Evidence-Grounded Multicriteria Prioritization of Methylcyclohexane Dehydrogenation Catalysts Using CRITIC–TOPSIS

Abstract

Liquid organic hydrogen carriers require catalysts that balance catalytic performance, operating severity, stability-related properties, and cost. This study applies a CRITIC-TOPSIS framework based on the literature to prioritize nine catalysts for methylcyclohexane dehydrogenation under a nominal scenario targeting 18,000 t H₂ yr⁻¹. Eight source-reported descriptors were used: operating temperature, H₂-release rate, fresh-catalyst cost, process cost, MCH conversion, reaction or test time, BET surface area, and toluene selectivity. The CRITIC weights reproduced from the supplied calculation annex were highest for selectivity (18.60%), BET surface area (15.72%), and reported time (13.95%). TOPSIS ranked 0.4 wt.% Pt/CB first and 1 wt.% Pt/AC second, with closeness coefficients of 66.09% and 58.94%, respectively. The result is a screening outcome rather than an intrinsic kinetic comparison because the source experiments were not harmonized and the reporting bases for release rate and reaction time were not uniform. Activation energy, turnover frequency, and active-site-specific activity were not calculated because common kinetic and site-density data were unavailable. No formal sensitivity analysis was performed because the supplied sources contained neither evidence-based uncertainty intervals nor a documented perturbation protocol. The process-cost criterion was strongly correlated with operating temperature (r = 0.959), indicating partial duplication of operating-severity information. The ranking therefore identifies candidates for standardized validation and does not establish universal catalyst superiority.

1. Introduction

Hydrogen can connect renewable-electricity production with industrial, transport, and energy-storage demand, but its low volumetric density complicates large-scale storage and long-distance transport. Liquid organic hydrogen carriers (LOHCs) address this constraint by reversibly binding hydrogen in liquids that can be handled with infrastructure similar to that used for conventional fuels [1,2]. The carrier is hydrogenated before transport and dehydrogenated at the point of use, where hydrogen and the hydrogen-lean liquid are recovered (Figure 1).

The methylcyclohexane–toluene (MCH-TOL) pair is widely studied because both compounds are liquid under practical handling conditions, the carrier has a hydrogen capacity of approximately 6.2 wt.%, and the chemistry is compatible with established liquid-fuel logistics [3,4]. Its main operational difficulty is the endothermic dehydrogenation step. Catalyst choice affects conversion, selectivity, operating temperature, hydrogen-release rate, deactivation, and the cost of releasing hydrogen.

Pt-based catalysts dominate the available MCH-dehydrogenation literature, but reported performance depends on metal loading, promoter, support, dispersion, pretreatment, reactor configuration, temperature, pressure, feed composition, catalyst mass, and time basis [5,6,7,8,9]. A catalyst that performs well according to conversion may require a higher temperature or have a less favorable cost. Conversely, a low-temperature alternative may have a lower reported release rate. A single metric therefore cannot represent the full decision problem.

Multicriteria decision-making has been used for catalyst selection in other reaction systems. Noyan Tekeli et al. [10] compared decision methods for catalyst selection in formic-acid production, and Nath et al. [11] used VIKOR to rank heterogeneous catalysts for biodiesel production. The present work does not propose a new MCDM algorithm. Its contribution is a traceable screening application for MCH dehydrogenation that integrates the technical descriptors and cost inputs recorded in the original study and calculation annex.

CRITIC was selected because the matrix is quantitative and no formal expert-elicitation record was available. It derives weights from criterion dispersion and intercriterion correlation [12]. TOPSIS then ranks alternatives according to their distance from positive and negative ideal solutions and can accommodate benefit and cost criteria [13]. This combination is fit for the supplied dataset, but it is not method-independent and does not eliminate uncertainty inherited from the source studies.

Figure 1. MCH-TOL liquid-organic-hydrogen-carrier pathway from hydrogenation and transport to endothermic dehydrogenation and hydrogen use, adapted from Rao and Yoon [14].

