CHEMICAL ENGINEERING · GAS-PHASE VOCs · TiO₂/SiO₂ · RESEARCH CASE STUDY

High-velocity fluidized-bed
photocatalysis

An experimental and modelling study of gas-phase n-hexane, n-pentane, benzene, toluene and butanone in a circulating annular photoreactor — from supported-catalyst synthesis and GC-PID sampling to Langmuir–Hinshelwood parameter estimation.

The dataset connects operating conditions, adsorption saturation, inlet/outlet concentration cycles, steady-state conversion, catalyst properties, kinetic constants and process limitations in one auditable engineering chain.

5 VOCs41 completed experiments20 wt% TiO₂/SiO₂UVC 254 nm23.8 L min⁻¹GC-PID/TIDPFR + Langmuir–HinshelwoodLevenberg–Marquardt

Explore the dataset View reactor architecture Open GitHub Download project package

Research dataset from FAPESP project 2020/09004-2, developed in the AdOx-CESQ/PQI-EPUSP research environment. Laboratory-scale results; no institutional endorsement or industrial-performance claim is implied.

Kinetic snapshot

Reconstructed model curves

RFLF · v1.0 Validated dataset

Completed matrix experiments

VOCs with fitted k and Ka

23.8 L min⁻¹

Airflow setpoint

430 m² g⁻¹

Catalyst surface area

C₀ (ppmv) vs X (%)n-hexanen-pentanebenzenetoluene

STUDY OVERVIEW

Experimental programme at a glance

Ten headline metrics summarise the completed experimental matrix, reactor operating point, analytical scope and catalyst used to estimate gas-phase VOC degradation kinetics.

Completed experiments

82.0%

Experimental plan delivered

VOCs studied

VOCs with kinetic fits

Fitted observations

1–103.4 ppmv

Inlet concentration window

23.8 L min⁻¹

Airflow

0.8 kg m⁻² s⁻¹

Solids circulation flux

75 W · 254 nm

Germicidal UVC source

430 m² g⁻¹

Coated catalyst surface area

20 wt%

TiO₂ loading

98%

Anatase phase

59 µm

Sauter mean d[3,2]

45–55%

Relative-humidity control band

The 41 completed matrix runs comprise 40 concentration–conversion observations used for four kinetic fits plus one successful butanone experiment. Control tests are documented separately and are not counted as additional matrix points.

This case study demonstrates a complete experimental modelling chain: catalyst synthesis, material qualification, controlled VOC generation, high-velocity solids recirculation, inlet/outlet GC measurement, steady-state validation, nonlinear parameter estimation and engineering interpretation.

Materials

20 wt% TiO₂/SiO₂, 98% anatase, 430 m² g⁻¹.

Experiments

41 completed runs, 40 fitted observations, 5 target VOCs.

Model

Plug-flow balance, Langmuir–Hinshelwood kinetics, fitted k and Kₐ.

ANALYTICAL CHAIN

From catalyst synthesis to kinetic interpretation

Each stage produces a traceable object or acceptance rule, preserving the link between material properties, operating conditions, analytical measurements, conversion results and fitted parameters.

Research framing

Method

Select representative aliphatic, aromatic and ketone VOCs; define the kinetic question and operating window.

Output

Five target VOCs and a 1–100+ ppmv study envelope.

Photocatalyst synthesis

Method

Coat SiO2 with 20 wt% TiO2 through controlled hydrolysis, dialysis, drying, calcination and sieving.

Output

Supported TiO2/SiO2 particles for fast fluidization.

Material characterization

Method

Validate crystalline phase, surface area and particle-size distribution.

Output

98% anatase, 430 m2/g and d32 = 59 µm.

Operating-point qualification

Method

Use the high-velocity plateau at Q = 23.8 L/min and Gs = 0.8 kg m−2 s−1 to minimize contact and irradiation limitations.

Output

Kinetically informative reactor operation.

Gas generation and conditioning

Method

Control VOC concentration through a 5 °C bubbler and humidity through a water branch.

Output

Stable contaminated-air feed at 45–55% RH.

Online analytical cycle

Method

Alternate two inlet and four outlet samples through a valve relay into GC-PID/TID.

Output

Calibrated inlet/outlet concentration series.

Steady-state degradation tests

Method

Activate UVC only after adsorption saturation and close runs after three stable cycles.

Output

41 completed matrix experiments and control evidence.

Kinetic estimation and interpretation

Method

Fit a PFR + Langmuir–Hinshelwood model by nonlinear least squares using Levenberg–Marquardt.

