Capacity and Selectivity for Carbon Dioxide on Lignin-Derived Adsorbents via Molecular Simulation Authors: Mohsen Samandari, David P. Harper, David J. Keffer Published: 2026 ABSTRACT In pursuit of cost-effective and environmentally benign materials for carbon capture applications, this study investigates the effectiveness of carbon quantum dots (CQDs) and lignin-derived carbon composites as sustainable materials for selective carbon dioxide capture. Classical grand canonical Monte Carlo (GCMC) and molecular dynamics (MD) simulations are employed to examine the adsorption behavior, binding energy, and selectivity of different functionalized CQDs on graphite and lignin-based carbon composite (LBCC) substrates. Four types of CQDs were studied: amine-functionalized, nitrogen-doped, carboxyl-functionalized, and hydroxyl-functionalized. Our simulations reveal that while all CQD types preferentially adsorb CO2 over N2 and O2, their performance varies significantly depending on functionalization type and surface density. Notably, bare LBCC substrates demonstrate remarkably high CO2/N2 and CO2/O2 selectivity (>50) in both dry and wet flue gas mixtures at 1 atm pressure. Addition of CQDs to LBCC substrates can increase capacity without reduction in selectivity in dry flue gases. Our findings suggest that an optimized lignin-based activated carbon decorated with specific CQDs could offer a renewable, cost-effective solution for industrial carbon capture. Introduction Background on Gas Separation and Carbon Capture Technologies The continued rise in greenhouse gas emissions, particularly carbon dioxide (CO2), is a major driver of climate change, prompting efforts to capture and utilize CO2 effectively. Despite efforts to reduce CO2 emissions, significant challenges remain, particularly due to cost and the need for materials that are affordable, scalable, and efficient in capturing CO2. − There are three main technology pathways for CO2 capture: precombustion capture, postcombustion capture, and oxy-fuel combustion technology. , Postcombustion capture is the most mature and widely adopted industry technology, known for its excellent CO2 selectivity and capture efficiency. Methods include chemical absorption, physisorption, membrane processes, cryogenics, hydrate technology, and microbial approaches. In chemical absorption, commonly studied absorbents include ammonia solutions, sodium hydroxide solutions, and organic amine solutionssuch as monoethanolamine (MEA) mixed with an ionic liquid, which remains one of the most effective and widely used methods. Solid-phase porous adsorption using materials like activated carbon, zeolites, and metal–organic frameworks is considered promising due to its high adsorption capacity and low energy consumption. ,− The optimal CO2 adsorbent should have high capacity, selectivity, and ease of regeneration for economic viability. − Current research studies focus on materials like activated carbon, zeolites, MOFs, COFs, M41 series mesoporous materials, hydrotalcites, and geopolymers. − Although solid materials effectively adsorb CO2, their capacity is lower than that of absorption methods, prompting efforts to enhance adsorption through material modifications. Activated carbon (AC) is a cost-effective, stable adsorbent with a high surface area (500–3200 m2/g), derived from economical biomass feedstock sources like lignin, coconut shells, and wood. ,,− Carbon nanomaterials, such as nanotubes and graphene, enhance CO2 capture and can be functionalized for improved adsorption. − Physisorption allows efficient regeneration but has selectivity and capacity limitations. ,, Materials like zeolites utilize Coulombic interactions to selectively adsorb gases, such as N2 over O2, and are used commercially for air separation. − This selectivity is based on the charge distribution of the adsorbent interacting with gas quadrupole moments, validated by industrial applications. Lignin, Carbon Quantum Dot, and Computational Screening Methods Lignin, a byproduct of the pulp and paper industry, is an optimal material for large-scale carbon capture due to its abundance and cost-effectiveness. − With annual production exceeding 50 million tons and expected to reach 200 million tons, lignin is an economical alternative to petroleum-based materials. , Pyrolyzing lignin produces activated carbon (AC), which is used in various applications, including CO2 capture. ,− Lignin-based materials are being developed for use in carbon fibers, anodes, and electrodes. ,− Although not always the top performers, Lignin-based materials offer a cost-effective and plentiful alternative to traditional counterparts. Lignin-based activated carbon has shown promising selectivity for CO2 over N2, with potential enhancements through modifications like activated carbon fibers, which provide handling benefits and minimize pressure drop. − ,− Ideally, the thermodynamic upper bound for CO2 separation is defined by ultramicroporous MOFs, such as the SIFSIX family, which exhibit selectivities exceeding 200; , however, cost and stability remain significant barriers. Nitrogen-doped carbons offer a robust alternative, with Selmert et al. achieving selectivities up to 194 in nanofibers. In the specific domain of sustainable materials, a recent review by Barker-Rothschild et al. highlights lignin as a premier precursor. Mechanistically, Saha et al. established that nitrogen content in lignin-derived carbons directly correlates with capacity, a strategy recently advanced by Tkachenko et al., who reported a competitive CO2:N2 selectivity of 17.2 under ambient conditions for N-doped Kraft lignin adsorbents. While these values are promising, further enhancing surface heterogeneity via functionalization is necessary to surpass these experimental limits and approach the high efficiency of chemisorbents. To bridge the performance gap toward high-efficiency chemisorbents, recent advancements emphasize the direct functionalization and structural tuning of lignin-derived platforms. − Specific amination strategies, such as phenolation-enhanced grafting, double amination, urea-modification, and in situ amino-lignin production via the Mannich reaction, successfully introduce dense, reactive basic sites onto the carbon framework. − When these highly accessible amine groups are integrated with tailored hierarchical pores and nanoconfinement, , synergistic metal codoping, , or advanced aerogel composites, , the capture mechanism shifts strongly toward targeted chemisorption. This drastically improves both absolute CO2 uptake and CO2:N2 selectivity under diverse conditions. Extending these targeted amination concepts to nanoscale materials, such as amine-functionalized carbon quantum dots (CQDs), presents a highly promising pathway. By combining the rich basic surface chemistry of polyamines with the high surface-to-volume ratio of CQDs, the resulting composites can maximize selective interaction sites, driving CO2 separation efficiencies well beyond standard