Specific Photochemical Dehydrocoupling of N-Heterocyclic Phosphanes and Their Use in the Photocatalytic Generation of Dihydrogen

Article abstract of DOI:10.1002/anie.201504504

N-Heterocyclic phosphanes react under UV irradiation in a highly selective dehydrocoupling reaction to diphosphanes and H₂. Computational studies suggest that the product formation is initiated by the formation of dimeric molecular associates w

Full text of DOI:10.1002/anie.201504504

Angewandte
Chemie
International Edition: DOI: 10.1002/anie.201504504
German Edition: DOI: 10.1002/ange.201504504
Photocatalysis
Specific Photochemical Dehydrocoupling of N-Heterocyclic
Phosphanes and Their Use in the Photocatalytic Generation of
Dihydrogen
Oliver Puntigam, Lµszló Kçnczçl, Lµszló Nyulµszi,* and Dietrich Gudat*
Dedicated to Professor Manfred Scheer on the occasion of his 60th birthday
À
Abstract: N-Heterocyclic phosphanes react under UV irradi-
ation in a highly selective dehydrocoupling reaction to
diphosphanes and H2. Computational studies suggest that the
product formation is initiated by the formation of dimeric
molecular associates whose electronic excitation yields H2 and
a diphosphane. Combining the dehydrocoupling of sterically
demanding phosphanes with Mg-reduction of the formed
diphosphanes allows constructing a reaction cycle for the
photocatalytic reductive generation of H2 from Et3NH+.
phosphorus–element bonds (including P P bonds).[3] Assem-
À
bly of P P bonds through uncatalyzed dehydrocoupling is
also known, but has received little interest. Thermally induced
reactions need high temperatures (175–2008C), which is
detrimental for selectivity and product stability.[4] Photo-
chemically induced dehydrocoupling seems to be limited to
parent phosphane, which reacts upon UV photolysis to give
H2 and P2H4 as primary products.[5] However, selectivity is
likewise low, as diphosphane is photochemically unstable and
decays to higher phosphanes and eventually elemental
phosphorus.
During earlier syntheses of secondary N-heterocyclic
phosphanes, we had noted that some of the formed products
were converted to diphosphanes when the reaction mixtures
were exposed to light (Scheme 2).[6] Inspired by this observa-
C
ondensation of two element–hydrogen bonds to afford
À
a new element–element (E E) bond and dihydrogen is
commonly denoted as “dehydrocoupling” (Scheme 1; the
Scheme 1. a) Homo- and b) hetero- dehydrocoupling reaction.
[1]
À
E E bond formed can be homo- or heteronuclear). The
Scheme 2. Reductive diphosphane synthesis via a spectroscopically
detectable secondary phosphane intermediate. R=tBu, Xyl (2,6-
Me2C6H3).
reaction is generally carried out with a catalyst, which permits
to apply less forcing conditions and widens the scope of
useable element–hydride substrates. Currently, dehydrocou-
pling routes draw growing attention in main-group chemistry
as complement to established protocols (e.g., Wurtz-type
tion, we have now carried out a detailed study of the
photolysis of N-heterocyclic phosphanes and could establish
that these species undergo photochemically induced dehy-
drocoupling which can be employed in a reaction cycle for
photocatalytic generation of H2.
reductions or dehydrohalogenation/salt elimination) for syn-
[1]
À
theses of E E-bonded frameworks. Lately, they have also
gained interest as a means to release H2 from chemical
storage materials.[2]
Whereas early studies mainly focused on the bond
formation between boron atoms or heavier Group 14 ele-
ments, catalyzed dehydrocoupling has meanwhile developed
into a valuable approach for the selective formation of
Photolysis experiments were carried out by irradiating
solutions of isolated, pure N-heterocyclic phosphanes 1a–c/
1’a,c in inert solvents (THF, hexane) with a medium-pressure
Hg-lamp. Monitoring the progress of the reactions by
31P NMR spectroscopy confirmed that a clean transformation
to the diphosphanes 2/2’ took place (Scheme 3 and Figure 1).
