Decarboxylative Coupling Reactions

Abstract

Transition metal-catalyzed decarboxylative couplings have recently emerged as a promising concept for C–C and C–heteroatom bond formation. Our contribution to this rapidly evolving field was the development of redox-neutral decarboxylative cross-couplings. In these catalytic transformations, carboxylic acids extrude CO₂ to give organometallic intermediates, which react with aryl electrophiles under regioselective formation of a new C–C bond. This reaction concept compares favorably to traditional cross-couplings involving preformed organometallic reagents, as it draws on stable and broadly available carboxylic acids as the source of carbon nucleophiles. In this chapter, we describe the invention process that resulted in the discovery of the first active catalyst systems and in the stepwise extension of this concept to a broadly applicable synthetic concept. A short overview on recent developments in this field is also provided.

1 Introduction

Catalytic decarboxylative cross-couplings have emerged within recent years as a powerful strategy to form carbon–carbon or carbon–heteroatom bonds starting from carboxylic acids. The impressive progress achieved in this field has been recently reviewed [1–4]. The carboxylate function is used as a leaving group to direct the coupling to an electrophilic center, with concurrent loss of carbon dioxide. This strategy allows for the regioselective coupling of specific types of carboxylic acids with aryl halides, Michael acceptors, nitriles, and other electrophiles to a range of valuable building blocks (Scheme 1). Under oxidative conditions, carboxylic acids can also decarboxylatively couple with nucleophiles. The basis for this rapidly developing field of research (170 publications since 2006) [5] was laid with the discovery of catalysts that effectively mediate the extrusion of carbon dioxide from aromatic carboxylic acids with formation of aryl-metal species and their incorporation into coupling reactions [6–10].

Scheme 1

Examples of decarboxylative coupling reactions

2 Defining the Research Target

The invention process of redox-neutral decarboxylative cross-couplings mediated by bimetallic catalysts began in 2003. While I was assistant professor (Habilitand) at the Max Planck Institute for Coal Research in Mülheim, my group had developed several sustainable metal-catalyzed synthetic methods based on carboxylic acid derivatives [11]. In April 2003, I gave a colloquium at Bayer AG on this topic. An animated discussion ensued with experienced industrial chemists, centered on synthetic problems that are regularly troublesome in industrial practice, but paradoxically viewed as “solved” by academic scientists. Examples of this were metal-catalyzed carbon–carbon bond formations, such as the Suzuki–Miyaura reaction. What particularly struck me that day was the difficulty of synthesizing boronic acids on industrial scales, as opposed to the ease of coupling them. In a subsequent conversation, Dr. Hugl (at that time head of Organic Synthesis Research at Bayer Chemicals) encouraged me to search for alternative strategies for carbon–carbon bond formation and promised generous funding for any convincing concept I would present for the regioselective synthesis of biaryls, however far-fetched it may be.

In the ensuing months, I drew up diverse approaches to novel biaryl syntheses on the blackboard and refined them in intense discussions with coworkers and colleagues in Mülheim. Most of the ideas had to be dismissed because we found no satisfactory catalyst cycle based only on thermodynamically feasible individual steps. The first concept to withstand these discussions consisted of using either arylphosphonic rather than arylboronic acids or aryl hexafluorophosphate rather than aryl trifluoroborate salts as the source of aryl nucleophiles in Suzuki-type coupling reactions. An analogous use of arylsulfinic acid derivatives also appeared to be feasible (Scheme 2). An even more speculative idea was to convert benzoate salts into aryl nucleophiles by extrusion of carbon dioxide.

Scheme 2

Alternative sources of carbon nucleophiles in Suzuki-type arylations

We used DFT calculations to clarify whether these substrate classes were at least theoretically suited as carbon nucleophiles. My Ph.D. student Debasis Koley investigated the polarity of the carbon–phosphorus and carbon–sulfur bonds and came to the conclusion that both reagents can be derivatized in a way that the adjacent carbon atoms would indeed become negatively charged. The decarboxylation of metal benzoates was predicted to be exergonic, but with a large activation barrier. A literature search revealed that Nilsson had shown already in 1968 that copper arenecarboxylates extrude carbon dioxide at high temperatures with formation of transient aryl-copper species. These are believed to act as intermediates in Cu-mediated protodecarboxylations [12–16]. All these pieces of evidence convinced us that at least one of these approaches had a chance of succeeding.