nitrogen-doped carbons. Quantum dots (QDs) are particles with nanoscale confinement in the three spatial dimensions, leading to unique physical and electronic behaviors. Carbon dots are biocompatible and fluorescent nanomaterials, offering low toxicity and facile synthesis. These zero-dimensional materials are promising tools for detecting food contaminants, additives, and microbes, with growing potential in food safety applications. CQDs are gaining attention for their optical properties and cost-effective functionalization,, making them suitable candidates for charge distribution control in membranes. When placed on activated carbon surfaces, CQDs can alter the charge distribution of materials, enhancing electrochemical performance. , CQDs, which can be synthesized from lignin, are graphene-like benzene rings attached to each other and expanded to form one-atom thick discs, that can be functionalized or doped to render the CQDs hydrophilic and improve their interaction with CO2. Creating a renewable composite material with increased CO2 selectivity can be achieved by integrating CQDs with a porous activated carbon surface. CQDs can contain various functional groups containing nitrogen, oxygen, etc. The functional groups are attached to the carbon atoms by removing hydrogen atoms from the terminal carbon atoms or by reducing the double-bond in the interior carbon atoms of the benzene rings to single-bonded ones and attaching hydroxyl functional groups instead. Broud et al. modeled nitrogen-containing CQDs either through amine functionalization or doping, having different sizes. Nitrogen and oxygen functional groups on CQDs can increase CO2 selectivity, with studies showing a selectivity ratio of 4.3 and 6.7 for CO2:N2 and CO2:O2, respectively, at 300 K. , Moreover, a system with a constant CO2:N2 selectivity of 4.3 can concentrate a 10% CO2 feed to 32.3% CO2 in one stage, 89.8% CO2 in three stages, and 99.4% in stage five. Oxygen-containing CQDs can be synthesized with carboxylic, hydroxyl, and epoxy functionality. , The types of functional group change in response to different temperature treatments. , On graphite oxide (GO) nanosheets, carboxylic groups are located at the edges, while hydroxyl and epoxy groups are on the basal plane. ,− Carboxylic groups are more stable but less abundant compared to epoxy or hydroxyl. ,,,− Furthermore, hydroxyl groups are more stable than epoxy groups. CQDs can be made on both defective and nondefective structures, with carbon-to-oxygen ratios ranging from 0.9 to 2.9. ,,,− Computational methods for screening materials for their potential as adsorbents with high capacity or selectivity , are categorized into quantum and classical approaches. Density Functional Theory (DFT) assesses CO2 binding energies. , For nonreactive separation processes, crystalline materials like Metal–Organic Frameworks (MOFs) and zeolites are explored. High-throughput screening uses classical simulations, like Molecular Dynamics (MD) and Monte Carlo (MC), typically under infinite dilution conditions. This study uses a combination of classical MC and MD simulations to consider temperature, pressure, and nonideal energetic factors in CO2 interactions , efficiently. Adsorption Isotherms Adsorption isotherms are essential for understanding adsorption processes, with various models tailored to different conditions and applications. The Langmuir isotherm models monolayer adsorption on homogeneous surfaces and is widely used in gas purification. The BET isotherm extends the Langmuir model to multilayer adsorption, crucial for characterizing porous materials. Hybrid models like Redlich-Peterson and Sips combine features of Langmuir and Freundlich, suitable for CO2 capture. Understanding these models is vital for optimizing adsorption processes in various applications. , Moreover, studies have highlighted the versatility of adsorption isotherms in evaluating novel materials, such as metal–organic frameworks (MOFs), for gas capture and separation processes. − It has been confirmed that through the strategic use of specific doping and functionalization, it is possible to customize the charge distribution of CQDs to exhibit a preference for interacting with the quadrupole moment of CO2 compared to N2 and O2. Modeling CQDs that are nitrogen-doped or amine-functionalized, as well as CQD-functionalization with carboxylic acid and hydroxyl groups in the basal plane, has been done based on experimental results. This work hypothesizes that increasing the surface density greatly increases the number of gas molecules interacting with the CQDs, strengthens the binding energy between the two, and increases the magnitude of the CO2:N2 and CO2:O2 selectivity due to the higher relative quadrupole moment of CO2. This work employs grand canonical Monte Carlo (GCMC) and classical molecular dynamics (MD) simulations to investigate the specified adsorption isotherms, followed by calculating the selectivity of CO2:N2 and CO2:O2 using various CQDs based on the size of the dots and their density on the substrate. Synthesizing an adsorbent composed of a lignin-based AC decorated with a lignin-based CQD, which would be sufficiently selective toward CO2 rather than N2 or O2, has the potential to result in a material that is environmentally friendly, industrially abundant, and cost-efficient for carbon capture globally. Simulation and Methods System Definition The CQDs examined in this study can be classified into two main groups: nitrogen-containing CQDs and oxygen-containing CQDs. Table provides details on the composition of these structures with various functionalities and doping, which are visually represented in Figure . The figure illustrates two types of N-containing models: (a) an amine-functionalized CQD with 32 amine groups and (b) an N-doped CQD containing 116 nitrogen-doped atoms, along with two types of O-containing models: (c) a carboxylic acid-functionalized dot with 12 carboxyl groups and (d) a hydroxyl-functionalized dot featuring 47 hydroxyl groups within the planar interior of the structure. These four CQD types have been selected based on their efficiency of capacity and selectivity for CO2 over other gas types according to the previous studies conducted on a large variety of surface chemistries. , Nitrogen-doped and amine-functionalized CQDs were selected because prior mechanistic studies establish that nitrogen content in lignin-derived carbons directly correlates with increased CO2 capacity and yields highly competitive CO2:N2 selectivity. , Carboxyl- and hydroxyl-functionalized CQDs were selected to represent oxygen-containing functional groups. During synthesis, carboxylic, hydroxyl, and epoxy functionalities occur. , However, we specifically modeled carboxyl and hydroxyl groups because the literature confirms they are significantly more stable than epoxy groups, ,,,− making them the most viable candidates for robust, real-world adsorption applications. Ultimately, our primary objective is to evaluate how increasing the surface density of CQDs on a substrate strengthens binding energy and increases CO2:N2 and CO2:O2 selectivity. By selecting these four specific, highly stable, and experimentally validated functional groups, our computational models provide realistic and targeted screening to identify the optimal chemistry for sustainable, lignin-based carbon capture materials. 