You are on page 1of 22

Supporting Information

for

An explicit reduced-order model of Cu-Zeolite SCR catalyst for


embedding in ECM

Rohil Daya*, Saurabh Y. Joshi, Rama Krishna Dadi, Yadan Tang, Dylan Trandal,
Anand Srinivasan, Antonius P. Nusawardhana, Michael Cunningham

Cummins Inc., 1900 McKinley Ave, Columbus, IN 47201, U.S.A.

E-mail : rohil.daya@cummins.com
1. Thiele modulus estimation for SCR catalysts
The dimensionless quantity 𝜙 is called the Thiele modulus, and is defined as the square root of the
characteristic diffusion time in the washcoat divided by the reaction time [1,2]. For a first order reaction,
it can be mathematically expressed as:

𝑘
𝜙 = 𝛿𝑤𝑐𝑎𝑣𝑔 √𝐷1 (1)
𝑒

Here 𝛿𝑤𝑐𝑎𝑣𝑔 represents the average washcoat thickness (m), 𝑘1 represents the first order rate constant

(1/s), and 𝐷𝑒 represents the effective washcoat diffusion coefficient (m2/s). In general, SCR mechanisms
consist of multiple coupled non-linear reactions, and as a result, equation (1) needs to be modified to
account for this nonlinearity. A generalized version of the thiele modulus (𝜙2 ) was proposed in [2],
defined as :

𝐶
𝛿𝑤𝑐𝑎𝑣𝑔 2 ∫0 𝑤𝑐 𝑟𝑙 (𝐶 ′ )𝑑𝐶 ′
𝜙2,𝑙 = √ (2)
𝐶𝑤𝑐 𝐷𝑒

A key limitation of the thiele modulus approach is the dependence on the reaction rates. In the present
work, the generalized thiele modulus was deduced from the high-fidelity 1+1D model. The SCR kinetics
were first developed using kinetic regime experimental data and a high-fidelity 1+1D model that solves
the full numerical species balance equation in the washcoat direction (𝑥). In this model, the solid-phase
species balance is given by the equation below:

𝜕𝑦𝑙,𝑠 1 𝜕2 𝑦𝑙,𝑠
𝜀𝑤𝑐 𝜕𝑡
= 𝐶 ∑𝐼𝑖 𝜈𝑖,𝑙 . 𝑟𝑖 + 𝐷𝑒,𝑙 𝜕𝑥 2
; 0 < 𝑥 < 𝛿𝑤𝑐 𝑎𝑣𝑔 (3)
0

Subject to the following boundary conditions:

𝑘𝑚𝑜,𝑙 𝜕𝑦𝑙,𝑠
𝛿𝑤𝑐𝑎𝑣𝑔
(𝑦𝑙,𝑔 − 𝑦𝑙,𝑠 ) = −𝐷𝑒,𝑙 𝜕𝑥
𝑎𝑡 𝑥 = 0 (4)

𝜕𝑦𝑙,𝑠
𝐷𝑒,𝑙 = 0 𝑎𝑡 𝑥 = 𝛿𝑤𝑐 𝑎𝑣𝑔 (5)
𝜕𝑥

The remaining species and energy balances are the same as the high-fidelity model described in the main
manuscript. The effective diffusion coefficient was modeled using the effective pore diffusion approach,
and is defined as :

𝜀
𝐷𝑒,𝑙 = 𝐷
𝜏 𝑓,𝑙
(6)
Here, the binary gas-phase diffusion coefficient (𝐷𝑓,𝑙 ) was modeled using the standard Fuller correlation
[3]. The porosity (void fraction) of the washcoat (𝜀𝑤𝑐 ) was fixed at 0.5, and the tortuosity (𝜏) was varied
to describe the NOx conversion in the mixed kinetic-washcoat diffusion regime, as shown in Figure S1.
Five grid points in the washcoat direction were sufficient to achieve convergence for all 𝜏 values.

Figure S1. NOx Conversion under standard SCR conditions (NO2/NOx = 0) at 73k/h GHSV, 930 ppm NOx and 1.06
ANR . Dots (red) represent experimental data, and solid lines represent 1+1D model results with different 𝜏 values
(𝜏 = 1.5 (red), 𝜏 = 5 (green), 𝜏 = 1.0 (black) and 𝜏 = 50 (magenta))

It is observed that a relatively low tortuosity value (𝜏 = 1.5) was necessary to describe the experimental
data under SCR conditions. This indicates minimal washcoat diffusion limitations in the catalyst, enabling
the utilization of simplified washcoat diffusion models. For known washcoat diffusivities, reaction rates
and species concentrations from the 1+1D model, the generalized thiele modulus calculation could be
performed. Expanding and integrating equation (2) for NO and NO2 leads to the following expressions for
the reaction mechanism considered in Table 1 of the main manuscript:

𝑘𝑁𝑂 𝑦𝑁𝑂 𝛺
𝑂𝑥 2𝑠 1
𝛿𝑤𝑐𝑎𝑣𝑔 𝜏( 𝑟𝑁𝑂𝑂𝑥 − − 4𝑟𝑁𝐻3 − 4𝑟𝑁𝐻3 + 4𝑟𝑆𝑆𝐶𝑅 + 4𝑟𝑆𝑆𝐶𝑅𝑁 𝑂 + 𝑟𝐹𝑆𝐶𝑅 + 𝑟𝐹𝑆𝐶𝑅𝑁 𝑂 + 𝑟𝐴𝑁𝑡𝑖𝑡𝑟 )
𝜙2,𝑁𝑂 = √ 𝐾𝑒𝑞 𝑂𝑥𝑁𝑂1 𝑂𝑥𝑁𝑂2 2 2
(7)
𝐶0 𝜀𝐷𝑓,𝑁𝑂 𝑦𝑁𝑂,𝑠

𝛿𝑤𝑐𝑎𝑣𝑔 𝜏(−𝑟𝑁𝑂𝑂𝑥 − 𝑘𝑁𝑂 𝑦0.5 𝛺 + 𝑟𝐹𝑆𝐶𝑅 + 𝑟𝐹𝑆𝐶𝑅𝑁 𝑂 + 3𝑟𝑁𝑆𝐶𝑅 + 2𝑟𝑁𝑆𝐶𝑅𝑁 𝑂 + 2𝑟𝐴𝑁𝑓𝑜𝑟𝑚 + 𝑟𝐴𝑁𝑡𝑖𝑡𝑟 )
𝑂𝑥 𝑂2 1
𝜙2,𝑁𝑂2 = √ 2 2
(8)
𝐶0 𝜀𝐷𝑓,𝑁𝑂2 𝑦𝑁𝑂2 ,𝑠

Figure S2 shows the generalized thiele modulus for NO and NO2 under standard SCR (NO2/NOx = 0) and
fast SCR (NO2/NOx = 0.5) conditions. The estimated generalized thiele modulus for SCR catalysts using
this approach is observed to be < 0.5 under for all relevant temperatures under standard and fast SCR
conditions. This is associated with the relatively low reaction rates, low washcoat thickness (a
consequence of high cell density) and low washcoat diffusion resistance (𝜏 < 10) observed for
commercial SCR catalysts. In addition, the generalized thiele modulus reaches a maximum around
300°C-350°C and decreases at higher temperatures due to a drop in SCR reaction rates associated with
NH3 desorption, as highlighted by us previously [4].

Figure S2. Generalized thiele modulus under standard SCR conditions (NO 2/NOx = 0, 73k/h GHSV, 930 ppm NOx
and 1.06 ANR) and fast SCR conditions (NO2/NOx = 0.5, 130k/h GHSV, 990 ppm NOx and 1 ANR). Dotted lines
represent generalized thiele modulus for NO (standard SCR (red) and fast SCR (blue)) and NO 2 (fast SCR (green))

These calculations are extended to estimate the internal Sherwood number, and compare the overall mass
transfer approach with the 1+1D model in section 2.

