Highly water-dispersible magnetite-supported Pd nanoparticles and single atoms as excellent catalysts for Suzuki and hydrogenation reactions †

The molecule 4-(diphenylphosphino)benzoic acid (dpa) anchored on the surface of magnetite nanoparticles permits the easy capture of palladium ions that are deposited on the surface of the magnetite nanoparticles after reduction with NaBH 4 . Unexpectedly, a signi ﬁ cant fraction of dpa is removed in this process. Samples of Fe 3 O 4 dpa@Pd x containing di ﬀ erent Pd loadings ( x ¼ 0.1, 0.3, 0.5 and 1.0 wt%) were prepared, and their catalytic e ﬃ ciency for the Suzuki C – C coupling reaction was studied. The best catalyst was Fe 3 O 4 dpa@Pd 0.5 , which gave the highest TOF published to date for the reaction of bromobenzene with phenylboronic acid in a mixture of ethanol/water (1/1). Interestingly, the same reaction carried out in water also produced excellent yields of the resulting C – C coupling product. The behaviour of other bromide aryl molecules was also investigated. The best catalytic results for the aqueous phase reduction of 4-nitrophenol (4-NP) to 4-aminophenol (4-AP) were obtained using Fe 3 O 4 dpa@Pd 0.1 . The presence of Pd SACs (single atom catalysts) seems to be responsible for this performance. In contrast, the same Fe 3 O 4 dpa@Pd 0.1 catalyst is absolutely inactive for the hydrogenation of styrene in ethanol.


Introduction
For the last few decades, Pd nanoparticles (NPs) have attracted enormous attention as transition metal catalysts in an array of transformation processes in organic chemistry, mainly for C-C cross-coupling reactions. 1Given that easy aggregation of nanoparticles can affect their catalytic behaviour, it is crucial to protect them with appropriate linkers, namely, thiol-protected linkers, 2,3 dendrimers, 4 or dendrons, 5 or by the use of ionic liquids. 6Another strategy involves the immobilization of the Pd NPs on solid supports, usually SiO 2 , TiO 2 and C, prepared by different procedures and having different areas and morphologies. 7This interest has increased in recent years with the easy deposition of Pd onto the surface of magnetite NPs, resulting in magnetically separable catalytic systems that avoid the requirement of catalyst ltration aer completion of the reaction.In addition, the nanocatalyst can be efficiently recycled and reused in the vast majority of examples. 8For anchoring small and homogeneous nanoparticles, it is worthwhile to previously functionalize the magnetite surface with linkers, generally terminated with amino 9,10 or phosphine units. 11owever, the benets produced by the linkers on the deposition of the metal catalysts should be contrasted with the potential catalytic inhibition produced by their steric hindrance, which obstructs the approach of the reagents to the catalyst.In this context, it is evident that the possibility of partial removal of the linker aer metal immobilization becomes a desirable condition.
Very recently, we have reported the synthesis and use of 4mercaptophenyldiphenylphosphine (Sdp) as a linker to immobilize Pd nanoparticles onto the surface of magnetite nanoparticles. 12Subsequent treatment of the nanoparticles with an aqueous hydrogen peroxide solution permitted the removal of 70 wt% of Sdp.The partial loss of the linker made the resulting nanoparticles catalytically more efficient for the Suzuki-Miyaura C-C cross coupling reaction and for hydrogenation of 4-nitrophenol and styrene.
In this paper, we expand this study to the formation of Pd/ magnetite nanoparticles and single atoms using partially water soluble and commercially available 4-(diphenylphosphino)benzoic acid (Fig. 1) as a linker in order to make the above processes "greener" by minimizing the use of (toxic) organic solvents.In addition to this, the increased demand for scarce metals such as palladium and their future availability 13 is a matter of concern that may be dispelled if control of the palladium particle size is achieved for proper atomic economy.

