Ground and low-lying excited state potential energy surfaces of diiodomethane in four dimensions

This paper presents accurate, four-dimensional adiabatic potential energy surfaces for diiodomethane's ground and low-lying excited states, constructed via spline interpolation to enable detailed molecular dynamics simulations of its photochemical dissociation and rearrangement pathways near 260 nm.

The Gist

Imagine a molecule of diiodomethane ($CH_2I_2$) as a tiny, three-armed dancer. It has a central carbon "body" holding two hydrogen "hands" and two heavy iodine "feet."

This paper is essentially a high-definition 3D map (actually, a 4D map!) of the dance floor this molecule performs on. The author, Yijue Ding, has created a detailed guide to show exactly how this dancer moves, breaks apart, or rearranges itself when hit by a flash of ultraviolet light.

Here is the breakdown of the research using simple analogies:

1. The Problem: The Dance Floor is Too Complex

In the real world, molecules wiggle in many directions at once. Trying to map every single wiggle of every atom is like trying to draw a map of a city while accounting for every single person's heartbeat. It's too messy and too expensive to calculate.

The Solution: The author decided to simplify the dance. He realized that when this molecule gets hit by light, the most important moves are:

• The two iodine "feet" stretching away from the body.
• The angles at which those feet swing.
• The "body" (the $CH_2$ part) mostly just stays stiff and watches.

By freezing the less important wiggles, the author reduced the problem from a chaotic 6-dimensional mess down to a manageable 4-dimensional map. Think of it as turning a complex, swirling tornado into a clear, straight highway to understand the traffic flow.

2. The Map-Making: Building the Terrain

To build this map, the author didn't just guess. He used a super-powerful computer method (called MS-CASPT2) that acts like a high-precision GPS.

• The Terrain: The map shows "hills" (high energy, unstable) and "valleys" (low energy, stable).
• The Light: When the molecule absorbs a UV photon (like a spark of energy), it jumps up a hill.
• The Spin: Because iodine is a heavy atom, it has a "spin" (like a spinning top). The author had to account for this spin, which splits the energy levels, making the map more like a multi-layered cake than a flat sheet.

3. What the Map Reveals: The Three Main Stories

The author mapped out 17 different "dance floors" (electronic states), but focused on the four most important ones. Here is what happens on them:

• The Ground State (The Calm Valley):
  This is where the molecule sits when it's resting. The map shows a deep valley where the molecule is happy. However, there is a tricky "bump" (a barrier) nearby. If the molecule gets enough energy, it can roll over this bump, break one leg off, and form a weird temporary shape called an isomer (like a dancer doing a handstand before falling). But this handstand is unstable; it quickly collapses, and the molecule breaks apart into a $CH_2I$ piece and a free Iodine atom.

• The Excited States (The Slippery Slopes):
  When the molecule absorbs light, it jumps to higher floors.

  • The Repulsive Slopes: On some floors, the map is just a steep, slippery slide. As soon as the molecule gets there, it slides down instantly, snapping a bond and flying apart. This explains why the molecule breaks so fast.
  • The Shallow Pools: On other floors, there are tiny, shallow puddles (local minima). The molecule might get stuck there for a split second, wobbling around, before finding a way out to break apart.

4. Why This Matters: Predicting the Future

Why go through all this trouble to draw a map?

• The Simulation: Now that the map exists, scientists can run computer simulations (like a video game) to watch thousands of these molecules "dance" in real-time. They can see exactly how and when the molecule breaks apart.
• The Experiment: This map helps explain real-world experiments where scientists use lasers to take "snapshots" of molecules breaking. The map acts as the script that tells the scientists what they should be seeing in their photos.

The Big Picture Analogy

Imagine you are trying to understand how a specific type of car crashes.

• Old way: You crash one car, take a photo, and guess what happened.
• This paper's way: The author built a perfect, physics-based crash test simulator. He mapped out every possible angle the car could hit a wall, every speed, and every way the metal could bend.

Now, instead of just guessing, we can run the simulation a million times to see that 90% of the time, the car flips left, and 10% of the time, it slides right. This paper provides the "crash test simulator" for the diiodomethane molecule, helping us understand the invisible, ultra-fast world of chemistry with crystal-clear precision.

