Experimental Evaluation of Commercial Molecular Sieves 13X, 4A, and JLPM3 for Sustainable Direct Air CO₂ Capture from Humid Air via Temperature-Swing Adsorption: “Sieve the Atmosphere”

Abstract

Direct air capture (DAC) of CO₂ via temperature-swing adsorption (TSA) can support sustainable carbon dioxide removal, but only if sorbents regenerate with low energy demand and maintain performance under humid ambient air. In this paper, we evaluate three commercial molecular sieves (JLPM3, 13X, and 4A) in packed-bed tests using humid ambient air. We compared 40 g samples as received with 200 g samples conditioned for 12 days at 100 °C to emulate prolonged exposure to regeneration temperature (the cumulative effect of many heating/desorption cycles); all cycle-stabilized uptake values are reported from the conditioned materials. JLPM3 delivered the highest stabilized CO₂ uptake (0.24 ± 0.01 mmol·g⁻¹), consistent with a combined physisorption/chemisorption mechanism. Its higher total porosity (26.190%) and smaller mesopores (7.569 nm width) promoted rapid mass transfer and site accessibility, while slightly greater micropore area (710.285 m²·g⁻¹) and volume (0.267 cm³·g⁻¹) than 13X supported its marginally higher capacity. Evidence of partial structural degradation under mechanical and thermal stress indicates that minimizing strain during cycling will be important for scale-up and for reducing sorbent replacement. Conditioning at 100 °C activated additional chemisorption sites across all sieves but reduced physisorption capacity. Importantly, a ~100 °C desorption step fully regenerated physisorbed CO₂ while purging moisture from zeolite pores, indicating that low-temperature TSA (compatible with low-grade or waste heat) can replace harsher 300 °C regeneration and lower energy demand. CO₂–H₂O competition experiments confirmed substantial site occupancy by water vapor, which limits capture under humid conditions and motivates water management strategies. Overall, maximizing DAC performance requires tailoring pore structure and operating conditions while preserving sorbent integrity; JLPM3 emerges as a promising candidate for more energy- and resource-efficient DAC.

1. Introduction

The ongoing rise in atmospheric carbon dioxide (CO₂) concentrations is one of the most pressing challenges of our time, with levels exceeding 420 ppm as of 2024, an alarming increase from pre-industrial ~280 ppm [1,2]. This rapid accumulation, driven by fossil fuel combustion, deforestation, and industrial processes, has already led to profound climatic changes, including rising global temperatures, more extreme weather events, and widespread ecological disruption [3]. While deep emissions cuts via renewable energy, efficiency improvements, and ecosystem stewardship are essential, mitigation alone will not be enough. Even if emissions ceased immediately, excess atmospheric CO₂ would continue to warm the climate for centuries, underscoring the need for carbon dioxide removal (CDR) alongside rapid decarbonization [4].

From a sustainability standpoint, DAC must deliver net-negative CO₂ when energy, water, and material inputs are considered. In adsorption-based DAC, regeneration heat and moisture management strongly influence operating energy, lifecycle emissions, and cost. Likewise, sorbent lifetime and mechanical robustness affect resource demand and waste generation, making cyclic stability a key sustainability metric for scalable deployment.

Within the CDR portfolio, direct air capture (DAC) offers a scalable, location-flexible route to remove CO₂ directly from ambient air and to address residual emissions from hard-to-abate sectors [5,6]. Although capturing CO₂ at ~420 ppm is technically demanding, DAC can be paired with permanent geological storage (CCS) or with utilization pathways (CCU) to enable negative emissions while supporting a range of end-uses [7]. Achieving practical DAC at scale hinges on sorbent materials that couple high capacity and selectivity with stability and regenerability under realistic operating conditions, including humidity [8]. Zeolitic molecular sieves are leading physisorbents for temperature-swing adsorption (TSA) because their framework chemistry, pore structure, and surface properties can be tuned for CO₂ capture [6,9].

Commercial zeolites remain key benchmark sorbents for adsorption-based CO₂ capture because their framework topology and high cation density can provide strong CO₂ affinity at low partial pressure, while pore architecture and binder-related textural porosity govern kinetics and working capacity. FAU-type materials such as 13X are among the most widely studied zeolites for CO₂ separation and serve as common reference sorbents due to their large-pore access and high micropore volume [6,9]. In contrast, LTA-type zeolites such as 4A are highly hydrophilic and widely used as desiccants; under mixed CO₂/H₂O feeds their performance is often dominated by water uptake and transport constraints associated with narrow pore windows, which can strongly limit effective CO₂ working capacity under humid conditions [6,10]. In addition to standard grades, industrial molecular sieves used for air pre-purification (e.g., JLPM3) are designed to remove CO₂/H₂O upstream of cryogenic air separation, but comparative datasets under humid ambient air and packed-bed TSA conditions remain less common, motivating direct benchmarking under realistic feeds.

However, real systems operate with humid ambient air and undergo many thermal cycles near typical regeneration temperatures, factors that can shift the balance between physisorption and chemisorption and adversely impact long-term sorbent performance. In particular, moisture in ambient air competitively occupies adsorption sites in hydrophilic zeolites, drastically curtailing CO₂ uptake and increasing regeneration energy requirements [8]. Repeated heating and pressurization cycles can also induce structural and capacity degradation in zeolites over time [11]. Accordingly, this study focuses on three commercial zeolites purchased from Jalonzeolite (Yanshi, Luoyang, China) (13X, 4A, and JLPM3) to: (i) benchmark CO₂ capture from humid ambient air in a packed-bed TSA process; (ii) quantify CO₂–H₂O competitive adsorption under varying conditions; (iii) determine the regeneration conditions needed to fully desorb physisorbed CO₂ while purging accumulated moisture; (iv) assess how prolonged exposure at 100 °C, used to emulate the cumulative effect of many desorption/heating cycles, alters adsorption mechanisms and mechanical robustness; and (v) relate performance differences to textural attributes (porosity, mesopore size, and micropore volume).

To meet these objectives, we conducted two complementary experimental phases. In Phase 1, ambient air was fed through a packed bed containing 40 g of each sieve (no pre-treatment) at 2.94 L/min (3.09 NL/min) under ambient conditions. Breakthrough curves established CO₂ uptake; desorption was quantified during regeneration up to 310–315 °C; and concurrent water capture was tracked to elucidate hygroscopic behavior. In Phase 2, the materials underwent 12 days of pre-conditioning at 100 °C prior to testing, after which one TSA cycle was performed with a 200 g bed using humid ambient air feed. This deliberate, long-duration conditioning was designed to simulate the multiple desorption heating cycles encountered in real DAC/TSA operation at typical regeneration temperatures. All cycle-stabilized uptake values reported refer to these 12-day/100 °C-conditioned samples.

By linking humidity effects, regeneration temperature, and prolonged thermal exposure to pore architecture and mechanical integrity, this work provides actionable guidance to improve the energy efficiency, durability, and overall sustainability of zeolite-based DAC and to translate laboratory metrics into durable, real-world operation.

2. Experimental Procedure

2.1. Experimental Setup and Procedure

The experimental setup (Figure 1a) consists of a homemade fixed-bed reactor consisting of a quartz tube. The ends of the quartz tube are sealed using vacuum flanges; one flange is equipped with a single valve, while the other has two valves for gas flow control. The quartz tube is oriented vertically (Figure 1b), and a metallic filter is placed inside to support the molecular sieves (see Appendix A).

