Homogeneous Reforming of Aqueous Ethylene Glycol to Glycolic Acid and Pure Hydrogen Catalyzed by Pincer-Ruthenium Complexes Capable of Metal–Ligand Cooperation

Article abstract of DOI:10.1002/chem.202005450

Glycolic acid is a useful and important α-hydroxy acid that has broad applications. Herein, the homogeneous ruthenium catalyzed reforming of aqueous ethylene glycol to generate glycolic acid as well as pure hydrogen gas, without concomitant CO₂ emission, is reported. This approach provides a clean and sustainable direction to glycolic acid and hydrogen, based on inexpensive, readily available, and renewable ethylene glycol using 0.5 mol % of catalyst. In-depth mechanistic experimental and computational studies highlight key aspects of the PNNH-ligand framework involved in this transformation.

Full text of DOI:10.1002/chem.202005450

Full Paper
doi.org/10.1002/chem.202005450
Chemistry—A European Journal
&
Homogeneous Catalysis
Homogeneous Reforming of Aqueous Ethylene Glycol to Glycolic
Acid and Pure Hydrogen Catalyzed by Pincer-Ruthenium
Complexes Capable of Metal–Ligand Cooperation
You-Quan Zou+,[a] Niklas von Wolff+,[a, b] Michael Rauch,[a] Moran Feller,[a] Quan-Quan Zhou,[a]
Aviel Anaby,[a] Yael Diskin-Posner,[c] Linda J. W. Shimon,[c] Liat Avram,[c] Yehoshoa Ben-
David,[a] and David Milstein*[a]
Abstract: Glycolic acid is a useful and important a-hydroxy
acid that has broad applications. Herein, the homogeneous
ruthenium catalyzed reforming of aqueous ethylene glycol
to generate glycolic acid as well as pure hydrogen gas, with-
out concomitant CO2 emission, is reported. This approach
provides a clean and sustainable direction to glycolic acid
and hydrogen, based on inexpensive, readily available, and
renewable ethylene glycol using 0.5 mol% of catalyst. In-
depth mechanistic experimental and computational studies
highlight key aspects of the PNNH-ligand framework in-
volved in this transformation.
Introduction
understanding of their reactivity and bond activation patterns
may help the design of new reaction schemes in the future to
access desirable molecules in an efficient manner.
The development of green and atom efficient protocols in or-
ganic synthesis is an important goal.[1] Indeed, over the last
decade, progress has been made in the design and discovery
of highly atom efficient transformations in organic redox
chemistry. In this respect, transition metal complexes that are
able to activate strong bonds via metal-ligand cooperation
(MLC) are particularly fruitful in catalysis,[2] enabling the dehy-
drogenation of readily available starting materials, such as al-
cohols and amines under mild conditions and very high atom
efficiency in the synthesis of useful amides and esters via ac-
ceptorless dehydrogenative coupling (ADC) reactions. Addi-
tionally, these complexes have been utilized to hydrogenate a
variety of oxidized substrates, such as amides, esters, carbo-
nates, carbamates and ureas to fine and useful chemicals.[3]
Given the great potential of ADC catalysts in greener synthetic
routes, it is desirable to extend the possible field of application
to key molecular targets. In addition, gaining a more thorough
Being used today in a variety of fields,[4] ranging from skin
care, adhesives and polymers to the textile industry, glycolic
acid is an essential chemical commodity with a forecasted US
market share of over 400 million USD by 2024.[5] Nevertheless,
the current protocols for its synthesis rely on high pressure
formaldehyde carbonylation using CO at high temperatures,[6]
or hydrolysis of chloroacetic acid or methyl 2-hydroxyace-
tate;[7,8] processes that are associated with the generation of
stoichiometric amounts of waste, multiple synthetic steps, and
the use of toxic reagents. The environmentally benign, direct
generation of glycolic acid from readily accessible starting ma-
terials under mild and atom efficient conditions is highly desir-
able.
Given the outstanding performance of pincer-type com-
plexes in the ADC reactions of alcohols,[3] we are interested in
understanding if and how pincer complexes might enable the
generation of glycolic acid from readily available alcohols. In
this respect, ethylene glycol[9] as a feedstock is of particular in-
terest, as it is potentially biomass-derived and already used on
an industrial scale.[10] Pioneering work by the Dumesic group
elegantly demonstrated that aqueous ethylene glycol can be
reformed to hydrogen, carbon dioxide and short alkanes at
[a] Dr. Y.-Q. Zou,+ Dr. N. von Wolff,+ Dr. M. Rauch, Dr. M. Feller, Dr. Q.-Q. Zhou,
Dr. A. Anaby, Y. Ben-David, Prof. Dr. D. Milstein
Department of Organic Chemistry
Weizmann Institute of Science, Rehovot 76100 (Israel)
E-mail: david.milstein@weizmann.ac.il
[b] Dr. N. von Wolff+
[11]
elevated temperatures using Pt/Al2O3 and Raney-NiSn[12] cat-
Present address: Laboratoire d’Electrochimie MolØculaire, CNRS
UniversitØ de Paris, 75006 Paris (France),
alysts at 2258C or 2658C (Scheme 1a). Despite the great inter-
est in the use of atom efficient reforming of ethylene glycol,
only a single heterogenous system developed by Bitter and
co-workers has been described to catalyze the transformation
of ethylene glycol to glycolic acid and hydrogen. At 150 or
1808C using carbon nanofiber (CNF)- supported copper and
nickel nanoparticles, glycolic acid could be obtained from eth-
[c] Dr. Y. Diskin-Posner, Dr. L. J. W. Shimon, Dr. L. Avram
Department of Chemical Research Support
Weizmann Institute of Science, Rehovot 76100 (Israel),
[+] These authors contributed equally to this work.
