endo-Tricyclo[5.2.1.02,6]deca-3,8-diene, or dicyclopentadiene (DCPD), hereinafter designated as (1), is a promising feedstock for petrochemical synthesis [1]. DCPD is essentially a dimer of 1,3-cyclopentadiene (CPD) and is formed by the Diels–Alder reaction even at room temperature.

As a by-product of crude oil pyrolysis, CPD is a readily available and low-cost compound [2, 3]. The CPD content in the C5 fraction of oil pyrolysis products may reach 25%.

The [2+4]-cyclodimerization of CPD is an exothermic reaction accelerated with heating. However, at 160°C DCPD thermally decomposes into the monomer form [4, 5]. The CPD dimerization produces both endo- and exo-isomers of (1) at a ratio of 99.5:0.5. This selectivity is associated with the kinetics and the steric properties of CPD [6]. This photochemical process results in the formation of an equimolar mixture of endo/exo-isomers of (1) and some amount of norbornane framework degradation products [7, 8].

DCPD and its hydrogenation products have been used as components of unsaturated polyester and hydrocarbon resins, in cyclic polyalkenamers, and as monomers to produce poly(dicyclopentadiene) and ethylene–propylene–diene elastomers [9, 10]. In agriculture they have been applied to synthesize low-molecular-weight sulfur-containing oligomers that are storage-stable and resistant to various solvents and acids [11]. The application range of derivatives of (1) includes: the pharmaceutical industry; synthesis of polymers [12, 13]; and the development of novel composite materials [14]. Furthermore, they have been used as components of high-density and high-energy rocket fuels [15] and as pesticides.

Selective hydrogenation of carbocyclic dienes along a double bond to produce cycloalkenes is of great importance for subsequent structural modification with various functional groups [16–18] and for evaluating the relative reactivity of double bonds [19, 20]. DCPD is one of the most interesting bicyclic diolefins. It contains two endocyclic double bonds significantly different in reactivity: one in the bicyclic norbornene moiety and the other in the cyclopentene ring condensed with this moiety [1, 9, 21, 22].

The wide application range of DCPD hydrogenation products is largely determined by their fomation selectivity. Therefore it is relevant, both from the scientific and practical perspectives, to develop a highly active heterogeneous catalyst and to identify process conditions optimal for selective hydrogenation of cycloalkenes and dienes with a strained carbon framework being retained.

The kinetics of liquid-phase and gas/liquid-phase hydrogenation of (1) has been examined in various studies [17, 19, 20, 22, 23]. The kinetics have been studied in the presence of Pd/C and Pd/γ-Al2O3 catalysts in n-hexane or toluene in the range of 40–160°C at both atmospheric and elevated hydrogen pressure, and under varied concentrations of the substrate and the catalyst.

In [17, 23], the kinetics of liquid-phase and gas/liquid-phase hydrogenation of (1) was identified, and an appropriate kinetic model was proposed. The process was found to consist of two steps, the first step involving hydrogenation into cycloalkene. It was shown that hydrogen pressure mostly affects the rate of this first step, but has a negligible effect on the maximum yield of cycloalkene. Although process temperature was not critical to DCPD hydrogenation, the maximum yield of intermediate cycloalkene increased with cooling. The initial rate of DCPD hydrogenation only weakly depends on its concentration, thus suggesting a zero kinetic order with respect to the substrate.

Both the shape and size of the catalyst pellets affect its activity in the hydrogenation of (1). Finely dispersed catalysts are more preferable to granular catalysts of the same type [22].

When palladium catalysts are used [23, 24], saturation of double bonds follows a consecutive mechanism. In (1), as in other norbornene derivatives, the double bond of the bicycloheptene moiety is hydrogenated first [25–30]. The ratio between the hydrogenation rates depends on the solvent type.