Output

k and Ka for four VOCs plus regime-level engineering conclusions.

MATERIAL → REACTOR → FEED → STEADY STATE → GC CYCLE → CONVERSION → FIT → INTERPRETATION

UVC is activated only after adsorption saturation is demonstrated by C ≈ C₀.

Stable pressure drop is used as evidence of stable solids holdup.

A run closes only after three stable inlet/outlet sampling cycles.

EXPERIMENTAL SYSTEM

A circulating annular photoreactor operated at the kinetic plateau

The selected operating point combines high gas velocity, maximum solids circulation and a small irradiated catalyst inventory to reduce mass-transfer, contact and irradiation limitations while maintaining measurable conversion.

23.8 L min⁻¹

Airflow Q

0.8 kg m⁻² s⁻¹

Solids flux Gs

200 g

Reservoir inventory

≈3 g

Irradiated-zone mass

45–55% RH

Humidity band

5 °C

VOC bubbler bath

The operating point was selected at the beginning of the high-velocity reaction-rate plateau. Above this region, contact and irradiation limitations are reduced, while excessively higher flow would lower conversion and increase uncertainty.

Analytical cycle

E · inlet sampleS · outlet sample

Three stable cycles required before closing a run.

airflow Qconstant

inlet concentration C0constant

outlet concentration CGC cycle / continuous

relative humidity RHGC cycle / continuous

pressure drop ΔPcontinuous

Control evidence

•Direct UVC photolysis was negligible for n-hexane, n-pentane, benzene and toluene under the study conditions.
•Butanone showed only about 5% removal under UVC alone at a doubled 5.4 s residence time, so photolysis was not material to the photocatalytic results.
•With catalyst present and UVC off, no relevant steady-state inlet/outlet concentration difference was observed for the four fitted VOCs.

MATERIAL SYSTEM

Supported TiO₂ engineered for fast fluidization and UVC activation

The catalyst combines a silica support suitable for fluidization with a photoactive TiO₂ coating, preserving high surface area while providing an anatase-dominant crystalline phase.

Synthesis flow

01Combine 8.5 mL TTIP with 1 M HNO3.

02Stir until the solution becomes transparent.

03Add NaOH to pH 2.

04Add SiO2 support.

05Agitate the suspension.

06Add NaOH to pH 3.

07Rest in dialysis tubing and repeat dialysis when required.

08Confirm supernatant pH above 5.

09Filter, dry, calcine at 450 °C and mechanically sieve the particles.

98% anatase / 2% rutile

XRD

430 m² g⁻¹

N₂ adsorption / BET

40–200 µm

Particle-size range

d[3,2] = 59 µm

Sauter mean diameter

Property	Value	Method
Crystalline phase	98% anatase, 2% rutile	X-ray diffraction
Surface area	430 m² g⁻¹	N₂ adsorption / BET
Particle size	40–200 µm; d[3,2] = 59 µm	Static light scattering
TiO₂ loading	20 wt%	Synthesis basis
Calcination	450 °C, 1 h	Thermal treatment

The final coated-catalyst value is 430 m² g⁻¹. The 460 m² g⁻¹ figure refers to the original silica support or earlier contextual characterisation, not the final dashboard KPI.

EXPERIMENTAL MATRIX

Forty fitted observations across four VOCs, plus one butanone case

The two project phases jointly cover ten inlet concentrations for each fitted pollutant. The combined matrix spans low-concentration adsorption control, a mixed kinetic regime and high-concentration surface-site limitation.

Planned matrix experiments

Completed

82.0%

Completion

Fitted observations

Successful butanone case

Nine planned butanone runs were not completed because strong adsorption extended operating time and caused excessive photocatalyst attrition and entrainment. The limitation is reported explicitly rather than hidden.

Planned concentration matrix (ppmv)

Source label normalised · n-pentane

Pollutant	C1	C2	C3	C4	C5	C6	C7	C8	C9	C10
n-hexane	2.4	5.9	12.0	22.8	30.1	43.6	52.6	61.4	72.3	91.7
n-pentane	7.1	14.3	29.1	37.2	48.3	62.2	70.6	85.3	86.5	103.4
benzene	2.8	5.1	9.6	17.5	29.5	32.1	41.2	56.3	59.8	61.9
toluene	4.9	7.4	10.0	19.9	24.6	32.3	45.8	47.8	52.5	63.4

Source label normalised: the final report matrix contains “cyclohexane”, but the concentration series, figures, narrative and kinetic constants consistently identify n-pentane.