Figure 1. MCH-TOL liquid-organic-hydrogen-carrier pathway from hydrogenation and transport to endothermic dehydrogenation and hydrogen use, adapted from Rao and Yoon [14].

Accordingly, the objective was to reproduce and critically interpret the CRITIC-TOPSIS prioritization of nine MCH-dehydrogenation catalysts without introducing new measurements, inferred values, or unreported uncertainty ranges. Particular attention was given to data comparability, the meaning of screening-level descriptors, the dependence between operating temperature and process cost, and the limits of the resulting ranking.

2. Materials and Methods

2.1. Study Design and Evidence Basis

The study is documentary and quantitative. The original work searched Scopus and ScienceDirect and followed references to identify experimental studies of catalysts used for MCH dehydrogenation. The retained catalyst set was restricted to alternatives represented by a complete row in the supplied decision matrix. Missing technical values were not imputed. The nine-row calculation annex was treated as the controlling numerical source for the CRITIC weights and TOPSIS ranking.

The original source did not preserve a complete search string, search-date log, or a study-screening flow. The literature acquisition should therefore be interpreted as a targeted review rather than a systematic review. The retained set is a data-qualified subset of the catalyst literature and may favor catalyst families that are more extensively characterized.

2.2. Evaluation Criteria and Screening-Level Descriptors

The eight descriptors were grouped as techno-economic feasibility, catalytic performance, and stability-related proxies (Figure 2). A screening-level descriptor is a value available for every retained alternative and usable in the decision matrix, while retaining the conditions and normalization basis of its source. Such a descriptor supports pre-experimental prioritization but is not equivalent to an intrinsic kinetic or durability parameter.

Figure 2. Evaluation hierarchy used in the supplied decision matrix. Only the eight descriptors included in the numerical calculations are shown.

Operating temperature, fresh-catalyst cost, process cost, and reported reaction or test time were treated as cost criteria. Reported H₂-release rate, MCH conversion, BET surface area, and TOL selectivity were treated as benefit criteria. Conversion and selectivity were entered as percentages, not fractions. The H₂-release values retain the nominal unit reported in the source matrix (mmol g⁻¹ min⁻¹); a common denominator based on catalyst mass, active-metal mass, or accessible sites could not be confirmed for every publication. Similarly, the time values may represent different experimental concepts and are therefore labelled as reported reaction or test time.

Activation energy, turnover frequency, and activity normalized by metal mass or accessible active sites were not calculated. The supplied sources do not provide a common set of kinetic measurements, Pt dispersion values, or site concentrations needed for those calculations. Nominal metal loading was not used as a substitute for active site density. BET area and selectivity were retained as the available stability-related proxies, but neither is a direct measure of catalyst lifetime, regenerability, coking rate, or sintering resistance.

The original source gives qualitative by-product observations for several catalysts, including no detected liquid by-products for cases with 100% reported TOL selectivity and low by-product formation for cases near 99–99.9% selectivity. It does not provide a common quantitative by-product profile across all nine alternatives. Consequently, no additional by-product criterion was created.

2.3. Source-Based Economic Inputs

2.3.1. Fresh-Catalyst Cost

The fresh-catalyst cost values were taken directly from the source study and calculation annex. The source reports that CatCost was used to represent raw materials, material losses, preparation steps, equipment, utilities, labor, and direct and indirect cost contributions for precommercial catalyst manufacture [15]. The criterion is expressed as US$ kg⁻¹ of fresh catalyst. It does not include service life, replacement frequency, regeneration, spent-catalyst disposal, or recovery of precious metals.

Detailed case files and route-by-route inputs for the nine catalyst estimates were not included among the supplied annexes. The reported final values were therefore used without recalculation. They should be interpreted as comparative screening inputs rather than supplier quotations or lifecycle catalyst costs.

