Research Article

MOF-derived cobalt nanoparticles catalyze a general synthesis of amines

See allHide authors and affiliations

Science  20 Oct 2017:
Vol. 358, Issue 6361, pp. 326-332
DOI: 10.1126/science.aan6245

A MOF sets the stage to make amines

Reductive amination is a common method that chemists use to make carbon-nitrogen bonds. The reaction, which often requires precious metal catalysts, couples ammonia or other amines with carbonyl compounds and then with hydrogen. Jagadeesh et al. report a class of nonprecious cobalt nanoparticles that catalyze this reaction across a very broad range of substrates, including complex molecules of pharmaceutical interest (see the Perspective by Chen and Xu). The cobalt was first embedded in a metal-organic framework (MOF), which, upon heating, transformed into a graphitic shell. The catalyst could be conveniently separated from products and recycled up to six times.

Science, this issue p. 326; see also p. 304


The development of base metal catalysts for the synthesis of pharmaceutically relevant compounds remains an important goal of chemical research. Here, we report that cobalt nanoparticles encapsulated by a graphitic shell are broadly effective reductive amination catalysts. Their convenient and practical preparation entailed template assembly of cobalt-diamine-dicarboxylic acid metal organic frameworks on carbon and subsequent pyrolysis under inert atmosphere. The resulting stable and reusable catalysts were active for synthesis of primary, secondary, tertiary, and N-methylamines (more than 140 examples). The reaction couples easily accessible carbonyl compounds (aldehydes and ketones) with ammonia, amines, or nitro compounds, and molecular hydrogen under industrially viable and scalable conditions, offering cost-effective access to numerous amines, amino acid derivatives, and more complex drug targets.

The development of nanostructured catalysts for innovative organic synthesis is crucial for the advancement of sustainable processes in the chemical, pharmaceutical, and agrochemical industries (18). However, most of the nanocatalysts known to date were developed for the activation of structurally simple molecules (38) and are scarcely applied for more challenging synthetic reactions or refinement of structurally complex compounds. Among the different classes of nanomaterials, supported metal particles are particularly valued for their low energy consumption and high activities and selectivities (114), which can be tuned by controlling the nature, size, distribution, and stability of the nanoparticles as well as surface composition of the support (914). In general, simple supported metal nanoparticles are prepared through thermal (via pyrolysis or calcination) or chemical reduction of the respective metal salts on heterogeneous supports (110). To improve the activity of these simple materials, well-defined organometallic complexes have also recently been explored as precursors for pyrolytic activation (13, 14). Although the resulting materials found application in catalytic hydrogenations and oxidations (13, 14), they largely exhibit poor reactivity and selectivity for other challenging synthetic organic reactions. One key to more rational preparation of active nanoparticulate catalysts might be the use of structure-controlling templates, which should strongly bind to the respective metal ions. In this respect, metal organic frameworks (MOFs) represent a stable class of porous compounds, which can be assembled in a highly modular manner from metal ions and organic linkers (1517). Recently, they have been used as self-sacrificing compounds for producing bulk materials via their direct pyrolysis without the use of external heterogeneous supports (1823). Here, we describe the use of MOFs as specific precursors for producing highly active and stable cobalt (Co) nanoparticles supported on commercial carbon (C). In contrast to previous works (1823), the preformed material (Co-diamine-dicarboxylic acid MOF) acted as a structure-directing template for pyrolysis on carbon. Compared with more conventional routes to heterogeneous materials prepared by impregnation or immobilization, in our procedure the pyrolysis precursor is better defined. We applied the resulting nanoparticles to reductive amination of carbonyl compounds for the general synthesis of amines.

Amines represent a privileged class of compounds used extensively in fine and bulk chemicals, pharmaceuticals, and materials (2429). In fact, 170 of the top 200 drugs of 2015 contained nitrogen (N) and/or amino groups (26). Catalytic reductive amination of carbonyl compounds with molecular hydrogen is widely performed in research laboratories and industrial facilities (3048). However, the preparation of primary amines by using ammonia relies on either precious metal–based catalysts (30, 31, 3339) or Raney nickel (Ni) (30, 31, 40), which often remains nonselective or otherwise problematic (30, 31, 3440). Hence, the development of more selective and earth-abundant metal–based catalysts continues to attract scientific interest.