Synthesis of the catalysts
The starting bare magnetite nanoparticles were synthesized following a literature procedure. 14Functionalized magnetite nanoparticles, Fe 3 O 4 dpa, were obtained by sonication of Fe 3 O 4 in methanol in the presence of dpa for 2 hours (Scheme 1a).Elemental analyses (C, H) of the resulting nanoparticles showed that the load of the organic component was about 3.2 wt%.The next step involved immobilizing Pd onto the magnetite surface (Scheme 1b).Thus, Fe 3 O 4 dpa was dispersed in water and sonicated for 10 min.Then, a water solution of K 2 [PdCl 4 ] followed by another of sodium borohydride were added to obtain the Fe 3 O 4 dpa@Pd x nanoparticles, where x is the weight percentage of Pd in the NPs.With this method, we prepared several samples with different Pd contents (0.1%, 0.3%, 0.5% and 1.0%).The rst one was included in order to obtain magnetite NPs decorated exclusively with isolated palladium single atoms.These species are denoted as SACs (single-atom catalysts) 15 and exhibit generally exceptional catalytic activity.
A crucial and surprising observation in the course of the fabrication of Fe 3 O 4 dpa@Pd x NPs was the signicant fraction of dpa lost during the reduction process with NaBH 4 .For example, Fe 3 O 4 dpa@Pd 0.5 only retained 1.7% dpa from 3.2% of the starting Fe 3 O 4 dpa NPs.That is, dpa favours the dispersion of Pd NPs on the magnetite surface, and then the ligand undergoes partial elimination without the use of other agents.
Consequently, the low linker content in Fe 3 O 4 dpa@Pd x NPs is expected to yield excellent catalytic activity, which we have explored in the Suzuki-Miyaura reaction, the reduction of 4nitrophenol and the styrene hydrogenation processes.

