**1. Introduction**

Sustainable solutions are required to reduce the greenhouse gas emissions of the transportation and industrial sectors, shrinking the dependency on fossil fuels. With the continuously increasing installed capacity of wind and solar power plants [1], adequate energy storage solutions have to be implemented in order to deal with their fluctuating nature. Therefore, the conversion of electricity into valuable chemicals and fuels, a concept often called power-to-fuels or power-to-X, can make an important contribution to the future energy system [2]. In this context, key process steps are hydrogen generation via electrolysis (primary conversion) [3,4] and methanol synthesis (secondary conversion) [5].

Methanol can be used as a fuel substitute or additive, either in fuel cells or via combustion [2], as a feedstock in the production of base chemicals (e.g., formaldehyde) and liquid fuels (e.g., gasoline, oxymethylene ethers, jetfuel) [6–8], and also as a solvent. Methanol fuel has recently attracted significant interest, especially in China, where the consumption of methanol for thermal applications (e.g., boilers, kilns, cooking stoves) and in the transportation section amounted to 5.7 Mton·a−<sup>1</sup> (year 2019) [9].

The current global capacity of methanol production is 164 Mton·a−<sup>1</sup> (year 2021), with an annual increase of 10% projected for the next decade [10]. Traditionally, methanol synthesis is fed by fossil-based syngas, which comes either from steam reforming of natural gas or from coal gasification [11]. However, sustainable syngas production is gaining

**Citation:** Lacerda de Oliveira Campos, B.; John, K.; Beeskow, P.; Herrera Delgado, K.; Pitter, S.; Dahmen, N.; Sauer, J. A Detailed Process and Techno-Economic Analysis of Methanol Synthesis from H2 and CO2 with Intermediate Condensation Steps. *Processes* **2022**, *10*, 1535. https://doi.org/10.3390/ pr10081535

Academic Editors: Elio Santacesaria, Riccardo Tesser and Vincenzo Russo

Received: 13 July 2022 Accepted: 3 August 2022 Published: 5 August 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

importance, such as from renewable electricity and captured CO2, and also from biomass. In Figure 1, a scheme is presented showing the intermediate position of methanol in the conversion of both fossil-based and sustainable syngas to added-value chemicals and fuels, as well as methanol end-use applications.

**Figure 1.** The key position of methanol to convert syngas sources into added-value chemicals and fuels. Icons from: Freepik, Flaticon [12].

The current world-scale technology for methanol synthesis is mostly based on the application of Cu/ZnO/Al2O3 (CZA) catalysts in either multi-tube reactors with boiling water as the cooling fluid, normally called isothermal reactors (e.g., the Lurgi process, the Linde process), or adiabatic reactors with intermediate cold syngas quenching, generally named quench reactors (e.g., ICI and the Casale process, the Haldor Topsoe process) [11,13]. Less common but also industrially applied are the adiabatic reactors with intermediate cooling (e.g., the Kellogg process, the Toyo process) [11,14]. Normally, temperatures between 200 and 300 ◦C and pressures between 50 and 100 bar are applied [13].

Methanol can be produced from either CO (Equation (1)) or CO2 (Equation (2)), with the reverse water–gas shift reaction (rWGSR, Equation (3)) also occurring. If the feed gas contains both CO and CO2, there is a prevailing opinion that direct CO hydrogenation (Equation (1)) on Cu/Zn-based catalysts is significantly slower than CO2 hydrogenation [15,16], and kinetic models not considering this reaction can adequately simulate experimental data [17–19].

$$\text{CO}\_{\text{(g)}} + 2\cdot\text{H}\_{2\text{(g)}} \rightleftharpoons \text{CH}\_{3}\text{OH}\_{\text{(g)}} \qquad \qquad \Delta H\_{25\text{ }^\circ \text{C}}^0 = -90.6 \text{ kJ}\cdot \text{mol}^{-1} \tag{1}$$

$$\text{CO}\_{2(g)} + 3\text{-H}\_{2(g)} \rightleftharpoons \text{CH}\_3\text{OH}^-\_{(g)} + \text{H}\_2\text{O}\_{(g)} \quad \Delta H^0\_{25\text{ }^\circ \text{C}} = -49.4 \text{ kJ}\cdot \text{mol}^{-1} \tag{2}$$

$$\text{CO}\_{2(g)} + \text{H}\_{2(g)} \rightleftharpoons \text{CO}\_{(g)} + \text{H}\_{2}\text{O}\_{(g)} \qquad \Delta H\_{25\text{ }^\circ \text{C}}^{0} = 41.2 \text{ kJ} \cdot \text{mol}^{-1} \tag{3}$$