GC-PID/TID settings

Oven70 °C, isothermal
Sample loop98 µL
Injector pressure4 kgf in⁻²
Carrier gasN₂
PID detector200 °C
ColumnRTX-Volatiles, 30 m, 0.32 mm ID, 2.0 µm film

Pollutant	LOD (ppmv)	LOQ (ppmv)	tR (min)
toluene	0.5	1.6	2.7
benzene	0.9	2.9	1.5
n-pentane	1.1	3.6	0.8
n-hexane	0.2	0.7	0.8
butanone	0.5	1.8	1

Experimental observations

Pollutant	Phase	C₀ (ppmv)	σ C₀	X (%)	σ X
n-hexane	R1	2.4	—	42.8	—
n-hexane	R2	5.9	0.4	31.4	3.3
n-hexane	R1	12.0	—	26.0	—
n-hexane	R1	22.8	—	15.9	—
n-hexane	R2	30.1	1.7	13.4	1.4
n-hexane	R2	43.6	1.4	17.8	3
n-hexane	R2	52.6	0.9	13.8	1.2
n-hexane	R1	61.4	—	12.4	—
n-hexane	R2	72.3	0.5	9.8	1.3
n-hexane	R1	91.7	—	7.6	—
n-pentane	R1	7.1	—	10.6	—
n-pentane	R1	14.3	—	8.6	—

KINETIC MODELLING

Langmuir–Hinshelwood fits reveal distinct adsorption and reaction regimes

A plug-flow material balance was combined with a single-component Langmuir–Hinshelwood rate law and fitted by nonlinear least squares to each ten-point pollutant dataset.

n-hexane

k 13.7·Kₐ 0.0204

Curve reconstructed from reported equations and constants.

n-pentane

k 13.4·Kₐ 0.00439

Curve reconstructed from reported equations and constants.

benzene

k 6.91·Kₐ 0.0752

Curve reconstructed from reported equations and constants.

toluene

k 11.5·Kₐ 0.076

Curve reconstructed from reported equations and constants.

Comparative model curves

Reconstructed · four VOCs

Pollutant	k (mg gcat⁻¹ h⁻¹)	Kₐ (m³ mg⁻¹)	Interpretation
n-hexane	13.7	0.0204	Fast surface reaction, moderate adsorption
n-pentane	13.4	0.00439	Fast surface reaction, weak adsorption
benzene	6.91	0.0752	Slower surface reaction, strong adsorption
toluene	11.5	0.076	Strong adsorption and high aromatic reactivity

Regime

C₀ < 10 ppmv · adsorption-sensitive

• Order: toluene > benzene > n-hexane > n-pentane.
• Kₐ strongly influences conversion.

Regime

10–70 ppmv · mixed control

• Order evolves toward toluene > n-hexane > benzene > n-pentane.
• Adsorption and surface reaction both matter.

Regime

C₀ > 70 ppmv · site-limited

• Available TiO₂ sites and residence time dominate.
• Modelled conversion remains below about 15%.

The aromatic compounds have adsorption constants about four times the n-hexane value and nearly twenty times the n-pentane value. Toluene combines strong adsorption with a substantially faster surface reaction than benzene.

OPERATIONAL EDGE CASE

Chemically favourable, operationally difficult

A single long-duration butanone run produced the highest conversion at a comparable inlet concentration, but strong adsorption greatly extended equilibration and amplified catalyst loss.

18 ppmv

Inlet · source 17.7 ppmv

43.5%

Steady-state conversion

2 h

Adsorption equilibration

7 h

Total experiment

91.3 g

Catalyst lost

108.7 g

Catalyst remaining

Incident-style timeline

t = 0 min
Start gas feed and adsorption equilibration
t = 120 min
Approximate adsorption equilibrium; UVC switched on
t = 180 min
Thermal desorption transient after lamp activation
t = 360 min
Long re-equilibration period continues
t = 420 min
Final steady state and mean conversion established

The report provides a plotted profile but not a tabulated digital time series; milestones are shown instead of a fabricated continuous curve.

The butanone run lost 91.3 g of photocatalyst, compared with approximately 10–15 g for a typical experiment with the other VOCs. Strong adsorption made the chemistry favourable but the high-velocity fluidized-bed configuration operationally costly for this compound.

No k or Kₐ is reported for butanone. Do not extrapolate a kinetic fit from this single run.