Preparation of MOF-derived cobalt nanoparticle-based catalysts

We initially explored MOFs assembled in dimethylformamide (DMF) at 150°C from Fe, manganese (Mn), Co, Ni, and copper (Cu) nitrates and organic linkers 1,4-diazabicyclo[2.2.2]octane (DABCO) and terephthalic acid (TPA). Next, these in situ–generated MOFs were immobilized on commercial Vulcan XC 72R carbon powder. Upon slow evaporation of the solvent followed by drying of the resulting MOFs, the template on C was formed (Fig. 1). Subsequent pyrolysis at 800°C for 2 hours under argon atmosphere produced the final nanoparticles. In addition, the Co-DABCO-TPA MOF was separately prepared and isolated before pyrolysis on carbon (detailed preparation is provided in the supplementary materials, materials and methods S2.2). Hereafter, all these materials are labeled as M-L1-L2@C-x, where M = Fe, Mn, Co, Ni, or Cu; L1 = DABCO; and L2 = TPA; and x denotes the pyrolysis temperature, respectively.

Fig. 1 Preparation of graphitic shell–encapsulated Co nanoparticles supported on carbon by using MOF precursors.

All of these materials were tested in the benchmark reductive amination of 3,4-dimethoxybenzaldehyde by using ammonia and hydrogen to give veratrylamine (3,4-dimethoxybenzylamine), which is present as a structural motif in several bioactive molecules. As shown in fig. S1, materials resulting from pyrolysis of Fe-, Mn-, Cu-, and Ni-DABCO-TPA@C at 800°C showed no or only little catalytic activity. Gratifyingly, the corresponding Co catalyst (Co-DABCO-TPA@C-800) from Co-MOFs gave a very good yield (88%) of the desired amine. The pyrolyzed materials prepared from the in situ–generated and the isolated Co-MOFs exhibited similar activity. Pyrolysis of cobalt nitrate with either DABCO (Co-DABCO@C-800) or TPA (Co-TPA@C-800) linkers alone led to less active catalysts (15 to 20% yields). Apart from the metal and the linkers, the pyrolysis temperature was important for activity. Specifically, pyrolysis at 400°C gave material with poor reactivity (16% yield), whereas pyrolysis at 600°C and 1000°C resulted in active catalysts (75 and 83% yields, respectively). As expected, the homogeneous catalyst and the pyrolyzed cobalt nitrate [Co(NO3)2@C-800] were not active. Similarly, the Co-MOFs with and without support showed no product formation (fig. S1).

Characterization of Co nanoparticle-based catalysts

We undertook detailed structural characterization of these Co-based catalysts. Aberration-corrected scanning transmission electron microscopy (STEM) analysis of the most active material (Co-DABCO-TPA@C-800) revealed the formation of mainly metallic Co particles, with diameters ranging from <5 to 30 nm (Fig. 2). The energy-dispersive x-ray spectroscopy map (EDXS) (Fig. 2A, left) shows mainly the presence of metallic Co particles within the carbon matrix. Most of these are surrounded by a combination of some graphitic layers and short-range ordered graphitic shells (Fig. 2A, middle). In addition, a smaller quantity of core-shell particles with cobalt oxide (Co3O4) shells at metallic Co is also present (fig. S2A). In regions of short-range ordered C, we detected single Co atoms as bright dots in high-angle annular dark field (HAADF) images (Fig. 2A, right). To get information on the Co/C/N relation, we performed the parallel mapping of EDXS for all elements and electron energy loss spectroscopy (EELS) (Fig. 2B) optimized for C, N, and oxygen (O). Because the N signal is superimposed on the C signal in EDXS, we used these maps only for the Co distribution. Shown in Fig. 2B are the maps of C, N, and Co (Fig. 2B, left image) in the neighborhood of a metallic particle wrapped by graphitic carbon. The C/N overlay in the HAADF image (Fig. 2B, middle image) gives evidence that N is not only located in the graphitic shell on the Co particle but surprisingly in short-range ordered C, which is not part of the graphitic shell and corresponds to features with single atoms shown in Fig. 2A. Co traces are detectable everywhere in the N containing C at low concentration. In contrast, the less active material Co-DABCO@C-800 contained mainly hollow Co3O4 particles (fig. S2B). In addition to Co3O4, some Co–Co3O4 core-shell particles were also present. Although subnanometer Co structures are found in this material, no single Co atoms were detected. Similarly, Co-TPA@C-800 (fig. S2C) also contained mainly Co3O4 particles encapsulated within graphitic shells, along with a small quantity of metallic cobalt in Co–Co3O4 core-shell structures. No single Co atoms or subnanometer Co structures were detected in this material. Cobalt nitrate@C-800, which was completely inactive, contained hollow Co3O4 with short-range ordered C from the support in the vicinity (fig. S2D).

