Scaling Dependencies in Irradiation-Driven Molecular Dynamics Simulations: Case Study of W(CO)$_6$ Fragmentation

This study utilizes irradiation-driven molecular dynamics simulations to investigate the electron-induced fragmentation of W(CO)$_6$ precursors, revealing how precursor density and electron fluence govern the formation of tungsten-rich clusters and establishing scaling relations to optimize simulation parameters for focused electron beam-induced deposition.

The Gist

The Big Picture: Building Tiny 3D Structures with Light

Imagine you are an architect trying to build a microscopic skyscraper, but you don't have a crane or a hammer. Instead, you have a super-precise "laser pointer" made of electrons. This technique is called FEBID (Focused Electron Beam-Induced Deposition).

To build your skyscraper, you spray a fog of special gas molecules (called precursors) into the air. In this study, the gas is W(CO)₆ (Tungsten Hexacarbonyl). Think of these molecules as little "Tungsten balls" wrapped in six "CO balloons."

When you shine your electron beam on the gas, the electrons hit the molecules, pop the balloons, and leave behind the Tungsten balls. These balls then stick together to form your tiny metal structure.

The Problem: How Fast Should We Shoot?

The scientists wanted to know: Does it matter how we deliver the electron beam?

Imagine you need to knock down a wall. You have two options:

1. Option A: A slow, gentle stream of water hitting the wall for a long time.
2. Option B: A massive, high-pressure firehose hitting the wall for a very short time.

If both options deliver the exact same total amount of water (the same "electron fluence"), will the wall fall down the same way? Or does the speed of the water change the outcome?

The researchers used a super-computer simulation (called IDMD) to answer this. They didn't just look at one molecule; they looked at a whole crowd of them (up to 207 molecules) to see how they interacted.

The Experiment: The "Demolition Derby"

The team ran simulations with three different "crowd densities":

• Low Density: A few molecules scattered in a large room (23 molecules).
• Medium Density: A moderate crowd (107 molecules).
• High Density: A packed mosh pit (207 molecules).

They hit these crowds with electrons, changing the "beam current" (how hard the electrons hit) and the "time" (how long they hit), while keeping the total energy delivered constant.

The Findings: What Happened?

1. The "Pop" Effect (Fragmentation)

When the electrons hit the W(CO)₆ molecules, they acted like pinballs, knocking the "CO balloons" off the Tungsten core.

• The Result: The molecules didn't just break into two pieces; they got stripped down layer by layer. They went from W(CO)₆ → W(CO)₅ → W(CO)₄... all the way down to bare Tungsten atoms (W).
• The Analogy: Imagine a snowman. The electrons are warm hands. First, the hat falls off. Then the scarf. Then the arms. Finally, you are left with just the snowball (the Tungsten).

2. The "Crowd" Effect (Density Matters)

This is where it got interesting.

• In the Empty Room (Low Density): The molecules broke apart, but they mostly stayed alone. They didn't have many neighbors to bump into.
• In the Mosh Pit (High Density): When the molecules broke apart, the pieces (fragments) were so close together that they immediately started bumping into each other. Instead of just floating away, the bare Tungsten atoms grabbed onto each other and formed clusters (little metal balls).
• The Lesson: If you want to build a solid metal structure, you need a dense crowd. If the gas is too thin, the pieces fly apart and don't stick together well.

3. The "Speed" Effect (Fluence vs. Time)

The researchers tested if hitting hard and fast (high current, short time) gave the same result as hitting soft and slow (low current, long time).

• The Surprise: It did not always give the same result.
• The Analogy: Think of a dance floor. If you play music slowly for a long time, people might just sway. If you blast the music at maximum volume for a split second, everyone might panic and jump.
• The Science: When the electron beam was very intense (high current), it broke the molecules apart too fast. This created a chaotic mix of tiny, broken pieces that didn't have time to settle into the "perfect" intermediate shapes (like W(CO)₃).
• The Takeaway: To get the best simulation of how these structures grow, you can't just speed up the clock. You have to match the simulation time to how long the real experiment takes (usually microseconds to milliseconds). If you simulate too fast, you might think you get more metal than you actually do.

4. The "Oxygen" Side Effect

Sometimes, the electrons didn't just pop the CO balloons; they broke the balloons themselves (breaking the bond between Carbon and Oxygen).