Supporting information and the ORCID identification number(s) for the
author(s) of this article can be found under:
https://doi.org/10.1002/chem.202005450.
Chem. Eur. J. 2021, 27, 4715 –4722
4715
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005450
Chemistry—A European Journal
Results and Discussion
During recent years, our group has reported several acceptor-
less dehydrogenative coupling reactions of alcohols with alco-
hols,[19] amines[20] and water[21] to the corresponding esters,
amides and acids. Very recently, we described a liquid organic
hydrogen carrier system based on ethylene glycol, which
enabled the efficient and reversible loading and discharge
of hydrogen using
a single ruthenium pincer complex
(Scheme 1b).[22] Based on this report, we envisioned that ethyl-
ene glycol would undergo dehydrogenation to glycolaldehyde
(GAL) in the presence of a ruthenium pincer complex via MLC,
accompanied by hydrogen release (Scheme 1c). Subsequently,
the interception of glycolaldehyde by water and base could
liberate a second molecule of hydrogen and enable access to
glycolic acid (GAC) after acid workup.
To examine the feasibility of the above-proposed reforming
process, we evaluated the acridine-type PNP complexes, Ru-1
and Ru-2, that gave excellent results in the dehydrogenative
coupling of ethylene glycol to oligoesters (Table 1).[22] The re-
forming of aqueous ethylene glycol was attempted using
0.5 mol% catalyst loading and 5 equivalents of potassium hy-
Scheme 1. (a) Representative examples of heterogeneously catalyzed re-
forming of aqueous ethylene glycol. (b) Ethylene glycol as a reversible hy-
drogen carrier. (c) Homogeneously catalyzed reforming of aqueous ethylene
glycol to glycolic acid and H2.
Table 1. Optimization of the reaction conditions using PNN and PNP
pincer complexes Ru-1 to Ru-7.
ylene glycol,[13] whereas a Ni/CNF catalyst led to further de-
composition to formic acid, hydrogen and carbon dioxide
(Scheme 1a).
Homogeneous reforming of ethylene glycol to generate hy-
drogen was reported by the groups of Cole-Hamilton[14] and
Beller,[15a] although the generation of glycolic acid was not con-
firmed. The group of Grützmacher reported the oxidative de-
hydrogenation of alcohols and polyalcohols, including ethylene
glycol, to carboxylic acids using a homogeneous rhodium cata-
lyst under mild conditions, albeit in the presence of sacrificial
hydrogen acceptors such as ketones or methyl methacrylate.[16]
Reforming of ethylene glycol to glycolic acid homogeneously
catalyzed by an iridium complex (4 mol%) was reported re-
cently, although the catalytic mechanism was not studied.[17]
Very recently, homogeneous iridium-catalyzed dehydrogena-
tive cross-coupling of ethylene glycol and methanol to lactic
acid was also disclosed by Tu, Xu and co-workers.[18] The scarci-
ty of reports on catalytic reforming of aqueous ethylene glycol
to glycolic acid, despite its significant promise, demonstrates
the need for new catalytic systems as well as a deeper under-
standing of the reaction mechanism. Herein, we report a fully
characterized homogeneously catalyzed reforming of aqueous
ethylene glycol to glycolic acid and exclusively hydrogen gas
at low catalyst loading (0.1 mol% on large scale). Mechanistic
studies, supported by synthesis of potential intermediates, X-
ray crystallography, nuclear magnetic resonance (NMR) experi-
ments and density functional theory (DFT) calculations are pre-
sented to highlight the key steps of this transformation.
Entry[a]
Ru
T [8C]
V (H2, mL)[b]
Yield [1, %][c]
1
2
3
4
5
6
7
8
Ru-1
Ru-1
Ru-1
Ru-2
Ru-2
Ru-2
Ru-3
Ru-4
Ru-5
Ru-6
Ru-7
115
135
150
115
135
150
115
115
115
115
115
14
22
71
<1
68
6
9
30
N.D.
28
30
75
81(78)[d]
61
71
181
195
146
113
170
9
10
11
47
71
[a] Reaction conditions: ethylene glycol (5.0 mmol), Ru cat. (0.5 mol%),
KOH (25 mmol), H2O (1.0 mL) and THF (1.0 mL) at 115 8C (bath tempera-
ture) for 48 h. [b] Hydrogen gas was collected every 24 h. [c] Yields are
based on the volume of collected hydrogen gas. [d] Yield of glycolic acid
is shown in parentheses, as determined by 1H NMR of the crude reaction
mixture after acid workup using acetonitrile as an internal standard. N.D.:
not determined.
Chem. Eur. J. 2021, 27, 4715 –4722
www.chemeurj.org
4716
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005450
Chemistry—A European Journal
droxide at 115 8C (oil bath temperature) in a 1:1 mixture of
water (1.0 mL) and THF (1.0 mL).[23] Somewhat surprisingly,
these catalysts showed only sluggish reactivity (Table 1, en-
tries 1 and 4) giving a maximum of 14 mL of hydrogen. In-
creasing the temperature to 1508C resulted in increased vol-
umes of H2 up to 71 mL (entries 3 and 6). On the other hand,
the PNN and PNP ruthenium complexes Ru-3 to Ru-7, devel-
oped by our group, gave more promising results (entries 7–11).