The formation of dihydrogen as by-product was estab-
lished by the direct observation of its characteristic signal
(d1H = 4.69) in the 1H NMR spectrum of an irradiated
solution of 1a in [D8]THF, and by trapping of the gas evolving
from the reaction mixture through adsorption on palladium
and elemental analysis.[7] The 1H and 31P NMR spectra of the
reaction mixture showed no signals of transient intermediates
or by-products. The reaction stopped when the photolysis was
interrupted, and diphosphane formation was not at all
detectable in control samples that were prepared and stored
[*] O. Puntigam, Prof. Dr. D. Gudat
Institut für Anorganische Chemie
University of Stuttgart
Pfaffenwaldring 55, 70550 Stuttgart (Germany)
E-mail: gudat@iac.uni-stuttgart.de
L. Kçnczçl, Prof. Dr. L. Nyulµszi
Department of Inorganic and Analytical Chemistry
Budapest University of Technology and Economy
Szent GellØrt tØr 4, 1111 Budapest (Hungary)
E-mail: nyulaszi@mail.bme.hu
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.201504504.
Angew. Chem. Int. Ed. 2015, 54, 11567 –11571
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
11567

.
Angewandte
Communications
aug-cc-pVDZ level to account for the dispersion effects
associated with the bulky substituents. Calculations of excited
states were carried out at the TD-DFT CAM-B3LYP/aug-cc-
pVDZ level, which was selected after a performance study of
different computational levels on 1’d and allows a suitable
modeling of Rydberg states.
For the electronic ground states of 1a and 1’a, three
conformational isomers with different alignment of N-tBu
and PH substituents were located. As the relative energies of
these conformers differ by less than 2 kcalmolÀ1, it is possible
that all three contribute to the observed spectrum. The lowest
(vertical) excitation is predicted to occur around 230 nm for
1a and at 280–300 nm for 1’a, depending on the conformation.
In addition, each species may undergo several further
excitations in the near UV region (7–9 out of the 15 lowest
energy transitions for each conformer of 1a and 12–15 tran-
sitions for each conformer for 1’a are predicted to occur at
wavelengths > 200 nm). Due to the high density of states and
the coexistence of several conformers, the spectra are
expected to exhibit rather little structure. This is in accord
with the habit of the experimental UV-spectra of 1a and 1’a,
which display an end-absorption with an inflection arising
from an unresolved absorption band. The band maxima (1a:
lmax ꢀ 237 nm; 1b: lmax ꢀ 274 nm as inferred from spectral
deconvolution) are red-shifted with respect to the spectral
bands of PH3 or simple trialkylphosphanes (lmax = 200–
210 nm).[9] Both the position of these bands and the ca.
40 nm redshift of the band onset of 1’a with respect to 1a are
in accord with the calculated wavelengths of the lowest
energy transitions.
Scheme 3. Photochemically induced dehydrocoupling of N-heterocyclic
secondary phosphanes at ambient temperature (Dipp=2,6-iPr2C6H3,
Cy =cyclohexyl, tBu=tert-butyl); [a] determined by integration of
31P NMR signals. [b] The P-deuterated derivative (1a-D1) gave 86%
conversion after 28 h.
This lowest energy excitation is of specific importance
because photochemical reactions originate from this state
(Kasha’s rule). Our calculations indicate that each conformer
of 1a and 1’a exhibits two close-lying excited states which are
of mixed valence/Rydberg character. Both excitations origi-
nate from the HOMO, which is a symmetrical combination of
the nitrogen lone pairs with some hyperconjugative inter-
action with the sPH and (for 1’a) pCC orbitals (for an
illustration see the HOMO of 1’d in Figure 2a). The 8–10
lowest unoccupied orbitals of each conformer of 1a and 1’a
are of diffuse nature. The LUMO is always an s-type and the
LUMO + 1 a p-type diffuse orbital with similar symmetry as
the lowest energy s*-valence orbital (LUvMO), which has
NPN antibonding character (Figure 2b shows the appropriate
orbital of 1’d). One of the two low-energy transitions is mainly
of HOMO–LUMO (Rydberg) nature with small contribu-
tions from a valence excitation, whereas the other one
Figure 1. Photochemical conversion 1a!2a in THF at ambient tem-
perature as monitored by 31P NMR spectroscopy.
under identical conditions, but protected from UV light.