3 Designing a Catalytic Process

Following intensive evaluation of the literature and of potential reaction mechanisms, we drafted a research plan that we presented to specialists at Bayer in October 2003. In it, we elaborated on the use of arylphosphonic, arylsulfinic, and arylcarboxylic acids as alternatives to arylboronic acids in cross-coupling reactions. The decarboxylative version was placed third because at the time, we viewed this approach as the least promising, fearing that the temperatures required for it to work would be prohibitively high for industrial applications. For each reaction, the proposal included draft catalytic cycles composed for the most part of elementary steps known from other metal-catalyzed reactions or from stoichiometric conversions of organometallic compounds [17, 18]. The inventive effort lies mostly in designing these novel combinations of elementary steps. We took our chances with this proposal, lacking the financial means and personnel to support our plans with experimental studies.

Dr. Hugl and his colleagues at Bayer carefully examined our research plan and ultimately offered to finance the project for two years. They did not request any experimental support for the validity of the postulated reaction mechanisms, but based their decision solely on the theoretical calculations, mechanistic considerations, and our previous track record of successfully developing new catalytic transformations [19–25]. While they rated the probability for this project to succeed as very low, they viewed the potential gain for Bayer as so high that they were prepared to take their chances.

This industrial support was critical for the success of this high-risk project considering it took us almost one year to furnish the proof of concept, which then allowed us to submit an experimentally validated proposal to public funding agencies. Method developers regularly face the challenge of demonstrating preliminary work to support their proposals because it may take years to reach proof-of-concept stage for one of the many reactions they investigate. Once there, the newly discovered reaction can often be led to preparative utility within weeks.

In Dr. Guojun Deng, I found an excellent coworker, as is indispensible for such a challenging project, and we began our experimental work in early 2004. The first months were taken up by intensive investigations of the arylphosphonate cross-coupling. Whereas we had some encouraging results, we had to admit that the corrosiveness of the required reaction conditions was likely to impede any industrial implementation of this approach. For the sulfinic acid derivatives, we suspected similar issues, so that we moved straight on to the decarboxylative reactions.

Our original mechanism for decarboxylative reactions included a monometallic palladium catalyst related to that postulated in 2002 by Myers for his oxidative decarboxylative Heck reactions (Scheme 3) [6].

Scheme 3

Oxidative decarboxylative Heck reaction

In this process, carboxylic acids and olefins are converted to vinylarenes in the presence of palladium in catalytic amounts and silver in overstoichiometric amounts (Scheme 4). It is initiated by formation of a silver arenecarboxylate salt, which then reacts with a palladium(II) halide complex (a) to give the silver halide and palladium carboxylate b. The extrusion of carbon dioxide, and thus the conversion of the carboxylate to the organometallic species c, occurs within the ligand sphere of palladium. The subsequent steps, i.e., insertion of the olefin into the palladium–carbon bond, internal rotation, and β-hydride elimination, correspond to the classical Heck reaction [26]. An additional oxidation step converts the palladium(0) species f back to palladium(II). It is this oxidation by silver that causes the arylpalladium species c generated via the extrusion of CO₂ from a carboxylate ion to react as an aryl electrophile rather than aryl nucleophile.