Complex Ru-4 performed best, delivering 195 mL of hydrogen
gas (81% yield, theoretical volume of hydrogen: 240 mL) after
48 h at 1158C (entry 8). Gas chromatographic (GC) analysis of
the gas phase indicated that the evolved hydrogen was pure
(Figure S28). NMR analysis of the crude reaction mixture indi-
cated that only potassium 2-hydroxyacetate was formed (see
Figures S12 and S13 for details). After workup with 6m HCl,
glycolic acid was formed in 78% yield (entry 8, Figures S20 and
S21).
glycolic acid and the amount of hydrogen evolved. Additional-
ly, Ru-9 and Ru-10 also catalyzed the reaction, albeit in moder-
ate yields of hydrogen and potassium 2-hydroxyacetate (en-
tries 3 and 4). Notably, doubling the water loading liberated
223 mL of pure hydrogen (Figure S30) and further increased
the yield of glycolic acid to 91% (entry 5, Figures S24 and 25).
The glycolic acid was isolated in 84% yield after removing the
water, followed by extraction with THF. Lowering the reaction
temperature to 1008C led to decreased reaction efficiency
(entry 6). Using NaOH instead of KOH provided comparable re-
sults, yielding 94% of hydrogen and 93% of glycolic acid after
workup (entry 7, Figures S26 and S27). Lower loadings of
NaOH resulted in formation of less hydrogen gas (193 mL,
entry 8). Using water as the only solvent (single phase) resulted
in a sluggish reaction (entry 9), but it proceeded well at higher
temperatures, generating 48% and 70% yields of potassium 2-
hydroxyacetate at 1358C and 1508C, respectively (entries 10
and 11).
To our delight, upon screening PNNH-based complexes
(Table 2) we found that using Ru-8 the yield of potassium 2-hy-
droxyacetate was increased to 89% and 214 mL of pure hydro-
gen (Figure S29) were collected after 48 h (Table 2, entry 2). At
the same time, glycolic acid was formed in 88% yield after
acid workup of the crude reaction mixture (Figures S22 and
S23) indicating an excellent correlation between the yield of
After establishing the optimized reaction conditions, we in-
vestigated the scalability of our methodology. As shown in
Figure 1, using only 0.1 mol% of Ru-8, 10 mmol of ethylene
glycol were efficiently dehydrogenated to form hydrogen gas
and sodium 2-hydroxyacetate in a mixture of THF (2.0 mL) and
water (4.0 mL). During the first 48 h, the reaction was relatively
fast, releasing 282 mL of hydrogen. Subsequently, the reaction
slowed down, likely a result of the decreased concentration of
ethylene glycol. After 120 h, 349 mL pure hydrogen gas (yield
of hydrogen at this stage: 73%) was collected (red curve, see
Figure S4 for details). The reforming reaction also proceeded
quite well on a 30 mmol scale, delivering 921 mL of pure hy-
drogen gas (yield of hydrogen at this stage: 64%) after 120 h
(blue curve, 3 mL THF and 6 mL water were used, see Figure S5
for details). A catalyst recycling experiment indicated that the
catalytic species was still active after 72 h, albeit at a reduced
rate (Figure S6, see Supporting Information for details).
Table 2. Production of glycolic acid and hydrogen from aqueous ethyl-
ene glycol catalyzed by PNNH complexes Ru-8, 9, 10.
Even though complexes Ru-1 and Ru-2 were shown to be
efficient in the dehydrogenative coupling of ethylene glycol to
oligoesters,[22] the PNNH complex Ru-8 showed superior per-
formance in the reforming reaction. In order to understand the
Entry[a]
Ru
MOH
V [H2, mL][b]
Yield [1, %][c]
1
2
3
4
5[e]
6[f]
7[e]
8[e,g]
9[h]
10[h,i]
11[h,j]
Ru-4
Ru-8
Ru-9
Ru-10
Ru-8
Ru-8
Ru-8
Ru-8
Ru-8
Ru-8
Ru-8
KOH
KOH
KOH
KOH
KOH
KOH
NaOH
NaOH
KOH
KOH
KOH
195
214
138
166
223
79
226
193
29
81(78)[d]
89(88)[d]
58
69
93(91)[d]
33
94(93)[d]
80
12
48
70
116
167
[a] Reaction conditions: ethylene glycol (5.0 mmol), Ru cat. (0.5 mol%),
MOH (25 mmol, M=K, Na), H2O (1.0 mL) and THF (1.0 mL) at 115 8C (bath
temperature) for 48 h. [b] Hydrogen was collected every 24 h. [c] Yields
are based on the volume of collected hydrogen gas. [d] Yield of glycolic
acid is shown in parentheses, as determined by 1H NMR of the crude re-
action mixture after acid workup using acetonitrile as an internal stan-
dard. [e] H2O (2.0 mL) was used. [f] Reaction was performed at 1008C.
[g] NaOH (15 mmol) was used. [h] H2O (3.0 mL) was used as the reactant
and solvent without THF. [i] Reaction was performed at 1358C. [j] Reaction
was performed at 1508C.
Figure 1. Reforming aqueous ethylene glycol to hydrogen and sodium 2-hy-
droxyacetate using 0.1% mol of Ru-8.
