Advertisement

Emission Control Science and Technology

, Volume 4, Issue 3, pp 172–197 | Cite as

Development and Validation of a Two-Site Kinetic Model for NH3-SCR over Cu-SSZ-13. Part 2. Full-Scale Model Validation, ASC Model Development, and SCR-ASC Model Application

  • Rohil Daya
  • Chintan Desai
  • Bruce Vernham
Special Issue: 2017 CLEERS October 3 - 5, Ann Arbor, MI, USA

Abstract

We present herein the final part in the development and validation of a two-site kinetic model for NH3-SCR over Cu-SSZ-13. To predict tailpipe emissions accurately, it was necessary to combine the kinetic model of the SCR catalyst developed in part I (Daya et al. 2018), with a reaction-diffusion model of the dual-layer ammonia slip catalyst (ASC). This dual-layer ASC model was developed following a three-step process, including development of the kinetic models of the individual layers, followed by parameterization of a parallel pore diffusion model of the dual-layer ASC. Reactor-scale validation of the dual-layer ASC model confirmed the kinetic model accuracy and highlighted the significance of intra-porous diffusion. Following this, the SCR model developed in part I of this paper was validated on an engine dynamometer through comprehensive steady-state experiments with inlet NH3 to NOx ratio (ANR) sweeps. The final SCR and ASC models were then evaluated on cold and hot heavy-duty transient (HDT) cycles, to examine the capability of predicting tailpipe NOx, NH3 slip as well as storage-based dynamics. Overall cycle-averaged NOx conversion was predicted within 3% using these models. Validated models have significant application in model-based control as well as improving catalyst design through improved functional understanding. The present Cu-SSZ-13 SCR model was simulated using the four-step SCR protocol (Kamasamudram et al. Catal. Today 151(3):212–222) to calculate the intra-catalyst dynamic capacity. These numerical experiments showed that the dynamic capacity decreases upon hydrothermal aging but leads to higher NOx conversion under standard and fast SCR conditions at 250 °C. This increase in NOx conversion is due to more uniform NH3 storage along the length of the catalyst, leading to higher NH3 utilization near the rear of the aged catalyst. Similar numerical experiments on the dual-layer ASC model demonstrated intra-layer washcoat distributions causing NO slip during transient drive cycles for both hydrothermal aging conditions.

Keywords

Global Kinetic Model Dual-Layer ASC Model Intra-Porous Diffusion Drive Cycle Validation Hydrothermal Ageing 

Nomenclature

Deff

effective diffusivity m2/s

Dgas

species bulk diffusivity m2/s

DKn

Knudsen diffusivity m2/s

dp

washcoat pore diameter m

τ

washcoat tortuosity

ε

washcoat porosity

1 Introduction

Oxides of nitrogen (NOx) represent one of the major sources of air pollution, and a significant portion of these emissions derive from heavy-duty mobile sources, contributing to formation of ground-level ozone, known as smog [1]. Therefore, a reduction of these emissions is crucial to reduce ground level ozone and meet the air quality standards set by the National Highway Traffic Safety Administration (NHTSA) [2]. Consequently, it is expected that future NOx emission regulations will be 90% lower than the present limits, requiring extremely high NOx conversions [3]. A secondary benefit of improved NOx conversion is the reduction in greenhouse gas emissions through more efficient engine operation with increased engine-out NOx. Selective catalytic reduction (SCR) is one promising technique that has been widely used to overcome the issue of NOx reduction under lean conditions [4]. The most convenient form of enabling SCR is through injection of a 32.5 wt.% urea water solution (UWS), followed by decomposition of UWS to form the reductant ammonia (NH3) [5]. Inside the SCR, NH3 stores and reacts with NOx according to a well-established global reaction chemistry. Typically, unless there is an excess of nitrogen dioxide (NO2), 1 mol of NH3 is required to reduce 1 mol of NOx to produce 1 mol of nitrogen (N2) [6]. However, it is often the case that over stoichiometric amounts of NH3 is supplied to maximize NOx conversion under certain conditions [7]. Additionally, pre-storing of NH3 at low temperatures is a common technique used to convert NOx before it is possible to inject UWS without deposit formation. In such instances, it is possible for excess NH3 to desorb out of the SCR catalyst unreacted, typically above 300 °C and/or during sudden transient spikes [8].

NH3 is colorless, toxic, and corrosive gas with an unpleasant odor, detectable at levels of 10 ppm or higher [9]. Overexposure to ammonia over 300 ppm can cause lung diseases [10]. Furthermore, NH3 can react with sulfuric acid (H2SO4) and water (H2O) in the atmosphere, enhancing the rate of aerosol formation over the binary formation of H2SO4 and H2O [11]. To minimize the amount of released NH3, 10 ppm is considered a typical threshold limit over transient test cycles, as regulated by the European Commission (EC) [12].

To keep the tailpipe NH3 levels below this threshold, and to allow for aggressive dosing strategies, an additional downstream catalyst is necessary. This catalyst, labeled the ammonia slip catalyst (ASC), typically consists of platinum on a washcoated alumina support (Pt/γ-Al2O3) and is highly active for NH3 oxidation above 200 °C [13]. However, one of the major downsides of Pt-based ASC catalysts is that the oxidation of NH3 is highly selective to nitrous oxide (N2O) and NOx, according to the global reaction chemistry shown below [13, 14]:

$$ 2{\mathrm{N}\mathrm{H}}_3+2{\mathrm{O}}_2\to {\mathrm{N}}_2\mathrm{O}+3{\mathrm{H}}_2\mathrm{O} $$
(1)
$$ 4{\mathrm{NH}}_3+5{\mathrm{O}}_2\to 4\mathrm{NO}+6{\mathrm{H}}_2\mathrm{O} $$
(2)
$$ 4{\mathrm{NH}}_3+7{\mathrm{O}}_2\to 4{\mathrm{NO}}_2+6{\mathrm{H}}_2\mathrm{O} $$
(3)

To overcome this issue, a dual-layer ASC was proposed and introduced commercially [15]. This modern ASC consists of a SCR catalyst coated in a layer on top of the Pt-based ASC catalyst. Products downstream of the main SCR catalyst (NH3 and/or NOx) first diffuse through the SCR layer. Any excess NOx is typically converted with NH3 in this layer. Excess NH3 then reaches the Pt layer, where it reacts with O2 to form N2, N2O, and/or NOx. The excess NOx then reacts with stored NH3 on the SCR layer, leading to high overall selectivity to N2. It has been demonstrated that the dual-layer ASC is highly effective in converting slipped NH3 to harmless N2 [7].

With this degree of complexity in the exhaust aftertreatment system, mathematical modeling plays an important role to reduce time and costs associated with development [16]. In the case of an ASC, these models are also useful for accurate tailpipe NOx estimation, along with N2O slip prediction, allowing for improved model-based control of the UWS dosing. Fundamental understanding of the physiochemical processes occurring inside the ASC derives from previous works on SCR and PGM catalysts. Due to its importance in the production of nitric acid (Ostwald process), the oxidation of NH3 on Pt is an extremely well-studied reaction [14]. The SCR layer on ASC catalysts is typically very similar to the state-of-the art SCR catalysts, which have received considerable attention [17, 18, 19].

These understandings from literature have recently been incorporated into different kinds of 1 + 1D reaction-diffusion models of the ASC [12, 13, 14, 15, 20, 21, 22]. Scheuer et al. [13, 14, 22] developed a micro kinetic model for NH3 oxidation over an ASC, accounting for adsorption and desorption of all relevant species. They later extended this model and included a global SCR mechanism along with a numerical intra-porous diffusion model to completely describe the behavior of a dual-layer ASC. Furthermore, they used a solution mapping approach to reduce the computation time issue present in the numerical 1 + 1D model. Sukumar et al. [15] presented a global dual-layer reaction model, with internal mass transfer coefficients to account for intra-porous diffusion. Colombo et al. [12, 20, 21] noticed that the concentration gradients in a dual layer ASC are almost always limited to the thicker SCR layer and used this understanding to develop a “layer + surface model (LSM)” to account for all the relevant physiochemical processes in a computationally efficient manner. They also proposed a new global reaction mechanism for the Pt-layer, with NO2 inhibition of NO oxidation and NH3 oxidation to N2, NO and N2O, along with NO2 SCR reactions. A similar kinetic approach as suggested in [12, 20, 21] is adapted in the present work.

In the first part of the paper, a detailed global kinetic model was developed to characterize the behavior of the Cu-SSZ-13 SCR catalyst [23]. This part extends the previous modeling work, following a similar approach to develop a quasi-1 + 1D reaction-diffusion model of the dual-layer ASC catalyst, utilizing an asymptotic pore diffusion approach [24], along with detailed global kinetics. The experimental section begins with a description of the reactor testing performed to develop the ASC model. The procedures used to validate the full-scale SCR and ASC models on the engine dynamometer are also outlined in this section. Next, the ASC reaction-diffusion modeling structure, including the reaction mechanism and the parallel pore diffusion model, is explained in detail. The results section illustrates the comparisons between experimental and simulated NH3 conversions and selectivities for the PGM-coated ASC and the SCR-coated ASC. Following this, intra-porous diffusion model development and reactor-scale validation of the final ASC model using dual-layer transient reactor data is demonstrated. The full-scale model validation section shows the applicability of the developed SCR and ASC models, simulated over steady-state and transient engine-out conditions with varying inlet ANRs. Finally, the validated SCR and ASC models are analyzed to understand internal mechanistic and diffusion behaviors. These models should serve as a starting point for development of state estimators, which can be embedded in DCUs for model-based UWS dosing control.