1Composition of the Carbon Quantum Dots Used in This Study #CQD type# of C# of H# of N# of O1amine163213632 2N-doped151692116 3carboxyl1644104 244hydroxyl (interior)1632151 47aBase diameter: 7.3 nm. Composition of the Carbon Quantum Dots Used in This Study Base diameter: 7.3 nm. Atomic architecture of the four 7.3 nm diameter CQDs. The planar dots include (a) an amine-functionalized CQD with 32 amine groups, (b) an N-doped CQD containing 116 nitrogen-doped atoms, (c) a carboxylic acid-functionalized dot with 12 carboxyl groups, and (d) a hydroxyl-functionalized dot featuring 47 hydroxyl groups within the planar interior of the structure. Color legend: H: white, C: green, O: red, and N: blue. The interaction potentials for the CQDs have been taken from the OPLS-AA force field. , The OPLS interaction potentials have been used and compared with DFT calculations for the adsorption of CO2 and other gas adsorbates. , The OPLS-AA potential encompasses modes that account for intramolecular stretching, bending, dihedral and improper torsion, along with nonbonded interactions based on the Lennard-Jones potential and electrostatic interactions. In this study, five distinct gas systems were simulated. These included three pure gas systems: CO2, N2, and O2. Additionally, there were two types of flue gas: a dry flue gas composed of 84% N2, 12% CO2, and 4% O2, and a wet flue gas consisting of 69.5% N2, 9% CO2, 2.5% O2, and 19% H2O. All the interaction potentials for CO2, N2, and O2 gas molecules were taken from the same source. H2O was simulated with the TIP4P potential. − For these gases, it has been shown that both rigid and flexible models have equivalent equilibrium properties. , The CQDs are modeled as fully flexible molecules. In CO2 molecules, the point charges are located directly on the atoms. For the diatomic molecules N2 and O2, an extra point charge is positioned at the center. , In water molecules, according to the TIP4P model, there is an additional point charge placed at a fixed distance from the oxygen atom along the bisector of the HOH bond angle. In practical applications, CQDs will be placed within the internal pore spaces of a porous activated carbon substrate. This study examined two different models of carbon substrate surfaces. The first model is a graphite surface composed of five graphite layers in the z-direction, with 72 unit cells along the x-axis and 73 unit cells along the y-axis. The second model is an atomistic depiction of a lignin-based carbon composite (LBCC), which includes both graphitic nanocrystalline and amorphous carbon regions. The atomic structure of lignin carbon composite was adopted from the literature. Information about these surfaces is presented in Table . Note that the surface area calculation is based on the nominal surface area, the product of the x and y lengths of the simulation box. This calculation does not account for local surface roughness in LBCC. 2Simulation Details for the Two Different Model Surfacessurface# of C# of Htotalsurface area [nm2]graphite56,557056,557291.65LBCC62,76838,224100,992289.86 Simulation Details for the Two Different Model Surfaces The default surface is graphite, and it will be used in the whole study solely unless specified otherwise. On all surfaces, the constituent atoms of the substrate were maintained in a fixed position. The interaction potentials for graphite were characterized by ε = 0.06554 kcal mol–1 and σ = 3.4 Å. The potential from graphite was also used for the LBCC. The graphite surface is uncharged. The charges of the LBCC are taken from previous work. The nonbonded intermolecular forces were calculated using standard mixing rules applied to the respective force field parameters. This approach is critical for accurately resolving the physisorption thermodynamics, particularly the interaction between the strong quadrupole moment of CO2 and the diverse topographic and functional environments of the carbonaceous surfaces. Each simulation included gas molecules, a different number of CQDs, and a carbon surface. The x–y plane area for both graphite and the lignin composite constituted a square (Table ). The system applied periodic boundary conditions in all three dimensions (x, y, and z). Exploring the surface density through a realistic gas pressure value (1 atm) has been done systematically by applying respective system parameters first in a GCMC simulation to find the correct number of gas molecules above (and below) each substrate surface and potentially different numbers of CQDs from our four main dot types. The volume of the gas phase was set to allow a reasonable number of gas molecules, which ranged from 161 to 2601 in these simulations. Grand Canonical Monte Carlo Simulation This work employs both Grand Canonical Monte Carlo (GCMC) simulations and classical Molecular Dynamics (MD) simulations. The GCMC simulations are conducted in order to generate portions of adsorption isotherms, which relate bulk pressure to a density in the adsorbed phase. The MD simulations are conducted in the canonical (NVT) ensemble at the densities provided by the GCMC simulations corresponding to equilibrium with a bulk fluid at 1 atm and the specified temperature. The GCMC simulations were performed using an in-house code that has previously been used to model H2, CO2, and CH4 in MOFs , and silsesquioxane nanoporous materials, , as well as RDX in MOFs. − The types of moves that have been conducted on to the gas molecules through GCMC are translation, deletion, insertion, rotation. The acceptance of these moves is evaluated by comparing the internal energy of the system before and after each trial move, following standard Metropolis acceptance criteria. The simulations utilized up to 22,000,000 GCMC steps, depending on the required equilibration period. Equilibration was verified by monitoring the total internal energy of the system until it reached a stable plateau with minimal fluctuations. Each point on the adsorption isotherm is accomplished with two GCMC simulations performed at the same temperature (T) and chemical potential (μ). The first simulation is of the bulk gas phase, which provides the pressure at T and μ. The second simulation is of the adsorbed phase, which provides the adsorbed phase density. Because T and μ are the same in both simulations, these two phases are in thermodynamic equilibrium. GCMC simulations were only performed for pure component gases. For MD simulations of mixtures, the number of each molecule (CO2, N2, and O2 in a dry flue gas, and CO2, N2, O2, and H2O in a wet flue gas) was determined as follows. Using the flue gas