Low Dimensional Topology

August 5, 2009

Immersed surfaces with injective fundamental group.

Filed under: 3-manifolds,Geometric Group Theory,Hyperbolic geometry — Jesse Johnson @ 2:08 pm

I just heard that Jeremy Kahn has announced joint work with Vladimir Markovic showing that every hyperbolic 3-manifold contains an immersed, \pi_1-injective surface.  This is equivalent to showing that the fundamental group of every hyperbolic 3-manifold contains a subgroup isomorphic to the fundamental group of a surface.  The abstract for Jeremy’s talk is on the conference web page.  This is a long standing open problem that Henry Wilton wrote about on this blog back in February.  I’m not at the conference, so I didn’t see the talk, but if any of you readers were ther and have anything to add, feel free to write in in the comments.

Update:  Danny Calegari has posted a discussion of some aspects of the proposed proof here and an outline written by Jeremy here.


  1. I would be very curious to know whether the proof also shows that every homologically trivial geodesic in a hyperbolic 3-manifold bounds an injective (non-closed) surface.

    Also, how is 3-dimensionality used? (or is it?) It’s not clear from the linked abstract.

    Comment by Danny Calegari — August 6, 2009 @ 11:24 pm | Reply

    • I wasn’t at Kahn’s talk, but my guess is their approach may be related to the one Lewis Bowen tried in [math.GT/0411662], especially as Kahn and Markovic have worked on the Ehrenpreis Conjecture which Bowen also attacks in that preprint. That is, one pieces together a large number of nearly-geodesic pairs of pants with very long cuff lengths. I think the tricky bit is that if the cuff lengths are large, the distances between cuffs is very small, which means the angles between adjacent pants have to be very very small it the resulting surface is going to be nearly geodesic.

      If this is the case, then I doubt that KM’s method would also answer your above question. I have no clue about the whether 3-dimensionality would be essential, though.

      Comment by Nathan Dunfield — August 7, 2009 @ 2:32 pm | Reply

      • Hi Nathan – I just heard from Ian that they glue pants with an almost-fixed “twist” of some definite length (so that the very short arcs joining pairs of long cusps do not match up) and so one doesn’t need the angle to be so very very small (just very small :).

        There is still the question of where the pants come from. At the risk of indulging in pure speculation, let me propose a method. One way to construct a pair of pants with edge cuffs of length close to R is to put a little geodesic letter H in your manifold (where the horizontal edge is perpendicular to the vertical one), then look for a pair of closed geodesics of complex length very close to R (i.e. with imaginary component of length very close to zero) and which are almost tangent to the two vertical sides of the letter H. For big R, such geodesics will be very close to wherever you look for them. The homotopy class of the H is a pair of pants with all three cuff lengths close to R (I think). One can choose in advance which the first cuff will be (providing R is big enough).

        The next obvious problem is pairing up the cuffs; if they have a way to do that, then I would guess it also lets one first prescribe some set of (sufficiently long) boundary curves. So maybe one can fill an arbitrary homologically trivial chain, providing the geodesics are sufficiently long . . .

        Anyway, this (completely speculative and uninformed) sketch does not use dimension 3. In fact, maybe it also works in dimension 2, and for finite volume but cusped manifolds, in which case I must have gotten something wrong . . .

        Comment by Danny Calegari — August 7, 2009 @ 5:47 pm

      • In fact, the crossbar of the “H” is the most delicate part of the pants (that’s where it’s thinnest) so this would not be a good place to try to pin down the geometry. I guess this is why they use tripods instead, centered at the most robust part of the pants (see Ian’s post below).

        Comment by Danny Calegari — August 7, 2009 @ 7:17 pm

  2. Do I infer correctly that, following this result, every LERF hyperbolic 3-manifold is virtually fibered?

    The implications would go as follows. Kahn and Markovic give you an essential immersed surface. LERF allows you to find a cover where the surface is embedded. So M is virtually Haken. Then, Ian Agol’s recent theorem tells you that that M is actually virtually fibered.

    Reducing virtual fibering to the LERF conjecture seems extremely strong…

    Comment by Dave Futer — August 7, 2009 @ 12:54 pm | Reply

    • @ Dave: Yes, you’re correct that [KM] + [A] + LERF = virtual fibering.

      Comment by Nathan Dunfield — August 7, 2009 @ 2:19 pm | Reply

  3. Here’s a bit of a report on their proof. I can scan my notes, if anyone is interested.

    With regard to Dave Futer’s comment, their result actually implies that hyperbolic manifolds are cubulated, so it supersedes my result (which required Haken & LERF).

    With regard to Danny’s comment, Kahn claims that 3-dimensionality is not important, although they still can’t prove the Ehrenpreis conjecture. But their method should produce surface subgroups in any rank 1 lattice of dimension > 2.

