistatic solutions
The static solutions of the Ricci flow equation
dgdt
= − 2 Ricg
are Ricci-flat.
The solutions that are static up to rescaling are Einsteinmetrics: Ric = const.g.
The solutions that are static up to rescaling anddiffeomorphisms are Ricci solitons : Ric = const.g + LV g.
Fact : On a compact 3-manifold, any such quasistatic solutionhas constant sectional curvature.
Apparent paradox : What happens to the Ricci flow if our3-manifold doesn’t admit a constant curvature metric?
![Page 103: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/103.jpg)
Quasistatic solutions
The static solutions of the Ricci flow equation
dgdt
= − 2 Ricg
are Ricci-flat.
The solutions that are static up to rescaling are Einsteinmetrics: Ric = const.g.
The solutions that are static up to rescaling anddiffeomorphisms are Ricci solitons : Ric = const.g + LV g.
Fact : On a compact 3-manifold, any such quasistatic solutionhas constant sectional curvature.
Apparent paradox : What happens to the Ricci flow if our3-manifold doesn’t admit a constant curvature metric?
![Page 104: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/104.jpg)
Quasistatic solutions
The static solutions of the Ricci flow equation
dgdt
= − 2 Ricg
are Ricci-flat.
The solutions that are static up to rescaling are Einsteinmetrics: Ric = const.g.
The solutions that are static up to rescaling anddiffeomorphisms are Ricci solitons : Ric = const.g + LV g.
Fact : On a compact 3-manifold, any such quasistatic solutionhas constant sectional curvature.
Apparent paradox : What happens to the Ricci flow if our3-manifold doesn’t admit a constant curvature metric?
![Page 105: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/105.jpg)
Quasistatic solutions
The static solutions of the Ricci flow equation
dgdt
= − 2 Ricg
are Ricci-flat.
The solutions that are static up to rescaling are Einsteinmetrics: Ric = const.g.
The solutions that are static up to rescaling anddiffeomorphisms are Ricci solitons : Ric = const.g + LV g.
Fact : On a compact 3-manifold, any such quasistatic solutionhas constant sectional curvature.
Apparent paradox : What happens to the Ricci flow if our3-manifold doesn’t admit a constant curvature metric?
![Page 106: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/106.jpg)
Quasistatic solutions
The static solutions of the Ricci flow equation
dgdt
= − 2 Ricg
are Ricci-flat.
The solutions that are static up to rescaling are Einsteinmetrics: Ric = const.g.
The solutions that are static up to rescaling anddiffeomorphisms are Ricci solitons : Ric = const.g + LV g.
Fact : On a compact 3-manifold, any such quasistatic solutionhas constant sectional curvature.
Apparent paradox : What happens to the Ricci flow if our3-manifold doesn’t admit a constant curvature metric?
![Page 107: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/107.jpg)
Nil geometry
Put NilZ =
1 a b
0 1 c0 0 1
: a,b, c ∈ Z
. Define NilR similarly.
Put M = NilR /NilZ. It is the total space of a nontrivial circlebundle over T 2.
Run the Ricci flow. The base torus expands like O(
t16
). The
circle fibers shrink like O(
t −16
).
With the rescaled metric g(t) = g(t)t , (M, g(t)) shrinks to a point.
![Page 108: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/108.jpg)
Nil geometry
Put NilZ =
1 a b
0 1 c0 0 1
: a,b, c ∈ Z
. Define NilR similarly.
Put M = NilR /NilZ. It is the total space of a nontrivial circlebundle over T 2.
Run the Ricci flow. The base torus expands like O(
t16
). The
circle fibers shrink like O(
t −16
).
With the rescaled metric g(t) = g(t)t , (M, g(t)) shrinks to a point.
![Page 109: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/109.jpg)
Nil geometry
Put NilZ =
1 a b
0 1 c0 0 1
: a,b, c ∈ Z
. Define NilR similarly.
Put M = NilR /NilZ. It is the total space of a nontrivial circlebundle over T 2.
Run the Ricci flow. The base torus expands like O(
t16
). The
circle fibers shrink like O(
t −16
).
