MultiOptPy 1.20.2__py3-none-any.whl → 1.20.4__py3-none-any.whl

multioptpy/optimization.py

@@ -1,11 +1,10 @@
-import sys
 import os
-import copy
+import sys
 import glob
+import copy
 import itertools
+import inspect
 import datetime
-
-
 import numpy as np
 
 from multioptpy.optimizer import CalculateMoveVector
@@ -22,32 +21,46 @@ from multioptpy.Utils.calc_tools import CalculationStructInfo, Calculationtools
 from multioptpy.Constraint.constraint_condition import ProjectOutConstrain
 from multioptpy.irc import IRC
 from multioptpy.Utils.bond_connectivity import judge_shape_condition
-from multioptpy.Utils.oniom import separate_high_layer_and_low_layer, specify_link_atom_pairs, link_number_high_layer_and_low_layer
+from multioptpy.Utils.oniom import (
+    separate_high_layer_and_low_layer,
+    specify_link_atom_pairs,
+    link_number_high_layer_and_low_layer,
+)
 from multioptpy.Utils.symmetry_analyzer import analyze_symmetry
 from multioptpy.Thermo.normal_mode_analyzer import MolecularVibrations
-
-# Responsibility 1: Holds the "Configuration" (immutable)
+from multioptpy.ModelFunction.opt_meci import OptMECI
+from multioptpy.ModelFunction.opt_mesx import OptMESX
+from multioptpy.ModelFunction.opt_mesx_2 import OptMESX2
+from multioptpy.ModelFunction.seam_model_function import SeamModelFunction
+from multioptpy.ModelFunction.conical_model_function import ConicalModelFunction
+from multioptpy.ModelFunction.avoiding_model_function import AvoidingModelFunction
+from multioptpy.ModelFunction.binary_image_ts_search_model_function import BITSSModelFunction
+# =====================================================================================
+# 1. Configuration (Immutable Settings)
+# =====================================================================================
 class OptimizationConfig:
     """
-    Holds all "settings" that do not change during the run.
-    Initialized from 'args'.
+    Immutable settings derived from CLI args.
     """
+
     def __init__(self, args):
-        # Constants like UVL
         UVL = UnitValueLib()
         np.set_printoptions(precision=12, floatmode="fixed", suppress=True)
+
+        # Physical constants
         self.hartree2kcalmol = UVL.hartree2kcalmol
         self.bohr2angstroms = UVL.bohr2angstroms
         self.hartree2kjmol = UVL.hartree2kjmol
 
-        # Port the logic from _set_convergence_criteria
+        # Base args
+        self.args = args
         self._set_convergence_criteria(args)
 
-        # Port all "immutable" settings from _initialize_variables
+        # Core parameters
         self.microiter_num = 100
-        self.args = args  # Keep a reference to args
         self.FC_COUNT = args.calc_exact_hess
-        self.temperature = 0.0
+        self.mFC_COUNT = args.calc_model_hess
+        self.temperature = float(args.temperature) if args.temperature else 0.0
         self.CMDS = args.cmds
         self.PCA = args.pca
         self.DELTA = "x" if args.DELTA == "x" else float(args.DELTA)
@@ -57,11 +70,11 @@ class OptimizationConfig:
         self.BASIS_SET = args.basisset
         self.FUNCTIONAL = args.functional
         self.excited_state = args.excited_state
-        
-        # Port the logic from _check_sub_basisset
+
+        # Sub-basis and ECP
         self._check_sub_basisset(args)
-        
-        self.mFC_COUNT = args.calc_model_hess
+
+        # Advanced settings
         self.DC_check_dist = float(args.dissociate_check)
         self.unrestrict = args.unrestrict
         self.irc = args.intrinsic_reaction_coordinates
@@ -82,40 +95,62 @@ class OptimizationConfig:
         self.software_path_file = args.software_path_file
         self.koopman_analysis = args.koopman
         self.detect_negative_eigenvalues = args.detect_negative_eigenvalues
+        self.excited_state = args.excited_state
+        self.spin_multiplicity = args.spin_multiplicity
+        self.electronic_charge = args.electronic_charge
+        self.model_function = args.model_function
+        
+        
 
     def _set_convergence_criteria(self, args):
-        # Original _set_convergence_criteria method code
         if args.tight_convergence_criteria and not args.loose_convergence_criteria:
             self.MAX_FORCE_THRESHOLD = 0.000015
             self.RMS_FORCE_THRESHOLD = 0.000010
             self.MAX_DISPLACEMENT_THRESHOLD = 0.000060
             self.RMS_DISPLACEMENT_THRESHOLD = 0.000040
+            
+            if len(args.projection_constrain) > 0:
+                self.MAX_DISPLACEMENT_THRESHOLD *= 4
+                self.RMS_DISPLACEMENT_THRESHOLD *= 4
+            
+            
+            
         elif not args.tight_convergence_criteria and args.loose_convergence_criteria:
             self.MAX_FORCE_THRESHOLD = 0.0030
             self.RMS_FORCE_THRESHOLD = 0.0020
             self.MAX_DISPLACEMENT_THRESHOLD = 0.0100
             self.RMS_DISPLACEMENT_THRESHOLD = 0.0070
+
+             
         else:
             self.MAX_FORCE_THRESHOLD = 0.0003
             self.RMS_FORCE_THRESHOLD = 0.0002
             self.MAX_DISPLACEMENT_THRESHOLD = 0.0015
             self.RMS_DISPLACEMENT_THRESHOLD = 0.0010
-
+            if len(args.projection_constrain) > 0: 
+                self.MAX_DISPLACEMENT_THRESHOLD *= 4
+                self.RMS_DISPLACEMENT_THRESHOLD *= 4   
+        
     def _check_sub_basisset(self, args):
-        # Original _check_sub_basisset method code
         if len(args.sub_basisset) % 2 != 0:
             print("invalid input (-sub_bs)")
             sys.exit(0)
-        self.electric_charge_and_multiplicity = [int(args.electronic_charge), int(args.spin_multiplicity)]
+
+        self.electric_charge_and_multiplicity = [
+            int(args.electronic_charge),
+            int(args.spin_multiplicity),
+        ]
         self.electronic_charge = args.electronic_charge
         self.spin_multiplicity = args.spin_multiplicity
-        
+
         if args.pyscf:
             self.SUB_BASIS_SET = {}
             if len(args.sub_basisset) > 0:
                 self.SUB_BASIS_SET["default"] = str(self.BASIS_SET)
                 for j in range(int(len(args.sub_basisset) / 2)):
-                    self.SUB_BASIS_SET[args.sub_basisset[2 * j]] = args.sub_basisset[2 * j + 1]
+                    self.SUB_BASIS_SET[args.sub_basisset[2 * j]] = args.sub_basisset[
+                        2 * j + 1
+                    ]
                 print("Basis Sets defined by User are detected.")
                 print(self.SUB_BASIS_SET)
             else:
@@ -125,10 +160,16 @@ class OptimizationConfig:
             if len(args.sub_basisset) > 0:
                 self.SUB_BASIS_SET += "\nassign " + str(self.BASIS_SET) + "\n"
                 for j in range(int(len(args.sub_basisset) / 2)):
-                    self.SUB_BASIS_SET += "assign " + args.sub_basisset[2 * j] + " " + args.sub_basisset[2 * j + 1] + "\n"
+                    self.SUB_BASIS_SET += (
+                        "assign "
+                        + args.sub_basisset[2 * j]
+                        + " "
+                        + args.sub_basisset[2 * j + 1]
+                        + "\n"
+                    )
                 print("Basis Sets defined by User are detected.")
                 print(self.SUB_BASIS_SET)
-        
+
         if len(args.effective_core_potential) % 2 != 0:
             print("invaild input (-ecp)")
             sys.exit(0)
@@ -136,640 +177,1352 @@ class OptimizationConfig:
         if args.pyscf:
             self.ECP = {}
             if len(args.effective_core_potential) > 0:
-                for j in range(int(len(args.effective_core_potential)/2)):
-                    self.ECP[args.effective_core_potential[2*j]] = args.effective_core_potential[2*j+1]
+                for j in range(int(len(args.effective_core_potential) / 2)):
+                    self.ECP[
+                        args.effective_core_potential[2 * j]
+                    ] = args.effective_core_potential[2 * j + 1]
         else:
             self.ECP = ""
 
-# Responsibility 2: Manages the "State" (mutable)
+# =====================================================================================
+# 2. State (Mutable Data)
+# =====================================================================================
 class OptimizationState:
     """
-    Holds all "state" variables that change during the optimization loop.
+    Dynamic state of the optimization.
     """
+
     def __init__(self, element_list):
         natom = len(element_list)
-        
-        # Current step state
         self.iter = 0
-        self.e = None  # Hartree
-        self.B_e = None  # Hartree
-        self.g = None  # Hartree/Bohr
-        self.B_g = None  # Hartree/Bohr
-        self.geom_num_list = None  # Bohr
-        self.Model_hess = np.eye(natom * 3) # Model_hess is treated as state
-        self.element_list = element_list
 
-        # Previous step state
-        self.pre_e = 0.0
-        self.pre_B_e = 0.0
-        self.pre_geom = np.zeros((natom, 3), dtype="float64")
-        self.pre_g = np.zeros((natom, 3), dtype="float64")
-        self.pre_B_g = np.zeros((natom, 3), dtype="float64")
+        # Energies and gradients
+        self.energies = {}
+        self.gradients = {}
+        self.raw_energy = 0.0
+        self.raw_gradient = np.zeros((natom, 3), dtype="float64")
+        self.bias_energy = 0.0
+        self.bias_gradient = np.zeros((natom, 3), dtype="float64")
+        self.effective_energy = 0.0
+        self.effective_gradient = np.zeros((natom, 3), dtype="float64")
+
+        # Geometry
+        self.geometry = None  # Bohr
+        self.initial_geometry = None  # Bohr
+
+        # Previous step
+        self.pre_geometry = np.zeros((natom, 3), dtype="float64")
+        self.pre_effective_gradient = np.zeros((natom, 3), dtype="float64")
+        self.pre_bias_gradient = np.zeros((natom, 3), dtype="float64")
+        self.pre_effective_energy = 0.0
+        self.pre_bias_energy = 0.0
+        self.pre_raw_gradient = np.zeros((natom, 3), dtype="float64")
         self.pre_move_vector = np.zeros((natom, 3), dtype="float64")
+        self.pre_raw_energy = 0.0
+
+        # Hessian
+        self.Model_hess = np.eye(natom * 3)
 
-        # Plotting / result lists
+        # Logs
+        self.history = {
+            "iter": [],
+            "energies": {},
+            "grad_rms": [],
+            "bias_grad_rms": [],
+        }
+
+        # For plotting and outputs
         self.ENERGY_LIST_FOR_PLOTTING = []
         self.BIAS_ENERGY_LIST_FOR_PLOTTING = []
         self.NUM_LIST = []
         self.grad_list = []
         self.bias_grad_list = []
-        self.cos_list = [] # Initialized properly in Optimize class
+        self.cos_list = []
 
-        # Final result placeholders
+        # Final results
         self.final_file_directory = None
         self.final_geometry = None
         self.final_energy = None
         self.final_bias_energy = None
         self.bias_pot_params_grad_list = None
         self.bias_pot_params_grad_name_list = None
+        self.symmetry = None
 
         # Flags
-        self.DC_check_flag = False
-        self.optimized_flag = False
         self.exit_flag = False
+        self.converged_flag = False
+        self.dissociation_flag = False
+        self.optimized_flag = False
+        self.DC_check_flag = False
 
-# Responsibility 3: Performs the "Execution" (main logic)
-class Optimize:
-    """
-    Main execution (Runner) class.
-    It holds the Config, creates and updates the State,
-    and runs the main optimization logic.
-    """
-    def __init__(self, args):
-        # 1. Set the Configuration (immutable)
-        self.config = OptimizationConfig(args)
-        
-        # 2. State will be created freshly inside the run() method for each job.
-        self.state = None 
+# =====================================================================================
+# 3. Potential Handlers (Strategy)
+# =====================================================================================
+class BasePotentialHandler:
+    def __init__(self, config, file_io, base_dir, force_data):
+        self.config = config
+        self.file_io = file_io
+        self.base_dir = base_dir
+        self.force_data = force_data
+        self.bias_pot_calc = BiasPotentialCalculation(base_dir)
+
+    def compute(self, state: OptimizationState):
+        raise NotImplementedError("Subclasses must implement compute()")
+
+    def _add_bias_and_update_state(self, state, raw_energy, raw_gradient):
+        # Store raw
+        state.raw_energy = raw_energy
+        state.raw_gradient = raw_gradient
+        state.energies["raw"] = raw_energy
+        state.gradients["raw"] = raw_gradient
+
+        # Bias
+        _, bias_e, bias_g, bpa_hess = self.bias_pot_calc.main(
+            raw_energy,
+            raw_gradient,
+            state.geometry,
+            state.element_list,
+            self.force_data,
+            state.pre_bias_gradient,
+            state.iter,
+            initial_geom_num_list=state.initial_geometry,
+        )
+        state.bias_energy = bias_e
+        state.bias_gradient = bias_g
+        state.energies["bias"] = bias_e
+        state.gradients["bias"] = bias_g
 
-        # 3. Helper instances and job-specific variables
-        self.BPA_FOLDER_DIRECTORY = None
-        self.START_FILE = None
-        self.element_list = None # Will be set in run()
-        self.CalcBiaspot = None # Shared helper
-        self.SP = None # Shared helper
+        # Effective
+        state.effective_energy = bias_e
+        state.effective_gradient = bias_g
+        state.energies["effective"] = state.effective_energy
+        state.gradients["effective"] = state.effective_gradient
 
-        # 4. Final results (for external access, mirrors original design)
-        self.final_file_directory = None
-        self.final_geometry = None
-        self.final_energy = None
-        self.final_bias_energy = None
-        self.irc_terminal_struct_paths = []
-        self.optimized_struct_file = None
-        self.traj_file = None
-        self.symmetry = None
+        # Attach bias Hessian for later use
+        state.bias_hessian = bpa_hess
+        return state
 
-    # --- Helper Methods ---
-    # (Ported from the original class)
-    # These methods must now read from self.config
-    # and read/write from self.state.
 
-    def _make_init_directory(self, file):
-        """
-        Create initial directory for optimization results.
-        Uses self.config to build the path.
-        """
-        self.START_FILE = file
-        timestamp = datetime.datetime.now().strftime("%Y_%m_%d_%H_%M_%S_%f")[:-2]
-        date = datetime.datetime.now().strftime("%Y_%m_%d")
-        base_dir = f"{date}/{self.START_FILE[:-4]}_OPT_"
+class StandardHandler(BasePotentialHandler):
+    """
+    Handles standard single-PES calculations.
+    """
 
-        if self.config.othersoft != "None":
-            self.BPA_FOLDER_DIRECTORY = f"{base_dir}ASE_{timestamp}/"
-        elif self.config.sqm2:
-            self.BPA_FOLDER_DIRECTORY = f"{base_dir}SQM2_{timestamp}/"
-        elif self.config.sqm1:
-            self.BPA_FOLDER_DIRECTORY = f"{base_dir}SQM1_{timestamp}/"
-        elif self.config.args.usextb == "None" and self.config.args.usedxtb == "None":
-            self.BPA_FOLDER_DIRECTORY = f"{base_dir}{self.config.FUNCTIONAL}_{self.config.BASIS_SET}_{timestamp}/"
+    def __init__(self, calculator, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.calculator = calculator
+
+    def compute(self, state: OptimizationState):
+        file_path = self.file_io.make_psi4_input_file(
+            self.file_io.print_geometry_list(
+                state.geometry * self.config.bohr2angstroms,
+                state.element_list,
+                self.config.electric_charge_and_multiplicity,
+            ),
+            state.iter,
+        )
+
+        self.calculator.Model_hess = copy.deepcopy(state.Model_hess)
+        e, g, geom_num_list, exit_flag = self.calculator.single_point(
+            file_path,
+            state.element_number_list,
+            state.iter,
+            self.config.electric_charge_and_multiplicity,
+            self.calculator.xtb_method,
+        )
+        state.geometry = np.array(geom_num_list, dtype="float64")
+        if exit_flag:
+            state.exit_flag = True
+            return state
+
+        state = self._add_bias_and_update_state(state, e, g)
+        state.Model_hess = copy.deepcopy(self.calculator.Model_hess)
+        return state
+
+class ModelFunctionHandler(BasePotentialHandler):
+    def __init__(self, calc1, calc2, mf_args, config, file_io, base_dir, force_data):
+        super().__init__(config, file_io, base_dir, force_data)
+        self.calc1 = calc1
+        self.calc2 = calc2
+        
+        self.method_name = mf_args[0].lower()
+        self.params = mf_args[1:]
+        
+        self.is_bitss = "bitss" in self.method_name
+        self.single_element_list = None 
+        
+        self.mf_instance = self._load_mf_class()
+        self._apply_config_params()
+
+        self.bitss_geom1_history = []
+        self.bitss_geom2_history = []
+        self.bitss_ref_geom = None
+        
+        self.bitss_initialized = False 
+
+        if self.is_bitss:
+            self._setup_bitss_initialization()
+
+    def _load_mf_class(self):
+        if self.method_name == "opt_meci": 
+            return OptMECI()
+        if self.method_name == "opt_mesx": 
+            return OptMESX()
+        if self.method_name == "opt_mesx_2":
+            return OptMESX2()
+        if self.method_name == "seam":
+            return SeamModelFunction()
+        if self.method_name == "conical": 
+            return ConicalModelFunction()
+        if self.method_name == "avoiding": 
+            return AvoidingModelFunction()
+        if self.method_name == "bitss": 
+            return BITSSModelFunction(np.zeros(1), np.zeros(1))
+        raise ValueError(f"Unknown Model Function: {self.method_name}")
+
+    def _apply_config_params(self):
+        if hasattr(self.config.args, 'alpha') and self.config.args.alpha is not None:
+            if hasattr(self.mf_instance, 'alpha'):
+                self.mf_instance.alpha = float(self.config.args.alpha)
+        if hasattr(self.config.args, 'sigma') and self.config.args.sigma is not None:
+            if hasattr(self.mf_instance, 'sigma'):
+                self.mf_instance.sigma = float(self.config.args.sigma)
+
+    def _setup_bitss_initialization(self):
+        if len(self.params) < 1:
+            raise ValueError("BITSS requires a reference geometry file path.")
+        
+        temp_io = FileIO(self.base_dir, self.params[0])
+        g_list, _, _ = temp_io.make_geometry_list(self.config.electric_charge_and_multiplicity)
+        
+        coords_ang = np.array([atom[1:4] for atom in g_list[0][2:]], dtype=float)
+        self.bitss_ref_geom = coords_ang / self.config.bohr2angstroms
+
+    def compute(self, state: OptimizationState):
+        iter_idx = state.iter
+        
+        if self.single_element_list is None:
+            self.single_element_list = state.element_list[:len(state.element_list)//2] if self.is_bitss else state.element_list
+
+        # --- 1. Prepare Geometries ---
+        if self.is_bitss:
+            n_atoms = len(self.single_element_list)
+            geom_1, geom_2 = state.geometry[:n_atoms], state.geometry[n_atoms:]
+            
+            if not self.bitss_initialized:
+                self.mf_instance = BITSSModelFunction(geom_1, geom_2)
+                self._apply_config_params()
+                self.bitss_initialized = True
         else:
-            method = self.config.args.usedxtb if self.config.args.usedxtb != "None" else self.config.args.usextb
-            self.BPA_FOLDER_DIRECTORY = f"{base_dir}{method}_{timestamp}/"
+            geom_1 = geom_2 = state.geometry
+
+        # State 1
+        e1, g1, ex1 = self._run_calc(self.calc1, geom_1, self.single_element_list, self.config.electric_charge_and_multiplicity, "State1", iter_idx)
         