The quartz reactor tube had inner diameter $D_i$ = 90 mm (inner radius $r_i$ = 45 mm) and total length of 800 mm. The corresponding internal cross-sectional area was A = $\pi r_i^2$ = 6.36 × 10⁻³ m² (63.6 cm²). A perforated metallic support plate (1 mm holes) retained the pellets. The bed volume was computed as $V_b$ = A·L, the packed-bed bulk density as ${\rho }_{bulk}$ = m/$V_b$, and the void fraction was estimated as $\epsilon$ = 1 − ${\rho }_{bulk}$/${\rho }_p$ using the skeletal (particle) density ${\rho }_p$ obtained from mercury porosimetry (Table 6).

Table 1. Fixed-bed reactor and packed-bed geometry/packing parameters.

Test	Sorbent	${\mathit{D}}_{\mathit{i}}$ (mm)	A (cm2)	L (cm)	${\mathit{\rho }}_{\mathit{b}\mathit{u}\mathit{l}\mathit{k}}$ (g·cm−3)	$\mathsf{\epsilon }$	${\mathit{u}}_{\mathit{s}}$ (cm·s−1)	${\mathit{R}\mathit{e}}_{\mathit{p}}$ *	$∆\mathit{P}$ (Pa) **
40 g	JLPM3	90	63.6	0.80	0.786	0.501	0.77	0.79–1.23	0.056–0.135
40 g	13X	90	63.6	0.80	0.786	0.474	0.77	0.79–1.23	0.073–0.177
40 g	4A	90	63.6	0.70	0.898	0.492	0.77	0.79–1.23	0.053–0.129
200 g	JLMP3	90	63.6	4.00	0.786	0.501	0.77	0.79–1.23	0.279–0.674
200 g	13X	90	63.6	4.00	0.786	0.474	0.77	0.79–1.23	0.365–0.883
200 g	4A	90	63.6	3.50	0.898	0.492	0.77	0.79–1.23	0.267–0.645

* ${Re}_p$ computed using air properties at ~25 °C and pellet diameter range d_p = 1.6–2.5 mm. ** ΔP estimated by the Ergun equation (not measured), using the ε values in this table.

summarizes all geometric and packing parameters. The molecular sieves used in the experiments include 13X, 4A, and JLPM3, with particle sizes ranging from 1.6 to 2.5 mm. Humid ambient air is pumped into the reactor from the bottom using a micro-diaphragm air pump that operates continuously at a flow rate of 2.94 L/min (3.09 NL/min). Based on the tube cross-section and the actual inlet volumetric flow, the superficial gas velocity was calculated as $u_s=Q/A$ [12,13]. Bed-scale flow regime was characterized using the particle Reynolds number $R{e}_p=\rho u_s{d}_p/\mu$ (air at ~25 °C, 1 atm), yielding $R{e}_p\approx 0.8$–1.2 for $d_p=1.6$–2.5 mm (laminar regime) [12,13]. The bed pressure drop was not measured directly; therefore, $\mathsf{\Delta }P$ was estimated using the Ergun equation with the measured bed height $L$, pellet size range, and the estimated void fraction $\epsilon$ (Table 1) [12,14]. Under the present conditions, the resulting $\mathsf{\Delta }P$ is predicted to be very small (sub-Pa even for the 200 g beds) [12,14]. The air passes through the packed bed of molecular sieves, and a portion of the exiting air is directed to an online NDIR CO₂ gas analyzer (Wuhan Enviro Solutions Technology Co., Ltd., Wuhan, China). The analyzer sampled the reactor effluent via a downstream tee (slipstream) connection; the gas was not dried prior to analysis, consistent with the use of humid ambient air feed. The instrument operates over 0–980 mg/m³ and 0–19,600 mg/m³, corresponding to maximum absolute errors of ±19.6 mg/m³ and ±392 mg/m³ (±2% of full scale in each range).

During desorption, the quartz tube was placed horizontally inside the tubular furnace (Model No.STG-100-12 tube furnace, Henan Sante Furnace Technology Co., Ltd., Luoyang, China) (Figure 1c) with the packed-bed region centered in the furnace hot zone. The temperature reported in the desorption plots corresponds to the furnace control thermocouple located at the center of the hot zone (i.e., the temperature signal available from the furnace controller). The furnace was programmed to ramp from ambient to 310–315 °C at an average heating rate of ~8 °C·min⁻¹. Axial and radial temperature gradients within the packed bed were not directly measured using multiple thermocouples at the bed inlet/outlet/center; therefore, transient features should be interpreted as referenced to the furnace thermocouple temperature, and small offsets may occur during the ramp. Because the bed lengths are short (Table 1) and the bed is located within the hot zone, temperature non-uniformity across the bed is expected to be limited near steady-state conditions.

2.2. Data Analysis

During the adsorption process, graphs like the one shown in Figure 2 were obtained. In these graphs, $C_{0,{\mathit{CO}}_2}$ represents the inlet concentration of CO₂. $C_{0,{\mathit{CO}}_2}$ is determined by passing air through the reactor without the molecular sieves and averaging the measured concentrations. To calculate the total amount of CO₂ adsorbed ($Q_{{\mathit{ads},\mathit{CO}}_2}$), an overall mass balance over the fixed-bed was applied, and the uptake was obtained from the area above the breakthrough curve, i.e., the time integral of the difference between inlet and outlet concentrations [15,16]:

$$Q_{{\mathit{ads},\mathit{CO}}_2}=F{\int }_{t=0}^{t=t_{\mathrm{sat}}}\left(C_{0,{\mathit{CO}}_2}-C_{{\mathit{CO}}_2}\right)dt$$

(1)

The equation (Equation (1)) to calculate the total amount of CO₂ adsorbed includes several key components. The volumetric flow rate of air, $F$, represents the amount of air passing through the reactor and is expressed in units such as liters per minute (L/min) or cubic meters per second (m³/s). In this study, $F$ was set at 2.94 L/min (3.09 NL/min), equivalent to 0.000049 m³/s (0.0000515 Nm³/s). The inlet CO₂ concentration, $C_{0,{\mathit{CO}}_2}$, reflects the amount of CO₂ in the air entering the reactor, measured in milligrams per cubic meter (mg/m³); this was determined before each run using a bypass procedure (reactor without sorbent), and this provided the baseline for the CO₂ available for adsorption. The outlet CO₂ concentration, $C_{{\mathit{CO}}_2}$, corresponds to the CO₂ content measured in the analyzer after passing through the packed bed of molecular sieves, indicating the CO₂ that remains unadsorbed. The time at which the adsorption process reaches saturation, $t_{\mathit{sat}}$, represents the point at which the adsorbent material can no longer effectively capture CO₂, serving as the upper limit of integration.

The CO₂ concentrations $C_0$ and $C_{{\mathit{CO}}_2}$ used in Equation (1) correspond to the NDIR analyzer readings under humid (wet) conditions, because the feed was unconditioned ambient air and no drying stage was installed upstream of the analyzer. Accordingly, the uptake integration uses wet-basis concentrations consistently for both inlet and outlet.