With the rescaled metric g(t) = g(t)t , (M, g(t)) shrinks to a point.
![Page 110: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/110.jpg)
Nil geometry
Put NilZ =
1 a b
0 1 c0 0 1
: a,b, c ∈ Z
. Define NilR similarly.
Put M = NilR /NilZ. It is the total space of a nontrivial circlebundle over T 2.
Run the Ricci flow. The base torus expands like O(
t16
). The
circle fibers shrink like O(
t −16
).
With the rescaled metric g(t) = g(t)t , (M, g(t)) shrinks to a point.
![Page 111: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/111.jpg)
Nil geometry
Put NilZ =
1 a b
0 1 c0 0 1
: a,b, c ∈ Z
. Define NilR similarly.
Put M = NilR /NilZ. It is the total space of a nontrivial circlebundle over T 2.
Run the Ricci flow. The base torus expands like O(
t16
). The
circle fibers shrink like O(
t −16
).
With the rescaled metric g(t) = g(t)t , (M, g(t)) shrinks to a point.
![Page 112: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/112.jpg)
Nil geometry
Put NilZ =
1 a b
0 1 c0 0 1
: a,b, c ∈ Z
. Define NilR similarly.
Put M = NilR /NilZ. It is the total space of a nontrivial circlebundle over T 2.
Run the Ricci flow. The base torus expands like O(
t16
). The
circle fibers shrink like O(
t −16
).
With the rescaled metric g(t) = g(t)t , (M, g(t)) shrinks to a point.
![Page 113: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/113.jpg)
Sol geometry
M fibers over a circle with 2-torus fibers. The monodromy is ahyperbolic element of SL(2,Z).
Run the Ricci flow. The base circle expands like O(
t12
). The
fiber sizes are O(t0).
With the rescaled metric, (M, g(t)) approaches a circle.
![Page 114: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/114.jpg)
Sol geometry
M fibers over a circle with 2-torus fibers. The monodromy is ahyperbolic element of SL(2,Z).
Run the Ricci flow. The base circle expands like O(
t12
). The
fiber sizes are O(t0).
With the rescaled metric, (M, g(t)) approaches a circle.
![Page 115: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/115.jpg)
Sol geometry
M fibers over a circle with 2-torus fibers. The monodromy is ahyperbolic element of SL(2,Z).
Run the Ricci flow. The base circle expands like O(
t12
). The
fiber sizes are O(t0).
With the rescaled metric, (M, g(t)) approaches a circle.
![Page 116: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/116.jpg)
˜SL(2,R) geometry
Suppose that M is the unit tangent bundle of a hyperbolicsurface Σ.
Run the Ricci flow. The base surface expands like O(
t12
). The
fiber sizes are O(t0).
With the rescaled metric, (M, g(t)) approaches the hyperbolicsurface Σ. As the fibers shrink, the local geometry of the totalspace becomes more product-like.
![Page 117: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/117.jpg)
˜SL(2,R) geometry
Suppose that M is the unit tangent bundle of a hyperbolicsurface Σ.
Run the Ricci flow. The base surface expands like O(
t12
). The
fiber sizes are O(t0).
With the rescaled metric, (M, g(t)) approaches the hyperbolicsurface Σ. As the fibers shrink, the local geometry of the totalspace becomes more product-like.
![Page 118: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/118.jpg)
˜SL(2,R) geometry
Suppose that M is the unit tangent bundle of a hyperbolicsurface Σ.
Run the Ricci flow. The base surface expands like O(
t12
). The
fiber sizes are O(t0).
With the rescaled metric, (M, g(t)) approaches the hyperbolicsurface Σ. As the fibers shrink, the local geometry of the totalspace becomes more product-like.
![Page 119: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/119.jpg)
Is there a common pattern?
To undo the collapsing, let’s pass to the universal cover. That is,we are looking at the Ricci flow on a Thurston geometry of typeR3,H3,H2 × R,Sol,Nil or SL2(R).
