### Clayton Shonkwiler PRO

Mathematician and artist

/osu18

This talk!

A polymer in solution takes on an ensemble of random shapes, with topology as the unique conserved quantity.

*Modern polymer physics is based on the analogy between a polymer chain and a random walk.*

– Alexander Grosberg

Protonated P2VP

Roiter/Minko

Clarkson University

Plasmid DNA

Alonso–Sarduy, Dietler Lab

EPF Lausanne

*Loop* \(\Leftrightarrow\) *point in some (nice!) conformation space*

Knowledge of the (differential, symplectic, algebraic) geometry of these conformation spaces leads to both

**theorems and fast numerical algorithms**.

As in Erik’s talk, points in the Grassmannian \(\mathrm{Gr}_2(\mathbb{C}^n)\) are conformations of closed, framed polygons in \(\mathbb{R}^3\).

Independently spinning the frames corresponds to the action by the torus of diagonal elements of \(U(n)\).

Think of this as a discrete, 3-dimensional version of Younes–Michor–Shah–Mumford’s shape space. Tom has developed the smooth version.

Scalar matrices act trivially on \(\mathrm{Gr}_2(\mathbb{C}^n)\), so there is an effective \(U(1)^{n-1}\) action, and \(\mathrm{Gr}_2(\mathbb{C}^n)/U(1)^{n-1}\) is the space of closed polygons in \(\mathbb{R}^3\).

This is not a symplectic or algebraic operation:

\mathrm{dim}\,\mathrm{Gr}_2(\mathbb{C}^n)/U(1)^{n-1}=4n-8-(n-1)=3n-7

$\mathrm{dim}\,\mathrm{Gr}_2(\mathbb{C}^n)/U(1)^{n-1}=4n-8-(n-1)=3n-7$

To get symplectic/algebraic geometry, need to take the symplectic reduction/GIT quotient

\mathrm{Gr}_2(\mathbb{C}^n)/\!/\!_r U(1)^{n-1} \quad \text{or} \quad \mathrm{Gr}_2(\mathbb{C}^n)/\!/\!_L(\mathbb{C}^*)^{n-1}

$\mathrm{Gr}_2(\mathbb{C}^n)/\!/\!_r U(1)^{n-1} \quad \text{or} \quad \mathrm{Gr}_2(\mathbb{C}^n)/\!/\!_L(\mathbb{C}^*)^{n-1}$

which depends on a choice of moment map fiber/line bundle, but is a symplectic manifold/projective variety.

Geometrically, this choice is equivalent to choosing the lengths \(r_1,\ldots , r_n\) of the edges in the polygon.

**Definition.** \(\mathrm{Pol}(n,r) := \mathrm{Gr}_2(\mathbb{C}^n)/\!/\!_r U(1)^{n-1}\) is the conformation space of closed polygons in \(\mathbb{R}^3\) with edgelength vector \(r=(r_1,\ldots , r_n)\).

In polymer physics, primarily interested in *equilateral polygons*, i.e., the space \(\mathrm{ePol}(n) := \mathrm{Pol}(n,(1,\ldots , 1))\).

This is the space of arclength-parametrized discrete closed curves in \(\mathbb{R}^3\). See Millson–Zombro for the generalization to smooth or Lipschitz closed curves.

The Gelfand–Tsetlin integrable system on \(\mathrm{Gr}_2(\mathbb{C}^n)\) descends to \(n-3\) commuting symmetries on \(\mathrm{Pol}(n,r)\).

The corresponding conserved quantities are the distances \(d_i\) from the first vertex to the \((i+2)\)nd vertex.

The \((n-3)\)-dimensional
* moment polytope* \(\mathcal{P}_n \subset \mathbb{R}^{n-3}\) is defined by the triangle inequalities

0 \leq d_i \leq 2

$0 \leq d_i \leq 2$

1 \leq d_i + d_{i-1}

$1 \leq d_i + d_{i-1}$

|d_i - d_{i-1}| \leq 1

$|d_i - d_{i-1}| \leq 1$

0 \leq d_{n-3} \leq 2

$0 \leq d_{n-3} \leq 2$

**Theorem (with Cantarella, 2016)**

The joint distribution of \(d_1,\ldots , d_{n-3}\) is uniform on \(\mathcal{P}_n\) and \(\theta_1, \ldots , \theta_{n-3}\) are each uniform on \([0,2\pi]\).