MODEL ARCHITECTURE

A plug-flow balance coupled to surface adsorption and reaction

The model separates reactor transport representation from the intrinsic Langmuir–Hinshelwood rate expression, then estimates pollutant-specific adsorption and surface-reaction constants.

Plug-flow material balance

Q · dC/dm = −rₘ

Concentration decreases along catalyst mass; the source report uses its own rate-sign convention.

Langmuir–Hinshelwood rate

rₘ = (k · Kₐ · C) / (1 + Kₐ · C)

Conversion

X = 1 − C/C₀

Variable dictionary

Q: gas volumetric flow, m³ h⁻¹
C₀: inlet VOC concentration, mg m⁻³ or ppmv before unit conversion
C: outlet/local VOC concentration
m: catalyst mass in the irradiated reactor zone
rₘ: mass-specific reaction rate, mg gcat⁻¹ h⁻¹
k: surface-reaction constant, mg gcat⁻¹ h⁻¹
Kₐ: adsorption constant, m³ mg⁻¹

Regression method

Nonlinear least squares

lsqnonlin routine

Levenberg–Marquardt algorithm

Ten observations per fitted VOC

≈3 g catalyst in irradiated zone

Q = 1.428 m³ h⁻¹ from 23.8 L min⁻¹

The local model_curves.json file is a transparent reconstruction for website visualisation. It uses the reported equations, k, Kₐ, Q = 1.428 m³ h⁻¹, approximately 3.0 g irradiated catalyst and a 24.45 L mol⁻¹ gas conversion basis at 25 °C. These points are model outputs, not additional measurements.

Single-component adsorption model.

Plug-flow representation rather than a full residence-time-distribution model.

No explicit radiation-field attenuation term.

No explicit void-fraction submodel inside this page.

Parent-VOC conversion, not mineralization or by-product yield.

No fitted-parameter confidence intervals reported.

Reproducible package

data/ → notebooks/ → model/ → figures/ → docs/ → evidence/ → qa/ → website/

Download experimental dataset Read final report

METHODOLOGY

A traceable method grounded in photocatalysis and fluidization literature

The project combines gas-phase VOC chemistry, supported-semiconductor preparation, circulating fluidized-bed operation, chromatographic measurement and nonlinear kinetic estimation.

VOC problem framing

Atmospheric relevance, difficult removal and treatment trade-offs.

Heterogeneous photocatalysis

UVC activation, electron–hole generation and surface oxidation.

Fast fluidization

Improved pollutant–catalyst–radiation contact with attrition trade-off.

Analytical measurement

Calibrated inlet/outlet GC-PID/TID cycles.

Steady-state validation

C≈C₀ before UVC, stable ΔP and RH, three stable cycles.

Kinetic estimation

PFR + Langmuir–Hinshelwood fitted by Levenberg–Marquardt.

Literature

SRC-01

Schneider et al. (2014), Understanding TiO2 Photocatalysis: Mechanisms and Materials, Chemical Reviews 114, 9919–9986.

Photocatalytic mechanism and TiO2 fundamentals.

SRC-02

Alberici & Jardim (1997), Photocatalytic destruction of VOCs in the gas phase using titanium dioxide, Applied Catalysis B 14, 55–68.

Gas-phase VOC photocatalysis and aromatic surface effects.

SRC-03

Diniz et al. (2021), Photocatalytic degradation of n-hexane in a circulating fluidized bed, Catalysis Today 361, 109–116.

High-velocity operating point and reactor hydrodynamics.

SRC-04

Matsumoto, Diniz & Teixeira (2020), kinetic modelling and experimental validation of an RFLF for n-hexane degradation, Brazilian Journal of Chemical Engineering 36.

Prior predictive model and motivation for intrinsic kinetic parameters.

SRC-05

Boulamanti & Philippopoulos (2009), Photocatalytic degradation of C5–C7 alkanes in the gas phase, Atmospheric Environment 43, 3168–3174.

Comparative alkane kinetics.

SRC-06

Mahmood et al. (2021), Langmuir–Hinshelwood kinetics of aromatic-ring VOCs, Journal of Environmental Chemical Engineering 9, 105069.

Aromatic adsorption and reactivity interpretation.

SRC-07

Korologos et al. (2012), Photocatalytic oxidation of benzene and toluene over TiO2-based catalysts, Journal of Photochemistry and Photobiology A 244, 24–31.

Toluene versus benzene reactivity and intermediates.

SRC-08

Kunii & Levenspiel (1991), Fluidization Engineering.

Fluidized-bed and circulating-solids fundamentals.