2. Validation of fixed internal Sherwood number approach for


modeling washcoat diffusion in SCR catalysts
Solution of the 2-D mass balance equation considering convection, diffusion and reaction in a monolithic
catalyst can be used to estimate internal Sherwood numbers for approximating washcoat diffusion, which
depend on the washcoat geometry and reaction mechanism. In general, the internal Sherwood number for
multiple coupled non-linear reactions can be approximated as [2]:
𝛬𝜙2
𝑆ℎ𝑖,𝑙 = 𝑆ℎ𝑖,∞ + 1+𝛬𝜙2,𝑙 (9)
2,𝑙

Here 𝛬 is a constant dependent on the washcoat property and reaction parameters. 𝛬 can vary from 0.25
for a zero-order irreversible reaction to 0.92 for a second-order irreversible reaction [2]. Utilizing the
generalized thiele modulus from Figure S2, and the 𝛬 value for first-order irreversible reactions (0.32)
[2], the internal Sherwood numbers for NO and NO2 are estimated, as shown in Figure S3. As expected,
low values of generalized thiele modulus translate to a nearly constant internal Sherwood number as a
function of temperature. This enables the utilization of the overall mass transfer approach with constant
(asymptotic) internal Sherwood numbers, which represents a useful simplification due to the non-
interference from reaction rates on washcoat diffusion flux calculations.

Figure S3. Internal Sherwood number under standard SCR conditions (NO 2/NOx = 0, 73k/h GHSV, 930 ppm NOx
and 1.06 ANR) and fast SCR conditions (NO2/NOx = 0.5, 130k/h GHSV, 990 ppm NOx and 1 ANR). Dotted lines
represent internal Sherwood number for NO (standard SCR (red) and fast SCR (blue)) and NO 2 (fast SCR (green))

To validate this constant internal Sherwood number approach, and identify the limitations, we simulated
the NOx conversion under fast SCR conditions for a range of 𝜏 values. The results (Figure S4) confirm
that a low tortuosity value is necessary to match the experimental data, which is reproduced by both the
1+1D and low-D models (the low-D model here refers to the simplified 1+1D model with constant
internal Sherwood number). The low-D model can also reproduce the washcoat diffusion resistance at
higher 𝜏 values, matching the 1+1D model predicted NOx conversion for 𝜏 = 10. (Figure S4).
Deviations are observed between the two models for 𝜏 = 50 , indicating violation of the constant internal
Sherwood number assumption. Further analysis (not shown here for brevity) confirmed the constant
𝐷
internal Sherwood number approximation generally works for 𝜏 ≤ 20 (i.e. 𝐷𝑓 ≤ 40), which is true for
𝑒

most commercial SCR catalysts.

As the reaction rates (and therefore, the corresponding internal Sherwood numbers) under fast SCR
conditions are higher than under standard and slow SCR conditions, the conclusions from this analysis are
also valid under varying NO2/NOx ratios. A similar inference can be made relating to catalyst aging, since
degreened catalysts are expected to have the highest intrinsic reaction rates.

Figure S4. NOx Conversion under fast SCR conditions (NO2/NOx = 0.5) at 130k/h GHSV, 990 ppm NOx and 1
ANR . Dots (red) represent experimental data, solid lines represent 1+1D low model results and dashed lines
represent Low-D model results with different 𝜏 values (𝜏 = 1.5 (red), 𝜏 = 1.0 (black) and 𝜏 = 50 (magenta))

3. Derivation of piecewise analytical solution from reduced-order


SCR catalyst model
The reduced-order model described in section 3.2 of the main manuscript consists of an array of 0-D CSTRs
in series. These simplified equations can have piecewise analytical solutions when subject to
approximations, as demonstrated below.

Energy Balance

We start with the solid-phase energy balance equation in the reduced-order model:
𝑑𝑇𝑠 ℎ ℎ𝑥 𝐴𝑥 𝛿𝑤𝑐𝑎𝑣𝑔
𝜀𝑤𝑐 = (𝑇 − 𝑇𝑠 ) + 𝐺𝑆𝐴.𝑉 (𝑇𝑥 − 𝑇𝑠 )(𝑇𝑥 − 𝑇𝑠 ) + ∑𝐼𝑖 𝜈𝑖,𝑙 𝑟𝑖 (−∆𝐻𝑖 ) (10)
𝑑𝑡 𝛿𝑠 𝐶𝑝,𝑠 𝜌𝑠 𝑔 𝑐𝑎𝑡 𝛿𝑠 𝐶𝑝,𝑠 𝜌𝑠 𝛿𝑠 𝐶𝑝,𝑠 𝜌𝑠

Rearranging this equation, we have:

ℎ𝑥 𝐴𝑥 𝑇𝑥 ℎ𝑥 𝐴𝑥
ℎ𝑇𝑔 + +𝛿𝑤𝑐𝑎𝑣𝑔 ∑𝐼𝑖 𝜈𝑖,𝑙 𝑟𝑖 (−∆𝐻𝑖 ) ℎ+
𝑑𝑇𝑠 𝐺𝑆𝐴.𝑉𝐶𝑎𝑡 𝐺𝑆𝐴.𝑉𝐶𝑎𝑡
𝜀𝑤𝑐 𝑑𝑡
= 𝛿𝑠 𝐶𝑝,𝑠 𝜌𝑠
− 𝑇𝑠 ( 𝛿𝑠 𝐶𝑝,𝑠 𝜌𝑠
) (11)

By separation of variables, we get:

𝑑𝑇𝑠 𝑑𝑡
ℎ 𝐴 𝑇 ℎ 𝐴 = (12)
ℎ𝑇𝑔 + 𝑥 𝑥 𝑥 +𝛿𝑤𝑐𝑎𝑣𝑔 ∑𝑖 𝜈𝑖,𝑙 𝑟𝑖 (−∆𝐻𝑖 )−𝑇𝑠 (ℎ+ 𝑥 𝑥 )
𝐼 𝜀𝑤𝑐 𝛿𝑠 𝐶𝑝,𝑠 𝜌𝑠
𝐺𝑆𝐴.𝑉𝐶𝑎𝑡 𝐺𝑆𝐴.𝑉𝐶𝑎𝑡

Here, we invoke the following approximations that decouple the solid temperature from the reaction rate
and gas temperature, to enable integration of equation (12):

𝑟𝑖𝑡+1 − 𝑟𝑖𝑡 → 0 for i = 1 to 14 (13)


𝑇𝑔𝑡+1 − 𝑇𝑔𝑡 → 0 for i = 1 to 14 (14)

Quasi Steady-State Approximation

In addition, we define the following intermediate variables:

ℎ.𝐺𝑆𝐴.𝑉𝐶𝑎𝑡
𝑅ℎ = ℎ𝑥 𝐴𝑥
(15)

ℎ𝑇𝑥
𝑆1 = ℎ𝑇𝑔 + + 𝛿𝑤𝑐𝑎𝑣𝑔 ∑𝐼𝑖 𝜈𝑖,𝑙 𝑟𝑖 (−∆𝐻𝑖 ) (16)
𝑅ℎ


𝑆2 = ℎ + 𝑅 (17)

Using these variables in equation (12), we have the integrable equation:

𝑇 𝑑𝑇𝑠 𝑡+1 𝑑𝑡
∫𝑇 𝑆𝑡+1 𝑆 −𝑇
= ∫𝑡 (18)
𝑆𝑡 1 𝑠 𝑆2 𝜀𝑤𝑐 𝛿𝑠 𝐶𝑝,𝑠 𝜌𝑠

Integrating equation (18) within the approximations of equations (13)-(14) gives us:

𝑆1𝑡 −𝑇𝑠𝑡+1 𝑆2𝑡 −𝑆2𝑡 𝛥𝑡


log 𝑒 ( 𝑆1 𝑡 −𝑇𝑠𝑡 𝑆2𝑡
) = 𝜀𝑤𝑐 𝛿𝑠 𝐶𝑝,𝑠 𝜌𝑠
(19)

Simplifying further yields:

−𝑆2𝑡 𝛥𝑡
𝑆1 𝑡 −(𝑆1 𝑡 −𝑇𝑠𝑡 𝑆2𝑡 )𝑒 𝜀𝑤𝑐 𝛿𝑠 𝐶𝑝,𝑠 𝜌𝑠
𝑇𝑠𝑡+1 = (20)
𝑆2𝑡
Substituting the definition of 𝑆1 and 𝑆2 from equations (16)-(17), and generalizing for 𝑛 CSTRs in series
with 𝑧 = 1,2 … 𝑛 we arrive at the final solution for solid-phase temperature:
1
− (1+ 𝑅 )
ℎ ℎ∆𝑡
𝑅ℎ 𝑅 𝜀𝑤𝑐 𝛿𝑠 𝐶𝑝,𝑠𝜌𝑠
𝑅ℎ 𝑇𝑔𝑡,𝑧 + 𝑇𝑥 + 𝛿 ∑𝐼 𝜈 𝑟 (−∆𝐻𝑖 ) − ( ℎ 𝛿𝑤𝑐𝑎𝑣𝑔 ∑𝐼𝑖 𝜈𝑖,𝑙 𝑟𝑖𝑡,𝑧 (−∆𝐻𝑖 ) +𝑅ℎ (𝑇𝑔𝑡,𝑧 − 𝑇𝑠𝑡,𝑧 ) +(𝑇𝑥 − 𝑇𝑠𝑡,𝑧 ))𝑒
ℎ 𝑤𝑐𝑎𝑣𝑔 𝑖 𝑖,𝑙 𝑖𝑡,𝑧
𝑇𝑠 𝑡+1,𝑧 = ℎ
𝑅ℎ +1
(21)

To estimate the gas-phase temperature, we start with the gas-phase energy balance in the reduced-order
model:

(𝑇𝑔 −𝑇𝑖𝑛 ) 4𝛼𝑔 𝑁𝑢


𝑣𝑔 𝐿
+ 2
𝑑ℎ
(𝑇𝑔 − 𝑇𝑠 ) = 0 (22)

For a known solid-phase temperature (equation (21)), the gas-phase temperature can be estimated directly
from rearrangement of equation (22). Generalizing this solution for 𝑛 CSTRs in series with 𝑧 = 1,2 … 𝑛
gives us:

𝑣𝑔 𝑇𝑔𝑡+1,𝑧−1 4𝛼𝑔 𝑁𝑢𝑇𝑠 𝑡+1,𝑧


+
∆𝑧 𝑑2

𝑇𝑔𝑡+1,𝑧 = 𝑣𝑔 4𝛼𝑔 𝑁𝑢 (23)
+ 2
∆𝑧 𝑑ℎ

Coverage

NH3

We start with the coverage balance equation for NH3 in the reduced-order model :

𝑑𝜃𝑁𝐻3 𝑆1
𝛺1 𝑑𝑡
= ∑𝐼𝑖 𝜈𝑖,𝑙 . 𝑟𝑖 (24)

Expanding the terms on the right-hand side for the reaction mechanism considered here (Table 1 in main
manuscript) gives us:

𝑑𝜃𝑁𝐻3 𝑆1
𝛺1 = (𝑟𝐴𝑑𝑠 − 𝑟𝐷𝑒𝑠 − 4 (𝑟𝑁𝐻3 + 𝑟𝑁𝐻3 ) − 4 (𝑟𝑁𝐻3 + 𝑟𝑁𝐻3 ) − 4𝑟𝑆𝑆𝐶𝑅 − 4𝑟𝑆𝑆𝐶𝑅𝑁2𝑂 −
𝑑𝑡 𝑂𝑥1 𝑂𝑥2 𝑂𝑥𝑁𝑂1 𝑂𝑥𝑁𝑂2

2𝑟𝐹𝑆𝐶𝑅 − 𝑟𝐹𝑆𝐶𝑅𝑁2𝑂 − 4𝑟𝑁𝑆𝐶𝑅 − 2𝑟𝑁𝑆𝐶𝑅𝑁2𝑂 − 2𝑟𝐴𝑁𝑓𝑜𝑟𝑚 ) (25)

We expand the reaction rates further:

𝑑𝜃𝑁𝐻3 𝑆1
𝛺1 = 𝛺1 (𝐸1 (1 − 𝜃𝑁𝐻3 𝑆1 − 𝜃𝐴𝑁𝑆1 ) − 𝐸1 𝜃𝑁𝐻3𝑆1 − 𝐸3 𝜃𝑁𝐻3𝑆1 − 𝐸4 𝜃𝑁𝐻3𝑆1 − 𝐸5 𝜃𝑁𝐻3𝑆1 − 𝐸6 𝜃𝑁𝐻3𝑆1 −
𝑑𝑡

𝐸7 𝜃𝑁𝐻3𝑆1 − 𝐸8 𝜃𝑁𝐻3𝑆1 − 𝐸9 𝜃𝑁𝐻3𝑆1 − 𝐸10 𝜃𝑁𝐻3𝑆1 − 𝐸11 𝜃𝑁𝐻3𝑆1 ) (26)

𝑑𝜃𝑁𝐻3 𝑆1
⇒ 𝑑𝑡
= 𝐸1 (1 − 𝜃𝐴𝑁𝑆1 ) − 𝜃𝑁𝐻3 𝑆1 ∑11
𝑖=1 𝐸𝑖 (27)
Here, we define the following intermediate variables:

𝐸1 = 𝑘𝐴𝑑𝑠 𝑦𝑁𝐻3 ,𝑠 (28)


𝐸2 = (𝑘𝐷𝑒𝑠1 + 𝑘𝐷𝑒𝑠2 ) (29)
𝐸3 = 4(𝑘𝑁𝐻3 𝑂𝑥1 𝑦𝑂0.5
2
+ 𝑘𝑁𝐻3 𝑂𝑥2 𝑦𝑂0.5
2
) (30)

𝐸4 = 4(𝑘𝑁𝐻3 𝑂𝑥𝑁𝑂1 𝑦𝑁𝑂,𝑠 𝑦𝑂0.2


2
+ 𝑘𝑁𝐻3 𝑂𝑥𝑁𝑂2 𝑦𝑁𝑂,𝑠 𝑦𝑁𝑂2 ,𝑠 (31)

𝐸5 = 4𝑘𝑆𝑆𝐶𝑅 𝑦𝑁𝑂,𝑠 𝑦𝑂0.32


2
(32)
𝐸6 = 4𝑘𝑆𝑆𝐶𝑅𝑁2𝑂 𝑦𝑁𝑂,𝑠 𝑦𝑂0.32
2
(33)
𝐸7 = 2𝑘𝐹𝑆𝐶𝑅 𝑦𝑁𝑂,𝑠 𝑦𝑁𝑂2 ,𝑠 (34)
𝐸8 = 𝑘𝐹𝑆𝐶𝑅𝑁2𝑂 𝑦𝑁𝑂,𝑠 𝑦𝑁𝑂2 ,𝑠 (35)
𝐸9 = 4𝑘𝑁𝑆𝐶𝑅 𝑦𝑁𝑂2 ,𝑠 (36)
𝐸10 = 2𝑘𝑁𝑆𝐶𝑅𝑁2𝑂 𝑦𝑁𝑂2 ,𝑠 (37)
𝐸11 = 2𝑘𝐴𝑁𝑓𝑜𝑟𝑚 𝑦𝑁𝑂2 ,𝑠 (38)
Now, we invoke the following approximations that decouple the NH3 coverage from rate constants and
surface mole fractions, to enable integration of equation (27)

𝜃𝐴𝑁𝑆1𝑡+1 − 𝜃𝐴𝑁𝑆1𝑡 → 0 (39)


𝑘𝑖𝑡+1 − 𝑘𝑖𝑡 → 0 for i = 1 to 11 (40)
𝑦𝑖,𝑠𝑡+1 − 𝑦𝑖,𝑠𝑡 → 0 for i = 1 to 11 (41)

Quasi Steady-State Approximation

The approximation in equation (39) is justified as the build-up of AN is typically a slow process,
occurring on the order of minutes and hours. In addition, the slow change in solid-phase temperature that
influences the rate constants also helps justify the approximation in equation (40). Separation of variables
on equation (27) gives us the following integrable equation:

𝜃𝑁𝐻3 𝑆1 𝑑𝜃𝑁𝐻3 𝑆1 𝑡+1


𝑡+1
∫𝜃 𝐸1 (1−𝜃𝐴𝑁𝑆1 )− 𝜃𝑁𝐻3 𝑆1 ∑11
= ∫𝑡 𝑑𝑡 (42)
𝑁𝐻3 𝑆1 𝑡 𝑖=1 𝐸𝑖

Integrating equation (42) within the approximations of equations (39)-(41) gives us:

𝐸1𝑡 (1−𝜃𝐴𝑁𝑆1𝑡 )− 𝜃𝑁𝐻3 𝑆1𝑡+1 ∑11


𝑖=1 𝐸𝑖𝑡
log 𝑒 ( 𝐸1𝑡 (1−𝜃𝐴𝑁𝑆1𝑡 )− 𝜃𝑁𝐻3 𝑆1𝑡 ∑11
) = −𝛥𝑡 ∑11
𝑖=1 𝐸𝑖𝑡 (43)
𝑖=1 𝐸𝑖𝑡

Simplifying further, and generalizing for 𝑛 CSTRs in series with 𝑧 = 1,2 … 𝑛 we arrive at the final
solution for NH3 coverage:
−𝛥𝑡 ∑11
𝑖=1 𝐸𝑖
𝐸1𝑡,𝑧 (1−𝜃𝐴𝑁𝑆1𝑡,𝑧 )−(𝐸1𝑡,𝑧 (1−𝜃𝐴𝑁𝑆1𝑡,𝑧 )− 𝜃𝑁𝐻3𝑆1𝑡,𝑧 ∑11
𝑖=1 𝐸𝑖𝑡,𝑧 ) 𝑒
𝑡,𝑧
𝜃𝑁𝐻3 𝑆1𝑡+1,𝑧 = ∑11
(44)
𝑖=1 𝐸𝑖𝑡,𝑧

AN

We start with the coverage balance equation for AN in the reduced-order model :

𝑑𝜃𝐴𝑁𝑆1
𝛺1 𝑑𝑡
= ∑𝐼𝑖 𝜈𝑖,𝑙 . 𝑟𝑖 (45)

Expanding the terms on the right-hand side for the reaction mechanism considered here (Table 1 in main
manuscript) gives us:

𝑑𝜃𝐴𝑁𝑆1
𝛺1 𝑑𝑡
= (𝑟𝐴𝑁𝑓𝑜𝑟𝑚 − 𝑟𝐴𝑁𝑡𝑖𝑡𝑟 − 𝑟𝐴𝑁𝑑𝑒𝑐𝑜𝑚𝑝𝑁2 − 𝑟𝐴𝑁𝑑𝑒𝑐𝑜𝑚𝑝𝑁2 𝑂 ) (46)

We expand the reaction rates further:

𝑑𝜃𝐴𝑁𝑆1
𝛺1 = 𝛺1 (0.5𝐸11 𝜃𝑁𝐻3 𝑆1 − 𝐸12 𝜃𝐴𝑁𝑆1 − 𝐸13 𝜃𝐴𝑁𝑆1 − 𝐸14 𝜃𝐴𝑁𝑆1 ) (47)
𝑑𝑡

𝑑𝜃𝐴𝑁𝑆1
⇒ 𝑑𝑡
= 0.5𝐸11 𝜃𝑁𝐻3 𝑆1 − 𝜃𝐴𝑁𝑆1 ∑14
𝑖=12 𝐸𝑖 (48)

Here, we define the following intermediate variables:

𝐸12 = 𝑘𝐴𝑁𝑡𝑖𝑡𝑟 𝑦𝑁𝑂,𝑠 (49)


𝐸13 = 𝑘𝐴𝑁𝑑𝑒𝑐𝑜𝑚𝑝𝑁 (50)
2

𝐸14 = 𝑘𝐴𝑁𝑑𝑒𝑐𝑜𝑚𝑝𝑁 (51)


2𝑂

The approximations listed in equations (40)-(41) are extended for i = 12 to 14, to decouple the AN
coverage from rate constants and surface mole fractions and enable integration of equation (48)

𝜃 𝑑𝜃𝐴𝑁𝑆1 𝑡+1
∫𝜃 𝐴𝑁𝑆1 𝑡+1 0.5𝐸 14 = ∫𝑡 𝑑𝑡 (52)
𝐴𝑁𝑆1 𝑡 11 𝜃𝑁𝐻3 𝑆1 − 𝜃𝐴𝑁𝑆1 ∑𝑖=12 𝐸𝑖

Integrating equation (52) within the approximations of equations (40)-(41) gives us:

0.5𝐸11𝑡 𝜃𝑁𝐻3 𝑆1𝑡+1 − 𝜃𝐴𝑁𝑆1𝑡+1 ∑14


𝑖=12 𝐸𝑖𝑡
log 𝑒 ( 0.5𝐸11 𝑡 𝜃𝑁𝐻3 𝑆1 − 𝜃𝐴𝑁𝑆1𝑡 ∑11
) = −𝛥𝑡 ∑14
𝑖=12 𝐸𝑖𝑡 (53)
𝑡+1 𝑖=1 𝐸𝑖𝑡

Simplifying further, and generalizing for 𝑛 CSTRs in series with 𝑧 = 1,2 … 𝑛 we arrive at the final
solution for AN coverage:
−𝛥𝑡 ∑14
𝑖=12 𝐸𝑖
0.5𝐸11𝑡,𝑧 𝜃𝑁𝐻3𝑆1𝑡+1,𝑧 −(0.5𝐸11𝑡,𝑧 𝜃𝑁𝐻3 𝑆1𝑡+1,𝑧 − 𝜃𝐴𝑁𝑆1𝑡,𝑧 ∑14
𝑖=12 𝐸𝑖𝑡,𝑧 ) 𝑒
𝑡,𝑧
𝜃𝐴𝑁𝑆1𝑡+1,𝑧 = ∑14
(54)
𝑖=12 𝐸𝑖𝑡,𝑧

Species Balance

We start with the gas-phase species balance equation in the reduced-order model:

(𝑦𝑙,𝑔 −𝑦𝑙,𝑖𝑛 ) 4𝑘𝑚𝑜,𝑙


𝑣𝑔 𝐿
+ 𝑑ℎ
(𝑦𝑙,𝑔 − 𝑦𝑙,𝑠 ) = 0 (55)

Rearranging the above equation gives us:

𝑣 4𝑘𝑚𝑜,𝑙 𝑣𝑔 𝑦𝑙,𝑖𝑛 4𝑘𝑚𝑜,𝑙 𝑦𝑙,𝑠


𝑦𝑙,𝑔 ( 𝐿𝑔 + 𝑑ℎ
)− 𝐿
− 𝑑ℎ
=0 (56)

𝑣𝑔 𝑦𝑙,𝑖𝑛 4𝑘𝑚𝑜,𝑙 𝑦𝑙,𝑠


+
𝐿 𝑑ℎ
⇒ 𝑦𝑙,𝑔 = 𝑣𝑔 4𝑘𝑚𝑜,𝑙 (57)
( + )
𝐿 𝑑ℎ

We use this result in the solid-phase species balance in the reduced-order model :

1 𝐼 𝑘𝑚𝑜,𝑙
∑ 𝜈 . 𝑟𝑖 + (𝑦𝑙,𝑔 − 𝑦𝑙,𝑠 ) = 0 (58)
𝐶0 𝑖 𝑖,𝑙 𝛿𝑤𝑐𝑎𝑣𝑔

𝑣𝑔 𝑦𝑙,𝑖𝑛 4𝑘𝑚𝑜,𝑙 𝑦𝑙,𝑠


1 𝐼 𝑘𝑚𝑜,𝑙 +
𝐿 𝑑ℎ
∑ 𝜈 . 𝑟𝑖 + 𝛿 ( − 𝑦𝑙,𝑠 ) = 0 (59)
𝐶0 𝑖 𝑖,𝑙 𝑤𝑐𝑎𝑣𝑔 (
𝑣𝑔 4𝑘𝑚𝑜,𝑙
+ )
𝐿 𝑑ℎ

𝑣𝑔 4𝑘𝑚𝑜,𝑙
𝛿𝑤𝑐𝑎𝑣𝑔 ( + )
𝐿 𝑑ℎ
Multiplying equation (59) by 𝑘𝑚𝑜,𝑙
, we get :