Despite the available information on the main steps and kinetics of DCPD hydrogenation, all attempts to model this process have proven unsuccessful. In particular, all the models for the initial and final process steps turned out to be inadequate. In all related studies, the process was kinetically controlled by the hydrogen consumption rate. This is definitely insufficient: to build an adequate mathematical model, complete chromatographic control of the reaction mixture composition must be provided during the experiments, with initial concentrations and temperatures being varied within a wide range. The modeling also requires a complete material balance to be conducted. Without these data, it is too difficult to choose appropriate catalysts, reactor type, and reaction conditions that would ensure selective hydrogenation with the norbornane framework being retained. We previously used a mesoporous core–shell catalyst named PK-25 with a low concentration of the active component (0.25% Pd/γ-Al2O3) [25–29].

The purpose of this study was to identify the kinetic patterns of liquid-phase hydrogenation of (1) in a batch reactor, to build a model that would adequately reflect the entire transformation sequence of (1) in the presence of PK-25, and to reveal the key steps of the reaction mechanism.

EXPERIMENTAL

endo-DCPD (CP grade, colorless crystals with a pungent odor, KhimMed, Russia) was used as a reagent; before testing, it was distilled at reduced pressure (65°C/50 mmHg). The solvent and internal standard were n-heptane (standard grade, KhimMed) and n-nonane (for chromatography, CP grade, ReaKhim, Russia), respectively. These compounds as well as gaseous hydrogen (A grade, 99.99% pure, GOST 3022-80) were used without additional treatment.

DCPD (1) occurs as spatial exo- and endo-isomers at a ratio that depends on external conditions. The endo-isomer was isolated from an isomer mixture by recrystallization (mp 32 and 19°C for endo- and exo-isomers, respectively [3]).

The reaction was monitored by GC on a Crystal 5000M chromatograph equipped with a 50 m×0.2 mm VS-101 column (dimethylpolysiloxane phase). The GC conditions were as follows: analysis time 27 min; detector and injector temperature 180°C; initial column temperature 70°C, followed by 10-min holding, cooling to 15°C, 12-min holding, heating to 250°C, and 3-min holding; carrier gas helium; flow rate 0.8 mL/min; and split ratio 1 : 125. The hydrogenation products of (1) were identified by GC/MS using an Agilent 5973N mass spectrometer equipped with an Agilent 6890 chromatograph (electron ionization, Agilent 122-5536 DB-5ms column).

The structures of both the reagents and hydrogenation products, including their spatial structure, were examined by 1H and 13C NMR on a Bruker Avance DPX 300 instrument, with CHCl3 (7.26 ppm) as an internal standard.

The reaction was carried out using a vibration bench in a 100 mL thermostated static batch reactor equipped with a reflux condenser and a sampler, in an n-heptane medium at 26–76°C and atmospheric pressure of hydrogen. Specialized preliminary experiments showed that the reaction mixture needed to be shaken at least at 380 rpm to ensure that the process occurred in the kinetic mode. For the hydrogenation of (1), a PK-25 commercial core–shell palladium catalyst (Russian Technical Specification TU 38.102178-96; 0.25% Pd/γ-Al2O3 ground to 0.1–0.2 mm pellets) [31] was used. Grinding to the specific pellet size was required to remove diffusion limitations during the hydrogenation process.

The specific surface area and pore sizes of the catalyst samples were evaluated by the Brunauer–Emmett–Teller (BET) method using a Beckman Coulter SA 3100 analyzer (the sample was degassed for 60 min at 120°C). The methods for catalyst activation and kinetic tests are described in [25–29].

RESULTS AND DISCUSSION

Table 1 presents the data on specific surface areas and pore size distributions for a series of catalyst samples, specifically the fresh sample, the sample reduced in a hydrogen flow, and the catalyst spent after the series of experiments. These samples had IUPAC type IV adsorption isotherms with a hysteresis loop in accordance with the classification first proposed by Brunauer et al. This isotherm type is typical of mesoporous catalysts. In all samples, the vast majority (about 80%) of pores were 10 to 80 nm in diameter.