**Corollary**

Any algorithm for sampling the convex polytope \(\mathcal{P}_n\subseteq \mathbb{R}^{n-3}\) gives an algorithm for sampling \(\mathrm{ePol}(n)\).

Introduce fake chordlengths \(d_0=1=d_{n-2}\) and make the linear change of variables

\(s_i = d_i - d_{i-1} \text{ for } 1 \leq i \leq n-2\).

Then \(\sum s_i = d_{n-2} - d_0 = 0\), so \(s_{n-2}\) is determined by \(s_1, \ldots , s_{n-3}\)

and the inequalities

0 \leq d_i \leq 2

$0 \leq d_i \leq 2$

1 \leq d_i + d_{i-1}

$1 \leq d_i + d_{i-1}$

|d_i - d_{i-1}| \leq 1

$|d_i - d_{i-1}| \leq 1$

0 \leq d_{n-3} \leq 2

$0 \leq d_{n-3} \leq 2$

become

\(-1 \leq s_i \leq 1, -1 \leq \sum_{i=1}^{n-3} s_i \leq 1\)

\(\sum_{j=1}^i s_j + \sum_{j=1}^{i+1}s_j \geq -1\)

Let \(\mathcal{C}_n \subset \mathbb{R}^{n-3}\) be determined by

\(-1 \leq s_i \leq 1, -1 \leq \sum_{i=1}^{n-3} s_i \leq 1\)

\(\sum_{j=1}^i s_j + \sum_{j=1}^{i+1}s_j \geq -1\)

\(\mathcal{C}_5\)

\(\mathcal{C}_6\)

**Proposition (with Cantarella, Duplantier, Uehara, 2016)**

The probability that a point in the hypercube lies in \(\mathcal{C}_n\) is asymptotic to

\(6 \sqrt{\frac{6}{\pi}}\frac{1}{n^{3/2}}\)

**Action-Angle Method**

**Theorem (with Cantarella, Duplantier, Uehara, 2016)**

The action-angle method directly samples polygon space in expected time \(\Theta(n^{5/2})\).

- Generate \((s_1,\ldots , s_{n-3})\) uniformly on \([-1,1]^{n-3}\)
- Test whether \((s_1,\ldots , s_{n-3})\in \mathcal{C}_n\)
- Change to \((d_1, \ldots , d_{n-3})\) coordinates
- Generate dihedral angles from \(T^{n-3}\)
- Build corresponding polygon

\(O(n)\) time

acceptance probability \(\sim \frac{1}{n^{3/2}}\)

```
RandomDiagonals[n_] :=
Accumulate[
Join[{1}, RandomVariate[UniformDistribution[{-1, 1}], n]]];
InMomentPolytopeQ[d_] :=
And[Last[d] >= 0, Last[d] <= 2,
And @@ (Total[#] >= 1 & /@ Partition[d, 2, 1])];
DiagonalSample[n_] := Module[{d},
For[d = RandomDiagonals[n], ! InMomentPolytopeQ[d], ,
d = RandomDiagonals[n]];
d[[2 ;;]]
];
```

**Definition.** A *frame* in a Hilbert space \(\mathcal{H}\) is a collection \(\{f_i\}\) of elements of \(\mathcal{H}\) so that

A\|x \|^2 \leq \sum |\langle f_i, x\rangle|^2 \leq B\|x\|^2

$A\|x \|^2 \leq \sum |\langle f_i, x\rangle|^2 \leq B\|x\|^2$

for all \(x \in \mathcal{H}\). If \(A=B\), the frame is *tight*. If \(\mathcal{H} = \mathbb{R}^d\) or \(\mathbb{C}^d\), the frame is *finite*. If \(\|f_i\|=1\) for all \(i\), the frame is *unit norm*. Finite unit norm tight frames are called FUNTFs.

**Lemma.** A frame \(F\) in \(\mathbb{C}^d\) is tight \(\Leftrightarrow\) \(T_F^*T_F = \lambda \mathbb{I}_d\).

A finite frame \(F\) in \(\mathbb{C}^d\) has corresponding *analysis operator* \(T_F: \mathbb{C}^d \to \mathbb{C}^n\) given by \(T_F(v) = (\langle v, f_1\rangle, \ldots , \langle x, f_n\rangle)\), *synthesis operator* \(T_F^*\), and *frame operator* \(T_F^*T_F\).