-        os.makedirs(self.BPA_FOLDER_DIRECTORY, exist_ok=True)
+        # State 2
+        if self.is_bitss:
+            chg_mult_2 = self.config.electric_charge_and_multiplicity
+        else:
+            if len(self.params) >= 2:
+                chg_mult_2 = [int(self.params[0]), int(self.params[1])]
+            else:
+                chg_mult_2 = self.config.electric_charge_and_multiplicity
+
+        e2, g2, ex2 = self._run_calc(self.calc2, geom_2, self.single_element_list, chg_mult_2, "State2", iter_idx)
+
+        if ex1 or ex2:
+            state.exit_flag = True
+            return state
+
+        h1 = self.calc1.Model_hess
+        h2 = self.calc2.Model_hess
+
+        # --- 3. Compute Model Function Energy, Gradient, Hessian ---
+        if self.is_bitss:
+            mf_E = self.mf_instance.calc_energy(e1, e2, geom_1, geom_2, g1, g2, iter_idx)
+            mf_G1, mf_G2 = self.mf_instance.calc_grad(e1, e2, geom_1, geom_2, g1, g2)
+            mf_G = np.vstack((np.array(mf_G1), np.array(mf_G2))).astype(np.float64)
+            
+            if hasattr(self.mf_instance, "calc_hess"):
+                 try:
+                     raw_H = self.mf_instance.calc_hess(e1, e2, g1, g2, h1, h2)
+                     if raw_H is not None:
+                         mf_H = raw_H
+                     else:
+                         mf_H = self._make_block_diag_hess(h1, h2)
+                 except:
+                     mf_H = self._make_block_diag_hess(h1, h2)
+            else:
+                 mf_H = self._make_block_diag_hess(h1, h2)
 
-    def _save_input_data(self):
-        with open(self.BPA_FOLDER_DIRECTORY+"input.txt", "w") as f:
-            f.write(str(vars(self.config.args))) # Read from config
-        return
+            self.bitss_geom1_history.append(geom_1 * self.config.bohr2angstroms)
+            self.bitss_geom2_history.append(geom_2 * self.config.bohr2angstroms)
 
-    def _constrain_flag_check(self, force_data):
-        # (This method is pure, no changes needed)
-        if len(force_data["projection_constraint_condition_list"]) > 0:
-            projection_constrain = True
         else:
-            projection_constrain = False
+            # Standard Mode (3N)
+            mf_E = self.mf_instance.calc_energy(e1, e2)
             
-        if len(force_data["fix_atoms"]) == 0:
-            allactive_flag = True
-        else:
-            allactive_flag = False
+            raw_output = self.mf_instance.calc_grad(e1, e2, g1, g2)
+            if isinstance(raw_output, (tuple, list)):
+                raw_G = np.array(raw_output[0]).astype(np.float64)
+            else:
+                raw_G = np.array(raw_output).astype(np.float64)
 
-        if "x" in force_data["projection_constraint_condition_list"] or "y" in force_data["projection_constraint_condition_list"] or "z" in force_data["projection_constraint_condition_list"]:
-            allactive_flag = False
+            if raw_G.ndim != 2:
+                 if raw_G.size == len(self.single_element_list) * 3:
+                     mf_G = raw_G.reshape(len(self.single_element_list), 3)
+                 else:
+                     mf_G = raw_G
+            else:
+                 mf_G = raw_G
+
+            mf_H = None
+            if hasattr(self.mf_instance, "calc_hess"):
+                try:
+                    raw_H = self._call_calc_hess_safely(self.mf_instance, e1, e2, g1, g2, h1, h2)
+                    if raw_H is not None:
+                        if isinstance(raw_H, (tuple, list)):
+                            mf_H = raw_H[0]
+                        else:
+                            mf_H = raw_H
+                except Exception as e:
+                    print(f"Note: calc_hess failed or not applicable ({e}), falling back to average.")
+            
+            if mf_H is None:
+                mf_H = 0.5 * (h1 + h2)
+
+        # --- 4. Apply Bias Potential ---
+        if self.is_bitss:
+            _, be1, bg1, bh1 = self.bias_pot_calc.main(
+                0.0, g1 * 0.0, geom_1, self.single_element_list, self.force_data, 
+                state.pre_bias_gradient[:len(geom_1)] if state.pre_bias_gradient is not None else None,
+                iter_idx
+            )
+            _, be2, bg2, bh2 = self.bias_pot_calc.main(
+                0.0, g2 * 0.0, geom_2, self.single_element_list, self.force_data,
+                state.pre_bias_gradient[len(geom_1):] if state.pre_bias_gradient is not None else None,
+                iter_idx
+            )
+            
+            final_E = mf_E + be1 + be2
+            final_G = mf_G + np.vstack((bg1, bg2))
+            bias_H = self._make_block_diag_hess(bh1, bh2)
+            
+        else:
+            _, final_E, final_G, bias_H = self.bias_pot_calc.main(
+                mf_E, mf_G, state.geometry, self.single_element_list, self.force_data,
+                state.pre_bias_gradient, iter_idx
+            )
+
+        # --- 5. Update State ---
+        state.raw_energy = mf_E
+        state.raw_gradient = mf_G
+        state.bias_energy = final_E
+        state.bias_gradient = final_G
+        state.effective_gradient = final_G
+        
+        state.energies["raw"] = mf_E
+        state.energies["effective"] = final_E
+        state.gradients["raw"] = mf_G
+        state.gradients["effective"] = final_G
+        
+        state.Model_hess = mf_H 
+        state.bias_hessian = bias_H
+        
+        return state
+
+    def _make_block_diag_hess(self, h1, h2):
+        d1 = h1.shape[0]
+        d2 = h2.shape[0]
+        full_H = np.zeros((d1 + d2, d1 + d2))
+        full_H[:d1, :d1] = h1
+        full_H[d1:, d1:] = h2
+        return full_H
+
+    def _call_calc_hess_safely(self, instance, e1, e2, g1, g2, h1, h2):
+        sig = inspect.signature(instance.calc_hess)
+        params = sig.parameters
+        if len(params) == 2:
+            return instance.calc_hess(h1, h2)
+        elif len(params) == 4:
+            return instance.calc_hess(g1, g2, h1, h2)
+        else:
+            return instance.calc_hess(e1, e2, g1, g2, h1, h2)
+
+    def _run_calc(self, calc_inst, geom, elems, chg_mult, label, iter_idx):
+        run_dir = os.path.join(self.base_dir, label, f"iter{iter_idx}")
+        os.makedirs(run_dir, exist_ok=True)
+        old_dir = calc_inst.BPA_FOLDER_DIRECTORY
+        calc_inst.BPA_FOLDER_DIRECTORY = run_dir
+        
+        # Charge/Multiplicity update for PySCF compatibility
+        calc_inst.electronic_charge = chg_mult[0]
+        calc_inst.spin_multiplicity = chg_mult[1]
+
+        geom_str = self.file_io.print_geometry_list(geom * self.config.bohr2angstroms, elems, chg_mult, display_flag=True)
+        inp_path = self.file_io.make_psi4_input_file(geom_str, iter_idx, path=run_dir)
+        
+        # Method string for xTB
+        method_str = getattr(calc_inst, "xtb_method", "")
+        if method_str is None:
+            method_str = ""
+
+        # [FIX] Convert list to numpy array (int) to avoid 'list has no attribute tolist' error in tblite tools
+        atom_nums = np.array([element_number(el) for el in elems], dtype=int)
+
+        e, g, _, ex = calc_inst.single_point(
+            inp_path, 
+            atom_nums,  # Passing numpy array instead of list
+            iter_idx, 
+            chg_mult, 
+            method=method_str
+        )
         
-        return projection_constrain, allactive_flag
+        calc_inst.BPA_FOLDER_DIRECTORY = old_dir
+        return e, g, ex
+
+    def finalize_bitss_trajectory(self):
+        if not self.is_bitss or not self.bitss_geom1_history: return
+        filename = os.path.join(self.base_dir, f"{self.file_io.NOEXT_START_FILE}_traj.xyz")
+        full_seq = self.bitss_geom1_history + self.bitss_geom2_history[::-1]
+        with open(filename, 'w') as f:
+            for s, g in enumerate(full_seq):
+                f.write(f"{len(g)}\nBITSS_Step {s}\n")
+                for i, atom in enumerate(g):
+                    f.write(f"{self.single_element_list[i]:2s} {atom[0]:12.8f} {atom[1]:12.8f} {atom[2]:12.8f}\n")
+  
+class ONIOMHandler(BasePotentialHandler):
+    """
+    Handles ONIOM calculations with microiterations.
+    """
 
-    def _init_projection_constraint(self, PC, geom_num_list, iter, projection_constrain, hessian=None):
-        # (This method is pure, no changes needed)
-        if iter == 0:
-            if projection_constrain:
-                PC.initialize(geom_num_list, hessian=hessian)
+    def __init__(
+        self,
+        high_calc,
+        low_calc,
+        high_atoms,
+        link_atoms,
+        *args,
+        **kwargs,
+    ):
+        super().__init__(*args, **kwargs)
+        self.high_calc = high_calc
+        self.low_calc = low_calc
+        self.high_atoms = high_atoms
+        self.link_atoms = link_atoms
+
+    def compute(self, state: OptimizationState):
+        raise "Not Implemented!"
+        # This handler runs one ONIOM iteration including microiterations.
+        # The logic mirrors optimize_oniom from the legacy code.
+        force_data = self.force_data
+        config = self.config
+        bohr2angstroms = config.bohr2angstroms
+
+        # Build mappings
+        linker_atom_pair_num = specify_link_atom_pairs(
+            state.geometry, state.element_list, self.high_atoms, self.link_atoms
+        )
+        real_2_highlayer_label_connect_dict, highlayer_2_real_label_connect_dict = (
+            link_number_high_layer_and_low_layer(self.high_atoms)
+        )
+
+        # Separate layers
+        high_layer_geom_num_list, high_layer_element_list = (
+            separate_high_layer_and_low_layer(
+                state.geometry, linker_atom_pair_num, self.high_atoms, state.element_list
+            )
+        )
+
+        # Prepare Hessians
+        LL_Model_hess = copy.deepcopy(state.Model_hess)
+        HL_Model_hess = np.eye(len(high_layer_element_list) * 3)
+
+        # Bias calculators
+        LL_Calc_BiasPot = BiasPotentialCalculation(self.base_dir)
+        CalcBiaspot_real = self.bias_pot_calc  # Use parent
+
+        # Previous vars
+        pre_model_HL_B_e = 0.0
+        pre_model_HL_B_g = np.zeros((len(high_layer_element_list), 3))
+        pre_model_HL_g = np.zeros((len(high_layer_element_list), 3))
+        pre_real_LL_B_e = state.pre_bias_energy
+        pre_real_LL_e = state.pre_raw_energy
+        pre_real_LL_B_g = copy.deepcopy(state.pre_bias_gradient)
+        pre_real_LL_g = copy.deepcopy(state.pre_raw_gradient)
+        pre_real_LL_move_vector = copy.deepcopy(state.pre_move_vector)
+        pre_model_HL_move_vector = np.zeros((len(high_layer_element_list), 3))
+
+        # High-layer optimizer
+        HL_CMV = CalculateMoveVector(
+            config.DELTA,
+            high_layer_element_list[: len(self.high_atoms)],
+            config.args.saddle_order,
+            config.FC_COUNT,
+            config.temperature,
+            max_trust_radius=config.max_trust_radius,
+            min_trust_radius=config.min_trust_radius,
+        )
+        HL_optimizer_instances = HL_CMV.initialization(force_data["opt_method"])
+        for inst in HL_optimizer_instances:
+            inst.set_hessian(
+                HL_Model_hess[: len(self.high_atoms) * 3, : len(self.high_atoms) * 3]
+            )
+            if config.DELTA != "x":
+                inst.DELTA = config.DELTA
+
+        # Model low-layer calc
+        self.low_calc.Model_hess = LL_Model_hess
+        model_LL_e, model_LL_g, high_layer_geom_num_list, finish_frag = (
+            self.low_calc.single_point(
+                self.file_io.make_psi4_input_file(
+                    self.file_io.print_geometry_list(
+                        state.geometry * bohr2angstroms,
+                        state.element_list,
+                        config.electric_charge_and_multiplicity,
+                    ),
+                    state.iter,
+                ),
+                high_layer_element_list,
+                state.iter,
+                config.electric_charge_and_multiplicity,
+                force_data["oniom_flag"][2],
+                geom_num_list=high_layer_geom_num_list * bohr2angstroms,
+            )
+        )
+        if finish_frag:
+            state.exit_flag = True
+            return state
+
+        # Microiterations on low layer
+        LL_CMV = CalculateMoveVector(
+            config.DELTA,
+            state.element_list,
+            config.args.saddle_order,
+            config.FC_COUNT,
+            config.temperature,
+        )
+        LL_optimizer_instances = LL_CMV.initialization(["fire"])
+        LL_optimizer_instances[0].display_flag = False
+
+        current_geom_num_list = copy.deepcopy(state.geometry)
+        real_initial_geom_num_list = copy.deepcopy(state.geometry)
+        real_pre_geom = copy.deepcopy(state.geometry)
+
+        low_layer_converged = False
+        for microiter in range(config.microiter_num):
+            self.low_calc.Model_hess = LL_Model_hess
+            real_LL_e, real_LL_g, current_geom_num_list, finish_frag = (
+                self.low_calc.single_point(
+                    self.file_io.make_psi4_input_file(
+                        self.file_io.print_geometry_list(
+                            current_geom_num_list * bohr2angstroms,
+                            state.element_list,
+                            config.electric_charge_and_multiplicity,
+                            display_flag=False,
+                        ),
+                        microiter,
+                    ),
+                    state.element_list,
+                    microiter,
+                    config.electric_charge_and_multiplicity,
+                    force_data["oniom_flag"][2],
+                    geom_num_list=current_geom_num_list * bohr2angstroms,
+                )
+            )
+            LL_Model_hess = copy.deepcopy(self.low_calc.Model_hess)
+            LL_Calc_BiasPot.Model_hess = LL_Model_hess
+            _, real_LL_B_e, real_LL_B_g, LL_BPA_hessian = LL_Calc_BiasPot.main(
+                real_LL_e,
+                real_LL_g,
+                current_geom_num_list,
+                state.element_list,
+                force_data,
+                pre_real_LL_B_g,
+                microiter,
+                real_initial_geom_num_list,
+            )
+
+            for inst in LL_optimizer_instances:
+                inst.set_bias_hessian(LL_BPA_hessian)
+                if microiter % config.FC_COUNT == 0:
+                    inst.set_hessian(LL_Model_hess)
+
+            if len(force_data["opt_fragment"]) > 0:
+                real_LL_B_g = self._calc_fragment_grads(
+                    real_LL_B_g, force_data["opt_fragment"]
+                )
+                real_LL_g = self._calc_fragment_grads(
+                    real_LL_g, force_data["opt_fragment"]
+                )
+
+            prev_geom = current_geom_num_list.copy()
+            current_geom_num_list_ang, LL_move_vector, LL_optimizer_instances = (
+                LL_CMV.calc_move_vector(
+                    microiter,
+                    current_geom_num_list,
+                    real_LL_B_g,
+                    pre_real_LL_B_g,
+                    real_pre_geom,
+                    real_LL_B_e,
+                    pre_real_LL_B_e,
+                    pre_real_LL_move_vector,
+                    real_initial_geom_num_list,
+                    real_LL_g,
+                    pre_real_LL_g,
+                    LL_optimizer_instances,
+                    print_flag=False,
+                )
+            )
+            current_geom_num_list = current_geom_num_list_ang / bohr2angstroms
+
+            # Fix high-layer atoms
+            for key, value in highlayer_2_real_label_connect_dict.items():
+                current_geom_num_list[value - 1] = copy.deepcopy(
+                    high_layer_geom_num_list[key - 1]
+                )
+
+            # Fix user-specified atoms
+            if len(force_data["fix_atoms"]) > 0:
+                for j in force_data["fix_atoms"]:
+                    current_geom_num_list[j - 1] = copy.deepcopy(
+                        real_initial_geom_num_list[j - 1]
+                    )
+
+            displacement_vector = current_geom_num_list - prev_geom
+            low_layer_grads = []
+            low_layer_displacements = []
+            for i in range(len(state.element_list)):
+                if (i + 1) not in self.high_atoms:
+                    low_layer_grads.append(real_LL_B_g[i])
+                    low_layer_displacements.append(displacement_vector[i])
+            low_layer_grads = np.array(low_layer_grads)
+            low_layer_displacements = np.array(low_layer_displacements)
+
+            low_layer_rms_grad = self._calculate_rms_safely(low_layer_grads)
+            max_displacement = np.abs(displacement_vector).max()
+            rms_displacement = self._calculate_rms_safely(displacement_vector)
+
+            if (
+                (low_layer_rms_grad < 0.0003)
+                and (low_layer_grads.max() < 0.0006 if len(low_layer_grads) > 0 else True)
+                and (max_displacement < 0.003)
+                and (rms_displacement < 0.002)
+            ):
+                low_layer_converged = True
+                break
+
+            # Update previous for microiter
+            pre_real_LL_B_e = real_LL_B_e
+            pre_real_LL_g = real_LL_g
+            pre_real_LL_B_g = real_LL_B_g
+            pre_real_LL_move_vector = LL_move_vector
+            real_pre_geom = current_geom_num_list
+
+        # Update state geometry after microiterations
+        state.geometry = current_geom_num_list
+        geom_num_list = current_geom_num_list
+
+        # Model high-layer calc
+        self.high_calc.Model_hess = HL_Model_hess
+        model_HL_e, model_HL_g, high_layer_geom_num_list, finish_frag = (
+            self.high_calc.single_point(
+                self.file_io.make_psi4_input_file(
+                    self.file_io.print_geometry_list(
+                        geom_num_list * bohr2angstroms,
+                        state.element_list,
+                        config.electric_charge_and_multiplicity,
+                    ),
+                    state.iter,
+                ),
+                high_layer_element_list,
+                state.iter,
+                config.electric_charge_and_multiplicity,
+                method="",
+                geom_num_list=high_layer_geom_num_list * bohr2angstroms,
+            )
+        )
+        HL_Model_hess = copy.deepcopy(self.high_calc.Model_hess)
+        if finish_frag:
+            state.exit_flag = True
+            return state
+
+        # Combine gradients
+        _, tmp_model_HL_B_e, tmp_model_HL_B_g, LL_BPA_hessian = LL_Calc_BiasPot.main(
+            0.0,
+            real_LL_g * 0.0,
+            geom_num_list,
+            state.element_list,
+            force_data,
+            pre_real_LL_B_g * 0.0,
+            state.iter,
+            real_initial_geom_num_list,
+        )
+        tmp_model_HL_g = tmp_model_HL_B_g * 0.0
+        for key, value in real_2_highlayer_label_connect_dict.items():
+            tmp_model_HL_B_g[key - 1] += model_HL_g[value - 1] - model_LL_g[value - 1]
+            tmp_model_HL_g[key - 1] += model_HL_g[value - 1] - model_LL_g[value - 1]
+
+        bool_list = []
+        for i in range(len(state.element_list)):
+            if i in self.high_atoms:
+                bool_list.extend([True, True, True])
             else:
-                pass
-            return PC
-        else:
-            return PC
+                bool_list.extend([False, False, False])
 
