Optimized tandem catalyst patterning for CO$_2$ reduction flow reactors

This study demonstrates that integrating continuum transport modeling with adjoint-based design optimization significantly enhances CO$_2$ reduction flow reactor performance by strategically patterning Ag and Cu catalysts to maximize ethylene current density, particularly at high voltages and with increased patterning sections.

The Gist

Imagine you are trying to bake the perfect cake, but your kitchen has a strange rule: you can't mix all the ingredients at once. Instead, you have two separate stations.

• Station A (Silver): This station is great at turning raw flour (Carbon Dioxide) into dough (Carbon Monoxide).
• Station B (Copper): This station is amazing at turning that dough into a delicious cake (Ethylene, a valuable chemical).

The problem? If you put Station A far away from Station B, the dough gets carried away by the wind (the flowing water in the reactor) before it can reach Station B. Or, if you put too much of Station A and not enough of Station B, you end up with a pile of dough and no cake.

This paper is about figuring out the perfect layout for these two stations to make the most cake possible.

The Big Idea: "Tandem Catalysis"

The researchers are studying a process called tandem catalysis. Think of it like an assembly line.

1. Silver (Ag) acts as the first worker, converting CO₂ into CO.
2. Copper (Cu) acts as the second worker, taking that CO and turning it into high-value products like ethylene (a building block for plastics and fuels).

In a traditional setup, these workers might be mixed together or placed in big, separate blocks. The researchers wanted to know: If we break the electrode into many small, alternating strips of Silver and Copper, and we can change the length of each strip, what is the best pattern to get the most cake?

The Experiment: A Digital "Tuning" Knob

Instead of building physical reactors and trying thousands of different patterns (which would take years), the team built a computer simulation.

They created a digital "flow reactor" where liquid flows over a flat surface. They used a smart computer algorithm (like a super-advanced GPS) to test millions of different patterns. The computer would:

1. Try a pattern (e.g., a long Silver strip, then a short Copper strip).
2. See how much "cake" (ethylene) was made.
3. Adjust the lengths of the strips slightly.
4. Repeat this over and over until it found the absolute best arrangement.

What They Found

The computer found that the "perfect" pattern depends heavily on how hard you push the system (the voltage) and how fast the liquid is flowing.

1. The "Strong Push" Scenario (High Voltage):
When they pushed the system hard (using a strong electrical voltage), the best design was to have many, many small strips (up to 12 sections) rather than just two big ones.

• The Result: This optimized pattern produced up to 65% more ethylene than a simple, unoptimized design.
• Why? At high speeds, the liquid moves fast. If the Copper section is too long, the "dough" (CO) gets used up at the very beginning of the strip, and the rest of the Copper strip sits idle (a "dead zone"). By making the strips shorter and more numerous, the fresh dough is constantly delivered to the Copper workers, keeping them busy the whole time.

2. The "Gentle Push" Scenario (Low Voltage):
When the push was weaker, the best pattern looked different. It favored a very long first Silver strip to make a huge pile of dough, followed by a very long last Copper strip to eat it all up, with tiny, fast-switching strips in the middle.

3. The Flow Rate Matters:

• Fast Flow: If the water is rushing by, you need the reaction to be very strong (high voltage) to keep the dough from washing away.
• Slow Flow: If the water is slow, the dough has time to settle, but you need to be careful not to run out of fresh ingredients.

The Secret Sauce: Avoiding "Dead Zones"

The main reason the optimized patterns worked so well is that they eliminated "dead zones."

Imagine a conveyor belt where the first few workers are busy, but the last few workers are standing around doing nothing because the parts ran out. In the old designs, the Copper sections often had these dead zones at the end where the CO ran out.

The computer's optimized designs rearranged the strips so that the "dough" (CO) was distributed evenly. It ensured that every inch of the Copper surface had enough dough to work on, maximizing the production of the final product.

Summary

This paper is a "proof of concept." It didn't build a physical factory, but it proved that using math and computers to design the layout of catalysts can significantly improve how well we turn CO₂ into useful chemicals.

• The Problem: CO₂ reduction is tricky; intermediate products get lost or wasted.
• The Solution: Use a computer to find the perfect pattern of alternating Silver and Copper strips.
• The Payoff: By simply changing the shape of the catalyst surface (not the chemicals themselves), they could boost production by up to 65% in their simulation.

It's like realizing that if you rearrange the furniture in a room, you can move around much faster, even if you don't buy any new furniture.

Technical Summary

Technical Summary: Optimized Tandem Catalyst Patterning for CO2 Reduction Flow Reactors

Problem Statement
Electrochemical CO2 reduction (CO2R) offers a pathway to convert excess renewable energy into sustainable fuels and chemicals while mitigating greenhouse gas emissions. However, commercial-scale deployment is hindered by performance limitations in copper (Cu) catalysts, specifically high overpotentials and limited selectivity toward high-value multi-carbon products (e.g., ethylene, ethanol). These limitations often stem from ineffective management of intermediate species transport. In traditional tandem catalysis, where a first catalyst (e.g., Ag) converts CO2 to CO and a second catalyst (e.g., Cu) converts CO to high-value products, intermediate species like CO can be lost to advection or diffusion before reaching the second catalyst. While previous studies have explored various tandem catalyst configurations (e.g., interspersed particles, alternating planar sections), they have largely relied on forward analysis—proposing and evaluating specific designs—rather than systematically optimizing the spatial arrangement of catalysts to maximize performance.