**Example.** The *Mercedes–Benz* frame in \(\mathbb{R}^2\) is a FUNTF.

**Lemma.** A FUNTF in \(\mathbb{C}^d\) has \(T_F^*T_F=\frac{n}{d}\mathbb{I}_d\).

\(\mathcal{F}^n_{\frac{n}{d} \mathbb{I}_d} =\{F \subseteq \mathbb{C}^d : T_F^*T_F=\frac{n}{d}\mathbb{I}_d\} \simeq \mathrm{St}_d(\mathbb{C}^n) \), the Stiefel manifold of orthonormal \(d\)-frames in \(\mathbb{C}^n\), and hence \(U(d)\backslash \mathcal{F}^n_{\frac{n}{d}\mathbb{I}_d} \simeq \mathrm{Gr}_d(\mathbb{C}^n)\).

If \(S\) is an invertible, positive definite, Hermitian \(d \times d\) matrix, let \(\mathcal{F}^n_S := \{F = \{f_i\}_{i=1}^n \subseteq \mathbb{C}^d : T_F^*T_F=S\}\)

Let \(\mathcal{F}^n_S(r)\subseteq\mathcal{F}^n_S\) be the space of frames with frame operator \(S\) and \(\|f_i\| = r_i\) for all \(i\).

**Proposition.** \(U(d)\backslash\mathcal{F}^n_{\frac{n}{d}\mathbb{I}_d}(r)/U(1)^{n-1} \simeq \mathrm{Gr}_d(\mathbb{C}^n)/\!/\!_r U(1)^{n-1}=\mathrm{Pol}(n,r)\)

**Theorem (2018)**

A modification of the action-angle algorithm directly samples FUNTFs in \(\mathbb{C}^2\) in \(\Theta(n^{5/2})\) time.

Equilateral polygons in \(\mathbb{R}^3\) lift to FUNTFs in \(\mathbb{C}^2\)!

Histogram of coherences of length-6 FUNTFs in \(\mathbb{C}^2\)

Histogram of coherences of length-4 FUNTFs in \(\mathbb{C}^2\)

Coherences of lifts of Sloane's optimal point packings \(S^2\), compared to the Toth bound

Similar symplectic reasoning leads to:

**Theorem (with Needham, 2018)**

If \(S\) is an invertible, positive-definite, \(d \times d\) matrix, the space \(\mathcal{F}^n_S(r)\) (of length-\(n\) frames in \(\mathbb{C}^d\) with frame operator \(S\) and \(\|f_i\|=r_i\)) is path-connected.

This generalizes Cahill, Mixon, and Strawn’s (2017) resolution of the *frame homotopy conjecture*, which showed that the space of FUNTFs in \(\mathbb{C}^d\) is path-connected.

- Does distance in the Grassmannian serve as a reasonable proxy for structural similarity physical loops (e.g., ring biopolymers)?
- What is going on with coherences of length-4 FUNTFs in \(\mathbb{C}^2\)?
- Is there a (fast) sampling algorithm for FUNTFs in \(\mathbb{C}^d\)?
- What is the corresponding 2D/real story?

The symplectic geometry of closed equilateral random walks in 3-space

J. Cantarella & C. Shonkwiler

*Annals of Applied Probability*
**26**
(2016), no. 1, 549–596

A fast direct sampling algorithm for equilateral closed polygons

J. Cantarella, B. Duplantier, C. Shonkwiler, & E. Uehara

*Journal of Physics A* **49** (2016), no. 27, 275202

Funding: Simons Foundation

Symplectic geometry of spaces of frames

T. Needham & C. Shonkwiler

In preparation

By Clayton Shonkwiler

Random flights in 3-space forming a closed loop, or random polygons, are a standard simplified model of so-called ring polymers like bacterial DNA. Equilateral random polygons, where all steps are the same length, are particularly interesting (and challenging) in this context. In this talk I will describe an (almost) toric Kähler structure on the moduli space of equilateral polygons and show how to exploit this structure to get a fast algorithm to directly sample the space. Using work of Hausmann–Knutson, the Kähler structure on the space of equilateral polygons can be realized as the Kähler reduction of the standard Kähler structure on the Grassmannian of 2-planes in complex n-space. This means that equilateral polygons in 3-space can be lifted to Finite Unit-Norm Tight Frames (FUNTFs) in complex 2-space. I will describe how to modify the polygon sampler to produce a FUNTF sampler and show that optimal packings in the 2-sphere lift to FUNTFs with low coherence.

- 1,533