A second, independent 2025 paper (Bolliger et al.) confirms differentiable MPM is a standing technique, closing a 26–42 year gap from Linnainmaa's algorithm and SAND-91-7095 to robotics applications

material-point-methoddifferentiable-simulationautomatic-differentiationlinnainmaasandiaroboticshistory-of-mlbridge-investigation

Six years after ChainQueen and from an unconnected author group — Bolliger, Fadini, Bambach & Rupenyan (ZHAW / ETH Zürich), different institutions, different authors — a second preprint builds a differentiable Material Point Method simulator for deformable-object control, citing ChainQueen itself as prior work (its reference [12]; "[12], [13] introduce differentiable MPM simulators") rather than rediscovering it, and stating plainly: "Using AD, we can additionally obtain the derivative of the cost with respect to the optimization variables." The automatic differentiation is supplied by JAX, whose grad transform implements the same reverse-mode family Linnainmaa's 1976 Algorithm T belongs to. This corroborates ChainQueen's earlier instance as a standing line of research rather than a one-off.

Dating the three fixed points already established at Tier 1 elsewhere in the vault — Linnainmaa's BIT paper (1976), SAND-91-7095 (OSTI-confirmed January 1992), and ChainQueen (submitted October 2018, published 2019) — yields a 42-year gap from Linnainmaa's algorithm to its first documented application to MPM, and a 26-year gap from SAND-91-7095 to the same event. Two techniques that originated independently, in unrelated institutions, for unrelated purposes (rounding-error analysis; penetration mechanics), were joined only decades later in a third domain — differentiable physics for robotics — that neither founding author group could have anticipated. As with ChainQueen, Bolliger et al. cite no numerical-analysis AD source by name; the technique arrives through JAX, not through direct attribution.