Update Colvars library to version 2025-04-18

The following is a list of pull requests relevant to LAMMPS in the Colvars repository since 2024-08-06:

- 752 New tool poisson_integrator_conv
  https://github.com/Colvars/colvars/pull/752 (@jhenin)

- 733 Custom grids for all biases
  https://github.com/Colvars/colvars/pull/733 (@giacomofiorin, @jhenin)

- 776 Avoid error in acos and asin with fast-math
  https://github.com/Colvars/colvars/pull/776 (@jhenin)

- 773 fix: fix the clang build test failure of OPES
  https://github.com/Colvars/colvars/pull/773 (@HanatoK)

- 768 fix: clamp the input values of asin and acos in case of fast math on aarch64
  https://github.com/Colvars/colvars/pull/768 (@HanatoK)

- 761 Add debug code for the Jacobi failure
  https://github.com/Colvars/colvars/pull/761 (@HanatoK)

- 759 min_image fix; Saves long runs from crashes;
  https://github.com/Colvars/colvars/pull/759 (@PolyachenkoYA)

- 757 Fix MSVC OpenMP issue
  https://github.com/Colvars/colvars/pull/757 (@HanatoK)

- 755 Fix indentation of 'Init CVC' message in standard output
  https://github.com/Colvars/colvars/pull/755 (@jhenin)

- 750 Optimize and simplify the calculation of dihedral gradients
  https://github.com/Colvars/colvars/pull/750 (@HanatoK)

- 749 Add references to new Colvars paper
  https://github.com/Colvars/colvars/pull/749 (@jhenin, @giacomofiorin)

- 740 Report the specific C++ standard at init time, stop warning about C++97/03
  https://github.com/Colvars/colvars/pull/740 (@giacomofiorin)

- 731 Improve detection of hard/mathematical boundaries
  https://github.com/Colvars/colvars/pull/731 (@giacomofiorin)

- 729 Optimize the fit gradients
  https://github.com/Colvars/colvars/pull/729 (@HanatoK, @jhenin)

- 728 Fix undefined behavior when getting the current working directory from std::filesystem
  https://github.com/Colvars/colvars/pull/728 (@giacomofiorin)

- 727 Add patchversion scripting command
  https://github.com/Colvars/colvars/pull/727 (@giacomofiorin)

- 724 Fix gradients and metric functions of distanceDir
  https://github.com/Colvars/colvars/pull/724 (@giacomofiorin)

- 715 Add missing rotation in orientation component
  https://github.com/Colvars/colvars/pull/715 (@giacomofiorin)

- 713 fix: try to solve #87 for non-scala components
  https://github.com/Colvars/colvars/pull/713 (@HanatoK)

- 709 Implementation of OPES in Colvars
  https://github.com/Colvars/colvars/pull/709 (@HanatoK, @giacomofiorin, @jhenin)

- 706 BUGFIX for Segmentation fault in colvarbias_meta::calc_energy() with useGrids off
  https://github.com/Colvars/colvars/pull/706 (@alphataubio)

- 570 enable use of CVs defined by PyTorch neural network models
  https://github.com/Colvars/colvars/pull/570 (@zwpku, @giacomofiorin, @HanatoK, @jhenin)

Authors: @alphataubio, @EzryStIago, @giacomofiorin, @HanatoK, @jhenin, @PolyachenkoYA, @zwpku

lib/colvars/colvarcomp_alchlambda.cpp

@@ -20,22 +20,46 @@ colvar::alch_lambda::alch_lambda()
 {
   set_function_type("alchLambda");
 
-  disable(f_cvc_explicit_gradient);
-  disable(f_cvc_gradient);
+  provide(f_cvc_explicit_gradient, false);
+  provide(f_cvc_gradient, false); // Cannot apply forces on this CVC
+  provide(f_cvc_collect_atom_ids, false);
+
+  provide(f_cvc_inv_gradient); // Projected force is TI derivative
+  provide(f_cvc_Jacobian);     // Zero
 
   x.type(colvarvalue::type_scalar);
-  // Query initial value from back-end
+
+  // Query initial value from back-end; will be overwritten if restarting from a state file
   cvm::proxy->get_alch_lambda(&x.real_value);
 }
 
 
+int colvar::alch_lambda::init_alchemy(int factor)
+{
+  // We need calculation every time step
+  // default in Tinker-HP and NAMD2, must be enforced in NAMD3
+  // Also checks back-end settings, ie. that alchemy is enabled
+  // (in NAMD3: alchType TI, computeEnergies at the right frequency)
+
+  // Forbid MTS until fully implemented
+  if (factor != 1) {
+    return cvm::error("Error: timeStepFactor > 1 is not yet supported for alchemical variables.");
+  }
+  cvm::proxy->request_alch_energy_freq(factor);
+
+  return COLVARS_OK;
+}
+
+
 void colvar::alch_lambda::calc_value()
 {
-  // Special workflow:
-  // at the beginning of the timestep we get a force instead of calculating the value
+  // By default, follow external parameter
+  // This might get overwritten by driving extended dynamics
+  // (in apply_force() below)
+  cvm::proxy->get_alch_lambda(&x.real_value);
 
   cvm::proxy->get_dE_dlambda(&ft.real_value);
-  ft.real_value *= -1.0; // Energy derivative to force
+  ft.real_value *= -1.0; // Convert energy derivative to force
 
   // Include any force due to bias on Flambda
   ft.real_value += cvm::proxy->indirect_lambda_biasing_force;
@@ -43,19 +67,24 @@ void colvar::alch_lambda::calc_value()
 }
 
 
-void colvar::alch_lambda::calc_gradients()
+void colvar::alch_lambda::calc_force_invgrads()
 {
+  // All the work is done in calc_value()
+}
+
+
+void colvar::alch_lambda::calc_Jacobian_derivative()
+{
+  jd = 0.0;
 }
 
 
 void colvar::alch_lambda::apply_force(colvarvalue const & /* force */)
 {
-  // new value will be cached and sent at end of timestep
-  cvm::proxy->set_alch_lambda(x.real_value);
+  // Forces, if any, are applied in colvar::update_extended_Lagrangian()
 }
 
 
-
 colvar::alch_Flambda::alch_Flambda()
 {
   set_function_type("alch_Flambda");