Fig. 2 Catalyst characterization.

(A) EDXS map and STEM images of Co-DABCO-TPA@C-800 catalyst. (Left) EDXS map. Co, red; oxygen, blue; carbon, green. (Middle) Black arrows highlight graphite-embedded metallic Co particles. (Right) White circles represent single Co atoms at C. (B) EELS/EDXS study of Co-DABCO-TPA@C-800 catalyst. (Left) Parallel detected space-resolved C (EELS), N (EELS), and Co (EDX) signals. (Middle) HAADF image overlaid by C/N elemental map at the position of measurement. (Right) N K-edge EELS spectrum of the white box in the middle image.

In order to understand the formation mechanism of the active catalyst, materials pyrolyzed at lower temperatures were also characterized (fig. S3). In Co-DABCO-TPA@C-400, a minor amount of metallic Co was present in the core of Co/Co oxide core-shell structures, and no formation of graphitic shells was observed. Co-DABCO-TPA@C-600 contained more metallic Co, and incipient formation of graphitic shells enveloping the metallic Co was evident. Apparently, this structural process is crucial for high activity and stability. For the most active Co-DABCO-TPA@C-800, single Co atoms within some of the graphitic structures were detected. In the case of Co-DABCO-TPA@C-1000, most of the Co was present in metallic crystallite morphology completely covered by graphitic structures. In all of the active catalysts, cloudy regions of Co species in the 1- to 2-nm range were detected. The different phases of Co in both active and less active catalysts have been also confirmed with x-ray powder diffraction data (figs. S7 and S8) that accorded with the TEM analysis.

The nature and quantity of nitrogen in these materials was further explored with x-ray photoelectron spectroscopy (XPS) (fig. S10 and S11). Surprisingly, the combination of the two linkers increased the quantity of N in the near-surface region significantly compared with either linker alone (fig. S12). The N content in Co-DABCO-TPA@C-800 was three times higher than in Co-DABCO@C-800, whereas in Co-TPA@C-800, only traces of N were observed. In both Co-DABCO-TPA@C-800 and Co-DABCO@C-800, two N-states could be detected (fig. S10): one correlating with imine-like N known from pyridine (~398 eV) (49, 50), the other manifesting a higher binding energy, corresponding to N bonded to the metal (49, 50). For the former sample, a clear separation of both peaks was observed because of the slightly higher binding energy. For Co-DABCO-TPA@C pyrolyzed at different temperatures, iminic N was observed, and the binding energy of the Co–N bond increased as pyrolysis temperature ascended up to 800°C (fig. S11). In comparison with all other samples, the observation of the two N states was specific to these active systems (fig. S11) (49, 50). It seems that for the optimal catalyst, the bonding between Co and N is most pronounced. In the material pyrolyzed at 1000°C, the formation of nitrides could be observed as well (fig. S11). In the unpyrolyzed and 1000°C samples, the amount of Co in the near-surface region was too low for a reasonable peak fitting; for all other samples, the metal content was nearly the same.

Synthesis of primary amines

With an active material (Co-DABCO-TPA@C-800; hereafter represented as catalyst I) in hand, we investigated its substrate scope in the reductive amination. Initially, various primary amines, which are easily further functionalized and therefore represent central building blocks, were prepared starting from carbonyl compounds and simple ammonia (Figs. 3 and 4). A crucial problem for the synthesis of primary amines has been the subsequent formation of secondary and tertiary amines. The few catalysts known for the chemoselective reductive amination of aldehydes and ketones to give primary amines are based on precious metals (30, 31, 3439) or Raney Ni (30, 31, 40). Although Raney nickel is available and comparably inexpensive, it is highly flammable, hazardous, and less selective for functionalized and structurally complex molecules.

