Automated Synthesis of a 96 Product-Sized Library of Triazole Derivatives Using a Solid Phase Supported Copper Catalyst

This article deal with the parallel synthesis of a 96 product-sized library using a polymer-based copper catalyst that we developed which can be easily separated from the products by simple filtration. This gave us the opportunity to use this catalyst in an automated chemical synthesis station (Chemspeed ASW-2000). Studies and results about the preparation of the catalyst, its use in different solvent systems, its recycling capabilities and its scope and limitations in the synthesis of this library will be addressed. The synthesis of the triazole library and the very good results obtained will finally be discussed.


Introduction
With creativity, the organic chemist is seeking new structures and methods to access molecular diversity. Over the years of organic synthesis adventure, several reagents and routes have been discovered. They were applied in total synthesis of natural compounds, or analogs, and to elaborate entirely new structures. Many efficient tools are now available to the chemist who wants to play with chemical architecture. Nevertheless the effectiveness of all the available approaches, there is still a need for new methods and approaches that can improve both efficiency and speed of preparation of organic derivatives.
We published our initial findings on a new and original catalytic system based on copper (I) iodide chelated on Amberlyst A-21 polymer for a use in automated solution synthesis of 1,2,3-triazoles from organic azides and terminal alkynes [43]. The advantages of this catalyst are the ease of preparation, a good catalytic activity and the simple separation from the reaction product by filtration. The first examples seemed to demonstrate that this system had some advantages, making it a good candidate for a use in automated synthesis. We wish to report here our findings and complete studies on this catalyst, as well as its use in the automated parallel synthesis of a 96 compound-sized library in solution.

Preparation of the catalyst
Most of the published methods involving the use of copper salts to catalyze Huisgen's reaction are conducted in the presence of an added base, which can be of organic or inorganic nature. The presence of a base was suggested to be needed in order to help the formation of an intermediate copper (I) acetylide, which may be formed and accounted for the observed regioselectivity [44,45]. In our initial communication, in order to choose a heterogeneous catalytic system, we decided to look for a polymer which can both chelate copper salts and act as a base. For this purpose we selected Amberlyst A-21, a dimethylaminomethyl-grafted polystyrene, bearing an amine group, which is meeting both criteria [43]. After a first evaluation of the solubility of simple copper (I) halides in various organic solvents, we selected copper (I) iodide in acetonitrile to prepare the catalyst.
From the experiments, the better set for the fixation protocol was the use of the ratio 2.4 mmol amine / 1.0 mmol CuI. This gave a light green colored polymer with a final composition of 1.23 mmol CuI/g of resin, indicative of a 2.98 mmol amines for 1 mmol CuI. Elemental analysis of this polymer finally gave a copper content of 8.64%, which gave a loading of 1.36 mmol CuI g -1 , close to the value measured by the weight increase (Scheme 2) [43].

Influence of the solvent
In our preliminary study, we found out that the catalyst worked better in methylene chloride than acetonitrile, not always in terms of yields due to triazoles' solubility but, most of all, this solvent prevented copper from leaching from the catalyst [43]. In order to determine if our catalytic system can be used under other conditions, a study of both conversion and yield, as a function of the solvent used, was done at 10 mol % in catalyst ( Figure 1). A panel of solvents as a function of their polarity and nature were selected and the results studied by NMR for conversion (C) and isolated yields (Y) for an over-night on the model reaction (16 h) for the formation of 3. In case of dichloromethane (C: 100%, Y: 100%), the product was totally formed and isolated pure after a simple filtration. For acetonitrile (C: 100%, Y: 94%), the product needed a purification step in order to remove some copper salts that leached. The reaction seemed to work well in toluene (C: 82%, Y: 63%) but the yield was lower after removal of the excess azide 2. In heptane (C: 20%, Y: 16%) and ethyl acetate (C: 27%, Y: 22%) both yields and conversions were low. This can be due to poor solubility of the reagents in these solvents.
Interestingly, the reaction proceeds with a good conversion but a modest yield in ethanol (C: 91%, Y: 63%). This can be due to a partial solubility of the reagents in ethanol. In aqueous systems such as water (C: 21%, Y: 13%) and t-BuOH/water (1/1) (C: 33%, Y: 14%) the conversions were low as well as yields after purification. For the polar and high boiling DMSO (Y: 31%) and DMF (Y: 25%) the yields were only measured after cumbersome extraction and residual solvent evaporation followed by purifications. From this study, it is obvious that all reagents have to be at least a little soluble in the reaction media in order to react at polymer's surface, where the catalyst is fixed. In very polar and protic solvents, not very suitable for use with poly(styrene) derivatives, the conversion and yields ranged from 10-30%. Furthermore, more difficult isolation and purification procedures were needed and some copper leaching was always observed when the crude samples were analyzed.