2.3.2. Process-Cost Criterion

The original study supplied three screening-level process-cost values associated with operating temperatures of 300, 350, and 400 °C under a nominal target of 18,000 t H₂ yr⁻¹. The reported accounting basis combines hydrogen used to provide additional heat with contributions assigned to a reactor, pump, two heaters, flash separator, and boiler. The source also reports a five-year evaluation period, an initial investment of US$2,693,720, a 10.5% return assumption, and a hydrogen value of US$10.70 kg⁻¹.

The final source values were US$1,014,539.29 at 300 °C, US$1,235,945.11 at 350 °C, and US$1,317,091.34 at 400 °C. Catalysts assigned the same operating temperature therefore received the same process-cost value. The supplied files do not contain sufficiently granular, fully linked records to reconstruct every total independently. For this reason, these values were used as fixed source-reported inputs and are not described as complete annual operating costs, levelized hydrogen costs, or detailed techno-economic results.

2.4. CRITIC Weighting

The supplied calculation annex was used without changing the decision matrix, the ideal and anti-ideal entries, or the recorded normalization conventions. After min–max transformation, the annex calculates the standard deviation of each criterion and the Pearson correlation coefficient between each pair of criteria. CRITIC defines the information content of criterion j as the product of its contrast intensity and its aggregate conflict with the remaining criteria [12]:

$$C_j={\sigma }_j\ast {\displaystyle \sum _{k=1}^m}\left(1-r_{\mathit{jk}}\right)$$

The objective weight is obtained by normalizing the information-content values:

$$w_j={\displaystyle \frac{C_j}{{\displaystyle {\sum }_{k=1}^m}{C}_k}}$$

The weights therefore describe the statistical structure of this nine-alternative matrix. They are not universal importance coefficients and must be recalculated if alternatives or criteria are added or removed.

2.5. TOPSIS Ranking

TOPSIS was applied to the original decision matrix using the CRITIC weights. Each column was vector-normalized and multiplied by its criterion weight:

$$r_{\mathit{ij}}={\displaystyle \frac{f_{\mathit{ij}}}{\sqrt{{\displaystyle {\sum }_{i=1}^m}{f}_{\mathit{ij}}^2}}}$$

$$v_{\mathit{ij}}=\left(w_j\right)\ast \left(r_{\mathit{ij}}\right)$$

The positive and negative ideal vectors were defined using the benefit or cost direction of each criterion. Euclidean separation distances were then calculated:

$$S_i^{+}=\sqrt{{\displaystyle \sum _{j=1}^n}{(v_{\mathit{ij}}-v_j^{+})}^2}{ S}_i^{-}=\sqrt{{\displaystyle \sum _{j=1}^n}{(v_{\mathit{ij}}-v_j^{-})}^2}$$

The relative closeness coefficient was calculated as:

$$C_i={\displaystyle \frac{S_i^{-}}{S_i^{+}+S_i^{-}}}$$

A larger CC_i indicates greater proximity to the positive ideal alternative.

2.6. Comparability and Robustness Boundary

The source experiments were conducted under different temperatures, pressures, feed compositions, catalyst loadings, space velocities, reactor configurations, pretreatments, and measurement protocols. Unit normalization supports numerical aggregation but does not remove these experimental differences. The ranking consequently compares the reported condition-specific values and should not be read as an intrinsic kinetic benchmark.

A formal sensitivity or uncertainty analysis was not performed. The supplied sources contain neither common experimental uncertainty intervals nor a documented perturbation protocol. Introducing arbitrary ± ranges, a probability distribution, or an unreported leave-one-out calculation would create results not present in the evidence base. Robustness is therefore discussed qualitatively from the supplied correlation matrix and the spacing of the TOPSIS coefficients.

3. Results and Discussion

3.1. Decision Matrix and Data Provenance

Table 1 reproduces the nine-alternative numerical matrix from the calculation annex. Technical values were reported in the cited catalyst studies, whereas the two cost columns were taken from the original study’s economic calculations. No values were interpolated or inferred for this revision.