Fig. 3 Co-DABCO-TPA@C-800–catalyzed reductive amination of aldehydes for the synthesis of linear primary amines.

Reaction conditions are 0.5 mmol aldehyde, 25 mg catalyst I [3.5 mole % (mol %) Co], 5 to 7 bar NH3, 40 bar H2, 3 ml t-BuOH, 120°C, and 15 hours. Isolated yields are reported unless otherwise indicated. Asterisk indicates GC yields by using n-hexadecane standard. Dagger symbol (†) indicates reaction for 8 hours. Isolated as free amines and converted to hydrochloride salts for measuring NMR and HRMS spectra.

Fig. 4 Co-DABCO-TPA@C-800–catalyzed reductive amination of ketones.

Reaction conditions are 0.5 mmol ketone, 25 mg catalyst I (3.5 mol % Co), 5 to 7 bar NH3, 40 bar H2, 3 ml tetrahydrofuran (THF) (dry), 120°C, and 15 hours. Isolated yields are reported unless otherwise indicated. Asterisk indicates GC yields by using n-hexadecane standard. Dagger symbol (†) indicates reaction for 20 hours. Double dagger (‡) symbol indicates reaction for 24 hours. Section symbol (§) indicates reaction for 30 hours, with 35 mg catalysts. Isolated as free amines and converted to hydrochloride salts for measuring NMR and HRMS. Parallel bars (||) indicate 5 to 50 g substrate, 25 mg catalyst I (3.5 mol % Co) for each 0.5 mmol substrate, 5-7 bar NH3, 40 bar H2, 120-150 ml dry THF, 120°C, 15 to 30 hours, and isolated yields.

By applying our Co catalyst, we performed selective reductive amination of 39 aldehydes to produce benzylic, heterocyclic, and aliphatic linear primary amines, in good to excellent yields (up to 92%) (Fig. 3). Sensitive functional groups including halides, esters, as well as challenging C–C double and triple bonds were well tolerated. We also prepared amino-salicin in 83% yield from the O-glucoside helicin (Fig. 3, product 41) to showcase the selective amination of bioactive compounds. Compared with benzaldehydes, the reductive amination of aliphatic aldehydes is hampered by unproductive aldol condensation, which can easily occur under basic conditions. Nevertheless, several aliphatic and araliphatic substrates gave the corresponding primary amines in good yields (up to 92%) and high selectivity.

The hydrogenation of the in situ–generated imine from aldehydes is faster than reduction of the corresponding ketone-derived imines. Thus, a more active catalyst or more drastic conditions tend to be required for efficient reductive amination of ketones. The generality of catalyst I was further demonstrated by the synthesis of branched primary amines starting from diverse ketones (Fig. 4). Gratifyingly, both industrially relevant and structurally challenging ketones underwent smooth amination and produced the corresponding primary amines in good to excellent yields (Fig. 4). Biologically active amphetamines (products 77 to 79), which are potent central nervous system (CNS)–stimulating drugs, were prepared in up to 91% yield. In general, amphetamines are synthesized through reductive amination of the corresponding phenyl-2-propanones by using ammonium formate at higher temperature (Leuckart-Wallach reaction) (51). More sensitive products are prepared through reductive amination reactions by using more expensive platinum- and palladium (Pd)–based catalysts (51). We also explored amination of nonsteroidal anti-inflammatory agents (80 to 82) and steroid derivatives (83 to 87). Introduction of amino groups into these bioactive molecules has been scarcely explored so far. Sulfur-containing compounds constitute a common poison for most heterogeneous catalysts, although they are found in more than 300 U.S. Food and Drug Administration–approved drugs (52). In this context, thioethers are the most common scaffolds. Gratifyingly, catalyst I tolerated several S-containing substrates, including thiophene, thioether, and sulfone (Figs. 3 and 4, products 28, 29, 30, 43, 46, and 50).