2 Experimental Methods

2.1 Laboratory Reactor Experiments

Reactor experiments were conducted to develop the dual-layer ASC model. The modeling approach was split into three steps, and governed the collection of reactor data. These steps were as follows:
  1. 1.

    Development of PGM-coated ASC kinetic model

     
  2. 2.

    Development of SCR-coated ASC kinetic model

     
  3. 3.

    Integration of PGM-coated and SCR-coated ASC kinetics, along with parameterization of a parallel pore diffusion model

     

Therefore, three sets of experimental protocols were developed to characterize three different ASCs, and each one is described briefly below. One of the requirements of the testing was to keep the reactor as close to isothermal operation as possible, to avoid influence of any axial temperature gradients on steady-state conversion results. The detailed reactor setup and instrumentation is described in part I of this paper [23].

Prior to reactor testing, all the ASC samples were hydrothermally aged at two different conditions in the presence of 10% H2O/air: 650 °C for 16-h (degreening) and 700 °C for 100-h (aging). H2O (4.5%) in balance Ar was utilized for all the reactor experiments.

2.1.1 PGM-Coated ASC

An ASC coated with Pt on γ-Al2O3 was used for the experiments. The monolith used for the core reactor testing was a square cross-section with the following physical properties: 18 mm height, 89 mm length, 300 cells per square inch (nominal cpsi), and a wall thickness of 0.14 mm. The reactor testing primarily consisted of isothermal experiments at five different temperatures between 200 and 550 °C with varying feed gas concentrations to sequentially characterize all the reactions expected to occur between NH3, NOx and O2 on a Pt/γ-Al2O3 catalyst [20]. This included NH3 storage, NO oxidation, NH3/O2 interactions, NH3/NO/O2 interactions, and NH3/NO2/O2 interactions. Typical concentrations used were 150 ppm NH3, 150 ppm NOx, and 10% O2. Note that wherever NOx was present, a fixed inlet NH3 to NOx ratio (ANR) of 1 was used in the reactor experiments, and no tests were performed with co-feeding NO and NO2, since they are not expected to react simultaneously in the presence of NH3. However, such a test could be used for validation of the developed kinetic model. For NO oxidation, the inlet NO2/NOx ratio was 0, since minimal NO2 slipped from the upstream SCR or formed in the SCR layer and diffused to the PGM layer. Additional TPR testing was conducted between 200 and 550 °C with 150 ppm NH3, 150 ppm NO and 10% O2 for first-stage validation purposes. The ramp rate used for the TPR experiments was 10 °C/min.

2.1.2 SCR-Coated ASC

The experiments were performed on an ASC coated with a Cu-SSZ-13 SCR catalyst. The monolith used for the core reactor testing was a square cross-section with the following physical properties: 18 mm height, 89 mm length, 300 cells per square inch (nominal cpsi), and a wall thickness of 0.14 mm. The reactor testing primarily consisted of isothermal experiments at five different temperatures between 200 and 550 °C with varying feed gas concentrations to sequentially characterize all the reactions expected to occur on a Cu-SSZ-13 catalyst [23]. This included NH3 storage, NO oxidation, NH3 oxidation, standard SCR, fast SCR, and NO2-SCR reactions. Typical concentrations used were 150 ppm NH3, 150 ppm NOx, and 10% O2. For NO oxidation, the inlet NO2/NOx ratio used 0.33, since some NO2 might be formed in the Pt layer and diffuse back to the SCR layer. Fast SCR experiments were performed using an inlet NO2/NOx ratio of 0.5, and NO2-SCR experiments were performed using a stoichiometric inlet ANR of 1.33, with 112.5 ppm NO2.

2.1.3 Dual-Layer ASC

A commercial dual-layer ASC coated with the SCR and PGM layer catalysts was obtained for the experiments. The monolith used for the core reactor testing was a square cross-section with the following physical properties: 17 mm height, 89 mm length, 300 cells per square inch (nominal cpsi), and a wall thickness of 0.14 mm. The length of the monolith used was identical to the full-scale ASC. An internal test protocol was developed that consisted of the following:
  • Isothermal experiments at four temperatures between 250 and 550 °C to sequentially characterize NH3 oxidation and NH3/NO/O2 interactions at an inlet ANR of 1

  • Isothermal experiments at five temperatures between 250 and 150 °C for NH3/NO/O2 and NH3/NO/NO2/O2 interactions at an inlet ANR of 1, followed by TPR experiments with a ramp-up to 550 °C at 10 °C/min

All the above sets of experiments were repeated for degreened and aged catalysts at two gas hourly space velocities (GHSV) at STP 70,000 and 140,000/h. Since multiple tests were conducted for 4 different conditions (2 GHSVs and 2 hydrothermal aging conditions), the following nomenclature is used in reporting the data and model prediction results throughout the paper:
  • D-70 degreened catalyst at GHSV of 70,000/h

  • D-140 degreened catalyst at GHSV of 140,000/h

  • A-70 aged catalyst at GHSV of 70,000/h

  • A-140 aged catalyst at GHSV 140,000/h

Experimental and modeling results are presented in terms of NH3 conversion and product yields, calculated using the following equations [20]:

$$ {\mathrm{NH}}_3\ \mathrm{conversion}\ \left(\%\right)=\frac{{\mathrm{NH}}_{3_{\mathrm{in}}}-{\mathrm{NH}}_{3_{\mathrm{out}}}}{{\mathrm{NH}}_{3_{\mathrm{in}}}}\times 100 $$
(4)
$$ {\mathrm{N}}_2\ \mathrm{yield}\ \left(\%\right)=\frac{2{\mathrm{N}}_{2_{\mathrm{out}}}}{{\mathrm{N}\mathrm{H}}_{3_{\mathrm{in}}}+{\mathrm{N}\mathrm{O}}_{{\mathrm{x}}_{\mathrm{in}}}}\times 100 $$
(5)
$$ {\mathrm{NO}}_{\mathrm{x}}\ \mathrm{yield}\ \left(\%\right)=\frac{{\mathrm{NO}}_{{\mathrm{x}}_{\mathrm{out}}}}{{\mathrm{NH}}_{3_{\mathrm{in}}}+{\mathrm{NO}}_{{\mathrm{x}}_{\mathrm{in}}}}\times 100 $$
(6)
$$ {\mathrm{N}}_2\mathrm{O}\ \mathrm{yield}\ \left(\%\right)=\frac{2{\mathrm{N}}_2{\mathrm{O}}_{\mathrm{out}}}{{\mathrm{N}\mathrm{H}}_{3_{\mathrm{in}}}+{\mathrm{N}\mathrm{O}}_{{\mathrm{x}}_{\mathrm{in}}}}\times 100 $$
(7)

2.2 Engine Dynamometer Experiments

Steady-state and transient engine dynamometer experiments were conducted on a 3.0-L turbocharged diesel engine to validate the developed SCR and ASC models. The steady-state experiments consisted of inlet ANR sweeps at fixed engine speeds and loads that covered the range of real-world operating temperatures, GHSVs, and inlet NO2/NOx ratios. The transient experiments included both the cold and hot heavy-duty transient (HDT) cycles, with a 20-min hot soak between them. The SCR and ASC monoliths used had the following physical properties:
  • SCR (Cu-SSZ-13): 19.05 cm diameter, 38.1 cm length (for HDT, and 31.6 cm length for steady-state), cell density of 400 CPSI, and a wall thickness of 0.114 mm

  • Dual-layer ASC (Cu-SSZ-13 on top of Pt/γ-Al2O3): 19.05 cm diameter, 8.89 cm length, cell density of 300 CPSI, and a wall thickness of 0.14 mm

Prior to engine dynamometer testing, both catalytic converters were aged hydrothermally at 650 °C for 100 h, with minimal on-engine aging. Although the hydrothermal aging condition was only slightly different from the reactor tests, this could possibly account for minor deviations.

The overall aftertreatment system consisted of a diesel oxidation catalyst (DOC), a catalyzed diesel particulate filter (cDPF) followed by the SCR and ASC catalysts. There was no NH3 pre-storage for the steady-state tests, while for the HDT cycle, NH3 was pre-stored to simulate the real certification cycle. UWS was injected between the cDPF and SCR catalysts, targeting a certain NH3 load depending on the inlet conditions. No UWS was injected until the SCR inlet temperature was greater than 200 °C. A chemiluminescence detector (Horiba MEXA-7000 series) measured the NOx concentration at the inlet of SCR and the outlet of ASC, while an FTIR (AVL SESAM i-60) was placed between the SCR and ASC catalysts. A paramagnetic analyzer measured the SCR-inlet and ASC-outlet O2 concentrations.

