How does the routing number scale for fully connected topologies versus lattice graphs?

This question asks about a theoretical computer science / network analysis topic—how routing complexity scales with network structure—rather than frontier physics. The preprints you've provided are almost entirely about quantum computing (qubit routing, quantum walks on networks), which is adjacent but not quite the same as classical graph routing.

That said, I can extract some relevant scaling insights from the quantum literature here, since the underlying network structure matters in both domains.

What the quantum routing literature suggests

The preprints do confirm that lattice topologies and fully connected (or near-fully connected) topologies have different scaling properties — but the metric depends on what you're optimizing.

For qubit routing in quantum processors:

On lattices (square, hexagonal), the communication overhead scales with the average path length between qubits. A square lattice with N qubits has diameter O(√N), meaning two distant qubits need O(√N) SWAP gates to interact [AG-2024.02-1271]. By contrast, a star topology (one central hub) has diameter 2, but becomes a severe bottleneck—all traffic flows through one qubit, creating hotspots [AG-2024.02-1271].

For quantum walks on complex networks:

The scaling depends on network properties like clustering and small-world structure. The average path length emerges as a universal parameter: quantum search times scale with path length distributions across topologies, from lattices to scale-free networks [AG-2024.01-1819]. This suggests that fully connected or random graphs (shorter average path) enable faster quantum information transfer than regular lattices, though at the cost of physical implementation complexity.

Practical trade-off:

The square lattice is actually favoured for real quantum architectures because it balances routing distance against physical fabrication constraints [AG-2024.02-1271], even though it has longer average path lengths than fully connected designs [AG-2024.03-2090].

---

The provided preprints don't address classical worst-case routing complexity (P-hardness, approximation ratios) for general graphs, so I cannot give you formal scaling laws like "fully connected is O(log n) versus lattice O(√n)" without speculating.