Chem. Eur. J. 2021, 27, 4715 –4722
www.chemeurj.org
4717
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005450
Chemistry—A European Journal
efficiency of the PNNH-family of complexes, we sought to in-
vestigate its possible modes of bond activation. An entry point
to a possible catalytic cycle is the deprotonation of Ru-8 by
base. Previous experimental and theoretical reports have
shown that in the presence of excess base (and absence of a
protic medium), a double deprotonated anionic species was
formed, in which the CH2-group of P-arm was still present.[24]
On the other hand, using only 1.1 equivalents of base 1H,
31P{1H}, 13C{1H} DEPTQ and 1H{13C} HSQC NMR clearly showed
that among the three acidic sites (P-arm, N-arm and NH-group)
only P-arm deprotonation took place (Figures S33–S37). This
finding is further corroborated by DFT calculations, showing
that the P-arm deprotonated species 3a is 3.6 and 6.8 kcal
molÀ1 more stable than the NH-deprotonated species 3b and
the N-arm deprotonated 3c, respectively (Scheme 2b, see the
Supporting Information for details). Furthermore, in a protic
medium (such as the crude reaction mixture), no double de-
protonated species is observed even in the presence of excess
base. Treatment of the resulting P-arm deprotonated mixture
with 1 atmosphere of H2 at room temperature in THF
Scheme 3. Dearomatization of complex Ru-4 and reaction with ethylene
glycol.
alkoxo group. A broad singlet appeared at d=103.4 ppm in
the 31P{1H} NMR spectrum (Figure S46), which is also consistent
with a saturated aromatic complex. The IR spectrum showed a
strong absorption band at 1907 cmÀ1 (Figure S47). Slow evapo-
ration of the mixture led to formation of crystals suitable for X-
ray diffraction. As shown in Figure 2, a neutral distorted octa-
hedral ruthenium complex was formed, with the alkoxide of
ethylene glycol coordinated to the metal center. In the unit
cell of the crystal, H-bonding with another molecule of ethyl-
ene glycol to the alkoxide ligand is observed (see Supporting
Information for details).
converged to the aromatized trans-dihydride complex
4
(Scheme 2a). Complex 4 was only stable under an atmosphere
of hydrogen, and was fully characterized under H2 (Figures
S38–S42, see Supporting Information for details).
Figure 2. X-ray crystal structure of complex 6. Atoms are presented as ther-
mal ellipsoids at 50% probability level. Hydrogen bonds are marked in
dashed blue lines. Hydrogen atoms are omitted for clarity except the hy-
dride H1A, H5, H3A and H4A. Selected bond lengths (Š) and angles (deg):
Ru1ÀC20, 1.834(7); Ru1ÀN1, 2.123(5); Ru1ÀN2, 2.081(5); Ru1ÀP1, 2.2738(17);
Ru1ÀH1A, 1.68(8); Ru1ÀO2, 2.243(4); H4AÀO2, 1.867; H5ÀO2, 1.787; C20-
Ru1-N2, 178.3(2); P1-Ru1-N1, 157.83(14); N2-Ru1-O2, 80.60(17).
Scheme 2. (a) Deprotonation of Ru-8 followed by reaction with H2. (b) Site
of deprotonation of Ru-8 by DFT calculations. Values are Gibbs free energies
in kcalmolÀ1 with respect to 3a in THF (values in parentheses are gas phase
values, 298 K).
Reaction of the dearomatized complex 5 with ten equiva-
lents of ethylene glycol in a 1:1 mixture of THF (0.25 mL) and
water (0.25 mL) with no added base was also performed at
115 8C in an open system, resulting in formation of the glycolic
acid adduct 7 as the major product along with minor byprod-
ucts (Scheme 4, Figures S53 and S54, see Supporting Informa-
tion for details). Complex 7 was also obtained by the addition
of two equivalents of glycolic acid to complex 5[25] in THF at
room temperature, and its structure was further confirmed by
X-ray crystallography (Figure 3). The results of the base-free ex-
periment indicate that the first cycle of the reforming reaction
to generate the glycolic acid adduct does not require base to
promote the CÀO bond formation.[23] Moreover, upon treat-
ment of sodium 2-hydroxyacetate as a substrate under the op-
timal conditions no hydrogen evolution occurred, clearly show-
ing that further dehydrogenation of sodium 2-hydroxyacetate
Since attempts to crystallize the product of the reaction be-
tween dearomatized intermediate 3 and ethylene glycol were
unsuccessful, we used the PNN complex Ru-4 as surrogate
(which exhibits good catalytic performance, see Table 1,
entry 8). Deprotonating Ru-4 (using 1.2 equivalent of tBuOK)
followed by reaction with ethylene glycol led to formation of
the aromatized complex 6 via the dearomatized species 5,[25]
1
as witnessed by NMR spectroscopy (Scheme 3). In the H NMR
spectrum (Figure S43), the hydride is shifted downfield to
2
À16.14 ppm (doublet, JPH =25 Hz), consistent with the vacant
position trans to the hydride of 5, now being occupied by the
Chem. Eur. J. 2021, 27, 4715 –4722
www.chemeurj.org
4718
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005450
Chemistry—A European Journal
Following the reaction under standard conditions by 31P{1H}
NMR showed the initial formation of a new species, assigned
to be the hydrido–hydroxo complex 8 shown in Scheme 5A
(see Supporting Information, Figure S55–S57, previously also
reported for an analogous PNN-based ruthenium complex).[26]
Importantly, this shows that even in the presence of excess
base, no double deprotonation takes place, excluding a previ-
ously proposed anionic reaction surface.[27] This is also in
agreement with the base-free reaction outlined earlier
(Scheme 4) and might be expected given the protic reaction
conditions. In an earlier study, we envisioned that a PNNH-
pincer complex might enable challenging reaction schemes by
switching between appropriate MLC modes (dearomatization
and amido-formation).[24]
Scheme 4. Dearomatization of complex Ru-4 and reaction with glycolic acid,
and reaction with ethylene glycol without base.