• The Result: This released free Oxygen atoms. In the crowded simulations, these Oxygen atoms found each other and formed O₂ (regular oxygen gas).
• The Lesson: In a dense crowd, chemical reactions happen between the broken pieces, not just because of the electron beam.

Why Does This Matter?

This paper is like a user manual for the future of 3D nanoprinting.

1. Better Simulations: It tells scientists exactly how to set up their computer models. If they simulate the process too quickly, they will get the wrong answer. They need to use the right "dwell time" (how long the beam stays on one spot).
2. Better Manufacturing: It explains why some 3D printed metal structures are strong and others are weak. If the gas density is too low, the metal won't clump together properly. If the electron beam is too intense, it might break the molecules into useless dust instead of building blocks.
3. Predicting the Future: By understanding how these molecules break and stick, scientists can design better "inks" (precursor gases) to print tiny wires, circuits, and sensors for the next generation of electronics.

The Bottom Line

The paper is a guide on how to control a microscopic demolition and construction crew. It teaches us that density (how crowded the gas is) and timing (how fast the electrons hit) are the two most important knobs to turn if you want to build perfect, tiny metal structures out of thin air.

Technical Summary

1. Problem Statement

Focused Electron Beam-Induced Deposition (FEBID) is a critical nanofabrication technique used to create 3D nanostructures with high resolution. The process relies on the electron-induced fragmentation of organometallic precursor molecules (e.g., Tungsten Hexacarbonyl, $W(CO)_6$) to deposit metal-rich structures.

Despite its utility, the fundamental physicochemical mechanisms governing precursor fragmentation and subsequent nanostructure growth under electron irradiation are not fully understood. A specific challenge in computational modeling of FEBID is the timescale discrepancy: experimental dwell times range from microseconds to milliseconds, while atomistic Molecular Dynamics (MD) simulations are typically limited to nanoseconds. To bridge this gap, previous studies rescaled electron beam currents to maintain a constant total electron fluence over short simulation times. However, it remained unclear whether the fragmentation dynamics are purely fluence-controlled (dependent only on the total number of electrons) or if they depend explicitly on irradiation parameters (beam current and dwell time) and precursor density. This study aims to resolve these scaling dependencies.

2. Methodology

The authors employed Irradiation-Driven Molecular Dynamics (IDMD) simulations using the MBN Explorer software package.

• System Setup: Three gas-phase systems were modeled containing $W(CO)_6$ molecules at different densities within a 20 nm cubic box:
  • Low density: 23 molecules ($\sim 2.9 \times 10^{-3} \text{ nm}^{-3}$)
  • Intermediate density: 107 molecules ($\sim 1.3 \times 10^{-2} \text{ nm}^{-3}$)
  • High density: 207 molecules ($\sim 2.6 \times 10^{-2} \text{ nm}^{-3}$)
• Force Fields: The reactive CHARMM (rCHARMM) force field was used, allowing for dynamic bond breaking and formation.
  • Bonded interactions were modeled using the Morse potential.
  • Angular interactions utilized a sigmoid-type function to account for bond rupture.
  • Non-bonded interactions used the Lennard-Jones potential.
  • Crucially, the model allowed for dynamic changes in atom types and interaction parameters (e.g., bond lengths and dissociation energies) based on the specific fragment formed, derived from DFT calculations.
• Irradiation Model:
  • Mechanism: Dissociative ionization (DI) was the primary mechanism.
  • Cross-Sections: Absolute partial ionization cross-sections (PICS) for $W(CO)_n^+$ and $WC(CO)_n^+$ fragments were derived by combining experimental data (up to 140 eV) with theoretical total ionization cross-sections (up to 5000 eV).
  • Simulation Parameters: Simulations were run at an electron energy of 100 eV. To test scaling, the beam current ($I$) and simulation time ($t$) were varied inversely to keep the total electron fluence ($\Phi$) constant (reference $\Phi \approx 15.6 \text{ nm}^{-2}$). Additional simulations varied the fluence (up to 4x) to study its direct impact.
  • Energy Deposition: Upon electron-molecule collision, a characteristic energy ($E_{dep} \approx 13 \text{ eV}$) was deposited to simulate bond rupture.

