Atomic Layer Etching: Cl2 Adsorption and Ar Ion Bombardment on Si(100)

This application demonstrates an advanced Surface Process Simulation (SPS) technique for studying atomic layer etching (ALE) processes involving sequential Cl2 adsorption and Ar ion bombardment on silicon surfaces. The simulation captures the complete ALE cycle including chlorine saturation, argon-induced etching, and surface regeneration to study selectivity, etch rates, and surface chemistry evolution in semiconductor processing applications.

Important

QuantumATK Version: This application is designed for QuantumATK X-2025.06.

This application requires QuantumATK’s Surface Process Simulation framework and includes advanced analysis scripts for ALE characterization. You can download the required files below:

Key Features

  • Complete ALE cycle: Sequential Cl2 adsorption followed by Ar ion bombardment

  • Saturation analysis: Real-time monitoring of chlorine surface coverage and saturation behavior

  • Selective etching: Study of preferential etching mechanisms and surface selectivity

  • Cyclic processing: Multiple ALE cycles to study steady-state behavior and surface evolution

  • Advanced analysis: Automated extraction of etch rates, selectivity factors, and surface composition

System Overview

ALE process cycle showing chlorine adsorption and argon bombardment

ALE Process Schematic: The ALE cycle consists of two distinct phases - chlorine adsorption (chemical modification) and argon bombardment (physical removal), enabling precise atomic-scale control of silicon etching.

The simulated system consists of a silicon (100) surface optimized for ALE studies:

  • Surface Structure: Si(100) 2×1 reconstructed surface with periodic boundary conditions

  • Dimensions: 15.36 Å × 15.36 Å × 48.01 Å supercell

  • Temperature Control: 300 K substrate temperature maintained throughout ALE cycles

  • Surface Preparation: Clean silicon surface with proper 2×1 dimer reconstruction

ALE Process Parameters:

  • Chlorine Adsorption Phase: Thermal energy applied for chlorine molecules (300 K)

  • Ar Bombardment Phase: 100 eV argon ions with controlled flux and angle

  • Cycle Timing: Sequential exposure with complete separation between chemical adsorption and physical removal phases

  • Surface Regeneration: Automatic surface cleaning and preparation between cycles

Simulation Workflow

Step 0: Choice of Calculator

Before starting the simulation workflow, select an appropriate calculator for energy, force, and trajectory calculations. The choice depends on the desired accuracy and available computational resources.

Common options include:

  • Classical Force Fields: Suitable for large-scale simulations and rapid prototyping (e.g., Tersoff, Stillinger-Weber for Si).

  • Density Functional Theory (DFT): Provides higher accuracy for electronic structure and reaction mechanisms, but is computationally more demanding.

  • Machine-Learned Potentials: Combine speed and accuracy for specific systems.

    • Targeted MLFF (e.g., MTP): Machine-learned force fields such as Moment Tensor Potentials (MTP) are trained on data specific to a given material or process, offering high accuracy for that system but limited transferability to other chemistries or environments.

    • Universal MLFF (e.g., MACE, MatterSim, etc.): Universal machine-learned force fields are trained on large, diverse datasets to provide broad applicability across many materials and chemistries. They offer greater transferability but may be less accurate for highly specialized systems compared to targeted MLFFs.

Assign the calculator to your configuration before proceeding with surface preparation and subsequent simulation steps.

potentialSet = TremoloXPotentialSet(name='TersoffBrenner_ClFSi_2003+Moliere')

The Moliere potential is used in this setup to model the short-range repulsive interactions between argon (Ar) and other elements in the system, ensuring accurate description of ion-surface collisions during bombardment.

Step 1: Surface Preparation and Equilibration

The simulation begins with proper preparation of the Si(100) surface including reconstruction and thermal equilibration:

../../_images/reconstructed_si.png

The silicon surface is initially optimized to achieve the correct 2×1 reconstruction characteristic of clean Si(100). Note that buckled reconstructed layer is not observed with the forcefield chosen for this study. Following optimization, a two-stage molecular dynamics equilibration is performed:

Stage 1: Heating from 10 K to 300 K

The optimized surface undergoes gradual heating using controlled molecular dynamics:

  • Ensemble: NVT

  • Initial temperature: 10 K with Maxwell-Boltzmann velocity distribution

  • Heating rate: 2.9 K/ps

  • Time step: 0.5 fs

  • Total steps: 200,000 (100 ps total simulation time)

  • Purpose: Controlled heating prevents structural damage while bringing the surface to the target temperature

Stage 2: Equilibration at 300 K

Following the heating phase, the surface is equilibrated at constant temperature:

  • Ensemble: NVT

  • Temperature: 300 K (constant)

  • Time step: 0.5 fs

  • Total steps: 200,000 (100 ps total simulation time)

  • Purpose: Ensures proper thermal motion, surface relaxation, and establishes realistic surface reactivity and adsorption site availability for subsequent ALE processing

Step 2: Chlorine Adsorption Phase

The first phase of each ALE cycle involves controlled exposure to Cl2 molecules:

Chlorine Molecular Setup:

Chlorine molecules are positioned 10 Å above the surface with thermal kinetic energy distributions. The molecular orientation is randomized to ensure statistical sampling of different adsorption pathways.