𝑣𝑔 4𝑘𝑚𝑜,𝑙
𝛿𝑤𝑐𝑎𝑣𝑔 ( + ) 𝑣𝑔 𝑦𝑙,𝑖𝑛 4𝑘𝑚𝑜,𝑙 𝑦𝑙,𝑠 𝑣𝑔 4𝑘𝑚𝑜,𝑙
𝐿 𝑑ℎ
∑𝐼𝑖 𝜈𝑖,𝑙 . 𝑟𝑖 + + −(𝐿 + ) 𝑦𝑙,𝑠 = 0 (60)
𝑘𝑚𝑜,𝑙 𝐶0 𝐿 𝑑ℎ 𝑑ℎ

𝑣𝑔 4𝑘𝑚𝑜,𝑙
𝛿𝑤𝑐𝑎𝑣𝑔 ( + ) 𝑣𝑔 𝑦𝑙,𝑖𝑛 𝑣𝑔 𝑦𝑙,𝑠
𝐿 𝑑ℎ
⇒ ∑𝐼𝑖 𝜈𝑖,𝑙 . 𝑟𝑖 + − =0 (61)
𝑘𝑚𝑜,𝑙 𝐶0 𝐿 𝐿

Rearranging equation (61) gives us:

𝑣𝑔 𝑑ℎ
4𝛿𝑤𝑐𝑎𝑣𝑔 ( +𝐿𝑘𝑚𝑜,𝑙 )
4
𝑦𝑙,𝑠 = 𝑦𝑙,𝑖𝑛 + ∑𝐼𝑖 𝜈𝑖,𝑙 . 𝑟𝑖 (62)
𝑣𝑔 𝑘𝑚𝑜,𝑙 𝐶0 𝑑ℎ

We now expand equation (62) for each of the solved surface species to derive the piecewise analytical
solution.
NO

Re-writing equation (62) for NO gives us:

𝑣𝑔 𝑑ℎ
4𝛿𝑤𝑐𝑎𝑣𝑔 ( +𝐿𝑘𝑚𝑜,𝑁𝑂 )
4
𝑦𝑁𝑂,𝑠 = 𝑦𝑁𝑂𝑖𝑛 + ∑𝐼𝑖 𝜈𝑖,𝑁𝑂 . 𝑟𝑖 (63)
𝑣𝑔 𝑘𝑚𝑜,𝑁𝑂 𝐶0 𝑑ℎ

Here, we define a mass-transfer intermediate variable, 𝐾𝑁𝑂 , with the units of m3wc/mol-s:

𝑣𝑔 𝑑ℎ
4𝛿𝑤𝑐𝑎𝑣𝑔 ( +𝐿𝑘𝑚𝑜,𝑁𝑂 )
4
𝐾𝑁𝑂 = 𝑣𝑔 𝑘𝑚𝑜,𝑁𝑂 𝐶0 𝑑ℎ
(64)

We expand the reaction rates on the right-hand side of equation (63) as shown below:

𝑦𝑁𝑂,𝑠 = 𝑦𝑁𝑂𝑖𝑛 + 𝐾𝑁𝑂 (−𝑟𝑁𝑂𝑂𝑥 + 4𝑟𝑁𝐻3𝑂𝑥𝑁𝑂1 + 4𝑟𝑁𝐻3𝑂𝑥𝑁𝑂2 − 4𝑟𝑆𝑆𝐶𝑅 − 4𝑟𝑆𝑆𝐶𝑅𝑁2𝑂 − 𝑟𝐹𝑆𝐶𝑅 − 𝑟𝐹𝑆𝐶𝑅𝑁2𝑂 − 𝑟𝐴𝑁𝑡𝑖𝑡𝑟 ) (65)

⇒ 𝑦𝑁𝑂,𝑠 = 𝑦𝑁𝑂𝑖𝑛 − 𝐷1 𝑦𝑁𝑂,𝑠 + 𝐷2 𝑦𝑁𝑂2 ,𝑠 + 𝐷3 𝑦𝑁𝑂,𝑠 + 𝐷4 𝑦𝑁𝑂,𝑠 𝑦𝑁𝑂2 ,𝑠 − 𝐷5 𝑦𝑁𝑂,𝑠 − 𝐷6 𝑦𝑁𝑂,𝑠 −


𝐷7 𝑦𝑁𝑂,𝑠 𝑦𝑁𝑂2 ,𝑠 − 𝐷8 𝑦𝑁𝑂,𝑠 𝑦𝑁𝑂2 ,𝑠 − 𝐷9 𝑦𝑁𝑂,𝑠 (66)

Here, we define the following intermediate variables:

𝐷1 = 𝐾𝑁𝑂 𝑘𝑁𝑂𝑂𝑥 𝑦𝑂0.5


2
𝛺1 (67)
𝐾𝑁𝑂 𝑘𝑁𝑂 𝛺
𝑂𝑥 1
𝐷2 = 𝐾𝑒𝑞
(68)

𝐷3 = 4𝐾𝑁𝑂 𝑘𝑁𝐻3 𝑂𝑥𝑁𝑂1 𝑦𝑂0.2 𝜃 𝛺


2 𝑁𝐻3 𝑆1 1
(69)

𝐷4 = 4𝐾𝑁𝑂 𝑘𝑁𝐻3 𝑂𝑥𝑁𝑂2 𝜃𝑁𝐻3 𝑆1 𝛺1 (70)

𝐷5 = 4𝐾𝑁𝑂 𝑘𝑆𝑆𝐶𝑅 𝑦𝑂0.32


2
𝜃𝑁𝐻3 𝑆1 𝛺1 (71)
𝐷6 = 4𝐾𝑁𝑂 𝑘𝑆𝑆𝐶𝑅𝑁 𝑂 𝑦𝑂0.32
2
𝜃𝑁𝐻3 𝑆1 𝛺1 (72)
2

𝐷7 = 𝐾𝑁𝑂 𝑘𝐹𝑆𝐶𝑅 𝜃𝑁𝐻3 𝑆1 𝛺1 (73)


𝐷8 = 𝐾𝑁𝑂 𝑘𝐹𝑆𝐶𝑅𝑁2 𝑂 𝜃𝑁𝐻3 𝑆1 𝛺1 (74)

𝐷9 = 𝐾𝑁𝑂 𝑘𝐴𝑁𝑇𝑖𝑡𝑟 𝜃𝐴𝑁𝑆1 𝛺1 (75)

Rearranging equation (66) leads to the following expression for NO solid-phase mole fraction:

𝑦𝑁𝑂𝑖𝑛 +𝐷2 𝑦𝑁𝑂2 ,𝑠


𝑦𝑁𝑂,𝑠 = 1+𝐷1 − 𝐷3 − 𝐷4 𝑦𝑁𝑂2 ,𝑠 + 𝐷5 + 𝐷6 + 𝐷7 𝑦𝑁𝑂2 ,𝑠 + 𝐷8 𝑦𝑁𝑂2,𝑠 + 𝐷9
(76)
NO2

Re-writing equation (62) for NO2 gives us:

𝑣𝑔 𝑑ℎ
4𝛿𝑤𝑐𝑎𝑣𝑔 ( +𝐿𝑘𝑚𝑜,𝑁𝑂2 )
4
𝑦𝑁𝑂2 ,𝑠 = 𝑦𝑁𝑂2 𝑖𝑛 + ∑𝐼𝑖 𝜈𝑖,𝑁𝑂2 . 𝑟𝑖 (77)
𝑣𝑔 𝑘𝑚𝑜,𝑁𝑂2 𝐶0 𝑑ℎ

Like NO, we define a mass-transfer intermediate variable for NO2, 𝐾𝑁𝑂2 , with the units of m3wc/mol-s:

𝑣𝑔 𝑑ℎ
4𝛿𝑤𝑐𝑎𝑣𝑔 ( +𝐿𝑘𝑚𝑜,𝑁𝑂2 )
4
𝐾𝑁𝑂2 = 𝑣𝑔 𝑘𝑚𝑜,𝑁𝑂2 𝐶0 𝑑ℎ
(78)