Table 1. Pore size distribution and specific surface area for a series of PK-25 samples
Full size table

The specific surface area of all catalyst samples ranged between 155 and 180 m2/g. The fresh catalyst after hydrogen activation had the highest area, and the spent sample after hydrogenation had the lowest. The surface area drop in the spent catalyst can be explained by the presence of products in the catalyst pores. Indeed, subsequent thermal treatment restored the specific surface area back to its initial value.

To investigate the kinetics of liquid-phase hydrogenation of (1), nine series of experiments were carried out in the presence of a finely ground (0.1–0.2 mm) PK-25 palladium catalyst. In these experiments, the initial concentration of (1) and temperature in n-heptane were varied within the ranges of 0.2–0.6 mol/L and 26–76°C, respectively. Each experiment was duplicated to evaluate the reproducibility of kinetic data and statistical parameters (description of the mathematical model). Furthermore, the material balance was calculated for all system components (with the margin of error not exceeding 5%).

The process kinetics was monitored both volumetrically (by measuring the hydrogen uptake) and chromatographically (by analyzing the samples being taken over the course of the experiment). The GC and GC/MS analysis of the reaction mixtures indicated that the hydrogenation occurred by a consecutive route that consisted of the formation of an intermediate (2) and a final hydrogenation product (3), with a complete lack of any products of isomerization or degradation of the norbornane framework:

structure 1

The kinetic data showed that, over the wide range of the initial concentration of (1), this parameter had almost no effect on the consumption rate (Fig. 1). A similar result is known to have also been observed for other norbornene derivatives such as 5-vinylnorbornene-2, norbornadiene-2,5, and norbornene-2 [25–29].

Fig. 1.
figure 1

Kinetic curves of DCPD consumption at different initial concentrations.

Full size image

The linear concentration decrease corresponds to a zero kinetic order. Figure 2 shows the typical consumption kinetic curves of (1) and its hydrogenation products—(2) and (3).

Fig. 2.
figure 2

Kinetic curves of hydrogenation of DCPD (1) and reaction products at an initial substrate concentration of 0.6 mol/L (76°C, n-heptane). The vertical line divides the graph into two regions: (I) conversion of DCPD into cycloalkene; and (II) hydrogenation of cycloalkene into cycloalkane.

Full size image

The liquid-phase hydrogenation of (1) is a distinct consecutive process. Its kinetics is well illustrated by the hydrogen uptake curve, which appears as a broken line composed of two different sections. The first section reflects the hydrogenation of the norbornene double bond of (1). Like for other norbornene derivatives, this reaction occurred selectively, obviously due to a stronger adsorption of the norbornene bond on the active sites (AS’s) of the catalyst. This resulted in nearly quantitative formation of (2). At this step, the hydrogen uptake occurred at a maximum rate consistent with the rates of substrate consumption and intermediate formation. As long as (1) was present in the system, (2) was not hydrogenated. The hydrogenation rate of (1) was almost equal to the hydrogenation rates for other norbornene derivatives under identical conditions. As previously noted, the quantitative hydrogenation of the norbornene bond, which did not involve any other double bonds (e.g., cyclopentene, vinyl, or ethylidene bonds), serves as clear evidence of a higher reactivity of the norbornene bond).

In the second section of the hydrogen uptake curve, the liquid-phase hydrogenation of (1) was noticeably inhibited: cycloalkene (2) slowly transformed into cycloalkane (3) (see Fig. 2). In the absence of (1), this step had a near-first kinetic order with respect to (2). This essentially differentiates the hydrogenation of (1) from the hydrogenation of norbornadiene-2,5, for which zero kinetic order was observed in both process steps.

An investigation of the rate of liquid-phase hydrogenation of (1) as a function of temperature over the range of 26–76°C revealed no clear patterns (Fig. 3). Temperature was found to have a very minor effect on the process rate. The apparent activation energy of the first hydrogenation step was nearly zero. Based on the assumption that this energy is the total of the true activation energy (Ea) of DCPD and the heat of its adsorption on the catalyst surface (ΔHads), it is very likely that these two parameters were similar in absolute values and opposite in signs.

Fig. 3.
figure 3

lnk as a function of inverse temperature for step 1: liquid-phase hydrogenation of DCPD.

