Electronic Supplementary Information Selectivity of the Lindlar catalyst in alkyne semi-hydrogenation: a direct liquid-phase adsorption study

Pd catalysts contain active sites that strongly adsorb alkyne and alkene molecules. The presence of the latter, alkene sites, defines the low semi-hydrogenation selectivity.


SI1.1. Details on the adsorption studies
Because of high adsorption constants observed (~10 4 L mol -1 ) and low adsorption capacity (~ 1 nmol per 50 mg sample), we had to work with low concentrations close to the analytical detection limits.
The system used for adsorption experiments is presented in Figure S1 contains a fraction collector and a syringe pump. The fraction collector had a built-in 3-way valve with a needle that allowed injecting a given solution into any vial. The syringe pump equipped with an SGE 1mL precision syringe acted as a liquid handling unit injecting pre-defined volumes of the solution into vials. Figure S1. Scheme of the automated system used for the adsorption experiments.
The solution was changed following the procedure of blowing air into all connecting tubes and the syringe, washing with the solvent, and blowing air. Afterwards, 5 full injections into an empty vial were performed to wash the needle. Solutions were changed in increasing concentration order starting from solvent (hexane, 99%, Fischer Scientific).

Electronic Supplementary Material (ESI) for Catalysis Science & Technology. This journal is © The Royal Society of Chemistry 2021
The vials for the adsorption experiment were divided into 4 parts: (i) control, (ii) reference, (iii) catalyst support, and (iv) catalyst. The control vials were injections of the solvent at the beginning and end of every run. This way, it was possible to control that the solution concentration was consistent during the injection and observe any problems of dead volume in the fluidics connections which would cause a lower concentration at the beginning of the experiment. The remaining three groups of vials contained the same volumes of the solutions. The reference vials were empty to study linearity of the response with the concentration. The vials with 50 mg catalyst support were used to study the adsorption of the compounds over the catalyst support considering that its surface area was significantly higher than that observed for the Pd nanoparticles. Lastly, the vials with 50 mg catalyst were used to study adsorption over both the Pd catalyst and the support.
The vials with the catalyst support and the catalysts were reduced in a flow of 20 mL min⁻¹ 5 vol% H₂ in N₂ at 150 o C for 1 h followed by flushing with N₂ and passivation with 1 vol% O₂ in N₂ before exposure to air. The adsorption experiments were performed within 2-4 h after exposure to air. A significant decrease in adsorption capacity was observed only after about 1 week of exposure to air.

Analysis of the samples obtained
The samples obtained were analysed directly by gas chromatography with a Shimadzu GC 2010 equipped with an FID detector. High adsorption observed required using trace substance concentrations (~ 0.1 ppmw) to determine the adsorption parameters. The GC analysis was optimised to ensure high reproducibility and low detection limits. The GC was equipped with a Stabilwax 60 m x 0.32 mm x 1 μm column, 8 μL of the sample were injected in a splitless mode at the column temperature of 40 o C and the inlet pressure of 100 kPa. After 4 minutes at 40 o C, the column temperature was ramped at 15 o C min -1 to 175 o C followed by holding this temperature for 3 min. The injector pressure was also ramped after 8.5 min from injection -30 kPa min -1 till 220 kPa followed by holding the pressure for 4.5 min.
The analysis method was extended when quinoline was injected. After reaching the column temperature of 175 o C, the second ramping state was started at 25 o C min -1 till 255 o C. Meanwhile, the injection pressure was also ramped after 8.5 min at 100 kPa -the ramping at 30 kPa min-1 till 380 kPa was performed to desorb quinoline faster. The method provided a detection limit of 0.2 μM with complete separation of all the studied compounds. The analysis was performed based on the absolute area of the peak corresponding to the repeatability of ± 2%. No internal standard was used to avoid interference with the adsorption such as displacement of the analysed species or adsorption of the internal standard itself.

Analysis of the data obtained
The concentrations in the reference solutions were processed using a weighted least-squares method. The weights were taking into account the experimental error of ±2% or a fixed error (whichever is larger). The fixed error was found to depend on the concentration of the species analysed and varies between 3 μM (when 500-800 μM MBY solution was used), 1 μM when 200 μM MBY or MBE solutions were used, or 0.3 μM when 40 μM MBE solution was taken.
Comparison of the amount of the substrate injected into the vials and determined by the analysis provided the amounts of adsorbed and equilibrium concentrations. These raw data were analysed using the Monte-Carlo regression method [1]. Sets of simulated experimental data (1000 for each case) were created (with errors introduced with a normal distribution and standard deviations as above). These data were fit with the Langmuir adsorption model analysing the parameters obtained statistically considering covariations of the parameters and reporting 90% confidence intervals.