For the graphs obtained during the desorption process, a similar integration was performed to calculate the total amount of CO₂ desorbed ($Q_{{\mathit{des},\mathit{CO}}_2}$). However, in this case, the integration was applied to the expression ($C_{{\mathit{CO}}_2}-C_{0,{\mathit{CO}}_2}$) from $t=0$ to the time when the desired temperature is reached. The outlet CO₂ concentration ($C_{{\mathit{CO}}_2}$) was adjusted by subtracting the inlet CO₂ concentration ($C_{0,{\mathit{CO}}_2}$) because during desorption, the CO₂ released from the molecular sieves due to heating is mixed with the air being pumped through the system. This results in an additional $C_{0,{\mathit{CO}}_2}$ being recorded by the analyzer, which must be accounted for to accurately quantify the CO₂ desorbed.

The inlet water-vapor concentration $\left(C_{0,H_{2}O}\right)$ was estimated from ambient temperature $\left(T\right)$ and relative humidity $\left(RH\right)$ using standard psychrometric relations. Because the adsorption tests used unconditioned ambient air and no inline humidity analyzer (or inlet hygrometer) was installed at the reactor inlet during the experimental campaign, hourly $T$ and $RH$ were obtained from the Meteostat platform for the nearest representative airport station, Tenerife/Los Rodeos [17]. Meteostat compiles and redistributes observational datasets from public meteorological sources (e.g., NOAA/WMO networks), enabling traceable post-processing of ambient test conditions.

Relative humidity is defined as $RH=100\hspace{0.17em}{p}_{H_{2}O}/p_{sat}\left(T\right)$, where $p_{H_{2}O}$ is the partial pressure of water vapor and $p_{sat}\left(T\right)$ is the saturation vapor pressure at temperature T [18]. Thus, $p_{H_{2}O}=(RH/100)\hspace{0.17em}{p}_{sat}\left(T\right)$. Assuming ideal-gas behaviour for water vapour, the inlet absolute humidity (mass concentration) is:

$$C_{0,H_{2}O}={\displaystyle \frac{\mathit{Relative} \mathit{Humidity} (\%)\times P_{\mathit{sat}}\left(T\right)\times M_{H_{2}O}}{100\times R\times T}}$$

(2)

where $M_{H_{2}O}$ is the molar mass of water (18.015 g·mol⁻¹), $R$ is the universal gas constant (8.314 Pa·m³·mol⁻¹·K⁻¹), and $T$ is in Kelvin. When $p_{sat}\left(T\right)$ is expressed in Pa, Equation (2) yields $C_{0,H_{2}O}$ in g·m⁻³ (multiply by ${10}^3$ to obtain mg·m⁻³). The saturation vapor pressure $p_{sat}\left(T\right)$ was calculated using the Buck correlation [19]. This estimated inlet humidity was used only to contextualize and compare inlet conditions between runs; no inlet–outlet H₂O mass balance is claimed in the absence of a dedicated humidity analyzer.

Uncertainties reported in

Table 2. Adsorption and desorption results for CO₂ and H₂O over 4 cycles for 40 g packed bed of molecular sieves JLPM3.

Cycles	0	1	2	3
$C_{0,{\mathit{CO}}_2}$ [mg/m3]	840.85 ± 19.60	750.81 ± 19.60	744.99 ± 19.60	736.61 ± 19.60
$C_{0,H_{2}O}$ [mg/m3]	11,662.40	5164.10	8014.29	8364.19
Weight before adsorption [g]	40.01	39.91	39.13	38.73
Weight after adsorption [g]	44.53	43.24	43.30	41.60
$Q_{{\mathit{ads},\mathit{CO}}_2}$ [mg]	214.91 ± 12.47	175.01 ± 14.08	117.48 ± 9.79	77.07 ± 7.08
$Q_{{\mathit{des},\mathit{CO}}_2}$ [mg]	312.27 ± 46.54	222.13 ± 43.15	145.47 ± 38.75	95.48 ± 37.80
$Q_{\mathit{des},H_{2}O}$ [mg]	4308 ± 46.54	3888 ± 43.15	4425 ± 38.75	2775 ± 37.80
CO2 uptake [mmol/g]	0.34 ± 0.05	0.25 ± 0.05	0.17 ± 0.04	0.11 ± 0.04
H2O uptake [mmol/g]	1.95 ± 0.02	1.79 ± 0.02	2.06 ± 0.02	1.29 ± 0.02
H2O/CO2 Selectivity	5.74 ± 0.85	7.16 ± 1.43	12.12 ± 2.85	11.73 ± 4.27

,

Table 3. Adsorption and desorption results for CO₂ and H₂O over 4 cycles for 40 g packed bed of molecular sieves 13X.

Cycles	0	1	2	3
$C_{0,{\mathit{CO}}_2}$ [mg/m3]	784.68 ± 19.60	770.93 ± 19.60	770.29 ± 19.60	832.49 ± 19.60
$C_{0,H_{2}O}$ [mg/m3]	12,878.63	6540.08	9520.27	14,366.6
Weight before adsorption [g]	40.01	37.49	37.07	36.88
Weight after adsorption [g]	43.25	39.91	39.38	39.36
$Q_{{\mathit{ads},\mathit{CO}}_2}$ [mg]	28.54 ± 3.62	120.10 ± 9.82	82.35 ± 7.35	62.64 ± 5.40
$Q_{{\mathit{des},\mathit{CO}}_2}$ [mg]	173.07 ± 47.28	144.87 ± 36.18	115.68 ± 34.06	75.45 ± 35.51
$Q_{\mathit{des},H_{2}O}$ [mg]	5587 ± 47.28	2695 ± 36.18	2384 ± 34.06	2405 ± 35.51
CO2 uptake [mmol/g]	0.20 ± 0.06	0.17 ± 0.04	0.14 ± 0.04	0.09 ± 0.04
H2O uptake [mmol/g]	2.68 ± 0.02	1.31 ± 0.02	1.17 ± 0.02	1.17 ± 0.02
H2O/CO2 Selectivity	13.40 ± 4.02	7.71 ± 1.82	8.36 ± 2.39	13.00 ± 5.78

,

Table 4. Adsorption and desorption results for CO₂ and H₂O over 4 cycles for 40 g packed bed of molecular sieves 4A.

Cycles	0	1	2	3
$C_{0,{\mathit{CO}}_2}$ [mg/m3]	735.32 ± 19.60	728.58 ± 19.60	771.57 ± 19.60	723.62 ± 19.60
$C_{0,H_{2}O}$ [mg/m3]	8991.64	5376.07	11,925.4	12,158.3
Weight before adsorption [g]	40.01	37.85	37.51	37.52
Weight after adsorption [g]	42.48	41.10	41.53	40.60
$Q_{{\mathrm{ads},\mathit{CO}}_2}$ [mg]	30.46 ± 6.03	104.06 ± 18.17	62.24 ± 10.38	34.42 ± 11.72
$Q_{{\mathrm{des},\mathit{CO}}_2}$ [mg]	240.57 ± 44.13	146.02 ± 35.86	85.02 ± 28.80	45.46 ± 24.47
$Q_{\mathit{des},H_{2}O}$ [mg]	4389 ± 44.13	3444 ± 35.86	3925 ± 28.80	3035 ± 24.47
CO2 uptake [mmol/g]	0.28 ± 0.05	0.17 ± 0.04	0.10 ± 0.03	0.05 ± 0.03
H2O uptake [mmol/g]	2.07 ± 0.02	1.65 ± 0.02	1.88 ± 0.01	1.46 ± 0.01
H2O/CO2 Selectivity	7.39 ± 1.32	9.71 ± 2.29	18.80 ± 5.64	29.20 ± 17.52

and

Table 5. Adsorption and desorption results for CO₂ and H₂O over one TSA cycle with 12 days thermal pre-treatment at 100 °C for molecular sieves JPM3, 13X, and 4A.