Surface Process Simulation Parameters:

# Chlorine adsorption - Surface Process Simulation
surface_process_simulation = SurfaceProcessSimulation(
    substrate=substrate_configuration,
    filename='SPS_Si100_Cl2_300K_Ar_100eV_ALE_results.hdf5',
    temperature=300.0 * Kelvin,
    fixed_thickness=1.0 * Angstrom,
    thermostat_thickness=5.0 * Angstrom,
    fixed_directions=(2,),
)

surface_process_simulation.addSequence(
    molecule=cl2,
    number_of_events=250,
    md_time=10.0 * picoSecond,
    time_step=1.0 * femtoSecond,
    mean_kinetic_energy=300.0 * kBK,
    mean_incident_angle=0.0 * Degrees,
    internal_energy=300.0 * Kelvin,
)

  • Substrate temperature: 300 K

  • Fixed layer thickness: 1.0 Å (bottom atoms held fixed)

  • Thermostat thickness: 5.0 Å (temperature control region)

  • Fixed directions: z-direction (prevents lateral drift)

Cl₂ Adsorption Sequence:

  • Molecule: Cl₂ (chlorine dimer)

  • Number of events: 250 (molecular impacts per cycle)

  • MD time per event: 10.0 ps

  • Time step: 1.0 fs

  • Mean kinetic energy: 300 K thermal energy (0.026 eV)

  • Incident angle: 0° (normal to surface)

  • Internal energy: 300 K (molecular vibration/rotation)

Adsorption Process:

The Surface Process Simulation tracks molecular dissociation, atomic adsorption, and surface saturation. Number of events are chosen to ensure saturation of the surface. More events would be needed if we increase the surface area. Cl2 molecules undergo dissociative adsorption on silicon dangling bonds, forming Si-Cl bonds and modifying surface reactivity.

Step 3: Argon Ion Bombardment Phase

The second phase involves controlled Ar ion bombardment to remove the chlorinated surface layer:

Ar Ion Parameters:

Argon ions are generated with 100 eV kinetic energy and directed toward the chlorinated surface. The ion flux and angular distribution are controlled to achieve selective removal of chlorinated species while minimizing damage to the underlying silicon.

# Argon bombardment - Surface Process Simulation
surface_process_simulation_1 = SurfaceProcessSimulation(
    substrate=substrate_configuration,
    filename='SPS_Si100_Cl2_300K_Ar_100eV_ALE_results.hdf5',
    temperature=300.0 * Kelvin,
    fixed_thickness=1.0 * Angstrom,
    thermostat_thickness=5.0 * Angstrom,
    fixed_directions=(2,),
)

surface_process_simulation_1.addSequence(
    molecule=ar,
    number_of_events=250,
    md_time=10.0 * picoSecond,
    time_step=1.0 * femtoSecond,
    mean_kinetic_energy=100.0 * eV,
    mean_incident_angle=0.0 * Degrees,
    internal_energy=300.0 * Kelvin,
)

Surface Process Simulation Parameters:

  • Substrate temperature: 300 K

  • Fixed layer thickness: 1.0 Å (bottom atoms held fixed)

  • Thermostat thickness: 5.0 Å (temperature control region)

  • Fixed directions: z-direction (prevents lateral drift)

Ar Bombardment Sequence:

  • Molecule: Ar (argon atom)

  • Number of events: 250 (ion impacts per cycle)

  • MD time per event: 10.0 ps

  • Time step: 1.0 fs

  • Mean kinetic energy: 100 eV (high-energy ion bombardment)

  • Incident angle: 0° (normal to surface)

Etching Mechanism:

The bombardment process preferentially removes Si-Cl species through momentum transfer and chemical enhancement. The etching selectivity can be quantified by comparing removal rates of chlorinated versus non-chlorinated surface species, providing insights into the chemical enhancement factor. Ar impacts on Si alone should result in lower or no etching compared to chlorinated surfaces.

Step 4: Post-Processing Analysis

After the complete simulation (all ALE cycles) has finished, comprehensive post-processing analysis is performed using the provided Python analysis scripts. These scripts read the saved trajectory data and extract key performance metrics from the HDF5 output file.