Scheme 4

Proposed mechanism for Myers’ decarboxylative Heck reaction

We reasoned that such a decarboxylation step could also be employed in a redox-neutral cross-coupling reaction with carbon electrophiles. On this basis, we drew up a catalytic cycle that starts with an oxidative addition of aryl halides or pseudohalides to a coordinatively unsaturated palladium(0) species f (Scheme 5). The more weakly coordinating the leaving group X, the easier should be its subsequent replacement by a carboxylate. At least for X = OTf, the palladium(II) carboxylate h should form quantitatively, whereas for X = halide, it should be possible to enforce this step by employing silver or thallium salts as species g. The ensuing thermal decarboxylation of the palladium(II) intermediate i represents the most critical step. Myers’ results indicated that certain palladium(II) carboxylates liberate carbon dioxide on heating. However, it remained unclear whether arylpalladium(II) carboxylate complexes such as i would display a similar reactivity. If this were to be the case, they would form Ar–Pd–Ar′ intermediates k, which in turn are known to eliminate the desired biaryls. This would close the catalytic cycle with regeneration of palladium in the original zero oxidation state (f).

Scheme 5

First design of a catalytic decarboxylative cross-coupling

This first plan for a decarboxylative cross-coupling carried with it certain weaknesses for potential industrial applications. It was to be expected that the salt metathesis between alkali metal carboxylates and late transition metal halides would be thermodynamically disfavored. We expected the formation of a palladium benzoate complex i from palladium bromide complexes c and potassium benzoate (g) to proceed well only in the presence of a stoichiometric quantity of silver to capture bromide ions [27]. However, we did not like the idea of using stoichiometric quantities of silver salts or of expensive aryl triflates in the place of aryl halides. Finally, the published substrate scope of the oxidative Heck reaction led to concerns that palladium catalysts mediate the decarboxylation only of a narrow range of carboxylates, precluding use of this reaction as a general synthetic strategy.

These concerns were supported by the first test reactions in which we tried to couple benzoic acid with 4-bromoanisole in the presence of palladium complexes. We encountered great difficulties with generating the palladium benzoate from palladium bromide complexes in the absence of silver salts. Phosphine-stabilized palladium benzoates, especially arylpalladium(II) benzoates similar to i, did not extrude CO₂ even at high temperatures. Only a small spectrum of palladium arenecarboxylates and phenylacetates was identified that lost CO₂ upon heating. This is in agreement with a study of the protodecarboxylation activity of palladium published subsequently by Kozlowski [28].

In none of our extensive test reactions of aryl halides with benzoic acids carried out in the presence of diverse palladium catalysts, we were able to detect even traces of the unsymmetrical biaryl. Instead, we observed mostly homocoupling and dehalogenation products in some cases along with phenol esters. The conversion of aryl triflates with potassium benzoates did not lead to the desired biaryls, either, but to the phenol esters instead.

4 Designing a Bimetallic Catalyst System

In view of this lengthy period of discouraging results, we returned to the drawing board to look for alternative reaction mechanisms. At the time, it appeared that palladium-based protocols would at best allow us to convert a limited range of benzoates with a few particularly robust aryl triflates. We therefore decided to alter our concept by introducing a cocatalyst designed to effectively mediate specifically the decarboxylation step and then transfer the aryl residue to the palladium cross-coupling catalyst.

Nilsson et al. had previously demonstrated that arylcopper species intermediately form in copper-mediated protodecarboxylation reactions [12–16]. Moreover, they isolated 2-nitrobiphenyl along with nitrobenzene and several by-products in the pyrolysis of copper 2-nitrobenzenecarboxylate at 250°C in the presences of excess iodobenzene. Due to the limitations of known copper-mediated Ullmann-type couplings [29], namely, the low selectivity for the heterocoupling and stoichiometric use of copper, we were very skeptical that this trapping experiment could be turned into a preparatively useful biaryl synthesis. We thus sketched a mechanism in which two different metal complexes cooperatively catalyze the decarboxylative cross-coupling. The first complex, a copper, silver, or gold catalyst, would mediate the extrusion of carbon dioxide from arenecarboxylates. The second, a two-electron catalyst such as palladium, nickel, or rhodium, would be responsible for the cross-coupling with an aryl halide (Scheme 6).