-    def _save_init_geometry(self, geom_num_list, element_list, allactive_flag):
-        # (This method is pure, no changes needed)
-        if allactive_flag:
-            initial_geom_num_list = geom_num_list - Calculationtools().calc_center(geom_num_list, element_list)
-            pre_geom = initial_geom_num_list - Calculationtools().calc_center(geom_num_list, element_list)
+        HL_BPA_hessian = LL_BPA_hessian[np.ix_(bool_list, bool_list)]
+        for inst in HL_optimizer_instances:
+            inst.set_bias_hessian(HL_BPA_hessian)
+            if state.iter % config.FC_COUNT == 0:
+                inst.set_hessian(
+                    HL_Model_hess[: len(self.high_atoms) * 3, : len(self.high_atoms) * 3]
+                )
+
+        if len(force_data["opt_fragment"]) > 0:
+            tmp_model_HL_B_g = self._calc_fragment_grads(
+                tmp_model_HL_B_g, force_data["opt_fragment"]
+            )
+            tmp_model_HL_g = self._calc_fragment_grads(
+                tmp_model_HL_g, force_data["opt_fragment"]
+            )
+
+        model_HL_B_g = copy.deepcopy(model_HL_g)
+        model_HL_B_e = model_HL_e + tmp_model_HL_B_e
+        for key, value in real_2_highlayer_label_connect_dict.items():
+            model_HL_B_g[value - 1] += tmp_model_HL_B_g[key - 1]
+
+        pre_high_layer_geom_num_list = high_layer_geom_num_list.copy()
+        high_layer_geom_num_list_ang, move_vector, HL_optimizer_instances = (
+            HL_CMV.calc_move_vector(
+                state.iter,
+                high_layer_geom_num_list[: len(self.high_atoms)],
+                model_HL_B_g[: len(self.high_atoms)],
+                pre_model_HL_B_g[: len(self.high_atoms)],
+                pre_high_layer_geom_num_list[: len(self.high_atoms)],
+                model_HL_B_e,
+                pre_model_HL_B_e,
+                pre_model_HL_move_vector[: len(self.high_atoms)],
+                high_layer_geom_num_list[: len(self.high_atoms)],
+                model_HL_g[: len(self.high_atoms)],
+                pre_model_HL_g[: len(self.high_atoms)],
+                HL_optimizer_instances,
+            )
+        )
+        high_layer_geom_num_list = high_layer_geom_num_list_ang / bohr2angstroms
+
+        # Update full system geometry with high layer changes
+        for l in range(len(high_layer_geom_num_list) - len(linker_atom_pair_num)):
+            geom_num_list[
+                highlayer_2_real_label_connect_dict[l + 1] - 1
+            ] = copy.deepcopy(high_layer_geom_num_list[l])
+
+        geom_num_list -= Calculationtools().calc_center_of_mass(
+            geom_num_list, state.element_list
+        )
+        geom_num_list, _ = Calculationtools().kabsch_algorithm(
+            geom_num_list, real_pre_geom
+        )
+        state.geometry = geom_num_list
+        high_layer_geom_num_list, high_layer_element_list = (
+            separate_high_layer_and_low_layer(
+                geom_num_list, linker_atom_pair_num, self.high_atoms, state.element_list
+            )
+        )
+
+        # Combine energies and gradients for real system
+        real_e = real_LL_e + model_HL_e - model_LL_e
+        real_B_e = real_LL_B_e + model_HL_B_e - model_LL_e
+        real_g = real_LL_g + tmp_model_HL_g
+        real_B_g = real_LL_B_g + tmp_model_HL_g
+
+        state.raw_energy = real_e
+        state.bias_energy = real_B_e
+        state.raw_gradient = real_g
+        state.bias_gradient = real_B_g
+        state.effective_energy = real_B_e
+        state.effective_gradient = real_B_g
+        state.Model_hess = LL_Model_hess  # keep low layer hess for next step
+        state.bias_hessian = HL_BPA_hessian
+        state.energies["raw"] = real_e
+        state.energies["bias"] = real_B_e
+        state.energies["effective"] = real_B_e
+        state.gradients["raw"] = real_g
+        state.gradients["bias"] = real_B_g
+        state.gradients["effective"] = real_B_g
+
+        # Update previous caches for next iteration
+        state.pre_raw_energy = real_e
+        state.pre_bias_energy = real_B_e
+        state.pre_raw_gradient = real_g
+        state.pre_bias_gradient = real_B_g
+        state.pre_move_vector = move_vector
+        return state
+
+    @staticmethod
+    def _calc_fragment_grads(gradient, fragment_list):
+        calced_gradient = gradient
+        for fragment in fragment_list:
+            tmp_grad = np.array([0.0, 0.0, 0.0], dtype="float64")
+            for atom_num in fragment:
+                tmp_grad += gradient[atom_num - 1]
+            tmp_grad /= len(fragment)
+            for atom_num in fragment:
+                calced_gradient[atom_num - 1] = copy.deepcopy(tmp_grad)
+        return calced_gradient
+
+    @staticmethod
+    def _calculate_rms_safely(vector, threshold=1e-10):
+        filtered_vector = vector[np.abs(vector) > threshold]
+        if filtered_vector.size > 0:
+            return np.sqrt((filtered_vector**2).mean())
         else:
-            initial_geom_num_list = geom_num_list 
-            pre_geom = initial_geom_num_list 
-        
-        return initial_geom_num_list, pre_geom
+            return 0.0
 
-    def _calc_eff_hess_for_fix_atoms_and_set_hess(self, allactive_flag, force_data, BPA_hessian, n_fix, optimizer_instances, geom_num_list, B_g, g, projection_constrain, PC):
-        # (Reads self.state.Model_hess, self.config.FC_COUNT, etc.)
-        if not allactive_flag:
-            fix_num = []
-            for fnum in force_data["fix_atoms"]:
-                fix_num.extend([3*(fnum-1)+0, 3*(fnum-1)+1, 3*(fnum-1)+2])
-            fix_num = np.array(fix_num, dtype="int64")
-            #effective hessian
-            tmp_fix_hess = self.state.Model_hess[np.ix_(fix_num, fix_num)] + np.eye((3*n_fix)) * 1e-10
-            inv_tmp_fix_hess = np.linalg.pinv(tmp_fix_hess)
-            tmp_fix_bias_hess = BPA_hessian[np.ix_(fix_num, fix_num)] + np.eye((3*n_fix)) * 1e-10
-            inv_tmp_fix_bias_hess = np.linalg.pinv(tmp_fix_bias_hess)
-            BPA_hessian -= np.dot(BPA_hessian[:, fix_num], np.dot(inv_tmp_fix_bias_hess, BPA_hessian[fix_num, :]))
+class EDEELHandler(BasePotentialHandler):
+    """
+    Handles EDEEL calculations for Electron Transfer.
+    Computes V11 (Reactant) and V22 (Product) diabatic potentials using
+    energy decomposition of Complex, Donor, and Acceptor.
+    # ref.: https://doi.org/10.1039/D3RA05784D
+    # This is under constraction.
+    Target function can be switched between:
+    - 'reactant': Minimize V11
+    - 'product':  Minimize V22
+    - 'sx':       Minimize Seam of Crossing penalty function (Default)
+    """
+    def __init__(self, complex_calc, donor_atoms, acceptor_atoms, ede_params, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.complex_calc = complex_calc
+        self.donor_atoms = np.array(donor_atoms)
+        self.acceptor_atoms = np.array(acceptor_atoms)
+        
+        # Charges and Multiplicities for all 5 states
+        # Expected ede_params keys: 
+        # 'complex': [chg, mult], 'd_ox': [chg, mult], 'd_red': [chg, mult], ...
+        self.params = ede_params
+        
+        # SX penalty weight (sigma) and target mode
+        self.sigma = kwargs.get('sigma', 2.0)
+        self.target_mode = kwargs.get('target_mode', 'sx') 
+
+        # Setup directories for components to avoid file collision
+        self.dirs = {
+            'complex': self.base_dir, # Complex runs in root
+            'd_ox': os.path.join(self.base_dir, "Components", "Donor_Ox"),
+            'd_red': os.path.join(self.base_dir, "Components", "Donor_Red"),
+            'a_ox': os.path.join(self.base_dir, "Components", "Acceptor_Ox"),
+            'a_red': os.path.join(self.base_dir, "Components", "Acceptor_Red"),
+        }
+        for d in self.dirs.values():
+            os.makedirs(d, exist_ok=True)
+
+    def compute(self, state: OptimizationState):
+        config = self.config
+        full_geom = state.geometry
+        
+        # 1. Extract Fragment Geometries
+        # Note: Atoms indices are 1-based, numpy is 0-based
+        d_indices = self.donor_atoms - 1
+        a_indices = self.acceptor_atoms - 1
+        
+        geom_d = full_geom[d_indices]
+        geom_a = full_geom[a_indices]
+        
+        # Helper to run calculation for one component
+        def run_component(label, calc_geom, atom_indices, chg_mult, target_dir):
+            # Prepare file IO for this component
+            # Note: We create a specific input file in the target directory
+            input_content = self.file_io.print_geometry_list(
+                calc_geom * config.bohr2angstroms,
+                [state.element_list[i] for i in atom_indices], 
+                chg_mult,
+                display_flag=False
+            )
+            
+            # Temporarily switch directory in calculator
+            original_dir = self.complex_calc.BPA_FOLDER_DIRECTORY
+            self.complex_calc.BPA_FOLDER_DIRECTORY = target_dir
+            
+            # Run Single Point
+            # Assuming make_psi4_input_file can accept a 'path' argument or we construct the path manually
+            # Here we assume the file_io or calculator handles the path based on BPA_FOLDER_DIRECTORY
+            e, g, _, exit_flag = self.complex_calc.single_point(
+                self.file_io.make_psi4_input_file(input_content, state.iter, path=target_dir),
+                [state.element_list[i] for i in atom_indices], 
+                state.iter,
+                chg_mult,
+                method="" 
+            )
+            
+            # Restore directory
+            self.complex_calc.BPA_FOLDER_DIRECTORY = original_dir
+            
+            return e, g, exit_flag
+
+        # 2. Run 5 Calculations
+        # A. Complex (n+m)
+        e_c, g_c, exit_c = run_component(
+            'complex', full_geom, range(len(full_geom)), 
+            self.params['complex'], self.dirs['complex']
+        )
         
-        for i in range(len(optimizer_instances)):
-                
-            if projection_constrain:
-                if np.all(np.abs(BPA_hessian) < 1e-20):
-                    proj_bpa_hess = PC.calc_project_out_hess(geom_num_list, B_g - g, BPA_hessian)
-                else:
-                    proj_bpa_hess = BPA_hessian
-                optimizer_instances[i].set_bias_hessian(proj_bpa_hess)
-            else:
-                optimizer_instances[i].set_bias_hessian(BPA_hessian)
-            
-            if self.state.iter % self.config.FC_COUNT == 0 or (self.config.use_model_hessian is not None and self.state.iter % self.config.mFC_COUNT == 0):
+        # B. Donor Ox (n)
+        e_do, g_do, exit_do = run_component(
+            'd_ox', geom_d, d_indices, 
+            self.params['d_ox'], self.dirs['d_ox']
+        )
         
-                if not allactive_flag:
-                    self.state.Model_hess -= np.dot(self.state.Model_hess[:, fix_num], np.dot(inv_tmp_fix_hess, self.state.Model_hess[fix_num, :]))
-                
-                
-                if projection_constrain:
-                    proj_model_hess = PC.calc_project_out_hess(geom_num_list, g, self.state.Model_hess)
-                    optimizer_instances[i].set_hessian(proj_model_hess)
-                else:
-                    optimizer_instances[i].set_hessian(self.state.Model_hess)
+        # C. Donor Red (n+1)
+        e_dr, g_dr, exit_dr = run_component(
+            'd_red', geom_d, d_indices, 
+            self.params['d_red'], self.dirs['d_red']
+        )
         
-        return optimizer_instances
+        # D. Acceptor Ox (m)
+        e_ao, g_ao, exit_ao = run_component(
+            'a_ox', geom_a, a_indices, 
+            self.params['a_ox'], self.dirs['a_ox']
+        )
         
-    def _apply_projection_constraints(self, projection_constrain, PC, geom_num_list, g, B_g):
-        # (This method is pure, no changes needed)
+        # E. Acceptor Red (m+1)
+        e_ar, g_ar, exit_ar = run_component(
+            'a_red', geom_a, a_indices, 
+            self.params['a_red'], self.dirs['a_red']
+        )
+
+        if any([exit_c, exit_do, exit_dr, exit_ao, exit_ar]):
+            state.exit_flag = True
+            return state
+
+        # 3. Construct V11 and V22
+        # V11 (Initial State) = E_Complex - E_Donor_Ox + E_Donor_Red
+        # V22 (Final State)   = E_Complex - E_Acceptor_Ox + E_Acceptor_Red
+        
+        V11 = e_c - e_do + e_dr
+        V22 = e_c - e_ao + e_ar
+        
+        state.energies['V11'] = V11
+        state.energies['V22'] = V22
+        state.energies['gap'] = V11 - V22
+        
+        # 4. Construct Gradients
+        # Helper to map fragment gradients back to full system dimensions
+        def map_grad(g_frag, indices):
+            g_full = np.zeros_like(g_c)
+            for i, idx in enumerate(indices):
+                g_full[idx] = g_frag[i]
+            return g_full
+
+        grad_V11 = g_c - map_grad(g_do, d_indices) + map_grad(g_dr, d_indices)
+        grad_V22 = g_c - map_grad(g_ao, a_indices) + map_grad(g_ar, a_indices)
+        
+        state.gradients['V11'] = grad_V11
+        state.gradients['V22'] = grad_V22
+
+        # 5. Determine Effective Energy/Gradient for Optimizer
+        if self.target_mode == 'reactant':
+            eff_E = V11
+            eff_G = grad_V11
+        elif self.target_mode == 'product':
+            eff_E = V22
+            eff_G = grad_V22
+        else: # 'sx' (Seam of Crossing Search)
+            # Penalty function approach: L = mean(V) + sigma * (V1 - V2)^2
+            mean_V = 0.5 * (V11 + V22)
+            diff_V = V11 - V22
+            penalty = self.sigma * (diff_V ** 2)
+            
+            eff_E = mean_V + penalty
+            
+            # Gradient of L: dL/dX = 0.5(g1 + g2) + 2*sigma*diff_V * (g1 - g2)
+            eff_G = 0.5 * (grad_V11 + grad_V22) + 2 * self.sigma * diff_V * (grad_V11 - grad_V22)
+            
+            state.energies['penalty'] = penalty
+
+        # Store results
+        state.raw_energy = eff_E
+        state.raw_gradient = eff_G
+        
+        # Bias processing (if any)
+        self._add_bias_and_update_state(state, eff_E, eff_G)
+        
+        return state
+
+# =====================================================================================
+# 4. Managers for constraints, convergence, Hessian, logging, result paths
+# =====================================================================================
+class ConstraintManager:
+    def __init__(self, config):
+        self.config = config
+
+    def constrain_flag_check(self, force_data):
+        
+        projection_constrain = len(force_data["projection_constraint_condition_list"]) > 0 and any(s.lower() == "crsirfo" for s in self.config.args.opt_method)
+     
+        allactive_flag = len(force_data["fix_atoms"]) == 0
+        if (
+            "x" in force_data["projection_constraint_condition_list"]
+            or "y" in force_data["projection_constraint_condition_list"]
+            or "z" in force_data["projection_constraint_condition_list"]
+        ):
+            allactive_flag = False
+        return projection_constrain, allactive_flag
+
+    def init_projection_constraint(self, PC, geom_num_list, iter_idx, projection_constrain, hessian=None):
+        if iter_idx == 0:
+            if projection_constrain:
+                PC.initialize(geom_num_list, hessian=hessian)
+            return PC
+        else:
+            return PC
+
+    def apply_projection_constraints(self, projection_constrain, PC, geom_num_list, g, B_g):
         if projection_constrain:
             g = copy.deepcopy(PC.calc_project_out_grad(geom_num_list, g))
             proj_d_B_g = copy.deepcopy(PC.calc_project_out_grad(geom_num_list, B_g - g))
             B_g = copy.deepcopy(g + proj_d_B_g)
-        
+            print("Projection was applied to gradient.")
         return g, B_g, PC
 
-    def _zero_fixed_atom_gradients(self, allactive_flag, force_data, g, B_g):
-        # (This method is pure, no changes needed)
+    def apply_projection_constraints_to_geometry(self, projection_constrain, PC, new_geometry, hessian=None):
+        if projection_constrain:
+            tmp_new_geometry = new_geometry / self.config.bohr2angstroms
+            adjusted_geometry = (
+                PC.adjust_init_coord(tmp_new_geometry, hessian=hessian)
+                * self.config.bohr2angstroms
+            )
+            return adjusted_geometry, PC
+        return new_geometry, PC
+
+    def zero_fixed_atom_gradients(self, allactive_flag, force_data, g, B_g):
         if not allactive_flag:
             for j in force_data["fix_atoms"]:
-                g[j-1] = copy.deepcopy(g[j-1]*0.0)
-                B_g[j-1] = copy.deepcopy(B_g[j-1]*0.0)
-        
+                g[j - 1] = copy.deepcopy(g[j - 1] * 0.0)
+                B_g[j - 1] = copy.deepcopy(B_g[j - 1] * 0.0)
         return g, B_g
 
-    def _project_out_translation_rotation(self, new_geometry, geom_num_list, allactive_flag):
-        # (Reads self.config.bohr2angstroms)
+    def project_out_translation_rotation(self, new_geometry, geom_num_list, allactive_flag):
         if allactive_flag:
-            # Convert to Bohr, apply Kabsch alignment algorithm, then convert back
             aligned_geometry, _ = Calculationtools().kabsch_algorithm(
-                new_geometry/self.config.bohr2angstroms, geom_num_list)
+                new_geometry / self.config.bohr2angstroms, geom_num_list
+            )
             aligned_geometry *= self.config.bohr2angstroms
             return aligned_geometry
         else:
-            # If not all atoms are active, return the original geometry
             return new_geometry
 
-    def _apply_projection_constraints_to_geometry(self, projection_constrain, PC, new_geometry, hessian=None):
-        # (Reads self.config.bohr2angstroms)
-        if projection_constrain:
-            tmp_new_geometry = new_geometry / self.config.bohr2angstroms
-            adjusted_geometry = PC.adjust_init_coord(tmp_new_geometry, hessian=hessian) * self.config.bohr2angstroms
-            return adjusted_geometry, PC
-        
-        return new_geometry, PC
-
-    def _reset_fixed_atom_positions(self, new_geometry, initial_geom_num_list, allactive_flag, force_data):
-        # (Reads self.config.bohr2angstroms)
+    def reset_fixed_atom_positions(
+        self, new_geometry, initial_geom_num_list, allactive_flag, force_data
+    ):
         if not allactive_flag:
             for j in force_data["fix_atoms"]:
-                new_geometry[j-1] = copy.deepcopy(initial_geom_num_list[j-1]*self.config.bohr2angstroms)
-        
+                new_geometry[j - 1] = copy.deepcopy(
+                    initial_geom_num_list[j - 1] * self.config.bohr2angstroms
+                )
         return new_geometry
 
-    def _initialize_optimization_tools(self, FIO, force_data):
-        """
-        Initializes all tools needed for the optimization loop.
-        This replaces the old _initialize_optimization_variables.
-        It assumes self.state is already created.
-        """
-        # Load modules
-        Calculation, xtb_method = self._import_calculation_module()
-        self._save_input_data() # Save input.txt
-        G = Graph(self.BPA_FOLDER_DIRECTORY)
-        
-        # Get atom info
-        file_directory, electric_charge_and_multiplicity, element_list = self.write_input_files(FIO)
-        self.element_list = element_list # Store on self for helper methods
-        self.state.element_list = element_list # Store in state
-        
-        element_number_list = np.array([element_number(elem) for elem in element_list], dtype="int")
-        natom = len(element_list)
-        
-        # Constraint setup
-        PC = ProjectOutConstrain(force_data["projection_constraint_condition_list"], 
-                                  force_data["projection_constraint_atoms"], 
-                                  force_data["projection_constraint_constant"])
-        projection_constrain, allactive_flag = self._constrain_flag_check(force_data)
-        n_fix = len(force_data["fix_atoms"])
+    @staticmethod
+    def calc_fragment_grads(gradient, fragment_list):
+        calced_gradient = gradient
+        for fragment in fragment_list:
+            tmp_grad = np.array([0.0, 0.0, 0.0], dtype="float64")
+            for atom_num in fragment:
+                tmp_grad += gradient[atom_num - 1]
+            tmp_grad /= len(fragment)
+            for atom_num in fragment:
+                calced_gradient[atom_num - 1] = copy.deepcopy(tmp_grad)
+        return calced_gradient
 
-        # Bias potential and calculation setup
-        self.CalcBiaspot = BiasPotentialCalculation(self.BPA_FOLDER_DIRECTORY)
-        self.SP = self.setup_calculation(Calculation) # SP is self.SP
-        
-        # Move vector calculation
-        CMV = CalculateMoveVector(self.config.DELTA, element_list, self.config.args.saddle_order, 
-                                  self.config.FC_COUNT, self.config.temperature, self.config.use_model_hessian, 
-                                  max_trust_radius=self.config.max_trust_radius, min_trust_radius=self.config.min_trust_radius)
-        optimizer_instances = CMV.initialization(force_data["opt_method"])
-        
-        # Check optimizer compatibility
-        for i in range(len(optimizer_instances)):
-            if CMV.newton_tag[i] is False and self.config.FC_COUNT > 0 and not "eigvec" in force_data["projection_constraint_condition_list"]:
-                print("Error: This optimizer method does not support exact Hessian calculations.")
-                print("Please either choose a different optimizer or set FC_COUNT=0 to disable exact Hessian calculations.")
-                sys.exit(0)
-        
-        # Initialize optimizer instances
-        for i in range(len(optimizer_instances)):
-            optimizer_instances[i].set_hessian(self.state.Model_hess) # From state
-            if self.config.DELTA != "x":
-                optimizer_instances[i].DELTA = self.config.DELTA
-                
-        if self.config.koopman_analysis:
-            KA = KoopmanAnalyzer(natom, file_directory=self.BPA_FOLDER_DIRECTORY)
+
+class ConvergenceChecker:
+    def __init__(self, config):
+        self.config = config
+
+    @staticmethod
+    def calculate_rms_safely(vector, threshold=1e-10):
+        filtered_vector = vector[np.abs(vector) > threshold]
+        if filtered_vector.size > 0:
+            return np.sqrt((filtered_vector**2).mean())
         else:
-            KA = None
+            return 0.0
 