Methodology
This study presents a proof-of-concept framework integrating continuum transport modeling with adjoint-based design optimization within a simplified two-dimensional (2D) flow reactor setup.

• System Setup: The computational domain models a shear flow of aqueous bicarbonate electrolyte (500 mM KHCO3, 34 mM CO2) over an active electrode surface. The electrode consists of alternating Ag and Cu sections. Ag sections facilitate the CO2 → CO reaction, while Cu sections facilitate the conversion of CO to ethylene (C2H4), ethanol (C2H6O), methane (CH4), and hydrogen (H2). CO is not present in the inlet stream; it is generated by Ag and consumed by Cu.
• Governing Equations: The model solves steady-state Nernst-Planck equations for species transport (CO2, H+, OH-, HCO3-, CO3 2-, K+, CO, H2, and hydrocarbons) coupled with bulk bicarbonate buffer reactions. Surface kinetics are modeled using Tafel expressions for both Ag and Cu catalysts. Fluid flow is approximated as a shear flow with a defined shear rate derived from the Péclet number.
• Optimization Formulation: The problem is formulated as a PDE-constrained optimization. The design variables are the lengths of $N$ alternating Ag/Cu sections ($l_j$). The objective function is the maximization of the spatially averaged ethylene current density ($i_{C2H4}$).
• Algorithm: The optimization utilizes the adjoint method to compute sensitivities of the objective function with respect to the catalyst section lengths. An L-BFGS-B quasi-Newton algorithm iteratively refines the section lengths to maximize performance. The study investigates the effects of applied voltage ($U_{app}$), electrolyte flow rate, and the number of patterning sections ($N$).

Key Contributions

1. Design Optimization Framework: The paper introduces a systematic optimization approach for tandem catalyst patterning, moving beyond forward analysis to iteratively refine catalyst spatial arrangements based on transport and reaction constraints.
2. Performance Enhancement via Patterning: The study demonstrates that optimizing the spatial distribution of Ag and Cu sections significantly enhances ethylene production, particularly when the number of sections ($N$) is increased and under conditions of strong surface reactions (more negative applied voltages).
3. Mechanistic Insight: Through concentration field analysis, the work elucidates how optimized patterns mitigate "dead zones" (regions of low CO concentration on Cu surfaces) and improve the utilization of the intermediate CO species.

Results

• Impact of Section Number ($N$): Increasing the number of sections from $N=2$ to $N=12$ significantly increases ethylene current density. For an applied voltage of $-1.7$ V vs. SHE, the optimized $N=12$ design achieves up to a 65% increase in ethylene current density compared to an unoptimized $N=2$ equal-length design.
• Optimization vs. Section Count: The performance gain is decomposed into contributions from increasing $N$ and from optimizing the pattern lengths. At less negative voltages ($-1.35$ V), optimization of the pattern is the dominant factor. At more negative voltages ($-1.7$ V), increasing the number of sections is the primary driver of improvement, though optimization still yields significant gains.
• Pattern Characteristics:
  • At $-1.35$ V (weaker reaction), optimized patterns for high $N$ feature long initial Ag and final Cu sections with short, rapidly alternating intermediate sections.
  • At $-1.7$ V (stronger reaction) and low flow rates, optimized patterns favor longer Ag sections and shorter Cu sections to maximize CO production and minimize CO starvation on Cu surfaces.
  • At $-1.7$ V and high flow rates, the optimized patterns remain close to equal-length configurations, suggesting the initial equal-length design is already near-optimal under these high-mass-transport conditions.
• Flow Rate Dependence: The effect of flow rate on performance is voltage-dependent. At $-1.35$ V, higher flow rates reduce CO utilization and ethylene production due to the sweeping away of weakly produced CO. Conversely, at $-1.7$ V, higher flow rates increase ethylene production by replenishing reactants more effectively, as the strong surface reactions maintain high local CO concentrations.

Significance and Claims
The authors claim that this work provides a proof of concept for using continuum modeling combined with design optimization to guide the spatial arrangement of catalysts in CO2R electrolyzers. The study establishes that:

• Optimal reactor design is highly dependent on operating conditions (voltage, flow rate) and the degree of patterning ($N$).
• The primary mechanism for performance improvement in optimized tandem catalysts is the minimization of low-reactant-concentration zones on the secondary catalyst surface, thereby maximizing the utilization of the key intermediate (CO).
• While the current 2D setup is a simplification that does not capture all complexities of industrial gas diffusion electrodes (GDEs), the methodology offers a foundational step toward optimizing more realistic, complex electrolyzer geometries.
• The integration of high-performance computing with informed continuum modeling serves as a powerful tool for narrowing the design space for experimental validation and improving electrochemical process efficiency.

The paper concludes that future work should extend these optimization capabilities to 3D settings (e.g., GDEs), explore other objective functions (selectivity, energy efficiency), and optimize operating conditions alongside catalyst patterning.