Scheme 6

Pathway for a decarboxylative coupling cooperatively catalyzed by two metals

In the new catalytic cycle, the carboxylic acid is deprotonated with a mild base, and the carboxylate group is transferred onto a copper- or silver-based decarboxylation catalyst h via salt exchange with formation of the corresponding carboxylate complex g. The metal would then have to insert into the C–C(O) bond under extrusion of CO₂ to form a stable arylcopper or arylsilver intermediate l. For these polarizable late transition metals, the metal–carbon bond should be particularly stable in comparison to the metal–oxygen bond, thus providing a driving force for this step in addition to the entropically favored loss of gaseous CO₂. The carbon residue is then transferred onto an arylpalladium species c, generated in the reaction of the aryl electrophile with the second catalyst component, a low-valent palladium(0) species f. Finally, the carbon–carbon bond is formed in a reductive elimination from k, regenerating the original palladium species f. For further turnover of the decarboxylation catalyst to occur, the copper salt h formed in the transmetallation step would have to undergo a salt metathesis with the next carboxylate molecule. However, in the initial test reactions, we used stoichiometric amounts of copper or silver salts, so that this step was not yet required.

In these test reactions, we preformed copper(II) benzoate by treating benzoic acid with basic copper(II) carbonate, dried the resulting salt, and heated it to 150°C with 4-bromoanisole in the presence of various palladium complexes. This led to several products (Scheme 7). The formation of benzene indicated that the decarboxylation step worked, but that the intermediate arylcopper species was protonated before transfer of the aryl group to the palladium could occur. Radical processes could account for the formation of biphenyl. Debromination of 4-bromoanisole and its homocoupling to 4,4′-dimethoxybiphenyl gave evidence for a successful oxidative addition step, which, however, was not in balance with the remainder of the complex catalytic cycle. The desired cross-coupling product 4-methoxybiphenyl was detected first only in traces, and after optimizing temperature and reaction time, we were able to isolate an isomerically pure product in up to 10% yield.

Scheme 7

Products observed in the attempted decarboxylative couplings

I was now fully convinced that this project would succeed. I wrote a proposal for a Heisenberg research fellowship in which I drafted a research plan for the upcoming years and enthusiastically explained the opportunities of decarboxylative cross-couplings. When I showed it to my wife, who is also a chemist, she said, “But you don’t really think that this is actually going to work, do you?” Dr. Deng still set up an average of 20 carefully planned reactions per day, but I realized that he, too, was losing hope, and I was not surprised when he informed me that he had started writing job applications.

5 The Right Test Substrate

In a series of protodecarboxylation experiments with several benzoic acids, Dr. Deng confirmed our suspicion that our model compound, benzoic acid, was particularly unreactive, but found that 2-nitrobenzoic acid seemed to decarboxylate much more easily. At that time, the fungicide boscalid was starting its success story [30], and I encouraged him to switch to the reaction of 4-bromochlorobenzene with 2-nitrobenzoic acid as the test system. This renewed our motivation, as a successful coupling of these substrates would be of immediate interest as an alternative entry to the biaryl building block of boscalid (meanwhile produced on 1,500 tpa scale). Scheme 8 illustrates how a decarboxylative coupling could save two steps in the overall synthesis.

Scheme 8

Boscalid synthesis via decarboxylative coupling and Suzuki reaction

This new test reaction was the strike of fortune that led to the breakthrough in the catalyst development. Using preformed copper(II) 2-nitrobenzoate, we started to obtain reproducible yields of around 30%, which finally permitted us to systematically study the influence of catalyst components, solvents, and additives. Based on these findings, we optimized the catalyst system and steadily improved the yields. Interestingly, the yields seemed to hit a ceiling at 50% – we had to be missing something important. We found a potential explanation for this in a study by Kaeding et al., who reported that in the pyrolysis of Cu(II) benzoate, phenyl benzoate is formed as a by-product [31]. Maybe, only one of the two carboxyl groups coordinated to the copper was extruding CO₂, precluding the reaction to proceed to completion. We thus added stoichiometric amounts of various potassium salts to the copper 2-nitrobenzoate reaction mixtures to facilitate the formation of copper complexes with a single coordinated carboxylate.

