Primitive rings and OVITO FindRings

A ring is primitive when every pair of its vertices is as close along the ring as it is through the rest of the graph. That is Franzblau’s shortest-path test (Phys. Rev. B 44, 4925, 1991). OVITO’s FindRingsModifier is a shortest-path census of the same kind of ring, and on the frames where both searches stop at six-membered rings the two counts agree. It is not a complete primitive census. FindRings can drop a chordless ring that the Franzblau test keeps.

The same bond graph

Both searches on this page use one cutoff, 3.5 Å, and the minimum image of the periodic cell. A pair is a bond when its squared distance is strictly below the squared cutoff. The comparison is the bond list, then the rings. A count that differs while the bond list differs is a neighbour-list mismatch, not a ring-definition mismatch.

K2,3

Five vertices and six edges. Label the two parts \(\{0,3\}\) and \(\{1,2,4\}\), and join every vertex of the first part to every vertex of the second. There is no edge inside either part, so none of the three 4-cycles has a chord:

  • \(0-1-3-2\)

  • \(0-1-3-4\)

  • \(0-2-3-4\)

Each pair of opposite vertices on those cycles is at distance 2 in the graph, which is the ring distance. All three cycles are primitive. ringNetwork with maximum size 4 returns the three. FindRings with the same six bonds and maximum size 4 returns two. The missing cycle is primitive, so the smaller count is not a stricter definition. It is a ring the search did not emit.

tests/test_franzblau_extra.cpp builds this adjacency and requires the three primitive 4-rings. tests/test_ovito_ring_counts.py places the same graph in a cell, checks that a 3.5 Å cutoff reproduces the six edges, and checks the OVITO count of 2.

The interface frame

input/traj/interface_f0.lammpstrj is orthorhombic, 864 atoms, all of one type. At 3.5 Å the cutoff graph has 1929 bonds on both sides. ringNetwork with maximum size 10 returns 2662 rings. Every one of them passes the Franzblau test: for every pair, the shorter arc is a breadth-first distance in that 1929-bond graph. FindRings on the same bonds, with the same maximum size, returns fewer rings at each size from 4 through 10. Those missing rings are among the 2662, so each of them is primitive.

The counts that agree with OVITO elsewhere stop at hexagonal rings and use other frames. They do not contradict this frame. A search that stopped at size 6 never walks the shortcuts that make a larger primitive ring visible, and K2,3 is already a counterexample at size 4.

What the test asserts

tests/test_ovito_ring_counts.py does not require the two histograms to be equal. It requires the bond counts to match, every d-SEAMS ring to be primitive, and each OVITO bin to be at most the primitive bin. The K2,3 case is the strict subset: 3 primitive 4-rings, and FindRings reports 2.