All operations were carried out under nitrogen atmosphere using standard Schlenk techniques. Solvents were distilled from the appropriate drying agents and degassed before use.

Complexes 1 - 3 were prepared according to literature protocols [27] [38] .

[RuCl₂(p-cymene)]₂, RuCl(p-cymene)[(R,R)-TsDPEN] , ethanol amine and (1R,2R)-(−)-N-p-Tosyl-1,2-diphenylethylenediamine (TsDPEN) were purchased from Aldrich.

For TEM analysis 3.5 microliter of 0.25 mM of the catalytic solution (with or without addition of 0.01 mmol of KOH) was deposited and dried onto a carbon coated 400 mesh copper TEM grid. The samples were analyzed with 200 kV FEG scanning-transmission electron microscope (STEM) Tecnai F20 G2 (FEI Company, USA). Imaging of the samples was done in bright field (BF) TEM mode and STEM mode with high angular annular dark-field STEM detector (HAADF). Elemental composition was analyzed with energy dispersion X-Ray spectroscopy (EDS) in STEM mode.

Transmission Electron Microscopy (TEM): Performed with the high resolution TEM Tecnai F20 G2. The microscope combines ultra-high resolution performance (point resolution 0.24 nm, line resolution 0.1 nm, limit of information 0.15 nm, HR STEM resolution < 0.2 nm) with extended analytical abilities and equipped with energy dispersive X-ray spectroscopy EDS detector.

For UV-visible and CD analyses, a sample of 3ml of the solution was poured into glass cuvette and directly analysed.

UV-Vis spectra were measured with a UV-Vis (Varian EL-03097225) spectrophotometer using isopropanol as the reference.

Circular dichroism (CD) spectra were recorded on a JASCO J-810 Spectrophotometer (JASCO, Japan) using the supplied Spectra-Manager software.

For H-NMR and diffusion NMR analysis, after the solution was evaporated and dried under vaacum at room temperature, a small sample is dissolved in CDCl3 in order to fill the NMR tube by 4 - 5 cm.

H-NMR spectra were recorded on a 500 MHz Burker Avance II instrument. Diffusion rates are measured by COSY-NMR.

Degassed propan-2-ol (40 ml) was added to a mixture of [RuCl₂(p-cymene)]₂ (6.124 mg, 0.01 mmol) and (1R, 2R)-(−)-N-p-Tosyl-1, 2-diphenylethylenediamine (14.66 mg, 0.04 mmol). The mixture was heated to 82˚C for 30 min under nitrogen (pretreatment). After cooling to 30˚C, acetophenone (0.48 g, 4.0 mmol) was added followed by of 1 ml of 0.1 M KOH solution in 2-propanol (0.01 mmol). The mixture was maintained at 30˚C for 24 hours. Samples were taken from the mixture at certain intervals and the conversion was determined by GC analyses.

Samples taken from the reaction mixtures were analyzed by GC and HPLC.

For GC we used a Thermo Focus Gas chromatograph (Thermo scientific) equipped with a 95%-dimethyl-po- lysiloxane-5%-diphenyl packed column (GC Cap. Column 30 m × 0.25 mm × 0.25 µm) (Zebron™ ZB-5) and an FID detector. Helium was used as a carrier gas at a pressure of 50 kPa. Retention time of acetophenone: 5.30 and of 1-phenylethanol: 5.19 minutes.

Scheme 1 was tested under the original protocol defined in the prior art [8] [9] . The kinetic profile of the reaction is shown in Figure S1 in appendix. The catalyst performed exactly as reported (TOF of 69 min⁻¹ after 30 min). The reaction stopped after 15 hours at 92% conversion. Using a chiral HPLC column [39] we verified the high ee (97%) of the product (R) 1-phenylethanol. However, upon addition of a batch of fresh substrate to the mixture after 15 hours, no further reaction was observed. Catalytic activity could be resumed only after the addition of a new batch of a catalyst (4 + TsDPEN + KOH). The lifetime of the catalyst in this system is thus limited to 12 - 15 hours at ambient temperature (Figure S1 in Appendix).

Remarkably, when elemental mercury (100:1 molar ratio Hg:Ru) was added to the above reaction mixture the reaction rate dropped by 52% to TOF of 33 min⁻¹. This is an indication that the active catalyst in this reaction is possibly not a homogeneous metal complex.