On carnival 2005, I went to our laboratory at the RWTH Aachen to pick up Dr. Deng and take him to the traditional Rosenmontag parade. However, when I opened the door, I immediately realized that something had happened. Dr. Deng was smiling broadly, and he appeared to be in the middle of a chromatographic purification. Next to him was a beautifully clean gas chromatogram which showed the desired product in near-quantitative yield along with the internal standard only. Needless to say, we spend this carnival in the lab optimizing the first protocol of a decarboxylative cross-coupling. The reaction that had given such clean conversion contained KF as the additive (Scheme 9), and we saw the sharply increased yield as evidence for the intermediate formation of ArC(O)OCuF, which seemed to have a much lower decarboxylation barrier that the homoleptic copper(II) carboxylate.

Scheme 9

The first high-yielding decarboxylative cross-coupling protocol

Within a few days, we found a reliable reaction protocol in which a combination of the benzoic acid and aryl bromide in NMP is stirred for several hours at 120°C in the presence of stoichiometric amounts of basic copper carbonate and potassium fluoride, molecular sieves, and 2 mol% of a Pd(acac)₂/P(i-Pr)Ph₂ catalyst [36, 37]. The addition of molecular sieves, which effectively trapped the reaction water, allowed to deprotonate the benzoic acid in situ with carbonate bases. It was thus no longer necessary to use preformed and carefully dried copper benzoates. Beside copper, silver carbonate was also found to be effective, but due to the higher cost of this metal, we initially did not follow up on this.

6 The Transition to a Doubly Catalytic Protocol

Our first reaction protocol proved to be broadly applicable with regard to aryl bromides, but only in combination with 2-nitrobenzoic acid and a few other highly activated carboxylic acids (Scheme 9). It was a decisive first step, but failed to fulfill my high expectations. I will never forget the face of Dr. Deng when I told him that I did not yet want to publish the results of his meanwhile more than 1,200 reactions. I did not want to draw the attention of big research groups to this concept until our two-man team had achieved the next critical step and made the reaction catalytic also with regard to the decarboxylation catalyst. Fortunately, we did not know that despite all our mechanistic considerations, this would require another 800 experiments.

Whenever we tried to reduce the amount of basic CuCO₃ to 30 mol% and supplement the rest of it with K₂CO₃, the yields dropped to below the amount of copper employed. According to our proposed mechanism (Scheme 6), the copper salt should be regenerated after each transmetallation step, so that a catalytic amount should theoretically suffice. However, it was the anion exchange step from h to g that proved to be difficult. Raising the temperature above 120°C or prolonging the reaction time did not help. In either case we observed a color shift from green to brown, which we attributed to a partial reduction of Cu(II) to Cu(I).

We were reluctant to switch from copper(II) to copper(I) complexes as decarboxylation catalysts, having observed that copper(I) carboxylates lose CO₂ only above 160°C. In a systematic study we found that chelating bipyridine ligands, and 1,10-phenanthroline in particular, dramatically enhance the decarboxylation activity of copper(I) salts. After many experiments, we achieved more than one turnover of the copper using phenanthroline as the ligand for both copper and palladium at a temperature of 160°C. A rigorous exclusion of water was crucial to avoid protodecarboxylation becoming the main reaction pathway. At this stage, the exact reaction conditions, including heating rates and the purity of reagents and solvents, often influenced the reaction outcome more than the composition of the catalyst. In a situation like this, it is crucial to reproduce systematic studies several times before reaching any conclusions. If one reaction works better than the other, the reason may not be that one of the phosphines tested is more effective than others, but simply that of one of the phosphines is more pure. If the trend observed in a systematic study is hard to rationalize, our experience is that the experiments differ in more than just the intended way.

It took us some time to discover that the optimal way to trap the reaction water during the in situ deprotonation of 2-nitrobenzoate by potassium carbonate was to continuously distill off part of the reaction solvent. However, this strategy could not be used in screening experiments, and we thus added an excess of rigorously dried, finely powdered molecular sieves. This way, the model reaction gave almost quantitative yields using only 1% CuI, 0.5% Pd(acac)₂ and 3% phenanthroline in NMP at 160°C (Scheme 10).