Proposition(L. 2007) For any initial globally homogeneous metric on such aThurston geometry, there is a limiting (blowdown) Ricci flowsolution, which is an expanding soliton. There is one suchsoliton for each homogeneity type. It is a universal attractor.
Ric +12LV g = − 1
2tg.
A subtlety : the limit is in the pointed sense. The soliton metricg is homogeneous but the vector field V need not behomogeneous. Also, the homogeneity type may change in thelimit.
![Page 120: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/120.jpg)
Is there a common pattern?
To undo the collapsing, let’s pass to the universal cover. That is,we are looking at the Ricci flow on a Thurston geometry of typeR3,H3,H2 × R,Sol,Nil or SL2(R).
Proposition(L. 2007) For any initial globally homogeneous metric on such aThurston geometry, there is a limiting (blowdown) Ricci flowsolution, which is an expanding soliton. There is one suchsoliton for each homogeneity type. It is a universal attractor.
Ric +12LV g = − 1
2tg.
A subtlety : the limit is in the pointed sense. The soliton metricg is homogeneous but the vector field V need not behomogeneous. Also, the homogeneity type may change in thelimit.
![Page 121: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/121.jpg)
Is there a common pattern?
To undo the collapsing, let’s pass to the universal cover. That is,we are looking at the Ricci flow on a Thurston geometry of typeR3,H3,H2 × R,Sol,Nil or SL2(R).
Proposition(L. 2007) For any initial globally homogeneous metric on such aThurston geometry, there is a limiting (blowdown) Ricci flowsolution, which is an expanding soliton. There is one suchsoliton for each homogeneity type. It is a universal attractor.
Ric +12LV g = − 1
2tg.
A subtlety : the limit is in the pointed sense. The soliton metricg is homogeneous but the vector field V need not behomogeneous. Also, the homogeneity type may change in thelimit.
![Page 122: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/122.jpg)
Is there a common pattern?
To undo the collapsing, let’s pass to the universal cover. That is,we are looking at the Ricci flow on a Thurston geometry of typeR3,H3,H2 × R,Sol,Nil or SL2(R).
Proposition(L. 2007) For any initial globally homogeneous metric on such aThurston geometry, there is a limiting (blowdown) Ricci flowsolution, which is an expanding soliton. There is one suchsoliton for each homogeneity type. It is a universal attractor.
Ric +12LV g = − 1
2tg.
A subtlety : the limit is in the pointed sense. The soliton metricg is homogeneous but the vector field V need not behomogeneous.
Also, the homogeneity type may change in thelimit.
![Page 123: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/123.jpg)
Is there a common pattern?
To undo the collapsing, let’s pass to the universal cover. That is,we are looking at the Ricci flow on a Thurston geometry of typeR3,H3,H2 × R,Sol,Nil or SL2(R).
Proposition(L. 2007) For any initial globally homogeneous metric on such aThurston geometry, there is a limiting (blowdown) Ricci flowsolution, which is an expanding soliton. There is one suchsoliton for each homogeneity type. It is a universal attractor.
Ric +12LV g = − 1
2tg.
A subtlety : the limit is in the pointed sense. The soliton metricg is homogeneous but the vector field V need not behomogeneous. Also, the homogeneity type may change in thelimit.
![Page 124: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/124.jpg)
The limiting solitons
Thurston type Expanding soliton
H3 4 t gH3
H2 × R or ˜SL(2,R) 2 t gH2 + gR
Sol e−2z dx2 + e2z dy2 + 4 t dz2
Nil1
3t13
(dx +
12
ydz − 12
zdy)2
+ t13
(dy2 + dz2
)R3 gR3
![Page 125: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/125.jpg)
A general convergence theorem
Theorem(L. 2010) Suppose that (M,g(t)) is a Ricci flow on a compactthree-dimensional manifold, that exists for t ∈ [0,∞). Supposethat the sectional curvatures are O
(t−1) in magnitude, and the
diameter is O(√
t). Then the pullback of the Ricci flow to Mapproaches one of the homogeneous expanding solitons.Remarks :
I By Bamler’s result, the sectional curvatures are alwaysO(t−1).