compositions from the experimental literature, we included the number of molecules of each species from their pure component simulation, weighted by the atom fraction in the gas mixture. Molecular Dynamics Simulation Classical Molecular Dynamics simulations were performed using the LAMMPS software. Separate configuration files, detailing positions and interaction potentials, were generated for each CQD, each type of gas (both pure and mixture), and both surface models. These elements were then combined into a single configuration for each simulation. This approach enabled rapid, automated integration of CQD, gas, and surface. Figure shows examples of such a configuration. Initial configuration of the MD simulation for the system with CO2, graphite substrate, and (a) functionalized CQDs of diameter 75.1 Å with 32 amine groups and (b) functionalized CQDs of diameter 32.1 Å with 13 amine groups. Color legend: H: white, C of CQD: green, O: red, N: blue, and C of graphite and CO2: gray. To avoid physical boundary conditions, the system is periodic in all three dimensions, including that normal to the surface, as shown in Figure . Thus, there are two equivalent surfaces on the top and bottom. Each surface contains the same number of CQDs. Capacity and selectivity results are averaged over both sides of the surface. The MD simulations are performed in the canonical ensemble. The number of gas molecules in each system has been taken from GCMC simulations to correspond to a bulk pressure of 1 atm. GCMC simulations were not performed in our previous study, in which the systems were compared under constant temperature and density (rather than constant temperature and pressure) leading to varying bulk pressures. Table S.1 provides GCMC results in a system with CO2 and one or four functionalized CQDs of diameter 7.5 nm with 32 amines at each side of the graphite at 300 K. A suite of simulations was performed with unique combinations of CQDs, gases, surfaces, and temperatures. Most simulations were performed at 300 K. Simulations of systems with wet flue gas were performed at 500 K to avoid condensation of water on the surface. Each simulation consisted of 3,000,000 steps, using a time step of 1 fs. The cutoff value for all interaction potentials was established at 12 Å. The Ewald method was used to calculate long-range electrostatic contributions to the potential. The r-ReSPA integrator was applied, with intramolecular modes (bond, angle, dihedral, and improper torsion) evaluated at 1/8, 14 , 12 , and 12 of the base time step for nonbonded interactions, respectively. , To control the temperature, the Nosé-Hoover thermostat was used. , The system underwent equilibration between 1,000,000 to 140,000,000 steps before data collection. Positions were saved to trajectory files every 1,000 steps. The MD simulations provide various thermodynamic and structural data. The interaction energy between the CQD and the surface is defined as the binding energy of the CQD. Gas molecules are distributed between a gas phase and an adsorbed phase, with continuous exchange occurring between these phases. To accurately estimate the gas binding energy, the number of interfacial gas molecules, and the selectivity, it is essential to have clear definitions. The number of gas molecules at the interface is defined as any molecule whose center-of-mass is within the cutoff distance (12 Å in this study) from any atom in the CQD or the surface. This 12 Å region accommodates multiple adsorption layers and a transition zone to the bulk, rather than a singular bound state. The instantaneous count of atoms is determined through postsimulation analysis of the trajectory files and is averaged over the frames saved after equilibration. In general, the selectivity of component 1 to component 2 is defined as − S=N1,aN1,bN2,aN2,b 1 where N 1,a is the number of interacting molecules of type 1 in the adsorbed (a) phase and N 1,b is the total number of molecules of type 1 in the bulk (b) phase. For MD simulations of gas mixtures (dry and wet flue gases) this formula is used directly. Selectivity can also be estimated from pure component simulations using Ideal Adsorption Solution Theory. , In this case, an estimate of the selectivity, which we hereafter call a selectivity index, was computed using S≈N1,aN2,a 2 Results and Discussion Adsorption Isotherms Adsorption isotherms of select systems were generated at 300 K from the GCMC simulations. For example, in Figure , the adsorption isotherms for CO2 in a system with a graphite substrate with either CQDs functionalized with 32 amine groups or CQDs doped with 116 nitrogen atoms are shown. Adsorption isotherm of CO2 on a graphite substrate decorated with four functionalized CQDs with 32 amine groups (blue) or nitrogen-doped CQDs with 116 nitrogen atoms (orange) on each side of the substrate surface. The loading is higher for the system with nitrogen-doped CQDs, indicating that those systems have more capacity for adsorbing CO2 gas molecules. Table S.2 shows the results of the adsorption Isotherm data related to the systems with CO2 gas molecules and 4 CQDs on each side of the graphite substrate at 300 K. The adsorption isotherm of the system with four CQDs (on each side of the substrate) functionalized with 32 amine groups, graphite substrate, and CO2 gas molecules is shown in Figure at 300 K. Because the MD simulation is two-phase, we have also shown snapshots of the system that provide a visual cue into the definition of gas in the adsorbed and bulk phases. As noted above, the adsorbed phase ends 12 Å from the surface, which is the cutoff in the Lennard-Jones potential. The adsorption isotherm is shown as the total amount of CO2 present (both adsorbed and bulk), and the difference. We also plot the corresponding amount of material in the bulk volume using the density obtained from single-phase bulk simulations in the absence of a surface. The amount of gas in the bulk volume from the two-phase simulation is similar to but slightly larger than that in a similar volume of bulk gas, indicating that the definition of the adsorbed gas molecules is conservative. (We have not overcounted molecules in the bulk gas as adsorbed.) Adsorption isotherm of the system with CO2, graphite substrate, and four amine-functionalized CQDs on each side of the substrate surface. Blue curve: the total number of CO2 molecules in the system. Orange curve: the number of adsorbed CO2 molecules. Yellow curve: the number of CO2 molecules that would occupy the bulk volume based on the density from the bulk gas GCMC simulation. Gray curve: the mathematical difference between the total and the adsorbed CO2 molecules from the MD simulation. Color legend as in Figure . Pure Gases on CQD-Decorated Graphite MD simulations of pure gases were performed on a graphite surface decorated with different surface densities of CQDs at 300 K. A majority of the simulations focused on CQDs with a nominal diameter of 7.5 nm. We conducted a preliminary investigation