Scheme 10

The first decarboxylative cross-coupling catalytic in two transition metals

Once we had found this protocol and acquired sufficient practice with the model reaction to reliably reproduce our own results, we went on to explore the generality of this protocol. We were delighted to find that it allowed coupling various aryl bromides and even some chlorides with 2-nitrobenzoic acid.

However, the extension to other carboxylic acids proved to be troublesome. Continuous fine-tuning of the copper–phenanthroline catalysts was required to extend the protocol catalytic in both metals first to carboxylic acids with other strongly coordinating groups in ortho position that significantly increase the copper-ligating quality of the carboxylate substrate (2-acyl, 2-formyl), then to carboxylic acids with weakly coordinating ortho substituents (2-fluoro, 2-cyanobenzoic acid), and finally to vinylic or heterocyclic derivatives (cinnamic, thiophenecarboxylic acid). Initially, we reoptimized the conditions for almost every carboxylic acid, until we eventually found more generally applicable catalyst systems [35, 37].

Our partners at Bayer, by now Lanxess, realized already at this early state of development that the new decarboxylative biaryl synthesis would open up opportunities for the industrial synthesis of high-value pharmaceutical intermediates and were pleased to find that it worked also on kilogram scales [32–34] (Scheme 11).

Scheme 11

Pilot plant synthesis of a bixafen intermediate on 50-kg scale

In October 2005, after Dr. Deng had already left my group and I had moved to the TU Kaiserslautern, we finally submitted our results for publication. However, the referees were skeptical, viewing our reaction as an extension of existing cross-coupling technology. In contrast, my wife was now so convinced of the wide-ranging opportunities of decarboxylative cross-couplings that she would not allow us to give up now rather than addressing each individual referee’s comment. It took more than a year and as many as three resubmissions to see the manuscript accepted [35]. One argument that was repeatedly held up against this reaction was that it releases a stoichiometric amount of CO₂ and thus contributes to global warming. Another was that the reaction scope was still very limited and that it might well be intrinsically limited to ortho-substituted carboxylic acids [36, 37]. The latter was indeed a critical issue, and we addressed it with highest priority.

7 Overcoming Substrate Limitations

In order to evaluate what caused the limitation to ortho-substituted carboxylic acids, we investigated the decarboxylation step separately from the cross-coupling step. In the presence of only the copper(I) phenanthroline/quinoline decarboxylation catalyst (15 mol%), a wide range of aromatic, heteroaromatic, and vinylic carboxylates protodecarboxylated to the corresponding hydrocarbons on heating to 160°C (Scheme 12). However, added halide ions, as would inevitably be released in a decarboxylative cross-coupling (Scheme 6), prevented the decarboxylation for all non-ortho-substituted or not otherwise activated aromatic carboxylic acids.

Scheme 12

Effect of halides on copper-catalyzed protodecarboxylations

The results suggested that the halides competed with the carboxylates for coordination sites at the copper. Therefore, a general process catalytic in both metals should be possible as soon as decarboxylation catalysts with a stronger preference for carboxylate over halide ions could be found. An alternative strategy to achieve coupling the full range of aromatic carboxylates was to avoid the formation of halide salts altogether, by employing electrophiles with non-coordinating leaving groups such as aryl triflates. Unfortunately, known palladium catalysts for the cross-couplings of aryl triflates require special catalyst systems that stabilize cationic palladium intermediates with chelating phosphines and/or added halides. As both these additives had previously found to impede the copper-mediated decarboxylation step, it was a real challenge to find a suitable catalyst system for the desired reaction.