The color changes observed at different stages of this run are shown in Figure 2(b). We noted that the reaction solution remained clear without any precipitation formed even after 15 hours although it turned into dark black accompanied with a total loss of activity. The color changes clearly suggest that the catalyst is not in a steady state through the process and its nature is varying with time.

As originally reported, in the absence of the TsDPEN ligand, precursor 4 exhibited a very poor catalytic activity. Thus, with 4 used as a sole catalyst, after the standard pretreatment and addition of KOH and substrate, conversion of barely 1% was measured in Scheme 1 after 1 hour and 8% after 15 hours (at 30˚C). Suspension of black particles appeared after 2.5 hours and precipitation was clearly observed in the mixture after 5 hours (Figure 2(a)). The catalyst was totally inactive after the first batch. Interestingly, the performance of 4 as a single catalyst in transfer hydrogenation was remarkably improved at reflux temperature (82˚C). Nonetheless the addition of TsDPEN under these conditions had no effect on the catalysis and no enantioselectivity [39] was observed. We also noticed that the lifetime of catalyst 4 in Scheme 1 at 82˚C is the same (approximately 12 hours) in the presence or in the absence of TsDPEN. As described above this is also the lifetime of the advocated catalyst at ambient temperature.

The “metal-ligand bifunctional” and the “outsphere” mechanisms proposed by Noyori were allegedly authenticated by physical isolation and characterization of the intermediate complexes 1 - 3 and by the stoichiometric reversible reaction of 2 with acetone [27] Nonetheless, none of these proclaimed intermediates was actually prepared under the genuine CTH or ATH reaction conditions, namely in the simultaneous presence of the alcohol donor (2-propanol) and the base (KOH).

We have replicated the original three steps synthetic protocol and recorded the UV-visible spectra of the three species 1 - 3. To a solution of 4 in methylene chloride (Figure 3(a)) 1 equivalent of TsDPEN and of solid KOH were added at ambient temperature to generate the orange solution of 1 (Figure 3(b)). Then, 1 in methylene chloride solution was converted to the deep purple complex 3 by contacting it at room temperature with an aqueous KOH solution (Figure 3(c)). Finally the methylene chloride solution of 3 was washed with water to

remove the KOH, the solvent was evaporated under vacuum, and upon addition of 2-propanol in the absence of a base, 3 was transformed to dark brown ruthenium hydride complex 2 (Figure 3(d)).

The UV-visible spectra of complexes 1 - 3 and the appearance of their solutions are shown in Figure 3 and in Figure 4 respectively. The original complex 4 in CH2Cl2 displays three absorbance peaks at 271, 346 and 459 nm (Figure 3(a), Figure 4 red line). After addition of TsDPEN and KOH in methylene chloride the peak at 271 nm disappeared, the peak at 459 nm moved to 452 nm and the peak at 346 nm did not change (Figure 3(b), Figure 4 green line).

Washing the above solution with aqueous KOH generated the purple complex 3 (Figure 3(c)) that displayed two new absorption peaks at 305 and 550 nm (Figure 4 blue line). After evaporation of the methylene chloride and addition of 2-propanol, the brown complex 2 (Figure 3(d)) was formed with absorption peaks at 356 and 459 nm (Figure 4 brown line). After exposing the as-prepared complex 2 (in 2-propanol) to the authentic conditions of the ATH/CTH reaction, by adding two equivalents of KOH to the solution, an instant color change from brown to black was observed (Figure 3(e)). The UV-Visible spectrum of the resulting solution showed a flat absorption line with no peaks (Figure 4 purple line).

In view of these observations we can safely argue that none of the intermediate complexes 1 - 3 is present in the actual ATH/CTH reaction mixture and evidently no other soluble catalytic intermediate is present in the system. We may conclude that the molecular mechanism proposed by Noyori and coworkers cannot be utterly accurate.

UV-visible absorption spectrophotometry is considered as a practical method for monitoring the evolution of metallic species in the course of the preparation of colloidal metal nanoclusters [40] -[43] . In view of the Mie theory [44] that allows the reckoning of the spectrum of a ruthenium (and other metallic) colloids, we may contend that Ru(0) nanoclusters were evidently formed in the above solution upon the addition of KOH.

Complex 1 can be prepared and purified in an alternative method according to an Organic Syntheses procedure [38] . This procedure utilizes conditions that are different from the standard ATH/CTH reaction settings e.g. the base used is triethylamine and not KOH. Complex 1 prepared via this procedure behaved similarly to what is described above, showing the same flat UV-visible absorption spectra, when exposed to 2-propanol in the presence of KOH.