of the effect of CQD size with a comparison of a system with (i) 22 functionalized CQDs with a diameter of 3.2 nm and 13 amine groups and with (ii) four CQDs of diameter 7.5 nm functionalized with 32 amine groups on each side of the substrate. The number of dots for the small-sized CQDs has been specified in a way that the total surface density of all the dots on the graphite substrate surface is equal to that of their counterparts with four large-sized CQDs. Figure shows a snapshot of these two systems. Table S.3 presents the three main results: the average adsorbent capacity, which represents the number of gas molecules interacting, the average binding energy between the adsorbent and the gas molecules, and the CO2:N2 and CO2:O2 selectivities for these two systems at 300 K. The capacity for CO2 and O2 of the adsorbents with the larger CQDs are higher than that with the smaller CQDs, but almost the same for N2. The binding energies and the selectivities do not have a significant difference in the two systems, indicating that the size of the CQDs is a significant factor in selectivity. The effect of the surface density of CQDs is studied for the systems with 0 to 4 CQDs of diameter 7.5 nm on each side of the graphite substrate. In the left column of Figure , results are shown for the amine-functionalized CQD. The top row reports the number of interacting gas molecules per surface area. The middle row (the binding energy values) includes two different calculations, namely the binding energy between the gas molecules and just the CQDs, and the binding energy between the gas molecules and the whole adsorbent (CQDs plus graphite substrate). The bottom row reports the selectivity index. Simulation results of pure gas simulations exposed to 0 to 4 N-containing CQDs on each side of the graphite substrate at 300 K. Left column (a–c): amine-functionalized CQDs with 32 amine groups. Right column (d–f): N-doped CQDs with 116 nitrogen atoms. Top row (a,d): number of gas molecules interacting with the adsorbent normalized by surface area of the adsorbent. Middle row (b,e): interaction energy of gas with CQD (solid lines) and of gas with CQDs + graphite (dashed lines) normalized by the number of interacting gas molecules. Bottom row (c,f): selectivity. For the models with amine-functionalized CQDs, we observed preferential adsorption of CO2, followed by O2 and then N2. Interestingly, the capacity of the adsorbents for CO2 gas molecules increases with increasing surface density of CQD, while the capacity for N2 and O2 does not change significantly. The behavior in the capacity can be explained by an examination of the binding energy. The binding energy of the gas (both to the CQD and total CQD and surface) is most favorable for CO2. For CO2, the contribution of CQDs to the total binding energy increases with the increasing CQD surface density, to a point that the gas-CQD binding energy is over 50% of the total. The behavior of the binding energy is different for N2 and O2. The contribution of the CQD to the binding energy increases with increasing CQD surface density, as was the case for CO2, though to a lesser extent. However, the total binding energy becomes less favorable for N2 and O2. These trends in capacity result in the selectivity index for these systems increasing with CQD density for the amine-functionalized CQDs. Moreover, the selectivity of CO2:N2 is higher than CO2:O2, with the highest numbers of 2.07 and 1.81, respectively. All in all, it can be concluded that CO2’s larger quadrupole moment compared to the other gas type molecules results in stronger interactions with amine groups, and this effect is enhanced as the number of CQDs on the surface increases due to the increased charge distribution of the adsorbent. In the right column of Figure , we report capacity, binding energy, and selectivity for N-doped CQDs as a function of CQD density. Again, CO2 binds preferentially, followed by O2, then N2. The magnitude of the capacity is larger for the N-doped CQDs, a maximum of 6.9 CO2/nm2 compared to 1.7 CO2/nm2 for the amine-functionalized CQD. Again, the capacity for CO2 gas molecules increases with increasing CQD density, while it does not have a significant change for N2 and O2. The binding energy for all the gas types is roughly the same for both the amine-functionalized and N-doped CQDs. Thus, the increase in capacity for CO2 is not due to energetically more favorable sites but rather due to more binding sites. In the amine-functionalized CQDs, the amine groups are bound to the edge of the dot, whereas in the N-doped CQDs, the doping occurs throughout the CQD. The selectivity index for these systems increases by inserting dots on the surface to a certain limit. The selectivity of CO2:N2 and CO2:O2 is 1.76 to 1.92 and 1.63 to 1.67, respectively. Simulations were also performed of CQDs functionalized with oxygen-containing groups. The carboxyl groups bind to the edges of the CQD, while the hydroxyl groups bind to the interior of the CQD. ,− ,− In Figure , we report the capacity, binding energy, and selectivity index for carboxyl-functionalized CQDs (left column) and hydroxyl-functionalized CQDs (right column). Simulation results of pure gas simulations exposed to 0 to 4 O-containing CQDs on each side of the graphite substrate at 300 K. Left column (a–c): carboxyl-functionalized CQDs with 12 carboxylic-acid groups. Right column (d–f): hydroxyl-functionalized CQDs with 47 interior hydroxyl groups. Top row (a,d): number of gas molecules interacting with the adsorbent normalized by surface area of the adsorbent. Middle row (b,e): interaction energy of gas with CQD (solid lines) and of gas with CQDs + graphite (dashed lines) normalized by the number of interacting gas molecules. Bottom row (c,f): selectivity. Similar to the systems that have CQDs functionalized with nitrogen-containing groups, the capacity for CO2 remains greater than that for N2 or O2 at all surface densities of CQDs, and the presence of the CQDs has no significant change for the systems with N2 and O2 gas molecules. The binding energy for CO2 becomes more favorable, to a greater extent, compared to the systems with N-containing CQDs with the addition of the CQDs. The selectivity index for these systems generally increases with increasing CQD surface density. The selectivity index of CO2:N2 is higher than that of CO2:O2, with maximum values for CO2:N2 and CO2:O2 of 1.98 and 1.79. The results for simulations including hydroxyl-functionalized CQDs are shown in the right column of Figure . As was the case for the previous three CQDs, in these systems, the capacity for CO2 is highest. The maximum CO2 capacity for the hydroxyl-functionalized CQDs was 2.4 CO2/nm2 compared to 1.8 CO2/nm2 for the systems with carboxyl-functionalized CQDs. The binding energy for CO2 in these systems increases in magnitude more rapidly with increasing CQD surface density compared to the other CQDs. The selectivity of CO2:N2 is higher than CO2:O2, with the