With regard to the general use of CO2 as a carbon source, several process simulations and techno-economic analyses of methanol synthesis from green hydrogen and captured CO2 have been reported, with the CO2 source being either a cleaned industrial flue gas or concentrated atmospheric CO2 (i.e., through carbon capture units, CCUs) [20–22]. Pérez-Fortes et al. [20] and Szima et al. [21] simulated a methanol synthesis plant with CZA, having a total production of 440 and 100 kton MeOH·a−1, respectively. Heat integration was considered in both studies, with the plant being energetically self-sufficient. Cordero-Lanzac et al. [22] simulated the production of 275 kton MeOH·a−<sup>1</sup> with an In2O3/Co catalyst. In all studies, it was concluded that economic viability can be achieved if reactant prices significantly decrease or if CO2 taxation is enforced. Nonetheless, it is expected that the costs of electrolysis and CCUs might considerably decrease in the foreseeable future [23,24], and the first industrial-scale plants to produce e-methanol and e-gasoline are expected to start operating in 2024–2025 [25,26].

A general argument is the thermodynamic restrictions of CO2 conversion to methanol compared to CO conversion (see Figure 2), whereby only limited methanol yields are achievable even at elevated pressures and lower temperatures. Consequently, a low CO2 single-pass conversion (*XCO*2,*SP*) is obtained independently of the reactor size, leading to large recycle streams, which increase operating costs and cause higher reactant losses in purge streams.

**Figure 2.** Methanol equilibrium yield as a function of temperature and pressure. Data generated with Aspen Plus. (**a**) H2/CO feed in a 2:1 ratio. (**b**) H2/CO2 feed in a 3:1 ratio.

If the products (i.e., methanol and water) are removed from the reacting system, the thermodynamic equilibrium is shifted towards a higher methanol yield. This strategy has been studied using alternative reactor designs with in situ condensation [27,28] or membrane reactors [29], but these technologies are not yet ready for commercialization. A feasible approach using commercially proven technology is the implementation of intermediate condensation steps between reactor units displaced in series. In the Davy series loop methanol process, two reactors with an intermediate condensation unit are proposed for large scale methanol production from CO-rich syngas [13,30]. Although the implementation of intermediate condensation steps is a promising strategy to increase methanol yield from H2/CO2 syngas, such an approach has still not received particular attention, and plant simulations with heat integration and techno-economic analyses are not available in the literature yet, to the best of our knowledge.

In this work, the conventional approach (named here the 'one-step process') is compared with a new alternative approach including two condensation steps (named here the 'three-step process'). Using our recently developed kinetic model for methanol synthesis [19], both processes were implemented in Matlab in order to critically analyze and select key process parameters (i.e., cooling fluid temperature, number of reactor modules, and purge fraction). With the optimized parameters, detailed methanol synthesis plants with heat integration were implemented in Aspen Plus, and techno-economic analyses were performed.

#### **2. Methodology**

#### *2.1. Process Overview*

In the present work, a methanol synthesis plant from H2/CO2 with a production of 145 ton·h−<sup>1</sup> is considered. This value is based on an ongoing power-to-gasoline project via H2/CO2 conversion to methanol [26], whose final goal is a gasoline production of 5.5·108 <sup>L</sup>·a<sup>−</sup>1, which corresponds to a methanol production of 1.16 Mton·a−<sup>1</sup> or 145 ton·h−<sup>1</sup> (assuming a yield of 80% in the methanol-to-gasoline process and plant operating hours of 8000 h·a<sup>−</sup>1).

In our simulations, feed carbon dioxide comes from the cleaned flue gas of nearby industries (e.g., a cement industry) at 25 ◦C and 1 bar, with a purity of 99.5% mol/mol (the rest was water). Feed hydrogen comes from water electrolysis at 25 ◦C and 30 bar, with a purity of 99.5% mol/mol (the rest was nitrogen). Although it is possible to obtain these feedstocks in an extremely high purity (e.g., 99.99% mol/mol) [31,32], we chose a more conservative scenario, which also allows a proper simulation of inert material accumulation in the plant.