3 Kinetic Modeling

3.1 Reactor Model

The reaction-kinetic modeling was performed using GT-SUITE v2017. The overall approach used here was analogous to the one described in the first part of the paper [23], and therefore only new aspects related to modeling of the dual-layer ASC are mentioned here. The reader should refer to [23] for detailed information of the reactor modeling approach for the single-layer ASCs, including the list of governing equations. In the dual-layer ASC model, each layer had its own set of reaction rates and coverages, although the species were common to both. The governing equations described in [23] still hold for the upper SCR layer and the lower PGM layer, with the necessary continuity boundary condition at the SCR-PGM layer interfaces, and the no-flux boundary condition at the PGM layer-wall interface.

An effective diffusivity of 5e−6 m2/s was used for the single-layer catalysts, due to lack of experimental data. However, due to the relatively increased significance of intra-porous diffusion for dual-layer ASCs, the parallel pore diffusion model was used for calculating effective diffusivities in each layer, as described below.

Intra-porous diffusion in the dual-layer ASC was accounted for using the recently developed asymptotic approach [24]. Due to the utilization of this approach, no significant increase in computation time was observed with the inclusion of the dual-layer intra-porous diffusion model. The effective diffusivity for each species in each layer was calculated using Eq. (8) [25]:

$$ \frac{1}{D_{eff}}=\frac{\tau }{\varepsilon}\left(\frac{1}{D_{gas}}+\frac{1}{D_{Kn}}\right) $$
(8)
where
$$ {D}_{Kn}=\frac{d_p}{3}\sqrt{\frac{8\overline{R}T}{{\pi M}_w}} $$
(9)

In Eq. (8), Dgas was calculated from the Fuller correlation [26]. The washcoat pore diameter determines the relative significance of Knudsen diffusion in the washcoat and in the present work was set to 5 μm, as recommended in [25]. While this is on the higher side as compared to values measured using mercury porosimetry, this value is primarily representative of mesopores which are expected to be the least resistive pathway for internal diffusion of gas molecules. The porosity/tortuosity ratio governs the overall effective diffusivity and was calibrated using dual-layer ASC reactor data.

All the governing equations mentioned in [23] were solved using GT-SUITE’s Advance Adaptive solver [25]. An initial discretization of 20 axial sub-volumes was used to generate a fixed uniform axial mesh for the method of lines (MOL) solution of the solid-phase temperature and the coverage equations. The MOL solution was obtained using LSODI as the variable time-step integrator [25, 27]. These temperature and coverage solutions were then coupled with the quasi-steady species governing equations, which utilize an adaptive mesh to accurately capture reaction fronts [25].

3.2 Kinetic Model

The approach for developing the kinetic model relied on detailed global reactions to describe the surface chemistry. Eley-Rideal-type kinetics were used, with reactions occurring between diffused gas-phase species and adsorbed NH3 [28]. As mentioned in the experimental section, the overall model was developed in three stages, utilizing three different ASCs, allowing for isolation of species reactions on each of two ASC layers, along with the intra-porous diffusion effect. The reaction mechanism used for each of the ASC layers is discussed below.

3.2.1 PGM-Coated ASC

The overall reaction mechanism for the PGM-coated ASC was adapted from [20], with minor modifications, and is shown in Table 1, where kf and kb refer to the forward and reverse rate constants respectively, described according to the Arrhenius Eq. (10). In Eq. (10), A and Ea refer to the frequency factor and activation energy respectively, and T refers to the catalyst temperature. Species concentrations in the reaction rate were described in units of mol/m3, and the overall reaction rate was described on a turnover number basis, in units of mol/m3reactor-s. The key difference between the mechanism developed by the authors in [20] and the one shown in Table 1 is the absence of a NO2-SCR-type reaction to selectively form N2. This was considered redundant in the present modeling work:
Table 1

PGM reactions and rate expressions

Reaction no.

Reaction

Rate expression

R1

NO + 0.5O2 ⇿ NO2

\( {r}_1=\frac{k_1\left({y}_{NO}{y}_{O_2}^{0.5}-\frac{y_{NO_2}}{K_{eq}}\right){\varOmega}_1}{1+{K}_{N{O}_2 PGM}.{C}_{N{O}_2}} \)

R2

NH3 + S1 ⇿ NH3 − S1

\( {r}_{2f}={k}_{2,f}{y}_{NH_3}{\theta}_{S1}{\varOmega}_1 \)

\( {r}_{2b}={k}_{2,b}{\theta}_{NH_3S1}{\varOmega}_1 \)

R3

2NH3 − S1 + 1.5O2 ⇾ N2 + 3H2O + 2S1

\( {r}_3=\frac{k_3{y}_{O_2}^{0.5}{\theta}_{NH_3S1}{\varOmega}_1}{1+{K}_{NN}.{C}_{N{O}_2}} \)

R4

2NH3 − S1 + 2.5O2 ⇾ 2NO + 3H2O + 2S1

\( {r}_4=\frac{k_4{y}_{O_2}^{0.5}{\theta}_{NH_3S1}{\varOmega}_1}{1+{K}_{NO}.{C}_{N{O}_2}} \)

R5

2NH3 − S1 + 2NO + 1.5O2 ⇾ 2N2O + 3H2O + 2S1

\( {r}_5=\frac{k_5{y}_{NO}{\theta}_{NH_3S1}{\varOmega}_1}{1+{K}_{NH NOOX}.{C}_{N{O}_2}} \)

R6

2NH3 − S1 + 2NO2 ⇾ N2 + N2O + 3H2O + 2S1

\( {r}_6={k}_6{y}_{NO_2}{\theta}_{NH_3S1}{\varOmega}_1 \)

$$ k=\mathrm{A}{e}^{\frac{- Ea}{RT}} $$
(10)

While it is known the NH3 does not store in significant amounts over PGM catalysts, this reaction was modeled explicitly, following the approach of [20]. The authors in [20] observed that exclusion of the NH3 adsorption/desorption reactions led to incorrect description in the high-temperature regime, where NH3 adsorption likely becomes rate controlling for the oxidation reactions. This conclusion will be revisited in the present work in the results section. NH3 adsorption was modeled as a non-activated process, while desorption followed Eq. (10), with constant desorption activation energy (reaction R2 in Table 1). In addition to these reactions, the mechanism included reversible NO oxidation (reaction R1 in Table 1), NH3 oxidation with selectivity to N2 and NO (reactions R3 and R4 in Table 1), along with an unselective SCR reaction in the presence of NO to form N2O (reaction R5 in Table 1). All these reactions were experimentally observed to be inhibited in the presence of NO2, and this effect was incorporated empirically with Langmuir-Hinshelwood inhibition functions. Finally, the model included a reaction between NH3 and NO2, leading to formation of N2O and N2 (reaction R6 in Table 1).

3.2.2 SCR-Coated ASC

The overall two-site reaction mechanism for the SCR-coated ASC was developed in a manner similar to [23], utilizing Eley-Rideal kinetics, and is shown in Table 2, where kf and kb refer to the forward and reverse rate constants respectively, described according to the Arrhenius Eq. (10). Species concentrations in the reaction rate were described in units of mol/m3, and the overall reaction rate was described on a turnover number basis, in units of mol/m3reactor-s.
Table 2

SCR reactions and rate expressions

Reaction no.

Reaction

Rate expression

R7

NH3 + S2 ⇿ NH3 − S2

\( {\mathrm{r}}_{7\mathrm{f}}={\mathrm{k}}_{7,\mathrm{f}}{\mathrm{y}}_{{\mathrm{NH}}_3}{\uptheta}_{\mathrm{S}2}{\Omega}_2 \)

\( {\mathrm{r}}_{7\mathrm{b}}={\mathrm{k}}_{7,\mathrm{b}}{\uptheta}_{{\mathrm{NH}}_3\mathrm{S}2}{\Omega}_2 \)

R8

NH3 + S3 ⇿ NH3 − S3

\( {\mathrm{r}}_{8\mathrm{f}}={\mathrm{k}}_{8,\mathrm{f}}{\mathrm{y}}_{{\mathrm{NH}}_3}{\uptheta}_{\mathrm{S}3}{\Omega}_3 \)

\( {\mathrm{r}}_{8\mathrm{b}}={\mathrm{k}}_{8,\mathrm{b}}{\uptheta}_{{\mathrm{NH}}_3\mathrm{S}3}{\Omega}_3 \)

R9

4NH3 − S2 + 3O2 ⇾ 2N2 + 6H2O + 4S2

\( {\mathrm{r}}_9={\mathrm{k}}_9{\mathrm{y}}_{{\mathrm{O}}_2}^{0.5}{\uptheta}_{{\mathrm{NH}}_3\mathrm{S}2}{\Omega}_2 \)

R10

2NH3 − S3 + 2O2 ⇾ N2O + 3H2O + 2S3

\( {\mathrm{r}}_{10}={\mathrm{k}}_{10}{\mathrm{y}}_{{\mathrm{O}}_2}^{0.5}{\uptheta}_{{\mathrm{NH}}_3\mathrm{S}3}{\Omega}_3 \)