Table 1. Source decision matrix for CRITIC-TOPSIS evaluation of MCH-dehydrogenation catalysts.

Catalyst/Primary Source	T (°C)	Reported H2 Release (mmol g−1 min−1) *	Fresh-Catalyst Cost (US$ kg−1)	Process Cost (US$)	MCH Conversion (%)	Reported Time (h)	BET Area (m2 g−1)
2 wt.% Pt-5 wt.% Sn/Mg-Al-O [16]	300	262.1	1053	1,014,539	90.5	10	157.4
0.5 wt.% Pt/10 wt.% TiO2-Al2O3 [17]	400	1711	293	1,317,091	93.2	1.5	148.2
Cu-Pt/S-1 [18]	350	288.9	318	1,235,945	59.35	10	415.1
2 wt.% Pt-5 wt.% Ir/Mg-Al-275 [19]	350	578.7	1614	1,235,945	99.9	10	173
3 wt.% Pt/Y2O3 [20]	350	958	453.85	1,235,945	95.9	1	6.37
0.4 wt.% Pt/CB [21]	300	342	201.36	1,014,539	95	5	1240
0.4 wt.% Pt/Ce-Mg-Al-O [22]	350	1358	479.35	1,235,945	98.5	10	163.1
1 wt.% Pt/AC [23]	300	428	587.64	1,014,539	88	3	1120
3.5 wt.% Pt-12.7 wt.% Mo/SiO2 [24]	400	500	800	1,317,091	90	2	1000

* The nominal unit is reproduced from the source matrix. A common denominator based on total catalyst, active-metal mass, or accessible sites could not be confirmed for all studies. Cost values are source-reported screening inputs and were not recalculated in this revision.

The matrix illustrates why direct performance claims require caution. Conversion ranges from 59.35% to 99.9%, but the values were measured under different conditions. Selectivity is narrowly distributed between 99% and 100%, while BET area ranges from 6.37 to 1240 m² g⁻¹. These spreads influence CRITIC even though BET area is only a structural proxy and selectivity is not a lifetime measurement.

3.2. CRITIC Weights and Intercriterion Dependence

Selectivity received the largest weight (18.60%), followed by BET area (15.72%) and reported time (13.95%). These weights reflect dispersion and nonredundancy within the nine-row matrix; they do not establish universal scientific importance. In particular, the high weights of BET area and selectivity should not be interpreted as proof that these variables directly measure long-term stability (Table 2).

Table 2. Criterion directions and CRITIC weights reproduced from the supplied calculation annex.

Criterion	Decision Direction	CRITIC Weight	Interpretive Role
Operating temperature	Cost	10.53%	Operating severity
Reported H2-release rate	Benefit	10.02%	Source-reported productivity
Fresh-catalyst cost	Cost	11.03%	Manufacturing-cost screen
Process cost	Cost	11.63%	Temperature-linked cost screen
MCH conversion	Benefit	8.51%	Condition-specific conversion
Reported reaction/test time	Cost	13.95%	Source-reported time basis
BET surface area	Benefit	15.72%	Structural proxy
TOL selectivity	Benefit	18.60%	Product-distribution proxy

The group totals recorded in the annex are 43.21% for techno-economic feasibility, 22.46% for catalytic performance, and 34.32% for stability-related proxies. Totals differ from 100.00% only through rounding.

The supplied correlation matrix gives r = 0.959 between operating temperature and process cost. This strong association follows from assigning one process-cost value to each temperature level. CRITIC includes correlation when calculating information content, but both variables remain separate TOPSIS criteria; their joint inclusion can therefore give operating severity a double representation in the final distance calculation. The source files do not contain a ranking with either criterion removed, so the magnitude of this influence could not be quantified without generating a new analysis.