Cl_saturation_plot.py - Chlorine Saturation Analysis

This script analyzes the chlorine adsorption behavior during the ALE cycles. This script can be executed while the simulation is running to monitor saturation in real-time.

Inputs:

  • HDF5 trajectory files: Cl_pulse_1.hdf5, Cl_pulse_2.hdf5, Cl_pulse_3.hdf5, Cl_pulse_4.hdf5

  • Monolayer reference: 16 atoms (number of surface Si atoms in the simulation cell)

  • Surface area: Calculated from primitive lattice vectors (15.36 × 15.36 Ų)

Analysis Performed:

  • Reads SurfaceProcessSimulation trajectories from each Cl₂ pulse cycle

  • Cleans vacuum region to remove unreacted molecules

  • Counts chlorine atoms after each molecular impact event

  • Calculates fluence (impacts/area) in units of nm⁻²

  • Normalizes chlorine uptake to monolayer coverage

Outputs:

  • Fluence vs. Uptake plot: Shows chlorine coverage evolution as a function of molecular dose for each pulse

  • Saturation behavior: Reveals self-limiting adsorption characteristic of ALE processes

  • Demonstrates convergence to ~0.8-1.0 monolayer saturation coverage

Etch_plot.py - Etching Depth Analysis

This script quantifies the silicon removal during Ar ion bombardment phases.

Inputs:

  • HDF5 trajectory files: Alternating Cl and Ar pulse files

    • Cl_pulse_1.hdf5, Ar_pulse_1.hdf5, Cl_pulse_2.hdf5, Ar_pulse_2.hdf5, Cl_pulse_3.hdf5, Ar_pulse_3.hdf5, Cl_pulse_4.hdf5, Ar_pulse_4.hdf5

  • Monolayer parameters: 16 atoms/monolayer, 0.136 nm/monolayer thickness

  • Substrate definition: 16 monolayers total in the pristine substrate

Analysis Performed:

  • Reads trajectories from both Cl₂ adsorption and Ar bombardment phases sequentially

  • Cleans vacuum region to count only substrate atoms

  • Counts remaining silicon atoms after each impact event

  • Calculates cumulative etch depth by comparing to pristine surface (16 monolayers)

  • Tracks material removal as a function of total impact number

Outputs:

  • Etch depth vs. Impact number plot: Shows cumulative silicon removal over all cycles

  • Etch per cycle (EPC) values: Quantifies the thickness removed per ALE cycle (typically in nm/cycle)

  • Steady-state behavior: Demonstrates reproducible etching after initial transient cycles

ImpactsVsuptake_plot.py - Impact-Uptake Correlation Analysis

This script correlates the number of molecular/ion impacts with chlorine uptake and material removal throughout the ALE cycles.

Inputs:

  • HDF5 trajectory files: Alternating Cl and Ar pulse files

    • Cl_pulse_1.hdf5, Ar_pulse_1.hdf5, Cl_pulse_2.hdf5, Ar_pulse_2.hdf5, Cl_pulse_3.hdf5, Ar_pulse_3.hdf5, Cl_pulse_4.hdf5, Ar_pulse_4.hdf5

  • Monolayer reference: 16 atoms (surface Si atoms in simulation cell)

  • Fuzz factor: 1.15 (for vacuum region identification)

Analysis Performed:

  • Reads trajectories from both Cl₂ and Ar phases sequentially

  • Cleans vacuum region after each impact event

  • Counts chlorine atoms throughout the entire ALE sequence

  • Tracks chlorine uptake normalized to monolayer coverage

  • Shows cyclic behavior of Cl adsorption (increasing) and removal (decreasing) phases

Outputs:

  • Uptake vs. Impact number plot: Shows cyclic chlorine coverage evolution across all impacts

  • Cyclic behavior visualization: Demonstrates the saw-tooth pattern of Cl adsorption during Cl₂ pulses and removal during Ar pulses

  • Self-limiting behavior: Shows saturation during adsorption and efficient removal during bombardment

Analysis and Observations

Chlorine Saturation Behavior

Chlorine uptake vs impacts showing cyclic adsorption and removal

Fig. 148 Chlorine Uptake Analysis: The plot shows the cyclic evolution of chlorine coverage throughout multiple ALE cycles. The saw-tooth pattern demonstrates Cl₂ adsorption (rising edges) and Ar-induced removal (falling edges). A rolling average is applied to smooth the curve fluctuations.

The chlorine saturation analysis reveals the self-limiting nature of the adsorption process. Initial rapid uptake occurs on high-reactivity dangling bond sites, followed by slower saturation as steric hindrance limits additional adsorption. The saturation coverage typically reaches 0.8-1.0 monolayers depending on surface temperature and molecular flux conditions.