We have examined the UV-Visible spectra of a solution of complex 4 in 2-propanol without added ligands (heated to 82˚C for 30 minutes followed by cooling and addition of 2 equivalents of KOH at 30˚C) as a function of time.

The spectrum is shown in Figure 5(a). It is evident that the original solution of 4 in 2-propanol exhibits two peaks with λ max at 341 and 453 nm (red line). 1 minute after the addition of KOH there is a clear shift of these peaks to 392 and 540 nm respectively (blue line). Traces of the two peaks remain after 15 and 40 minutes but both completely disappear after one hour where a flat absorption curve is observed (brown line). Ru(0) nanoclusters were seemingly quantitatively formed in the above solution after one hour. This was also corroborated by TEM analysis of the sample where particles of the size 3 - 4 nm are clearly observed as shown in Figure 6(a).

Remarkably, when TsDPEN was added to the above experiment, the original absorbance peaks disappeared much faster and after merely 4 minutes, a flat absorption line was apparent. This is shown in Figure 5(b). The same observation was made when a ketonic substrate is present [45] (Figure S3 in Appendix). TEM and STEM analyses of these mixtures clearly show that no particles are observed when 4 and TsDPEN are blended in 2-propanol at 82˚C (Figure 6(b)) but upon cooling and addition of KOH, particles of the size of 1 nm are blurrily detected (Figures 6(c) and Figure 6(d)). Immediate disappearance of the UV-Vis peaks is detected when the solution is maintained 82˚C after addition of KOH. No change in the particles size is observed (see TEM Figure S4 in Appendix). When we tracked the conduct of the UV-Visible spectra of the 4 + TsDPEN + KOH mixture beyond the 4 minutes required for the total disappearance of the original peaks, we observed an interesting dynamic behavior as a function of time as shown in Figure 5(c). During the process the flat absorption line (attributed to scattering absorbance of the nanoclusters solution between 300 - 600 nm) monotonically moved upwards for a period of 2 hours and then stopped. It is known that the shape, size and surface composition of nanoclusters in solution influence the nature of the absorption spectra in the UV-Visible range [46] -[48] .

Although we cannot realize any quantitative information from this observed phenomenon, we may safely assume that nanoparticles are evidently formed in the ATH reaction mixture after 6 - 10 minutes. These particles are gradually varying in nature for the next 2 hours. As noted previously, the absence of any peak in the UV-Vis region may suggest that NPs are the only catalytic species present in solution during the reaction. These species lose their catalytic activity after approximately 10 - 15 hours. No solid precipitation and no change in the particle size accompany this loss of catalytic activity.

It is therefore evident that the TsDPEN ligand advocated for the ATH reactions actually strongly enhance the transformation of complex 4 via reduction into ruthenium nanoclusters. At the same time the clusters formed in the presence of this ligand are far more catalytically active in comparison with the nanoparticles formed from complex 4 in the absence of the TsDPEN. It seems that with TsDPEN the in-situ formed active nanoclusters are particularly stable in solution as no precipitation is formed even after prolonged time (weeks). Nonetheless, the catalytic activity in transfer hydrogenation is lost within 10 - 15 hours at 30˚C. Our attempts to isolate these soluble nanoclusters failed. Our experiments at 82˚C confirm that TsDPEN fails to stabilize the ruthenium clusters at this temperature. Swift reaction is observed but black precipitate is observed at the end of the reaction.

We also realized that the pretreatment previously proposed (boiling 4 + TsDPEN in 2-propanol for 30 minutes) prior to the addition of KOH and the substrate, is actually redundant. The same kinetic results and general behavior were observed when 4, TsDPEN, KOH, 2-propanol and acetophenone were directly mixed at ambient temperature under inert atmosphere without any pretreatment.

We measured the CD spectrum of (R,R) TsDPEN in 2-propanol. This is shown in Figure 7(a). Upon addition of one equivalent of 4 and boiling the solution at 82˚C for 30 minutes, we obtained the CD spectrum presented in Figure 7(b). The latter spectrum exhibits two cotton effect peaks at 235 and 245 nm that in general looks quite similar to the CD spectrum of the free TsDPEN ligand. However upon cooling of the above solution and addition of 5 equivalents of KOH the CD spectrum observed is totally different. This is shown in Figure 7(c) where new peaks at lower wavelength (220 - 235 nm) are apparent.