Influence of the catalyst amount
In our first publication and studies presented here, we selected arbitrarily an amount of 10 mol % to conduct reactions [43]. We then found out that the catalyst was better working in methylene chloride, rather than other solvents, as discussed in the previous section. Thus, a study of conversion as a function of the catalyst ratio was conducted in deuteriochloroform and the results are presented in Figure 2. When the catalyst amount was varied from 1 to 8 mol %, the conversion gradually increased  Even if 6 mol % was enough to obtain a quantitative yield of 3, the value of 8 mol % was selected to insure a complete transformation in most of the reactions using this procedure and reaction time. Of course, an amount of 10 mol %, as we previously used, can guarantee as well a complete transformation of most of the reaction partners.

Reaction kinetics: homogeneous vs. heterogeneous conditions
In order to compare the efficiency of our catalytic system, kinetics studies were done on the model reaction with Amberlyst A-21•CuI and the Et 3 N/CuI couple at the same catalytic level, i.e. at 10 mol % (Scheme 3). The reactions were followed-up by NMR analysis of the reaction mixtures using deuteriochloroform as the solvent.

Et 3 N / CuI
The NMR spectra for both experiments are presented in Figures 3 and 4. The reaction is easily followed by the disappearance of the alkyne 1 proton (triplet at 2.27 ppm) that is replaced by the H-5 proton of the formed triazole 3 (singlet, 7.51 ppm). The methylene protons of the starting azide 2 (singlet, 4.37 ppm) and alkyne 1 (doublet, 4.50 ppm) are shifted downfield and appear as singlets at 5.00 and 5.51 ppm respectively.   For the homogeneous conditions (Et 3 N/CuI) the reaction started as soon as the reagents were mixed (Figures 3, 5). After only five minutes, there was already around 5% conversion. The reaction is half completed after one hour and total conversion was reached in a little more than three hours (quantitative for the 3.5 h sample). In the case of the Amberlyst A-21•CuI catalyst, in heterogeneous conditions (Figures 4, 5), there was an induction time of around an hour (0.5% conversion after fortyfive minutes). The reaction then proceeded smoothly, reaching half completion after three hours, and was over after six hours and a half. By comparison, using the model reaction leading to the formation of 3, the homogeneous conditions were twice as fast as the heterogeneous ones. However, the isolation procedures were different and in favor of the insoluble catalyst. The use of A-21•CuI only necessitated a filtration and evaporation of the solvent, the product 3 being isolated in quantitative yield. In the case of the reaction with Et 3 N/CuI, extractions and purification were needed but the product 3 was isolated in a good yield of 75%.

Recycling and stability of the catalyst
Since the catalyst is recovered at the end of reaction, its reuse (recycling) and stability were finally evaluated (Table 1). Another model reaction was used for these tests, namely the formation of 1carboethoxymethyl-4-carbomethoxy-1,2,3-triazole (6) from methyl propiolate (4) and ethyl azidoacetate (5).
Recycling of the catalyst was done in two ways, but for both of them the catalyst was recovered by filtration, washed with the reaction solvent (CH 2 Cl 2 ) and dried under vacuum before being reused immediately for another cycle or kept in a vial before to reuse it.
In the first series of experiments (Entries 1 to 6), the catalyst was immediately reused after a first reaction. The catalyst was thus reused over five days (one reaction a day) without losing its efficiency or activity leading to 6 in quantitative yields.
For the second series (Entries 7 to 12), the catalyst was kept on the shelve between the reactions. The reactions were then done each week over a five-week period. Once again, the catalyst leads to the formation of 6 in a quantitative manner with no differences between the cycles. Interestingly, the stability of this catalyst seem very good, since it can even be used after one year without lost in the activity and selectivity. A possible explanation is that chelation of the copper by the polymer amine "ligands" (-CH 2 NMe 2 ) could be accounted for this stability, maybe by preventing oxidation of the copper (I) state [12].