3.3. TOPSIS Ranking

Within the supplied matrix, 0.4 wt.% Pt/CB and 1 wt.% Pt/AC provide the highest overall proximity to the positive ideal. Both combine a 300 °C operating temperature with large BET areas and relatively favorable cost inputs. Their first and second positions are outcomes of the eight-criterion aggregation, not proof of superior intrinsic kinetics or durability (Table 3).

Table 3. TOPSIS ranking reproduced from the formula output in the supplied calculation annex.

Rank	Catalyst	Closeness Coefficient (%)
1	0.4 wt.% Pt/CB	66.09
2	1 wt.% Pt/AC	58.94
3	3.5 wt.% Pt-12.7 wt.% Mo/SiO2	52.38
4	0.4 wt.% Pt/Ce-Mg-Al-O	51.76
5	Cu-Pt/S-1	51.59
6	0.5 wt.% Pt/10 wt.% TiO2-Al2O3	45.35
7	2 wt.% Pt-5 wt.% Sn/Mg-Al-O	38.64
8	2 wt.% Pt-5 wt.% Ir/Mg-Al-275	35.54
9	3 wt.% Pt/Y2O3	34.76

The ranking does not support a general monometallic-versus-bimetallic conclusion. The Pt-Mo/SiO₂ alternative ranks third, whereas Pt/Y₂O₃ ranks ninth. In addition, Pt-Mo/SiO₂, Pt/Ce-Mg-Al-O, and Cu-Pt/S-1 have closely spaced coefficients (52.38%, 51.76%, and 51.59%). Their relative order should therefore be treated as tentative until a source-grounded robustness analysis is available.

3.4. Scientific Interpretation and Limitations

The scientific value of the comparison lies in screening. It organizes heterogeneous evidence, makes trade-offs visible, and identifies candidates for testing under a common protocol. It does not convert values from the literature into intrinsic kinetic constants. Normalization changes scale, not experimental comparability, and a high TOPSIS score cannot substitute for direct measurements of turnover frequency, activation energy, accessible-site density, time-on-stream deactivation, coke deposition, or regenerability.

The principal limitations are the small complete-case dataset, heterogeneous reaction and reporting conditions, incomplete by-product information, proxy-based representation of stability, source-reported cost inputs that cannot be fully reconstructed from the supplied annexes, strong dependence between temperature and process cost, and the absence of a formal sensitivity analysis. These constraints prevent claims of universal catalyst superiority or ranking invariance.

Future work should first test the leading alternatives under the same temperature, pressure, feed composition, space velocity, catalyst loading, particle size, and analytical method. The resulting dataset should report conversion and carbon balance, identified by-products, rates normalized by catalyst and active metal, Pt dispersion, turnover frequency, activation energy, time-on-stream loss, coke content, and regeneration performance. Economic inputs should include catalyst inventory, lifetime, replacement and recovery. Sensitivity analysis should then use measured uncertainty ranges and should test criterion removal, alternative weighting, and alternative ranking methods.

4. Conclusions

The source-grounded CRITIC-TOPSIS analysis integrates eight technical and economic descriptors for nine reported MCH-dehydrogenation catalysts. In the supplied matrix, 0.4 wt.% Pt/CB and 1 wt.% Pt/AC rank first and second, with closeness coefficients of 66.09% and 58.94%. The result is useful for prioritizing standardized experiments, but it is conditional on the selected alternatives, source-specific test conditions, the recorded normalization and weighting procedure, and the source-reported cost inputs.

The review questions reveal two material limits. The process-cost criterion is strongly correlated with operating temperature and may duplicate part of the same operating-severity signal. In addition, the evidence base does not support a formal sensitivity analysis, activation-energy calculation, turnover-frequency comparison, or a quantitative by-product assessment. No unsupported values were introduced to fill these gaps. The ranking should therefore be presented as a transparent hypothesis-generation tool whose leading candidates require validation under a harmonized experimental and economic protocol.