As pressure has a significant influence on the thermodynamic equilibrium of methanol synthesis (see Figure 2), the reactor operating pressure was set to 70 bar. Although higher pressures are reported to have potential in methanol synthesis [33,34], they were out of the scope of this work, since considerable extrapolations in the kinetic model would be necessary, and condensation inside the reactor might have to be taken into account. Besides, higher pressures increase compression costs and might also require more expensive materials to build the equipment.

The dimensions of the reactor modules were chosen to be close to the upper size limits that are currently commercially available. That is, each reactor module consisted of a shell containing 33,000 tubes with 12.5 m length and an inner diameter of 3.75 cm. Since the heat generation in CO2 hydrogenation is lower than in CO hydrogenation (Equations (1) and (2)), less heat transfer area is necessary. Because of that, the tube inner diameter chosen in this work (3.75 cm) was larger than the size typically used for CO conversion to methanol (2.5 cm). Considering 1050 kg·m−<sup>3</sup> as the apparent catalyst bed density [35], the total CZA catalyst loading of each reactor module was 478.13 ton. A total pressure loss of 0.75 bar was considered for each reactor module [36].

#### 2.1.1. One-Step Approach—Process Description

In Figure 3, a detailed flowsheet of the one-step process is presented. This is an adapted version from a concept reported in the literature [37–39]. Feed CO2 is mixed with a low-pressure recycle stream, and then compressed from 1 to 70 bar in a three-stage process, including intermediate cooling (reducing compression work) and intermediate phase separation (to remove condensed methanol and water from the recycle stream). The resulting compressed stream is mixed with feed H2 (compressed from 30 to 70 bar in one stage) and with a high pressure recycle stream. The mixed stream is preheated with the product gas and enters the inner tubes of parallel reactor modules, with the temperature being controlled by boiling water on the shell side.

The product stream is cooled down to 30 ◦C in four heat exchangers, condensing water, methanol, and some CO2, which are separated from the light gases in a flash drum. A fraction of the gas stream is purged, and the remaining stream is recompressed to 70 bar and recycled. The liquid stream from the flash drum is depressurized to 1 bar and heated to 30 ◦C, vaporizing most remaining CO2. A liquid–gas separation is performed in another flash drum. A fraction of the gas stream from the low-pressure flash drum is purged, and the rest is recycled by mixing with feed CO2. The liquid stream from the low-pressure flash drum is preheated and fed to a packed column, where methanol in high purity (>99.5% m/m) is recovered in the liquid distillate, water is recovered in the bottom, and most of the remaining CO2 is recovered in the gas distillate.

**Figure 3.** One-step process—detailed flowsheet with a total of three reactor modules. Cooling water streams are omitted.

The purge streams are burned with 15% air excess in a fired heater [40]. The heat of reaction of both the purge combustion and the methanol synthesis are used in a water Rankine cycle to produce electricity. The cycle starts with liquid water at 1 bar and 99.6 ◦C being pumped to a certain pressure, whose boiling temperature corresponds to the desired reactor temperature. Pressurized water reaches its boiling temperature in two steps (heat exchanger and fired heater) and vaporizes inside the reactor modules. The produced saturated steam is further heated in the fired heater and then performs work in a turbine, with a discharge pressure of 1.43 bar (*Tboiling* = 110 ◦C). The resulting low-pressure steam condenses partially in the column reboiler, and total condensation is completed in a heat exchanger, closing the water cycle.

#### 2.1.2. Three-Step Approach—Process Description

In Figure 4, a detailed flowsheet of the three-step process is presented. In this approach, the feed compression and recycling of non-converted reactants occurs similarly to the onestep process. The mixed feed stream is preheated and enters the first reactor module. The product gas is cooled down to 45 ◦C in three steps, and the condensed stream (mostly water, methanol, and some CO2) is separated from the light gases in a flash drum. The gas stream is preheated and enters the second reactor module. The second product gas is cooled down to 30 ◦C in three steps, and the condensed stream (mostly water, methanol, and some CO2) is separated from the gas stream in another flash drum.