This suggests the creation of a new type of chirality. The latter can feasibly be attributed to the generation of chiral nanoparticles. Interestingly, upon heating the above solution to 82˚C the chirality is lost (see Figure S5 in Appendix). This explains the lack of enantioselectivity when the ATH reaction is carried out at 82˚C.

Diffusion ordered spectroscopy (DOSY) [49] -[51] obtained through pulsed field gradient (PFG) NMR experiments [52] [53] has become a multi-purpose tool in the chemical laboratory due to its capacity to gain insight into the behaviour of solution systems through the measurement of diffusion coefficients (D) and their correlation to molecular properties such as size, weight, shape etc. or to the nature of interaction between different molecules [54] [55] .

DOSY techniques were previously applied to study the dynamics of formation of nanoparticles, exhibiting lower diffusion rate than the starting molecules [56] [57] . Although the metallic nanoparticles themselves cannot be detected by NMR, measurements of diffusion coefficients of species adsorbed on the particle’s surface, allow

to perform the test and draw certain critical and definite conclusions concerning the formation and behaviour of metallic nanoclusters. In the present case these adsorbed species probably comprise adsorbed cymene and TsDPEN molecules or fragments thereof and possibly, metallic hydrides. The self-diffusion constant is measured in m² s⁻¹ and for metallic nanoparticles it is expected to be smaller, by several orders of magnitude, than that of typical organic or organometallic molecules.

We initially determined the self-diffusion coefficients of complex 4 and of TsDPEN in chloroform-d at 25˚C. We measured D = 8.13 × 10 − 10 m² s⁻¹ for complex 4 and 9.55 × 10 − 10 m² s⁻¹ for TsDPEN. Using the Stokes-Einstein equation we estimated the molecular diameter of complex 4 at 9.9A and of TsDPEN at 8.4 A which is in accordance with molecular models (note the viscosity of chloroform-d: 0.542 mPa s).

Then, 4 and TsDPEN (2 equivalents) mixture in 2-propanol, was heated for 30 minutes to 82˚C under nitrogen (the standard pretreatment conditions) and was analysed by diffusion ¹H NMR spectroscopy. The NMR spectrum of the mixture is shown in Figure S6 in appendix. The DOSY spectra are shown in Figure 8(a) and Figure 8(b).

It is apparent that the original peaks of the coordinated cymene in complex 4 at δ = 5.0 (for aromatic H); 2.9 (for CH(CH3)₂); 2.1 (for CH₃) and 1.2 (for CH(CH₃)₂) ppm and of the TsDPEN ligand at δ = 7.3 − 7.0 (for aromatic H); 4.0 (for CHN); 2.3 (for CH₃ in p-Ts) ppm are all recognized with a small shift to higher field indicating formation of a complex. This is also supported by a new hydride peak clearly observed at −9.0 ppm. The latter was measured to have a diffusion coefficient of 5 × 10 − 10 m² s⁻¹. Traces of fragments at δ = 4 and 1.4 ppm with diffusion coefficient of 1.5 × 10 − 9 m² s⁻¹ are also observable.

Using the Stokes-Einstein equation we estimated the van der Waals diameter of the main product to be 1.60 nm and of the smaller fragment as 5.3 A. The hydride is evidently a constituent of the larger molecule.

It is clear that a new complex is indeed formed between 4 and TsDPEN even prior to the addition of KOH. Based on the diameter measured we assume that this complex maintains the dimeric nature of 4 and that at least one of the original two chlorine bridges is not broken. This intermediate is clearly not one of the complexes 1 - 3. Examining the ¹H NMR spectra of pure complex 1 revealed that it is different in nature from the adduct formed between 4 and TsDPEN in 2-propanol at 82˚C. As noted in view of the TEM analyses no nanoclusters are formed at this stage.

A dramatic change in the NMR spectrum is observed when KOH is added to the above solution at 30˚C (after it was heated in 2-propanol to 82˚C under nitrogen for 30 minutes). This is shown in Figure 9(b). It is clear that the original characteristic peaks of the coordinated cymene in complex 4 and of the added TsDPEN ligand have almost entirely disappeared in this spectrum. Instead, new peaks at 1.7, 1.3 and 0.9 ppm appeared which can be attributed to unidentified organic fragments adsorbed on the metallic ruthenium cluster. The latter were measured to have a diffusion coefficient of 7.9 × 10 − 10 m² s⁻¹ which corresponds to a van der Waals diameter of 1.0 nm (Figure 9(a)).