À
These findings confirm that P P bond formation requires no
other reactants (which may be present in the original reaction
mixtures of Scheme 2) than the phosphanes 1/1’, that the
conversion is a photochemical reaction which does not go
along with a thermal background process, and that the
transformation proceeds as a selective, photochemically
induced dehydrocoupling as shown in Scheme 3.
The reaction is amenable to diazaphospholenes (1’a,c) and
diazaphospholidines (1a–c) with N-aryl and N-alkyl substitu-
ents (Scheme 3). Low-pressure Hg-lamps can also be used as
radiation source. Preliminary studies indicate that the initial
rate of diphosphane formation decreases with increasing
concentration of the phosphane precursor, suggesting that the
reaction does not follow a simple rate law. Unlike P2H4[5] and
P3H5,[8] the diphosphanes 2/2’ did not undergo detectable
follow-up reactions and seem thus photochemically stable.
To gain insight into the electronic excitation and excited-
state characteristics of the N-heterocyclic phosphanes, we
carried out DFT calculations on 1a, 1’a, and the N-Me-
substituted model compound 1’d. Ground-state relative
energies and geometries were computed at the wB97xD/
involves
a mixture of HOMO–LUvMO and HOMO–
LUMO+1 excitations. For two conformers of 1a, the s*PH
orbital is only slightly higher than the LUvMO. It is note-
Figure 2. MOLDEN representation of a) the Kohn–Sham HOMO and
b) the lowest-energy unoccupied valence MO (LUvMO) of 1’d.
11568 www.angewandte.org
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 11567 –11571

Angewandte
Chemie
worthy that for all 1’a conformers and for one 1a conformer
the HOMO–LUMO (pure Rydberg) excitation is not the one
at lowest energy.
reaction was calculated to be weakly exergonic (DG298 =
À1.5 kcalmolÀ1) and the second one slightly endergonic
(DG298 = 2.4 kcalmolÀ1). NMR studies on solutions of 2c
and D2 (which was used instead of H2 in order to distinguish
a possible bond activation product from species formed by
diphosphane hydrolysis[6]) in THF or [D8]THF gave no
evidence for the formation of D2-activation products between
ambient temperature and 608C. We conclude that the
dehydrocoupling of 1c is therefore indeed exergonic, or that
reaction of H2 with 2c (or the phosphanyl radicals formed by
TD-DFT geometry optimization of the first two excited
states of 1a and 1’a,d gave no evidence for a PH-dissociative
nature which could explain the observed photochemical
reactivity. Excitations with significant contribution from the
HOMO–LUvMO electron configuration are accompanied by
distinct PN bond lengthening (in accord with the PN
antibonding character of this orbital),[10] but even geometry
optimization from a starting configuration with an extended
P P bond fission[6,13]) is impeded by a substantial activation
À
À
P H distance (1.6 Š) gave no indication for a possible decay
barrier.
through PH bond cleavage. In contrast, geometry optimiza-
As alternative to the direct activation of H2, regeneration
of a secondary phosphane from a diphosphane is also feasible
through a reductive pathway. The Et3NHCl/Mg reduction of
[11]
tion of the first excited state of PH3 at the same computa-
tional level converged to a conical intersection which opened
a pathway to PH bond fission, in accord with the known
photochemical behavior of phosphine. It seems thus that the
H-elimination channel in 1a and 1’a can only be reached from
higher excited states.