I The hypotheses imply that M admits a locallyhomogeneous metric.
ConjectureFor a long-time 3D Ricci flow, the diameter is O(
√t) if and only
if M admits a locally homogeneous metric.
![Page 126: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/126.jpg)
A general convergence theorem
Theorem(L. 2010) Suppose that (M,g(t)) is a Ricci flow on a compactthree-dimensional manifold, that exists for t ∈ [0,∞). Supposethat the sectional curvatures are O
(t−1) in magnitude, and the
diameter is O(√
t). Then the pullback of the Ricci flow to Mapproaches one of the homogeneous expanding solitons.
Remarks :I By Bamler’s result, the sectional curvatures are always
O(t−1).
I The hypotheses imply that M admits a locallyhomogeneous metric.
ConjectureFor a long-time 3D Ricci flow, the diameter is O(
√t) if and only
if M admits a locally homogeneous metric.
![Page 127: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/127.jpg)
A general convergence theorem
Theorem(L. 2010) Suppose that (M,g(t)) is a Ricci flow on a compactthree-dimensional manifold, that exists for t ∈ [0,∞). Supposethat the sectional curvatures are O
(t−1) in magnitude, and the
diameter is O(√
t). Then the pullback of the Ricci flow to Mapproaches one of the homogeneous expanding solitons.Remarks :
I By Bamler’s result, the sectional curvatures are alwaysO(t−1).
I The hypotheses imply that M admits a locallyhomogeneous metric.
ConjectureFor a long-time 3D Ricci flow, the diameter is O(
√t) if and only
if M admits a locally homogeneous metric.
![Page 128: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/128.jpg)
A general convergence theorem
Theorem(L. 2010) Suppose that (M,g(t)) is a Ricci flow on a compactthree-dimensional manifold, that exists for t ∈ [0,∞). Supposethat the sectional curvatures are O
(t−1) in magnitude, and the
diameter is O(√
t). Then the pullback of the Ricci flow to Mapproaches one of the homogeneous expanding solitons.Remarks :
I By Bamler’s result, the sectional curvatures are alwaysO(t−1).
I The hypotheses imply that M admits a locallyhomogeneous metric.
ConjectureFor a long-time 3D Ricci flow, the diameter is O(
√t) if and only
if M admits a locally homogeneous metric.
![Page 129: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/129.jpg)
A more refined result
What happens to the Ricci flow on a 3-torus?
Theorem(L.-Sesum 2014) Let g0 be a warped product metric on T 3, withrespect to the circle fibering T 3 → T 2 and any Riemannianmetric on T 2.
Then under the Ricci flow, g(t) approaches a flat metric g∞ onT 3 exponentially fast.
Question : is this true for all initial metrics on T 3?
![Page 130: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/130.jpg)
A more refined result
What happens to the Ricci flow on a 3-torus?
Theorem(L.-Sesum 2014) Let g0 be a warped product metric on T 3, withrespect to the circle fibering T 3 → T 2 and any Riemannianmetric on T 2.
Then under the Ricci flow, g(t) approaches a flat metric g∞ onT 3 exponentially fast.
Question : is this true for all initial metrics on T 3?
![Page 131: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/131.jpg)
A more refined result
What happens to the Ricci flow on a 3-torus?
Theorem(L.-Sesum 2014) Let g0 be a warped product metric on T 3, withrespect to the circle fibering T 3 → T 2 and any Riemannianmetric on T 2.
Then under the Ricci flow, g(t) approaches a flat metric g∞ onT 3 exponentially fast.
Question : is this true for all initial metrics on T 3?
![Page 132: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/132.jpg)
A more refined result
What happens to the Ricci flow on a 3-torus?
Theorem(L.-Sesum 2014) Let g0 be a warped product metric on T 3, withrespect to the circle fibering T 3 → T 2 and any Riemannianmetric on T 2.
Then under the Ricci flow, g(t) approaches a flat metric g∞ onT 3 exponentially fast.
Question : is this true for all initial metrics on T 3?