-        # Pack and return all initialized tools
-        tools = {
-            'Calculation': Calculation, 'xtb_method': xtb_method,
-            'SP': self.SP, 'CMV': CMV, 'optimizer_instances': optimizer_instances,
-            'FIO': FIO, 'G': G, 'file_directory': file_directory,
-            'element_number_list': element_number_list, 'natom': natom,
-            'electric_charge_and_multiplicity': electric_charge_and_multiplicity,
-            'PC': PC, 'projection_constrain': projection_constrain,
-            'allactive_flag': allactive_flag, 'force_data': force_data, 'n_fix': n_fix,
-            'KA': KA
-        }
-        return tools
-
-    def check_negative_eigenvalues(self, geom_num_list, hessian):
-        # (This method is pure, no changes needed)
-        proj_hessian = Calculationtools().project_out_hess_tr_and_rot_for_coord(hessian, geom_num_list, geom_num_list, display_eigval=False)
-        if proj_hessian is not None:
-            eigvals = np.linalg.eigvalsh(proj_hessian)
-            if np.any(eigvals < -1e-10):
-                print("Notice: Negative eigenvalues detected.")
-                return True
-        return False
+    def check_convergence(self, state, displacement_vector, optimizer_instances):
+        max_force = np.abs(state.effective_gradient).max()
+        rms_force = self.calculate_rms_safely(state.effective_gradient)
 
-    def judge_early_stop_due_to_no_negative_eigenvalues(self, geom_num_list, hessian):
-        # (Reads self.config)
-        if self.config.detect_negative_eigenvalues and self.config.FC_COUNT > 0:
-            negative_eigenvalues_detected = self.check_negative_eigenvalues(geom_num_list, hessian)
-            if not negative_eigenvalues_detected and self.config.args.saddle_order > 0:
-                print("No negative eigenvalues detected while saddle_order > 0. Stopping optimization.")
-                with open(self.BPA_FOLDER_DIRECTORY+"no_negative_eigenvalues_detected.txt", "w") as f:
-                    f.write("No negative eigenvalues detected while saddle_order > 0. Stopping optimization.")
-                return True
-        return False
+        delta_max_force_threshold = max(
+            0.0, self.config.MAX_FORCE_THRESHOLD - 1 * max_force
+        )
+        delta_rms_force_threshold = max(
+            0.0, self.config.RMS_FORCE_THRESHOLD - 1 * rms_force
+        )
 
-    def optimize(self):
-        # 1. Initialize State.
-        #    write_input_files needs FIO, FIO needs BPA_FOLDER_DIRECTORY.
-        #    This is complex. Let's initialize FIO and element_list first.
-        FIO = FileIO(self.BPA_FOLDER_DIRECTORY, self.START_FILE)
-        
-        # This will read the file and set self.element_list
-        file_directory, electric_charge_and_multiplicity, element_list = self.write_input_files(FIO)
-        self.element_list = element_list
+        max_displacement = np.abs(displacement_vector).max()
+        rms_displacement = self.calculate_rms_safely(displacement_vector)
+
+        max_displacement_threshold = max(
+            self.config.MAX_DISPLACEMENT_THRESHOLD,
+            self.config.MAX_DISPLACEMENT_THRESHOLD + delta_max_force_threshold,
+        )
+        rms_displacement_threshold = max(
+            self.config.RMS_DISPLACEMENT_THRESHOLD,
+            self.config.RMS_DISPLACEMENT_THRESHOLD + delta_rms_force_threshold,
+        )
+
+        if (
+            max_force < self.config.MAX_FORCE_THRESHOLD
+            and rms_force < self.config.RMS_FORCE_THRESHOLD
+            and max_displacement < max_displacement_threshold
+            and rms_displacement < rms_displacement_threshold
+        ):
+            return True, max_displacement_threshold, rms_displacement_threshold
         
-        # Now we can create the State
-        self.state = OptimizationState(element_list)
-        self.state.cos_list = [[] for i in range(len(force_data_parser(self.config.args)["geom_info"]))] # Init cos_list
+        for opt in optimizer_instances:
+            if getattr(opt, 'proj_grad_converged', False):
+                return True, max_displacement_threshold, rms_displacement_threshold
+                
         
-        # 2. Initialize all other tools, passing FIO
-        force_data = force_data_parser(self.config.args)
-        tools = self._initialize_optimization_tools(FIO, force_data)
-
-        # 3. Unpack tools into local variables for the loop
-        # (This is better than the giant vars_dict at the end)
-        xtb_method = tools['xtb_method']
-        SP = tools['SP']
-        CMV = tools['CMV']
-        optimizer_instances = tools['optimizer_instances']
-        FIO = tools['FIO']
-        G = tools['G']
-        file_directory = tools['file_directory']
-        element_number_list = tools['element_number_list']
-        electric_charge_and_multiplicity = tools['electric_charge_and_multiplicity']
-        PC = tools['PC']
-        projection_constrain = tools['projection_constrain']
-        allactive_flag = tools['allactive_flag']
-        force_data = tools['force_data']
-        n_fix = tools['n_fix']
-        KA = tools['KA']
         
-        # 4. Main Optimization Loop
-        for iter in range(self.config.NSTEP):
-            
-            self.state.iter = iter
-            self.state.exit_flag = os.path.exists(self.BPA_FOLDER_DIRECTORY+"end.txt")
-            if self.state.exit_flag:
-                break
-            
-            self.state.exit_flag = judge_shape_condition(self.state.geom_num_list, self.config.shape_conditions)
-            if self.state.exit_flag:
-                break
-                
-            print(f"\n# ITR. {iter}\n")
-            
-            # --- Perform Single Point Calculation ---
-            SP.Model_hess = copy.deepcopy(self.state.Model_hess)
-            e, g, geom_num_list, exit_flag = SP.single_point(file_directory, element_number_list, iter, electric_charge_and_multiplicity, xtb_method)
-            
-            # Update state
-            self.state.e = e
-            self.state.g = g
-            self.state.geom_num_list = geom_num_list
-            self.state.exit_flag = exit_flag
-            self.state.Model_hess = copy.deepcopy(SP.Model_hess)
-            
-            if self.state.exit_flag:
-                break
-                
-            # --- Update Model Hessian (if needed) ---
-            if iter % self.config.mFC_COUNT == 0 and self.config.use_model_hessian is not None and self.config.FC_COUNT < 1:
-                SP.Model_hess = ApproxHessian().main(geom_num_list, self.element_list, g, self.config.use_model_hessian)
-                self.state.Model_hess = SP.Model_hess 
-                
-            if iter == 0:
-                initial_geom_num_list, pre_geom = self._save_init_geometry(geom_num_list, self.element_list, allactive_flag)
-                # Save initial geometry to state
-                self.state.pre_geom = pre_geom
-
-            # --- Bias Potential Calculation ---
-            _, B_e, B_g, BPA_hessian = self.CalcBiaspot.main(e, g, geom_num_list, self.element_list, force_data, self.state.pre_B_g, iter, initial_geom_num_list)
-            # Update state
-            self.state.B_e = B_e
-            self.state.B_g = B_g
-
-            # --- Check Eigenvalues (if first iter) ---
-            Hess = BPA_hessian + self.state.Model_hess
-            if iter == 0:
-                if self.judge_early_stop_due_to_no_negative_eigenvalues(geom_num_list, Hess):
-                    break
-            
-            # --- Constraints ---
-            PC = self._init_projection_constraint(PC, geom_num_list, iter, projection_constrain, hessian=Hess)
-            optimizer_instances = self._calc_eff_hess_for_fix_atoms_and_set_hess(allactive_flag, force_data, BPA_hessian, n_fix, optimizer_instances, geom_num_list, B_g, g, projection_constrain, PC)
-            
-            if not allactive_flag:
-                B_g = copy.deepcopy(self.calc_fragement_grads(B_g, force_data["opt_fragment"]))
-                g = copy.deepcopy(self.calc_fragement_grads(g, force_data["opt_fragment"]))
-            
-            self.save_tmp_energy_profiles(iter, e, g, B_g)
-            
-            g, B_g, PC = self._apply_projection_constraints(projection_constrain, PC, geom_num_list, g, B_g)
-            g, B_g = self._zero_fixed_atom_gradients(allactive_flag, force_data, g, B_g)
+        return False, max_displacement_threshold, rms_displacement_threshold
 
-            # Update state with final gradients for this step
-            self.state.g = g
-            self.state.B_g = B_g
-            
-            if self.config.koopman_analysis:
-                _ = KA.run(iter, geom_num_list, B_g, self.element_list)
+    def judge_early_stop_due_to_no_negative_eigenvalues(self, geom_num_list, hessian, saddle_order, FC_COUNT, detect_negative_eigenvalues, folder_dir):
+        if detect_negative_eigenvalues and FC_COUNT > 0:
+            proj_hessian = Calculationtools().project_out_hess_tr_and_rot_for_coord(
+                hessian, geom_num_list, geom_num_list, display_eigval=False
+            )
+            if proj_hessian is not None:
+                eigvals = np.linalg.eigvalsh(proj_hessian)
+                if not np.any(eigvals < -1e-10) and saddle_order > 0:
+                    print("No negative eigenvalues detected while saddle_order > 0. Stopping optimization.")
+                    with open(
+                        folder_dir + "no_negative_eigenvalues_detected.txt",
+                        "w",
+                    ) as f:
+                        f.write("No negative eigenvalues detected while saddle_order > 0. Stopping optimization.")
+                    return True
+        return False
 
-            # --- Calculate Move Vector ---
-            new_geometry, move_vector, optimizer_instances = CMV.calc_move_vector(
-                iter, geom_num_list, B_g, self.state.pre_B_g, self.state.pre_geom, B_e, self.state.pre_B_e,
-                self.state.pre_move_vector, initial_geom_num_list, g, self.state.pre_g, optimizer_instances, projection_constrain)
-            
-            # --- Post-step Geometry Adjustments ---
-            new_geometry = self._project_out_translation_rotation(new_geometry, geom_num_list, allactive_flag)
-            new_geometry, PC = self._apply_projection_constraints_to_geometry(projection_constrain, PC, new_geometry, hessian=Hess)
-
-            # --- Update State Lists ---
-            self.state.ENERGY_LIST_FOR_PLOTTING.append(e * self.config.hartree2kcalmol)
-            self.state.BIAS_ENERGY_LIST_FOR_PLOTTING.append(B_e * self.config.hartree2kcalmol)
-            self.state.NUM_LIST.append(int(iter))
-            
-            self.geom_info_extract(force_data, file_directory, B_g, g) # This updates self.state.cos_list
-            
-            if self.state.iter == 0:
-                displacement_vector = move_vector
+
+class HessianManager:
+    def __init__(self, config):
+        self.config = config
+
+    def calc_eff_hess_for_fix_atoms_and_set_hess(
+        self,
+        state,
+        allactive_flag,
+        force_data,
+        BPA_hessian,
+        n_fix,
+        optimizer_instances,
+        geom_num_list,
+        B_g,
+        g,
+        projection_constrain,
+        PC,
+    ):
+        if not allactive_flag:
+            fix_num = []
+            for fnum in force_data["fix_atoms"]:
+                fix_num.extend([3 * (fnum - 1) + 0, 3 * (fnum - 1) + 1, 3 * (fnum - 1) + 2])
+            fix_num = np.array(fix_num, dtype="int64")
+            tmp_fix_hess = state.Model_hess[np.ix_(fix_num, fix_num)] + np.eye(
+                (3 * n_fix)
+            ) * 1e-10
+            inv_tmp_fix_hess = np.linalg.pinv(tmp_fix_hess)
+            tmp_fix_bias_hess = BPA_hessian[np.ix_(fix_num, fix_num)] + np.eye(
+                (3 * n_fix)
+            ) * 1e-10
+            inv_tmp_fix_bias_hess = np.linalg.pinv(tmp_fix_bias_hess)
+            BPA_hessian -= np.dot(
+                BPA_hessian[:, fix_num], np.dot(inv_tmp_fix_bias_hess, BPA_hessian[fix_num, :])
+            )
+
+        for inst in optimizer_instances:
+            if projection_constrain:
+                if np.all(np.abs(BPA_hessian) < 1e-20):
+                    proj_bpa_hess = PC.calc_project_out_hess(geom_num_list, B_g - g, BPA_hessian)
+                else:
+                    proj_bpa_hess = BPA_hessian
+                inst.set_bias_hessian(proj_bpa_hess)
             else:
-                displacement_vector = new_geometry / self.config.bohr2angstroms - geom_num_list
-            
-            # --- Check Convergence ---
-            converge_flag, max_displacement_threshold, rms_displacement_threshold = self._check_converge_criteria(B_g, displacement_vector)
-            self.print_info(e, B_e, B_g, displacement_vector, self.state.pre_e, self.state.pre_B_e, max_displacement_threshold, rms_displacement_threshold)
-            
-            self.state.grad_list.append(self.calculate_rms_safely(g))
-            self.state.bias_grad_list.append(self.calculate_rms_safely(B_g))
-            
-            new_geometry = self._reset_fixed_atom_positions(new_geometry, initial_geom_num_list, allactive_flag, force_data)
-            
-            # --- Dissociation Check ---
-            DC_exit_flag = self.dissociation_check(new_geometry, self.element_list)
+                inst.set_bias_hessian(BPA_hessian)
 
-            if converge_flag:
-                if projection_constrain and iter == 0:
-                    pass
+            if state.iter % self.config.FC_COUNT == 0 or (
+                self.config.use_model_hessian is not None
+                and state.iter % self.config.mFC_COUNT == 0
+            ):
+                if not allactive_flag:
+                    state.Model_hess -= np.dot(
+                        state.Model_hess[:, fix_num],
+                        np.dot(inv_tmp_fix_hess, state.Model_hess[fix_num, :]),
+                    )
+                if projection_constrain:
+                    proj_model_hess = PC.calc_project_out_hess(
+                        geom_num_list, g, state.Model_hess
+                    )
+                    inst.set_hessian(proj_model_hess)
                 else:
-                    self.state.optimized_flag = True
-                    print("\n=====================================================")
-                    print("converged!!!")
-                    print("=====================================================")
-                    break
-            
-            if DC_exit_flag:
-                self.state.DC_check_flag = True
-                break
-            
-            # --- Save State for Next Iteration ---
-            self.state.pre_B_e = B_e
-            self.state.pre_e = e
-            self.state.pre_B_g = B_g
-            self.state.pre_g = g
-            self.state.pre_geom = geom_num_list
-            self.state.pre_move_vector = move_vector
-            
-            # --- Write Next Input File ---
-            geometry_list = FIO.print_geometry_list(new_geometry, self.element_list, electric_charge_and_multiplicity)
-            file_directory = FIO.make_psi4_input_file(geometry_list, iter+1)
-            
-        else: # Loop ended (no break)
-            self.state.optimized_flag = False
-            print("Reached maximum number of iterations. This is not converged.")
-            with open(self.BPA_FOLDER_DIRECTORY+"not_converged.txt", "w") as f:
-                f.write("Reached maximum number of iterations. This is not converged.")
+                    inst.set_hessian(state.Model_hess)
+        return optimizer_instances
 
-        # --- 5. Post-Optimization Analysis ---
-        
-        # Check if exact hessian is already computed.
-        if self.config.FC_COUNT == -1:
-            exact_hess_flag = False
-        elif self.state.iter % self.config.FC_COUNT == 0 and self.config.FC_COUNT > 0:
-            exact_hess_flag = True
-        else:
-            exact_hess_flag = False
-        
-        if self.state.DC_check_flag:
-            print("Dissociation is detected. Optimization stopped.")
-            with open(self.BPA_FOLDER_DIRECTORY+"dissociation_is_detected.txt", "w") as f:
-                f.write("Dissociation is detected. Optimization stopped.")
 
-        if self.config.freq_analysis and not self.state.exit_flag and not self.state.DC_check_flag:
-            self._perform_vibrational_analysis(SP, geom_num_list, self.element_list, initial_geom_num_list, force_data, exact_hess_flag, file_directory, iter, electric_charge_and_multiplicity, xtb_method, e)
+class RunLogger:
+    def __init__(self, config):
+        self.config = config
 
-        # --- 6. Finalize and Save Results ---
-        self._finalize_optimization(FIO, G, self.state.grad_list, self.state.bias_grad_list, file_directory, force_data, geom_num_list, e, B_e, SP, self.state.exit_flag)
-        
-        # Copy final results from state to self
-        self._copy_final_results_from_state()
-        return
+    def log_dynamic_csv(self, state, folder_dir, convergence_checker):
+        csv_path = os.path.join(folder_dir, "energy_profile.csv")
+        keys = sorted(state.energies.keys())
+        if state.iter == 0:
+            with open(csv_path, "w") as f:
+                f.write("iter," + ",".join(keys) + "\n")
+        values = [str(state.energies[k]) for k in keys]
+        with open(csv_path, "a") as f:
+            f.write(f"{state.iter}," + ",".join(values) + "\n")
 
-    def _perform_vibrational_analysis(self, SP, geom_num_list, element_list, initial_geom_num_list, force_data, exact_hess_flag, file_directory, iter, electric_charge_and_multiplicity, xtb_method, e):
-        # (Reads self.state, self.config)
-        print("\n====================================================")
-        print("Performing vibrational analysis...")
-        print("====================================================\n")
-        print("Is Exact Hessian calculated? : ", exact_hess_flag)
-        
-        if exact_hess_flag:
-            g = np.zeros_like(geom_num_list, dtype="float64")
-            exit_flag = False
-        else:
-            print("Calculate exact Hessian...")
-            SP.hessian_flag = True
-            e, g, geom_num_list, exit_flag = SP.single_point(file_directory, element_list, iter, electric_charge_and_multiplicity, xtb_method)
-            SP.hessian_flag = False
-            
-        if exit_flag:
-            print("Error: QM calculation failed.")
-            return
-            
-        _, B_e, _, BPA_hessian = self.CalcBiaspot.main(e, g, geom_num_list, element_list, force_data, pre_B_g="", iter=iter, initial_geom_num_list="")
-        tmp_hess = copy.deepcopy(SP.Model_hess) # SP.Model_hess holds the latest hessian
-        tmp_hess += BPA_hessian
-        
-        MV = MolecularVibrations(atoms=element_list, coordinates=geom_num_list, hessian=tmp_hess)
-        results = MV.calculate_thermochemistry(e_tot=B_e, temperature=self.config.thermo_temperature, pressure=self.config.thermo_pressure)
-        
-        MV.print_thermochemistry(output_file=self.BPA_FOLDER_DIRECTORY+"/thermochemistry.txt")
-        MV.print_normal_modes(output_file=self.BPA_FOLDER_DIRECTORY+"/normal_modes.txt")
-        MV.create_vibration_animation(output_dir=self.BPA_FOLDER_DIRECTORY+"/vibration_animation")
-        
-        if not self.state.optimized_flag:
-            print("Warning: Vibrational analysis was performed, but the optimization did not converge. The result of thermochemistry is useless.")
-        
-        return
+        grad_profile = os.path.join(folder_dir, "gradient_profile.csv")
+        if state.iter == 0:
+            with open(grad_profile, "w") as f:
+                f.write("gradient (RMS) [hartree/Bohr] \n")
+        with open(grad_profile, "a") as f:
+            f.write(str(convergence_checker.calculate_rms_safely(state.raw_gradient)) + "\n")
 