R11

NO + 0.5O2 ⇿ NO2

\( {\mathrm{r}}_{11}={\mathrm{k}}_{11}\left({y}_{NO}{\mathrm{y}}_{{\mathrm{O}}_2}^{0.5}-\frac{y_{NO_2}}{K_{eq}}\right){\Omega}_2 \)

R12

4NH3 − S2 + 4NO + O2 ⇾ 4N2 + 6H2O + 4S2

\( {\mathrm{r}}_{12}={\mathrm{k}}_{12}{y}_{NO}{\mathrm{y}}_{{\mathrm{O}}_2}^{0.5}{\uptheta}_{{\mathrm{NH}}_3\mathrm{S}2}{\Omega}_2 \)

R13

4NH3 − S3 + 4NO + O2 ⇾ 4N2 + 6H2O + 4S3

\( {\mathrm{r}}_{13}={\mathrm{k}}_{13}{y}_{NO}{\mathrm{y}}_{{\mathrm{O}}_2}^{0.5}{\uptheta}_{{\mathrm{NH}}_3\mathrm{S}3}{\Omega}_3 \)

R14

2NH3 − S2 + NO + NO2 ⇾ 2N2 + 3H2O + 2S2

\( {\mathrm{r}}_{14}={\mathrm{k}}_{14}{y}_{NO}{y}_{NO_2}{\uptheta}_{{\mathrm{NH}}_3\mathrm{S}2}{\Omega}_2 \)

R15

2NH3 − S3 + NO + NO2 ⇾ 2N2 + 3H2O + 2S3

\( {\mathrm{r}}_{15}={\mathrm{k}}_{15}{y}_{NO}{y}_{NO_2}{\uptheta}_{{\mathrm{NH}}_3\mathrm{S}3}{\Omega}_3 \)

R16

8NH3 − S2 + 6NO2 ⇾ 7N2 + 12H2O + 8S2

\( {\mathrm{r}}_{16}={\mathrm{k}}_{16}{y}_{NO_2}{\uptheta}_{{\mathrm{NH}}_3\mathrm{S}2}{\Omega}_2 \)

R17

8NH3 − S3 + 6NO2 ⇾ 7N2 + 12H2O + 8S3

\( {\mathrm{r}}_{17}={\mathrm{k}}_{17}{y}_{NO_2}{\uptheta}_{{\mathrm{NH}}_3\mathrm{S}3}{\Omega}_3 \)

R18

2NH3 − S2 + 2NO2 ⇾ N2 + N2O + 3H2O + 2S2

\( {r}_{18}={k}_{18}{y}_{NO_2}{\theta}_{NH_3S2}{\varOmega}_2 \)

In the present model, desorption from site S2 was modeled using the Langmuir approach, while desorption from site S3 was modeled using the Temkin approach (R7-R8 in Table 2). Since the present kinetic mechanism adapted the Eley-Rideal kinetics approach, the adsorption and desorption of species apart from NH3 was assumed to be instantaneous, and therefore not rate limiting. Thus, the reversible NO oxidation to NO2 was modeled in the gas phase using the partial equilibrium approach (reaction R11 in Table 2) [29]. Standard, fast, and NO2 SCR reactions were modeled in a global manner on both sites, with different activation energies (reactions R12-R17 in Table 2). Finally, noticeable N2O formation was observed in the presence of excess NO2, and this was accounted for globally using reaction R18 in Table 2.

The calibration of the rate constants was done in a sequential manner, with specific isolated reactor data. The optimization algorithm used was primarily the well-known genetic algorithm NSGA-III [30], in combination with a gradient-based Nelder-Mead Simplex method [31] once the optimal domain was identified.

4 Results and Discussion

4.1 ASC Model Development

4.1.1 PGM-Coated ASC Model

As mentioned in the experimental section, the PGM-coated ASC model was developed using steady-state isothermal experiments of varying feed gas concentrations between 200 and 550 °C. An example of the inlet feed gas sequence is shown in Fig. 1 for D-70 at 250 °C.
Fig. 1

PGM-Coated ASC reactor isothermal experiment: D-70 at 250 °C; 10% O2 and 4.5% H2O used in all steps

All model calibration was performed using data for degreened and aged catalysts at GHSVs of 70,000 and 140,000/h. The reaction rate constants were kept identical for both hydrothermal aging conditions, and only the PGM dispersion factors were modified. The low PGM loading made it difficult to measure dispersion factors via standard CO chemisorption tests. Since no useful measurements were available, the dispersion factor of the degreened ASC was set to 1, and the aged ASC dispersion factor was calibrated to the outlet measured NO2/NOx ratio during the NO oxidation test with no inlet NO2. The results of these fits are shown in Fig. 2. As expected, the rate of NO oxidation decreased upon hydrothermal aging, and this was captured by calibrating the Arrhenius rate constants of the reversible NO oxidation reaction (R1 in Table 1), along with the aged ASC dispersion factor. In the present context, this factor can be thought of as a relative dispersion factor to express the degree of hydrothermal aging, rather than an absolute value representing the fraction of total PGM sites exposed to the channel gas.
Fig. 2

Steady-state NO oxidation results from 250 to 550 °C

Following NO oxidation, the rate constants of the remaining reactions (R2-R6 in Table 1) were calibrated in a sequential manner. NH3/O2 interactions were captured using reactions R2-R4, while N2O formation in the absence of NO2 was captured indirectly as a combination of reactions R4 and R5. Reactions R2-R5 also covered all the NH3/NO/O2 interactions, and each of these reactions was inhibited in the presence of NO2. Finally, NH3/NO2/O2 interactions were captured using R6, with fixed NH3 storage rate constants. The results for GHSV 70,000/h are shown in the remaining portion of this section, and the results for GHSV 140,000/h can be found in the Appendix.

The steady-state conversion results for the degreened catalyst in Fig. 3 show that in the absence of NOx, NH3 oxidation began to light-off around 200 °C, with 100% conversion efficiency by 250 °C. The product selectivity changed with temperature. Peaks in N2 and N2O yields were seen at 250 °C, with increased NOx formation at higher temperatures. With the introduction of NO in the feed gas, NH3 light-off was shifted to a lower temperature, with increased conversion seen at 200 °C. Most of this converted NH3 at 200 °C formed N2O, with a noticeable decrease in peak N2 yield at 250 °C. Increased NOx yield was also observed in the entire temperature range, attributed to incomplete conversion of inlet NO. Replacing NO with NO2 in the feed gas led to delayed NH3 light-off, with complete conversion observed only around 350 °C (note that no data was collected between 250 and 350 °C). This was indicative of NO2 inhibition of NH3 oxidation below 350 °C. Both N2 and N2O yields were reduced in the presence of NO2, with increased low temperature NOx yields. The peak in N2O yield was also shifted from 250 to 350 °C. Most of these conclusions were in line with observations from authors in [20], and justified the utilization of a similar reaction mechanism structure.
Fig. 3

PGM-coated ASC D-70 steady-state NH3 conversion and product yield results for NH3-O2 feed, NH3-NO-O2 feed, and NH3-NO2-O2 feed between 200 and 550 °C

The impact on hydrothermal aging on ASC NH3 conversion and product selectivity has been reported previously in literature [7], with the major impact being the reduction of N2 selectivity at similar overall NH3 conversions. In the present work, Fig. 4 shows that hydrothermal aging in the absence of NOx led to decreased N2 yield at 200 °C and from 350 to 550 °C, with increased peak N2 yield at 250 °C. N2O yield was also reduced, with a minor increase in NOx yield beyond 250 °C and similar overall NH3 conversion in the entire temperature range. The increased peak N2 yield could be associated with reduced NO2 inhibition due to a reduction in NO formation an oxidation to NO2. With the introduction of NO, the aging trends with respect to N2 and N2O yields did not change (i.e., similar trends as NH3-O2 feed). A noticeable increase in NOx yield was observed at 200 °C, which may be attributed to the overall reduction in NH3 conversion at 200 °C, leading to delayed NH3 light-off. Finally, hydrothermal aging in the presence of NH3 and NO2 led to shift in peak N2 yield to 350 °C, with reduced N2 yield below 350 °C and increased N2 yield from 350 to 450 °C. Minor decrease in peak N2O yield was observed at 350 °C, with similar NOx yield and reduced NH3 conversion at 250 °C (delayed NH3 light-off). The increase in N2 yield from 350 to 450 °C upon hydrothermal aging could be attributed to increased resistance to NO2 poisoning of the Pt surface due to presence of larger agglomerated Pt particles, leading to higher site-specific activity, compensating for the loss of exposed Pt surface area. However, this needs to be confirmed with further dedicated experiments.
Fig. 4

PGM-coated ASC A-70 steady-state NH3 conversion and product yield results for NH3-O2 feed, NH3-NO-O2 feed, and NH3-NO2-O2 feed between 200 and 550 °C

Kinetic fits for D-70 and A-70 are shown in Figs. 3 and 4, respectively. Considering the range of complex interactions occurring with different feed gas compositions, and the modification of these interactions upon hydrothermal aging, the relatively simple global kinetic model captured all the trends reasonably well. More specifically, the kinetic model successfully accounted for the NO2 inhibition of NH3 conversion, NO enhancement of NH3 conversion, and the trends in product yields as a function of temperature. Furthermore, the impact of hydrothermal aging was captured using solely the aged dispersion factor, with the same reaction kinetics.

