Gas-phase Transport Limitations and Activated-Carbon Filtration in Enclosed FDM Printing Chambers

Abstract

Desktop fused deposition modeling (FDM) 3D printers are increasingly used in homes, schools, and laboratories, but high-temperature materials such as ABS can release volatile organic compounds (VOCs) that accumulate inside enclosed spaces. Activated-carbon filters are commonly used to reduce these emissions, yet many filtration designs focus primarily on carbon quantity while giving less attention to how contaminated air moves through the chamber and reaches the carbon bed. This study investigates whether gas-phase transport is an important limiting factor in VOC mitigation during enclosed extrusion.

The objective of this study was to determine whether active chamber recirculation reduces VOC sensor response at matched activated-carbon mass. ABS filament was extruded in a sealed acrylic chamber under conditions with varying carbon masses and circulation activity. VOC levels were measured using two SGP41 sensors placed at different chamber locations, with additional gas resistance and temperature recordings as secondary measurements.

The findings show that active recirculation produced substantial VOC reductions even when carbon mass was held constant. At both 30 g and 60 g carbon, active recirculation lowered endpoint response, cumulative response, and differences between sensor locations. In particular, 30 g carbon with active recirculation performed similarly to or better than 60 g carbon under passive circulation, suggesting that airflow and chamber mixing can be as important as sorbent quantity under these conditions. Overall, this study supports the interpretation that enclosed-printer VOC filtration is not only capacity-limited, but also transport-limited, and that airflow structure should be considered in future emission-control designs.

Keywords: fused deposition modeling; activated carbon; volatile organic compounds; adsorption; gas-phase transport; recirculation; enclosure filtration

Introduction

Desktop fused deposition modeling has made polymer additive manufacturing accessible in classrooms, homes, small laboratories, and hobbyist environments. This access has also increased the number of printers operating in rooms without engineered ventilation or industrial emission controls. During extrusion, thermoplastics are heated to temperatures high enough to release airborne pollutants, including ultrafine particles and volatile organic compounds. Emission rate and chemical composition depend on printer design, filament material, nozzle temperature, chamber conditions, and operating parameters.^1-4

ABS is an important material for enclosed-printer emission studies because it is commonly printed at elevated temperatures and has been associated with styrene and other VOC emissions. Prior studies on desktop 3D printers have shown that both particles and VOCs can accumulate during operation, motivating control strategies such as lower-emission materials, room ventilation, printer enclosures, and filtration.^2-5 Enclosures can reduce direct release into the surrounding room, but containment alone does not remove gas-phase contaminants. If air inside the chamber is not treated effectively, VOCs may accumulate during the print and then be released when the enclosure is opened.

Activated carbon is widely used for gas-phase emission control because its porous structure provides high internal surface area for physisorption. Porous carbon materials can adsorb a range of VOCs through interactions between gas molecules and internal pore surfaces, making them useful for broad gas-phase filtration in enclosed systems.^6,7 In practical printer enclosures, filtration is often discussed primarily in terms of the amount of carbon present. A larger carbon bed provides more potential adsorption sites and can increase total capacity. However, capacity is not the same as utilization. A VOC molecule emitted near the hotend must first move through the chamber, reach the working region of the filter, pass through or near the carbon bed, and then diffuse to internal sorbent surfaces before adsorption can occur.^7,9

This distinction is especially important in small enclosures, where chamber geometry, fan placement, pressure drop, and airflow path may control how often contaminated air actually contacts the carbon. Increasing carbon mass can improve capacity, but it may also increase resistance through the carbon bed. Fan performance depends on the pressure-flow operating point; as static pressure rises, delivered airflow can decrease.⁸ Therefore, adding more carbon in the same location may produce diminishing returns if VOC-rich air is not transported efficiently to the sorbent. Under those conditions, active chamber recirculation could improve filtration by increasing air-carbon contact rather than by increasing carbon quantity.