3. Key Contributions

• Scaling Analysis: The study systematically investigates whether FEBID precursor fragmentation is governed solely by total electron fluence or if it exhibits dependencies on beam current, dwell time, and molecular density.
• Ensemble Dynamics: Unlike previous studies focusing on single molecules, this work analyzes the fragmentation dynamics of molecular ensembles, capturing intermolecular interactions and secondary chemical reactions.
• Advanced Fragmentation Model: The implementation of a dynamic rCHARMM model that updates atom types and interaction parameters upon bond cleavage (specifically distinguishing between $W(CO)_n$ and $WC(CO)_n$ species) provides a more accurate representation of chemical evolution than static force fields.
• Cluster Formation Mechanism: The paper elucidates the transition from simple ligand loss to the aggregation of tungsten atoms into metal-rich molecular clusters.

4. Key Results

A. Scaling with Irradiation Time and Current (Fluence Control)

• Low-Density Systems (23 molecules):
  • At short simulation times (high beam current, $<10$ ns), the system produces an overabundance of highly fragmented species ($W(CO)_2$, $W(CO)$, atomic $W$).
  • At longer simulation times (lower beam current, $>10$ ns), the system reaches a steady state where fragment abundances stabilize (dominated by $W(CO)_3$, $W(CO)_4$, $W(CO)_5$).
  • Conclusion: For low densities, fragmentation dynamics are not strictly fluence-controlled at short timescales. Short dwell times in simulations overestimate the production of highly fragmented species. A minimum dwell time ($\sim 10$ ns in simulation) is required to reach a realistic steady state.

B. Influence of Precursor Density

• High-Density Systems (207 molecules):
  • Fragmentation is significantly more extensive than in low-density systems. Parent molecules ($W(CO)_6$) and intermediate species ($W(CO)_5$, $W(CO)_4$) deplete rapidly.
  • The system evolves toward a steady state dominated by smaller fragments ($W(CO)_3$) and atomic tungsten.
  • Mechanism: Higher density increases the probability of intermolecular collisions and secondary reactions between fragments and free ligands, facilitating further dissociation.

C. Effect of Electron Fluence

• Increasing electron fluence accelerates fragmentation, shifting the distribution toward smaller species and atomic tungsten.
• High fluence leads to a higher population of under-coordinated tungsten fragments, which increases the probability of aggregation.

D. Post-Fragmentation Aggregation and Cluster Formation

• Cluster Growth: Tungsten atoms aggregate to form W-rich molecular clusters.
  • In low-density systems, cluster formation is negligible ($\sim 2\%$ of W atoms).
  • In high-density systems, up to 30% of tungsten atoms incorporate into clusters.
• Cluster Characteristics:
  • Clusters are mixed organometallic structures (e.g., $W_{12}C_3O_3$, $W_8C_2O$).
  • C–O bond cleavage occurs, leading to clusters with reduced oxygen content compared to the parent molecule.
  • The largest clusters observed contained up to 29 atoms, with tungsten content reaching 65–70 at.%.
• Oxygen Recombination: While $O_2$ molecules form via recombination of oxygen atoms from C–O cleavage, their abundance remains low compared to W-containing species.

5. Significance and Implications

• Simulation Parameter Selection: The study provides critical guidelines for setting up IDMD simulations of FEBID. It demonstrates that using excessively high beam currents (to compress simulation time) can lead to inaccurate predictions of deposit composition (overestimating tungsten content) if the system has not reached a steady state.
• Process Optimization: The findings suggest that precursor density is a key control parameter. Higher densities promote the formation of metal-rich clusters, which is desirable for conductive FEBID structures, whereas low densities may result in more volatile, metal-poor deposits.
• Experimental Validation: The results align with recent FEBiMS (Focused Electron Beam Induced Mass Spectrometry) experiments, validating the IDMD approach for studying gas-phase fragmentation and cluster formation.
• Fundamental Insight: The work bridges the gap between quantum mechanical bond breaking and macroscopic nanostructure growth, offering a quantitative description of how irradiation parameters dictate the chemical evolution of organometallic precursors.

In summary, this paper establishes that while total electron fluence is a primary driver, the temporal profile of irradiation (dwell time vs. current) and local precursor density are equally critical in determining the final chemical composition and morphology of FEBID deposits.