Molecular Sieve	JLPM3	13X	4A
Bed length ($l_b$) [cm]	4.0	4.0	3.5
$C_{0,{\mathit{CO}}_2}$ [mg/m3]	718.27 ± 19.60	745.33 ± 19.60	732.91 ± 19.60
$C_{0,H_{2}O}$ [mg/m3]	8726.44	3382.61	9072.79
Weight Before Adsorption [g]	200.01	200.01	200.01
Weight After Adsorption [g]	214.22	207.39	208.83
Weight After Desorption [g]	187.14	184.6	188.44
$Q_{{\mathit{ads},\mathit{CO}}_2}$ [mg]	408.40 ± 24.95	128.95 ± 8.89	22.57 ± 19.62
$Q_{{\mathit{des},\mathit{CO}}_2}$ [mg]	1009.93 ± 45.35	857.48 ± 43.45	588.55 ± 37.76
$Q_{\mathit{des},H_{2}O}$ [mg]	26,070 ± 43.35	21,933 ± 43.45	19,802 ± 37.76
CO2 uptake [mmol/g]	0.24 ± 0.01	0.20 ± 0.01	0.14 ± 0.01
H2O uptake [mmol/g]	2.51 ± 0.00	2.14 ± 0.00	1.89 ± 0.00
H2O/CO2 Selectivity	10.46 ± 0.44	10.70 ± 0.53	13.50 ± 0.96
$T_1$ [°C]	103.1	103.7	92.3
$T_2$ [°C]	152.6	136.9	126.6
$T_3$ [°C]	275.3	273.2	236.6

correspond to combined measurement uncertainty (propagated from instrument specifications and data reduction) and are not standard deviations from independent replicate experiments. CO₂ uptake uncertainty was obtained by propagation of the NDIR analyzer accuracy (±2% of full scale for the selected range) together with numerical integration discretization error. Because cycles 0–3 were performed sequentially on the same packed bed (no repacking), cycle-to-cycle variation reflects cyclic conditioning/degradation rather than replicate-to-replicate variability.

Baseline correction was not based on a fixed “laboratory baseline” across days. Instead, each run was treated under its own steady conditions and referenced to $C_0$, defined as the inlet CO₂ concentration measured immediately before that run using the bypass procedure (reactor without sorbent). During data processing, the concentration time series was screened for isolated spikes/outliers (e.g., occasional analyzer glitches or switching transients); points that deviated strongly from the local neighborhood were excluded so they did not bias the baseline estimate or the subsequent numerical integration.

2.3. Characterization Techniques

2.3.1. Nitrogen Adsorption

The samples were degassed at 200 °C for 16 h under vacuum to remove moisture and other physisorbed species, ensuring that subsequent N₂ uptake reflects the accessible pore volume. Adsorption–desorption isotherms were then measured on a (Micromeritics Instrument Corporation, Norcross, GA, USA) with the MicroActive 5.02 software. Liquid nitrogen cooling maintained a bath temperature of 77.3 K, and high-purity N₂ gas was used as the adsorptive. Data points were collected over a relative pressure range of p/p₀ ≈ 0.001–0.987 with a fixed 20 s equilibration interval at each step.

The instrument software generated, for each sample: single-point BET surface area, multi-point BET surface area, t-plot micropore area and volume, BJH adsorption/desorption pore-size distribution, and Horváth–Kawazoe micropore width. These textural parameters were used to correlate structural properties with adsorption performance.

2.3.2. Mercury Porosimetry

Mercury intrusion–extrusion measurements were performed with a high-pressure porosimeter, Micromeritics AutoPore IV (Micromeritics Instrument Corporation, Norcross, GA, USA), controlled by the MicroActive software. Powder samples were loaded into a 3-bulb, 0.412-stem glass penetrometer (volume = 3.1276 mL). Intrusion was recorded from vacuum up to ~60,000 psia. Calculations assumed a mercury contact angle of 141°. The resulting mercury intrusion data provided macropore size distributions and bulk density measurements, supplementing the nitrogen physisorption data to provide a full picture of the pore architecture in each material.

2.3.3. X-Ray Diffraction (XRD)

Powder specimens (10 mm track length) were examined on a θ–2θ diffractometer operated in Bragg–Brentano geometry. Cu K-α radiation was used (λ = 1.54060 Å; generator setting 45 kV/40 mA). Key instrument parameters were: divergence slit = 0.1799°, incident-beam Soller slit = 0.040°, 10 mm beam mask, goniometer radius = 240 mm and specimen spinning “Yes”. Patterns were collected continuously from 5.013 to 79.97° 2θ with a step of 0.026° 2θ and 56.865 s counting time per step. Peak lists produced by the control software were matched against ICDD PDF-4+ entries, providing semi-quantitative phase abundances by the reference intensity ratio (RIR) routine included in the same package.

2.3.4. X-Ray Fluorescence (XRF)

Elemental concentrations were obtained from wavelength-dispersive XRF measurements. The instrument acquired two excitation ranges automatically: Range 3 at 20 kV and Range 2 at 40 kV (ranges shown in the report for every oxide). The three samples return analytical totals of 99.99–100.02 wt %, confirming measurement completeness within the stated precision.

3. CO₂/H₂O Adsorption–Desorption Performance

3.1. The 40 g Packed Bed of Molecular Sieves JLPM3, 13X, and 4A (No Pre-Treatment)

Four TSA cycles were carried out on separate days in a fixed-bed reactor packed with 40 g of JLPM3, 13X, and 4A molecular sieves. The first adsorption cycle (cycle 0) served as a pre-treatment to activate the zeolites. Adsorption was performed under ambient conditions, while desorption was induced by a temperature swing from ambient ($T_{amb}$) up to 310–315 °C. The packed beds were prepared in a tube of inner diameter $D_i$ = 90 mm (inner radius $r_i$ = 45 mm). For the 40 g tests, the resulting packed-bed heights were L = 0.80 cm (JLPM3), 0.80 cm (13X), and 0.70 cm (4A) (see Table 1 for full geometric and packing parameters).

Four TSA cycles were performed on separate days with 40 g beds of JLPM3, 13X, and 4A using humid ambient air as feed (Section 2.1). Adsorption proceeded at ambient conditions; desorption used a temperature swing to 310–315 °C. The corresponding breakthrough and temperature-programmed desorption (TPD) profiles are shown in Figure 3, Figure 4 and Figure 5 and cycle-wise uptakes in Table 2, Table 3 and Table 4.