Dr. Nuria Rodríguez, who joined my group as postdoc in 2006, led this project to success. She carefully investigated how the individual components of the decarboxylation and cross-coupling catalysts interact with each other and found that the clue to success lies in the choice of the phosphine ligand. A catalyst system consisting of copper(I) oxide, 1,10-phenanthroline, and palladium(II) iodide along with the sterically demanding, moderately electron-rich chelating phosphine Tol-BINAP ideally strikes the delicate balance of stabilizing the palladium on one hand and maintaining the decarboxylation activity of the copper at the highest possible level on the other. As anticipated, the triflate ions released in the process were unable to block the carboxylates out of the coordination sphere of the decarboxylation catalyst, so that this protocol allowed the conversion of a much broader range of carboxylate substrates, now also including meta- and para-substituted benzoates (Scheme 13) [38, 39]. This gave the experimental proof that decarboxylative cross-couplings are not intrinsically limited to a narrow range of substrates, but indeed held the potential to become a generally applicable technology for regioselective C–C and C–heteroatom bond formation.

Scheme 13

Decarboxylative coupling of benzoate salts with aryl triflates

8 State of the Art in Decarboxylative Couplings

Since 2006, the efficiency of decarboxylative couplings has steadily been improved, and they are on the way to becoming standard tools for organic synthesis [1–4]. A growing number of researchers have developed a wealth of catalytic transformations starting from carboxylic acids with release of CO₂, providing new synthetic entries to various valuable product classes. Some of the new decarboxylative couplings seem to have been inspired mainly by the work of Saegusa [8] and Tsuji [9] on decarboxylative allylations [11, 40–45], others by Myers’ oxidative Heck reaction [6], and yet others by the redox-neutral cross-coupling processes discussed above. Only a few of them can be introduced here.

The applicability of the decarboxylative biaryl synthesis was rapidly extended to a broad range of aryl electrophiles with the help of new catalyst generations, including not only aryl iodides, bromides, and triflates but also the inexpensive but unreactive aryl chlorides and tosylates [46, 47]. Its preparative utility was demonstrated, e.g., in the synthesis of telmisartan and valsartan [48, 49]. The key factor in these advances was the identification of ligands that strongly activate the palladium catalysts toward oxidative addition steps while not interfering with the decarboxylation activity of the copper cocatalysts (Scheme 14).

Scheme 14

Decarboxylative couplings of aryl chlorides

Besides bimetallic palladium/copper systems, palladium/silver catalysts have been employed. The silver catalysts allow for lower reaction temperatures than the copper catalysts, but can be used in substoichiometric amounts only in couplings of aryl triflates in combination with certain ortho-substituted benzoic acids [50, 51]. The design of low-temperature decarboxylation catalysts was guided by DFT calculations, in which the influence of the central metal and ligand environment on activation energies was systematically studied [52–54]. Calculations revealed that 2-fluorobenzoates of silver or gold should decarboxylate much more easily than the corresponding copper complexes. These findings, which were confirmed in experimental studies, led to the development of a palladium/silver-catalyzed decarboxylative coupling that proceeds already at moderate temperatures (Scheme 15) [55]. This is a good example of how DFT calculations can support rational catalyst development and help to reduce the number of screening experiments needed. In this and several other decarboxylative couplings, microwave heating was demonstrated to have a beneficial effect [76]. Flow reactors have also been shown to be advantageous, as they mimize the exposure of the products to thermal stress [77].

Scheme 15

Ag-/Pd-catalyzed low-temperature decarboxylative biaryl synthesis

The ability of palladium to mediate decarboxylations of certain heteroaromatic carboxylates was demonstrated by Steglich et al. for intramolecular couplings, as well as by Forgione and Bilodeau for intermolecular couplings (Scheme 16) [56–58]. However, the scope of this reaction indeed seems to be more limited than its bimetallic counterpart. It so far includes mainly five-ring heteroarenes with carboxylate groups in the 2-position. Alkynylcarboxylic acids have also been decarboxylatively coupled with palladium alone [59].

Scheme 16

Arylation of heteroarenecarboxylates catalyzed by monometallic Pd systems

Monometallic copper complexes were also found to mediate decarboxylative couplings, but these reactions have a rather limited scope as well [60]. They are advantageous particularly for the coupling of polyfluorinated benzoic acids with aryl halides (Scheme 17).