As mentioned in the kinetic model description, NH3 adsorption and desorption reactions on PGM were included based on observations of authors in [20] that NH3 adsorption became rate controlling at higher temperatures. The reaction rates in the presently developed kinetic model were analyzed to verify this conclusion. At high temperatures (450 and 550 °C), it was observed that NH3 adsorption and desorption rates were higher than the NO formation rates by a factor of ~ 2. This means that the rate of adsorption and desorption is normally always fast and never the rate determining step. Therefore, it should be possible to have a global kinetic mechanism without the inclusion of NH3 adsorption/desorption reactions and still achieve reasonable modeling accuracy.

The first stage validation of the developed reaction scheme in Table 1 with calibrated rate constants was performed with TPR experiments conducted in the presence of 150 ppm NH3, 150 ppm NO, 10% O2, and 4.5% H2O in Ar from 200 to 550 °C at a ramp rate of 10 °C/min. NH3 was consumed completely early in the experiment, and the outlet concentration did not change from 250 to 550 °C, as expected. The model prediction results for these experiments, shown in Figs. 5 and 6, clearly demonstrate the capability of the model to capture transient dynamics associated with product selectivity for NH3 oxidation, for different temperatures and hydrothermal aging conditions.
Fig. 5

D-70 transient NO, NO2, N2O, and N2 concentrations during temperature ramp from 200 to 550 °C for model validation with nominal inlet NH3 = 150 ppm and nominal inlet NO = 150 ppm in 10% O2, 4.5% H2O, and balance Ar

Fig. 6

A-70 transient NO, NO2, N2O, and N2 concentrations during temperature ramp from 200 to 550 °C for model validation with nominal inlet NH3 = 150 ppm and nominal inlet NO = 150 ppm in 10% O2, 4.5% H2O, and balance Ar

4.1.2 SCR-Coated ASC Model

The SCR-coated ASC model was developed using steady-state isothermal experiments of varying feed gas concentrations between 200 and 550 °C. An example of the inlet feed gas sequence is shown in Fig. 7 for D-70 at 250 °C.
Fig. 7

SCR-coated ASC reactor isothermal experiment: D-70 at 250 °C; 10% O2 and 4.5% H2O used in all steps

The two-site kinetic model was calibrated using data for degreened and aged catalysts at GHSVs of 70,000 and 140,000/h. The reaction rate constants were kept identical for both hydrothermal aging conditions, and only the active site densities were modified.

The active site densities were first calibrated along with the storage and NH3 oxidation rate constants (reactions R7-R10 in Table 2), utilizing the isothermal NH3 oxidation data between 200 and 550 °C with 0.2 and 10% O2. Following this, the rate constants of the remaining reactions (R11-R18 in Table 2) were calibrated in a sequential manner.

Figures 8 and 9 show the experimental and simulation results for D-70 and A-70 respectively, with varying feed gas mixture compositions between 200 and 550 °C. For the degreened catalyst, NH3 oxidation in the absence of NOx led to almost 100% selectivity to N2, with minor amounts of NOx and N2O formed at 550 °C. The SCR reactions followed the expected trends, with maximum low temperature NOx conversion (shown as equivalent N2 yield) occurring for fast SCR, along with a decrease in conversion at high temperatures due to unselective NH3 oxidation. N2O yield increased with increase in inlet NO2/NOx ratio and showed a bi-modal behavior, associated with low temperature nitrate decomposition and high temperature NH3 oxidation.
Fig. 8

SCR-coated ASC D-70 steady-state NH3 conversion and product yield results for NH3-O2 feed, NH3-NO-O2 feed, and NH3-NO2-O2 feed between 200 and 550 °C

Fig. 9

SCR-coated ASC A-70 steady-state NH3 conversion and product yield results for NH3-O2 feed, NH3-NO-O2 feed, and NH3-NO2-O2 feed between 200 and 550 °C

Hydrothermal aging led to significant reduction in NH3 conversion and NOx conversion in the entire temperature range. Despite the reduction in parasitic NH3 oxidation, reduction in active sites led to significantly reduced NOx conversion, accounting for the high temperature behavior. Low temperature N2O slip also reduced, with an increase in high-temperature N2O and NOx yield, indicating relatively poor selectivity for NH3 oxidation.

The two-site SCR-coated ASC model successfully described species conversion behavior for both hydrothermal aging conditions, as seen in Figs. 8 and 9 respectively. Minor deviations were observed in high-temperature N2 yield for the degreened catalyst, due to underprediction of unselective NH3 oxidation. The reduction in NH3 conversion and N2 yield upon hydrothermal aging was predicted accurately by using the same kinetics with reduced active site densities, justifying this modeling approach. High-temperature conversion errors were in part due to the assumption of constant species effective diffusivities through the SCR layer. This was fixed using the dual-layer data and is explained in Sect. 4.2. Finally, low-temperature N2O formation was only modeled in the presence of excess N2O, leading to minor underprediction for standard and fast SCR reactions. However, a 3% yield error corresponds to an error of 4.5 ppm, which lies within the limits of measurement discrepancies. N2O slip beyond 350 °C was generally underpredicted for the aged catalyst and requires inclusion of an additional site with increased site density upon hydrothermal aging. However, the overall calculated RMS errors from the two-site model were considered sufficiently low for the present application.

4.2 ASC Model Validation—Reactor Scale Dual-Layer ASC

The integrated set of SCR and PGM kinetics along with the dual-layer intra-porous diffusion model represented the final model. Isothermal dual-layer ASC reactor data between 250 and 450 °C (range of temperatures where intra-porous diffusion was expected to be most significant) was utilized to calibrate the porosity/tortuosity ratios of both layers in the parallel pore diffusion model, with the kinetics from the single-layer models. Figure 10 shows the NOx yield prediction comparisons between the constant effective diffusivity (Deff = 5e−6 m2/s for all species in both layers) model and the parameterized parallel pore diffusion model. The noticeable increase in NOx yield as a function of temperature for the dual-layer ASC could be described accurately by the parallel pore diffusion model. This increase in simulated NOx yield was associated with increased effective diffusivities for NH3 through the SCR layer as a function of temperature, leading to higher conversion, along with formation of NO in the PGM layer. Similar trends were seen for other species. The variation between the two simulation results with the same set of kinetics further emphasizes the significance of pore diffusion in dual-layer catalysts. The importance of predicting NOx slip at high temperatures will be further highlighted in the transient model validation section (Sect. 4.3.3).
Fig. 10

Dual-layer ASC steady-state NOx yield results for NH3-O2 feed and NH3-NO-O2 feed between 250 and 450 °C

The final dual-layer model was initially validated with transient reactor data for a commercial dual-layer ASC. The reactor testing primarily consisted of isothermal experiments at three temperatures between 250 and 150 °C in the presence of 150 ppm NH3, 150 ppm NO, and 10% O2, followed by a transient ramp-up to 550 °C at 10 °C/min. Figure 11 shows the TPR results at both hydrothermal aging conditions.
Fig. 11

Dual-layer ASC transient NH3 and NO concentrations for model validation under NH3-NO-O2 feed conditions

For the degreened catalyst, low-temperature NH3 and NO conversions were identical. During the ramp-up phase, the NH3 outlet concentration decreased to zero, while the overall rate of NO conversion decreased due to formation of additional NO in the PGM layer. Hydrothermal aging led to increased NO formation, such that the outlet NO concentration exceeded the inlet value around 400 °C.

The developed kinetic model predicted these key trends accurately. However, NH3 conversion was over-predicted at 200 °C and requires further investigation. The peak in N2O slip during the ramp-up phase of the experiment was also captured correctly but is not shown here for brevity.

4.3 Full-Scale SCR and ASC Model Validation

The developed SCR and dual-layer ASC models were validated using steady-state and transient engine dynamometer tests. Detailed steady-state validation of the SCR model will be discussed first, followed by the transient validation of the SCR and ASC models. It is important to note that the present validation only applies to the aged catalysts, and further testing must be performed to validate the degreened SCR and ASC models.

All SCR and ASC validation simulations were performed utilizing full-scale models of the SCR and dual-layer ASC catalysts in GT-SUITE v2017. The full-scale model also included detailed modeling of the outer insulation layers for accurate temperature predictions. For model validation, a key assumption was that all the injected UWS converted to NH3. This is not always true, and more accurate UWS-NH3 conversion maps will be obtained from validated CFD in the future.

4.3.1 Steady-State SCR Model Validation

Steady-state SCR model validation was performed with inlet ANR sweeps for 30 different engine operating conditions. For each point at a fixed engine-out condition, UWS was injected in a stepped manner starting from no dosing until noticeable NH3 slip was observable at an inlet ANR greater than 1. SCR-outlet NH3, NO, NO2, and N2O concentrations were monitored using an FTIR. The initial FTIR data with no dosing was used to calculate the inlet NO2/NOx ratio, and the total inlet NOx was obtained from the CLD NOx analyzer. Finally, a mid-bed thermocouple was used to confirm near isothermal reactor operation at steady-state conditions. Figure S1 in the supplementary material shows the range of temperatures, GHSVs, and inlet NO2/NOx ratios for the steady-state points.