Technical Summary

1. Problem Statement

Diiodomethane ($CH_2I_2$) is a prototypical molecule for studying ultrafast reaction dynamics and photochemistry due to its structural simplicity and complex dissociation pathways. Upon UV photoexcitation (near 260 nm), $CH_2I_2$ undergoes C–I bond cleavage, leading to competing dissociation channels correlated with spin–orbit split atomic iodine states ($I(^2P_{3/2})$ and $I^*(^2P_{1/2})$). It also exhibits potential photoisomerization to a $CH_2I-I$ intermediate.

Despite extensive experimental studies using ultrafast electron diffraction (UED), time-resolved photoelectron spectroscopy (TRPES), and Coulomb explosion imaging (CEI), a comprehensive theoretical framework has been missing. Specifically, no accurate, pre-constructed potential energy surfaces (PESs) existed for the ground or low-lying excited states of $CH_2I_2$. While ab initio molecular dynamics (AIMD) is often used, it is computationally expensive for simulating rare events or statistically significant reaction channels. The lack of high-fidelity PESs hinders the ability to perform efficient molecular dynamics (MD) simulations to interpret experimental data and map complex reaction pathways involving non-adiabatic transitions.

2. Methodology

Electronic Structure Calculations

• Method: The authors employed Multi-State Complete Active Space Second-Order Perturbation Theory (MS-CASPT2). This method was chosen over single-reference approaches (like CCSD) and static correlation-only methods (SA-MCSCF) because it accurately captures both static and dynamic electron correlations essential for bond breaking and formation.
• Active Space: A (12, 8) active space was used, comprising 12 electrons in 8 orbitals. This includes the C–I bonding ($\sigma$) and anti-bonding ($\sigma^*$) orbitals for both C–I bonds, plus the remaining 5p orbitals of the two iodine atoms. This space is necessary to describe both C–I bond breaking and I–I bond formation.
• Spin-Orbit Coupling (SOC): Given the heavy iodine atoms, SOC effects were critical. A state-interaction approach was used, where the Hamiltonian ($\hat{H}_{el} + \hat{H}_{SO}$) was diagonalized using a basis of SA-MCSCF wavefunctions, but with diagonal elements replaced by MS-CASPT2 energies. This yielded 17 spin-orbit coupled states (ground state + 16 excited states).
• Basis Sets: Dunning's correlation-consistent double-zeta basis sets (cc-pVDZ) were used for C and H. For Iodine, a relativistic pseudo-potential (cc-pVDZ-PP) replaced the inner 28 core electrons.

Dimensionality Reduction and Coordinate Selection

To make the PES construction computationally feasible while retaining physical accuracy, the system was reduced to four dimensions (4D):

1. $R_1, R_2$: The two C–I bond lengths.
2. $\alpha_1, \alpha_2$: The angles between the C–I vectors and the $C_2$ axis of the $CH_2$ group.

• Constraints: The $CH_2$ moiety was treated as a rigid body (fixed C–H lengths and H–C–H angle), and the $CI_2$ plane was constrained to remain perpendicular to the $CH_2$ plane, preserving $C_s$ symmetry. This approximation is valid because $CH_2$ vibrational frequencies are much higher than C–I modes.

PES Construction

• Grid: Over 70,000 ab initio energy points were computed on a structured grid within the relevant configuration space ($R_{max} < 5.8$ Å, $R_{II} < 5$ Å).
• Interpolation: A multivariate spline interpolation algorithm was extended to 4D. This algorithm ensures:
  • $C_1$ Continuity: Smooth first-order derivatives (forces) are guaranteed, which is critical for MD simulations.
  • Permutation Invariance: The surface respects the symmetry $V(R_1, \alpha_1, R_2, \alpha_2) = V(R_2, \alpha_2, R_1, \alpha_1)$.
  • Local Feature Resolution: The method accurately captures avoided crossings without the need for diabatization.