**Figure 4.** Three-step process—detailed flowsheet with a total of three reactor modules. Cooling water streams are omitted.

The gas phase is preheated and enters the third reactor module. The third product gas is cooled down, mixed with the condensed streams from the first and second reaction stages, and further cooled down to 30 ◦C.

Similar to the one-step process, component separation of the product stream is performed with one flash drum at high pressure, one flash drum at ambient pressure, and one distillation column.

The purge stream is burned in a fired heater with preheated air. In the water cycle, pressurized water is preheated and distributed to the reactor modules. A fraction of the produced saturated steam is split and used to preheat the water while the remaining steam is further heated in the fired heater. Supersaturated steam performs work in a turbine, with a discharge pressure of 1.43 bar (*Tboiling* = 110 ◦C). The resulting low-pressure steam is partially condensed in the column reboiler, and total condensation is completed in a heat exchanger, closing the water cycle.

#### *2.2. Process Simulation in Matlab*

Before implementing the final version of each plant in Aspen Plus, different scenarios were investigated in Matlab. Therefore, optimal key parameters were selected, such as the total number of reactor modules, the purge fraction, and the temperature of the cooling fluid in the reactor.

In order to simulate the reactor, the following considerations were made: there are only variations along the length of the reactor (1D assumption), the influence of back-mixing is neglected (plug flow assumption), and the cooling fluid temperature (*Tw*, in K) is constant. Starting from mass and energy balances, the differential equations of the total mole flow of a single tube ( . *<sup>n</sup>*, in mol·s<sup>−</sup>1), the mole fraction of each component *<sup>j</sup>* (*yj*), and the temperature (*T*, in K) in the axial direction *z* are shown as follows:

$$\frac{\mathbf{d}\dot{m}}{\mathbf{d}z} = \frac{m\_{\rm Cat}}{L} \cdot \sum\_{j=1}^{6} \sum\_{k=1}^{2} (\boldsymbol{\nu}\_{jk} \cdot \boldsymbol{r}\_{k})\tag{4}$$

$$\frac{\mathrm{d}y\_{\circ}}{\mathrm{d}z} = \frac{1}{\dot{n}} \cdot \left\{ \frac{m\_{\mathrm{Cat}}}{L} \cdot \sum\_{k=1}^{2} (\nu\_{\mathrm{jk}} \cdot r\_{k}) - y\_{\circ} \cdot \frac{\mathrm{d}\dot{n}}{\mathrm{d}z} \right\} \tag{5}$$

$$\frac{\mathrm{d}T}{\mathrm{d}z} = \frac{1}{\left(\dot{n} \cdot \mathbb{C}\_{P,f}\right)} \cdot \left[ -\frac{\mathrm{d}\dot{n}}{\mathrm{d}z} \cdot \mathrm{h}\_f - \dot{n} \cdot \sum\_{j=1}^6 \left( h\_j \cdot \frac{\mathrm{d}y\_j}{\mathrm{d}z} \right) + \mathrm{l}I \cdot \pi \cdot D\_{\dot{i}} \cdot \left( T\_w - T \right) \right] \tag{6}$$

where *mCat* is the catalyst mass (kg), *L* is the reactor length (m), *yjk* is the stoichiometric coefficient of component *<sup>j</sup>* in reaction *<sup>k</sup>*, *rk* is the rate of reaction *<sup>k</sup>* (mol·kgcat·s<sup>−</sup>1), *CP*, *<sup>f</sup>* is the heat capacity of the fluid (J·mol·K<sup>−</sup>1), *hf* is the specific enthalpy of the fluid (J·mol<sup>−</sup>1), *hj* is the specific enthalpy of component *<sup>j</sup>*, *<sup>U</sup>* is the global heat transfer coefficient (W·m−2·K<sup>−</sup>1), and *Di* is the inner diameter of a single tube (m).

The temperature-dependent parameters (*CP*, *<sup>f</sup>* , *hf* , *hj*, *U*) were updated in each integration point in the axial direction. Heat capacity and enthalpy were calculated with the thermodynamic functions provided by Goos et al. [41], which are detailed in the Supplementary Material (Section A) along with the derivation of the differential equations. The global heat transfer coefficient was estimated (*U*) by summing the heat transfer resistances in the axial direction, according to the methodology described in the literature [42,43] (see Section B of the Supplementary Material).