Application: Automated synthesis of a library of 96 triazoles
In order to evaluate scope and limitations of this catalytic system, the automated synthesis of a library of 96 triazoles was undertaken (Scheme 4). The synthetic operations were conducted using a ChemSpeed ASW-2000 machine by loading the catalyst (8 mol%) in the reactor and injecting dichloromethane solutions of twelve alkynes a n and eight azides b n . After 18 h of orbital agitation at room temperature, the reaction mixtures were filtered and the corresponding ninety-six triazoles a n b n recuperated in solution. The results are presented in Table 2.

Scheme 4.
General scheme for automated triazoles a n b n synthesis from alkynes a n and azides b n .   When looking at the alkynes a n used, propargyl alcohol (a1) and N-acetylpropargylamine (a6) gave equal average yields of 74%; methyl propiolate (a4), phenylacetylene (a10), 1-decyne (a11) and trimethylsilylacetylene (a12) average yields between 80-85%; and finally phenyl propargyl ether (a2), propiolaldehyde diethyl acetal (a3), N-propargylphthalimide (a5), N-trifluoroacetylpropargylamine (a7) and N-(t-butoxycarbonyl)propargylamine (a8) average yields within 90-95% range. Tripropargylamine (a9) was the only alkyne that did not give good results. The problem is arising from the fact that the three alkyne functions have to react to form the corresponding tris(triazolylmethyl) amines. These conditions did not permit the complete formation of the tris compounds giving various ratios of tris (T), di (D) and mono (M) triazole derivatives. Furthermore, the presence of the amine center in the molecules was found to cause copper leaching at some extent. This outcome is quite different from the results obtained without solvent with this catalyst, on in solution with a clay supported one, where the tris(triazolylmethyl) amines were exclusively obtained [23,46].
Most of the reactions for this library generation gave good yield of triazoles, with an average of 85%. For the 96 products isolated, excluding tripropargylamine (a9), 74% of the library gave yields between 80-100%, 18% of it yields between 60-80%, and only 4% and 3% of the yields were between 50-60% and below 50% respectively. The lowest yields observed here are usually due to a poor solubility of the triazole in dichloromethane, especially for oily products that tend to coat the polymer beads. We indicated in Table 2 yields obtained after filtration and washings of the polymer by dichloromethane. The yields can be increased in the case of lower solubility by doing acetonitrile washes, but with the cost of variable levels of pollution of the triazole by copper leaching from the catalyst.
In all reactions, only the 1,4-isomer was observed and all products save some exceptions were pure. As previously observed with this system, no azide excess was found into the reaction products obtained in quantitative yields, suggesting a possible sequestration by the polymer. No copper leaching was observed in most cases. The products were therefore isolated as pure after a simple filtration/ evaporation procedure. For reactions were the excess azide was still present or some copper leaching occurred, the impurities can be easily removed from products' solutions in dichloromethane using polymer-supported triphenylphosphine and thiourea [47][48][49][50][51].
Solvents: Acetonitrile (spectrometric grade, low water), toluene, heptane, ethyl acetate, t-butanol and ethanol was purchased from SDS France and used as such. Dimethylsulfoxide and dimethylformamide (spectrometric grade, low water) were placed oven 4 Å molecular sieves and kept under nitrogen. Dichloromethane (SDS France) was treated with phosphorus pentoxide at reflux (1 h) before being distilled.
Melting points (m.p.): Melting points were determined using a Kofler apparatus after a first evaluation, calibration with a reference sample of a mp near the observed fusion and final measure of the melting point.

Infrared spectroscopy (IR):
Infrared spectra were recorded neat on a Jasco FT/IR-4100 in ATR mode (PIKE-MIRacle) between 4 000 and 400 cm -1 and are given in ν (cm -1 ) Nuclear Magnetic Resonance (NMR): NMR spectra were recorded on a Bruker Avance DRX at 300 MHz for 1 H and 75.5 for 13 C. Chemical shifts (δ) are reported in part per million (ppm) relative to tetramethylsilane signal as an internal reference. Couplings constants (J) are in hertz and signal multiplicities indicated as s (singulet), d (doublet), t (triplet), q (quadruplet), m (multiplet), dd (doublet of doublet).