-    def _finalize_optimization(self, FIO, G, grad_list, bias_grad_list, file_directory, force_data, geom_num_list, e, B_e, SP, exit_flag):
-        # (Writes to self.state)
-        self._save_opt_results(FIO, G, grad_list, bias_grad_list, file_directory, force_data)
-        
-        self.state.bias_pot_params_grad_list = self.CalcBiaspot.bias_pot_params_grad_list
-        self.state.bias_pot_params_grad_name_list = self.CalcBiaspot.bias_pot_params_grad_name_list
-        self.state.final_file_directory = file_directory
-        self.state.final_geometry = geom_num_list  # Bohr
-        self.state.final_energy = e  # Hartree
-        self.state.final_bias_energy = B_e  # Hartree
-        
-        if not exit_flag:
-            self.symmetry = analyze_symmetry(self.element_list, self.state.final_geometry)
-            self.state.symmetry = self.symmetry # Save to state too
-            with open(self.BPA_FOLDER_DIRECTORY+"symmetry.txt", "w") as f:
-                f.write(f"Symmetry of final structure: {self.symmetry}")
-            print(f"Symmetry: {self.symmetry}")
-    
-    def _save_opt_results(self, FIO, G, grad_list, bias_grad_list, file_directory, force_data):
-        # (Reads self.state)
-        G.double_plot(self.state.NUM_LIST, self.state.ENERGY_LIST_FOR_PLOTTING, self.state.BIAS_ENERGY_LIST_FOR_PLOTTING)
-        G.single_plot(self.state.NUM_LIST, grad_list, file_directory, "", axis_name_2="gradient (RMS) [a.u.]", name="gradient")
-        G.single_plot(self.state.NUM_LIST, bias_grad_list, file_directory, "", axis_name_2="bias gradient (RMS) [a.u.]", name="bias_gradient")
-        
+        bias_grad_profile = os.path.join(folder_dir, "bias_gradient_profile.csv")
+        if state.iter == 0:
+            with open(bias_grad_profile, "w") as f:
+                f.write("bias gradient (RMS) [hartree/Bohr] \n")
+        with open(bias_grad_profile, "a") as f:
+            f.write(str(convergence_checker.calculate_rms_safely(state.bias_gradient)) + "\n")
+
+    def save_energy_profiles(self, state, folder_dir):
+        with open(folder_dir + "energy_profile_kcalmol.csv", "w") as f:
+            f.write("ITER.,energy[kcal/mol]\n")
+            for i in range(len(state.ENERGY_LIST_FOR_PLOTTING)):
+                f.write(
+                    str(i)
+                    + ","
+                    + str(
+                        state.ENERGY_LIST_FOR_PLOTTING[i]
+                        - state.ENERGY_LIST_FOR_PLOTTING[0]
+                    )
+                    + "\n"
+                )
+
+    def geom_info_extract(self, state, force_data, file_directory, B_g, g, folder_dir):
         if len(force_data["geom_info"]) > 1:
+            CSI = CalculationStructInfo()
+            data_list, data_name_list = CSI.Data_extract(
+                glob.glob(file_directory + "/*.xyz")[0], force_data["geom_info"]
+            )
+
             for num, i in enumerate(force_data["geom_info"]):
-                G.single_plot(self.state.NUM_LIST, self.state.cos_list[num], file_directory, i)
+                cos = CSI.calculate_cos(B_g[i - 1] - g[i - 1], g[i - 1])
+                state.cos_list[num].append(cos)
 
-        FIO.make_traj_file()
-        FIO.argrelextrema_txt_save(self.state.ENERGY_LIST_FOR_PLOTTING, "approx_TS", "max")
-        FIO.argrelextrema_txt_save(self.state.ENERGY_LIST_FOR_PLOTTING, "approx_EQ", "min")
-        FIO.argrelextrema_txt_save(grad_list, "local_min_grad", "min")
+            if state.iter == 0:
+                with open(folder_dir + "geometry_info.csv", "a") as f:
+                    f.write(",".join(data_name_list) + "\n")
 
-        self._save_energy_profiles()
+            with open(folder_dir + "geometry_info.csv", "a") as f:
+                f.write(",".join(list(map(str, data_list))) + "\n")
         return
 
-    def _copy_final_results_from_state(self):
-        """Copy final results from the State object to the main Optimize object."""
-        if self.state:
-            self.final_file_directory = self.state.final_file_directory
-            self.final_geometry = self.state.final_geometry
-            self.final_energy = self.state.final_energy
-            self.final_bias_energy = self.state.final_bias_energy
-            self.symmetry = getattr(self.state, 'symmetry', None)
-            
-            # These were not in the original _finalize, but probably should be
-            self.bias_pot_params_grad_list = self.state.bias_pot_params_grad_list
-            self.bias_pot_params_grad_name_list = self.state.bias_pot_params_grad_name_list
-            self.optimized_flag = self.state.optimized_flag
 
-    def _check_converge_criteria(self, B_g, displacement_vector):
-        # (Reads self.config)
-        max_force = np.abs(B_g).max()
-        max_force_threshold = self.config.MAX_FORCE_THRESHOLD
-        
-        rms_force = self.calculate_rms_safely(B_g)
-        rms_force_threshold = self.config.RMS_FORCE_THRESHOLD
-        
-        delta_max_force_threshold = max(0.0, max_force_threshold -1 * max_force)
-        delta_rms_force_threshold = max(0.0, rms_force_threshold -1 * rms_force)
-        
-        max_displacement = np.abs(displacement_vector).max()
-        max_displacement_threshold = max(self.config.MAX_DISPLACEMENT_THRESHOLD, self.config.MAX_DISPLACEMENT_THRESHOLD + delta_max_force_threshold)
-        rms_displacement = self.calculate_rms_safely(displacement_vector)
-        rms_displacement_threshold = max(self.config.RMS_DISPLACEMENT_THRESHOLD, self.config.RMS_DISPLACEMENT_THRESHOLD + delta_rms_force_threshold)
-        
-        if max_force < max_force_threshold and rms_force < rms_force_threshold and max_displacement < max_displacement_threshold and rms_displacement < rms_displacement_threshold:
-            return True, max_displacement_threshold, rms_displacement_threshold
-        return False, max_displacement_threshold, rms_displacement_threshold
-    
-    def _import_calculation_module(self):
-        # (Reads self.config)
+class ResultPaths:
+    @staticmethod
+    def get_result_file_path(folder_dir, start_file):
+        try:
+            if folder_dir and start_file:
+                base_name = os.path.splitext(os.path.basename(start_file))[0]
+                optimized_filename = f"{base_name}_optimized.xyz"
+                traj_filename = f"{base_name}_traj.xyz"
+
+                optimized_struct_file = os.path.abspath(
+                    os.path.join(folder_dir, optimized_filename)
+                )
+                traj_file = os.path.abspath(
+                    os.path.join(folder_dir, traj_filename)
+                )
+
+                print("Optimized structure file path:", optimized_struct_file)
+                print("Trajectory file path:", traj_file)
+                return optimized_struct_file, traj_file
+            else:
+                print(
+                    "Error: BPA_FOLDER_DIRECTORY or START_FILE is not set. Please run optimize() first."
+                )
+                return None, None
+        except Exception as e:
+            print(f"Error setting result file paths: {e}")
+            return None, None
+
+# =====================================================================================
+# 5. Optimize Runner
+# =====================================================================================
+class Optimize:
+    """
+    Main runner orchestrating config, state, handler, and loop.
+    """
+
+    def __init__(self, args):
+        self.config = OptimizationConfig(args)
+        self.state = None
+        self.handler = None
+        self.file_io = None
+        self.BPA_FOLDER_DIRECTORY = None
+        self.START_FILE = None
+        self.element_list = None
+        self.SP = None
+        self.final_file_directory = None
+        self.final_geometry = None
+        self.final_energy = None
+        self.final_bias_energy = None
+        self.symmetry = None
+        self.irc_terminal_struct_paths = []
+        self.optimized_struct_file = None
+        self.traj_file = None
+        self.bias_pot_params_grad_list = None
+        self.bias_pot_params_grad_name_list = None
+
+        # Managers
+        self.constraints = ConstraintManager(self.config)
+        self.convergence = ConvergenceChecker(self.config)
+        self.hessian_mgr = HessianManager(self.config)
+        self.logger = RunLogger(self.config)
+        self.result_paths = ResultPaths()
+
+    # ------------------------------------------------------------------
+    # Setup helpers
+    # ------------------------------------------------------------------
+    def _setup_directory(self, input_file):
+        self.START_FILE = input_file
+        timestamp = datetime.datetime.now().strftime("%Y_%m_%d_%H_%M_%S_%f")[:-2]
+        date = datetime.datetime.now().strftime("%Y_%m_%d")
+        base_dir = f"{date}/{os.path.splitext(input_file)[0]}_OPT_"
+
+        if self.config.othersoft != "None":
+            suffix = "ASE"
+        elif self.config.sqm2:
+            suffix = "SQM2"
+        elif self.config.sqm1:
+            suffix = "SQM1"
+        elif self.config.args.usextb == "None" and self.config.args.usedxtb == "None":
+            suffix = f"{self.config.FUNCTIONAL}_{self.config.BASIS_SET}"
+        else:
+            method = (
+                self.config.args.usedxtb
+                if self.config.args.usedxtb != "None"
+                else self.config.args.usextb
+            )
+            suffix = method
+        self.BPA_FOLDER_DIRECTORY = f"{base_dir}{suffix}_{timestamp}/"
+        os.makedirs(self.BPA_FOLDER_DIRECTORY, exist_ok=True)
+
+        with open(os.path.join(self.BPA_FOLDER_DIRECTORY, "input.txt"), "w") as f:
+            f.write(str(vars(self.config.args)))
+
+    def _init_calculation_module(self):
         xtb_method = None
         if self.config.args.pyscf:
             from multioptpy.Calculator.pyscf_calculation_tools import Calculation
@@ -784,14 +1537,17 @@ class Optimize:
                 print("Use Lennard-Jones cluster potential.")
             elif self.config.othersoft.lower() == "emt":
                 from multioptpy.Calculator.emt_calculation_tools import Calculation
-                print("Use ETM potential.")
+                print("Use EMT potential.")
             elif self.config.othersoft.lower() == "tersoff":
                 from multioptpy.Calculator.tersoff_calculation_tools import Calculation
                 print("Use Tersoff potential.")
             else:
                 from multioptpy.Calculator.ase_calculation_tools import Calculation
                 print("Use", self.config.othersoft)
-                with open(self.BPA_FOLDER_DIRECTORY + "use_" + self.config.othersoft + ".txt", "w") as f:
+                with open(
+                    self.BPA_FOLDER_DIRECTORY + "use_" + self.config.othersoft + ".txt",
+                    "w",
+                ) as f:
                     f.write(self.config.othersoft + "\n")
                     f.write(self.config.BASIS_SET + "\n")
                     f.write(self.config.FUNCTIONAL + "\n")
@@ -804,24 +1560,18 @@ class Optimize:
                 xtb_method = self.config.args.usextb
             else:
                 from multioptpy.Calculator.psi4_calculation_tools import Calculation
-                
         return Calculation, xtb_method
-    
-    def setup_calculation(self, Calculation):
-        # (Reads self.config, self.state)
-        # Note: Model_hess is passed from state, but SP is re-created per job.
-        # This assumes the initial Model_hess (eye) is what's needed.
-        # This might be a flaw if SP needs the *current* Model_hess.
-        # Let's assume self.state.Model_hess is correct at time of call.
-        
-        SP = Calculation(
+
+    def _create_calculation(self, Calculation, xtb_method, model_hess, override_dir=None):
+        target_dir = override_dir if override_dir else self.BPA_FOLDER_DIRECTORY
+        calc = Calculation(
             START_FILE=self.START_FILE,
             N_THREAD=self.config.N_THREAD,
             SET_MEMORY=self.config.SET_MEMORY,
             FUNCTIONAL=self.config.FUNCTIONAL,
             FC_COUNT=self.config.FC_COUNT,
-            BPA_FOLDER_DIRECTORY=self.BPA_FOLDER_DIRECTORY,
-            Model_hess=self.state.Model_hess, # Reads from state
+            BPA_FOLDER_DIRECTORY=target_dir, 
+            Model_hess=model_hess,
             software_type=self.config.othersoft,
             unrestrict=self.config.unrestrict,
             SUB_BASIS_SET=self.config.SUB_BASIS_SET,
@@ -830,724 +1580,874 @@ class Optimize:
             electronic_charge=self.config.electronic_charge,
             excited_state=self.config.excited_state,
             dft_grid=self.config.dft_grid,
-            ECP = self.config.ECP,
-            software_path_file = self.config.software_path_file
+            ECP=self.config.ECP,
+            software_path_file=self.config.software_path_file,
         )
-        SP.cpcm_solv_model = self.config.cpcm_solv_model
-        SP.alpb_solv_model = self.config.alpb_solv_model
-        return SP
-
-    def write_input_files(self, FIO):
-        # (Reads self.config)
-        # (This method sets self.element_list and self.state.Model_hess,
-        #  which is a bit of a side-effect, but we'll keep it)
-        
-        if os.path.splitext(FIO.START_FILE)[1] == ".gjf":
-            print("Gaussian input file (.gjf) detected.")
-            geometry_list, element_list, electric_charge_and_multiplicity = FIO.read_gjf_file(self.config.electric_charge_and_multiplicity)
-        elif os.path.splitext(FIO.START_FILE)[1] == ".inp":
-            print("GAMESS/Orca/Q-Chem input file (.inp) detected.")
-            geometry_list, element_list, electric_charge_and_multiplicity = FIO.read_gamess_inp_file(self.config.electric_charge_and_multiplicity)
-        elif os.path.splitext(FIO.START_FILE)[1] == ".mol":
-            print("MDL Molfile (.mol) detected.")
-            geometry_list, element_list, electric_charge_and_multiplicity = FIO.read_mol_file(self.config.electric_charge_and_multiplicity)
-        elif os.path.splitext(FIO.START_FILE)[1] == ".mol2":
-            print("MOL2 file (.mol2) detected.")
-            geometry_list, element_list, electric_charge_and_multiplicity = FIO.read_mol2_file(self.config.electric_charge_and_multiplicity)
-        else:
-            geometry_list, element_list, electric_charge_and_multiplicity = FIO.make_geometry_list(self.config.electric_charge_and_multiplicity)
-        
-        file_directory = FIO.make_psi4_input_file(geometry_list, 0)
+        calc.cpcm_solv_model = self.config.cpcm_solv_model
+        calc.alpb_solv_model = self.config.alpb_solv_model
+        calc.xtb_method = xtb_method
+        return calc
+
+    def _initialize_handler(self, element_list, force_data):
+        Calculation, xtb_method = self._init_calculation_module()
         
-        if self.config.args.pyscf:
-            electric_charge_and_multiplicity = self.config.electric_charge_and_multiplicity
-            
-        self.element_list = element_list # Set self.element_list
-        # self.Model_hess = np.eye(len(element_list) * 3) # This is now done in OptimizationState
+   
+        self.SP = self._create_calculation(Calculation, xtb_method, self.state.Model_hess)
         
-        return file_directory, electric_charge_and_multiplicity, element_list
-
-    def save_tmp_energy_profiles(self, iter, e, g, B_g):
-        # (This method is pure, no changes needed, writes to files)
-        if iter == 0:
-            with open(self.BPA_FOLDER_DIRECTORY+"energy_profile.csv","a") as f:
-                f.write("energy [hartree] \n")
-        with open(self.BPA_FOLDER_DIRECTORY+"energy_profile.csv","a") as f:
-            f.write(str(e)+"\n")
-        #-------------------gradient profile
-        if iter == 0:
-            with open(self.BPA_FOLDER_DIRECTORY+"gradient_profile.csv","a") as f:
-                f.write("gradient (RMS) [hartree/Bohr] \n")
-        with open(self.BPA_FOLDER_DIRECTORY+"gradient_profile.csv","a") as f:
-            f.write(str(self.calculate_rms_safely(g))+"\n")
-        #-------------------
-        if iter == 0:
-            with open(self.BPA_FOLDER_DIRECTORY+"bias_gradient_profile.csv","a") as f:
-                f.write("bias gradient (RMS) [hartree/Bohr] \n")
-        with open(self.BPA_FOLDER_DIRECTORY+"bias_gradient_profile.csv","a") as f:
-            f.write(str(self.calculate_rms_safely(B_g))+"\n")
-            #-------------------
-        return
-    
-    def _save_energy_profiles(self):
-        # (Reads self.state)
-        with open(self.BPA_FOLDER_DIRECTORY+"energy_profile_kcalmol.csv","w") as f:
-            f.write("ITER.,energy[kcal/mol]\n")
-            for i in range(len(self.state.ENERGY_LIST_FOR_PLOTTING)):
-                f.write(str(i)+","+str(self.state.ENERGY_LIST_FOR_PLOTTING[i] - self.state.ENERGY_LIST_FOR_PLOTTING[0])+"\n")
-        return
-
-    def geom_info_extract(self, force_data, file_directory, B_g, g):
-        # (Writes to self.state.cos_list)
-        if len(force_data["geom_info"]) > 1:
-            CSI = CalculationStructInfo()
-            
-            data_list, data_name_list = CSI.Data_extract(glob.glob(file_directory+"/*.xyz")[0], force_data["geom_info"])
-            
-            for num, i in enumerate(force_data["geom_info"]):
-                cos = CSI.calculate_cos(B_g[i-1] - g[i-1], g[i-1])
-                self.state.cos_list[num].append(cos)
-            
-            # Need to use self.state.iter to check
-            if self.state.iter == 0:
-                with open(self.BPA_FOLDER_DIRECTORY+"geometry_info.csv","a") as f:
-                    f.write(",".join(data_name_list)+"\n")
-            
-            with open(self.BPA_FOLDER_DIRECTORY+"geometry_info.csv","a") as f:    
-                f.write(",".join(list(map(str,data_list)))+"\n")          
-        return
-    
-    
-    def dissociation_check(self, new_geometry, element_list):
-        """
-        Checks if the molecular geometry has dissociated into multiple fragments
-        based on a distance threshold.
-        """
-        # (Reads self.config.DC_check_dist)
-        atom_label_list = list(range(len(new_geometry)))
-        fragm_atom_num_list = []
-
-        # 1. Identify all molecular fragments (connected components)
-        while len(atom_label_list) > 0:
-            tmp_fragm_list = Calculationtools().check_atom_connectivity(new_geometry, element_list, atom_label_list[0])
-            atom_label_list = list(set(atom_label_list) - set(tmp_fragm_list))
-            fragm_atom_num_list.append(tmp_fragm_list)
+        self.state.element_list = element_list
+        self.state.element_number_list = np.array(
+            [element_number(elem) for elem in element_list], dtype="int"
+        )
         
-        # 2. Check distances only if there is more than one fragment
-        if len(fragm_atom_num_list) > 1:
-            fragm_dist_list = []
-            
-            # Ensure geometry is a NumPy array for efficient slicing
-            geom_np = np.asarray(new_geometry)
-            
-            # Iterate through all unique pairs of fragments
-            for fragm_1_indices, fragm_2_indices in itertools.combinations(fragm_atom_num_list, 2):
+        # --- EDEEL Mode ---
+        # Assuming the parser sets an 'edeel' flag in args or force_data
+        if hasattr(self.config.args, 'edeel') and self.config.args.edeel:
+            print("Mode: EDEEL (Energy Decomposition and Extrapolation-based Electron Localization)")
+            
+            # Note: The user handles the parser. 
+            # We assume 'ede_params', 'donor_atoms', and 'acceptor_atoms' are available in force_data.
+            # Example structure for ede_params: {'complex': [0,1], 'd_ox': [0,1], ...}
+            
+            ede_params = force_data.get('edeel_params')
+            d_atoms = force_data.get('donor_atoms', [])
+            a_atoms = force_data.get('acceptor_atoms', [])
+            
+            if not ede_params or not d_atoms or not a_atoms:
+                raise ValueError("EDEEL mode requires 'edeel_params', 'donor_atoms', and 'acceptor_atoms' in input.")
+
+            return EDEELHandler(
+                self.SP, 
+                d_atoms, 
+                a_atoms, 
+                ede_params,
+                self.config, 
+                self.file_io, 
+                self.BPA_FOLDER_DIRECTORY, 
+                force_data
+            )
+        # --- Model Function Mode (NEW) ---
+        elif len(self.config.args.model_function) > 0:
+            print("Mode: Model Function Optimization")
+            Calculation, xtb_method = self._init_calculation_module()
+            
+            # Create independent base directories for State 1 and State 2
+            dir1 = os.path.join(self.BPA_FOLDER_DIRECTORY, "State1_base")
+            dir2 = os.path.join(self.BPA_FOLDER_DIRECTORY, "State2_base")
+            os.makedirs(dir1, exist_ok=True)
+            os.makedirs(dir2, exist_ok=True)
+
+            # Initialize two independent calculators to prevent state contamination
+            calc1 = self._create_calculation(Calculation, xtb_method, self.state.Model_hess, override_dir=dir1)
+            calc2 = self._create_calculation(Calculation, xtb_method, self.state.Model_hess, override_dir=dir2)
+
+            handler = ModelFunctionHandler(
+                calc1, calc2, 
+                self.config.args.model_function, 
+                self.config, 
+                self.file_io, 
+                self.BPA_FOLDER_DIRECTORY, 
+                force_data
+            )
+            
+            # --- BITSS Mode ---
+            if handler.is_bitss:
+                print("BITSS Mode detected: Expanding state to 2N atoms.")
+                geom1 = self.state.geometry
+                geom2 = handler.bitss_ref_geom
                 
-                # Get the coordinates for all atoms in each fragment
-                coords1 = geom_np[fragm_1_indices] # Shape (M, 3)
-                coords2 = geom_np[fragm_2_indices] # Shape (K, 3)
+                if geom1.shape != geom2.shape:
+                    raise ValueError("BITSS: Input and Reference geometries must have the same dimensions.")
                 