Characterization of Fe 3 O 4 dpa@Pd x NPs
The FT-IR spectra of the catalysts are practically identical, showing a sharp peak at 594 cm À1 , which is the IR signature for ferrite particles, and weak signals at 1100 cm À1 and 2921 cm À1 due to aryl C-H vibrations.The Pd content in Fe 3 O 4 dpa@Pd x was determined by ICPoes (Inductive Coupled Plasma optical emission spectroscopy) measurements, and the dpa content was estimated by OEA (Organic Elemental Analysis).The analytical composition of the hybrids, as well as of the NPs loaded with Pd in the absence of linker, is shown in Table 1.
Images of the hybrid Fe 3 O 4 dpa@Pd x nanoparticles were obtained by TEM (for x ¼ 1.0) and HAADF-STEM (High-Angle Annular Dark-Field Scanning Transmission Electron Microscopy) (for x ¼ 0.1, 0.3, and 0.5).The samples display spherical Fe 3 O 4 nanoparticles with an average particle size of around 12 nm.In addition, smaller Pd species appear to be immobilized on the magnetite surface.As expected, the total Pd load determines the Pd nanoparticle size.Thus, for x ¼ 1.0, the images clearly reveal the presence of nanoparticles with an average size of 4.5 nm (Fig. 2).In contrast, when the Pd loading is 0.1 wt%, all Pd species exist exclusively as isolated single atoms; neither subnanometre clusters nor nanoparticles were detected (Fig. 3).Samples containing intermediate palladium contents, x ¼ 0.3 and 0.5, evidenced numerous single metal atoms accompanied by other subnanometer clusters and larger nanoparticles.This   fact is somewhat surprising because the presence of large nanoparticles of $5 nm along with a signicant number of Pd SACs is unprecedented (Fig. 4).The presence of palladium was conrmed by energy-dispersive X-ray spectroscopy (EDS) measurements; sharp Pd La1 and Lb1 peaks are clearly visible at 2.84 and 2.99 keV, respectively (Fig. 4c and f).X-ray Photoelectron Spectroscopy (XPS) was used to determine the surface elemental composition of the Fe 3 O 4 dpa@Pd x nanocomposites (Fig. S1 †).The surface Pd/Fe atomic ratios are compiled in Table 2.For the Fe 3 O 4 dpa@Pd x catalysts with x ¼ 0.1, 0.3 and 0.5, the Pd/Fe ratio progressively increases from Pd/ Fe ¼ 0.003 to 0.010, as expected, since all the Pd single atoms and subnanometer clusters observed by HAADF-STEM are sampled.However, for the Fe 3 O 4 dpa@Pd 1.0 sample, the Pd/Fe ratio is lower because not all the Pd in the Pd nanoparticles is sampled due to the larger dimensions of the Pd nanoparticles, in total accordance with the high-resolution TEM results discussed above.The oxidation state of Pd in each catalyst is also included in Table 2, where the relative contributions from Pd(0), Pd(II) and Pd(IV) species and their binding energies (Pd 3d 5/2 ) are indicated.There is a clear trend between the oxidation state of Pd and its content in the catalysts; the higher the Pd loading, the more oxidized the Pd.The sample with the highest Pd content, Fe 3 O 4 dpa@Pd 1.0 , exhibits photoemission peaks ascribed to Pd(IV), which conrms the tendency of Pd oxidation in this series.On the other hand, the sample Fe 3 O 4 dpa@Pd 0.1 , which only contains Pd single atoms, exhibits a Pd(0) component at a signicantly higher binding energy value (335.8 vs. 335.2 to 335.3 eV), which suggests that there is an electronic transfer from the Pd single atoms to the Fe 3 O 4 support, in accordance with previous reports. 16,17talytic studies Suzuki-Miyaura reaction.Palladium-catalyzed Suzuki-Miyaura cross-coupling is a well-known process for the synthesis of biaryls using aryl halides with arylboronic   (Scheme 2) acids; it has many applications in agrochemicals, natural products and pharmaceutical intermediates. 18To test the catalytic activity of Fe 3 O 4 dpa@Pd x NPs, the reaction between bromobenzene and phenylboronic acid (1 : 1) was analysed and optimized in terms of time, temperature (Table S1 †), base (Table S2 †) and solvent (Table S3 †).The best results were obtained with Fe 3 O 4 dpa@Pd 0.5 NPs aer 20 min at 65 C in EtOH/water (1 : 1) and K 3 PO 4 .Table 3 lists the catalytic results.Note that Fe 3 O 4 dpa NPs in the absence of Pd (entry 1) showed catalytic inactivity, while entry 2 revealed the poor catalytic behaviour of the NPs loaded with Pd in the absence of linker.
Entries 3 to 6 display the results obtained with the remaining Pd-containing NPs.Clearly, the best catalyst is the sample containing 0.5% Pd (entry 5).The catalyst consisting only of single atoms (Fe 3 O 4 dpa@Pd 0.1 ) (entry 3) and Fe 3 O 4 dpa@Pd 0.3 (entry 4) are also excellent, although the sample loaded with 1.0 Pd% is clearly less efficient (entry 6) due the large nanoparticle size.
We next proceeded to examine the scope and limitation of the catalysts.Using Fe 3 O 4 dpa@Pd 0.5 NPs, several bromo-aryl substrates were reacted with phenylboronic acid (1 : 1), and the coupling products were obtained in excellent yields (Table 4).Note that the Pd used is as low as 9.4 Â 10 À5 eq.To our knowledge, the TOFs exhibited with Fe 3 O 4 dpa@Pd 0.5 NPs for the Suzuki-Miyaura C-C coupling are by far the highest reported to date.
The lifetime of catalytic systems and their reusability is important for practical applications.The reusability of the catalyst was examined following the reaction of bromobenzene     and phenylboronic acid.In detail, the NPs were collected aer completion of the rst reaction using an external magnet and were washed with ethanol and dried.Then, the recovered catalyst nanoparticles were used for the next round by mixing them with a new substrate, base, and solvent.The third round was carried out following the same procedure, and the results are displayed in Table 5.As can be seen, the loss of the catalytic efficiency is remarkably high and greater than that detected, for example, by employing Sdp linker. 12The strong decrease of the catalytic activity is concomitant with the reduction of Pd concentration (33 wt%) found in the residual NPs.The measured leaching of Pd suggests that the catalytic process is essentially homogeneous, which was subsequently conrmed by a hot ltration test.Thus, bromobenzene and phenylboronic acid were reacted at 65 C. Aer 10 min, the solid catalyst was separated magnetically and the ltrate was transferred to another Schlenk ask.Without the presence of catalyst, the reaction progressed to 95% conversion aer 30 min.This result is in good accord with recently reported data arguing that palladium species leached out from nanoparticles into the reaction mixture are indeed the actual catalytic species instead of the original Pd NPs. 19In addition, the HAADF-STEM images of the catalyst aer 3 cycles showed the presence of only palladium single atoms (Fig. 5), strengthening the idea that strong bonding between Pd single atoms and naked Fe 3 O 4 takes place. 16This behavior explains the retained activity of the catalyst aer several cycles.The Suzuki reaction is difficult to achieve with chloroaryl substrates.In this paper, only traces of the nal product were observed by reaction of chlorobenzene and 4-chlorotoluene with phenylboronic acid in the presence of Fe 3 O 4 dpa@Pd 0.5 .