Each steady-state inlet ANR sweep was simulated transiently. The downstream normalized NO, NO2, NH3, and N2O concentration profiles for all the ANR sweeps can be found in Figures S2-S31 in the supplementary material. As the inlet ANR was increased, the NO and NO2 outlet concentrations gradually decreased as expected, with an increase in N2O slip. In general, the model successfully followed these trends. The most difficult parameter to predict was the critical inlet ANR beyond which NH3 started slipping, and this was compounded by the assumptions of 100% UWS-NH3 conversion, along with uniform inlet distribution of NH3, velocity, and temperature. N2O slip was always below 30 ppm, and the errors in model prediction were always less than 10 ppm.

The absolute error in NOx conversion as a function of temperature and GHSV at an inlet ANR ~ 1 is shown in Fig. 12. Overall, the average error in NOx conversion for 30 steady-state points was 4.6%. This is in comparison to an error of 3.9% over 123 different reactor operating points used for model development [23]. The reason for relatively high error (> 8%) at three operating points could be associated with uncertainty in experimental data, along with some of the modeling assumptions. Further improvement might be possible with modification of kinetic rate constants, but this was beyond the scope of the present work.
Fig. 12

NOx conversion error as a function of inlet NO2/NOx ratio and SCR temperature at an inlet ANR~1

4.3.2 Transient SCR Model Validation

The developed SCR model was further validated via drive cycle simulations of the cold and hot heavy-duty transient (HDT) cycles. Prior to the cold HDT cycle, the pre-conditioning HDT cycle and overnight soak were also simulated, to obtain the appropriate NH3 pre-storage and axial distribution before the start of the cold HDT cycle. Figure 13 shows the NH3 coverage on both sites at the start of the cold HDT cycle, indicating that most of the pre-stored NH3 was near the front of the catalyst. Minor scaling corrections were applied to these coverages to begin the cold HDT cycle with the same NH3 storage in grams as indicated by the experimentally calibrated mass-balance model in the dosing control unit (DCU). Finally, a mid-bed thermocouple was used to confirm the thermal model predictions. The results for the cold HDT cycle are shown in the remaining portion of this section, while similar results for the hot HDT cycle can be found in the Appendix.
Fig. 13

NH3 storage axial distribution on S1 and S2 sites at the beginning of the cold HDT cycle

The SCR mid-bed temperature and GHSV during the cold HDT cycle are shown in Fig 14, demonstrating the temperature and flow range of operation for this SCR catalyst. The measured mid-bed temperature showed an exotherm in the initial warm-up phase, associated with adsorption of H2O on the surface. Accurate mid-bed temperature predictions are required to obtain the correct local reaction rates that govern the outlet species concentrations. In this case, the SCR mid-bed temperatures were estimated correctly beyond the first 200 s of the cold HDT cycle. This highlights the need for including H2O storage as part of the global kinetic model, and this sub-model can be added in the future.
Fig. 14

Cold HDT cycle SCR mid-bed temperature and GHSV

SCR outlet NO, NO2, and N2O concentrations were compared with the FTIR measurements, and the normalized traces are shown in Fig. 15. The experimental outlet NO concentration demonstrated an initial dead time, associated with low temperature storage, followed by a large peak. Prediction of this specific trend in the first 100 s requires explicit modeling of NO storage. Nevertheless, the model was successful in following the NO trace beyond the first 100 s, with all the peak locations and magnitudes predicted accurately. Negligible NO2 slip was observed, and any value beyond the measurement error threshold was predicted by the model. The trend in N2O slip was also captured by the model, associated with the nitrate formation and decomposition reactions. No NH3 measurements were available due to measurement issues with the FTIR. However, NH3 slip was indirectly inferred from the NH3 storage results shown in Fig. 16. The bottom left quadrant of this figure shows the normalized NH3 storage comparison between the calibrated DCU model prediction and the quasi-1 + 1D model. Around 720 s into the cold HDT cycle, both models demonstrated a sudden drop in stored NH3, associated with NH3 slip. This slip was also confirmed by simple mass balance calculations. However, the magnitude of the predicted peak slip requires further experimental validation.
Fig. 15

Cold HDT cycle normalized instantaneous species concentrations at SCR outlet

Fig. 16

Cold HDT cycle normalized cumulative species mass and NH3 storage

Figure 16 also shows the integrated normalized NO, NO2, and N2O masses at the inlet and outlet of the SCR catalyst. Mass-based NOx conversion was predicted within 3% for both the cold and hot HDT (shown in Figure A6), highlighting the predictive capability of the kinetic model. N2O slip was also predicted accurately with minor underestimation in outlet integrated mass.

Finally, Fig. 17 shows the simulated coverage of NH3 and NH4NO3 on both sites during the cold HDT cycle. NH4NO3 was estimated to be present at the start of the cycle due to the NH3 pre-conditioning and the low temperatures and evolved slowly throughout. The predicted NH4NO3 surface coverage was three orders of magnitude smaller than the NH3 surface coverage. These results have not been reported previously and provide quantitative relative comparisons between these two-coverage species.
Fig. 17

Cold HDT cycle simulated NH3 and NH4NO3 coverages on S1 and S2 sites

4.3.3 Transient ASC Model Validation

The validated SCR model was combined with the dual-layer ASC model and simulated over the same cold and hot HDT cycles. ASC outlet NOx measurements were compared with the simulation results and served as a transient validation for the dual-layer ASC model. The SCR-out and ASC-out simulated NOx values were almost identical for the first 700 s, due to minimal NH3 slip out of the SCR. Beyond 700 s, an increase in measured ASC-outlet NOx was observed, associated with unselective conversion of NH3 across the ASC. This was predicted quantitatively by the dual-layer ASC model, as shown in Fig. 18. This increase in NOx concentration across the ASC impacts the overall system NOx conversion and confirms the necessity for developing dual-layer ASC models to accurately predict this key parameter.
Fig. 18

Cold and hot HDT cycle normalized instantaneous NOx concentrations from 600 to 1200 s

4.4 Application of Validated SCR and ASC Models

4.4.1 Dynamic and Total Capacities during Standard and Fast SCR

The validated model of the Cu-SSZ-13 SCR catalyst was probed to obtain intra-catalyst mechanistic information that is difficult to measure with standard instrumentation. Understanding of NH3 capacity distribution and utilization provides useful insights that can be utilized for real-time catalyst control. In this regard, the concepts of dynamic capacity (DC), total capacity (TC), and unused capacity (UC) were first introduced along with the four-step SCR protocol in 2010 [32]. The total capacity refers to the NH3 storage capacity in the absence of NOx, while the unused capacity refers to the NH3 storage capacity during SCR. The difference between these two gives the dynamic capacity, which is utilized during SCR reaction.

Spatially resolved intra-catalyst measurements have enabled calculation of the axial distribution of these three capacities, and some of the recent work in this area is summarized in [33]. It has also been shown in [33] that the impact of real-world field aging is simply to reduce the number of active sites, with similar adsorption energetics. Thus, the information related to total capacity can be scaled with data from different aging conditions, enabling feedback adaptive control to maximize catalyst and NH3 utilization.

In the present work, the four-step SCR protocol was simulated at different temperatures, inlet NO2/NOx ratios, and hydrothermal aging conditions, and the conclusions were compared with those obtained from the above-mentioned literature. The simulations were performed until steady state at a GHSV of 60,000/h with catalysts representative of the reactor hydrothermal aging conditions used to develop the kinetics (i.e., degreening at 650 °C for 16 h in 10% H2O/air and aging at 700 °C for 100 h in 10% H2O/air). The inlet feed gas consisted of a combination of 300 ppm NH3 and 300 ppm NOx in 7% H2O, 8% CO2 and 10% O2.

Axially integrated capacities along with NH3 coverage on S1 are shown at 150 and 250 °C in Figs. 19 and 20, respectively. At 150 °C, the standard SCR reaction is below light-off conversion, and the entire catalyst is active for NOx conversion, leading to DC ≈ TC along the reactor length. In this case, the NH3 coverage slope is almost flat, following the Langmuir isotherm. Fast SCR light-off occurs earlier than standard SCR, and this is confirmed by the reduction in the SCR front zone (defined as the zone where DC ≈ TC). In the second half of the catalyst, the integrated unused capacity (UC) increases, and there is a sudden increase in the coverage slope. The total capacity (TC) and dynamic capacity (DC) decrease upon hydrothermal aging. Note that at such a low temperature, AN interferes with the TC and DC calculations under fast SCR conditions.
Fig. 19

Axially integrated total capacity (TC), dynamic capacity (DC), and NH3 coverage on S1 for standard and fast SCR reactions at 150 °C

Fig. 20

Axially integrated total capacity (TC), dynamic capacity (DC), and NH3 coverage on S1 for standard and fast SCR reactions at 250 °C

At 250 °C, a significant increase in reaction rate relative to 150 °C is observed, leading to a shift in the SCR zone to the catalyst front. There is a plateauing of the dynamic capacity (DC) near the rear of the catalyst due to reduction in locally available gas-phase NH3, leading to a steep decrease in local NH3 coverage. Hydrothermal aging leads to a decrease in dynamic capacity at 250 °C for both SCR reactions. However, the dynamic capacity slope near the rear of the catalyst is higher, implying higher NH3 utilization. This is attributed to more even NH3 storage upon hydrothermal aging, as shown by the total capacity (TC) in Fig. 20. This consequently results in slightly higher NOx conversion for the aged catalyst, despite the reduced overall dynamic capacity. The NOx concentration along the length of the SCR catalyst is shown in Fig. 21, confirming that hydrothermal aging leads to reduced conversion near the front of the catalyst and increased conversion near the rear of the catalyst. These conclusions hold true for both standard and fast SCR reactions.
Fig. 21

Axial NOx concentration at 250 °C under standard SCR and fast SCR conditions

The cause for more axially uniform NH3 storage upon hydrothermal aging is hypothesized to be related to the distribution of Z2Cu and ZCuOH active sites along the length of this Cu-SSZ-13 SCR catalyst. Since hydrothermal aging leads to an increase in the number of Z2Cu sites, it implies that Z2Cu sites are spread more uniformly relative to ZCuOH sites. This requires further experimental verification. Nevertheless, the above figures confirm that in addition to the overall dynamic capacity, the local dynamic capacity distribution along the reactor length (i.e., DC (z)) is an important parameter governing overall NOx conversion.