This study tested whether gas-phase transport limits activated-carbon VOC mitigation in an enclosed FDM printing chamber. The hypothesis was that, at matched activated-carbon mass, active chamber recirculation would reduce VOC sensor response compared with passive circulation because additional chamber mixing would transport contaminated air more effectively toward the carbon bed. The experiment compared five conditions during fixed ABS extrusion: no carbon with passive circulation, 30 g carbon with passive circulation, 60 g carbon with passive circulation, 30 g carbon with active recirculation, and 60 g carbon with active recirculation. The main outcome measures were endpoint VOC response, cumulative VOC response, post-opening recovery, and differences between two sensor locations.

Methods

An 8 L acrylic chamber measuring 200 x 200 x 200 mm was used to contain emissions from a fixed ABS extrusion source. The chamber had a removable top panel that remained closed during extrusion and was opened after the 900 s generation period. The chamber contained the activated-carbon treatment, the carbon filtration module, auxiliary chamber-mixing fans when active recirculation was used, two SGP41 sensor housings, one BME688 sensor, and a fixed hotend assembly.

Black ABS filament (Elegoo, US-EL-3D-PA01) was extruded from a fixed Revo Hemera XS hotend through a 0.6 mm brass nozzle at 260 °C and 5 mm/s for 900 s. The hotend was fixed rather than mounted on a moving toolpath so that source position, extrusion rate, nozzle temperature, and chamber geometry remained consistent across trials. The same spool, nozzle, extrusion temperature, and extrusion speed were used for all treatment conditions. Before each trial, the open chamber stabilized for 10 min. Filament was stored in a drybox with silica gel packets before use.

Five treatment conditions were tested: 0 g carbon passive, 30 g carbon passive, 60 g carbon passive, 30 g carbon active, and 60 g carbon active. Four replicate trials were performed for each condition. Passive trials used only the filtration module’s internal blowers. Active trials used the same filtration module and additionally activated two auxiliary chamber-mixing fans positioned in opposite corners and directed toward the filter region. Fresh acid-free activated carbon was weighed after taring the carbon bed and was replaced for every trial.

The carbon filtration module was a Nevermore Micro V5 Duo containing two 5015 blowers rated at 4.8 CFM and 189.3 Pa. During active trials, two 4028 axial fans rated at 11 CFM and 82.2 Pa were used as auxiliary chamber-recirculation fans. The purpose of the active condition was not to add sorbent capacity, but to change chamber-scale gas transport while holding the carbon mass fixed within each matched comparison.

VOC levels were tracked using two SGP41 sensors at different chamber locations, one near the filter output region and one farther from the filter. A BME688 recorded gas resistance and temperature as secondary measurements. The sensors were connected through a TCA9548A multiplexer and FT232H USB breakout board and were enclosed behind size 40 stainless-steel mesh in custom housings to reduce direct turbulence and stabilize flow across the sensor surfaces. The SGP41 was treated as a broadband VOC sensor, and its raw SRAW output was converted into relative log₂-scaled response units using manufacturer documentation indicating approximately 870 SRAW ticks per doubling of gas signal.¹⁰

Each trial followed the same timing procedure. At t = 0 s, ABS extrusion began and continued for 900 s while the chamber remained closed. At t = 900 s, extrusion was stopped and the top acrylic panel was removed. The chamber was monitored for an additional 300 s, until t = 1200 s, to measure post-opening decay.

A separate 8 min empty-enclosure trial with no ABS extrusion was used to define clean-air reference values for each SGP41 channel. The reference SRAW values were 30297.02 ticks for channel P4 and 30592.39 ticks for channel P7. Because SRAW decreases as gas response increases, response was calculated as the decrease from the clean-air reference divided by 870.

R_SGP,i(t) = [SRAW_ref − SRAW_treatment(t)] / 870    (Eq. 1)

where R_SGP,i(t) is the transformed response for sensor channel i at time t, SRAW_ref is the clean-air reference for that channel, and SRAW_treatment(t) is the raw sensor value during a treatment trial.

R_mean(t) = [R_P4(t) + R_P7(t)] / 2    (Eq. 2)

where R_mean(t) is the primary chamber VOC response and R_P4(t) and R_P7(t) are the transformed responses from the two SGP41 channels.