Catalysis in neat water.There are very few articles on the application of magnetically separable Pd-catalyzed Suzuki-Miyaura reaction in neat water; in some of the examples, the process required the use of large amounts of ionic liquids or phase transfer reagents. 20That is, there is still much room for developing new magnetically recoverable catalysts that can be used in pure water.Here, we report the catalytic behaviour of Fe 3 O 4 dpa@Pd 0.5 NPs in water for the Suzuki coupling under air atmosphere.Although the boronic acids are soluble in water, the efficiency of the Suzuki reaction in neat water can be affected by side reactions, mainly, homocoupling reactions of the boronic acids.The reaction of bromobenzene and phenylboronic acid was used as a model reaction and was optimized in terms of time and temperature.We next examined the scope of this process using several substrates (Table 6).It is well known that the coupling reactions of aryl bromides containing   electron-donating groups proceeds more efficiently than that of those having electron-withdrawing groups.This trend is not observed in our results, because the solubility of the substrates is the predominant factor in water.The formation of homocoupling products was negligible.The overall results shown in Table 7 are exceptional in terms of TOFs and selectivity, surpassing all those reported earlier; however, the reusability is very limited.For example, using bromobenzaldehyde as a reagent, it was found that aer the third cycle the TOF decreased from 5030 to 106 h À1 .Palladium leaching was clearly evidenced by a hot ltration test in which the important loss (28 wt%) of Pd content (by ICPoes analysis) in residual NPs conrmed the homogeneous nature of the catalytic process.Hydrogenation of 4-nitrophenol.Nitrophenol is a by-product produced from pesticides and synthetic dyes that causes damage to the human central nervous system; thus, its removal from the environment is a crucial task. 33Recently, FeO x -supported platinum SAC 34 and carbon nitride-palladium SAC 35 have been reported to show excellent activity for the hydrogenation of nitroarenes; however, to our knowledge, Pd SACs supported on magnetite NPs have not been explored for the reduction of 4-nitrophenol in water.The aqueous phase reduction reaction of 4-nitrophenol (4-NP) to 4-aminophenol (4-AP) (Scheme 3) using metal-supported NPs is a model reaction that permits a comparison of the efficiency of the employed catalysts.Esumi et al. have proposed that the catalytic reduction of 4-NP proceeds in two steps: (i) diffusion and adsorption of 4-NP to the catalyst surface and (ii) electron transfer mediated by the catalyst surface from BH 4 À to 4-NP. 36The rapid reduction generally is carried out using NaBH 4 and monitored by timeresolved UV-visible spectra.We have tested this reaction with the Fe 3 O 4 dpa@Pd x NPs catalysts, and the results are listed in Table 8.The measured TOFs are also very high, including that obtained with a sample with Pd deposited on its magnetite surface without the assistance of dpa (entry 1).In all cases, the reaction takes place with rapid release of H 2 bubbles accompanied by a decrease of the absorption at 400 nm and the appearance of another at 317 nm, conrming the transformation of 4-nitrophenol to 4-aminophenol (Fig. 6a).
The UV-vis spectra showed an isosbestic point (313 nm), indicating that no by-product is formed during the reduction of 4-NP.Since the concentration of NaBH 4 exceeds that of 4-nitrophenol (c(NaBH 4 )/c(nitrophenol) ¼ 100 : 1), the reduction can be considered as a pseudo-rst-order reaction with regard to 4nitrophenol only.The plot matches rst-order reaction kinetics, and the rate constant is calculated from the equation ln(c t /c 0 ) ¼ kt.Fig. 6b shows the linear relationship between ln(c t /c 0 ) and the reaction time for Fe 3 O 4 dpa@Pd 0.3 .For a quantitative comparison, we have introduced the activity parameter k 0 ¼ k(s À1 )/m Pd , where m Pd is the total mass of the palladium added as catalyst.From the catalytic results (Table 8), it can be seen that the best sample is Fe 3 O 4 dpa@Pd 0.1 , which contains only 0.10% palladium.As commented above, this catalyst consists of Pd SACs, whose enormous catalytic activity has been recently recognized in different reactions.
Table 9 permits us to compare the efficiency of numerous catalytic species, and it can be seen that the efficiency of the Fe 3 O 4 dpa@Pd 0.1 NPs surpasses most of them.The reusability of the catalysts has been examined.The catalytic efficiency was maintained during the rst six cycles for Fe 3 O 4 dpa@Pd 0.1 (Fig. 7).Then, it began to decrease.ICPoes analysis indicated that the Pd content of the resulting NPs aer six cycles  decreased by only 6%.This fact, along with the constant activity shown for the catalyst through the cycles, seems to indicate that the reduction of nitrophenol occurs heterogeneously.Hydrogenation of styrene.Very recently, we have described the catalytic activity of Fe 3 O 4 dopPPh 2 @Pd x NPs supported on magnetite nanoparticles using styrene hydrogenation (Scheme 4) as a model catalytic reaction in isopropanol as solvent at room temperature in a hydrogen atmosphere. 17Surprisingly, we observed that a sample containing a very low Pd weight content (0.18 wt%) and constituted uniquely of Pd SACs was absolutely inactive for the hydrogenation of styrene and other alkenes.This is the rst time that Pd single atom catalysts (SACs) were shown to be catalytically inactive.In contrast, samples with higher Pd content, exhibiting small clusters and nanoparticles, gave exceptional catalytic results.These ndings prompted us to analyse the catalytic activity of Fe 3 O 4 dpa@Pd x NPs for the hydrogenation of styrene in order to check whether our catalysts are able to mimic the catalytic behaviour shown by the Fe 3 O 4 dopPPh 2 @Pd x NPs.The results are compiled in Table 10.The most remarkable result was the null activity shown by the Fe 3 O 4 dpa@Pd 0.1 NPs (entry 2).This result conrms the inactivity of Pd SACs for the hydrogenation of styrene, and it is in contrast to the excellent catalytic efficiency reported in a number of processes. 44Signicantly, the remaining samples followed the expected catalytic trend, that is, excellent performances that deteriorated with increasing Pd content and Pd nanoparticle size.
Reusability studies have been carried out on the best catalytic sample, Fe 3 O 4 dpa@Pd 0.5 .The loss of activity is very small (Table 11); this is conrmed by the nal Pd content aer 3 rounds, which is less than 3% in relation to the starting sample.These results strongly support heterogeneous catalytic behaviour for this process.Table 12 lists selected catalysts for the hydrogenation of styrene, including reaction conditions and TOFs; the Pd catalysts reported here are among the best examples.