-                # Reshape coords1 to (M, 1, 3) and coords2 to (1, K, 3)
-                # This allows NumPy broadcasting to create all pairs of differences
-                # The result (diff_matrix) will have shape (M, K, 3)
-                diff_matrix = coords1[:, np.newaxis, :] - coords2[np.newaxis, :, :]
+                # Expand geometry and gradients to 6N dimensions
+                self.state.geometry = np.vstack((geom1, geom2))
+                self.state.initial_geometry = copy.deepcopy(self.state.geometry)
+                self.state.pre_geometry = copy.deepcopy(self.state.geometry)
                 
-                # Square the differences and sum along the last axis (axis=2)
-                # This calculates the squared Euclidean distance for all pairs
-                # The result (sq_dist_matrix) will have shape (M, K)
-                sq_dist_matrix = np.sum(diff_matrix**2, axis=2)
+                n_atoms = len(element_list) # Original N
                 
-                # Find the minimum value in the squared distance matrix
-                min_sq_dist = np.min(sq_dist_matrix)
+                # Current gradients
+                self.state.raw_gradient = np.zeros((2 * n_atoms, 3))
+                self.state.bias_gradient = np.zeros((2 * n_atoms, 3))
+                self.state.effective_gradient = np.zeros((2 * n_atoms, 3))
                 
-                # Take the square root of only the minimum value to get the final distance
-                min_dist = np.sqrt(min_sq_dist)
-           
-                fragm_dist_list.append(min_dist)
                 
-            # 3. Check if the closest distance between any two fragments
-            #    is greater than the dissociation threshold.
-            min_interfragment_dist = min(fragm_dist_list)
-            
-            if min_interfragment_dist > self.config.DC_check_dist:
-                print(f"Minimum fragment distance (ang.) {min_interfragment_dist:.4f} > {self.config.DC_check_dist}")
-                print("These molecules are dissociated.")
-                DC_exit_flag = True
+                self.state.pre_raw_gradient = np.zeros((2 * n_atoms, 3))
+                self.state.pre_bias_gradient = np.zeros((2 * n_atoms, 3))
+                self.state.pre_effective_gradient = np.zeros((2 * n_atoms, 3))
+                self.state.pre_move_vector = np.zeros((2 * n_atoms, 3))
+               
+
+                self.state.Model_hess = np.eye(2 * n_atoms * 3)
+                
+                # Double the element lists
+                self.state.element_list = element_list + element_list
+                self.state.element_number_list = np.concatenate((self.state.element_number_list, self.state.element_number_list))
             else:
-                DC_exit_flag = False
-        else:
-            # Only one fragment, so it's not dissociated
-            DC_exit_flag = False
+                print(f"Standard Model Function Mode ({handler.method_name}): Using {len(element_list)} atoms.")
+                
+            return handler
+        
+        # --- ONIOM Mode ---
+        elif len(self.config.args.oniom_flag) > 0:
+            # ONIOM
+            high_atoms = force_data["oniom_flag"][0]
+            link_atoms = force_data["oniom_flag"][1]
             
-        return DC_exit_flag
-    
-    def calculate_rms_safely(self, vector, threshold=1e-10):
-        # (This method is pure, no changes needed)
-        filtered_vector = vector[np.abs(vector) > threshold]
-        if filtered_vector.size > 0:
-            return np.sqrt((filtered_vector**2).mean())
-        else:
-            return 0.0
             
-    def print_info(self, e, B_e, B_g, displacement_vector, pre_e, pre_B_e, max_displacement_threshold, rms_displacement_threshold):
-        # (Reads self.config)
-        rms_force = self.calculate_rms_safely(np.abs(B_g))
-        rms_displacement = self.calculate_rms_safely(np.abs(displacement_vector))
-        max_B_g = np.abs(B_g).max()
-        max_displacement = np.abs(displacement_vector).max()
-        print("caluculation results (unit a.u.):")
-        print("                         Value                     Threshold ")
-        print("ENERGY                 : {:>15.12f} ".format(e))
-        print("BIAS  ENERGY           : {:>15.12f} ".format(B_e))
-        print("Maximum  Force         : {0:>15.12f}                 {1:>15.12f} ".format(max_B_g, self.config.MAX_FORCE_THRESHOLD))
-        print("RMS      Force         : {0:>15.12f}                 {1:>15.12f} ".format(rms_force, self.config.RMS_FORCE_THRESHOLD))
-        print("Maximum  Displacement  : {0:>15.12f}                 {1:>15.12f} ".format(max_displacement, max_displacement_threshold))
-        print("RMS      Displacement  : {0:>15.12f}                 {1:>15.12f} ".format(rms_displacement, rms_displacement_threshold))
-        print("ENERGY SHIFT           : {:>15.12f} ".format(e - pre_e))
-        print("BIAS ENERGY SHIFT      : {:>15.12f} ".format(B_e - pre_B_e))
-        return
-    
-    def calc_fragement_grads(self, gradient, fragment_list):
-        # (This method is pure, no changes needed)
-        calced_gradient = gradient
-        for fragment in fragment_list:
-            tmp_grad = np.array([0.0, 0.0, 0.0], dtype="float64")
-            for atom_num in fragment:
-                tmp_grad += gradient[atom_num-1]
-            tmp_grad /= len(fragment)
-
-            for atom_num in fragment:
-                calced_gradient[atom_num-1] = copy.deepcopy(tmp_grad)
-        return calced_gradient
+            hl_dir = os.path.join(self.BPA_FOLDER_DIRECTORY, "High_Layer")
+            ll_dir = os.path.join(self.BPA_FOLDER_DIRECTORY, "Low_Layer")
+            os.makedirs(hl_dir, exist_ok=True)
+            os.makedirs(ll_dir, exist_ok=True)
+            
+          
+            high_calc = self._create_calculation(Calculation, xtb_method, self.state.Model_hess, override_dir=hl_dir)
+            low_calc = self._create_calculation(Calculation, xtb_method, self.state.Model_hess, override_dir=ll_dir)
+           
 
-    def optimize_oniom(self):
+            return ONIOMHandler(
+                high_calc,
+                low_calc,
+                high_atoms,
+                link_atoms,
+                self.config,
+                self.file_io,
+                self.BPA_FOLDER_DIRECTORY,
+                force_data,
+            )
+        else:
+            # Standard
+            return StandardHandler(
+                self.SP, self.config, self.file_io, self.BPA_FOLDER_DIRECTORY, force_data
+            )
+            
+            
+    # ------------------------------------------------------------------
+    # Geometry parsing helper
+    # ------------------------------------------------------------------
+    def _extract_geom_from_geometry_list(self, geometry_list):
         """
-        Perform ONIOM optimization using a high-level QM method for a subset of atoms
-        and a low-level method for the entire system.
-        
-        Refactored to use self.config and self.state.
+        Extract geometry (Bohr) from geometry_list returned by FileIO.make_geometry_list.
+        Assumes geometry_list[0][0] and [0][1] hold charge/multiplicity,
+        and atom records start from geometry_list[0][2], each formatted as
+        [element, x, y, z] in Angstrom.
         """
-        # 1. Parse input parameters and initialize file IO
-        force_data = force_data_parser(self.config.args)
-        high_layer_atom_num = force_data["oniom_flag"][0]
-        link_atom_num = force_data["oniom_flag"][1]
-        calc_method = force_data["oniom_flag"][2]
-        
-        FIO = FileIO(self.BPA_FOLDER_DIRECTORY, self.START_FILE)
-        
-        # 2. Write input files and create the State object
-        geometry_list, element_list, electric_charge_and_multiplicity = self.write_input_files(FIO)
-        self.element_list = element_list # Set on self for helpers
-        
-        # Create the main State object for the "Real" system
-        self.state = OptimizationState(element_list)
-        self.state.cos_list = [[] for i in range(len(force_data["geom_info"]))]
+        try:
+            atom_entries = geometry_list[0][2:]
+            coords_ang = np.array([atom[1:4] for atom in atom_entries], dtype=float)
+            geom_bohr = coords_ang / self.config.bohr2angstroms
+            return geom_bohr
+        except Exception as e:
+            raise ValueError(f"Failed to parse geometry_list: {geometry_list}") from e
 
-        file_directory = FIO.make_psi4_input_file(geometry_list, 0)
-        
-        # 3. Import appropriate calculation modules
-        if self.config.args.pyscf:
-            from multioptpy.Calculator.pyscf_calculation_tools import Calculation as HL_Calculation
-        else:
-            from multioptpy.Calculator.psi4_calculation_tools import Calculation as HL_Calculation
-        
-        if calc_method in ["GFN2-xTB", "GFN1-xTB", "IPEA1-xTB"]:
-            from multioptpy.Calculator.tblite_calculation_tools import Calculation as LL_Calculation
-        else:
-            from multioptpy.Calculator.ase_calculation_tools import Calculation as LL_Calculation
-        
-        # Save ONIOM configuration to file
-        with open(self.BPA_FOLDER_DIRECTORY+"ONIOM2.txt", "w") as f:
-            f.write("### Low layer ###\n")
-            f.write(calc_method+"\n")
-            f.write("### High layer ###\n")
-            f.write(self.config.BASIS_SET+"\n")
-            f.write(self.config.FUNCTIONAL+"\n")  
-        
-        # 4. Initialize geometries and ONIOM setup
-        geom_num_list = []
-        for i in range(2, len(geometry_list[0])):
-            geom_num_list.append(geometry_list[0][i][1:4])
-        geom_num_list = np.array(geom_num_list, dtype="float64") / self.config.bohr2angstroms
-        self.state.geom_num_list = geom_num_list # Set initial geometry in state
-        
-        linker_atom_pair_num = specify_link_atom_pairs(geom_num_list, element_list, high_layer_atom_num, link_atom_num)
-        print("Boundary of high layer and low layer:", linker_atom_pair_num)
-        
-        high_layer_geom_num_list, high_layer_element_list = separate_high_layer_and_low_layer(
-            geom_num_list, linker_atom_pair_num, high_layer_atom_num, element_list)
-        
-        real_2_highlayer_label_connect_dict, highlayer_2_real_label_connect_dict = link_number_high_layer_and_low_layer(high_layer_atom_num)
-        
-        # 5. Initialize model Hessians (local state for ONIOM)
-        LL_Model_hess = np.eye(len(element_list)*3)
-        HL_Model_hess = np.eye((len(high_layer_element_list))*3)
-        
-        # Create mask for high layer atoms
-        bool_list = []
-        for i in range(len(element_list)):
-            if i in high_layer_atom_num:
-                bool_list.extend([True, True, True])
+    # ------------------------------------------------------------------
+    # Main run
+    # ------------------------------------------------------------------
+    def run(self):
+        input_list = (
+            [self.config.args.INPUT]
+            if isinstance(self.config.args.INPUT, str)
+            else self.config.args.INPUT
+        )
+        job_file_list = []
+        for job_file in input_list:
+            if "*" in job_file:
+                job_file_list.extend(glob.glob(job_file))
             else:
-                bool_list.extend([False, False, False])
-        
-        # 6. Initialize bias potential calculators
-        # (self.CalcBiaspot will be used for the "Real" system bias)
-        LL_Calc_BiasPot = BiasPotentialCalculation(self.BPA_FOLDER_DIRECTORY)
-        # HL_Calc_BiasPot = BiasPotentialCalculation(self.BPA_FOLDER_DIRECTORY) # Seems unused in original
-        self.CalcBiaspot = BiasPotentialCalculation(self.BPA_FOLDER_DIRECTORY) # For main state
+                job_file_list.append(job_file)
 
-        with open(self.BPA_FOLDER_DIRECTORY+"input.txt", "w") as f:
-            f.write(str(vars(self.config.args)))
-        
-        # 7. Initialize ONIOM-specific previous-step variables (as local vars)
-        pre_model_HL_B_e = 0.0
-        pre_model_HL_B_g = np.zeros((len(high_layer_element_list), 3))
-        pre_model_HL_g = np.zeros((len(high_layer_element_list), 3))
-        # pre_model_LL_B_g = np.zeros((len(high_layer_element_list), 3)) # Seems unused
-        pre_real_LL_B_e = 0.0
-        pre_real_LL_e = 0.0
-        pre_real_LL_B_g = np.zeros((len(element_list), 3))
-        pre_real_LL_g = np.zeros((len(element_list), 3))
-        pre_real_LL_move_vector = np.zeros((len(element_list), 3))
-        pre_model_HL_move_vector = np.zeros((len(high_layer_element_list), 3))
-        
-        # 8. Initialize HL optimizer
-        HL_CMV = CalculateMoveVector(self.config.DELTA, high_layer_element_list[:len(high_layer_atom_num)], 
-                                     self.config.args.saddle_order, self.config.FC_COUNT, self.config.temperature, 
-                                     max_trust_radius=self.config.max_trust_radius, min_trust_radius=self.config.min_trust_radius)
-        HL_optimizer_instances = HL_CMV.initialization(force_data["opt_method"])
+        for file in job_file_list:
+            self._run_single_job(file)
+
+        print("All calculations were completed.")
+        self.get_result_file_path()
+        return
+
+    # ------------------------------------------------------------------
+    # Extracted single-job runner to simplify run()
+    # ------------------------------------------------------------------
+    def _run_single_job(self, file):
+        print("********************************")
+        print(file)
+        print("********************************")
+        if not os.path.exists(file):
+            print(f"{file} does not exist.")
+            return
+
+        self._setup_directory(file)
+        self.file_io = FileIO(self.BPA_FOLDER_DIRECTORY, file)
+
+        # Read geometry
+        geometry_list, element_list, chg_mult = self.file_io.make_geometry_list(
+            self.config.electric_charge_and_multiplicity
+        )
+        geom = self._extract_geom_from_geometry_list(geometry_list)
+        self.element_list = element_list
+
+        # Initialize state
+        # NOTE: element_list is passed here, but self.state.element_list might be modified later (e.g. for BITSS)
+        self.state = OptimizationState(element_list)
+        self.state.geometry = copy.deepcopy(geom)
+        self.state.initial_geometry = copy.deepcopy(geom)
+        self.state.element_list = element_list
+        self.state.element_number_list = np.array(
+            [element_number(elem) for elem in element_list], dtype="int"
+        )
+        self.state.cos_list = [
+            [] for _ in range(len(force_data_parser(self.config.args)["geom_info"]))
+        ]
+
+        force_data = force_data_parser(self.config.args)
         
-        for i in range(len(HL_optimizer_instances)):
-            HL_optimizer_instances[i].set_hessian(HL_Model_hess[:len(high_layer_atom_num)*3, :len(high_layer_atom_num)*3])
+        # Initialize Handler (This may update self.state.element_list and self.state.geometry for BITSS)
+        self.handler = self._initialize_handler(element_list, force_data)
+
+        # Constraint setup
+        PC = ProjectOutConstrain(
+            force_data["projection_constraint_condition_list"],
+            force_data["projection_constraint_atoms"],
+            force_data["projection_constraint_constant"],
+        )
+        projection_constrain, allactive_flag = self.constraints.constrain_flag_check(force_data)
+        n_fix = len(force_data["fix_atoms"])
+
+        # Move vector and optimizer
+        # FIX: Use self.state.element_list which handles the 2N size in BITSS mode
+        CMV = CalculateMoveVector(
+            self.config.DELTA,
+            self.state.element_list,
+            self.config.args.saddle_order,
+            self.config.FC_COUNT,
+            self.config.temperature,
+            self.config.use_model_hessian,
+            max_trust_radius=self.config.max_trust_radius,
+            min_trust_radius=self.config.min_trust_radius,
+            projection_constraint=PC
+        )
+        optimizer_instances = CMV.initialization(force_data["opt_method"])
+        for opt in optimizer_instances:
+            opt.set_hessian(self.state.Model_hess)
             if self.config.DELTA != "x":
-                HL_optimizer_instances[i].DELTA = self.config.DELTA     
-        
-        # 9. Initialize calculation instances
-        HLSP = HL_Calculation(START_FILE=self.START_FILE,
-                              SUB_BASIS_SET=self.config.SUB_BASIS_SET,
-                              BASIS_SET=self.config.BASIS_SET,
-                              N_THREAD=self.config.N_THREAD,
-                              SET_MEMORY=self.config.SET_MEMORY,
-                              FUNCTIONAL=self.config.FUNCTIONAL,
-                              FC_COUNT=self.config.FC_COUNT,
-                              BPA_FOLDER_DIRECTORY=self.BPA_FOLDER_DIRECTORY,
-                              Model_hess=HL_Model_hess[:len(high_layer_atom_num)*3, :len(high_layer_atom_num)*3],
-                              unrestrict=self.config.unrestrict,
-                              excited_state=self.config.excited_state,
-                              electronic_charge=self.config.electronic_charge,
-                              spin_multiplicity=self.config.spin_multiplicity
-                              )
-        
-        LLSP = LL_Calculation(START_FILE=self.START_FILE,
-                              SUB_BASIS_SET=self.config.SUB_BASIS_SET,
-                              BASIS_SET=self.config.BASIS_SET,
-                              N_THREAD=self.config.N_THREAD,
-                              SET_MEMORY=self.config.SET_MEMORY,
-                              FUNCTIONAL=self.config.FUNCTIONAL,
-                              FC_COUNT=self.config.FC_COUNT,
-                              BPA_FOLDER_DIRECTORY=self.BPA_FOLDER_DIRECTORY,
-                              Model_hess=LL_Model_hess,
-                              unrestrict=self.config.unrestrict,
-                              software_type=calc_method,
-                              excited_state=self.config.excited_state)
-        
-        # 10. Initialize result tracking (uses self.state)
-        real_grad_list = []
-        real_bias_grad_list = []
-        
-        # 11. Main optimization loop
-        for iter in range(self.config.NSTEP):
-            self.state.iter = iter
-            
-            exit_file_detect = os.path.exists(self.BPA_FOLDER_DIRECTORY+"end.txt")
-            if exit_file_detect:
-                self.state.exit_flag = True
+                opt.DELTA = self.config.DELTA
+            # Gracefully handle optimizers without newton_tag
+            supports_exact_hess = getattr(opt, "newton_tag", True)
+            if (not supports_exact_hess) and self.config.FC_COUNT > 0 and (
+                "eigvec" not in force_data["projection_constraint_condition_list"]
+            ):
+                print(
+                    "Error: This optimizer does not support exact Hessian calculations."
+                )
+                sys.exit(0)
+
+        # Koopman
+        if self.config.koopman_analysis:
+            # FIX: Use self.state.element_list
+            KA = KoopmanAnalyzer(len(self.state.element_list), file_directory=self.BPA_FOLDER_DIRECTORY)
+        else:
+            KA = None
+
+        # Initial files
+        # FIX: Use self.state.element_list
+        self.file_io.print_geometry_list(
+            self.state.geometry * self.config.bohr2angstroms,
+            self.state.element_list,
+            chg_mult,
+        )
+
+        # Main loop
+        for iter_idx in range(self.config.NSTEP):
+            self.state.iter = iter_idx
+            self.state.exit_flag = os.path.exists(
+                self.BPA_FOLDER_DIRECTORY + "end.txt"
+            )
+            if self.state.exit_flag:
                 break
-            
-            print(f"\n# ITR. {iter}\n")
-            
-            if iter == 0:
-                high_layer_initial_geom_num_list = high_layer_geom_num_list.copy()  # Bohr
-                high_layer_pre_geom = high_layer_initial_geom_num_list.copy()  # Bohr
-                real_initial_geom_num_list = geom_num_list.copy()  # Bohr
-                real_pre_geom = real_initial_geom_num_list.copy()  # Bohr
-            
-            # --- Model Low Layer Calc ---
-            print("Model low layer calculation")
-            model_LL_e, model_LL_g, high_layer_geom_num_list, finish_frag = LLSP.single_point(
-                file_directory, high_layer_element_list, iter, electric_charge_and_multiplicity, 
-                calc_method, geom_num_list=high_layer_geom_num_list*self.config.bohr2angstroms)
-            
-            if finish_frag:
-                self.state.exit_flag = True
+
+            # shape condition check
+            self.state.exit_flag = judge_shape_condition(
+                self.state.geometry, self.config.shape_conditions
+            )
+            if self.state.exit_flag:
                 break
-            
-            # --- Microiterations ---
-            print("Processing microiteration...")
-            LL_CMV = CalculateMoveVector(self.config.DELTA, element_list, self.config.args.saddle_order, self.config.FC_COUNT, self.config.temperature)
-            LL_optimizer_instances = LL_CMV.initialization(["fire"])
-            LL_optimizer_instances[0].display_flag = False
-            
-            low_layer_converged = False
-            
-            # Use geom_num_list from main state
-            current_geom_num_list = self.state.geom_num_list.copy()
-            
-            for microiter in range(self.config.microiter_num):
-                LLSP.Model_hess = LL_Model_hess
-                
-                real_LL_e, real_LL_g, current_geom_num_list, finish_frag = LLSP.single_point(
-                    file_directory, element_list, microiter, electric_charge_and_multiplicity, 
-                    calc_method, geom_num_list=current_geom_num_list*self.config.bohr2angstroms)
-                
-                LL_Model_hess = LLSP.Model_hess
-                
-                LL_Calc_BiasPot.Model_hess = LL_Model_hess
-                _, real_LL_B_e, real_LL_B_g, LL_BPA_hessian = LL_Calc_BiasPot.main(
-                    real_LL_e, real_LL_g, current_geom_num_list, element_list, 
-                    force_data, pre_real_LL_B_g, microiter, real_initial_geom_num_list)
-                
-                for x in range(len(LL_optimizer_instances)):
-                    LL_optimizer_instances[x].set_bias_hessian(LL_BPA_hessian)
-                    if microiter % self.config.FC_COUNT == 0: # Using FC_COUNT, not mFC_COUNT
-                        LL_optimizer_instances[x].set_hessian(LL_Model_hess)
-                
-                if len(force_data["opt_fragment"]) > 0:
-                    real_LL_B_g = copy.deepcopy(self.calc_fragement_grads(real_LL_B_g, force_data["opt_fragment"]))
-                    real_LL_g = copy.deepcopy(self.calc_fragement_grads(real_LL_g, force_data["opt_fragment"]))
-                
-                prev_geom = current_geom_num_list.copy()
-                
-                current_geom_num_list_ang, LL_move_vector, LL_optimizer_instances = LL_CMV.calc_move_vector(
-                    microiter, current_geom_num_list, real_LL_B_g, pre_real_LL_B_g, 
-                    real_pre_geom, real_LL_B_e, pre_real_LL_B_e, 
-                    pre_real_LL_move_vector, real_initial_geom_num_list, 
-                    real_LL_g, pre_real_LL_g, LL_optimizer_instances, print_flag=False)
-                
-                current_geom_num_list = current_geom_num_list_ang / self.config.bohr2angstroms
 