The methodology to calculate the reaction rates (*rk*) is described in Section 2.3. The system of differential equations was solved with the Matlab function ode45, with absolute and relative tolerances set to 10<sup>−</sup>10.

In order to simplify the simulation of the separation steps in Matlab, the following procedure was applied. Both processes were implemented in Aspen Plus, considering a total of six reactor modules, a purge fraction of 2%, and *Tw* = 235 ◦C. The values of the split ratio of each component in the liquid and gas phase of each flash drum and the distillation column were extracted. For example, in the column of the one-step process, the methanol distribution in the outlet streams was: 3.80% in the gas distillate, 96.13% in the liquid distillate, and 0.06% in the bottom. The split ratio of all the components were taken from Aspen Plus and were considered constant for the different scenarios investigated in Matlab (i.e., variations in the number of reactor modules, purge fraction, and *Tw*). These split ratios are provided in the Supplementary Material (Section C).

Flowsheet convergence was achieved in Matlab by an iterative method, as there were two cycles of streams due to recycling unconverted reactants. First, educated initial guesses of the composition and total mole flow of each recycle stream were given. In each iteration, the recycle stream mole flow and its composition were calculated and used in the next iteration until the tolerance criterion was fulfilled:

$$\frac{\left(\dot{n}\_{\mathrm{Rcf},k+1} - \dot{n}\_{\mathrm{Rcf},k}\right)^2}{\left(\dot{n}\_{\mathrm{Rcf},k+1}\right)^2} \le \text{Tolerance} \tag{7}$$

where . *nRef* ,*<sup>k</sup>* is the total mole flow of the recycle stream at iteration *k*. The tolerance of the inner cycle and the outer cycle were set to 10−<sup>9</sup> and 10<sup>−</sup>8, respectively.

#### *2.3. Kinetic Modeling of the Methanol Synthesis*

The kinetic simulation of the methanol synthesis was performed with our previously published six-parameter model (Model-6p) [19], whose considerations regarding the reaction mechanism, the assumption of the rate determining steps, and the most abundant surface species were based on our detailed microkinetic model [15]. This six-parameter model was validated with 496 experimental points from different laboratory plants [15,18,44], which contemplated temperatures between 210 and 260 ◦C, pressures between 20 and 60 bar, gas hourly space velocities (GHSV) between 1.8 and 40 LS·h−1·gcat−1, and a variety of syngas (H2/CO/CO2/N2) feed compositions, including 126 points with only H2/CO2/N2 in feed.

In this model, two main reactions are considered: CO2 hydrogenation (Equation (2)) and the rWGSR (Equation (3)). The reaction rates (*rCO*<sup>2</sup> *hyd*., *rrWGSR*) in mol·kgcat−1·s−<sup>1</sup> are described as follows:

$$\left(r\_{\text{CO}\_2\text{ hyd.}} = \exp\left(A\_2 - \frac{E\_{A,2}}{R \cdot T}\right) \cdot \phi\_{\text{Zn}} \cdot \theta\_b \cdot \theta\_c \cdot f\_{H\_2}^{1.5} \cdot f\_{\text{CO}\_2} \cdot \left(1 - \frac{f\_{\text{CH}\_3\text{OH}^-} f\_{\text{H}\_2\text{O}}}{f\_{\text{H}\_2}^3 \cdot f\_{\text{CO}\_2} \cdot K\_{\text{P,CO}\_2\text{ hyd.}}^0}\right) \right)$$

$$r\_{rWGSR} = \exp\left(A\_3 - \frac{E\_{A,3}}{R \cdot T}\right) \cdot \Phi\_{\text{Zn}} \cdot \theta\_b \cdot \theta\_c \cdot f\_{\text{CO}\_2} \cdot f\_{H\_2O} \cdot \left(1 - \frac{f\_{\text{CO}} \cdot f\_{H\_2O}}{f\_{H\_2} \cdot f\_{\text{CO}\_2} \cdot K\_{P,rWGSR}^0}\right) \tag{9}$$

Here, *A*2−<sup>3</sup> and *EA*,2−<sup>3</sup> are kinetic parameters, *R* is the universal gas constant, *φZn* is the zinc coverage on the surface, *θ<sup>b</sup>* and *θ<sup>c</sup>* are the free active sites b and c, *fj* is the fugacity of gas component *j* (bar), and *K*<sup>0</sup> *<sup>P</sup>*,*<sup>k</sup>* is the global equilibrium constant of reaction *k*.