Scheme 17

Decarboxylative cross-couplings mediated by monometallic copper systems

The application range of decarboxylative couplings has continuously been extended from aromatic carboxylic acids and β-oxocarboxylic acids to propiolic, phenylacetic, α-oxocarboxylic acids and oxalic acid monoesters [1–4]. The use of α-carbonyl carboxylic acids is particularly surprising as it is counterintuitive that unprotected acyl anion equivalents can be generated by decarboxylation. This allows a polarity reversal of the bond formation of traditional ketone syntheses (Scheme 18, right side, bottom) [61]. In the presence of amines, the corresponding azomethines are formed (Scheme 18, right side, top) [62, 63]. The analogous coupling of oxalic acid monoesters allows the convenient one-step synthesis of benzoate esters from aryl halides (Scheme 18, left side) [64].

Scheme 18

Decarboxylative couplings of α-oxocarboxylic acid derivatives

The in situ generation of carbon nucleophiles via extrusion of CO₂ from benzoates cannot only be combined with cross-coupling processes but also with 1,2- and 1,4-addition reactions. An example is the rhodium-catalyzed decarboxylative conjugate addition of activated benzoic acids to acrylic esters or amides developed by Zhao et al. (Scheme 19, right side) [65]. A nice application is the decarboxylative addition of aromatic carboxylic acids to nitriles in the presence of a rhodium catalyst by Larhed et al., which proceeds under neutral conditions and allows a novel, waste-minimized synthesis of aryl ketones (Scheme 19, left side) [66].

Scheme 19

Decarboxylative conjugate addition reactions

Myers’ oxidative decarboxylative Heck reaction became the prototype for a whole series of regiospecific oxidative couplings in which carboxylic acids adopt the reactivity of aryl electrophiles in the corresponding redox-neutral processes [67–72]. Crabtree et al. developed a process in which arenes react with aromatic carboxylates under C–H activation in the presence of a palladium catalyst and excess silver carbonate to yield biaryls. This reaction is useful especially for intramolecular couplings (Scheme 20) [73, 74]. Recently, a palladium-free, silver-catalyzed radical variant has been disclosed [78].

Scheme 20

Decarboxylative coupling with C–H activation

Under oxidative conditions, α-oxocarboxylic acids seem to decarboxylate at particularly low temperatures. This reactivity was utilized by Fang et al. in palladium-catalyzed decarboxylative α-acylations of acetanilides, which proceed already at room temperature (Scheme 21) [75]. This last example illustrates once again that decarboxylative couplings do not inherently require high temperatures, which nurtures the hope that someday, such low temperatures can be reached also for redox-neutral decarboxylative couplings. Other examples of oxidative decarboxylative couplings are biaryl syntheses starting from boronic esters and carboxylic acids [79], and homo- and heterocouplings of arenecarboxylic acids [80, 81].

Beyond C–C bond forming reactions, decarboxylative couplings have recently found application in regiospecific formations of carbon-halogen [82, 83], carbon-sulfur [84, 85], carbon-phosphorus [86, 87], carbon-nitrogen [88], and carbon-oxygen bonds [89].

Scheme 21

Decarboxylative ortho arylation of acetanilides

9 Conclusion and Outlook

The invention of decarboxylative cross-couplings was driven by a combination of rational catalyst and process design. In addition, it required large numbers of individually planned serial experiments. These reiterative series of parallel reactions were indispensable to generate data points laying the basis for our understanding of how each single component influenced the complex catalyst system. Based on these systematic studies, increasingly effective protocols could be devised. One of the indispensable strikes of luck was a talk on boscalid at a critical moment in time that inspired us to change our model system to a substrate combination, that turned out to be ideal.

The selected examples demonstrate that only few decarboxylative couplings are close to synthetic maturity. Many inventive steps are still required to overcome remaining limitations and to unleash the full potential of this new reaction concept. New decarboxylation catalysts have to be designed that operate at low temperatures and tolerate a broad variety of carboxylic acids. For several decarboxylative reactions, the proof of concept has been achieved only for a very limited scope of substrates. Extending these processes to the full range of carboxylic acids remains a challenging task.

Most importantly, many decarboxylative reactions are still to be invented as a replacement of traditional coupling reactions based on preformed organometallic reagents.