4.4.2 Apparent Negative Activation Energy for High-Temperature SCR

It is widely known that beyond 400 °C, the NOx conversion of most SCR catalysts decreases as a function of temperature. This reduction in NOx conversion at high temperatures has been attributed to the parasitic NH3 oxidation reaction which competes with the SCR reactions [6]. However, it has been shown recently that the reduction in NOx conversion is primarily due to reduction in the rate of the SCR reaction itself, due to the reduced NH3 surface coverage [34]. This means that under most realistic conditions, the SCR reaction has kinetic limitations, reducing the assumed significance of internal and external mass transfer. This behavior was confirmed with the present kinetic model of Cu-SSZ-13 by simulating standard and fast SCR conditions at GHSV 60,000/h and an inlet ANR of 1 and plotting the reaction rates on S2 as shown in Fig. 22. For the present catalyst, the SCR rates reach a maximum between 200 and 250 °C and continuously decrease beyond this temperature due to the strong dependence on NH3 coverage.
Fig. 22

Standard and fast SCR reaction rates on S2 site

4.4.3 Intra-Layer Washcoat Gradients in Dual-Layer ASC

As mentioned in the transient ASC model validation section (Sect. 4.3.3), noticeable NO slip was observed during the later portions of the cold and hot HDT cycles, caused by NH3 slip from the upstream SCR catalyst. This NO slip was successfully replicated by the model, confirming its validity. The conditions encountered during this NO slip were analyzed using the steady-state reactor model, and the resulting intra-layer NO concentration in the washcoat is shown in Fig. 23. The simulations were performed until steady-state at a GHSV of 140,000/h and an ASC temperature of 350 °C with catalysts representative of the reactor hydrothermal aging conditions used to develop the kinetics (i.e., degreening at 650 °C for 16 h in 10% H2O/air and aging at 700 °C for 100 h in 10% H2O/air). The inlet feed gas consisted of 150 ppm NH3 4.5% H2O and 10% O2.
Fig. 23

ASC intra-layer NO concentration at 350 °C with 150 ppm inlet NH3 in 4.5% H2O and 10% O2

Figure 23 shows that at 350 °C, NH3 diffuses through the SCR layer to reach the PGM layer, where it oxidizes unselectively to form NO. The formed NO then diffuses through the PGM and SCR layers, reacting with stored NH3 in the SCR layer. Over the length of the catalyst, the amount of NO formed increases cumulatively and slips out of the catalyst. Hydrothermal aging leads to increased NO formation in the PGM layer and reduced SCR layer NO conversion, resulting in an overall increase in NO slip. This slip could be reduced with an increase in activity and hydrothermal durability of the SCR layer catalyst.

5 Conclusions

The work presented here concludes the efforts in the development and validation of a global kinetic model to describe the NH3-SCR behavior over Cu-SSZ-13. Building on the SCR model developed in part I, this part focuses on development of a mathematical model for the downstream ASC, full-scale validation of both the SCR and ASC models, and their application to calculate NH3 utilization via dynamic capacities, along with analysis of ASC intra-layer washcoat concentrations.

A reaction-diffusion model of the commercial dual-layer ASC was developed to accurately estimate tailpipe NO, NO2, NH3, and N2O concentrations. The approach to develop this model was split into three steps. Firstly, kinetic models of the PGM-coated and SCR-coated ASCs were developed utilizing reactor data for single-layer-coated monoliths. The mechanism for these individual layers utilized Eley-Rideal-type kinetics, and the rate constants were parameterized to reactor data for these catalysts. Successful model predictions were obtained for each of these catalysts. The PGM-layer and SCR-layer kinetics were then integrated along with a parametrized parallel pore diffusion model for each layer. The resulting dual-layer ASC model was validated with transient reactor data of a commercial dual layer ASC. Excellent model predictions were obtained for the dual-layer catalyst without any kinetic modifications, with an average error of 3.4% in NH3 conversion, 6.6% in NOx formation, and 7.9% in N2O formation across 32 different operating points covering the range of temperatures, space velocities, aging conditions, and inlet NO2/NOx ratios seen in real-world driving conditions.

The aged versions of the developed SCR and ASC reactor models were then validated with a full-scale application on an engine dynamometer. Steady-state SCR model validation involved inlet ANR sweep tests at 30 different engine operating points covering the range of temperature, GHSV, and inlet NO2/NOx ratios seen by this SCR catalyst in real-world applications. The individual plots for each of these points can be found in the supplementary material, showing that the model was able to capture the overall trends in species concentrations, including NH3 and N2O slip. In general, the model over predicted NOx conversion, with an average error of 4.6%.

Transient validation of both the SCR and ASC models was performed via cold and hot heavy-duty transient (HDT) cycles, with comparisons of instantaneous mole and cumulative mass concentrations of all the important trace species. The SCR model showed very high accuracy in NOx conversions, with a maximum error in cumulative NOx conversion estimation of 3%. This model also quantified the relative magnitudes of NH3 and NH4NO3 coverages during a cold HDT cycle, demonstrating that the NH4NO3 coverage is three orders smaller than the NH3 coverage in this certification cycle. Improvements in SCR model development were identified as inclusion of H2O and NO storage sub-models and can be incorporated in the future. The ASC model was also validated on the same transient cycles by demonstrating its ability to predict minor NOx formation when NH3 slipped from the upstream SCR catalyst.

Finally, the validated SCR and ASC models were utilized to elucidate intra-catalyst functional understanding. More specifically, intra-catalyst dynamic and total capacities were calculated for the SCR catalyst by simulating the four-step protocol [32], revealing information related to catalyst and NH3 utilization under different temperatures, inlet NO2/NOx ratios, and hydrothermal aging conditions. A new finding from these numerical experiments was the increase in NOx conversion at 250 °C upon hydrothermal aging, despite the reduction in overall dynamic capacity. This was attributed to more uniform NH3 storage upon hydrothermal aging, likely related to distributions of the two chemically distinct Cu monomer active sites, leading to increased NH3 utilization near the rear of the aged catalyst.

Analysis of washcoat NO gradients in dual-layer ASC model under conditions encountered during transient HDT cycles provided useful and unmeasurable information related to the cause of NO slip in the presence of excess NH3. These insights can be used to improve catalyst design and control.

The comprehensive approach presented in both parts of the paper attempts to consolidate the widespread applicability of these models in design and control. It must be stressed that if the appropriate modeling approach is followed to develop these global kinetic models, and fit them to reactor data from detailed test protocols, they can essentially replace the physical hardware for all development purposes and be embedded in the dosing control unit for real-time state estimations in closed-loop feedback control. To accomplish this ambitious task, it is also necessary to implement an aging model that accounts for continuous deterioration due to hydrothermal aging and for the reversible and irreversible effects of chemical poisoning due to sulfur and hydrocarbons.

Notes

Acknowledgments

The authors would like to acknowledge Cormetech Inc. for executing the test protocol, supplying the reactor data, and assisting with the reactor setup description. Furthermore, the Gamma Technologies Aftertreatment support team helped us with useful discussions and continuous assistance with modeling work.