H(t) = |R_P4(t) − R_P7(t)|    (Eq. 3)

where H(t) is the sensor-location heterogeneity at time t. Mean heterogeneity was calculated by averaging H(t) over the relevant trial interval.

R(b) = R_∞ + Ae^−kb    (Eq. 4)

where R(b) is the response after the chamber was opened, b is time after opening in minutes, R_∞ is the fitted final response, A is the response amplitude, and k is the recovery constant in min⁻¹.

For each replicate trial, peak response, endpoint response at 900 s, mean response from 0-900 s, cumulative response from 0-900 s, recovery constant, 50% recovery time, and mean heterogeneity were calculated. Cumulative response was calculated by integrating the baseline-referenced SGP41 response over the 0-900 s extrusion interval and is reported in log₂-min.

Statistical analysis was performed on replicate-level metrics. One-way ANOVA was used to compare treatment effects across the passive groups and across the complete treatment set for key response metrics. The four carbon-containing groups were also analyzed using two-way ANOVA with carbon mass and recirculation state as factors. Statistical significance was defined as p < 0.05.

Results

The passive conditions established the effect of carbon mass without auxiliary chamber recirculation. During the 900 s extrusion interval, the no-carbon passive control produced the largest VOC response, while both 30 g and 60 g passive carbon conditions produced lower response curves. At 900 s, SGP41 response decreased from 5.01 ± 0.17 log₂ units with no carbon to 2.00 ± 0.18 with 30 g carbon and 1.43 ± 0.20 with 60 g carbon. These values corresponded to reductions of 59.9% and 71.4% relative to the no-carbon passive condition.

Carbon mass also reduced the overall response accumulated during the extrusion interval. The mean 0-900 s response was 4.03 ± 0.16 in the no-carbon passive group, 1.58 ± 0.29 in the 30 g passive group, and 1.21 ± 0.24 in the 60 g passive group. Cumulative response decreased from 60.40 ± 2.45 log₂-min in the no-carbon group to 23.75 ± 4.29 log₂-min with 30 g carbon and 18.19 ± 3.54 log₂-min with 60 g carbon. Treatment group significantly affected peak response, endpoint response, and mean response across passive conditions (one-way ANOVA; peak response: F = 896.64, p = 1.19 × 10⁻¹⁷; 900 s response: F = 911.13, p = 1.05 × 10⁻¹⁷; mean response: F = 786.54, p = 3.37 × 10⁻¹⁷). The additional improvement from 30 g to 60 g passive carbon was smaller than the improvement from 0 g to 30 g. This pattern was consistent with a diminishing return from adding carbon mass under filter-only operation.

BME688 gas resistance followed the same general pattern as the SGP41 output, with lower overall response in the carbon-containing passive conditions. Chamber temperature rose during extrusion and reached approximately 34 °C near the end of the 900 s generation period across passive groups, indicating similar thermal conditions during the main interval.

Active recirculation tested whether changing gas-phase transport affected VOC response beyond the effect of carbon mass alone. At 30 g carbon, active recirculation reduced endpoint response from 2.00 ± 0.18 to 1.30 ± 0.21 and cumulative response from 23.75 ± 4.29 to 15.98 ± 3.14 log₂-min. At 60 g carbon, active recirculation reduced endpoint response from 1.43 ± 0.20 to 0.96 ± 0.03, a 32.6% decrease. The 60 g active condition had the lowest endpoint and cumulative response values among all measured conditions.

Two-way ANOVA on the four carbon-containing groups showed that both carbon mass and active recirculation significantly affected SGP41 response. Carbon mass significantly affected peak response (F = 73.81, p = 1.80 × 10⁻⁶), endpoint response (F = 69.94, p = 2.38 × 10⁻⁶), and mean response (F = 62.27, p = 4.65 × 10⁻⁶). Active recirculation also significantly affected peak response (F = 118.87, p = 1.40 × 10⁻⁷), endpoint response (F = 115.19, p = 1.66 × 10⁻⁷), and mean response (F = 86.31, p = 7.09 × 10⁻⁷). These results show that the active-recirculation effect was measurable across multiple response metrics rather than limited to a single comparison.