We assume that these are the catalytically active ruthenium (0) nanoclusters. Since the catalysis is enantioselective we believe that the organic fragments adsorbed to the nanoparticles surface maintain some of the chirality introduced by the original TsDPEN ligand.

Remarkably we did not detect in Figure 9(b) any peaks with δ < 0 that could be assigned to an intermediate hydride species such as complex 2. We may assume that at the point, where the ATH reaction actually starts, the original TsDPEN and p-cymene ligands associated with the ruthenium core are evidently transformed into different species. We may infer that under the ATH reaction conditions, with the formation of catalytically active ruthenium nanoclusters, the p-cymene and the TsDPEN ligands disintegrate to form fragments that remain

partially adsorbed to the catalyst surface. This reduction-crumbling process, where TsDPEN has a pivotal role, is vital for the formation of the catalytically active chiral nanoclusters.

Upon heating the above solution to 82˚C total destruction of ligand fragments was detected, explaining the loss of the chirality (see Figure S7 in Appendix).

As was shown above, the added TsDPEN ligand empowers and facilitates the reduction of complex 4, in the presence of 2-propanol and KOH, into catalytic nanoclusters and stabilizes the latter against precipitation as long as the temperature is maintained below 30˚C (black suspension was observed at 82˚C after several hours). The catalytic activity is nonetheless totally lost after approximately 12 - 15 hours at 30˚C.

Aiming to determine whether one should deem TsDPEN as a ligand or as a non-stoichiometric additive, we studied the kinetic behaviour of Scheme 1 with different 4: TsDPEN ratios under the standard reaction conditions. Results are summarized in Table 1.

In view of the data presented in Table 1 it is clear that the 4: TsDPEN molar ratio advocated by Noyori et al. (1:1 or 1:2, entries 2 and 3) results in effective catalysis. Nonetheless, entry 4 shows that a ratio of 1:0.5 is not less effectual. If the active catalyst is indeed a stoichiometric complex formed between 4 and TsDPEN we would expect that the experiments in entry 5 and in entry 4 would give similar results and that experiment 6 would show much inferior performance.

This is evidently not the case and it seems that TsDPEN should conceivably be considered as a critical additive for the generation of the true active catalyst and not as a stoichiometric ligand. On the other hand we noted that with excess of TsDPEN the reaction is critically retarded (entry 1).

The significance of the base as a cocatalyst in CTH/ATH reactions is well established. It was argued that when 4 is used as a precursor, at least two equivalents of a base are required to convert complex 4 into complex 3. Nonetheless, in the cited procedures up to 1:5 Ru:KOH molar ratio was typically used. At the same time several authors claimed for CTH reactions “without base” [58] [59] . We have now confirmed that when 4 and TsDPEN are used in Scheme 1, a base is indeed essential for both the active catalyst in-situ formation and for promoting the CTH reaction. However, for the former function stoichiometric amount (relative to Ru) of base is needed while a trace dose is sufficient for the latter task.

Consequently, we noted that if at least one equivalent of KOH is provided, the critical factor is not the Ru:base molar ratio, but rather the pH of the reaction mixture. As long as the medium is acidic no CTH reaction is observed. We found that the normal pH of a solution of complex 4 + TsDPEN in 2-propanol is 3.6. Upon addition of 0.5 equivalent of KOH the pH rises to 5.4 - 5.8 and with 1.0 equivalent it increases to 6.3. Evidently this is not sufficient for the CTH reaction to initiate.

In a series of experiments we have examined the conversion in Scheme 1 with catalyst 4 + TsDPEN under standard conditions but with different Ru:KOH ratios and diverse KOH concentrations (controlled by different volumes of 2-propanol used). Results are summarized in Table 2.

Entries 1 - 5 (where no reaction is observed) clearly imply that at least one equivalent of a base is critical for the reaction to start. This is however not sufficient as long as the pH is below 7 (entry 5). In fact, even with Ru:KOH of 1:2 there is only inconsiderable reaction when the pH is 6.4 (entry 8). Once basic pH is attained the process advances smoothly. The color of the reaction mixtures is another indicator to the status of the catalyst. As long as the clear orange-brown solution is observed no reaction is apparent. At higher pHs (>8) the solution turns dark black indicating that productive reaction is taking place.