We expand the reaction rates on the right-hand side of equation (77) as shown below:

𝑦𝑁𝑂2 ,𝑠 = 𝑦𝑁𝑂2 𝑖𝑛 + 𝐾𝑁𝑂2 (𝑟𝑁𝑂𝑂𝑥 − 𝑟𝐹𝑆𝐶𝑅 − 𝑟𝐹𝑆𝐶𝑅𝑁2𝑂 − 3𝑟𝑁𝑆𝐶𝑅 − 2𝑟𝑁𝑆𝐶𝑅𝑁2𝑂 − 2𝑟𝐴𝑁𝑓𝑜𝑟𝑚 + 𝑟𝐴𝑁𝑡𝑖𝑡𝑟 ) (79)

⇒ 𝑦𝑁𝑂2 ,𝑠 = 𝑦𝑁𝑂2 𝑖𝑛 + 𝐷10 𝑦𝑁𝑂,𝑠 − 𝐷11 𝑦𝑁𝑂2 ,𝑠 − 𝐷12 𝑦𝑁𝑂,𝑠 𝑦𝑁𝑂2 ,𝑠 − 𝐷13 𝑦𝑁𝑂,𝑠 𝑦𝑁𝑂2 ,𝑠 − 𝐷14 𝑦𝑁𝑂2 ,𝑠 −

𝐷15 𝑦𝑁𝑂2 ,𝑠 − 𝐷16 𝑦𝑁𝑂2 ,𝑠 + 𝐷17 𝑦𝑁𝑂,𝑠 (80)

Here, we define the following intermediate variables:

𝐷10 = 𝐾𝑁𝑂2 𝑘𝑁𝑂𝑂𝑥 𝑦𝑂0.5


2
𝛺1 (81)
𝐾𝑁𝑂2 𝑘𝑁𝑂𝑂𝑥 𝛺1
𝐷11 = 𝐾𝑒𝑞
(82)

𝐷12 = 𝐾𝑁𝑂2 𝑘𝐹𝑆𝐶𝑅 𝜃𝑁𝐻3 𝑆1 𝛺1 (83)


𝐷13 = 𝐾𝑁𝑂2 𝑘𝐹𝑆𝐶𝑅𝑁 𝑂 𝜃𝑁𝐻3 𝑆1 𝛺1 (84)
2

𝐷14 = 3𝐾𝑁𝑂2 𝑘𝑁𝑆𝐶𝑅 𝜃𝑁𝐻3 𝑆1 𝛺1 (85)


𝐷15 = 2𝐾𝑁𝑂2 𝑘𝑁𝑆𝐶𝑅𝑁2 𝑂 𝜃𝑁𝐻3 𝑆1 𝛺1 (86)

𝐷16 = 2𝐾𝑁𝑂2 𝑘𝐴𝑁𝑓𝑜𝑟𝑚 𝜃𝑁𝐻3 𝑆1 𝛺1 (87)


𝐷17 = 𝐾𝑁𝑂2 𝑘𝐴𝑁𝑇𝑖𝑡𝑟 𝜃𝐴𝑁𝑆1 𝛺1 (88)
Rearranging equation (80) leads to the following expression for the NO2 solid-phase mole fraction:

𝑦𝑁𝑂2 +𝐷10 𝑦𝑁𝑂,𝑠 + 𝐷17 𝑦𝑁𝑂,𝑠


𝑖𝑛
𝑦𝑁𝑂2 ,𝑠 = (89)
1+ 𝐷11 + 𝐷12 𝑦𝑁𝑂,𝑠 + 𝐷13 𝑦𝑁𝑂,𝑠 + 𝐷14 + 𝐷15 +𝐷16

Using equation (76) for the NO surface mole fraction in equation (89) leads to:

𝑦𝑁𝑂 +𝐷2 𝑦𝑁𝑂 ,𝑠


𝑖𝑛 2
𝑦𝑁𝑂2 +(𝐷10 + 𝐷17 )
𝑖𝑛 1+𝐷1 − 𝐷3 − 𝐷4 𝑦𝑁𝑂 ,𝑠 + 𝐷5 + 𝐷6 + 𝐷7 𝑦𝑁𝑂 ,𝑠 + 𝐷8 𝑦𝑁𝑂 ,𝑠 + 𝐷9
2 2 2
𝑦𝑁𝑂2 ,𝑠 = 𝑦𝑁𝑂 +𝐷2 𝑦𝑁𝑂 ,𝑠 (90)
𝑖𝑛 2
1+ 𝐷11 ++ 𝐷14 + 𝐷15 +𝐷16 +(𝐷12 + 𝐷13 )
1+𝐷1 − 𝐷3 − 𝐷4 𝑦𝑁𝑂 ,𝑠 + 𝐷5 + 𝐷6 + 𝐷7 𝑦𝑁𝑂 ,𝑠 + 𝐷8 𝑦𝑁𝑂 ,𝑠 + 𝐷9
2 2 2
𝑦𝑁𝑂2,𝑠 ( 𝐷 ′5 𝑦𝑁𝑂2 +𝐷 ′ 2 𝐷2 )+ 𝑦𝑁𝑂2 𝐷 ′ 1 + 𝑦𝑁𝑂𝑖𝑛 𝐷 ′ 2
𝑖𝑛 𝑖𝑛
⇒ 𝑦𝑁𝑂2 ,𝑠 = 𝐷′3 (𝐷 ′ 1+ 𝐷′ 5 𝑦𝑁𝑂2,𝑠 )+𝐷′ 4 (𝑦𝑁𝑂𝑖𝑛 +𝐷2 𝑦𝑁𝑂2,𝑠 )
(91)

Here, we further define the following intermediate variables:

𝐷′1 = 1 + 𝐷1 − 𝐷3 + 𝐷5 + 𝐷6 + 𝐷9 (92)
𝐷′2 = 𝐷10 + 𝐷17 (93)
𝐷′3 = 1 + 𝐷11 + 𝐷14 + 𝐷15 + 𝐷16 (94)
𝐷′4 = 𝐷12 + 𝐷13 (95)
𝐷′5 = −𝐷4 + 𝐷7 + 𝐷8 (96)

Now, rearranging equation (91) leads to a quadratic equation in the mole fraction of NO2:
2 (𝐷′3 𝐷′5 + 𝐷′4 𝐷2 ) + 𝑦𝑁𝑂2,𝑠 (𝐷′3 𝐷′1 + 𝐷′4 𝑦𝑁𝑂𝑖𝑛 − 𝐷′ 5 𝑦𝑁𝑂2 − 𝐷′2 𝐷2 ) − (𝑦𝑁𝑂2 𝐷′1 + 𝑦𝑁𝑂𝑖𝑛 𝐷′2 ) = 0
𝑦𝑁𝑂2,𝑠 𝑖𝑛 𝑖𝑛
(97)

The coefficients of this quadratic equation are defined as:

𝑜 = 𝐷′3 𝐷′5 + 𝐷′4 𝐷2 (98)



𝑝 = 𝐷′3 𝐷′1 + 𝐷′4 𝑦𝑁𝑂𝑖𝑛 − 𝐷 5 𝑦𝑁𝑂2 𝑖𝑛 − 𝐷′2 𝐷2 (99)

𝑞 = −(𝑦𝑁𝑂2 𝑖𝑛 𝐷 ′1 + 𝑦𝑁𝑂𝑖𝑛 𝐷 ′ 2 ) (100)

Generalizing this solution for 𝑛 CSTRs in series with 𝑧 = 1,2 … 𝑛 from time 𝑡 to 𝑡 + 1 gives us:

𝑜𝑡+1,𝑧 = 𝐷′3𝑡+1,𝑧 𝐷′5𝑡+1,𝑧 + 𝐷′4𝑡+1,𝑧 𝐷2𝑡+1,𝑧 (101)