Synthesis of secondary and tertiary amines

We next explored the synthesis of secondary and tertiary amines (Fig. 5), which are found in a large number of biologically active natural products (2629). In contrast to the preparation of primary amines by using ammonia (vide supra), a few non-noble-metal–based catalysts have already been used for reductive aminations to prepare secondary and tertiary amines; however, in such cases functional group tolerance and broad substrate scope have been limited (4148). Both nitroarenes and amines reacted with aldehydes to give the corresponding secondary amines selectively in up to 92% yields (Fig. 5A, products 88 to 103). Compared with the traditional reaction of anilines, the one-pot reductive amination of nitroarenes and carbonyl compounds is straightforward and ensures a better step economy. This direct process shows that the Co nanocatalyst can also be used for the selective hydrogenation of nitroarenes to anilines, which are of additional interest for dye formation and materials synthesis. The reductive alkylation of N-containing heterocycles as well as primary and secondary amines gave the corresponding derivatives in 81 to 88% yields. To demonstrate the amination of ketones, phenyl-2-propanone was reacted with 4-aminocyclohexane or 4-aminocyclohexanol (aliphatic amines) and produced the corresponding secondary amines in 84 and 89%, respectively (Fig. 5A, products 104 and 105). Further, we also tried the reductive amination of acetophenone using nitroarenes or aromatic amines but did not observe any substantial desired product formation in this case. Competition reactions of aniline with 4-bromobenzaldehyde and acetophenone were examined. Here, reductive amination of the aldehyde occurred with excellent conversion and high selectivity, whereas the ketone remained untouched. Similarly, the reaction of 4-bromobenzaldehyde with aniline proceeded selectively in the presence of phenyl-2-propanone. Thus, this catalyst allows for selective amination of aldehydes in the presence of ketones. In general, for the majority of reductive amination reactions high chemoselectivity is obtained. However, in a few cases we observed minor amounts of the corresponding alcohol (<5%) and/or secondary imine/amine (<10%). In the preparation of secondary amines, up to 10% of the imine was observed. In bromo-substituted substrates (products 6, 7, 89, 112, 117, 122, and 125), we also observed small amounts of dehalogenation. In general, we did not detect any over-alkylation products.

Fig. 5 Co-DABCO-TPA@C-800–catalyzed synthesis of secondary and tertiary amines.

(A) Survey of nitroarenes and amines. Reaction conditions are 0.5 mmol nitroarene, 0.75 to 1 mmol aldehyde, 25 mg catalyst I (3.5 mol % Co), 20 mg amberlite IR-120, 3 ml t-BuOH, 120°C, 24 hours, and isolated yields. Asterisk indicates 0.5 mmol amine, 0.75 mmol aldehyde. Dagger symbol indicates 30 mg catalyst and 30 hours. Double dagger symbol indicates 0.5 mmol amine and 0.75 mmol ketone. (B) Reductive N-alkylation of chiral amines. Reaction conditions are 10 mmol amine, 15 mmol benzaldehyde, 500 mg catalyst I (3.5 mol % Co), 400 mg amberlite IR-120, 15 ml t-BuOH, 120°C, 24 hours, and isolated yields. (C) Reductive N-alkylation of amino acid esters. Reaction conditions are 0.5 mmol amino acid ester, 0.75 mmol aldehyde, 25 mg catalyst (3.5 mol % Co), 20 mg amberlite IR-120, 3 ml t-BuOH, 120°C, 24 hours, and isolated yields. (D) Preparation of existing drug molecules. Reaction conditions are 1 mmol amine, 1.5 mmol aldehyde, 50 mg catalyst I (3.5 mol % Co), 3 ml t-BuOH, 120°C, 24 hours, and isolated yields. Asterisk indicates 75 mg catalyst for 30 hours. Dagger symbol (†) indicates 2 mmol amine and 1 mmol aldehyde. Double dagger (‡) symbol indicates synthesis same as daggered products followed by acylation with acid chlorides (detailed procedure is provided in the supplementary materials, materials and methods S6.5).