In search for an alternative mechanism, we considered the
formation of a phosphane dimer, either in the ground state or
an excited state. During geometry optimization of dimeric
structures (1’d)2 at the wB97XD/aug-cc-PVDZ level, we
located one minimum with a stabilization energy of 5.2 kcal
molÀ1 relative to the monomers and a P–P distance of 3.326 Š.
RI-CC2/aug-cc-pVDZ geometry optimization gave a struc-
ture with a P–P distance of 3.089 Š and 8.9 kcalmolÀ1
dimerization energy. Its LUvMO is a bonding combination
of the LUvMOs of two monomers (Figure 3), and the first
À
sterically strained N-heterocyclic diphosphanes with weak P
P bonds[14] like 2c/2’c seems particularly interesting for this
purpose: the reaction proceeds at ambient temperature,[6] and
its combination with the photochemical dehydrocoupling of
the phosphanes (1c/1’c) allows one to design a reaction cycle
that should enable the reductive generation of H2 from
Et3NH+ with
(Scheme 4).
a catalytic amount of the diphosphane
Scheme 4. Photocatalytic generation of H2 from Et3NH+ with magne-
sium as stoichiometric reductant and sterically demanding N-hetero-
cyclic phosphanes/diphosphanes as photocatalysts. R=Dipp.
In line with this idea, we studied the reduction of Et3NHCl
with Mg in THF in the presence of 2c[6] or P-bromo-
diazaphospholene 3’c,[12] which is reduced in situ to 2’c[6] and
may thus serve as precatalyst. In a first experiment, a mixture
of the reactants in [D8]THF was photolyzed in an NMR tube
Figure 3. MOLDEN representation of the CAM-B3LYP/aug-cc-PVDZ
LUvMO of a dimeric structure (1’d)2.
1
excited state has a significant contribution from the HOMO–
LUvMO electron transition. The excitation energy of the
dimer (at the RI-CC2/aug-cc-pVDZ geometry) is by 0.14 eV
lower than that of the monomer, indicating a slight stabiliza-
tion upon excimer formation. CC2 geometry optimization of
this state converged to a conical intersection which enables
the split of H2 and formation of the diphosphane 2’d. We
conclude that photochemical excitation of a dimer provides
thus a suitable explanation of the observed reaction behavior.
Since it is of interest whether the photochemically induced
H2 formation is reversible, we investigated computationally
the thermochemistry of the dehydrocoupling reactions
(Scheme 3) for 1’d and its saturated analogue 1d. The first
and the amount of liberated Et3N monitored by H NMR
spectroscopy (Figure 4a). The observed molar ratio
n(Et3N):n(2’c) ꢀ 8.8 clearly exceeds the value of two expected
for the reaction[6]
20c þ 2 Et3NHCl þ Mg ! 2 10c þ MgCl2 þ 2 Et3N
which implies that the regeneration of 2’c by photolytic
dehydrocoupling enables a turnover number (TON) > 1 and
the reaction is thus catalytic in diphosphane. This was
confirmed by further experiments which were conducted at
higher dilution with 3’c or 2c as (pre)catalyst and monitored
by measuring the volume of evolved H2 (Figure 4b). The
Angew. Chem. Int. Ed. 2015, 54, 11567 –11571
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 11569

.
Angewandte
Communications
dehydrocoupling as specific reaction channel for functional
phosphanes suggests that this transformation may also play
a role in other photochemical processes (cf. the photolytic
high-pressure reaction of red phosphorus with water to give
H2 and phosphorus oxyacids[15]), and we are currently study-
ing possible applications for further types of phosphanes.
Experimental Section
The starting materials 1a–c,[6] 1’a,c,[16] and 3c[12] were prepared as
described. Photolysis experiments with pure phosphanes were carried
out in 3 mmol solutions in degassed and rigorously dried solvents
(THF, [D8]THF, hexane) in Pyrex NMR tubes, and the reactions were
monitored by 31P and—where appropriate—1H NMR spectroscopy.