Conclusions
In conclusion, in this work we demonstrate that the use of the dpa molecule as a linker for the deposition of palladium species on the surface of magnetite permitted partial solubilization of the resulting hybrid in water.This fact, along with the partial removal of dpa during the reduction process, facilitates the     approach of the substrates to the catalyst.In addition to this, the small size of the Pd NPs supported on magnetite in Fe 3 O 4dba@Pd 0.3 and Fe 3 O 4 dba@Pd 0.5 , along with the presence of Pd SACs, explains the exceptional results for the Suzuki C-C coupling reaction and hydrogenation of 4-nitrophenol both in neat water and in a mixture of ethanol/water.

Suzuki-Miyaura coupling
In a typical Suzuki-Miyaura reaction, 3.6 mmol of phenylboronic acid, 9 mmol of base, and a quantity of Pd catalyst (e.g.2.0 mg of Fe 3 O 4 Sdp@Pd 0.5 , 9.4 Â 10 À5 mmol) were weighed into a Schlenk tube.The mixture was purged with nitrogen and 60 mL of solvent was added.Aer that, the tube was brought to a preheated plate at 65 C, and 3 mmol (1 eq.) of substrate was added.When the desired reaction time was reached, the mixture was allowed to cool.Extraction with ethyl acetate was performed.The organic phase was removed, dried over sodium sulphate and analyzed by GC.The identity of the products was conrmed by GC-MS analysis.
Catalytic reduction of p-nitrophenol 30 mL of p-nitrophenol (7.4 mM) and 30 mL of NaBH 4 (0.40 M) were added to a quartz cuvette containing 2 mL of water.Then, 30 mL of an aqueous suspension containing the catalyst nanocomposite (approximately 3 to 5 mg in 5 mL of water) was injected into the cuvette to start the reaction.The intensity of the absorption peak at 400 nm in the UV-vis spectra was used to monitor the process of the conversion of p-nitrophenol to paminophenol.Aer each cycle of reaction, another 30 mL of pnitrophenol and 30 mL of NaBH 4 were added to the reaction to study the reuse of the catalyst.The catalytic reduction reactions were conducted at room temperature.