There is evidently no advantage in using larger excess of KOH. Ru:KOH ratio of 1:5 resulted in the highest conversion (95%). Additional base was not beneficial and actually lower conversion was obtained (entries 12 - 13).

The UV-visible spectra of the samples listed in Table 2 (after 1 hour reaction time) plainly demonstrate that a critical Ru:KOH ratio and a minimal pH is essential for the formation of the catalytic nanoparticles and for initiation of the CTH process. In Figure S8(a) in Appendix the spectra of the solution in entries 1, 2, 5, 9, 10 in Table 2 are shown as function of time. In Figure S8(b) in appendix the UV-Visible spectrum of entry 8 is shown as function of time. In the latter experiment the pH is below 7.0 (6.4) and the absorbance peaks are shift

ing as function of time from 458 nm to 386 nm but they do not disappear even after one hour.

In view of these findings we may suggest that the ruthenium complexes with bridging halides are resilient to reduction by 2-propanol in the absence of a base. These dimeric complexes are reduced by alcohols only in the presence of KOH or similar strong base which is consumed in the process.

Consequently one equivalent of e.g. KOH is required per equivalent of ruthenium to accomplish the reduction of Ru(II) to Ru(0) nanoparticles. Interestingly, monomeric ruthenium complexes (such as 1 - 3) do not require a base for the latter reduction step and pure 2-propanol is sufficient for this task. However, the ATH/CTH reaction that follows necessitates the presence of traces of base just to secure pH > 7. If this requirement is met, an ATH/CTH reaction “without base” will swiftly take place.

We examined the consequence of the pretreatment originally advocated for Scheme 1 (20 - 30 minutes reflux of 4 with the amino ligand in 2-propanol prior to the addition of KOH and the substrate). For that aim Scheme 1 was performed without any pretreatment. Complex 4 was mixed with the standard amounts of 2-propanol, TsDPEN, KOH and acetophenone at room temperature. The CTH process started immediately and conversion of 90% was measured after 15 hours. This is the very same result that was attained when conducting the suggested pretreatment prior to the addition of KOH and the substrate. The recommended pretreatment is evidently not necessary for the formation of the active catalytic species and for the CTH reaction to take place. This was also substantiated by spectroscopy as described below.

The UV-Visible spectra of the ternary mixture of 4, TsDPEN and KOH in 2-propanol at room temperature as function of time is shown in Figure S9 in appendix. It is clear that the original homogeneous complex decomposes and disappears in less than 15 minutes, identical to Figure 5(b).

Conversely, when 4 was mixed with KOH in 2-propanol at 30˚C for one hour prior to the addition of TsDPEN and the acetophenone substrate, a very slow CTH reaction was observed with less than 12% conversion after 8 hours and with total deactivation of the catalyst.

Consequently it is obvious that the ruthenium nanoclusters formed in the absence of TsDPEN are different in nature in comparison with the clusters shaped when the amino ligand is present. The former species are significantly less active and considerably less robust catalysts in the CTH/ATH reaction.

Surprisingly, however, when 4 in 2-propanol was mixed with TsDPEN and KOH at 30˚C for 24 hours, and only then the acetophenone substrate was added, the reaction started normally and reached 90% conversion (and catalyst deactivation) at the regular time (15 hours). This remarkable observation clearly suggests that the active nanocluster catalyst is stable in the absence of the substrate and that acetophenone evidently has a critical role in the deactivation process that follows.

In our previous study [60] on RuCl₂(PPh₃)₃ precatalyst in CTH reactions the cause for the catalyst deactivation was evidently observed by the naked eye: agglomeration and precipitation of metallic nanoclusters took place concurrently with the loss of catalytic activity. In the latter example the base was found to be the main drive for the agglomeration and deactivation of the nanocatalyst. The case of the present catalyst (4 + TsDPEN + KOH) is visibly different. No solid precipitation is observed at 30˚C and the solution remained clear and stable (though eventually catalytically inactive) for several weeks. Pretreatment of complex 4 with TsDPEN and KOH during 24 hours at 30˚C did not alter the normal behaviour of Scheme 1 after the addition of acetophenone. The TsDPEN ligand distinctly prevents the agglomeration and precipitation of the in-situ formed catalytic nanoparticles but did not avert the loss of catalytic activity after a certain time. Addition of more TsDPEN in the course of the reaction did not avert the catalyst deactivation. Reducing the amount and the concentration of the KOH used also did not extend the lifespan of the catalyst (Table 2) Attempts to enforce agglomeration by adding flocculants such as aluminium sulphate (15 equivalents relative to the catalyst) did not affect the catalysis of Scheme 1.