![Page 133: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/133.jpg)
3D Ricci flow since Perelman
Homogeneous spaces and the geometrization conjecture
Geometrization conjecture and Ricci flow
Finiteness of the number of surgeries
Long-time behavior
Flowing through singularities
![Page 134: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/134.jpg)
Can one flow through a singularity?
Is there a natural way to extend the flow beyond the singularity?
![Page 135: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/135.jpg)
Deus Ex Machina
Surgery
![Page 136: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/136.jpg)
Deus Ex Machina
Surgery
![Page 137: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/137.jpg)
Deus Ex Machina
Surgery
![Page 138: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/138.jpg)
Surgery scale
In the definition of Ricci flow with surgery, there is a functionh(t) so that the time-t surgeries are done on 2-spheres whoseradius is approximately h(t).
From Perelman’s first Ricci flow paper : It is likely that bypassing to the limit in this construction one would get acanonically defined Ricci flow through singularities, but at themoment I don’t have a proof of that.
![Page 139: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/139.jpg)
Surgery scale
In the definition of Ricci flow with surgery, there is a functionh(t) so that the time-t surgeries are done on 2-spheres whoseradius is approximately h(t).
From Perelman’s first Ricci flow paper : It is likely that bypassing to the limit in this construction one would get acanonically defined Ricci flow through singularities, but at themoment I don’t have a proof of that.
![Page 140: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/140.jpg)
A toy model
dxdt
= x2, x(0) = 1.
Try to solve it on a computer. Get x(t) = 11−t , as long as t < 1.
How can we extend this beyond the singularity at t = 1?
![Page 141: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/141.jpg)
A toy model
dxdt
= x2, x(0) = 1.
Try to solve it on a computer. Get x(t) = 11−t , as long as t < 1.
How can we extend this beyond the singularity at t = 1?
![Page 142: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/142.jpg)
A toy model
dxdt
= x2, x(0) = 1.
Try to solve it on a computer. Get x(t) = 11−t , as long as t < 1.
How can we extend this beyond the singularity at t = 1?
![Page 143: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/143.jpg)
Regularization
Regularize the equation. For ε 6= 0, consider
dxdt
= (x + iε)2, x(0) = 1.
Its solution
xε(t) =1 + iεt − ε2t1− t − iεt
is defined for all t .
If t 6= 1 then
limε→0
xε(t) =1
1− t.
This gives a way to extend the solution beyond (or around)t = 1.
![Page 144: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/144.jpg)
Regularization
Regularize the equation. For ε 6= 0, consider
dxdt
= (x + iε)2, x(0) = 1.
Its solution
xε(t) =1 + iεt − ε2t1− t − iεt
is defined for all t .
If t 6= 1 then
limε→0
xε(t) =1
1− t.
This gives a way to extend the solution beyond (or around)t = 1.
![Page 145: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/145.jpg)
Regularization
Regularize the equation. For ε 6= 0, consider
dxdt
= (x + iε)2, x(0) = 1.
Its solution
xε(t) =1 + iεt − ε2t1− t − iεt
is defined for all t .
If t 6= 1 then
limε→0
xε(t) =1
1− t.
This gives a way to extend the solution beyond (or around)t = 1.
![Page 146: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/146.jpg)
Regularization
Regularize the equation. For ε 6= 0, consider
dxdt
= (x + iε)2, x(0) = 1.
Its solution
xε(t) =1 + iεt − ε2t1− t − iεt
is defined for all t .
If t 6= 1 then
limε→0
xε(t) =1
1− t.
This gives a way to extend the solution beyond (or around)t = 1.
![Page 147: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/147.jpg)
Analogy
dxdt
= x2 −→ dgdt
= −2 Ricg
dxdt
= (x + iε)2 −→ Ricci-flow-with-surgery algorithm
ε −→ Surgery parameter h(t)
xε(t) −→ Ricci-flow-with-surgery solution
x(t) =1
1− t−→ A Ricci flow through singularities (?)
![Page 148: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/148.jpg)
Sublimits of Ricci flows with surgery
Theorem(Kleiner-L. 2014) Let hi : [0,∞)→ R be a sequence ofdecreasing continuous functions that tend uniformly to zero.