Source normalisation

Pollutant label in the final experiment matrix

Use n-pentane. The row labelled cyclohexane contains the n-pentane concentration series and conflicts with every narrative, chart and kinetic-parameter table.

Specific surface area

Use 430 m2/g for the final coated TiO2/SiO2 catalyst. Keep 460 m2/g only as the original silica-support or contextual value.

Airflow notation

Use 23.8 L/min as the exact experimental setpoint. Use 24 L/min only when reproducing a rounded caption or summary.

Butanone inlet concentration

Display 18 ppmv as in the detailed results narrative and retain 17.7 ppmv as the source experiment-matrix value in tooltips or notes.

Model curves

State that smooth curves are reconstructed from the reported PFR–Langmuir–Hinshelwood equations, k and Ka, Q = 1.428 m3/h, approximately 3 g irradiated catalyst and a 24.45 L/mol gas conversion basis at 25 °C.

Laboratory-scale research dataset.

No industrial-scale performance guarantee.

No institutional endorsement implied.

No claim of complete mineralization.

No k/Kₐ for butanone.

No invented R², confidence interval, raw monitoring time series or by-product profile.

QUALITY ASSURANCE

Every headline number has a reconciliation rule

Fifteen checks reconcile the experiment counts, pollutant scope, operating conditions, material properties, fitted parameters, source normalisations and model-curve provenance.

ID	Control	Expected	Status
QA-01	Completed-run reconciliation	41 = 40 fitted observations + 1 butanone run	PASS
QA-02	Planned-run reconciliation	41 / 50 = 82.0% completion	PASS
QA-03	Fitted dataset cardinality	4 VOCs × 10 inlet concentrations = 40 observations	PASS
QA-04	Pollutant scope	n-hexane, n-pentane, benzene, toluene and butanone	PASS
QA-05	Kinetic parameter fidelity	k and Ka match the final report table exactly	PASS
QA-06	Butanone boundary	No k or Ka is shown for butanone	PASS
QA-07	Reactor setpoint fidelity	Q = 23.8 L/min; 24 L/min appears only as a rounded label	PASS
QA-08	Solids-flux fidelity	Gs = 0.8 kg m−2 s−1	PASS
QA-09	Humidity control	45–55% RH	PASS
QA-10	Catalyst phase balance	98% anatase + 2% rutile = 100%	PASS
QA-11	Surface-area normalization	430 m2/g for coated catalyst; 460 m2/g retained only as support/context value	PASS
QA-12	Source nomenclature normalization	Final matrix row is labelled n-pentane, not cyclohexane, because all experiments, figures and fitted constants identify n-pentane	PASS
QA-13	Mass distinction	200 g reservoir inventory is not confused with approximately 3 g in the irradiated zone	PASS
QA-14	Model-curve provenance	Curves are labelled reconstructed from reported equations and constants, not additional measurements	PASS
QA-15	No unsupported claim	No R2, mineralization, by-product yield, industrial scale-up factor or fitted uncertainty is invented	PASS

40 / 40

Fitted observations present

4 × 10

Concentration points reconciled

41 / 50

Experimental plan

VOC identities resolved

Exact k/Kₐ pairs

Invented R² or fitted uncertainties

$ npm run build

$ npm run test:data

# all dataset reconciliation checks passed

ENGINEERING READ-OUT

What the results change in reactor interpretation

C01

Adsorption dominates pollutant ranking at low inlet concentration.

C02

n-Pentane is limited primarily by weak adsorption despite a surface-reaction constant similar to n-hexane.

C03

Toluene combines strong adsorption and higher surface reactivity than benzene.

C04

At high concentration, site availability and residence time constrain all four fitted VOCs to low conversion.

C05

Butanone is chemically easy to remove but operationally unsuitable for long high-velocity runs without attrition mitigation.

Future modelling directions

Void-fraction prediction.

Radiation-field attenuation as solids concentration rises.

Hydrodynamics more complex than ideal PFR.

Attrition-aware reactor configuration for strongly adsorbing VOCs.

REPRODUCIBILITY

Data, report, method and model connected to the same evidence chain

The final report, structured dataset, model reconstruction, technical specification and QA notes are exposed through stable local paths and a single project configuration.

Final scientific report

Complete FAPESP final report with methodology, results and discussion.

Experimental dataset

Structured concentration–conversion data used for the kinetic fits.

Complete project package

Data, notebooks, model reconstruction, figures and QA in one archive.

GitHub repository

Source, data files and reproducible model reconstruction.

Methodology Technical specification QA report