SI1.2. Details on the Computational Modelling
We used the Vienna Ab Initio Simulation Package (VASP) to perform spin-polarised periodic density functional theory-based calculations using the projector augmented wave method [2][3][4][5]. The cut-off energy for the expansion of the plane-wave basis sets was set to 550 eV, which gave bulk energies converged to within 10 -5 eV. For the structural optimization, the convergence criterion was set to 0.01 eV Å -1 . The ideal Pd(111), Pd(110), and Pd(210) surfaces were modelled by 4x4 cell with 5 atomic layers and of the five atomic layers bottom three layers were fixed to mimic the bulk of the material.
The calculated Pd lattice constant of 3.904 Å agreed with the experimental value of 3.891 Å [6].
The adsorption of MBY and MBE molecules on Pd surfaces was allowed only on one of the exposed surfaces and the spurious dipole moment, due to the adsorbed molecule on one of the two surfaces was taken into account using methods implemented in VASP according to the procedures of Makov et al. and Neugebauer et al. [7,8]. For the interaction of MBY and MBE with Pd surfaces, the dispersive effects may be significant; therefore, in all the calculations, we included Grimme's dispersion correction (DFT+3) [9]. To choose an appropriate k-point grid for these calculations we performed benchmark calculations on the adsorption property of MBY on Pd(111) surface with 3x3x1 and 5x5x1 K-point grid, which yielded a difference in adsorption energy of 7.690x10 -4 eV, which is within the DFT errors. Additionally, our previous studies on the interaction of organic molecules such as furfural on Pd surfaces have shown that the use of a K-point grid of 3x3x1 was sufficient, therefore, in this study, a K-point grid of 3x3x1 was used [10]. The adsorption energy was calculated as a difference in the total energy of the molecule on the surface and the energies of the isolated molecule and the pristine surface.

Catalyst preparation
The catalyst, 2 wt% Pd/CaCO3 catalyst was prepared by wet impregnation -a solution of palladium (II) acetate (98% Fisher Scientific) was dissolved in acetone and impregnated into the CaCO3 support (99%, Sigma-Aldrich) to provide the Pd loading of 2 wt%. Acetone was evaporated in a rotavapour and the solid material was calcined at 400 o C in the flow of air for 1 hour, 100 mL min -1 . (All gas flow rates in the work are referred to normal conditions of temperature and pressure). Afterwards, the catalyst was reduced in a flow of 20 mL min -1 5 vol% H₂ in N₂ at 150 o C for 1 hour, gas was replaced with N₂ and a small amount of air (1 vol% O₂) was introduced to passivate the catalyst surface.
The Pd catalysts obtained was separated into two parts. One part was poisoned with Pb by mixing

Catalyst characterisation
Elemental analysis performed on a Rigaku Primus IV WD-XRF instrument using a fundamental parameters model. For the powder experiments, the samples were mounted on a filter paper holder and the weight/diameter used to correct the model for thickness effects. The original Pd catalyst was found to contain 2.1 wt% Pd balance CaCO3 and below 0.05 wt% other elements; the Pd-Pd catalyst contained 1.8 wt% Pd and 3.7 wt% Pb.
CO chemisorption studies were performed in a modified flow adsorption system described in reference [11]. The catalysts (500 mg) were placed into a glass tube between quartz wool plugs.
After the reduction at 350 o C for 2 h in a flow of 1 vol% H2 in He 5 mL min -1 , purging with He for 2 h and cooling to 35 o C, a flow of 4 mL min -1 1.2 vol% CO/ 1.2 vol% Ar in He was admitted into the catalyst with the outlet concentration monitored with a quadrupole mass-spectrometer (m/z=12 to avoid interference with traces of N2). The CO chemisorption capacity was measured related to the signal of Ar and checked against a reference 0.5 wt% Pt/Al2O3 catalyst (provided by Micromeritics).

Liquid-phase hydrogenation
Hydrogenation experiments were carried out in a Parr 160 mL autoclave. In a typical experiment, the catalyst (50 mg) was placed into the reactor and 90 mL hexane (95%, Sigma-Aldrich) solvent was added. The reaction mixture was heated to the desired temperature, purged 5 times with N₂, then 5 times with H₂. The substrate, 2-methyl-3-butyn-2-ol (98%, Fisher Scientific) diluted to 10 mL was added into a separate vessel, degassed purging 5 times with H₂ and injected into the reaction vessel.
Reaction time started and the liquid samples (0.7 mL) were collected periodically during the reaction.  In this model, we neglect dimer formation which was below 0.2% compared to either specie -well in agreement with literature data [12]. The elementary steps of the reactions considering adsorption sites for organic molecules (σ) and hydrogen (ϵ) are shown in Table S1. Considering the material balance of the σ and ϵ adsorption sites, free coverage of these sites can be calculated according to equations S1, S2: The resulting rate equations are shown in equations S4-S6: The experimental concentration profiles were integrated with an in-house program written in Matlab using consumptions of the reaction species shown in equations S10-S12: where VL is the total volume of liquid in the reactor, mcat is mass of the catalyst added, is mass loading of Pd in the catalyst, and MPd is the molar mass of Pd metal. The system of these 3 equations was integrated numerically, not considering the material balance of organic species to identify lowaccuracy integrations by deviations from the material balance. Figure S3 shows that the values differing by more than a factor of 3 can be obtained fitting the same experimental data with high accuracy -the sum of differences between the models based on the k1 values of 0.3·10 8 L molPd -1 s -1 and 1.1·10 8 L molPd -1 s -1 is vanishingly low! Figure S3.