Given an initial compact Riemannian 3-manifold, let gi(t) be theensuing Ricci flow with surgery, as constructing using thesurgery scale hi(t).
Then after passing to a subsequence, there is a limit
limi→∞
gi(t) = g∞(t). (3)
![Page 149: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/149.jpg)
Sublimits of Ricci flows with surgery
Theorem(Kleiner-L. 2014) Let hi : [0,∞)→ R be a sequence ofdecreasing continuous functions that tend uniformly to zero.
Given an initial compact Riemannian 3-manifold, let gi(t) be theensuing Ricci flow with surgery, as constructing using thesurgery scale hi(t).
Then after passing to a subsequence, there is a limit
limi→∞
gi(t) = g∞(t). (3)
![Page 150: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/150.jpg)
Sublimits of Ricci flows with surgery
Theorem(Kleiner-L. 2014) Let hi : [0,∞)→ R be a sequence ofdecreasing continuous functions that tend uniformly to zero.
Given an initial compact Riemannian 3-manifold, let gi(t) be theensuing Ricci flow with surgery, as constructing using thesurgery scale hi(t).
Then after passing to a subsequence, there is a limit
limi→∞
gi(t) = g∞(t). (3)
![Page 151: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/151.jpg)
Spacetime of a flow through a neckpinch singularity
![Page 152: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/152.jpg)
Idea of the proof
There’s a theorem of Hamilton that lets you take convergentsubsequences of Ricci flow solutions, but unfortunately itdoesn’t apply in our case. Two ideas :
1. Look at the spacetime of a Ricci-flow-with-surgery. It is a4-manifold with boundary. Restrict to its interior X . The latter isequipped with a time function t : X → [0,∞) and a Riemannianmetric dt2 + g(t). Take limits of such spacetimes.
2. In 3D Ricci flow, the scalar curvature R controls the localgeometry, i.e. the curvature tensor and the injectivity radius(Hamilton-Ivey, Perelman). Fix R <∞. Consider the sublevelset
X≤R = {x ∈ X : R(x) ≤ R}.
Take a sublimit of these regions as the surgery parameter goesto zero. Then take a sublimit as R →∞.
![Page 153: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/153.jpg)
Idea of the proof
There’s a theorem of Hamilton that lets you take convergentsubsequences of Ricci flow solutions, but unfortunately itdoesn’t apply in our case. Two ideas :
1. Look at the spacetime of a Ricci-flow-with-surgery. It is a4-manifold with boundary. Restrict to its interior X . The latter isequipped with a time function t : X → [0,∞) and a Riemannianmetric dt2 + g(t). Take limits of such spacetimes.
2. In 3D Ricci flow, the scalar curvature R controls the localgeometry, i.e. the curvature tensor and the injectivity radius(Hamilton-Ivey, Perelman). Fix R <∞. Consider the sublevelset
X≤R = {x ∈ X : R(x) ≤ R}.
Take a sublimit of these regions as the surgery parameter goesto zero. Then take a sublimit as R →∞.
![Page 154: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/154.jpg)
Idea of the proof
There’s a theorem of Hamilton that lets you take convergentsubsequences of Ricci flow solutions, but unfortunately itdoesn’t apply in our case. Two ideas :
1. Look at the spacetime of a Ricci-flow-with-surgery. It is a4-manifold with boundary. Restrict to its interior X . The latter isequipped with a time function t : X → [0,∞) and a Riemannianmetric dt2 + g(t). Take limits of such spacetimes.
2. In 3D Ricci flow, the scalar curvature R controls the localgeometry, i.e. the curvature tensor and the injectivity radius(Hamilton-Ivey, Perelman). Fix R <∞. Consider the sublevelset
X≤R = {x ∈ X : R(x) ≤ R}.
Take a sublimit of these regions as the surgery parameter goesto zero. Then take a sublimit as R →∞.
![Page 155: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/155.jpg)
Some properties of the limit
The limiting spacetimeM∞ is a smooth 4-manifold equippedwith a time function t :M∞ → R, a time vector field ∂t and aRiemannian metric which can locally be written as dt2 + g∞(t).