Typically, zeolites are considered physical adsorbents, but formation of chemisorbed species (like carbonates or bicarbonates) on zeolite cation sites is known under certain conditions [7]. For example, Amrit Kumar et al. (2015) observed that some physisorbents can bind CO₂ strongly enough to require elevated temperatures for release, hinting at chemisorptive interactions in MOFs/zeolites during DAC [20]. In all three sieves, the CO₂ TPD curves show a reproducible bi-modal shape: a low-temperature peak (≈75–82 °C) followed by a broader peak at ≈214–220 °C (Figure 3b, Figure 4b and Figure 5b). We attribute the first to readily reversible physisorption in micropores and the second to stronger binding (e.g., bicarbonate/carbonate-like species on cationic/defect sites) that requires higher temperature to desorb, consistent with reports on FAU/LTA zeolites that separate weakly vs. strongly bound CO₂ in TPD features [6,7]. A fraction of stronger sites can also be introduced or amplified by composition/defects or metal modification (e.g., Fe@13X), which shifts part of the desorption to higher T [21].

As heat is applied from ambient up to 310–315 °C, a first peak appears between 75 °C and 82 °C (77.9 ± 1.2 °C for JLPM3; 75.3 ± 2.8 °C for 13X; 81.9 ± 0.6 °C for 4A), signaling the release of weakly adsorbed CO₂ (physisorption or surface interactions). Following this peak, the CO₂ concentration falls sharply (flush-out of residual gas) until roughly 156–168 °C. Then, a second, broader peak emerges at 214–220 °C (214.1 ± 7.3 °C for JLPM3; 216.6 ± 4.5 °C for 13X; 219.6 ± 5.1 °C for 4A), attributed to the desorption of more strongly bound CO₂ (chemisorption or internal zeolite interactions). A small, marginal rise near 300 °C is observed in all cases, likely due to combustion of trace organics rather than continued desorption.

Integrating the breakthrough curves, the initial CO₂ uptakes rank JLPM3 > 13X > 4A (Table 2, Table 3 and Table 4). JLPM3 shows the steepest front, indicative of faster kinetics, consistent with its higher porosity and smaller mesopores that shorten diffusion paths to CO₂-selective micropores (Section 4). Recent studies confirm that appropriately engineered zeolites can capture CO₂ at air-level concentrations with fast kinetics and modest heats of adsorption, particularly MOR-type frameworks under dry or controlled-humidity conditions [16]. These results align with our ranking (JLPM3 ≥ 13X ≫ 4A) and the observed sensitivity to moisture. All three materials, however, display substantial co-adsorption of H₂O; the measured H₂O/CO₂ selectivity increases across cycles (Figure 3, Figure 4 and Figure 5), reflecting the strong hydrophilicity of low-Si/Al zeolites and site blocking by water under ambient humidity [6,22]. The particularly low CO₂ capacity of 4A is consistent with its small LTA windows that limit CO₂ access and its strong affinity for H₂O [6,22]. Notably, while competitive adsorption by water is generally dominant in hydrophilic zeolites, specific K-MER sites have been shown to allow CO₂ to displace pre-adsorbed H₂O at 30 °C, highlighting framework/cation site effects under humid DAC conditions [23].

Quantitatively, the fraction of CO₂ released via the first mechanism (“low temperature” peak) varies by sorbent and cycle. For JLPM3, it remains around 80% across cycles (with no statistically significant change), whereas 13X rises from 43% in cycle 0 (preloaded and burning off organics) to about 85% thereafter, and 4A from 55% up to ~80–82%. Weight measurements (Table 2, Table 3 and Table 4) confirm a large drop after the first cycle (40.01 g → 37.49 g for 13X; 40.01 g → 37.85 g for 4A), then only minor losses in subsequent cycles, corroborating removal of pre-existing CO₂/organics and stabilization of the material. Across all sorbents, after pre-treatment (after cycle 0), adsorption/desorption capacities ($Q_{{\mathit{ads},\mathit{CO}}_2}$ and $Q_{{\mathit{des},\mathit{CO}}_2})$ match within margin of error each cycle, aside from cycle 0, but this is because, during the desorption in cycle 0, there was still CO₂ adsorbed in the zeolites and also part of the excess of CO₂ comes from organic matter being burnt, demonstrating full reversibility once pre-treatment effects are accounted for.

Except for the first cycle, the integrated CO₂ desorbed ($Q_{{\mathit{des},\mathit{CO}}_2}$) matches the CO₂ adsorbed ($Q_{{\mathit{ads},\mathit{CO}}_2}$) within the combined experimental uncertainty of the flow, baseline subtraction, and analyzer calibration. The apparent $Q_{{\mathit{des},\mathit{CO}}_2}$ > $Q_{{\mathit{ads},\mathit{CO}}_2}$ observed in cycle 0 arises from (i) pre-loaded CO₂ in the as-received zeolites (CO₂ weakly bound and/or bicarbonate/carbonate species accumulated) that is released upon the first heat-up and (ii) oxidation of trace organics on the pellets/binder, evidenced by the small shoulder near ~300 °C, which contributes additional CO₂ not associated with the immediately preceding adsorption step. After this conditioning cycle, baseline-corrected desorption profiles close the CO₂ mass balance in every cycle, confirming reversible uptake once pre-treatment effects are removed.

Despite this reversibility, total CO₂ uptake and release steadily decline with each TSA cycle (Figure 3a,c, Figure 4a,c and Figure 5a,c), indicative of thermal degradation at 310–315 °C. This is consistent with the thermally induced degradation of hydrophilic zeolites under repeated TSA exposure in humid feeds [11]. As high-quality CO₂ sites/accessibility decline, the H₂O/CO₂ selectivity rises because polar H₂O increasingly dominates competitive adsorption [6,22]. JLPM3 shows the greatest performance drop and physical fragmentation, producing dust, whereas 13X retains its structure best, and 4A lies in between. By cycle 3, JLPM3 and 13X converge to similar uptake levels, underscoring how JLPM3’s structural breakdown drives its loss. When comparing all three sorbents (Figure 6), JLPM3 initially delivers the highest CO₂ uptake, followed by 13X and then 4A, but all three diminish over cycles. Water uptake likewise converges, and the H₂O/CO₂ selectivity ratio increases cycle-by-cycle: as the sorbents’ electromagnetic adsorption strength weakens (affecting non-polar CO₂ more than polar H₂O), given that they preferentially capture water. These trends highlight that, while the dual-peak desorption signature and reversible adsorption mechanisms persist, high-temperature cycling inexorably erodes sorbent capacity and selectivity, especially for JLPM3, suggesting a trade-off between initial performance and thermal durability.

The humidity penalty observed in this paper and in the literature [10,22,23,24] underscores the need for water management (e.g., layered beds or pre-drying) to enable moderate-temperature TSA (≈100–120 °C) with zeolites [22]. Framework topology and extra-framework cations critically tune DAC performance at 400 ppm, especially in LTA/4A where cation charge/size governs low ppm uptake and moisture sensitivity [10]. These comparisons support our mechanistic interpretation and guide process/material strategies to mitigate water competition and thermal stress in DAC cycles.

3.2. Comparison of Molecular Sieves JLPM3, 13X, and 4A (No Pre-Treatment)

Figure 6 consolidates the cycle-wise CO₂ uptake, H₂O uptake, and H₂O/CO₂ selectivity for the three sieves (see Table 2, Table 3 and Table 4 for values and uncertainties). Trends are discussed in terms of framework topology, extra-framework cations, and competitive adsorption with water, which dominate DAC performance at ~400 ppm CO₂.