3. Key Contributions

1. First 4D PESs for $CH_2I_2$: The paper presents the first set of adiabatic PESs covering the ground state and all 16 low-lying excited states accessible via single-photon absorption near 260 nm.
2. High-Fidelity Interpolation: The development and application of a 4D spline interpolation algorithm that reproduces ab initio data with high accuracy (RMSE < 20 meV for most states) while providing smooth forces.
3. Inclusion of Spin-Orbit Coupling: The PESs explicitly include SOC effects, resolving the splitting between $I(^2P_{3/2})$ and $I^*(^2P_{1/2})$ channels, which is crucial for understanding the branching ratios of dissociation products.
4. Mapping Reaction Pathways: Identification of stationary points (global minima, local minima, transition states) and the mapping of adiabatic minimum energy paths (MEPs) for dissociation and isomerization.

4. Results

Spectroscopic Properties

• The calculated absorption spectrum (3.4–5.3 eV) shows good agreement with experimental data, reproducing the first broad absorption peak.
• Discrepancies in the second peak suggest limitations in describing higher excited states or transition dipole moments, particularly for states with large oscillator strengths ($9\Gamma_1$ and $8\Gamma_2$).
• The 17 states are grouped by dissociation asymptotes:
  • Group 1: $CH_2I + I$ (Dissociative/Repulsive)
  • Group 2: $CH_2I + I^*$ (Dissociative/Repulsive)
  • Group 3: $CH_2I^* + I$ (Adiabatically Bound)

Quality of PESs

• Accuracy: Testing on 3,000 random points showed Root Mean Square Errors (RMSE) of 5–10 meV for the ground state and key excited states ($4\Gamma_1, 5\Gamma_1$). The $7\Gamma_1$ state had a slightly higher RMSE (29 meV) overall, but within the energetically accessible region (< 4.76 eV), the error dropped to 18 meV.
• Derivatives: The first-order derivatives (forces) are smooth, except for minor unphysical wiggles in specific regions due to noise in the underlying ab initio data. However, these are not expected to significantly impact dynamics.

Stationary Points and Dynamics

• Ground State ($1\Gamma_1$):
  • Global Minimum (GM) at equilibrium geometry.
  • A Local Minimum (LM) corresponding to the $CH_2I-I$ isomer.
  • A Transition State (TS) separates the GM and LM. The TS energy is slightly above the $CH_2I + I$ dissociation threshold, implying the isomer has a very short lifetime and the isomerization path leads directly to dissociation.
• Excited States ($4\Gamma_1, 5\Gamma_1$):
  • These states are strongly repulsive, leading to rapid $CH_2I + I$ and $CH_2I + I^*$ dissociation.
  • Shallow local minima exist near the Franck-Condon point, but no stable isomers are found along the dissociation path.
• Excited State ($7\Gamma_1$):
  • This state is adiabatically bound with a deep local minimum (~1.5 eV below the $CH_2I^* + I$ threshold).
  • No adiabatic dissociation pathway is accessible via single-photon absorption; dissociation would require non-adiabatic transitions to repulsive states.
• Avoided Crossings: Significant avoided crossings occur in the $R_1 = 2.0–2.6$ Å region, indicating strong non-adiabatic coupling. This suggests that molecules on bound excited states (Group 3) can transition to repulsive states, enabling delayed dissociation.

5. Significance

• Resource for Dynamics: These PESs provide a computationally efficient and accurate foundation for future molecular dynamics simulations (e.g., surface hopping) of $CH_2I_2$ photodissociation. This allows for the statistical sampling of rare reaction channels that are too costly for direct AIMD.
• Experimental Interpretation: The surfaces enable detailed interpretation of ultrafast experimental data (UED, CEI, TRPES), particularly regarding the competition between direct dissociation and photoisomerization.
• Benchmarking: The study serves as a benchmark for coupled electronic and nuclear motion in polyatomic systems with heavy atoms, demonstrating the necessity of including SOC and multi-reference character in photochemical modeling.
• Future Work: The author plans to use these surfaces to simulate $CH_2I_2$ photodissociation dynamics in conjunction with Coulomb Explosion Imaging experiments to specifically investigate the photoisomerization mechanism.