To demonstrate compatibility with preexisting stereochemistry, we performed the reductive N-alkylation with chiral amines. A plethora of chiral amines is available from biological and industrial sources. As shown in Fig. 5B, the N-alkylations of (S)-(−)-1-methylbenzylamine and (R)-(+)-α-methylbenzylamine proceeded efficiently and offered the corresponding chiral amine derivatives with retention of the original stereochemistry (Fig. 5B, products 106 to 108), as assessed with chiral high-performance liquid chromatography (HPLC) [procedure and HPLC data are available in the supplementary materials, materials and methods S6.4(a)]. Further, the reductive alkylation of two amino acid esters (phenylalanine and tyrosine) was performed in 82 to 89% yields (Fig. 5C, products 109 to 114). However, the chirality of these N-alkylated amino acid esters was not retained, and we obtained the corresponding racemic mixtures.

Synthesis of N-methylamines

Among the various alkylated amines, N-methylamines are of special interest because of their role in regulating biological functions (53, 54). In general, this class of amines is prepared either through Pd/C–catalyzed reductive amination with formaldehyde or by using active (toxic) methylation reagents. By applying our Co catalyst I, we prepared selected N-methylamines starting from aldehydes and N,N-dimethylamine (DMA), which is also a readily available bulk chemical (Fig. 6A). In addition, the direct reductive methylation of nitroarenes or amines with aqueous formaldehyde produced selectively the corresponding N-methylamines (Fig. 6B). Compared with traditional alkylations by using methyl-X compounds, advantageously the presented Co-catalyzed synthesis of N-methylamines is either more cost effective or waste-free.

Fig. 6 Co-DABCO-TPA@C-800–catalyzed preparation of N-methylamines.

(A) Reactions of aldehydes with dimethylamine. Conditions are 0.5 mmol aldehyde, 100 μl aqueous dimethylamine (40%), 25 mg catalyst I (3.5 mol % Co), 3 ml t-BuOH, 120°C, and 24 hours. Isolated yields are reported unless otherwise indicated. Asterisk indicates GC yields by using n-hexane standard. (B) Reactions of nitroarenes or amines with formaldehyde. Conditions are 0.5 mmol nitroarene, 100 to 200 μl aqueous formaldehyde (37%), 1:1 THF- H2O (3ml), and isolated yields. Asterisk indicates 0.5 mmol amine, 100 to 200 μl aqueous formaldehyde (37%), and 1:1 THF- H2O (3 ml).

Demonstrating catalyst recycling and gram-scale reactions

Stability and recyclability are crucial features for any heterogeneous catalyst. In addition to the obvious cost advantages, the use of recyclable heterogeneous catalyst can considerably facilitate product purification. As shown in the reductive amination of phenylacetone, our supported Co nanoparticles are highly stable and are conveniently recycled up to six times without any significant loss of catalytic activity (fig. S14). After that, slight decrease of activity is observed.

Next, gram-scale reactions were performed for several interesting substrates. Apart from amphetamine, related 4(-4-hydroxyphenyl)-2-amino-butane, and 1,1-diphenyl-2-amino-propane as well as 17-amino-estrone were synthesized in up to 50-g scale (Fig. 4). In all the cases, similar yields to those of 50- to 100-mg–scale reactions were obtained.

Preparation of existing drug molecules

Last, we undertook preparation of 10 existing drug molecules (Fig. 5D). These syntheses showcase the applicability of the supported Co nanoparticles for selective preparation of amine-based drugs. Previously, these molecules were prepared via the reductive amination reactions by using either precious metal–based catalysts or sodium borohydride (5559). In addition, some have also been prepared through nucleophilic substitution reactions of corresponding amines with halogenated compounds (5560). Here, the corresponding aromatic aldehydes were treated with piperazine and morpholine derivatives or primary amines to give the desired products in 82 to 92% yields according to our standard procedure, which further demonstrates the general utility of this single Co–based catalyst.

Supplementary Materials

Materials and Methods

Figs. S1 to S14

NMR and HRMS Spectra

References and Notes

  1. Acknowledgments: We gratefully acknowledge the support of the European Research Council (ERC), Federal Ministry of Education and Research (BMBF), the State of Mecklenburg-Vorpommern, and King Abdulaziz City for Science and Technology (KACST). We thank the analytical staff of the Leibniz-Institute for Catalysis, Rostock, for their excellent service. All data are available in the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article