CO₂ uptake (Figure 6a): In the first cycle, the ranking is JLPM3 > 13X > 4A. JLPM3’s advantage is consistent with its higher accessible porosity and shorter diffusion paths to CO₂-selective micropores (Section 4), which sharpen the breakthrough front and improve bed utilization. 13X (FAU) provides large pore windows (~0.74 nm) and a high density of Na⁺ sites that polarize CO₂ effectively, explaining its robust, though lower, uptake. By contrast, 4A (LTA) has ~0.40–0.42 nm apertures that restrict CO₂ access and are readily occupied by H₂O under ambient humidity, depressing its effective capacity [6,11,16]. These observations align with DAC-specific studies showing zeolites can capture air-level CO₂ with fast kinetics under dry/controlled-humidity feeds (e.g., MOR- and FAU-type) and that framework/cation pairing strongly governs low ppm performance [6,24,25].

H₂O uptake (Figure 6b): All three sieves co-adsorb substantial water, with 4A and 13X exhibiting the strongest moisture loading, consistent with their low Si/Al and hydrophilicity [6,22]. Preferential H₂O occupation of cationic sites reduces the number of electrostatically favorable CO₂ sites and lengthens mass-transfer paths, explaining earlier CO₂ breakthrough and lower integrated uptake relative to dry conditions [6,22]. While water typically “wins” on hydrophilic zeolites, site-specific exceptions exist (e.g., K-MER), where CO₂ can displace pre-adsorbed H₂O at 30 °C; this underscores that cation identity and local ring geometry can flip CO₂/H₂O selectivity under certain conditions [23].

H₂O/CO₂ (Figure 6c) increases with cycle number for all sieves. Two effects compound: (i) progressive loss of high-quality CO₂ sites and/or pore accessibility due to thermal/mechanical stress at 310–315 °C in humid feeds, well documented for 4A/13X-class materials [11], and (ii) H₂O “hogging” of the remaining cationic sites as the CO₂-affine population shrinks [6,22]. The net result is a rising water fraction in the working capacity even as both absolute uptakes decline.

The monotonic decrease in CO₂ capacity across cycles, the visible fragmentation in JLPM3, and the relative robustness of 13X are consistent with the literature on the thermo-humid cyclic degradation of hydrophilic zeolites [11]. By the last cycle, JLPM3 and 13X converge to similar CO₂ uptakes, implicating mechanical attrition and/or partial amorphization in JLPM3 as the cause of its steeper decline (see Section 4). 4A remains transport-limited and water-poisoned throughout, consistent with LTA access constraints and strong cation–H₂O interactions [6,22,24]. The pronounced cation sensitivity of LTA-type zeolites at 400 ppm reported elsewhere further explains 4A’s low working capacity under humid air [10].

Mechanistic link to desorption profiles. The bi-modal CO₂-TPD response discussed in Section 3.1, low T, readily reversible physisorption followed by a higher-T contribution from stronger binding on cationic/defect sites, rationalizes the selective loss of CO₂ capacity under harsh cycling. Introducing or exposing a stronger-site fraction (e.g., via composition/defects or metal incorporation) shifts part of the desorption to higher temperature, as seen for Fe-modified FAU (Fe@13X) [21].

These results point to two levers for DAC with zeolites: (1) Water management, e.g., upstream drying or layered beds with a desiccant pre-layer [22]. (2) Framework/cation tuning, especially in LTA/FAU, controls the CO₂/H₂O balance at 400 ppm; cations with appropriate charge density and placement can improve low ppm CO₂ uptake yet should avoid excessive water affinity [6,10,24]. These principles are consistent with the stronger initial performance of FAU-type materials in this paper and with recent demonstrations of low ppm CO₂ capture on optimized zeolites [21,24].

3.3. Effect of Prolonged Thermal Conditioning (12 Days at 100 °C) on CO₂ Uptake and Mechanism

A 12-day conditioning at 100 °C was applied to 200 g beds of JLPM3, 13X, and 4A to emulate cumulative desorption exposure and obtain cycle-stabilized behavior. After this treatment, one TSA cycle was performed under humid ambient air. The results (Figure 7a–f; Table 5) show (i) similar total CO₂ uptake for JLPM3 and 13X relative to their first cycles in Section 3.1, (ii) a redistribution of CO₂ between weak and strong binding (a larger high T desorption contribution), and (iii) persistent H₂O co-adsorption that continues to limit working capacity under humidity.

From Figure 7b,d,f, the desorption profiles of the three molecular sieves as a function of temperature exhibit a distinct pattern: an initial peak ($T_1$), followed by a relative minimum ($T_2$), and finally, another peak ($T_3$). These temperature points, detailed in Table 5, correspond to different desorption mechanisms. At $T_1$, the zeolites release all CO₂ stored via physisorption. The intermediate point $T_2$ marks the onset of desorption for CO₂ bound by stronger interactions, while $T_3$ indicates the complete release of all stored CO₂.

In the pre-treated 200 g beds, the CO₂-TPD traces display a low-temperature release centered near 95–105 °C (JLPM3, 103.1 °C; 13X, 103.7 °C; 4A, 92.3 °C), followed by an intermediate minimum (~137–153 °C) and a pronounced high-temperature tail/peak extending to ~235–275 °C (Table 5; Figure 7b,d,f). We attribute the first feature to readily reversible physisorption in micropores and the higher T release to stronger binding (bicarbonate/carbonate-like species on cationic/defect sites) that requires additional thermal input, consistent with established assignments for FAU/LTA zeolites and other physisorbents showing dual-mode CO₂ retention [6,7,20]. The growth of the high T fraction after conditioning indicates that mild, extended heating can activate or expose stronger sites (e.g., by cation re-distribution or defect formation) [6,7], a trend aligned with observations that metal-modified FAU (e.g., Fe@13X) holds a portion of CO₂ more strongly and desorbs at higher T [21].

The total CO₂ and H₂O uptakes in Table 5 were calculated using the post-desorption weights of the molecular sieves, representing the fully regenerated samples. Table 5 shows that the weight of the molecular sieves after desorption is significantly lower than before adsorption, confirming the presence of CO₂ and H₂O stored in the materials even after 12 days at 100 °C. Post-conditioning CO₂ uptakes are 0.24 ± 0.01 mmol g⁻¹ (JLPM3), 0.20 ± 0.01 mmol g⁻¹ (13X), and 0.14 ± 0.01 mmol g⁻¹ (4A) (Table 5), preserving the rank JLPM3 > 13X > 4A seen in the 40 g tests. This hierarchy matches expectations from framework access and micropore volume (FAU ≫ LTA) and the kinetic advantage of hierarchical/mesoporous architectures [6,24,25]. Nevertheless, H₂O uptake remains substantial (2.51, 2.14, and 1.89 mmol g⁻¹ for JLPM3, 13X, and 4A), keeping H₂O/CO₂ selectivity ≈ 10–14 (Table 5). This confirms that, under humid feeds, water preferentially occupies cationic sites, depressing CO₂ working capacity unless moisture is managed [6,8,22]. While water usually predominates on hydrophilic, low-Si/Al zeolites, site-specific motifs (e.g., K-MER) can allow CO₂ to displace pre-adsorbed H₂O at 30 °C, underscoring the role of cation identity and ring geometry in tuning competition [23].