Hydrogenation of styrene
Hydrogenation reactions were carried out in a hydrogen atmosphere at room temperature.Typically, under nitrogen, ferrite catalyst (5.0 mg approximately) was dispersed in freshly distilled ethanol (20 mL), and then styrene (2.5 mmol) was added by pouring.Aer that, the mixture was transferred into a Fisher-Porter reactor, lled with hydrogen (3 bar), and stirred at constant speed.The conversion was determined by GC analysis.

Fig. 3
Fig. 3 (a) HAADF-STEM images of Fe 3 O 4 dpa@Pd 0.1 .Isolated Pd atoms (white circles) are uniformly dispersed on the magnetite support.(b) EDS spectrum.The Cu and C signals originate from the TEM grid.

Fig. 4
Fig. 4 (a and b) HAADF-STEM images of Fe 3 O 4 dpa@Pd 0.3 showing the presence of different Pd species.(c) EDS spectrum of Fe 3 O 4 dpa@Pd 0.3 .(d and e) HAADF-STEM images of Fe 3 O 4 dpa@Pd 0.5 showing the presence of different Pd species.(f) EDS spectrum of Fe 3 O 4 dpa@Pd 0.5 .The circles and squares in the images indicate Pd single atoms and clusters/nanoparticles, respectively.

Table 1
Composition of magnetite-dpa-Pd hybrids

Table 2
Pd/Fe surface atomic ratios, percentage of Pd species and binding energies obtained by XPS

Table 3
Activity of different catalysts in Suzuki-Miyaura coupling a

Table 6
Suzuki-Miyaura coupling with different substrates using Fe 3 O 4 dpa@Pd 0.5 as the catalyst a

Table 12
Comparison of styrene hydrogenation catalyzed by Pd NPs from the literature

Table 10
Activity of the catalyst in styrene hydrogenation a