Photocatalytic reduction of Et3NHCl was carried out a) by irradiating
a mixture of magnesium powder (13.4 mg, 0.55 mmol), Et3NHCl
(140 mg, 1.02 mmol), and 3’c: 5 mg, 1.13 ” 10À2 mmol) in 0.6 mL
[D8]THF, or b) by irradiating a slurry of magnesium turnings (100 mg,
4.2 mmol), Et3NHCl (700 mg, 5.1 mmol), and the phosphane (pre)-
catalyst (3’c: 100 mg, 0.20 mmol; 2c: 73 mg, 0.090 mmol) in THF
(25 mL) with a medium-pressure Hg-lamp. The progress of the
1
reaction was either monitored by H NMR spectroscopy (reaction a,
Figure 4a) or volumetric determination of H2 (reaction b, Figure 4b).
Further experimental details and characterization data (UV spectra)
as well as a comprehensive account of the computational studies are
given in the Supporting Information.
Acknowledgements
Financial support from the Deutsche Forschungsgemeinschaft
(GU415/16-1), the Hungarian Scientific Research Fund
OTKA NN 113772 and COST CM10302 (SIPS) is gratefully
acknowledged. We thank Mark Wiedenmeyer (IAC) for
carrying out the analytical determination of H2.
Figure 4. a) Plot of mole of Et3N per mole of diphosphane produced
during photolysis of 2’c (formed in situ from Mg and 3’c) in the
presence of excess Mg/Et3NHCl versus reaction time. Data were
derived from integration of 1H NMR spectra recorded during the
reaction. b) Plot of mole of H2 generated per mole of diphosphane
during photolysis of 2’c (formed in situ by Mg reduction of 3’c, blue
circles) or 2c (red diamonds) in the presence of excess Mg/Et3NHCl
versus reaction time. Grey areas indicate time periods when the
irradiation was switched off and gas evolution ceased. Open green
squares indicate the result of a control measurement without phos-
phane. The dashed horizontal lines mark the amount of Et3N or H2
formed in a stoichiometric reaction Mg+2’c+2Et3NH+!
1’c+2Et3N+Mg2+ +H2.
Keywords: computational studies · dehydrocoupling ·
hydrogen · N-heterocyclic phosphanes · photolysis
How to cite: Angew. Chem. Int. Ed. 2015, 54, 11567–11571
Angew. Chem. 2015, 127, 11730–11734
[1] a) F. Gauvin, J. F. Harrod, H. G. Woo, Adv. Organomet. Chem.
1998, 42, 363 – 405; b) D. W. Stephan, Angew. Chem. Int. Ed.
2000, 39, 314 – 329; Angew. Chem. 2000, 112, 322 – 338; c) J. Y.
Corey, Adv. Organomet. Chem. 2004, 51, 1 – 52; d) T. J. Clark, K.
Lee, I. Manners, Chem. Eur. J. 2006, 12, 8634 – 8648; e) V. Pons,
R. T. Baker, Angew. Chem. Int. Ed. 2008, 47, 9600 – 9602; Angew.
Chem. 2008, 120, 9742 – 9744; f) S. Greenberg, D. W. Stephan,
Chem. Soc. Rev. 2008, 37, 1482; g) R. Waterman, Dalton Trans.
2009, 18 – 26.
results illustrate that up to 18 turnovers (with 2’c as catalyst)
were reached, and that the activity of the phosphane–
diphosphane pair acting as photocatalyst can be switched on
and off by control of the radiation source. The higher
conversion rate for the diazaphospholene-based photocata-
lyst indicates that the high tendency of 2’c toward homolytic
[2] A. Staubitz, A. P. M. Robertson, I. Manners, Chem. Rev. 2010,
110, 4079 – 4124.
[13]
À
P P bond cleavage may have some impact on its catalytic
activity.
[3] a) R. Waterman, Curr. Org. Chem. 2008, 12, 1322 – 1339; b) R. J.
Less, R. L. Melen, V. Naseri, D. S. Wright, Chem. Commun.