Employing DFT calculations, we tried to shed light into plau-
sible reaction pathways for this transformation, in particular on
three key events: 1) liberation of H2, 2) formation of the CÀO
bond, as well as 3) CÀH bond scission. DFT calculations were
performed at the wB97M-V/def2-TZVPP/RIJCOSX/SMD//M06-L/
def2TZVP/GD3/W0 level of theory (see Supporting Information
for details). In order to correctly address possible concerted-
ness of transition states, all structures were optimized in solu-
tion (THF).[28,29] The trans dihydride 4 is connecting both the H2
liberation step, as well as the CÀH bond activation step and
was thus chosen as a starting point of the mechanistic investi-
gation. Transition states not involving side-arm protons (see
Supporting Information, section 8.5) are linked to unreasonably
high barriers and were thus excluded. As observed for many
other pincer complexes, a proton shuttle is needed for the lib-
eration of H2 from dihydride 4 (Step 1, TS1A, Scheme 5B). Al-
though highly asynchronous, extended IRC-calculations con-
firm the concertedness of this step (see Supporting Informa-
tion, section 8.7). Indeed, no stable intermediate is formed
Figure 3. X-ray crystal structure of complex 7. Atoms are presented as ther-
mal ellipsoids at 50% probability level. Hydrogen atoms are omitted for
clarity except the hydride and H3. Selected bond lengths (Š) and angles
(deg): Ru(1)ÀC(22), 1.845(4); Ru(1)ÀN(1), 2.086(3); Ru(1)ÀN(2), 2.118(3);
Ru(1)ÀO(1), 2.207(3); Ru(1)ÀP(1), 2.2653(9); Ru(1)ÀH, 1.56(5); C(22)-Ru(1)-N(1),
178.50(14); N(2)-Ru(1)-P(1), 159.96(9); N(1)-Ru(1)-O(1), 77.43(11).
did not take place (see Supporting Information, section 6.6, for
details).
Scheme 5. (A) Proposed catalytic cycle for the reforming of ethylene glycol with water. Values correspond to Gibbs free energies in kcalmolÀ1 with respect to
the starting material (0.0). Level of theory: M06-L/def2TZVP/W06/GD3//wBP97M-V/def2TZVPP/RIJCOSX, structures optimized in THF (SMD). (B) Key transition
states of the proposed mechanism. Bond lengths in the computed structures in Š.
Chem. Eur. J. 2021, 27, 4715 –4722
www.chemeurj.org
4719
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005450
Chemistry—A European Journal
after initial protonation of the hydride by the proton shuttle,
with subsequent side-arm deprotonation by the conjugated
tions.[28a,33] Following the approach of Dub and Gordon, we
computed selected key intermediates and transition states in
the presence of an explicit water cluster (Scheme 6A).[28b] Ex-
pectedly, the hydrido hydroxo compound 8 is highly stabilized
by the H-bonding network (À9.9 kcalmolÀ1 with respect to the
dihydride 4, stabilization of 16.1 kcalmolÀ1). The H-bonding
has also a stabilizing effect on the TDTS TS4BEXP (17.1 kcal
molÀ1, stabilization of 11.8 kcalmolÀ1) and the H2 liberating
TS1DEXP (11.7 kcalmolÀ1, stabilization of 12.4 kcalmolÀ1). As ob-
served in DFT-MD calculations, the H-bond network will de-
crease the hydricity of the alkoxide species, explaining the
lesser degree of stabilization in TS4BEXP. Including the explicit
water cluster yields an overall energetic span dE of 27.0 kcal
molÀ1, in good agreement with the observed experimental re-
activity. Furthermore, the explicit model predicts the hydrido
hydroxo complex to be the resting state, which is confirmed
experimentally (see Supporting Information) and predicted for
Ru-4 theoretically.[32] The inclusion of explicit solvent molecules
is thus critical to correctly identify rate-limiting states on the
potential energy surface.
base being linked to the highest energy of this step.[28b]
A
more basic proton-shuttle (ethylene glycol) thus gives slightly
lower barriers (Scheme 5B, TS1A vs. TS1C). As for related PNP-
complexes, the dearomatized species will quickly react with
the present protic species, to give either the alkoxo or the hy-
droxo compounds. The reversibility of the HÀH and OÀH acti-
vation events is demonstrated experimentally by running the
reaction in D2O instead of H2O. D2 liberation was observed by
2
GC-MS, and H NMR after 12 h indicates formation of deuterat-
ed deuterio–deuterioxo complex 8 (see Supporting Informa-
tion, Figure S58). After a first dehydrogenation event (see Sup-
porting Information),[30] the hydroxo complex 8 is nucleophilic
enough to attack the aldehyde intermediate (Step 3, TS3A).
This transition state is thus somewhat reminiscent of CÀN
bond cleavage events in Fe-and Mo-catalyzed hydrogenation
of amides.[31] The subsequent final CÀH bond dissociation
(Step 4, Scheme 5B) from the geminal diol is found to be step-
wise, that is, the CÀH bond abstraction involves formally the
alkoxide and a cationic Ru-species (TS4B) to regenerate the di-
hydride 4. It is important to note that no base is needed for
the transient generation of the alkoxide in the TOF-determin-
ing (TDTS) TS4B. This is in agreement with our experimental
results (Scheme 4), where the addition of water and ethylene
glycol (but no base) to complex 5 generates the carboxylate 7.