highest numbers of 1.93 and 1.83, respectively. Taking the entire data set of the systems with pure gases on CQD-decorated graphite, the highest selectivity indexes for CO2:N2 and CO2:O2 belong to the one with 4 amine-functionalized and 4 hydroxyl-functionalized CQDs on each side of the surface, respectively. Furthermore, the highest capacity in CO2 capture belongs to the system with 4 N-doped CQDs on each side of the surface by a significant amount. The simulation data for the systems with pure gas components on graphite can be found in Table S.4. Flue Gases on CQD-Decorated Graphite Simulations for dry and wet flue gas have been conducted in a system with a graphite substrate surface and 0 to 4 amine-functionalized CQDs with 32 amine groups on each side of the substrate. In Figure , we report results for dry and wet flue gases. We discuss the dry flue gas results first. As a reminder, the composition of the dry flue gas was 84% N2, 12% CO2, and 4% O2. In Figure a, we observe that the number of interacting molecules reflects the composition of the flue gas, with more N2 being adsorbed, followed by CO2, then O2. In Figure b, the number of molecules of each species in the bulk phase is also reported, which aids in putting into proper perspective the relative preference for the adsorption of CO2. The contribution of the CQDs to the average binding energy increases with increasing surface density, while the total binding energy remains in a certain range. The selectivity for CO2 is much higher, on average, than predicted from the selectivity index based on the pure component simulations. The highest CO2:N2 and CO2:O2 selectivities in these dry flue gas simulations are 12.1 and 7.8 respectively. We observed this same trend in earlier work and it is not unexpected. In the Supporting Information document, we provide a theoretical rationale for this observation. Using a single-component Langmuir adsorption model and a binary mixture Langmuir adsorption model, we show that the selectivities obtained from the simulation of the mixture will always be greater than the selectivities obtained from simulations of single-component gases. The larger error bars in the selectivity reflect the uncertainty of the small composition of CO2 and O2 in the flue gas. Table S.5 reports the selectivity values for systems with flue gases on CQD-decorated graphite. Simulation results of flue gas simulations exposed to 0 to 4 amine-functionalized CQDs with 32 amine groups on each side of the graphite substrate. Left column (a–d): dry flue gas at 300 K. Right column (e–h): wet flue gas at 500 K. Top row (a,e): number of gas molecules interacting with the adsorbent normalized by the surface area of the adsorbent. Second row (b,f): number of gas molecules NOT interacting with the adsorbent normalized by the surface area of the adsorbent. Third row (c,g): interaction energy of gas with CQD (solid lines) and of gas with CQDs + graphite (dashed lines) normalized by the number of interacting gas molecules. Bottom row (d,h): selectivity. Simulations of a wet flue gas at 500 K were conducted in a system with a graphite substrate surface and 0 to 4 functionalized CQDs with 32 amine groups on each side of the substrate. The composition of the wet flue gas was 69.5% N2, 9% CO2, 2.5% O2, and 19% H2O. In Figure e,f, the number of molecules in the adsorbed and bulk phase is reported as a function of CQD surface density. Interestingly, the number of interacting gas molecules does not change significantly with the surface density, which remains at very low values for O2. The average binding energy again hovers within a range versus the CQD surface density. The selectivity for CO2 is not so different from the values predicted from the selectivity index based on the pure component simulations, which are much lower than those in the systems with dry flue gas. Assessing these results, the efficiency of the systems with dry and wet flue gas depends less on the surface density, while the systems with dry flue gas have overall higher values for selectivity, compared to both the ones with wet flue and pure gas components. Clearly, the presence of water on the surface disrupts the selectivity for CO2. In Figure S.1a, the radial distribution functions (RDFs) are plotted for each component of the dry flue gas on graphite without CQDs at 300 K. By definition, RDFs are normalized by the average density of each component, so the magnitude of the RDF does not visually correlate with the amount of each gas, interacting with the surface, which is instead given by the first points (at 0 CQD surface density) in Figure a. The regular features in the RDFs correspond to the layered structure of the graphite and are sharpest for CO2, indicating a slightly more regular structure on the surface relative to N2 and O2. In Figure S.1b, the RDFs are plotted for the wet flue gas on the LBCC without CQDs at 500 K. This RDF corresponds to the first points (at 0 CQD surface density) in Figure e. The combination of the presence of water and increase in temperature serves to reduce the amount and degree of structure in all of the other adsorbed gas molecules. Lignin-Based Carbon Composite The LBCC surface models a carbonized lignin material and was generated to reproduce experimental characterization based on microscopy and total neutron scattering. The material is composed of two domainsgraphitic nanocrystallites and an amorphous domain formed of disordered fragments of graphene. The LBCC surface differs from graphite in two important ways. First, the edges of both graphitic nanocrystallites and graphene fragments are passivated with hydrogen. This passivation induces a partial charge separation in the surface at the C–H bonds. Second, the LBCC surface is rough at the atomic scale (Figure ). In previous work, a comparison of graphite and LBCC was performed using MD simulation at fixed density (rather than at a constant pressure of 1 atm as performed in this current work). In that work, it was observed that the capacity for all gases was higher for the LBCC substrate compared to graphite due to a more favorable binding energy. The selectivity of CO2 to N2 was also higher for the LBCC system. Snapshot of a lignin-based carbon composite substrate and CO2 gas. The system is periodic in all three dimensions. The top and bottom surfaces are equivalent, as in Figure . Color legend: H: white, O: red, and C of LBCC and CO2: gray. The current study compares adsorption on the LBCC surface much more extensively, including mixtures for the first time, and at a common pressure of 1 atm. In Figure , results for MD simulations of pure gases, CO2 and N2, are reported on the graphite and LBCC surface without the addition of CQDs. The capacity of the LBCC surface is greater than that of graphite for both gases. The binding energy is more favorable on the LBCC surface for both gases. The CO2:N2 selectivity is lower on the LBCC surface because it adsorbs more N2 compared to graphite. Simulation results