Full size image

In contrast, the liquid-phase hydrogenation rate of (2) heavily depended on the temperature: it dropped by a factor of ten as the temperature was lowered from 76 to 26°C (Fig. 4). In the second step, the apparent activation energy was 29 kJ/mol. The resultant data confirmed a markedly higher reactivity of the norbornene bond than that of the cyclopentene bond.

Fig. 4.
figure 4

lnk as a function of inverse temperature for step 2: liquid-phase hydrogenation of cycloalkene.

Full size image

Kinetic Modeling

To build a structural kinetic model, we followed the stepwise approach described in [29, 31]. The first stage was setting the compositions of all reaction components, including a variety of AS complexes with reagents, products, and intermediates. Then, to reasonably limit the number of AS’s, we applied certain rules (these being essentially restrictions or even prohibitions) based on relevant laws of physics, chemistry, thermodynamics, etc. This second stage was of a deductive nature, since its purpose was to eliminate “odd” AS’s.

For the process under study, this elimination was achieved by limiting the number of molecules adsorbed on one AS to 1, 2, and 3. Assuming that the generation of AS’s predominantly involves adsorption of mono- and diolefins, these numbers correspond to 2, 5, and 9 AS’s with different compositions. In this stage, the possible differences in the structure and activity of AS’s having identical compositions were not taken into account because this differentiation level was not required for an adequate description.

Next, the steps of the process mechanism were configured, the approach still combining techniques of both induction and deduction as described above. For a coordination number (CN) of three, the number of these steps is 15. Without going into generally known details, it is worth noting that solutions of the inverse kinetic problem are widely used to eliminate/deduce the steps. If this solution irrefutably indicates that some step(s) of the kinetic model make(s) a negligible contribution to the process, this/these step(s) should be eliminated. At the same time, it may be, in fact, worth discussing the causes of the negligible contribution of the eliminated step(s), as these causes may involve some issues of theoretical importance bearing on questions of energy, quantum-chemical, geometrical, or other aspects of the process. If ambiguity cannot be eliminated, it is reasonable to use scientific information sources for the deduction of process steps.

All hypotheses that fail to provide an adequate description of the experimental data according to Fisher’s criterion must be discarded. Importantly, we did not apply non-standard confidence probability levels, such as 0.95. These levels dramatically reduce the rigidity of Fisher’s criterion, thus permitting as acceptable (for the next round of consideration) certain hypotheses that even experimenters—by sight—suggest are unable to adequately describe the experimental data. The margin of experimental error was estimated by duplicating the experiments, in line with common, standard practice. The expert estimation method, which involves a calculation of expected error, is also, in every cases and purposes, acceptable.

Table 2 summarizes general data on the kinetic models that were considered. The limitation CN = 1 was discarded because it resulted in inadequate models. When the number of unsaturated molecules was limited to CN = 2 or CN = 3, the models were adequate; of these two cases, a significantly better description was obtained with CN = 3. The root-mean-square deviation (RMSD) was 3.8 and 3.5%, respectively. Fisher’s criterion was below one. A numerical assessment showed that the number of steps can be reduced to five (5) for CN = 2 and to three (3) for CN = 3 without increasing Fisher’s criterion. This makes further application of Occam’s razor ambiguous. Namely, following the principle of Occam’s razor, a model with either a lower limitation of CN or fewer steps can be assumed to be simpler. We chose the Fisher-optimal model description with fewer steps and CN = 3. It should be noted that there may exist more than one model satisfying this choice. These models will have the same structure but differ in the set of parameters with equal responses, i.e. will be characterized by output equivalence. Accordingly, a transition from one set of parameters to another with preservation of the model structure and output equivalence can be called an equivalent transformation. This approach can also be referred to as numerical reparametrization. It consists of a numerical assessment of the relationships between parameter estimates [29, 31] and, based on its results, elimination of insignificant steps in terms of process rate: for example, elimination of overly slow parallel steps or overly rapid consecutive steps. At the end of this procedure, a specific software algorithm identifies functionally related estimates of model parameters and represents them as linear or logarithmic (exponential) combinations. These two forms of combinations or nonlinear parametric functions (NPFs) are currently the most common in chemical kinetic modeling. The procedure concludes by inverting the full rank Fisher matrix to evaluate the standard errors of the numerical parameter estimates and the above-mentioned combinations. Below is a detailed representation of the chosen model, with only significant components being shown.

