Lower bound on the number of patterns

99.2. Lower bound on the number of patterns

Proposition 99.2

Let Nw,dN_{w,d} be the number of port graphs of width ww, depth dd and maximum degree Δ4\Delta \geq 4. We can lower bound

Nw,d>(w2e)Θ(wd),N_{w,d} > \left(\frac{w}{2e}\right)^{\Theta(wd)},

assuming wo(2d)w \leq o(2^d).

In the regime of interest, ww is small, so the assumption wo(2d)w \leq o(2^d) is not a restriction.

Proof

Let w,d>0w, d > 0 and Δ4\Delta \geq 4 be integers. We wish to lower bound the number of port graphs of depth dd, width ww and maximum degree Δ\Delta. It is sufficient to consider a restricted subset of such port graphs, whose size can be easily lower bounded. We will count a subset of CX quantum circuits, i.e. circuits with only CXCX gates, a two-qubit non-symmetric gate. Because we are using a single gate type, this is equivalent to counting a subset of port graphs with vertices of degree 4. Assume w.l.o.g that ww is a power of two. We consider CX circuits constructed from two circuits with ww qubits composed in sequence:

  • Fixed tree circuit: A log2(w)\log_2(w)-depth circuit that connects qubits pairwise in such a way that the resulting port graph is connected. We fix such a tree-like circuit and use the same circuit for all CX circuits. We can use this common structure to fix an ordering of the ww qubits, that refer to as qubits 1,,w1,\dots,w.
  • Bipartite circuit: A CX circuit of depth D=dlog2(w)D = d - \log_2(w) with exactly CX gates, each gate acting on a qubit and a qubit .

The following circuit illustrates the construction:

All that remains is to count the number of such bipartite circuits. Every slice of depth 1 must have w/2w / 2 CX gates acting on distinct qubits. Every qubit 11 to w/2w/2 must interact with one of the qubits w/2+1w/2+1 to ww, so there are (w/2)!(w/2)! such depth 1 slices. Repeating this depth 1 construction DD times and using Sterling’s approximation, we obtain a lower bound for the number of port graphs of depth dd, width ww and maximum degree at least 4:

((w2)!)D>wπ(w2e)wD/2=(w2e)Θ(wd)\left(\left(\frac{w}2\right)!\right)^D > \sqrt{w\pi}\left(\frac{w}{2e}\right)^{wD/2} = \left(\frac{w}{2e}\right)^{\Theta(w\cdot d)}

where we used w=o(2d)w = o(2^d) to obtain Θ(D)=Θ(d)\Theta(D) = \Theta(d) in the last step.