The reason for the deactivation of this catalytic system is probably not related to agglomeration. In fact the deactivation is most likely not caused by change in size of the catalytic nanoparticles. TEM analyses of the catalyst 4 + TsDPEN + KOH at different stages of Scheme 1 suggest that once a size of 2 - 3 nm is attained, there is no further growth in the size of the nanoclusters. This is also supported by the NMR and the UV-Visible measurements.

In order to determine whether the catalyst deactivation is a result of a product or a substrate poisoning we treated 4 + TsDPEN with KOH in 2-propanol under the standard conditions with excess of the product 1-phenylethanol for 10 hours prior to addition of the substrate. This pretreatment did not affect the normal behaviour of Scheme 1. Same was realized after pretreatment with acetone. Pretreatment with excess of substrate (acetophenone) also did not disturb the performance of Scheme 1.

At this stage we cannot point at the specific reason for the catalyst deactivation. However since the deactivation is taking place only when all the reaction components are present namely complex 4, the TsDPEN ligand, KOH, 2-propanol and the substrate (acetophenone) we presume that a certain time dependent surface modification process of the catalytic nanoclusters is taking place in presence of the above ingredients which render the catalyst inactive within 10 - 15 hours. Interestingly we have also found that this deactivation process is essentially independent of temperature. When Scheme 1 was conducted under reflux (82˚C) a much faster reaction was monitored (100% conversion after 95 minutes) but the catalyst deactivation rate remained essentially the same. Consequently 4 catalytic cycles could be carried out at 82˚C before the catalyst stopped working after 12 hours.

The hypothesis that ruthenium nanoparticle is the true catalyst in these CTH/ATH reactions calls for rationalization of an intriguing issue: what is the origin of the exceptionally high enantioselectivity typically observed in the presence of the chiral TsDPEN ligand.

Reviewing previous studies where chirally modified nanoclusters were deliberately prepared we realized that normally only moderate enantioselectivities were achieved. Ruthenium colloidal systems stabilized by chiral ligands were found to be active in the hydrogenation of arene derivatives [61] -[64] . Aplication of preprepared ruthenium NPs stabilized by chiral aminoalcohols and with oxazoline derived ligands in AH and ATH reactions were reported by Chaudert and coworkers [65] . Average ees were observed. Han et al. [66] have shown that (R)-BINAP ligands imprint chirally active sites on rhodium nanoparticles. These chirally stabilized rhodium NPs exhibited fairly high regioselectivity and chiral induction ability for the asymmetric hydroformylation of olefins (59% ee) [67] . Chirally modified Pd nanoparticles were used in asymmetric allylic alkylation again with only mediocre success [68] . Conversely, very high ees were reported by Jiang et al. who used cinchona and phosphine modified iridium catalyst for the enantioselective hydrogenation of aromatic ketones [69] .

Although we have presented only circumstantial evidences for the chirality issue (which is not the main aim of this study) we assert that the in-situ formed ruthenium nanoclusters in the presence of TsDPEN are chiral in nature and are thus accountable for inducing the asymmetric transfer hydrogenation. Chiral nanoscale materials have attracted intensive interest in recent years [70] -[76] . Several models were proposed to account for the optical activity found in chiral ligand-protected metallic nanoclusters (CLPMC) [77] [78] . One theory assumes a chiral metallic core, [79] second offers an achiral core but with chiral ligands adsorbed on the surface resulting in dissymmetric field effects [80] . A third model suggests a chiral footprint caused by a local deformation due to the ligand-cluster interactions [81] . Nonetheless, to date no plausible explanation of the physicochemical origin of this phenomenon was provided. This was attributed to the lack of precise structural information on CLPMCs and the difficulty in quantifying their chiroptical response. However, in a recent circular dichroism study on chiral gold nanoparticles it was concluded that the chirality results from mixing of ligand orbitals with those of the surface gold atoms of the nanoparticles [82] . We believe that a similar phenomenon takes place in the in-situ formation of chiral ruthenium nanoparticles stabilized with TsDPEN ligand. Concurrent with the reduction of the precursor 4, the TsDPEN ligand evidently functions simultaneously as a template, a structure directing agent and also as a stabilizing surface ligand leading to the formation of chiral ruthenium nanoclusters which are responsible for the asymmetric catalysis in these ATH reactions.