of pure gas simulations exposed to the graphite (blue) and LBCC (orange) surface at 300 K. Top (a): number of gas molecules interacting with the surface normalized by the total number of gas molecules in the system. Middle (b): interaction energy of gas with the surface normalized by the number of interacting gas molecules. Bottom (c): selectivity index. Simulations of dry and wet flue gases were also performed with an LBCC substrate with and without CQDs. The results for the dry flue gas over the amine-functionalized and N-doped CQDs on LBCC are shown in Figure . The number of interacting molecules for all species and the binding energies shown in Figure appear qualitatively similar to those shown for the pure component gases in Figure or for the dry flue gas over graphite shown in Figure a,c. However, the selectivities are an order of magnitude higher. It is true that the LBCC adsorbent adsorbs more N2 compared to graphite, and an increase in the amount of N2 adsorbed should decrease the selectivity. However, the reason that the selectivity is higher for the LBCC is due to the fact that the amount of CO2 in the gas phase is much smaller for the LBCC adsorbent than for graphite. Per eq S2, this drives the rise in selectivity. The cause of the high selectivity is due to the fact that virtually no CO2 is left in the bulk phase in some cases. Simulation results of dry flue gas simulations exposed to 0, 1, or 4 N-containing CQDs on each side of the LBCC substrate at 300 K. Left column (a–d): amine-functionalized CQDs with 32 amine groups. Right column (e–h): N-doped CQDs with 116 nitrogen atoms. Top row (a,e): number of gas molecules interacting with the adsorbent normalized by the surface area of the adsorbent. Second row (b,f): number of gas molecules NOT interacting with the adsorbent normalized by the surface area of the adsorbent. Third row (c,g): interaction energy of gas with CQD (solid lines) and of gas with CQDs + graphite (dashed lines) normalized by the number of interacting gas molecules. Bottom row (d,h): selectivity. Note the nonlinear scale on the x-axis. The size of the bulk reservoir in the two-phase MD simulations was increased for the gas simulations over LBCC in order to get better statistics on just how little CO2 actually was left in the bulk phase. Still, at some instances during a simulation there was not a single molecule of CO2 in the bulk phase, resulting in a selectivity of infinity. We have capped high selectivities at 50 to avoid artifacts based on simulation size. These high selectivities are observed in the LBCC surface both in the absence and presence of CQDs. In equilibrium with an infinite reservoir, some CO2 would remain in the bulk state. However, given the finite size of the reservoir in the MD simulations, the fact that virtually all CO2 is adsorbed is reasonable and indicative of very strong selectivity for CO2. In the case of the N-containing CQDs (both amine-functionalized and N-doped), the presence of the CQDs at any surface density does not impact the selectivity. The results for the dry flue gas over the O-containing CQDs (carboxyl-functionalized and hydroxyl-functionalized) on LBCC are shown in Figure . For the hydroxyl-functionalized CQDs, we observe similar qualitative behavior in the selectivity at any CQD surface density as was observed with the N-containing CQDs. However, in the cases with carboxyl-functionalized CQDs, while the overall qualitative behavior in the CO2:N2 selectivity at any CQD surface density remains high (though with a nonmonotonic trend due to the statistical noise), the CO2:O2 decreases with the increase in CQD surface density. Simulation results of dry flue gas simulations exposed to 0, 1, or 4 O-containing CQDs on each side of the LBCC substrate at 300 K. Left column (a-d): carboxylic acid-functionalized CQDs with 12 carboxyl groups. Right column (e–h): hydroxyl-functionalized CQDs with 47 internal hydroxyl groups. Top row (a,e): number of gas molecules interacting with the adsorbent normalized by the surface area of the adsorbent. Second row (b,f): number of gas molecules NOT interacting with the adsorbent normalized by the surface area of the adsorbent. Third row (c,g): interaction energy of gas with CQD (solid lines) and of gas with CQDs + graphite (dashed lines) normalized by the number of interacting gas molecules. Bottom row (d,h): selectivity. Note the nonlinear scale on the x-axis. A wet flue gas was also simulated over the LBCC surface at 500 K while varying the surface density of CQDs. The results for the wet flue gas over the amine-functionalized and N-doped CQDs on LBCC are shown in Figure . The selectivity for CO2:H2O is near zero, indicating preferential adsorption of H2O. The charged groups on the CQDs act as an electrostatic attraction to the H2O; hence, there is virtually no H2O left in the gas phase. Despite the presence of water in the system, in the absence of CQDs, we observe that LBCC surface is able to maintain a high selectivity for CO2:N2 and CO2:O2. This is due to the roughness and charged nature of the LBCC surface. We also observe for the wet flue gas with N-containing CQDs that the addition of either the amine-functionalized or N-doped CQDs results in a drastic decrease in both CO2:N2 and CO2:O2 selectivities. The presence of both water and CQDs simultaneously on the surface results in a degradation of selectivity for CO2 that does not result when either water or CQDs alone are present. Simulation results of wet flue gas simulations exposed to 0, 1, or 4 N-containing CQDs on each side of the LBCC substrate at 500 K. Left column (a–d): amine-functionalized CQDs with 32 amine groups. Right column (e–h): N-doped CQDs with 116 nitrogen atoms. Top row (a,e): number of gas molecules interacting with the adsorbent normalized by the surface area of the adsorbent. Second row (b,f): number of gas molecules NOT interacting with the adsorbent normalized by the surface area of the adsorbent. Third row (c,g): interaction energy of gas with CQD (solid lines) and of gas with CQDs + graphite (dashed lines) normalized by the number of interacting gas molecules. Bottom row (d,h): selectivity. Note the nonlinear scale on the x-axis. In Figure S.1c, the radial distribution functions (RDFs) are plotted for each component of the dry flue gas on the LBCC without CQDs at 300 K. These RDFs correspond to the first points (at 0 CQD surface density) in Figure a. As with graphite, we observe the most structure in the adsorbed CO2. The adsorbed O2 shows the least structure. In Figure S.1d, the RDFs are plotted for the wet flue gas on the LBCC without CQDs at 500 K. This RDF corresponds to the first points (at 0 CQD surface density) in Figure a. Here we observe clear evidence that the significant adsorption of water to the LBCC surface does not strongly effect CO2 adsorption but does strongly reduce N2 and O2 adsorption. In Figure S.2, RDFs are presented for the dry (S.2a) and wet (S.2b) flue gases on the LBCC decorated with the highest