Figure 7a,c,e provide insights into the adsorption behavior after extended pre-conditioning. For JLPM3 (Figure 7a), the CO₂ concentration surpasses the inlet concentration ($C_{0,{\mathit{CO}}_2}$) before stabilizing at a higher level. This phenomenon may result from the heat released during adsorption, which partially desorbs CO₂, and from competition between H₂O and CO₂ for adsorption sites due to JLPM3’s strong electromagnetic interactions. For 13X (Figure 7c), the CO₂ concentration initially exceeds $C_{0,{\mathit{CO}}_2}$ before oscillating around this value, suggesting lower heat being emitted during adsorption and fast cooling of the sorbent, which yields a rapid equilibration. In contrast, the behavior of 4A (Figure 7e) reflects a significant loss of adsorption capacity following thermal pre-treatment, with a markedly anomalous adsorption curve compared to JLPM3, 13X, and the results from Section 3.1 for the 40 g packed bed of molecular sieve 4A.

Prolonged exposure at 100 °C under humid air likely dehydrates residual hydroxyls, drives limited cation migration, and stabilizes defect-adjacent sites, modestly increasing the population of stronger CO₂-binding sites while slightly reducing purely physisorptive capacity [6,7,20]. This dual effect explains the larger high T desorption contribution in Figure 7, and the observation that JLPM3 and 13X have similar capacities relative to their unconditioned cycle 1 (Section 3.1), whereas 4A remains transport-limited by LTA apertures under humid air [6,10,22,24]. The trend is consistent with the literature where framework/cation tuning at DAC concentrations adjusts the balance between weak and strong binding, particularly for LTA/FAU families [6,10].

4. Textural Properties and Characterization

4.1. Nitrogen Adsorption

Table 6. BET/BJH data for molecular sieves JLPM3, 13X, and 4A.

Molecular Sieve	BET Surface Area (m2 g−1)	Micropore Area (m2 g−1)	External Surface Area (m2 g−1)	Total Pore Volume (cm3 g−1)	Micropore Volume (cm3 g−1)	Median Pore Width (nm)	BJH Average Mesopore Width (nm)
JLPM3	727.901	710.285	17.616	0.301 (at p/p0 = 0.977)	0.267	0.758	7.569 (des.)
13X	710.088	672.618	37.470	0.359 (at p/p0 = 0.975)	0.253	0.765	13.353 (des.)
4A	24.087	2.809	21.278	0.086 (at p/p0 = 0.976)	0.001	41.560	15.699 (des.)

summarizes BET areas and pore volumes derived from N₂ sorption at 77 K. As measured, 4A shows negligible measured micropore area/volume, while JLPM3 and 13X exhibit large micropore areas (≈710 and ≈673 m² g⁻¹) and micropore volumes (≈0.267 and ≈0.253 cm³ g⁻¹), respectively. The disparity primarily reflects aperture accessibility at 77 K: LTA (4A) eight-ring windows (~0.40–0.42 nm) impose diffusion limitations for N₂ at 77 K; so, conventional N₂ sorption “sees” little internal surface, whereas FAU (13X/JLPM3) 12-ring windows (~0.74 nm) are accessible, yielding the much larger measured microporous surface [6]. Consequently, measured N₂ microporosity correlates with the higher DAC CO₂ uptake observed for FAU vs. LTA in Section 3.1, Section 3.2 and Section 3.3. This trend is consistent with zeolite structure, property relationships discussed in [6], and with the hierarchical transport benefits reported for small-mesopore chabazite analogues [25].

The HK micropore width for JLPM3 and 13X (~0.75 nm) indicates channels readily accessible to CO₂, while the BJH mesopore peaks (JLPM3 ≈ 7.6 nm; 13X ≈ 13.3 nm) suggest shorter diffusion paths in JLPM3 than in 13X, which helps explain JLPM3’s sharper breakthrough fronts and slightly higher working capacities under humid air (Section 3.1 and Section 3.2) [6,24,25]. In contrast, 4A’s low measured microporosity at 77 K aligns with its low working capacity at ~400 ppm in humid feeds due to combined access limitations and water competition [6,10,22].

The N₂ adsorption–desorption isotherms at 77.3 K (Figure 8) allow classification of the porous texture and the origin of the BET/BJH parameters following IUPAC recommendations [26]. (a) JLPM3 exhibits a very steep uptake at low relative pressure $(p/p^0\lesssim 0.05)$, which is characteristic of micropore filling and corresponds to an IUPAC Type I behavior typical of zeolitic adsorbents [26]. At intermediate pressures, the uptake increases only slightly, indicating limited additional adsorption on external surfaces. A small hysteresis loop at high $p/p^0$ (near $p/p^0\gtrsim 0.8$) suggests the presence of textural mesoporosity (e.g., intercrystalline voids or packing/binder-related mesopores), commonly described as H4-like for micro–mesoporous solids [26]. Consequently, BET surface area provides a consistent comparative metric [27], while BJH analysis should be interpreted as describing textural mesopores rather than intrinsic zeolite micropores [26,28]. (b) 13X also shows a sharp low-pressure uptake $(p/p^0\lesssim 0.05)$, again consistent with dominant microporosity (Type I) [26]. Compared with JLPM3, the higher adsorbed amount across the pressure range indicates a larger accessible pore volume and/or a greater contribution from external/textural porosity in the commercial form. The more evident high-pressure hysteresis towards $p/p^0\to 1$ indicates additional adsorption associated with textural mesopores/interparticle voids, compatible with an H4-like loop in a micro–mesoporous solid [26]. Accordingly, BET remains appropriate for surface-area comparison [27], and BJH-derived pore size distributions should be considered primarily within the mesopore range (2–50 nm) [26,28]. (c) 4A differs markedly, showing much smaller apparent uptake at very low $p/p^0$ and a strong increase only at high relative pressure, together with a pronounced hysteresis loop. This shape indicates that, in the measured sample form, adsorption is strongly influenced by textural porosity (interparticle/binder voids) that undergoes capillary condensation/evaporation at high $p/p^0$, often associated with H3-like hysteresis for aggregates/slit-like interparticle voids [26]. In addition, for narrow-pore zeolites such as LTA/4A, N₂ at 77 K can be kinetically restricted in accessing the smallest micropores, leading to an underestimation of micropore filling; IUPAC notes that alternative probe molecules/conditions (e.g., CO₂ at higher temperature) may be required when diffusion limitations occur [26,29]. Therefore, BET/BJH results for 4A should be interpreted cautiously: BJH mainly reflects textural mesoporosity [26,28], while the intrinsic micropore volume may not be fully captured by N₂ at 77 K under diffusion-limited conditions [26,29].

4.2. Mercury Porosimetry

Table 7. Mercury intrusion porosimetry data for molecular sieves JLPM3, 13X, and 4A.

Molecular Sieve	Total Intrusion Volume at 59,862.34 Psia (mL g−1)	Total Pore Area at 59,862.34 Psia (m2 g−1)	Bulk Density at 1.02 Psia (g mL−1)	Apparent (Skeletal) Density at 59,862.34 Psia (g mL−1)	Porosity (%)
JLPM3	0.225	4.875	1.162	1.575	26.190
13X	0.229	16.305	1.114	1.495	25.491
4A	0.176	14.703	1.349	1.768	23.721

shows all three samples have similar bulk porosity (≈24–26 vol %), but distinct meso/macropore architectures. JLPM3 exhibits a narrower BJH mesopore distribution (~7–8 nm) than 13X and 4A (≈13–16 nm), which is consistent with shorter intraparticle paths to CO₂-selective micropores. Numerous studies on hierarchical zeolites (e.g., SSZ-13 with 5–8 nm mesopores) report accelerated uptake kinetics relative to purely microporous analogues, because a mesopore network reduces diffusion lengths into the micropore network [25]. These textural features rationalize JLPM3’s faster kinetics and higher initial utilization vs. 13X/4A under humid ambient air, while also explaining why larger mesopores alone (13X, 4A) do not compensate for insufficient accessible micropore volume (notably in 4A) [6,25] (Figure 9).