2009, 4929 – 4937; c) R. J. Less, R. L. Melen, D. S. Wright, RSC
Adv. 2012, 2, 2191 – 2199; d) V. Naseri, R. J. Less, M. McPartlin,
R. E. Mulvey, D. S. Wright, Chem. Commun. 2010, 46, 5000;
e) M. S. Hill, M. F. Mahon, T. P. Robinson, Chem. Commun.
2010, 46, 2498.
[4] A. B. Burg, R. I. Wagner, J. Am. Chem. Soc. 1953, 75, 3872 –
3877.
[5] a) J. P. Ferris, R. Benson, Nature 1980, 285, 156 – 157; b) J. P.
Ferris, R. Benson, J. Am. Chem. Soc. 1981, 103, 1922 – 1927;
In conclusion, we have shown that functional phos-
phanes 1 and 1’ undergo, unlike PH3 or P2H4, selective
photochemically induced dehydrocoupling at room temper-
ature. Computational studies suggest that the reaction is
initiated by formation of phosphane dimers whose photo-
chemical excitation yields H2 and the diphosphane. This
process can be implemented into a reaction cycle which
enables photocatalytic generation of H2 on demand upon
irradiation with UV light. The feasibility of photolytic
11570 www.angewandte.org
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 11567 –11571

Angewandte
Chemie
c) J. P. Ferris, A. Bossard, H. Kwaja, J. Am. Chem. Soc. 1984, 106,
318 – 324; d) J. Blazejowski, F. W. Lampe, J. Phys. Chem. 1981,
85, 1856 – 1864.
[12] J. W. Dube, G. J. Farrar, E. L. Norton, K. L. S. Szekely, B. F. T.
Cooper, C. L. B. Macdonald, Organometallics 2009, 28, 4377 –
4384.
[13] D. Fçrster, H. Dilger, F. Ehret, M. Nieger, D. Gudat, Eur. J.
Inorg. Chem. 2012, 3989 – 3994.
[6] O. Puntigam, D. Fçrster, N. A. Giffin, S. Burck, J. Bender, F.
Ehret, A. D. Hendsbee, M. Nieger, J. D. Masuda, D. Gudat, Eur.
J. Inorg. Chem. 2013, 2041 – 2050.
[7] A. Gurlo, D. R. Clarke, Angew. Chem. Int. Ed. 2011, 50, 10130 –
10132; Angew. Chem. 2011, 123, 10312 – 10314.
[8] T. P. Fehlner, J. Am. Chem. Soc. 1968, 90, 6062 – 6066.
[9] A. M. Halpern, B. R. Ramachandran, J. Mol. Spectrosc. 1992,
151, 26 – 32.
[10] It is noteworthy that a preliminary B3LYP/cc-PVDZ TD-DFT
geometry optimization on 1’d resulted in (experimentally not
observed) phosphinidene elimination, showing the importance
of the participation of the Rydberg excitation accounted for by
the diffuse functions.
À
[14] The bond weakening is evidenced by P P bond elongation
( ꢀ 2.30 Š) and partial dissociation into persistent radicals even
at room temperature; cf. Refs. [6] and [13]. Computational
studies (see the Supporting Information) suggest that the
diphosphanes exhibit nonetheless no biradicaloid character.
[15] M. Ceppatelli, R. Bini, M. Caporali, M. Peruzzini, Angew. Chem.
Int. Ed. 2013, 52, 2313 – 2317; Angew. Chem. 2013, 125, 2369 –
2373.
[16] S. Burck, D. Gudat, M. Nieger, W.-W. du Mont, J. Am. Chem.
Soc. 2006, 128, 3946 – 3955.
[11] This state is a mixture of a HOMO!s*PH and a HOMO!
Rydberg excitation (M. B. Robin, Higher Excited States of
Polyatomic Molecules, Vol. III, Academic Press, Orlando, 1985).
Received: May 18, 2015
Revised: June 14, 2015
Published online: July 17, 2015
Angew. Chem. Int. Ed. 2015, 54, 11567 –11571
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 11571