-                # Fix high layer atoms
-                for key, value in highlayer_2_real_label_connect_dict.items():
-                    current_geom_num_list[value-1] = copy.deepcopy(high_layer_geom_num_list[key-1]) # Already in Bohr
-                
-                # Fix user-specified atoms
-                if len(force_data["fix_atoms"]) > 0:
-                    for j in force_data["fix_atoms"]:
-                        current_geom_num_list[j-1] = copy.deepcopy(real_initial_geom_num_list[j-1]) # Already in Bohr
-                
-                displacement_vector = current_geom_num_list - prev_geom
-                
-                # Calculate convergence metrics for low layer atoms only
-                low_layer_grads = []
-                low_layer_displacements = []
-                for i in range(len(element_list)):
-                    if (i+1) not in high_layer_atom_num:
-                        low_layer_grads.append(real_LL_B_g[i])
-                        low_layer_displacements.append(displacement_vector[i])
-                
-                low_layer_grads = np.array(low_layer_grads)
-                low_layer_displacements = np.array(low_layer_displacements)
-                
-                low_layer_rms_grad = self.calculate_rms_safely(low_layer_grads)
-                max_displacement = np.abs(displacement_vector).max() if len(displacement_vector) > 0 else 0
-                rms_displacement = self.calculate_rms_safely(displacement_vector)
-                energy_shift = -1 * pre_real_LL_B_e + real_LL_B_e
-            
-                if microiter % 10 == 0:
-                    print(f"M. ITR. {microiter}")
-                    print("Microiteration results:")
-                    print(f"LOW LAYER BIAS ENERGY : {float(real_LL_B_e):10.8f}")
-                    print(f"LOW LAYER ENERGY      : {float(real_LL_e):10.8f}")
-                    print(f"LOW LAYER MAX GRADIENT: {float(low_layer_grads.max() if len(low_layer_grads) > 0 else 0):10.8f}")
-                    print(f"LOW LAYER RMS GRADIENT: {float(low_layer_rms_grad):10.8f}")
-                    print(f"MAX DISPLACEMENT      : {float(max_displacement):10.8f}")
-                    print(f"RMS DISPLACEMENT      : {float(rms_displacement):10.8f}")
-                    print(f"ENERGY SHIFT          : {float(energy_shift):10.8f}")
-                
-                # Check convergence (using hardcoded values from original)
-                if (low_layer_rms_grad < 0.0003) and \
-                   (low_layer_grads.max() < 0.0006 if len(low_layer_grads) > 0 else True) and \
-                   (max_displacement < 0.003) and \
-                   (rms_displacement < 0.002):
-                    print("Low layer converged... (microiteration)")
-                    low_layer_converged = True
-                    break
-                
-                # Update previous values for next microiteration
-                pre_real_LL_B_e = real_LL_B_e
-                pre_real_LL_g = real_LL_g
-                pre_real_LL_B_g = real_LL_B_g
-                pre_real_LL_move_vector = LL_move_vector
-            
-            # End of microiteration loop
-            if not low_layer_converged:
-                print("Reached maximum number of microiterations.")
-            print("Microiteration complete.")
-            
-            # Update the main geometry state
-            self.state.geom_num_list = current_geom_num_list
-            geom_num_list = current_geom_num_list # Use for this iter
-            
-            # --- Model High Layer Calc ---
-            print("Model system (high layer)")
-            HLSP.Model_hess = HL_Model_hess
-            model_HL_e, model_HL_g, high_layer_geom_num_list, finish_frag = HLSP.single_point(
-                file_directory, high_layer_element_list, iter, electric_charge_and_multiplicity,
-                method="", geom_num_list=high_layer_geom_num_list*self.config.bohr2angstroms)
-            
-            HL_Model_hess = HLSP.Model_hess
-            
-            if finish_frag:
-                self.state.exit_flag = True
+            print(f"# ITR. {iter_idx}")
+
+            # Compute potentials
+            self.state = self.handler.compute(self.state)
+            if self.state.exit_flag:
                 break
-            
-            # --- Combine Gradients ---
-            # Use LL_Calc_BiasPot to get bias gradient on "Real" system
-            _, tmp_model_HL_B_e, tmp_model_HL_B_g, LL_BPA_hessian = LL_Calc_BiasPot.main(
-                0.0, real_LL_g*0.0, geom_num_list, element_list, force_data, pre_real_LL_B_g*0.0, iter, real_initial_geom_num_list)
-            
-            tmp_model_HL_g = tmp_model_HL_B_g * 0.0
-            
-            for key, value in real_2_highlayer_label_connect_dict.items():
-                tmp_model_HL_B_g[key-1] += model_HL_g[value-1] - model_LL_g[value-1]
-                tmp_model_HL_g[key-1] += model_HL_g[value-1] - model_LL_g[value-1]
-            
-            HL_BPA_hessian = LL_BPA_hessian[np.ix_(bool_list, bool_list)]
-            
-            for i in range(len(HL_optimizer_instances)):
-                HL_optimizer_instances[i].set_bias_hessian(HL_BPA_hessian)
-                if iter % self.config.FC_COUNT == 0:
-                    HL_optimizer_instances[i].set_hessian(HL_Model_hess[:len(high_layer_atom_num)*3, :len(high_layer_atom_num)*3])
-            
-            if len(force_data["opt_fragment"]) > 0:
-                tmp_model_HL_B_g = copy.deepcopy(self.calc_fragement_grads(tmp_model_HL_B_g, force_data["opt_fragment"]))
-                tmp_model_HL_g = copy.deepcopy(self.calc_fragement_grads(tmp_model_HL_g, force_data["opt_fragment"]))
-            
-            model_HL_B_g = copy.deepcopy(model_HL_g)
-            model_HL_B_e = model_HL_e + tmp_model_HL_B_e
-            
-            for key, value in real_2_highlayer_label_connect_dict.items():
-                model_HL_B_g[value-1] += tmp_model_HL_B_g[key-1] # This seems incorrect logic, but mirrors original
-            
-            pre_high_layer_geom_num_list = high_layer_geom_num_list
-            
-            # --- Calculate HL Move Vector ---
-            high_layer_geom_num_list_ang, move_vector, HL_optimizer_instances = HL_CMV.calc_move_vector(
-                iter, high_layer_geom_num_list[:len(high_layer_atom_num)], model_HL_B_g[:len(high_layer_atom_num)], pre_model_HL_B_g[:len(high_layer_atom_num)], 
-                pre_high_layer_geom_num_list[:len(high_layer_atom_num)], model_HL_B_e, pre_model_HL_B_e, 
-                pre_model_HL_move_vector[:len(high_layer_atom_num)], high_layer_pre_geom[:len(high_layer_atom_num)], 
-                model_HL_g[:len(high_layer_atom_num)], pre_model_HL_g[:len(high_layer_atom_num)], HL_optimizer_instances)
-            
-            high_layer_geom_num_list = high_layer_geom_num_list_ang / self.config.bohr2angstroms
-            
-            # --- Update Full System Geometry ---
-            for l in range(len(high_layer_geom_num_list) - len(linker_atom_pair_num)):
-                geom_num_list[highlayer_2_real_label_connect_dict[l+1]-1] = copy.deepcopy(high_layer_geom_num_list[l])
-            
-            # Project out translation and rotation
-            geom_num_list -= Calculationtools().calc_center_of_mass(geom_num_list, element_list)
-            geom_num_list, _ = Calculationtools().kabsch_algorithm(geom_num_list, real_pre_geom)
-            
-            # Update high layer geometry after alignment
-            high_layer_geom_num_list, high_layer_element_list = separate_high_layer_and_low_layer(
-                geom_num_list, linker_atom_pair_num, high_layer_atom_num, element_list)
-            
-            # --- Combine Energies and Gradients for REAL system ---
-            real_e = real_LL_e + model_HL_e - model_LL_e
-            real_B_e = real_LL_B_e + model_HL_B_e - model_LL_e # Original uses model_LL_e, not B_e
-            real_g = real_LL_g + tmp_model_HL_g
-            real_B_g = real_LL_B_g + tmp_model_HL_g
-            
-            # --- Update Main State ---
-            self.state.e = real_e
-            self.state.B_e = real_B_e
-            self.state.g = real_g
-            self.state.B_g = real_B_g
-            self.state.geom_num_list = geom_num_list
-            
-            self.save_tmp_energy_profiles(iter, real_e, real_g, real_B_g)
-            self.state.ENERGY_LIST_FOR_PLOTTING.append(real_e*self.config.hartree2kcalmol)
-            self.state.BIAS_ENERGY_LIST_FOR_PLOTTING.append(real_B_e*self.config.hartree2kcalmol)
-            self.state.NUM_LIST.append(iter)
-            
-            self.geom_info_extract(force_data, file_directory, real_B_g, real_g)
-            
-            if len(linker_atom_pair_num) > 0:
-                tmp_real_B_g = model_HL_B_g[:-len(linker_atom_pair_num)].reshape(-1,1)
-            else:
-                tmp_real_B_g = model_HL_B_g.reshape(-1,1)
-            
-            abjusted_high_layer_geom_num_list, _ = Calculationtools().kabsch_algorithm(high_layer_geom_num_list, pre_high_layer_geom_num_list)
-            
-            if len(linker_atom_pair_num) > 0:
-                tmp_displacement_vector = (abjusted_high_layer_geom_num_list - pre_high_layer_geom_num_list)[:-len(linker_atom_pair_num)].reshape(-1,1)
-            else:
-                tmp_displacement_vector = (abjusted_high_layer_geom_num_list - pre_high_layer_geom_num_list).reshape(-1,1)
-            
-            # --- Check Convergence (on HL model) ---
-            converge_flag, max_displacement_threshold, rms_displacement_threshold = self._check_converge_criteria(tmp_real_B_g, tmp_displacement_vector)
-            
-            self.print_info(real_e, real_B_e, tmp_real_B_g, tmp_displacement_vector, self.state.pre_e, self.state.pre_B_e, 
-                            max_displacement_threshold, rms_displacement_threshold)
-            
-            real_grad_list.append(self.calculate_rms_safely(real_g))
-            real_bias_grad_list.append(self.calculate_rms_safely(real_B_g))
-            
-            # Update state lists
-            self.state.grad_list = real_grad_list
-            self.state.bias_grad_list = real_bias_grad_list
-            
-            if converge_flag:
-                self.state.optimized_flag = True
-                print("\n=====================================================")
-                print("converged!!!")
-                print("=====================================================")
+
+            # Exact model Hessian update
+            if (
+                iter_idx % self.config.mFC_COUNT == 0
+                and self.config.use_model_hessian is not None
+                and self.config.FC_COUNT < 1
+            ):
+                self.state.Model_hess = ApproxHessian().main(
+                    self.state.geometry,
+                    self.state.element_list, # FIX: Use self.state.element_list
+                    self.state.raw_gradient,
+                    self.config.use_model_hessian,
+                )
+                if isinstance(self.handler, StandardHandler):
+                    self.handler.calculator.Model_hess = self.state.Model_hess
+
+            # Initial geometry save
+            if iter_idx == 0:
+                # FIX: Use self.state.element_list
+                initial_geom_num_list, pre_geom = self._save_init_geometry(
+                    self.state.geometry, self.state.element_list, allactive_flag
+                )
+                self.state.pre_geometry = pre_geom
+
+            # Build combined hessian
+            Hess = (
+                self.state.bias_hessian + self.state.Model_hess
+                if hasattr(self.state, "bias_hessian")
+                else self.state.Model_hess
+            )
+            if iter_idx == 0 and self.convergence.judge_early_stop_due_to_no_negative_eigenvalues(
+                self.state.geometry,
+                Hess,
+                self.config.args.saddle_order,
+                self.config.FC_COUNT,
+                self.config.detect_negative_eigenvalues,
+                self.BPA_FOLDER_DIRECTORY,
+            ):
                 break
-            
-            # Fix user-specified atoms
-            if len(force_data["fix_atoms"]) > 0:
-                for j in force_data["fix_atoms"]:
-                    geom_num_list[j-1] = copy.deepcopy(real_initial_geom_num_list[j-1]) # Bohr
-            
-            DC_exit_flag = self.dissociation_check(geom_num_list, element_list)
+
+            # Constraints / projection
+            PC = self.constraints.init_projection_constraint(
+                PC, self.state.geometry, iter_idx, projection_constrain, hessian=Hess
+            )
+            optimizer_instances = self.hessian_mgr.calc_eff_hess_for_fix_atoms_and_set_hess(
+                self.state,
+                allactive_flag,
+                force_data,
+                self.state.bias_hessian if hasattr(self.state, "bias_hessian") else np.zeros_like(Hess),
+                n_fix,
+                optimizer_instances,
+                self.state.geometry,
+                self.state.bias_gradient,
+                self.state.raw_gradient,
+                projection_constrain,
+                PC,
+            )
+
+            if not allactive_flag and len(force_data["opt_fragment"]) > 0:
+                self.state.bias_gradient = self.constraints.calc_fragment_grads(
+                    self.state.bias_gradient, force_data["opt_fragment"]
+                )
+                self.state.raw_gradient = self.constraints.calc_fragment_grads(
+                    self.state.raw_gradient, force_data["opt_fragment"]
+                )
+
+            # logging
+            self.logger.log_dynamic_csv(self.state, self.BPA_FOLDER_DIRECTORY, self.convergence)
+            self.state.ENERGY_LIST_FOR_PLOTTING.append(
+                self.state.raw_energy * self.config.hartree2kcalmol
+            )
+            self.state.BIAS_ENERGY_LIST_FOR_PLOTTING.append(
+                self.state.bias_energy * self.config.hartree2kcalmol
+            )
+            self.state.NUM_LIST.append(iter_idx)
+
+            # Geometry info extract
+            # FIX: Use self.state.element_list
+            self.logger.geom_info_extract(
+                self.state,
+                force_data,
+                self.file_io.make_psi4_input_file(
+                    self.file_io.print_geometry_list(
+                        self.state.geometry * self.config.bohr2angstroms,
+                        self.state.element_list,
+                        chg_mult, 
+                        display_flag=False
+                    ),
+                    iter_idx,
+                ),
+                self.state.bias_gradient,
+                self.state.raw_gradient,
+                self.BPA_FOLDER_DIRECTORY,
+            )
+
+            # Apply constraints to gradients
+            g = copy.deepcopy(self.state.raw_gradient)
+            B_g = copy.deepcopy(self.state.bias_gradient)
+            g, B_g, PC = self.constraints.apply_projection_constraints(
+                projection_constrain, PC, self.state.geometry, g, B_g
+            )
+            g, B_g = self.constraints.zero_fixed_atom_gradients(allactive_flag, force_data, g, B_g)
+            
+            self.state.raw_gradient = g
+            self.state.bias_gradient = B_g
+            self.state.effective_gradient = g + (B_g - g)
+
+            if self.config.koopman_analysis and KA is not None:
+                # FIX: Use self.state.element_list
+                _ = KA.run(iter_idx, self.state.geometry, B_g, self.state.element_list)
+
+            # Move vector
+            new_geometry, move_vector, optimizer_instances = CMV.calc_move_vector(
+                iter_idx,
+                self.state.geometry,
+                B_g,
+                self.state.pre_bias_gradient,
+                self.state.pre_geometry,
+                self.state.bias_energy,
+                self.state.pre_bias_energy,
+                self.state.pre_move_vector,
+                self.state.initial_geometry,
+                g,
+                self.state.pre_raw_gradient,
+                optimizer_instances,
+                projection_constrain,
+            )
+
+            # Projection / alignment
+            new_geometry = self.constraints.project_out_translation_rotation(
+                new_geometry, self.state.geometry, allactive_flag
+            )
+            new_geometry, PC = self.constraints.apply_projection_constraints_to_geometry(
+                projection_constrain, PC, new_geometry, hessian=Hess
+            )
+
+            displacement_vector = (
+                move_vector
+                if iter_idx == 0
+                else new_geometry / self.config.bohr2angstroms - self.state.geometry
+            )
+
+            # convergence
+            converge_flag, max_disp_th, rms_disp_th = self.convergence.check_convergence(
+                self.state, displacement_vector, optimizer_instances
+            )
+
+            # track gradients
+            self.state.grad_list.append(self.convergence.calculate_rms_safely(g))
+            self.state.bias_grad_list.append(self.convergence.calculate_rms_safely(B_g))
+
+            # reset fixed atoms
+            new_geometry = self.constraints.reset_fixed_atom_positions(
+                new_geometry, self.state.initial_geometry, allactive_flag, force_data
+            )
+
+            # dissociation
+            # FIX: Use self.state.element_list
+            DC_exit_flag = self.dissociation_check(new_geometry, self.state.element_list)
+
+            # print info
+            self._print_info(
+                self.state.raw_energy,
+                self.state.bias_energy,
+                B_g,
+                displacement_vector,
+                self.state.pre_raw_energy,
+                self.state.pre_bias_energy,
+                max_disp_th,
+                rms_disp_th,
+            )
+
+            if converge_flag:
+                if projection_constrain and iter_idx == 0:
+                    pass
+                else:
+                    self.state.optimized_flag = True
+                    print("\n=====================================================")
+                    print("converged!!!")
+                    print("=====================================================")
+                    break
+
             if DC_exit_flag:
                 self.state.DC_check_flag = True
                 break
-            
-            # --- Update Previous State Variables ---
-            self.state.pre_B_e = real_B_e
-            self.state.pre_e = real_e
-            self.state.pre_geom = geom_num_list
-            real_pre_geom = geom_num_list # Update local var
-            
-            pre_model_HL_B_g = model_HL_B_g
-            pre_model_HL_g = model_HL_g
-            pre_model_HL_B_e = model_HL_B_e
-            pre_model_HL_move_vector = move_vector
-            
-            # Create input for next iteration
-            geometry_list = FIO.print_geometry_list(geom_num_list*self.config.bohr2angstroms, element_list, electric_charge_and_multiplicity)
-            file_directory = FIO.make_psi4_input_file(geometry_list, iter+1)
-        
-        else: # Loop finished without break
+
+            # update previous
+            self.state.pre_bias_energy = self.state.bias_energy
+            self.state.pre_raw_energy = self.state.raw_energy
+            self.state.pre_bias_gradient = B_g
+            self.state.pre_raw_gradient = g
+            self.state.pre_geometry = self.state.geometry
+            self.state.pre_move_vector = move_vector
+            self.state.geometry = new_geometry / self.config.bohr2angstroms
+
+            # write next input
+            # FIX: Use self.state.element_list
+            self.file_io.print_geometry_list(
+                new_geometry, self.state.element_list, chg_mult, display_flag=False
+            )
+
+        else:
             self.state.optimized_flag = False
             print("Reached maximum number of iterations. This is not converged.")
-            with open(self.BPA_FOLDER_DIRECTORY+"not_converged.txt", "w") as f:
+            with open(
+                self.BPA_FOLDER_DIRECTORY + "not_converged.txt", "w"
+            ) as f:
                 f.write("Reached maximum number of iterations. This is not converged.")
-        
+
+        # Post steps
         if self.state.DC_check_flag:
-            with open(self.BPA_FOLDER_DIRECTORY+"dissociation_is_detected.txt", "w") as f:
-                f.write("These molecules are dissociated.")
-        
-        # --- 12. Finalize and Save Results ---
-        G = Graph(self.BPA_FOLDER_DIRECTORY)
-        
-        # Finalize plots and save results (using self.state lists)
-        self._finalize_optimization(FIO, G, self.state.grad_list, self.state.bias_grad_list, file_directory, force_data, geom_num_list, e, B_e, SP, self.state.exit_flag) # Pass LLSP as dummy
-        
-        # Copy final results from state to self
+            print("Dissociation is detected. Optimization stopped.")
+            with open(
+                self.BPA_FOLDER_DIRECTORY + "dissociation_is_detected.txt", "w"
+            ) as f:
+                f.write("Dissociation is detected. Optimization stopped.")
+
+        # Vibrational analysis
+        if self.config.freq_analysis and not self.state.exit_flag and not self.state.DC_check_flag:
+            self._perform_vibrational_analysis(
+                self.SP,
+                self.state.geometry,
+                self.state.element_list, # FIX: Use self.state.element_list
+                self.state.initial_geometry,
+                force_data,
+                self._is_exact_hessian(iter_idx),
+                self.file_io.make_psi4_input_file(
+                    self.file_io.print_geometry_list(
+                        self.state.geometry * self.config.bohr2angstroms,
+                        self.state.element_list, # FIX: Use self.state.element_list
+                        chg_mult,
+                    ),
+                    iter_idx,
+                ),
+                iter_idx,
+                chg_mult,
+                self.SP.xtb_method if hasattr(self.SP, "xtb_method") else None,
+                self.state.raw_energy,
+            )
+
+        # Finalize
+        # FIX: Use self.state.element_list
+        self._finalize_optimization(
+            self.file_io,
+            Graph(self.BPA_FOLDER_DIRECTORY),
+            self.state.grad_list,
+            self.state.bias_grad_list,
+            self.file_io.make_psi4_input_file(
+                self.file_io.print_geometry_list(
+                    self.state.geometry * self.config.bohr2angstroms,
+                    self.state.element_list,
+                    chg_mult,
+                ),
+                iter_idx,
+            ),
+            force_data,
+            self.state.geometry,
+            self.state.raw_energy,
+            self.state.bias_energy,
+            self.SP,
+            self.state.exit_flag,
+        )
         self._copy_final_results_from_state()
-        return
-        
-    def get_result_file_path(self):
-        """
-        Sets the absolute file paths for optimization results.
-        Relies on self.BPA_FOLDER_DIRECTORY and self.START_FILE
-        which are set by _make_init_directory().
-        """
-        try:
-            if (hasattr(self, 'BPA_FOLDER_DIRECTORY') and self.BPA_FOLDER_DIRECTORY and
-                hasattr(self, 'START_FILE') and self.START_FILE):
-                
-                base_name = os.path.splitext(os.path.basename(self.START_FILE))[0]
-                optimized_filename = f"{base_name}_optimized.xyz"
-                traj_filename = f"{base_name}_traj.xyz"
 