As was already postulated in the case of Ru-4,[32] excess base is
still necessary in the catalytic system with a twofold role. On
the one hand, the overall reaction is endergonic (+5.5 kcal
molÀ1) and formation of the carboxylate salt would serve as a
driving force for the reaction. On the other hand, the forma-
tion of carboxylate 10 from 4 and glycolic acid is highly exer-
gonic (À18.7 kcalmolÀ1) and efficient quenching of glycolic
acid by excess base might help to prevent the formation of
this off-cycle product, while at the same time increasing the
concentration of the hydrido hydroxo complex 8. Several
groups highlighted the importance of explicit solvent models
for accurate DFT assessment of alcohol dehydrogenation reac-
Finally, we were interested in analyzing if the proposed
model could account (and explain) the observed difference be-
tween PNNH complex Ru-8 and PNN Ru-4 (Scheme 6B).
Indeed, the proposed mechanism reproduces qualitatively the
slightly higher activity of Ru-8, compared to Ru-4 (TS4B vs.
TS4Bbpy). Importantly, a step-wise deprotonation/hydride trans-
fer sequence can be excluded in the case of Ru-4, highlighting
the importance of branching points on the potential energy
surface.[34] Whereas in the case of the PNNH compound, depro-
tonation of the side-arm NH group is energetically accessible
(Scheme 5B, TS4A, 26.1 kcalmolÀ1), this is not the case for Ru-
4 (TS4Abpy =44.1 kcalmolÀ1, see Supporting Information). In-
stead, the CÀH bond activation via TS4Bbpy (Scheme 6B) might
be reached after CÀO bond formation (see Supporting Infor-
mation 8.4, TS3Abpy =28.0 kcalmolÀ1) via a O/H slippage, as
proposed previously.[35] Although the ligand framework in the
bipyridyl dihydride 11 is overall sterically less crowded, com-
pared to 4 (Scheme 6B, %Vfree =41.6 vs. %Vfree =36.1)[36] and
Scheme 6. (A) Selected structures of the potential energy surface in the presence of an explicit water cluster. Values correspond to Gibbs free energies in kcal
molÀ1 with respect to the starting material (0.0). Bond lengths in the computed structures in Š. (B) Comparison of Ru-4 and Ru-8. Steric topographic analysis
using SambVca 2.1[34] with the ruthenium atom as the center of the sphere (3.5), scaled bond radii (1.17) and 0.1 mesh spacing.
Chem. Eur. J. 2021, 27, 4715 –4722
www.chemeurj.org
4720
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005450
Chemistry—A European Journal
the Ru-centers have similar positive NBO-charges of +0.33, the
hydride abstraction in the case of the PNNH-ligand is around
4 kcalmolÀ1 lower in energy compared to the bipyridyl-ana-
logue (Scheme 6B, TS4B=26.9 vs. TS4Bbpy =30.9 kcalmolÀ1).
This difference might be explained by the additional stabiliza-
tion of the formal alkoxide by the NÀH moiety in 4. Indeed,
the N-H proton is not only more polarized (NBO-charge of
+0.41 vs. +0.27 in 11) but also more accessible in the parent
dihydride compound 4, as demonstrated by the steric topo-
graphic analysis. In the PNNH-ligand, the north-west (NW)
quadrant accommodating the NÀH-moiety has an accessible
free volume of %Vfree =22.9, whereas the quadrants occupied
by the ligand backbone PCH2 protons in both ligands are steri-
cally more hindered (e.g. Scheme 6B, %Vfree (NE)=10.6 and
%Vfree (SE)=14.6 in 11). Indeed, these electronic and steric anal-
yses together indicate that various factors likely govern the dif-
ference in activity of the catalysts employed. It might be
argued that the NÀH bonding is made redundant in the pres-
ence of protic reagents/solvents. We nevertheless observe that
the stabilization persists in the presence of explicit solvent
molecules, with TS4Bexp being around 4 kcalmolÀ1 lower than
the corresponding PNN-analogue (TS4Bexp =17.1 kcalmolÀ1 vs.
TS4Bexp,bpy =21.9 kcalmolÀ1, see Supporting Information 8.6).
Our mechanistic model indeed reproduces qualitatively the
trends observed in initial rate experiments run on representa-
tive candidates of the PNN and PNNH families (see Supporting
Information 8.8).
Acknowledgements
D.M. holds the Israel Matz Professorial Chair of Organic
Chemistry. Y.-Q.Z. acknowledges the Sustainability and Energy
Research Initiative (SAERI) of Weizmann Institute of Science for
a research fellowship. N.v.W. is supported by the Foreign Post-
doctoral Fellowship Program of the Israel Academy of Sciences
and Humanities. M.R. acknowledges the Zuckerman STEM
Leadership Program for a research fellowship. N.v.W and M.R.
thank Dr. Mark Iron (department of chemical research support)
for fruitful discussions regarding the DFT calculations. Compu-
tations were performed using HPC resources from GENCI-
CINES (Grant 2019AP010811227).
Conflict of interest
The authors declare no conflict of interest.