𝑝𝑡+1,𝑧 = 𝐷′3𝑡+1,𝑧 𝐷′1𝑡+1,𝑧 + 𝐷′4𝑡+1,𝑧 𝑦𝑁𝑂,𝑔𝑡+1,𝑧−1 − 𝐷 ′ 5 𝑦𝑁𝑂2 ,𝑔𝑡+1,𝑧−1 − 𝐷′2𝑡+1,𝑧 𝐷2𝑡+1,𝑧 (102)
𝑞𝑡+1,𝑧 = −(𝑦𝑁𝑂2 ,𝑔𝑡+1,𝑧−1 𝐷 ′1𝑡+1,𝑧 + 𝑦𝑁𝑂,𝑔𝑡+1,𝑧−1 𝐷 ′ 2𝑡+1,𝑧 ) (103)
The real root of the quadratic equation (97) gives the final solution for the NO2 solid-phase mole fraction:
2
−𝑝𝑡+1,𝑧 + √𝑝𝑡+1,𝑧 −4𝑜𝑡+1,𝑧 𝑞𝑡+1,𝑧
𝑦𝑁𝑂2 𝑠 = 2𝑜𝑡+1,𝑧
(104)
𝑡+1,𝑧

Using this solution for NO2 in equation (76) leads to the piecewise analytical solution for the NO solid-
phase mole fraction:

𝑦𝑁𝑂,𝑔𝑡+1,𝑧−1 +𝐷2𝑡+1,𝑧 𝑦𝑁𝑂2 ,𝑠


𝑡+1,𝑧
𝑦𝑁𝑂,𝑠𝑡+1,𝑧 = (105)
1+𝐷1𝑡+1,𝑧 − 𝐷3𝑡+1,𝑧 − 𝐷4𝑡+1,𝑧 𝑦𝑁𝑂2 ,𝑠𝑡+1,𝑧 + 𝐷5𝑡+1,𝑧 + 𝐷6𝑡+1,𝑧 + 𝐷7𝑡+1,𝑧 𝑦𝑁𝑂2 ,𝑠𝑡+1,𝑧 + 𝐷8𝑡+1,𝑧 𝑦𝑁𝑂2 ,𝑠𝑡+1,𝑧 + 𝐷9𝑡+1,𝑧
NH3

Rewriting equation (62) for NH3 gives us:

𝑣𝑔 𝑑ℎ
4𝛿𝑤𝑐𝑎𝑣𝑔 ( +𝐿𝑘𝑚𝑜,𝑁𝐻3 )
4
𝑦𝑁𝐻3 ,𝑠 = 𝑦𝑁𝐻3 𝑖𝑛 + ∑𝐼𝑖 𝜈𝑖,𝑁𝐻3 . 𝑟𝑖 (106)
𝑣𝑔 𝑘𝑚𝑜,𝑁𝐻3 𝐶0 𝑑ℎ

Here, we define a mass-transfer intermediate variable, 𝐾𝑁𝐻3 , with the units of m3wc/mol-s:

𝑣𝑔 𝑑ℎ
4𝛿𝑤𝑐𝑎𝑣𝑔 ( +𝐿𝑘𝑚𝑜,𝑁𝑂2 )
4
𝐾𝑁𝐻3 = 𝑣𝑔 𝑘𝑚𝑜,𝑁𝐻3 𝐶0 𝑑ℎ
(107)

We expand the reaction rates on the right-hand side of equation (106) as shown below:

𝑦𝑁𝐻3 ,𝑠 = 𝑦𝑁𝐻3 𝑖𝑛 + 𝐾𝑁𝐻3 (−𝑟𝐴𝑑𝑠 + 𝑟𝐷𝑒𝑠 ) (108)

⇒ 𝑦𝑁𝐻3 ,𝑠 = 𝑦𝑁𝐻3 𝑖𝑛 + 𝐾𝑁𝐻3 (−𝑘𝐴𝑑𝑠 𝑦𝑁𝐻3 ,𝑠 𝜃𝑆1 𝛺1 + (𝑘𝐷𝑒𝑠1 + 𝑘𝐷𝑒𝑠2 )𝜃𝑁𝐻3 𝑆1 𝛺1 ) (109)

Rearranging equation (109), and generalizing this solution for 𝑛 CSTRs in series with 𝑧 = 1,2 … 𝑛 from
time 𝑡 to 𝑡 + 1 gives us the piecewise analytical solution for NH3 solid-phase mole fraction:

𝑦𝑁𝐻3 ,𝑔𝑡+1,𝑧−1 + 𝐾𝑁𝐻3 (𝑘𝐷𝑒𝑠1𝑡+1,𝑧 +𝑘𝐷𝑒𝑠2𝑡+1,𝑧 )𝜃𝑁𝐻3 𝑆1𝑡+1,𝑧 𝛺1


𝑡+1,𝑧
𝑦𝑁𝐻3 ,𝑠𝑡+1,𝑧 = 1+𝐾𝑁𝐻3 𝑘 𝜃 𝛺
(110)
𝑡+1,𝑧 𝐴𝑑𝑠 𝑆1𝑡+1,𝑧 1

N2O

Rewriting equation (62) for N2O gives us:

𝑣𝑔 𝑑ℎ
4𝛿𝑤𝑐𝑎𝑣𝑔 ( +𝐿𝑘𝑚𝑜,𝑁2 𝑂 )
4
𝑦𝑁2 𝑂,𝑠 = 𝑦𝑁2 𝑂𝑖𝑛 + ∑𝐼𝑖 𝜈𝑖,𝑁2 𝑂 . 𝑟𝑖 (111)
𝑣𝑔 𝑘𝑚𝑜,𝑁2 𝑂 𝐶0 𝑑ℎ

Here, we define a mass-transfer intermediate variable, 𝐾𝑁2 𝑂 , with the units of m3wc/mol-s:

𝑣𝑔 𝑑ℎ
4𝛿𝑤𝑐𝑎𝑣𝑔 ( +𝐿𝑘𝑚𝑜,𝑁2 𝑂 )
4
𝐾𝑁2 𝑂 = (112)
𝑣𝑔 𝑘𝑚𝑜,𝑁2 𝑂 𝐶0 𝑑ℎ

We expand the reaction rates on the right-hand side of equation (111) as shown below:

𝑦𝑁2 𝑂,𝑠 = 𝑦𝑁2 𝑂𝑖𝑛 + 𝐾𝑁2 𝑂 (4𝑟𝑆𝑆𝐶𝑅𝑁 + 1.5𝑟𝐹𝑆𝐶𝑅𝑁 + 𝑟𝑁𝑆𝐶𝑅𝑁 + 𝑟𝐴𝑁𝑑𝑒𝑐𝑜𝑚𝑝𝑁 𝑂 ) (113)
2𝑂 2𝑂 2𝑂 2

As the right-hand side of equation (113) is completely known, this solution can directly be generalized for
𝑛 CSTRs in series with 𝑧 = 1,2 … . . 𝑛 from time 𝑡 to 𝑡 + 1, giving us the piecewise analytical solution for
N2O solid-phase mole fraction:
𝑦𝑁2 𝑂,𝑠𝑡+1,𝑧 = 𝑦𝑁2 𝑂𝑡+1,𝑧−1 + 𝐾𝑁2 𝑂𝑡+1,𝑧 (4𝑘𝑆𝑆𝐶𝑅𝑁2 𝑂 𝑦𝑁𝑂,𝑠𝑡+1,𝑧 𝑦𝑂0.32
2
𝜃𝑁𝐻3 𝑆1𝑡+1,𝑧 𝛺1 +
𝑡+1,𝑧

𝑘𝐹𝑆𝐶𝑅𝑁2 𝑂 𝑦𝑁𝑂,𝑠𝑡+1,𝑧 𝑦𝑁𝑂2 ,𝑠𝑡+1,𝑧 𝜃𝑁𝐻3 𝑆1𝑡+1,𝑧 𝛺1 + 𝑘𝑁𝑆𝐶𝑅𝑁2 𝑂 𝑦𝑁𝑂2 ,𝑠𝑡+1,𝑧 𝜃𝑁𝐻3 𝑆1𝑡+1,𝑧 𝛺1 +
𝑡+1,𝑧 𝑡+1,𝑧

𝑘𝐴𝑁𝑑𝑒𝑐𝑜𝑚𝑝𝑁2 𝑂 𝜃𝐴𝑁𝑆1𝑡+1,𝑧 𝛺1 ) (114)


𝑡+1,𝑧

These solid-phase species mole fractions are then used to calculate the gas-phase mole fractions using
equation (57). Generalizing this solution for 𝑛 CSTRs in series with 𝑧 = 1,2 … 𝑛 from time 𝑡 to 𝑡 + 1 leads
to:

𝑣𝑔 𝑦𝑙,𝑔 4𝑘𝑚𝑜,𝑙 𝑦𝑙,𝑠


𝑡+1,𝑧−1 𝑡+1,𝑧
+
∆𝑧 𝑑ℎ
𝑦𝑙,𝑔𝑡+1,𝑧 = 𝑣𝑔 4𝑘𝑚𝑜,𝑙 ; 𝑙 = 𝑁𝑂, 𝑁𝑂2 , 𝑁𝐻3 , 𝑁2 𝑂 (115)
( + )
∆𝑧 𝑑ℎ

Finally, the reaction rates (in Table 1 of the main manuscript) are updated using the solid-phase species
mole fraction.