surface density of N-doped CQDs. The RDFs for the dry flue gas correspond to the last points (8 N-doped CQDs) in Figure e. These RDFs show an increase in the adsorption of all three gases relative to the LBCC without CQDs. This results in a loss of selectivity. However, the resulting selectivities were still greater than 50, because virtually all the CO2 in the simulation was adsorbed. See Figure h. The RDFs for the wet flue gas correspond to the last points (8 N-doped CQDs) in Figure e. The RDFs show that water binds most closely to the adsorbent at the expense of all other gases, including CO2. While virtually all CO2 was adsorbed in the absence of CQDs, some CO2 molecules are now shifted to the bulk phase, which results in the decrease in selectivity as shown in Figure h. The results for the wet flue gas over the carboxyl-functionalized and hydroxyl-functionalized CQDs on LBCC are shown in Figure . The behavior is mostly the same as that observed with N-containing CQDs on an LBCC substrate. The surface preferentially adsorbs H2O. The addition of O-containing CQDs serves to reduce the selectivity for CO2:N2 and CO2:O2. The decrease in selectivity does not occur until higher surface loadings for the carboxyl-functionalized CQDs. Table S.6 presents the selectivity values for the systems with flue gases on CQD-decorated LBCC. Simulation results of wet flue gas simulations exposed to 0, 1, or 4 O-containing CQDs on each side of the LBCC substrate at 500 K. Left column (a–d): carboxylic acid-functionalized CQDs with 12 carboxyl groups. Right column (e–h): hydroxyl-functionalized CQDs with 47 interior hydroxyl groups. Top row (a,e): number of gas molecules interacting with the adsorbent normalized by the surface area of the adsorbent. Second row (b,f): number of gas molecules NOT interacting with the adsorbent normalized by the surface area of the adsorbent. Third row (c,g): interaction energy of gas with CQD (solid lines) and of gas with CQDs + graphite (dashed lines) normalized by the number of interacting gas molecules. Bottom row (d,h): selectivity. Note the nonlinear scale on the x-axis. To summarize the results of the flue gas mixtures on the LBCC, we observe high CO2:N2 and CO2:O2 selectivities on the bare LBCC in the absence or presence of water. While it is true that, in the case of the wet flue gas, H2O adsorbs to the surface, it does not impact the CO2:N2 and CO2:O2 selectivities on bare LBCC. The addition of functionalized CQDs in the dry flue gas systems has little effect, apart from carboxyl-functionalized CQDs, which lowers CO2:O2 selectivity. The addition of functionalized CQDs in the wet flue gas systems negatively impacts the selectivity for all CQDs. However, in a system with a dry flue gas, the capacity of the adsorbent for capturing CO2 increases with increasing surface density in almost all cases, especially in the cases with N-doped dots. Conclusions This comprehensive investigation into lignin-derived adsorbents for carbon dioxide capture has yielded several significant findings with important implications for sustainable carbon capture technology. Our simulations demonstrate that both graphite and lignin-based carbon composite (LBCC) substrates functionalized with CQDs can selectively adsorb CO2 from gas mixtures, though their mechanisms and effectiveness differ considerably. On systems with pure gas components and graphite substrates, all four types of CQDs (amine-functionalized, nitrogen-doped, carboxyl-functionalized, and hydroxyl-functionalized) showed preferential adsorption of CO2 over N2 and O2, with selectivity indexes ranging from 1.53 to 2.07. The surface density of CQDs significantly influenced both capacity and selectivity, with higher densities generally improving performance. Notably, nitrogen-doped CQDs exhibited the highest capacity for gas adsorption while maintaining good selectivity. Simulations of dry and wet flue gas on graphite showed that the calculated selectivity based on mixture simulations was higher than that based on pure component simulations. This enhancement is fundamentally due to competitive adsorption effects. A full statistical mechanical derivation based on Langmuir adsorption isotherms, demonstrating why mixture selectivity mathematically exceeds pure component selectivity under these conditions, is provided in Section S1 of the Supporting Information. The surface density of CQDs on graphite does not make a noticeable change in the efficiency of these systems with respect to either the capacity or the selectivity. The most striking discovery was the performance of the bare LBCC substrate, which demonstrated CO2:N2 and CO2:O2 selectivity values exceeding 50 in flue gas mixtures at 1 atm and 300 K and in wet flue gas mixtures at 1 atm and 500 K. This performance is particularly notable when contrasted with state-of-the-art experimental N-doped lignin carbons, which report ambient CO2:N2 selectivities of approximately 17.2. The LBCC models suggest that atomistic roughness and partial charge distribution can significantly increase selectivity. The addition of CQDs to LBCC substrates with dry flue gas in most cases does not impact selectivity. The exception is the CO2:O2 selectivity in cases with carboxyl-functionalized CQDs, in which those selectivity values decrease with increasing the CQD surface density. The capacity of the adsorbent for capturing CO2 molecules, however, increases by adding CQDs to the surface for the systems with dry flue on LBCC substrates. This finding is consistent with experimental trends for bulk amine- and nitrogen-functionalized lignin networks, ,,,− confirming that nanoscale amination successfully increases capacity without compromising the intrinsic selectivity of the substrate. In the systems with wet flue gas, the addition of CQDs to LBCC substrates, in most cases, has no effect on CO2 capacity of the adsorbent while degrades the selectivity of CO2:N2 and CO2:O2 in most cases, adding to the fact that CO2:H2O is almost zero in all cases, owing to the hydrophilic characteristic of an LBCC surface. Interestingly, while bulk amine-functionalized lignin maintains or enhances experimental performance under humid conditions, a critical feature for ambient applications like direct air capture, − our atomistic models reveal a fundamental nanoscale limitation: the intrinsic hydrophilicity of the underlying LBCC substrate actively outcompetes the CQD amine sites for H2O, ultimately degrading CO2 selectivity. These results suggest that optimizing lignin-based activated carbon with CQDs can yield environmentally friendly, cost-effective, and industrially viable materials for carbon capture. The LBCC substrate has high selectivity for CO2:N2 and CO2:O2 in both dry and wet flue gases. In the case of the dry flue gas, adsorptive capacity can be increased with the decoration of CQDs, without practical loss in selectivity. Future work should focus on experimental validation of these computational findings. Supplementary Material