-                self.optimized_struct_file = os.path.abspath(os.path.join(self.BPA_FOLDER_DIRECTORY, optimized_filename))
-                self.traj_file = os.path.abspath(os.path.join(self.BPA_FOLDER_DIRECTORY, traj_filename))
-                
-                print("Optimized structure file path:", self.optimized_struct_file)
-                print("Trajectory file path:", self.traj_file)
-            
+        # Analyses
+        if self.state and self.config.CMDS:
+            CMDPA = CMDSPathAnalysis(
+                self.BPA_FOLDER_DIRECTORY,
+                self.state.ENERGY_LIST_FOR_PLOTTING,
+                self.state.BIAS_ENERGY_LIST_FOR_PLOTTING,
+            )
+            CMDPA.main()
+        if self.state and self.config.PCA:
+            PCAPA = PCAPathAnalysis(
+                self.BPA_FOLDER_DIRECTORY,
+                self.state.ENERGY_LIST_FOR_PLOTTING,
+                self.state.BIAS_ENERGY_LIST_FOR_PLOTTING,
+            )
+            PCAPA.main()
+
+        if self.state and len(self.config.irc) > 0:
+            if self.config.args.usextb != "None":
+                xtb_method = self.config.args.usextb
             else:
-                print("Error: BPA_FOLDER_DIRECTORY or START_FILE is not set. Please run optimize() or optimize_oniom() first.")
-                self.optimized_struct_file = None
-                self.traj_file = None
-                
-        except Exception as e:
-            print(f"Error setting result file paths: {e}")
-            self.optimized_struct_file = None
-            self.traj_file = None
+                xtb_method = "None"
+            if self.state.iter % self.config.FC_COUNT == 0:
+                hessian = self.state.Model_hess
+            else:
+                hessian = None
+            
+            # NOTE: IRC logic for BITSS might need special handling, but keeping standard
+            EXEC_IRC = IRC(
+                self.BPA_FOLDER_DIRECTORY,
+                self.state.final_file_directory,
+                self.config.irc,
+                self.SP,
+                self.state.element_list, # FIX: Use self.state.element_list
+                self.config.electric_charge_and_multiplicity,
+                force_data_parser(self.config.args),
+                xtb_method,
+                FC_count=int(self.config.FC_COUNT),
+                hessian=hessian,
+            )
+            EXEC_IRC.run()
+            self.irc_terminal_struct_paths = EXEC_IRC.terminal_struct_paths
+        else:
+            self.irc_terminal_struct_paths = []
 
-        return
+        print(f"Trial of geometry optimization ({file}) was completed.")
+        
+    # ------------------------------------------------------------------
+    # Secondary helpers reused from legacy
+    # ------------------------------------------------------------------
+    def _save_init_geometry(self, geom_num_list, element_list, allactive_flag):
+        if allactive_flag:
+            initial_geom_num_list = geom_num_list - Calculationtools().calc_center(
+                geom_num_list, element_list
+            )
+            pre_geom = initial_geom_num_list - Calculationtools().calc_center(
+                geom_num_list, element_list
+            )
+        else:
+            initial_geom_num_list = geom_num_list
+            pre_geom = initial_geom_num_list
+        return initial_geom_num_list, pre_geom
 
-    def run(self):
-        # (Reads self.config)
-        if type(self.config.args.INPUT) is str:
-            START_FILE_LIST = [self.config.args.INPUT]
+    def dissociation_check(self, new_geometry, element_list):
+        atom_label_list = list(range(len(new_geometry)))
+        fragm_atom_num_list = []
+
+        while len(atom_label_list) > 0:
+            tmp_fragm_list = Calculationtools().check_atom_connectivity(
+                new_geometry, element_list, atom_label_list[0]
+            )
+            atom_label_list = list(set(atom_label_list) - set(tmp_fragm_list))
+            fragm_atom_num_list.append(tmp_fragm_list)
+
+        if len(fragm_atom_num_list) > 1:
+            fragm_dist_list = []
+            geom_np = np.asarray(new_geometry)
+            for fragm_1_indices, fragm_2_indices in itertools.combinations(
+                fragm_atom_num_list, 2
+            ):
+                coords1 = geom_np[fragm_1_indices]
+                coords2 = geom_np[fragm_2_indices]
+                diff_matrix = coords1[:, np.newaxis, :] - coords2[np.newaxis, :, :]
+                sq_dist_matrix = np.sum(diff_matrix ** 2, axis=2)
+                min_sq_dist = np.min(sq_dist_matrix)
+                min_dist = np.sqrt(min_sq_dist)
+                fragm_dist_list.append(min_dist)
+
+            min_interfragment_dist = min(fragm_dist_list)
+            if min_interfragment_dist > self.config.DC_check_dist:
+                print(
+                    f"Minimum fragment distance (ang.) {min_interfragment_dist:.4f} > {self.config.DC_check_dist}"
+                )
+                print("These molecules are dissociated.")
+                return True
+        return False
+
+    def _perform_vibrational_analysis(
+        self,
+        SP,
+        geom_num_list,
+        element_list,
+        initial_geom_num_list,
+        force_data,
+        exact_hess_flag,
+        file_directory,
+        iter_idx,
+        electric_charge_and_multiplicity,
+        xtb_method,
+        e,
+    ):
+        print("\n====================================================")
+        print("Performing vibrational analysis...")
+        print("====================================================\n")
+        print("Is Exact Hessian calculated? : ", exact_hess_flag)
+
+        if exact_hess_flag:
+            g = np.zeros_like(geom_num_list, dtype="float64")
+            exit_flag = False
         else:
-            START_FILE_LIST = self.config.args.INPUT
-        
-        job_file_list = []
-        
-        for job_file in START_FILE_LIST:
-            print()
-            if "*" in job_file:
-                result_list = glob.glob(job_file)
-                job_file_list = job_file_list + result_list
-            else:
-                job_file_list = job_file_list + [job_file]
-        
-        for file in job_file_list:
-            print("********************************")
-            print(file)
-            print("********************************")
-            if os.path.exists(file) == False:
-                print(f"{file} does not exist.")
-                continue
-            
-            # This creates the directory and sets self.START_FILE
-            self._make_init_directory(file)
-            
-            # Run the main optimization, which will create its own state
-            if len(self.config.args.oniom_flag) > 0:
-                self.optimize_oniom()
-            else:
-                self.optimize()
-            
-            # Post-processing (relies on self.state being set by optimize())
-            if self.state and self.config.CMDS:
-                CMDPA = CMDSPathAnalysis(self.BPA_FOLDER_DIRECTORY, self.state.ENERGY_LIST_FOR_PLOTTING, self.state.BIAS_ENERGY_LIST_FOR_PLOTTING)
-                CMDPA.main()
-            if self.state and self.config.PCA:
-                PCAPA = PCAPathAnalysis(self.BPA_FOLDER_DIRECTORY, self.state.ENERGY_LIST_FOR_PLOTTING, self.state.BIAS_ENERGY_LIST_FOR_PLOTTING)
-                PCAPA.main()
-            
-            if self.state and len(self.config.irc) > 0:
-                if self.config.args.usextb != "None":
-                    xtb_method = self.config.args.usextb
-                else:
-                    xtb_method = "None"
-                
-                if self.state.iter % self.config.FC_COUNT == 0:
-                    hessian = self.state.Model_hess
-                else:
-                    hessian = None
-                    
-                EXEC_IRC = IRC(self.BPA_FOLDER_DIRECTORY, self.state.final_file_directory, 
-                               self.config.irc, self.SP, self.element_list, 
-                               self.config.electric_charge_and_multiplicity, # This was from config
-                               force_data_parser(self.config.args), # Re-parse force_data
-                               xtb_method, FC_count=int(self.config.FC_COUNT), hessian=hessian) 
-                EXEC_IRC.run()
-                self.irc_terminal_struct_paths = EXEC_IRC.terminal_struct_paths
-            else:
-                self.irc_terminal_struct_paths = []
+            print("Calculate exact Hessian...")
+            SP.hessian_flag = True
+            e, g, geom_num_list, exit_flag = SP.single_point(
+                file_directory, element_list, iter_idx, electric_charge_and_multiplicity, xtb_method
+            )
+            SP.hessian_flag = False
+
+        if exit_flag:
+            print("Error: QM calculation failed.")
+            return
+
+        _, B_e, _, BPA_hessian = self.handler.bias_pot_calc.main(
+            e,
+            g,
+            geom_num_list,
+            element_list,
+            force_data,
+            pre_B_g="",
+            iter=iter_idx,
+            initial_geom_num_list="",
+        )
+        tmp_hess = copy.deepcopy(SP.Model_hess)
+        tmp_hess += BPA_hessian
+
+        MV = MolecularVibrations(
+            atoms=element_list, coordinates=geom_num_list, hessian=tmp_hess
+        )
+        MV.calculate_thermochemistry(
+            e_tot=B_e,
+            temperature=self.config.thermo_temperature,
+            pressure=self.config.thermo_pressure,
+        )
+        MV.print_thermochemistry(
+            output_file=self.BPA_FOLDER_DIRECTORY + "/thermochemistry.txt"
+        )
+        MV.print_normal_modes(
+            output_file=self.BPA_FOLDER_DIRECTORY + "/normal_modes.txt"
+        )
+        MV.create_vibration_animation(
+            output_dir=self.BPA_FOLDER_DIRECTORY + "/vibration_animation"
+        )
+
+        if not self.state.optimized_flag:
+            print(
+                "Warning: Vibrational analysis was performed, but the optimization did not converge. The result of thermochemistry is unreliable."
+            )
+
+    def _is_exact_hessian(self, iter_idx):
+        if self.config.FC_COUNT == -1:
+            return False
+        elif iter_idx % self.config.FC_COUNT == 0 and self.config.FC_COUNT > 0:
+            return True
+        return False
+
+    def _finalize_optimization(
+        self,
+        FIO,
+        G,
+        grad_list,
+        bias_grad_list,
+        file_directory,
+        force_data,
+        geom_num_list,
+        e,
+        B_e,
+        SP,
+        exit_flag,
+    ):
+        
+        G.double_plot(
+            self.state.NUM_LIST,
+            self.state.ENERGY_LIST_FOR_PLOTTING,
+            self.state.BIAS_ENERGY_LIST_FOR_PLOTTING,
+        )
+        G.single_plot(
+            self.state.NUM_LIST,
+            grad_list,
+            file_directory,
+            "",
+            axis_name_2="gradient (RMS) [a.u.]",
+            name="gradient",
+        )
+        G.single_plot(
+            self.state.NUM_LIST,
+            bias_grad_list,
+            file_directory,
+            "",
+            axis_name_2="bias gradient (RMS) [a.u.]",
+            name="bias_gradient",
+        )
+
+        if len(force_data["geom_info"]) > 1:
+            for num, i in enumerate(force_data["geom_info"]):
+                G.single_plot(self.state.NUM_LIST, self.state.cos_list[num], file_directory, i)
+
+        FIO.make_traj_file()
+        FIO.argrelextrema_txt_save(self.state.ENERGY_LIST_FOR_PLOTTING, "approx_TS", "max")
+        FIO.argrelextrema_txt_save(self.state.ENERGY_LIST_FOR_PLOTTING, "approx_EQ", "min")
+        FIO.argrelextrema_txt_save(grad_list, "local_min_grad", "min")
+
+        self.logger.save_energy_profiles(self.state, self.BPA_FOLDER_DIRECTORY)
+
+        self.state.bias_pot_params_grad_list = self.handler.bias_pot_calc.bias_pot_params_grad_list
+        self.state.bias_pot_params_grad_name_list = self.handler.bias_pot_calc.bias_pot_params_grad_name_list
+        self.state.final_file_directory = file_directory
+        self.state.final_geometry = geom_num_list
+        self.state.final_energy = e
+        self.state.final_bias_energy = B_e
+
+        if not exit_flag:
+            self.symmetry = analyze_symmetry(self.element_list, self.state.final_geometry)
+            self.state.symmetry = self.symmetry
+            with open(self.BPA_FOLDER_DIRECTORY + "symmetry.txt", "w") as f:
+                f.write(f"Symmetry of final structure: {self.symmetry}")
+            print(f"Symmetry: {self.symmetry}")
+
+        if isinstance(self.handler, ModelFunctionHandler) and self.handler.is_bitss:
+            # We need original single-N element list for writing single frames
+            # But state.element_list is doubled.
+            # Use cached or slice.
+            # The handler.finalize_bitss_trajectory needs access to single element list
+            single_elem_len = len(self.state.element_list) // 2
+            real_elems = self.state.element_list[:single_elem_len]
+            self.config.args.element_list_cache = real_elems # Ensure correct list is used
             
-            print(f"Trial of geometry optimization ({file}) was completed.")
-        
-        print("All calculations were completed.")
-        
-        self.get_result_file_path()
-        return
+          
+            self.handler.finalize_bitss_trajectory()
+
+    def _copy_final_results_from_state(self):
+        if self.state:
+            self.final_file_directory = self.state.final_file_directory
+            self.final_geometry = self.state.final_geometry
+            self.final_energy = self.state.final_energy
+            self.final_bias_energy = self.state.final_bias_energy
+            self.symmetry = getattr(self.state, "symmetry", None)
+            self.bias_pot_params_grad_list = self.state.bias_pot_params_grad_list
+            self.bias_pot_params_grad_name_list = self.state.bias_pot_params_grad_name_list
+            self.optimized_flag = self.state.optimized_flag
+
+    def geom_info_extract(self, force_data, file_directory, B_g, g):
+        # kept for backward compatibility; delegate to logger
+        return self.logger.geom_info_extract(self.state, force_data, file_directory, B_g, g, self.BPA_FOLDER_DIRECTORY)
+
+    def _print_info(
+        self,
+        e,
+        B_e,
+        B_g,
+        displacement_vector,
+        pre_e,
+        pre_B_e,
+        max_displacement_threshold,
+        rms_displacement_threshold,
+    ):
+        rms_force = self.convergence.calculate_rms_safely(np.abs(B_g))
+        rms_displacement = self.convergence.calculate_rms_safely(np.abs(displacement_vector))
+        max_B_g = np.abs(B_g).max()
+        max_displacement = np.abs(displacement_vector).max()
+        print("caluculation results (unit a.u.):")
+        print("                         Value                     Threshold ")
+        print("ENERGY                 : {:>15.12f} ".format(e))
+        print("BIAS  ENERGY           : {:>15.12f} ".format(B_e))
+        print(
+            "Maximum  Force         : {0:>15.12f}                 {1:>15.12f} ".format(
+                max_B_g, self.config.MAX_FORCE_THRESHOLD
+            )
+        )
+        print(
+            "RMS      Force         : {0:>15.12f}                 {1:>15.12f} ".format(
+                rms_force, self.config.RMS_FORCE_THRESHOLD
+            )
+        )
+        print(
+            "Maximum  Displacement  : {0:>15.12f}                 {1:>15.12f} ".format(
+                max_displacement, max_displacement_threshold
+            )
+        )
+        print(
+            "RMS      Displacement  : {0:>15.12f}                 {1:>15.12f} ".format(
+                rms_displacement, rms_displacement_threshold
+            )
+        )
+        print("ENERGY SHIFT           : {:>15.12f} ".format(e - pre_e))
+        print("BIAS ENERGY SHIFT      : {:>15.12f} ".format(B_e - pre_B_e))
+
+    def get_result_file_path(self):
+        self.optimized_struct_file, self.traj_file = self.result_paths.get_result_file_path(
+            self.BPA_FOLDER_DIRECTORY, self.START_FILE
+        )
+
+if __name__ == "__main__":
+    # The actual CLI interface is expected to supply args.
+    pass