The comparison between 30 g active and 60 g passive was especially important for evaluating transport limitation. The 30 g active condition used half as much carbon as the 60 g passive condition, but produced a lower 900 s response (1.30 ± 0.21 vs. 1.43 ± 0.20) and a lower cumulative response (15.98 ± 3.14 vs. 18.19 ± 3.54 log₂-min). This result indicates that improved chamber mixing and air-carbon contact can compensate for, and in this case exceed, the response reduction produced by doubling carbon mass under passive circulation.

Recovery and heterogeneity metrics supported the same interpretation. At 30 g carbon, active recirculation shortened the 50% recovery time from 66.00 ± 8.52 s to 53.00 ± 3.44 s and increased the fitted recovery constant from 0.65 ± 0.09 to 0.82 ± 0.07 min⁻¹. At 60 g carbon, 50% recovery time changed from 52.00 ± 4.50 s to 49.25 ± 9.40 s, while k remained similar within variability. Mean sensor heterogeneity decreased from 0.30 ± 0.06 to 0.08 ± 0.06 at 30 g carbon and from 0.12 ± 0.10 to 0.04 ± 0.03 at 60 g carbon. Lower heterogeneity indicated that active airflow made the chamber response more spatially uniform.

Discussion

The results support the interpretation that activated-carbon filtration in the enclosed chamber was affected by both sorbent quantity and gas-phase transport. Passive trials showed that increasing carbon mass reduced VOC response, which is expected from an adsorption-capacity perspective. However, active recirculation produced additional reductions at matched carbon masses, meaning the response differences could not be explained by carbon quantity alone. The strongest evidence was the 30 g active condition, which produced lower endpoint and cumulative response than the 60 g passive condition despite using half the carbon mass.

The mechanism most consistent with these results is improved air-carbon contact. In the passive conditions, the filter’s internal blowers moved air locally through the carbon module, but chamber-scale flow may have remained uneven. Stagnant regions and weak natural mixing can allow VOC-rich air to persist away from the filter, while air near the carbon bed is cleaned more repeatedly. Under that condition, additional carbon in the same location may not be fully utilized during the 900 s extrusion interval because contaminated air does not reach the adsorption sites fast enough.

Active recirculation changed this transport environment. The auxiliary fans likely increased chamber-scale circulation, reduced stagnant regions, and moved VOC-containing air toward the filtration region more frequently. This interpretation is consistent with three observations: lower endpoint response, lower cumulative response, and reduced differences between the two sensor locations. The reduction in heterogeneity is particularly important because it indicates that active airflow did not simply lower one local sensor reading. Instead, it made the two sensor locations more similar, suggesting a more uniform chamber atmosphere.

The pressure-drop and fan-performance tradeoff helps explain why increasing carbon mass alone may produce limited returns. A larger carbon bed increases available adsorption capacity, but it also increases flow resistance. If the filtration fan operates against higher static pressure, total delivered airflow can decrease, limiting the amount of chamber air that reaches the sorbent per unit time.⁸ This creates a design tradeoff between capacity and throughput. The results suggest that active chamber recirculation can improve the utilization of available carbon capacity by increasing transport to the bed, rather than increasing the capacity of the bed itself.

The recovery constant provides secondary support but should not be treated as the primary measure of filtration during closed-chamber operation. After the chamber was opened, decay reflected room exchange, passive dispersion, sensor behavior, and the lower response already achieved at the end of extrusion. Even so, the shorter half-recovery time at 30 g active carbon was consistent with lower residual chamber response. The closed-chamber endpoint, cumulative response, and heterogeneity metrics are more direct evidence for the transport-limited interpretation.