The cyclic behavior clearly shows:

  • Adsorption phase: Rapid Cl uptake during Cl₂ pulse reaching near-monolayer saturation

  • Removal phase: Efficient chlorine removal during Ar bombardment, resetting the surface

  • Reproducibility: Consistent cycle-to-cycle behavior after initial transients

  • Self-limiting nature: Saturation plateau indicating controlled ALE process

Note

The curve fluctuations visible in the plot are due to the limited sampling in this small-scale demonstration. Smoother curves can be obtained by using larger surface areas (increasing supercell dimensions) which provide better statistical sampling with the same number of impact events per unit area. Also, increaing the thickness of the substrate can allow for multiple cycles of bombardment without reaching the bottom fixed layer, improving accuracy.

Etching Yield and Material Removal

Etch depth vs impacts showing cumulative material removal

Fig. 149 Etch Yield Analysis: The plot shows cumulative silicon removal as a function of total impact number (both Cl₂ and Ar). The step-like behavior reflects alternating adsorption (no etching) and bombardment (active etching) phases. A rolling average highlights the overall etching trend.

The etching analysis demonstrates significant chemical enhancement during the Ar bombardment phases:

  • Chemical enhancement: High etch rates during Ar bombardment of chlorinated surfaces compared to negligible etching during Cl₂ exposure alone

  • Selectivity: There is a 1:1 correspondence to Cl uptake and Si removal, indicating preferential etching of chlorinated species. In the final pulse, not a lot of Cl is adsorbed and hence minimal etching is observed.

Note

The step-like features in the Ar pulse region of raw data reflect the discrete nature of molecular/ion impact events in this small simulation cell. Larger systems with more surface atoms would provide smoother curves due to improved statistical averaging, though the underlying physics and cyclic behavior remain the same.

Customization Options

Process Parameter Optimization

Large Systems: The simulation framework can be scaled to larger surface areas by increasing the supercell dimensions and adjusting the number of molecular/ion impacts accordingly. This should give smoother curves in the analysis due to increased sampling.

Temperature Effects: The simulation can be adapted for different substrate temperatures (200-500 K) to study temperature-dependent adsorption and desorption kinetics.

Ion Energy Studies: Argon bombardment energy can be varied (50-200 eV) to optimize the balance between selectivity and etch rate.

Alternative Chemistries: The framework supports other halogen chemistries (Br2, I2) and different ion species (Ne, Kr) for comparative studies.

Cycle Timing Optimization

Adsorption Time: The Cl2 exposure time can be optimized to achieve complete saturation while minimizing process time.

Bombardment Dose: The argon ion dose per cycle and impact angle can be optimized to acieve faster etching while preserving surface integrity.

Output Files

The simulation generates comprehensive output files for detailed analysis:

  • SPS_Si100_Cl2_300K_Ar_100eV_ALE_results.hdf5: Complete simulation trajectories and configuration data

  • uptake_plot.hdf5: Chlorine uptake and saturation curves as a function of impact number and cycle

  • etch_plot.hdf5: Silicon removal rates as a function of impact number and cycle

Using the NanoLab Workflow

The application includes pre-configured NanoLab workflow blocks for:

  • Surface preparation and equilibration setup

  • Cl2 molecule configuration and bombardment parameters

  • Ar ion setup with energy and flux specifications

  • looping over multiple ALE cycles

SPS_Si100_Cl2_300K_Ar_100eV_ALE.hdf5

Running the Script

Execute the complete ALE simulation:

atkpython SPS_Si100_Cl2_300K_Ar_100eV_ALE_results.py

We recommend using the Jobs tool and run on HPC cluster for faster execution.

Generate analysis plots in the same folder as output files from the above command:

atkpython Cl_saturation_plot.py
atkpython Etch_plot.py
atkpython ImpactsVsuptake_plot.py

Performance Considerations

Computational Requirements: The ALE simulation requires significant computational resources due to the multiple cycles and many impact events.

Memory Usage: ~11 GB harddisk space required for trajectory and analysis files. Since force field calculations are used, memory requirements are moderate compared to ab initio methods.

Computational Time: It took 21 hours using 1 MPI and 20 OpenMP threads.

Summary

This application example provides a complete framework for studying ALE processes using Surface Process Simulation. The sequential Cl2 adsorption and Ar bombardment approach enables detailed investigation of:

  • Chemical enhancement mechanisms in atomic layer etching

  • Saturation behavior and self-limiting surface chemistry

  • Cyclic process optimization for precise atomic-scale control

  • Surface morphology evolution during ALE processing

The included analysis scripts provide automated extraction of key ALE performance metrics, making this a valuable tool for both fundamental research and process development applications in semiconductor manufacturing.

Note

Large-Scale Simulation Results: check out the Synopsys QuantumATK whitepaper which demonstrates large-scale surface process simulations for Si ALE.