Some structural properties of the limit :
1. M∞ satisfies the Hamilton-Ivey pinching condition. It isκ-noncollapsed and satisfies the r -canonical neighborhoodassumption.
2. The scalar curvature function R : M∞ → R is boundedbelow, and is a proper function on any time slice.
3. The time slices may not be complete, but the quasiparabolicmetric (1 + R2) dt2 +
√1 + R2 g∞(t) is complete and has
bounded curvature.
One can study such singular Ricci flows in their own right.
![Page 156: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/156.jpg)
Some properties of the limit
The limiting spacetimeM∞ is a smooth 4-manifold equippedwith a time function t :M∞ → R, a time vector field ∂t and aRiemannian metric which can locally be written as dt2 + g∞(t).
Some structural properties of the limit :
1. M∞ satisfies the Hamilton-Ivey pinching condition. It isκ-noncollapsed and satisfies the r -canonical neighborhoodassumption.
2. The scalar curvature function R : M∞ → R is boundedbelow, and is a proper function on any time slice.
3. The time slices may not be complete, but the quasiparabolicmetric (1 + R2) dt2 +
√1 + R2 g∞(t) is complete and has
bounded curvature.
One can study such singular Ricci flows in their own right.
![Page 157: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/157.jpg)
Some properties of the limit
The limiting spacetimeM∞ is a smooth 4-manifold equippedwith a time function t :M∞ → R, a time vector field ∂t and aRiemannian metric which can locally be written as dt2 + g∞(t).
Some structural properties of the limit :
1. M∞ satisfies the Hamilton-Ivey pinching condition. It isκ-noncollapsed and satisfies the r -canonical neighborhoodassumption.
2. The scalar curvature function R : M∞ → R is boundedbelow, and is a proper function on any time slice.
3. The time slices may not be complete, but the quasiparabolicmetric (1 + R2) dt2 +
√1 + R2 g∞(t) is complete and has
bounded curvature.
One can study such singular Ricci flows in their own right.
![Page 158: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/158.jpg)
Some properties of the limit
The limiting spacetimeM∞ is a smooth 4-manifold equippedwith a time function t :M∞ → R, a time vector field ∂t and aRiemannian metric which can locally be written as dt2 + g∞(t).
Some structural properties of the limit :
1. M∞ satisfies the Hamilton-Ivey pinching condition. It isκ-noncollapsed and satisfies the r -canonical neighborhoodassumption.
2. The scalar curvature function R : M∞ → R is boundedbelow, and is a proper function on any time slice.
3. The time slices may not be complete, but the quasiparabolicmetric (1 + R2) dt2 +
√1 + R2 g∞(t) is complete and has
bounded curvature.
One can study such singular Ricci flows in their own right.
![Page 159: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/159.jpg)
Some properties of the limit
The limiting spacetimeM∞ is a smooth 4-manifold equippedwith a time function t :M∞ → R, a time vector field ∂t and aRiemannian metric which can locally be written as dt2 + g∞(t).
Some structural properties of the limit :
1. M∞ satisfies the Hamilton-Ivey pinching condition. It isκ-noncollapsed and satisfies the r -canonical neighborhoodassumption.
2. The scalar curvature function R : M∞ → R is boundedbelow, and is a proper function on any time slice.
3. The time slices may not be complete, but the quasiparabolicmetric (1 + R2) dt2 +
√1 + R2 g∞(t) is complete and has
bounded curvature.
One can study such singular Ricci flows in their own right.
![Page 160: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/160.jpg)
Summary
We still have a lot to learn.
Thank you.
![Page 161: 3D Ricci flow since Perelman - UCB Mathematicsmath.berkeley.edu/~lott/fullerton.pdf · 2015. 11. 2. · Poincare conjecture´ The geometrization conjecture implies the Poincare conjecture](https://reader034.fdocuments.in/reader034/viewer/2022051912/6002dbf83dd7751d4d13bc17/html5/thumbnails/161.jpg)
Summary
We still have a lot to learn.
Thank you.