Why the results are so ambiguous?
The problem with the results comes from the form of equations S7-S9 and the fact that adsorption is strong. Strong adsorption in mathematical terms means that ≫1. Taking equation S7 as an example, we could extract from the denominator to obtain equation S13: In this equation, it is clear that 1/ is a vanishingly small value that could be neglected resulting in equation S14: It could be simplified eliminating to obtain equation S15: As a result, there are only 2 independent parameters ( 1 ′/ and / ) in the equation. For example, the triplets of ( 1 ′, , ) of (1,2,1) and (100,200,100) will generate exactly the same 1 ′/ and / ratios and the same reaction rate in equation (S15) -hence equations S7-S9. Therefore, it is the notation of equations S7-S9 creates a problem and leads to poorly defined results.

SI2.2. Model2. Simplification of model 1 considering relative adsorption constants.
Model 2 considers difficulties in determining the absolute values of the adsorption constants. Relative constants were uses relative adsorption constants for a specie X ( QX=KX/KMBY ) as in equations S16-S18: where ki'' are the corresponding constants ki' divided by KMBY excluding 2 ′′ = 2 2 , KMBE was taken out to form QMBE into the dividend. If the adsorption constant KMBY is high, the value of 1/KMBY can be neglected providing equations S19-S20 used in model 2. (S19)

SI3. Experiments on alkyne displacement with alkene
In a set of experiments, we aimed to verify another basis of the typical Langmuir-Hinshelwood hydrogenation model based on thermodynamic explanation of selectivity -displacement of alkene species with alkyne. An excess of MBE was added to provide full MBE adsorption onto the catalysts before the addition of small amounts of MBY. Figure S4A

Figure S4. MBY adsorption over the (A) 2 wt% Pd/CaCO3 and (B) Pb-poisoned 2 wt% Pd/CaCO3 catalysts with pre-adsorbed MBE at the initial concentration of 34 μmol L -1 . Solid lines represent adsorption isotherms observed without MBE added. Change in equilibrium MBE concentration with MBY addition onto the (C) 2 wt% Pd/CaCO3 and (D) Pb-poisoned 2 wt% Pd/CaCO3 catalysts compared to initial MBE concentration (dashed line).
Therefore, we can conclude that there are two sites over the Pd catalysts: (i) alkyne sites that strongly adsorb alkyne molecules and weakly alkene molecules, (ii) alkene sites that strongly adsorb only alkenes. The alkene sites do not seem to displace adsorbed alkene species by the excess of alkyne creating a pathway for non-selective reaction. Poisoning the surface of Pd catalyst with Pb significantly decreases the number of alkene adsorption sites -the relative decrease agrees with the corresponding alkyne semi-hydrogenation selectivity. that good fit of the experimental data does not confirm that all model assumptions are valid [13][14][15].

SI4. Adsorption isotherm of quinoline over the Pd catalysts
In particular, we cannot estimate the adsorption energy which may depend on the surface coverage as known in the gas-phase adsorption studies of ethylene and acetylene over the Pd catalysts [16].
Temkin isotherm may be a better fit of the alkyne adsorption over the Pd catalysts and it is possible to speculate that Pb poisons high-energy sites of Pd catalyst. Such speculations, however, are difficult to support further; therefore, Langmuir model was used throughout.    Equations S22-S24 are obtained based on these assumptions and include nalkyne_sites and nalkene_sites are the numbers of alkyne and alkene adsorption sites determined by the adsorption measurements, KMBY and KMBE are the adsorption constants of MBY and MBE over the alkyne sites. The model assumes adsorption of the organic molecules over "ensembles" of several neighbouring Pd surface sites for simplicity. KMBY was determined experimentally, while KMBE was assumed to be significantly lower with the value of 10 L mol -1 (a factor of 5,000 lower than KMBY). Hydrogenation of MBE over the alkene sites was considered as a zero-order reaction because no competition with MBY adsorption and high adsorption constant resulted in complete surface coverage.
A sensitivity analysis of the model was performed -a pair of parameters was perturbed from its optimal value (the most accurate description of the experiment) and the effect on the weighed residual was analysed. It is worth pointing out that using statistical weighing (inversely proportional to experimental uncertainties) provided dimensionless residual values -the values independent of the concentration units.