4.3. X-Ray Diffraction (XRD)

Figure 10 confirms the expected FAU topology for 13X and JLPM3 and LTA topology for 4A. Because reference intensity ratio (RIR) matches are semi-quantitative, the phase percentages should be interpreted cautiously; minor reflections may also include binder/amorphous components and carbonate/hydroxide species introduced during handling. The key point for DAC is structural: FAU frameworks (JLPM3, 13X) provide 12-ring access to large supercages, whereas LTA (4A) provides 8-ring access to α-cages, driving the accessibility differences discussed above [6]. The trace Fe indicated by XRF (Section 4.4) is consistent with small iron-bearing domains or extra-framework cations that can slightly modify local electrostatics, a factor known to influence low ppm CO₂ binding on FAU-type materials [6].

4.4. X-Ray Fluorescence (XRF)

Table 8. Elemental composition of molecular sieve JLPM3, 13X, and 4A.

Composition (wt.%)	JLPM3	13X	4A
Na2O	17.69	16.63	15.26
MgO	0.14	2.68	2.43
SiO2	45.99	48.34	45.71
Al2O3	33.82	29.33	32.62
K2O	0.21	0.24	0.35
Fe2O3	0.60	1.13	1.44
CaO	0.97	0.83	1.15
Other *	1.58	0.82	1.04
Si/Al **	1.15	1.40	1.19

* “Other” includes P₂O₅, SO₃, Cl, TiO₂, MnO, SrO, ZrO₂ and (where present) V₂O₅, ZnO, Rb₂O; each individual component is ≤0.37 wt.%. ** Si/Al = Silicon/Aluminum.

shows all three are low-silica Na-zeolites (Si/Al ≈ 1.15–1.40), implying a high density of extra-framework cations that enhance adsorption of polarizable CO₂ via electrostatic/quadrupolar interactions [6]. The Si/Al trend helps explain strong CO₂ affinity in JLPM3 and 13X provided access is not rate-limiting, while 4A, despite similar Si/Al, remains transport-limited by its narrow windows under humid DAC conditions [6,10]. Minor levels of divalent/other cations (e.g., Mg²⁺) may adjust local fields or create defect-adjacent sites, but the dominant effects observed in this paper are framework access (FAU vs. LTA) and water competition [6,10,22] (Section 3.1, Section 3.2 and Section 3.3).

5. Conclusions

In conclusion, this comparative study has demonstrated distinct performance trade-offs among zeolite sorbents for direct air capture under humid TSA conditions. JLPM3 exhibited the highest cycle-stabilized CO₂ uptake (~0.24 mmol·g⁻¹) owing to a synergistic physisorption–chemisorption capacity and a more accessible pore architecture. However, JLPM3’s initial advantage was tempered by a pronounced capacity decline over repeated cycles due to mechanical and thermal strain, leading its uptake to converge toward that of the more robust 13X by the final cycles. Zeolite 13X achieved a slightly lower CO₂ capacity (~0.20 mmol·g⁻¹) but maintained superior structural integrity, showing minimal degradation across cycles. In contrast, 4A consistently displayed the smallest CO₂ uptake (~0.14 mmol·g⁻¹) because its narrow LTA-type pore windows and high intrinsic affinity for H₂O severely limited CO₂ access. All three sorbents showed substantial co-adsorption of water, which competed with CO₂ for adsorption sites and exacerbated performance differences.

Mechanistically, the sorbents were found to capture CO₂ via dual binding modes. We observed a readily reversible physisorbed component and a strongly bound chemisorbed component in all materials. The presence of a high-temperature desorption tail (extending up to ~250 °C in TPD profiles) confirms that a fraction of CO₂ forms stronger bonds (likely carbonate or bicarbonate species on cationic or defect sites) requiring higher thermal input to desorb. Mild thermal preconditioning (100 °C for 12 days) accentuated this effect by activating additional strong binding sites (increasing the chemisorbed fraction) at the expense of some physisorptive capacity. Notably, a regeneration temperature of ~100 °C was sufficient to fully desorb the physisorbed CO₂ and to purge accumulated water from the zeolites. Notably, the dominant low-temperature desorption feature (centered near ~75–105 °C) indicates that most of the reversible, physisorbed CO₂ can be released at moderate temperatures and that co-adsorbed moisture can be substantially purged under these conditions. However, the persistence of a high-temperature tail/peak extending to ~235–275 °C shows that a non-negligible fraction of CO₂ is more strongly bound and would require higher temperatures (or periodic high T regeneration) for complete removal. Humidity proved to be a critical factor: water vapor was shown to strongly compete for cationic adsorption sites, driving H₂O/CO₂ selectivity to ~10–14 in these low-Si/Al zeolites. Consequently, under humid ambient air, a significant portion of each sieve’s capacity was occupied by H₂O, directly limiting the available sites for CO₂ uptake unless moisture is rigorously managed.

From a practical standpoint, our findings underscore the importance of coupling water management strategies with gentle regeneration conditions to sustain DAC performance. Co-adsorbed water must be minimized, for example, by incorporating upstream dehumidification or layered bed configurations with a dedicated desiccant pre-layer—to prevent site blockage and preserve CO₂ capacity. Implementing such water management measures enables effective CO₂ capture at moderate desorption temperatures (~100–120 °C), instead of having to heat adsorbent beds to ~310 °C. Indeed, maintaining milder TSA conditions was sufficient in this work to restore the sieves’ working capacity on each cycle while avoiding the structural degradation observed at higher temperatures. This gentle regeneration not only reduces energy requirements but also helps preserve sorbent integrity over extended operation.

For future development, sorbent tuning and hybrid bed configurations emerge as promising pathways to enhance performance. Tailoring the zeolite framework and cation composition can modulate the balance between CO₂ affinity and H₂O selectivity. For instance, adjusting Si/Al ratios or exchanging extra-framework cations may reduce excessive water binding while maintaining strong CO₂ adsorption at 400 ppm. Such compositional modifications, alongside creating hierarchical porosity, could produce materials that combine the high initial capacity of JLPM3 with the long-term stability of 13X. Equally important is improving the mechanical robustness of candidate sorbents: the partial attrition observed for JLPM3 indicates that strengthening particle morphology or binder formulation will be critical for durable cyclic use. Additionally, hybrid adsorption system designs (e.g., layered beds or sorbent blends) should be explored to leverage the complementary strengths of different materials—for example, using a moisture-tolerant layer to safeguard a CO₂-selective layer. Overall, the insights gained in this study provide a foundation for optimizing zeolitic sorbents and process conditions. By managing water uptake and limiting thermal stress, and by tuning sorbent properties to favor dual-mode CO₂ binding without incurring fragility, next-generation DAC systems can achieve more efficient and stable CO₂ capture from humid air in practical applications.