Procedures
Dry Amberlyst A-21 [43]: Commercial wet Amberlyst A-21 resin (Aldrich, 20-50 mesh, 100 g) was suspended in MeOH (500 mL) for 0.5 h and filtered (3 times) and then soaked in methylene chloride (500 mL) for 0.5 h and again filtered (3 times). The resulting resin was placed in a round-bottom flask on a rotatory evaporator and dried at 50 ºC under 10 mm Hg until it was free flowing. The dried resin was then kept overnight in vacuo in a desiccator over P 2 O 5 . Specifications from the manufacturer indicate that the polymer contains 4.8 mequiv of amine/g of dry resin. [43]: Dry Amberlyst A-21 (1.0 g, 4.8 mmol amine) was added to a solution of copper (I) iodide (381 mg, 2.00 mmol) in acetonitrile (15 mL) and gently shaken on an orbital stirrer for 17 h. The solvent was drawn off and the resin washed with CH 3 CN (2 × 15 mL), CH 2 Cl 2 (2 × 15 mL) and dried in vacuo (0.01 mm Hg) at 40 ºC. The weight increase was of 0,307 g (1.61 mmol CuI) that gave a polymer loading of 1.23 mmol CuI.g -1 . Elemental analyses (Service Central d'Analyses du CNRS, Solaize, France) gave a copper content of 8.64%, indicative of a loading of 1.35 mmol CuI.g -1 .

General procedure for automated synthesis of triazoles
Amberlyst A-21·CuI (1.35 mmol/g, 30 mg, 0.040 mmol, 8 mol %) was placed in one of the Chemspeed ASW-2000 reactor equipped with a plunging filter, leaving the paired reactor empty. The azide (0.55 mmol) and alkyne (0.50 mmol), both dissolved in 1 mL dichloromethane were sequentially added at 1mL·min -1 . The reactors were orbitally shaken at 600 rpm for 12 hours at room temperature. The product's solution was separated from the catalyst by filtration followed by washing of the catalyst by dichloromethane (2 × 2.5 mL). The combined extracts were evaporated to obtain the product.

CAUTION! Organic azides are potentially explosive and should be handle with care. Even if no incident occurred in this reaction on this scale, the cycloaddition can be exothermic and should not be attempted on a larger scale, without being aware of explosion risks.
1-Benzyl-4-hydroxymethyl-1,2,3-triazole (a1b1). Prepared from 28 mg (0.50 mmol) of a1 and 73 mg (0.55 mmol) of b1. The product was obtained as a white solid (92 mg, 97%). C 10

Conclusions
We have presented in this article our findings concerning a polymer-supported copper (I) catalyst for the Huisgen reaction. The results include the study of the polymeric catalyst in terms of loading, efficiency of different samples, its use in different solvents and catalytic levels. The catalyst was found to be more efficient in apolar aprotic solvents at 6-8 mol%. The catalyst can be recycled at least five time without loss of activity, and seems to be stable enough to use it straight from the shelf over a long time. Kinetics of this heterogeneous catalyst was compared to a homogeneous version of it. It was found that the solid-phase catalyst was two times slower than the soluble one, the reactions being complete in 6 hours, but it was giving better results due to the simple product isolation procedure. The catalyst was finally used in automated synthesis of a 96-triazole library. It gave very good results for most of the awaited compounds having a wide scope, but with some limitations for amine bearing alkynes, as an example. Despite this imitation, Amberlyst A-21•CuI is a very efficient tool for the quick preparation of the triazole nucleus with an isolation procedure by simple filtration and evaporation.
Amberlyst A-21•CuI has already found other applications in triazole synthesis in solvent-free approaches, modular flow reactor chemistry -directly from alkynes or via a sequential Seyferth-Gilbert homologation/A-21•CuI catalyzed Huisgen reaction, one-pot synthesis of triazoles with in situ preparation of azides and synthesis of triazole-containing macrocycles, as well as formation of αalkylidene cyclic carbonates from propargyl alcohols in super critical carbon dioxide [46][47][48][49][50][51]. Amberlyst A-21 was also used to support other metals, such as ytterbium and palladium, opening the way to other polymer-supported catalysts based of this low cost commercially available macroreticular resin [52][53][54].
We are sure that this approach and catalysts will find other very interesting applications in several reactions where a metallic catalyst is needed, improving the tools of the organic chemist in both regular synthetic and click chemistry.