Keywords: ethylene glycol · glycolic acid · homogeneous
catalysis · hydrogen · sustainable chemistry
[1] a) B. M. Trost, Science 1991, 254, 1471–1477; b) P. A. Wender, V. A.
Verma, T. J. Paxton, T. H. Pillow, Acc. Chem. Res. 2008, 41, 40–49; c) T.
Newhouse, P. S. Baran, R. W. Hoffmann, Chem. Soc. Rev. 2009, 38, 3010–
3021.
[2] Selected reviews: a) T. Ikariya, K. Murata, R. Noyori, Org. Biomol. Chem.
2006, 4, 393–406; b) H. Grützmacher, Angew. Chem. Int. Ed. 2008, 47,
1814–1818; Angew. Chem. 2008, 120, 1838–1842; c) J. I. van der Vlugt,
J. N. H. Reek, Angew. Chem. Int. Ed. 2009, 48, 8832–8846; Angew. Chem.
2009, 121, 8990–9004; d) J. I. van der Vlugt, Eur. J. Inorg. Chem. 2012,
363–375; e) J. R. Khusnutdinova, D. Milstein, Angew. Chem. Int. Ed.
2015, 54, 12236–12273; Angew. Chem. 2015, 127, 12406–12445.
[3] Selected reviews: a) D. Milstein, Top. Catal. 2010, 53, 915–923; b) C. Gu-
nanathan, D. Milstein, Acc. Chem. Res. 2011, 44, 588–602; c) C. Gunana-
than, D. Milstein, Science 2013, 341, 1229712–1229711/11; d) R. H. Crab-
tree, Chem. Rev. 2017, 117, 9228–9246; e) G. E. Dobereiner, R. H. Crab-
tree, Chem. Rev. 2010, 110, 681–703; f) S. M. A. H. Siddiki, T. Toyao, K.-i.
Shimizu, Green Chem. 2018, 20, 2933–2952; g) R. Ferraccioli, Curr. Org.
Chem. 2018, 22, 1875–1892; h) N. Gorgas, K. Kirchner, Acc. Chem. Res.
2018, 51, 1558–1569; i) F. Kallmeier, R. Kempe, Angew. Chem. Int. Ed.
2018, 57, 46–60; Angew. Chem. 2018, 130, 48–63; j) L. Alig, M. Fritz, S.
Schneider, Chem. Rev. 2019, 119, 2681–2751; k) K. Sordakis, C. Tang, L.
Vogt, H. Junge, P. J. Dyson, M. Beller, Chem. Rev. 2018, 118, 372–433.
[4] J. Kubitschke, H. Lange, H. Strutz in Carboxylic Acids, Aliphatic in Ull-
mann’s Encyclopedia of Industrial Chemistry, Wiley-VCH, Weinheim, 2014.
[5] Glycolic Acid Market-Forecasts from 2018 to 2023. https://www.re-
searchandmarkets.com/reports/4542547/glycolic-acid-market-forecasts-
from-2018-to-2023.
Conclusions
In summary, we have developed a homogeneous, ruthenium
catalyzed, efficient reforming of aqueous ethylene glycol under
mild conditions. This reforming process produces glycolic acid
(up to 93% yield) and pure hydrogen gas, without CO2 or CO
emission. The reaction can also be performed on a large scale
of 30 mmol of ethylene glycol. Moreover, this reforming pro-
cess can also proceed well in water with no additional organic
solvent. These characteristics make this protocol environmen-
tally friendly, atom-economical, renewable, and sustainable. A
plausible reaction mechanism is proposed based on stoichio-
metric reactions, NMR studies, X-ray crystallography and DFT
calculations. The mechanistic studies allow for rationalization
of the high reactivity of the PNNH ligand framework. It is envi-
sioned that the understanding of the reaction pathways at
play might enable future development of new reaction
schemes and catalysts.
[6] S. D. Barnicki, R. T. Hembre, S. N. Falling, M. E. Janka, K. M. Moran, US
8703999 B2, 2014.
[7] L. Prager, P. Dowideit, H. Langguth, H.-P. Schuchmann, C. von Sonntag,
J. Chem. Soc. Perkin Trans. 2 2001, 1641–1647.
[8] D. Roberts, D. J. Watson, J. Swinney, WO 2018/051115 A1, 2018.
[9] H. Yue, Y. Zhao, X. Ma, J. Gong, Chem. Soc. Rev. 2012, 41, 4218–4244.
[10] A. Wang, T. Zhang, Acc. Chem. Res. 2013, 46, 1377–1386.
[11] R. D. Cortright, R. R. Davda, J. A. Dumesic, Nature 2002, 418, 964–967.
[12] G. W. Huber, J. W. Shabaker, J. A. Dumesic, Science 2003, 300, 2075–
2077.
Experimental Section
All computed structures and energies can be accessed on the
ioChem-BD.org online repository under the following link: https://
doi.org/10.19061/iochem-bd-6-66.
[13] T. van Haasterecht, T. W. van Deelen, K. P. de Jong, J. H. Bitter, Catal. Sci.
Technol. 2014, 4, 2353–2366.
Deposition numbers 1884387 and 1994638 contain the supplemen-
tary crystallographic data for this paper. These data are provided
free of charge by the joint Cambridge Crystallographic Data Centre
and Fachinformationszentrum Karlsruhe Access Structures service.
[14] D. Morton, D. J. Cole-Hamilton, J. Chem. Soc. Chem. Commun. 1988,
1154–1156.
[15] a) Y. Li, M. Nielsen, B. Li, P. H. Dixneuf, H. Junge, M. Beller, Green Chem.