Supplementary material

40825_2018_94_MOESM1_ESM.docx (2.5 mb)
ESM1 (DOCX 2596 kb)

References

  1. 1.
    Mondt, J.R.: Cleaner Cars: the History and Technology of Emission Control since the 1960s. (2000)Google Scholar
  2. 2.
    Johnson, T., Joshi, A.: Review of vehicle engine efficiency and emissions (No. 2017–01-0907). SAE Technical Paper. (2017)Google Scholar
  3. 3.
    Sharp, C., Webb, C.C., Yoon, S., Carter, M., Henry, C.: Achieving ultra low NOX emissions levels with a 2017 heavy-duty on-highway TC diesel engine-comparison of advanced technology approaches. SAE Int. J. Engines. 10(2017-01-0956) (2017)Google Scholar
  4. 4.
    Koebel, M., Madia, G., Elsener, M.: Selective catalytic reduction of NO and NO2 at low temperatures. Catal. Today. 73(3), 239–247 (2002)CrossRefGoogle Scholar
  5. 5.
    Yim, S.D., Kim, S.J., Baik, J.H., Nam, I.S., Mok, Y.S., Lee, J.H., et al.: Decomposition of urea into NH3 for the SCR process. Ind. Eng. Chem. Res. 43(16), 4856–4863 (2004)CrossRefGoogle Scholar
  6. 6.
    Nova, I., Tronconi, E. (eds.): Urea-SCR Technology for deNOx after Treatment of Diesel Exhausts. Springer, New York (2014)Google Scholar
  7. 7.
    Kamasamudram, K., Yezerets, A., Currier, N., Castagnola, M., Chen, H.Y.: New insights into reaction mechanism of selective catalytic ammonia oxidation technology for diesel aftertreatment applications. SAE Int. J. Engines. 4(2011–01-1314), 1810–1821 (2011)CrossRefGoogle Scholar
  8. 8.
    Chi, J.N.: Control challenges for optimal NOx conversion efficiency from SCR aftertreatment systems (No. 2009–01-0905). SAE Technical Paper. (2009)Google Scholar
  9. 9.
    Jabłońska, M., Palkovits, R.: Copper based catalysts for the selective ammonia oxidation into nitrogen and water vapour—recent trends and open challenges. Appl. Catal. B Environ. 181, 332–351 (2016)CrossRefGoogle Scholar
  10. 10.
    Park, S.J., Jin, S.Y.: Effect of ozone treatment on ammonia removal of activated carbons. J. Colloid Interface Sci. 286(1), 417–419 (2005)CrossRefGoogle Scholar
  11. 11.
    Girard, J.W., Cavataio, G., Lambert, C.K.: The influence of ammonia slip catalysts on ammonia, N2O and NOx emissions for diesel engines (No. 2007–01-1572). SAE Technical Paper. (2007)Google Scholar
  12. 12.
    Colombo, M., Nova, I., Tronconi, E.: A simplified approach to modeling of dual-layer ammonia slip catalysts. Chem. Eng. Sci. 75, 75–83 (2012)CrossRefGoogle Scholar
  13. 13.
    Scheuer, A., Votsmeier, M., Schuler, A., Gieshoff, J., Drochner, A., Vogel, H.: NH3-slip catalysts: experiments versus mechanistic modelling. Top. Catal. 52(13–20), 1847–1851 (2009)CrossRefGoogle Scholar
  14. 14.
    Scheuer, A., Hauptmann, W., Drochner, A., Gieshoff, J., Vogel, H., Votsmeier, M.: Dual layer automotive ammonia oxidation catalysts: experiments and computer simulation. Appl. Catal. B Environ. 111, 445–455 (2012)CrossRefGoogle Scholar
  15. 15.
    Sukumar, B., Dai, J., Johansson, A., Markatou, P., Ahmadinejad, M., Watling, T., ..., Szailer, T.: Modeling of dual layer ammonia slip catalysts (ASC) (No. 2012–01-1294). SAE Technical Paper. (2012)Google Scholar
  16. 16.
    Colombo, M., Nova, I., Tronconi, E., Schmeißer, V., Bandl-Konrad, B., Zimmermann, L.: NO/NO 2/N 2 O–NH 3 SCR reactions over a commercial Fe-zeolite catalyst for diesel exhaust aftertreatment: intrinsic kinetics and monolith converter modelling. Appl. Catal. B Environ. 111, 106–118 (2012)CrossRefGoogle Scholar
  17. 17.
    Shrestha, S., Harold, M.P., Kamasamudram, K.: Experimental and modeling study of selective ammonia oxidation on multi-functional washcoated monolith catalysts. Chem. Eng. J. 278, 24–35 (2015)CrossRefGoogle Scholar
  18. 18.
    Shrestha, S., Harold, M.P., Kamasamudram, K., Kumar, A., Olsson, L., Leistner, K.: Selective oxidation of ammonia to nitrogen on bi-functional Cu–SSZ-13 and Pt/Al 2 O 3 monolith catalyst. Catal. Today. 267, 130–144 (2016)CrossRefGoogle Scholar
  19. 19.
    Gao, F., Kwak, J.H., Szanyi, J., Peden, C.H.: Current understanding of cu-exchanged Chabazite molecular sieves for use as commercial diesel engine DeNOx catalysts. Top. Catal. 56(15–17), 1441–1459 (2013)CrossRefGoogle Scholar
  20. 20.
    Colombo, M., Nova, I., Tronconi, E., Schmeißer, V., Bandl-Konrad, B., Zimmermann, L.: Experimental and modeling study of a dual-layer (SCR+ PGM) NH 3 slip monolith catalyst (ASC) for automotive SCR aftertreatment systems. Part 1. Kinetics for the PGM component and analysis of SCR/PGM interactions. Appl. Catal. B Environ. 142, 861–876 (2013)CrossRefGoogle Scholar
  21. 21.
    Colombo, M., Nova, I., Tronconi, E., Schmeißer, V., Bandl-Konrad, B., Zimmermann, L.R.: Experimental and modeling study of a dual-layer (SCR+ PGM) NH 3 slip monolith catalyst (ASC) for automotive SCR after treatment systems. Part 2. Validation of PGM kinetics and modeling of the dual-layer ASC monolith. Appl. Catal. B Environ. 142, 337–343 (2013)CrossRefGoogle Scholar
  22. 22.
    Scheuer, A., Drochner, A., Gieshoff, J., Vogel, H., Votsmeier, M.: Runtime efficient simulation of monolith catalysts with a dual-layer washcoat. Catal. Today. 188(1), 70–79 (2012)CrossRefGoogle Scholar
  23. 23.
    Daya, R., Desai, C., Vernham, B.: Development and validation of a two-site kinetic model for NH3-SCR over Cu-SSZ-13. Part 1. Detailed global kinetics development based on mechanistic considerations. Emission Control Science and Technology. (2018).  https://doi.org/10.1007/s40825-018-0095-5
  24. 24.
    Bissett, E.J.: An asymptotic solution for washcoat pore diffusion in catalytic monoliths. Emission Control Sci Technol. 1(1), 3–16 (2015)CrossRefGoogle Scholar
  25. 25.
    GT-Suite Exhaust Aftertreatment Application Manual v2017Google Scholar
  26. 26.
    Fuller, E.N., Schettler, P.D., Giddings, J.C.: New method for prediction of binary gas-phase diffusion coefficients. Ind. Eng. Chem. 58(5), 18–27 (1966)CrossRefGoogle Scholar
  27. 27.
    Hindmarsh, A.C.: LSODE and LSODI, two new initial value ordinary differential equation solvers. ACM Signum Newslett. 15(4), 10–11 (1980)CrossRefGoogle Scholar
  28. 28.
    Centeno, M.A., Carrizosa, I., Odriozola, J.A.: In situ DRIFTS study of the SCR reaction of NO with NH 3 in the presence of O 2 over lanthanide doped V 2 O 5/Al 2 O 3 catalysts. Applied Catalysis B : Environ. 19(1), 67–73 (1998)Google Scholar
  29. 29.
    Chatterjee, D., Burkhardt, T., Weibel, M., Nova, I., Grossale, A., Tronconi, E.: Numerical simulation of zeolite-and V-based SCR catalytic converters (No. 2007–01-1136). SAE Technical Paper. (2007)Google Scholar
  30. 30.
    Deb, K., & Jain, H.: An evolutionary many-objective optimization algorithm using reference-point-based nondominated sorting approach, part I: Solving problems with box constraints. IEEE Trans. Evol. Comput. 18(4), 577–601 (2014)Google Scholar
  31. 31.
    Nelder, J.A., & Mead, R.: A simplex method for function minimization. Comput. J. 7(4), 308–313 (2014)Google Scholar
  32. 32.
    Kamasamudram, K., Currier, N.W., Chen, X., Yezerets, A.: Overview of the practically important behaviors of zeolite-based urea-SCR catalysts, using compact experimental protocol. Catal. Today. 151(3), 212–222 (2010)CrossRefGoogle Scholar
  33. 33.
    Partridge Jr, W.P., Choi, J.S., Parks, I.I., James, E., Currier, N., Yezerets, A., ..., Kamasamudram, K.: Cummins/ORNL-FEERC CRADA: NOx control & measurement technology for heavy-duty diesel engines (No. ORNL/TM-2017/76). Oak Ridge National Laboratory (ORNL), Oak Ridge, TN (United States). Fuels, Engines and Emissions Research Center. (2017)Google Scholar
  34. 34.
    Joshi, S.Y., Kumar, A., Luo, J., Kamasamudram, K., Currier, N.W., & Yezerets, A.: New insights into the mechanism of NH 3-SCR over Cu-and Fe-zeolite catalyst: apparent negative activation energy at high temperature and catalyst unit design consequences. Appl. Catal. B Environ. (2018)Google Scholar

Copyright information

© Springer Nature Switzerland AG 2018

Authors and Affiliations

  1. 1.Isuzu Technical Center of AmericaPlymouthUSA

Personalised recommendations