4. NO Oxidation, NH3 Oxidation and Dynamic NOx Conversion


NO oxidation experimental data (in the absence of feed NO2) as a function of temperature (red dots in
Figure S5) shows very low oxidation rates, as expected on Cu-SSZ-13. Both the high-fidelity and reduced-
order models can successfully describe this NO oxidized fraction, with an activation energy of ~40 kJ/mol.

Figure S5. NO Oxidation at 73k/h GHSV, NO2/NOx = 0 and 930 ppm NOx. Dots (red) represent experimental data,
and solid lines represent model results (High-fidelity model in red, Reduced-order model in blue)
Figure S6 shows the NH3 oxidation in the absence of NOx as a function of temperature, derived from Step-
3 of the 4-Step protocol [5]. Degreened Cu-SSZ-13 catalysts show an unusual NH3 oxidation behavior,
characterized by an initial steep increase in the oxidation rate around 300°C, followed by a plateauing up
to 400°C and a sharp increase above 400°C. This is likely associated with distinct active sites responsible
for low and high temperature NH3 oxidation, with varying activation energies [6]. NH3 oxidation could be
modeled reasonably well by considering two activation energies on a single-site high-fidelity model, as
demonstrated in Figure S6 (solid red line). The reduced-order model (solid blue line in Figure S6) showed
nearly identical NH3 conversion relative to the high-fidelity model.

Figure S6. NH3 Oxidation at 73k/h GHSV and 9930 ppm NH3. Dots (red) represent experimental data, and solid
lines represent model results (High-fidelity model in red, Reduced-order model in blue)

Figure S7 shows the model predicted build-up of NOx conversion as a function of NH3 storage at 190°C.
As expected, the NOx conversion increases monotonically with increased NH3 storage. Both the high-
fidelity and reduced-order models shows a very similar dynamic NOx conversion profile, validating the
approximations used to derive the reduced-order model solution.
Figure S7. 190°C NOx Conversion as a function of NH3 storage under standard SCR conditions (NO2/NOx = 0) at
73k/h GHSV, 930 ppm NOx and 1.06 ANR (NH3 to NOx ratio) . Solid lines represent model results (High-fidelity
model in red, Reduced-order model in blue)

Figure S8 shows the N2O slip observed under standard SCR conditions. Using the same activation energy
as standard SCR (86 kJ/mol) led to accurate high-fidelity model predictions below 300°C (solid red line in
Figure S8). However, the drop in N2O slip from 250°C onwards was not explained by the global model,
likely associated with decomposition of surface intermediates. The reduced-order model gave identical
results as the high-fidelity model (solid blue line in Figure S8).
Figure S8. N2O Slip under standard SCR conditions (NO2/NOx = 0) at 73k/h GHSV, 930 ppm NOx and 1.06 ANR
(NH3 to NOx ratio) . Dots (red) represent experimental data, and solid lines represent model results (High-fidelity
model in red, Reduced-order model in blue)

5. NH3 and AN coverage during low temperature Fast SCR


Figure S9 shows the mean NH3 and AN coverage at 150°C and 180°C under fast SCR conditions. The
inhibition of fast SCR caused by AN accumulation is captured through a competitive adsorption
mechanism, which leads to a drop in NH3 coverage (as observed in Figure S9) and a corresponding loss in
NOx conversion. The reduced-order model produces the same transient coverage as the high-fidelity
model.

A B

Figure S9. A. 150°C and B. 180°C NH3 and AN coverage under fast SCR conditions (NO2/NOx = 0.5) at 130k/h
GHSV, 990 ppm NOx and 1 ANR . Solid lines represent model predicted mean NH3 (High-fidelity model in red,
Reduced-order model in dark red) and mean AN coverage (High-fidelity model in blue, Reduced-order model in
dark blue)

6. NO2/NOx Sweep
A NO2/NOx sweep reactor experiment was conducted at 250°C to validate the kinetic model, and estimate
the NOx and N2O make from slow SCR. Figure S10 shows the inlet NOx and NH3 mole fractions, as the
NO2/NOx ratio is varied from 0 to 0.7
Figure S10. 250°C NO2/NOx sweep test at 130k/h GHSV, 990 ppm NOx and 1ANR. Dashed lines represent NO
(red), NO2 (blue) and NH3 (green) inlet mole fractions

Figures S11 and S12 plot the transient NO, NO2, NH3 and N2O slip during the NO2/NOx sweep test. As
expected, the reactor data shows a progressive increase in NOx conversion with increasing NO2/NOx up to
0.5. A further increase in NO2/NOx to 0.7 leads to a significant increase in NH3 and NOx slip, due to
decrease of overall SCR rate. Both the high-fidelity and reduced-order models successfully capture this
trend. However, NOx conversion and NH3 slip is overpredicted by ~10% at NO2/NOx equal to 0.7,
requiring further kinetic model improvements. The N2O slip in Figure S12 monotonically increases with
increasing NO2/NOx and is captured accurately by the high-fidelity (solid dark grey line) and reduced
order models (solid light grey line).
Figure S11. Outlet NOx and NH3 mole fractions during 250°C NO2/NOx Sweep Test at 130k/h GHSV, 990 ppm
NOx and 1 ANR. Solid lines represent experimental data (NO : red, NO2 : green, NH3 : blue), high-fidelity model
results (NO : dark red, NO2 : dark green, NH3 : dark blue) and reduced order model results (NO : dark yellow, NO 2 :
dark cyan, NH3 : dark magenta)

Figure S12. Outlet N2O during 250°C NO2/NOx Sweep Test at 130k/h GHSV, 990 ppm NOx and 1 ANR. Solid
lines represent experimental data (black), high-fidelity model results (dark grey) and reduced order model results
(light grey)
References
[1] Hayes, R. E., & Kolaczkowski, S. T. (1998). Introduction to catalytic combustion. CRC Press.

[2] Joshi, S. Y., Harold, M. P., & Balakotaiah, V. (2009). On the use of internal mass transfer coefficients in modeling of diffusion
and reaction in catalytic monoliths. Chemical Engineering Science, 64(23), 4976-4991.

[3] Fuller, E. N., Schettler, P. D., & Giddings, J. C. (1966). New method for prediction of binary gas-phase diffusion
coefficients. Industrial & Engineering Chemistry, 58(5), 18-27.

[4] Joshi, S. Y., Kumar, A., Luo, J., Kamasamudram, K., Currier, N. W., & Yezerets, A. (2018). New insights into the mechanism
of NH3-SCR over Cu-and Fe-zeolite catalyst: Apparent negative activation energy at high temperature and catalyst unit design
consequences. Applied Catalysis B: Environmental, 226, 565-574.

[5] Kamasamudram, K., Currier, N. W., Chen, X., & Yezerets, A. (2010). Overview of the practically important behaviors of
zeolite-based urea-SCR catalysts, using compact experimental protocol. Catalysis Today, 151(3-4), 212-222.

[6] Olsson, L., Wijayanti, K., Leistner, K., Kumar, A., Joshi, S. Y., Kamasamudram, K., ... & Yezerets, A. (2015). A multi-site
kinetic model for NH3-SCR over Cu/SSZ-13. Applied Catalysis B: Environmental, 174, 212-224.

You might also like