2015, 17, 193–198. b) For homogeneously reforming aqueous metha-
Chem. Eur. J. 2021, 27, 4715 –4722
www.chemeurj.org
4721
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005450
Chemistry—A European Journal
nol to hydrogen and carbon dioxide, see: M. Nielsen, E. Alberico, W.
Baumann, H.-J. Drexler, H. Junge, S. Gladiali, M. Beller, Nature 2013, 495,
85–89.
[27] C. Hou, J. Jiang, Y. Li, C. Zhao, Z. Ke, ACS Catal. 2017, 7, 786–795.
[28] a) D. G. Gusev, Organometallics 2020, 39, 258–270; b) P. A. Dub, J. C.
Gordon, ACS Catal. 2017, 7, 6635–6655.
[16] a) T. Zweifel, J.-V. Naubron, H. Grützmacher, Angew. Chem. Int. Ed. 2009,
48, 559–563; Angew. Chem. 2009, 121, 567–571; b) M. Trincado, K. Kuh-
lein, H. Grützmacher, Chem. Eur. J. 2011, 17, 11905–11913.
[17] Y. Zhan, W. Hou, G. Li, Y. Shen, Y. Zhang, Y. Tang, ACS Sustainable Chem.
Eng. 2019, 7, 17559–17564.
[18] J. Wu, L. Shen, Z.-N. Chen, Q. Zheng, X. Xu, T. Tu, Angew. Chem. Int. Ed.
2020, 59, 10421–10425; Angew. Chem. 2020, 132, 10507–10511.
[19] J. Zhang, G. Leitus, Y. Ben-David, D. Milstein, J. Am. Chem. Soc. 2005,
127, 10840–10841.
[20] C. Gunanathan, Y. Ben-David, D. Milstein, Science 2007, 317, 790–792.
[21] a) E. Balaraman, E. Khaskin, G. Leitus, D. Milstein, Nat. Chem. 2013, 5,
122–125; b) P. Hu, Y. Ben-David, D. Milstein, J. Am. Chem. Soc. 2016,
138, 6143–6146.
[29] Note however, that the experimentally observed site of deprotonation
is only correctly predicted following the original procedure for parame-
trization of Cramer and Truhlar, that is, optimizing in the gas phase:
a) R. F. Ribeiro, A. V. Marenich, C. J. Cramer, D. G. Truhlar, J. Phys. Chem. B
2011, 115, 14556–14562; b) A. V. Marenich, C. J. Cramer, D. G. Truhlar, J.
Phys. Chem. B 2009, 113, 6378–6396. No standard state correction for 1
atm to 1m was applied, nor H2 solubility taken into account. These cor-
rections can be applied subsequently.
[30] Step 2 resembles Step 4, but is significantly lower in energy.
[31] a) L. A. Suàrez, Z. Culakova, D. Balcells, W. H. Bernskoetter, O. Eisenstein,
K. I. Goldberg, N. Hazari, M. Tilset, A. Nova, ACS Catal. 2018, 8, 8751–
8762; b) T. Leischner, L. A. Suarez, A. Spannenberg, K. Junge, A. Nova,
M. Beller, Chem. Sci. 2019, 10, 10566–10576.
[22] Y.-Q. Zou, N. von Wolff, A. Anaby, Y. Xie, D. Milstein, Nat. Catal. 2019, 2,
415–422.
[32] H. Li, M. B. Hall, J. Am. Chem. Soc. 2014, 136, 383–395.
[33] a) V. Sinha, N. Govindarajan, B. de Bruin, E. J. Meijer, ACS Catal. 2018, 8,
6908–6913 b) N. Govindarajan, V. Sinha, M. Trincado, H. Grützmacher,
E. J. Meijer, B. de Bruin, ChemCatChem 2020, 12, 2610.
[34] P. A. Dub, J. C. Gordon, Dalton Trans. 2016, 45, 6756–6781.
[35] F. Hasanayn, L. M. Al-Assi, R. N. Moussawi, B. S. Omar, Inorg. Chem. 2016,
55, 7886–7902.
[23] It is important to note that the reaction mixture is biphasic. Excess base
is used to quench any formed acidic product, both to drive the reaction
to completion and to prevent catalyst degradation or deactivation.
Indeed, DFT calculations predict that the reaction is endergonic and
hence formation of the glycolate salt from glycolic acid might be an ad-
ditional driving force of the reaction (besides irreversible H2 release).
[24] E. Fogler, J. A. Garg, P. Hu, G. Leitus, L. J. W. Shimon, D. Milstein, Chem.
Eur. J. 2014, 20, 15727–15731.
[36] L. Falivene, Z. Cao, A. Petta, L. Serra, A. Poater, R. Oliva, V. Scarano, L.
Cavallo, Nat. Chem. 2019, 11, 872–879.
[25] D. Milstein, E. Balaraman, C. Gunanathan, B. Gnanaprakasam, J. Zhang,
US 2013/0281664 A1, 2013.
[26] S. W. Kohl, L. Weiner, L. Schwartsburd, L. Konstantinovski, L. J. W.
Shimon, Y. Ben-David, M. A. Iron, D. Milstein, Science 2009, 324, 74–77.
Manuscript received: December 23, 2020
Accepted manuscript online: December 28, 2020
Version of record online: February 11, 2021
Chem. Eur. J. 2021, 27, 4715 –4722
www.chemeurj.org
4722
ꢀ 2020 Wiley-VCH GmbH