The Peng–Robinson equation of state is used to calculate the fugacities [45], considering the binary interaction parameters reported by Meng et al. [46] and Meng and Duan [47], and an effective hydrogen acentric factor of −0.05 proposed by Deiters et al. [48].

The zinc coverage is dependent on temperature and gas concentration [49], and its exact quantification under reaction conditions is difficult to predict. The zinc coverage is then considered to be constant and equal to *φZn* = 0.50 for a general case, while it is set to *φZn* = 0.10 for CO2-rich feed (CO2/COx > 0.90). The free active sites are calculated with the following equations:

$$\theta\_b = \left(\overline{\mathcal{K}\_2} \cdot f\_{H\_2}^{0.5} \cdot f\_{\text{CO}\_2} + 1\right)^{-1} \tag{10}$$

$$\theta\_c = \left(\overline{\mathcal{K}\_3} \cdot f\_{H\_2}^{-0.5} \cdot f\_{H\_2O} + 1\right)^{-1} \tag{11}$$

where *K*2−<sup>3</sup> are adsorption parameters. In Table 1, the equilibrium constants as well as the previously estimated kinetic and adsorption parameters are summarized [19].

The side products of methanol synthesis on Cu/Zn-based catalysts (e.g., hydrocarbons or dimethyl ether) are typically at low concentrations [13,50]. Several studies reported that syngas conversion to hydrocarbons or dimethyl ether on commercial CZA at moderate temperatures (T ≤ 260 ◦C) is significantly low or even below detection range [15,18,44], while Condero-Lanzac et al. [22] reported low methane production from H2/CO2 on CZA at high temperatures (T ≥ 275 ◦C). Saito et al. [51] observed that side product formation is further reduced by increasing CO2/COX feed concentration. Therefore, the generation of side products is not considered in this work.


**Table 1.** Equilibrium constants, kinetic and adsorption parameters of Model-6p [19]. Reprinted with permission from [19]. Copyright 2021 American Chemical Society.

#### *2.4. Process Analysis and Optimization*

Considering the fixed methanol production of 145 ton·h−<sup>1</sup> or 1257.1 mol·s−<sup>1</sup> and the 99.5% mol/mol purity of the reactants, the minimum required feed is 1263.4 mol·s−<sup>1</sup> of CO2 and 3790.3 mol·s−<sup>1</sup> of H2, totalizing . *nf eed*,*min* = 5053.7 mol·s<sup>−</sup>1. Since there are reactant losses in the purge and product streams, an excess of feed is required. With a fix feed ratio H2:CO2 of 3:1, the excess of feed (*Exc*) is defined here as:

$$Exc = \frac{\left(\dot{n}\_{feed} - \dot{n}\_{feed,min}\right)}{\dot{n}\_{feed,min}} \cdot 100\% \tag{12}$$

It is, of course, of interest to minimize feed consumption, due to its high costs. Feed consumption is affected by key variables, such as reactor temperature and pressure, the number of reactor modules (which defines the total catalyst mass), and purge fraction. Avoiding large recycle streams is also important, as compressor work is required to get the pressure back to 70 bar, and larger equipment (i.e., heat exchangers, flash drums, compressors) are required to process higher flows.

Simulations were performed for a different number of reactor modules (from 3 to 12) and different purge fractions (from 0.5 to 5%). For each case, an initial guess for the feed excess was given (*Exc* = 5%), and a fix feed ratio H2:CO2 of 3:1 (a stoichiometric ratio) was applied. Then, an optimization problem was solved in Matlab with the function fminsearch (function tolerance = 0.1 mol·s<sup>−</sup>1, step tolerance = 0.1 ◦C), whose objective was to maximize methanol production by varying the reactor coolant temperature (*Tw*).

With the optimum *Tw*, the required excess of feed was calculated to meet the methanol demand (1257.1 mol·s−1) with Newton's method (function tolerance: 0.1 mol·s−1). The steps of the temperature optimization and *Exc* calculations were repeated until the temperature update was lower than 0.25 ◦C.