Several limitations define the scope of the study. First, the SGP41 and BME688 are broadband metal-oxide sensors rather than compound-specific analytical instruments. The experiment therefore compared VOC response across controlled treatment conditions rather than identifying individual compounds. Future work could pair this chamber setup with gas chromatography-mass spectrometry to determine whether the active-recirculation effect differs across specific compounds such as styrene or other ABS-related VOCs.¹¹ Second, the study used fixed ABS extrusion rather than a full moving print. This improved repeatability by removing toolpath motion, part geometry, bed interaction, and print-cooling variables, but complete prints may create different source distributions and flow interactions. Third, the 8 L chamber is smaller than many commercial printer enclosures, so larger systems should be tested to evaluate scale effects.

Future studies should directly map the airflow mechanism. Smoke visualization, tracer-gas experiments, larger sensor arrays, or computational fluid dynamics could identify stagnant regions and quantify how fan placement changes transport to the carbon bed. Additional trials could test different carbon-bed geometries, filter locations, fan speeds, and enclosure volumes. The present study nevertheless shows that enclosure filtration should not be evaluated only by sorbent loading. Airflow structure is a core design variable that can determine whether available carbon capacity is actually used during short-duration printing.

Conclusion

This study tested whether gas-phase transport limits activated-carbon VOC mitigation during enclosed ABS extrusion. Adding activated carbon reduced VOC response, but active chamber recirculation produced additional reductions when carbon mass was held constant. The 30 g active condition produced lower endpoint and cumulative response than the 60 g passive condition, indicating that improving chamber mixing and air-carbon contact can rival simply adding more sorbent under the tested conditions.

The findings support the conclusion that enclosed-printer VOC filtration is both capacity-limited and transport-limited. Activated carbon quantity remains important, but airflow path, fan placement, pressure drop, and chamber recirculation determine how efficiently contaminated air reaches the carbon bed. Future enclosure designs should evaluate carbon mass and gas-phase transport together rather than treating filter capacity as the only performance variable.

References

1. Stephens, B. et al. Ultrafine particle emissions from desktop 3D printers. Atmos. Environ. 79, 334-339 (2013). doi:10.1016/j.atmosenv.2013.06.050.
2. Azimi, P. et al. Emissions of ultrafine particles and volatile organic compounds from commercially available desktop three-dimensional printers with multiple filaments. Environ. Sci. Technol. 50, 1260-1268 (2016). doi:10.1021/acs.est.5b04983.
3. Baguley, D. A. et al. Review of volatile organic compound (VOC) emissions from desktop 3D printers and associated health implications. J. Expo. Sci. Environ. Epidemiol. 36, 149-166 (2026). doi:10.1038/s41370-025-00778-y.
4. Stefaniak, A. B. et al. Characterization of chemical contaminants generated by a desktop fused deposition modeling 3-dimensional printer. J. Occup. Environ. Hyg. 14, 540-550 (2017). doi:10.1080/15459624.2017.1302589.
5. Azimi, P. et al. Predicting concentrations of ultrafine particles and volatile organic compounds resulting from desktop 3D printer operation and the impact of potential control strategies. Ann. Work Expo. Health 61, 789-802 (2017). doi:10.1093/annweh/wxx054.
6. Choma, J. et al. A concise review on porous adsorbents for benzene and other volatile organic compounds. Molecules 29, 5677 (2024). doi:10.3390/molecules29235677.
7. U.S. Environmental Protection Agency. Carbon Adsorbers Chapter. EPA Air Pollution Control Cost Manual, 7th edn. (U.S. EPA, 2018).
8. Corona, J. J. Jr. et al. The best efficiency point of an axial fan at low-pressure conditions. Adv. Mech. Eng. 13, 1-17 (2021). doi:10.1177/16878140211001188.
9. Khararoodi, M. G. et al. Modelling of sorbent-based gas filters for indoor environment: a comprehensive review. Build. Environ. 208, 108579 (2022). doi:10.1016/j.buildenv.2021.108579.
10. Sensirion AG. SGP41 VOC and NOx Sensor Datasheet and Gas Index Algorithm Documentation. Sensirion AG (2021).
11. Lovestead, T. M. & Urness, K. N. Gas chromatography-mass spectrometry (GC-MS). In ASM Handbook, Volume 10: Materials Characterization (ASM International, 2018).