Table 2. Compositions and stability constants of adsorption complexes of unsaturated reagents with catalyst ASa
Full size table

Cat + P ↔ Cat·P (reversible binding of the catalyst grain surface with the total number of polycyclic reactants); K6 = 2.3.

Figures 5a–5c illustrate the modeling results for the liquid-phase hydrogenation of (1) at different initial concentrations.

Fig. 5.
figure 5

Modeled kinetic curves for hydrogenation of DCPD (1) and reaction products at different initial substrate concentrations (mol/L): (a) 0.15; (b) 0.33; and (c) 0.57 mol/L (76°C, n-heptane).

Full size image

Figure 1 clearly shows that the hydrogenation slightly slowed down as the total concentration of (1), (2), and (3) increased. It is the total concentration, rather than any other possible combination of concentrations, that—using Fisher’s criterion—was finally chosen after the elimination of other hypotheses. In this case, a minor portion of the total surface of the catalyst grain was blocked due to reversible nonselective adsorption of unsaturated polycyclic compounds:

(1) + H2 → (2).

The first hydrogenation step occurred by a single route:

Z·(1)·(1)·(2) + H2 → Z·(1)·(2)·(2); k1 = 1.4×105 min–1,

(2) + H2 → (3).

The second hydrogenation step followed two different routes:

Z·(2)·(2)·+ H2 → Z·(2) +·(3); k2 = 2.4×104 mon–1,

Z·(1)·(2)·(2) + H2 → Z·(1)·(2) +·(3); k1 = 5.9×109 min–1,

where Z is the active site.

A quantification of the ambiguity of parameter estimates demonstrated that K6 was the only equilibrium constant that was estimated unambiguously. The margin of error was ±25%. The other equilibrium constants were estimated with ratio-limited precision because AS’s were almost completely bound with adsorbates:

$${{{K_5}} \over {{K_8}}} \pm 30\% ;\;{{{K_4}} \over {{K_2}}} \pm 25\% .$$

The situation with the rate constants was as follows. Two rate constants were estimated unambiguously: k1±25%; k2±30%.

One rate constant was determined as part of an NPF. This refers to active sites low-filled with adsorbed molecules, a case typical for heterogeneous catalysis:

$${{{k_3}{K_9}} \over {{K_8}}} \pm 35\% .$$

The other equilibrium constants were estimated with poor accuracy, either due to high values (almost complete shift of the equilibrium to the right), or due to very low values (negligible binding of olefins with an AS). It should also be noted that the insignificance of any particular parameter may be caused by the trivial inability of an adsorption complex to participate in the related chemical transformation. For example, adsorption complexes based on (1) might only exist in the first reaction step, until (1) was depleted. Therefore, these complexes could not take part in the second reaction step. An adsorption constant of this kind could not be estimated unambiguously unless the corresponding complex was significantly involved in the balance of the total number of AS’s. It is worth noting that, to provide concrete and definite measurements, we assumed the total AS concentration in the system as 10–6 mol/L.

CONCLUSIONS

Based on a series of kinetic studies, a consecutive mechanism was proposed for hydrogenation of DCPD (1). The effect of prevalent adsorption of a strained norbornene double bond due to its increased reactivity was confirmed. An adequate kinetic model was developed based on the Langmuir–Hinshelwood approach and the concept of multiple adsorption of substrates on a single AS of the catalyst. Three process steps that occurred by two routes significantly contributed to the reaction rate. The rate constants of these reaction steps and the adsorption constants of AS complexes with unsaturated compounds were estimated within the framework of the kinetic model. The hydrogenation rates obtained in the first process step were found to be similar to those for a number of other norbornene derivatives.