    Jeremy spoke at the conference on Tuesday and Thursday since he only got thru part of the argument on Tuesday. The argument looks
    pretty good to me, modulo several technical details which
    I haven’t checked.
    The idea is similar to Bowen’s approach (and I’m
    told Thurston also suggested gluing pants together).
    Of course, the obstructions to making this approach
    work are well-known: if one has a bunch of almost
    geodesic pants, one must make sure that they are glued
    together to form an almost geodesic surface, and one
    must make sure that there are enough of them to glue
    up to form a closed surface. They seem to have overcome
    both of these obstacles.
    Jeremy first described a one parameter family of
    closed genus 2 surfaces. Take two pants where all three
    boundary curves have length 2R, and glue them together
    along the boundary so that the seams are shifted by 1
    (this could be expressed in Dehn-Thurston coordinates).
    Jeremy claims that this family of genus 2 surfaces
    stays in a bounded subset of moduli space of genus 2
    surfaces (as opposed to doubling the pants, which
    creates short geodesics from the doubled seams).
    As R gets large, the seams of the pants get
    short. But with the twist, the genus 2 surface will still have
    injectivity radius bounded below. He points out that
    any transverse geodesic with length 1 will cross
    the cuffs at most R times. I haven’t checked these claims, but they allow them to control the quasi-isometry of the embedding of surfaces.
    Now, he makes some claims about having pants in
    a hyperbolic 3-manifold which are almost geodesic and
    symmetric, up to some tolerances. I’m not going to write
    down the exact details, but the lengths of the boundary
    curves should be close to 2R, and the rotational part
    should be close to zero, measured by some small constant
    They want to find a lot of these almost symmetric
    and almost geodesic pants in a closed hyperbolic 3-manifold.
    They construct these by taking pairs of tripods, i.e.
    a triple of geodesics which make angles of 120 degrees
    with each other. One should think of these as the
    vertices of the spine of a symmetric pants with cuffs of
    length 2R. They shoot out geodesic rays of length r
    (the length of the edges in said spine), and measure
    how close they get to the other tripod, within some
    epsilon’ (depending on epsilon). If they are close
    (actually, this is done in the frame bundle), they
    get a nearly geodesic pants whose cuff lengths are within
    epsilon of 2R. For a fixed epsilon, this probability
    is non-zero, and gets distributed quite evenly because
    of exponential mixing of the geodesic flow (cf. Moore or
    Now one considers for a geodesic of length
    close to 2R and with small twist number, the collection
    of such almost geodesic symmetric pants with this geodesic
    as a cuff. Because the geodesic flow is exponentially
    mixing, these get equidistributed around the geodesic
    as R -> infinity (this is seen by considering the tripods, and noticing that one may rotate the tripods around the core of the geodesic by its centralizer. Exponential mixing implies that the tripods will close up within epsilon in an evenly distributed fashion about the geodesic).
    Thus, there is a pants very close to differing
    by 1+i*pi translation along a boundary curve of the
    pants. One uses a version of
    Hall’s Marriage Theorem to pair these pants up, and
    show that they can be glued together to form an almost
    geodesic closed surface.
    As I said, I haven’t checked all the details, but
    they certainly seem to have overcome some of the obstacles
    that Lewis Bowen came up against. First, by gluing with
    a twist, the tolerances for gluing pants together to
    get a nearly geodesic surface become manageable. Second,
    they use exponential mixing of the geodesic flow to
    show that nearly geodesic pants are equidistributed
    about a geodesic of length 2R, and get the required
    symmetry to allow one to glue the pants together.

    The reason they can’t prove the 2-dimensional case (i.e. Ehrenpreis) is that the nearly symmetric pants on both sides of a closed geodesic of length close to 2R won’t necessarily match up in number exactly.
    This isn’t a problem in 3-D, because the normal bundle to a geodesic is a torus, and one only needs a matching pants within some tolerance.


    Comment by Ian Agol — August 7, 2009 @ 5:56 pm | Reply

    • Since the argument does not (apparently) discriminate between closed and complete, it is good to hear that there is a clear reason why it doesn’t work in dimension 2!

      Comment by Danny Calegari — August 7, 2009 @ 6:06 pm | Reply

  4. […] pants, surface groups, Waldhausen | by Danny Calegari I just learned from Jesse Johnson’s blog that Vlad Markovic and Jeremy Kahn have announced a proof of the surface subgroup conjecture, that […]

    Pingback by Surface subgroups in hyperbolic 3-manifolds « Geometry and the imagination — August 7, 2009 @ 8:34 pm | Reply

  5. […] his argument with Vlad Markovic, that I blogged earlier about here (also see Jesse Johnson’s blog for other commentary). With his permission, this is reproduced below in its […]

    Pingback by Surface subgroups – more details from Jeremy Kahn « Geometry and the imagination — August 9, 2009 @ 9:32 pm | Reply

RSS feed for comments on this post. TrackBack URI

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Blog at WordPress.com.

%d bloggers like this: