In the format provided by the authors and unedited.
©
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Supplementary
Information
A potassium
tert
-
butoxide and hydrosilane system for ultra
-
deep
desulfurization
of fuels
Anton A. Toutov,
1
Mike Salata,
2
Alexey Fedorov,
1, 3
Yun
-Fang Yang,
4
Yong Liang,
5
Renan Cariou,
2
Kerry N. Betz,
1
Erik P. A. Couzijn,
3
John W
. Shabaker,
2
K. N. Houk
4
& Robert H. Grubbs
1
john.shabaker@bp.com
, houk@chem.ucla.edu,
rhg@caltech.edu
Supplementary
Note
1:
Optimization of the hydrodesulfurization (HDS) of
dibenzothiophene (1S)
................................
................................
......................................
2
Supplementa
ry Note
2: Robustness evaluation.
................................
.............................
3
Supplementary
Note
3: Optimization of the
KOSi
HDS
of 4,6
-Me2DBT in ULSD (3).
................................................................
...........................
3
Supplementary
Note
4: Quantification of retained Si and K in the
KOSi
HDS of dibenzothiophene (1S).
................................
..............................................
4
Supplementary
Note
5: Observation of
deoxygenation of
4,6-
dimethyldibenzofuran (3O) at high temperatures.
..................................................
4
Supplementary
Note
6:
Free energies and transition states for the
KOSi
HDS
of 4,6
-dimethyldibenzothiophene (3).
. ................................
.............................................
5
Supplementary
Note
7:
Free energies and transition states for the a
ryl C
–
O
bond cleavage of 4,6
-dimethyldibenzofuran (3O) by the
KOSi
method
.. .....................
6
Supplementary Figures
. ................................
................................
................................
... 7
Supplementary
Tables
. ................................
................................
...................................
11
1
Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, California 91125,
United States.
2
BP
Products North America, 150 West Warrenville Rd, Naperville, IL 60563
.
3
Department of Chemistry and Applied Biosciences, ETH Zürich, Vladimir
-Prelog
-Weg 2, CH
-8093, Zürich,
Switzerland.
4
Department of Chemistry and Biochemistry, University of California,
Los Angeles, California 90095, United
States
.
5
State Key Laboratory of Coordination Chemistry, School of Chemistry and Chemical Engineering, Nanjing
University, Nanjing, 210023, China
.
SUPPLEMENTARY INFORMATION
ARTICLE NUMBER: 17008
|
DOI: 10.1038/NENERGY.2017.8
NATURE ENERGY
|
www.nature.com/natureenergy
1
©
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Supplementary Note
1:
Optimization of the hydrodesulfurization (HDS) of
dibenzothiophene (1S).
Investigating the HD
S of dibenzothiophene under a variety of conditions revealed several
important features of the reaction. First, the reaction appears to require a strong inorganic
base
(Supplementary Figure 1,
entries 4
–
10). The reaction does not occur with potassium
hydro
xide
(entry 3
) or with organic bases (entries 11
–
13). Second, the use of a potassium base appears to be
vital for HDS to occur; the reaction does not proceed with lithiu
m- or sodium alkoxide bases
(entries 1 & 2). Potassium
tert
-butoxide (KO
t
-Bu) proved to
be the optimal base for the HDS
(entry 7) as KOMe (entry 5) and KOEt (entry 6
) affor
ded the product
in significantly lower
yield
. It appeared then that the
effectiveness of the
KO
R/
Si
-H (
KOSi
) system correlates with the
basicity of the alkoxide; however,
the ease of solubility of the base in the reaction medium
cannot be discounted as a contributing factor. Third, the yield of biphenyl (
1a)
increases with
increasing temperature (entries 7→10). Although considerations of energy input (especially on
an indus
trial scale) led us to run the HDS
reactions
at 165 °C, it is gratifying to observe that
increasing the temperature to 200 °C further improves the yield of
1a
(entry 10). Finally,
triethylsilane (Et
3
SiH) demonstrated the highest HDS activity; a dihydrosila
ne, Et
2
SiH
2
,
performed poorly compared to Et
3
SiH (entry 14).
The light and low
-boiling EtMe
2
SiH was
also
less effective (entry 15).
No reaction occurs with Et
4
Si (entry 16) clearly demonstrating, as anticipated, that a silane with
an Si
–
H bond is required
to induce HDS reactivity. A brief solvent investigation reveal
ed that
besides
mesitylene, other simple aromatic (entries 18 & 21) as well as aliphatic (entry 19)
hydrocarbons are superior solvents for this reaction, whereas ethers (entries 17 & 20) are poo
r or
shut down reactivity (entry 23). Alcohols (entries 22 & 26), and amides (entries 24 & 25)
lik
ewise shut down the reactivity.
NATURE ENERGY
|
www.nature.com/natureenergy
2
SUPPLEMENTARY INFORMATION
DOI: 10.1038/NENERGY.2017.8
©
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Supplementary Note
2: Robustness evaluation.
Practical considerations of the
KOSi
method will have an important impact on
the likelihood of
its eventual implementation in industry. Fortunately, a brief robustness evaluation demonstrates
that the reaction
is tolerant of conditions that
model
a general operating environment in a
refinery setting (Supplementary Figure 2
).
The ro
bustness investigation conducted in the context of the
KOSi
HDS of
4,6-
Me
2
DBT (
3
)
shows that the reaction tolerates impurities such as those that would be found in bulk, unpurified
solvents and reagents and
that it
can be performed under air.
A 10
-hour rea
ction time
leads to
slightly lower conversion.
The reaction shows some sensitivity to water, as would be expected
given the sensitivity of KO
t
-Bu to moisture.
Although
impacting the yield (see
IV
), the reaction
pro
ceeds in the presence of ambient moisture
both on the
surface of the
glassware as well as in
the solvent and reagents.
The reaction also proceeds well in ultra
-low sulfur diesel (ULSD) as the
solvent (see
V
).
The scalability of the reaction was also evaluated using dibenzothiophene (
1S
) due to its
much
lower cost and greater ease of synthesis compared to
3
. The reaction scales well
with gram
quantities of
1S
to give the hydrocar
bon (
1a
)
product in 53 % yield after
10 h at 165 °C.
Supplementary
Note
3:
Optimization of the
KOSi
HDS of 4,6
-Me
2
DBT
in
ULSD
(3).
Having
determined that
the
KOSi
HDS of 4,6
-Me
2
DBT
proceeds well in ULSD
(Supplementary
Figure 2,
V
) we proceeded to investigate the effect of various parameters on the reaction
using
ULSD as a practically
-relevant model solvent
. Decreasing the nu
mber of equivalents each of
base and silane
from 3 (
Supplementary Figure 3,
standard conditions;
entry 1) to 2 (e
ntry 2) to
1
(entry 3) shows
a step
-wise decrease of the yield.
With 3 equivalents each of KO
t
-Bu and Et
3
SiH
the reaction time can be lowered t
o 10 hours with
a slight
drop in the yield (entry 4)
as was
shown in the case of mesitylene
as the
solvent (Supplementary Figure 2)
. Further lowering
reaction time to 5 hours results in a large corresponding decrease in yield (entry 5). Et
3
SiH could
be rep
laced by
polymethylhydrosiloxane (PMHS)
–
an inexpensive, non
-toxic, air
- and water
stable polymeric hydrosiloxane which is a byp
roduct of the silicone industry
–
though with
decreased yield.
NATURE ENERGY
|
www.nature.com/natureenergy
3
SUPPLEMENTARY INFORMATION
DOI: 10.1038/NENERGY.2017.8
©
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
In this latter case, despite the non
-negligible decrease in yiel
d, the fact that the
KOSi
reaction
proceeds in the presence of such a simple, inexpensive, and abundantly available Si
–
H source is
very surprising since PMHS is generally employed for facile reductions, most often of carbonyl
derivatives.
Most importantly,
KOSi
with PMHS
is a vital
proof of principle for future
improvements
toward eventual implementation.
Supplementary
Note
4:
Quantification of retained Si
and K
in the
KOSi
HDS
of dibenzothiophene (1S)
.
The analysis demonstrated that the silicon remaining
in the feed is very high. This is of course
expected since the eventual fate of the Si atom is sequestra
tion of the sulfur (see Figure 3A
,
TS6_S
–
formation of TMS
2
S, and the discussion in the main text) and
, like the potassium,
no
efforts were made to re
move any of the silicon from the mixture so as not to affect the
ultimate
[S] quantification in any way. Any unreacted or oxidized silane will also remain in the feed due
to low volatility. In order to advance the
KOSi
method from a decisive proof of princ
iple to a
practical polishing refining technology, the silicon source would have to be inexpensive such as
PMHS (Supplementary
Figure 3, entry 6). Conversely, homogeneous silicon species
–
such as
those used and formed herein
–
could be removed from the fe
ed by scrubbing or by distillation
and ideally be reactivated and reused. Regarding the
latter, since the major Si product is an Si
–
S
–
Si species, it may be possible to hydrogenate the two Si
–
S bonds using H
2
to reform the active
hydrosilane (i.e., reform
the Si
–
H bond) with concomitant release of H
2
S. Work is currently
ongoing toward this end.
Supplementary
Note
5:
Observation of deoxygenation of 4,6
-
dimethyldibenzofuran (3O) at high temperatures.
Having observed the efficient desulfurization activity of
dibenzothiophene derivatives under the
standard
KOSi
conditions, we became interested in whether O
-heterocycles could potentially be
deoxygenated as well. However, our earlier studies into reductive C
–
O cleavage emploing
hydrosilane/base reductive systems
showed no evidence of deoxygenation in dibenzofuran
substrates such as
3O
and only single C
–
O bond cleavage was observed giving
3a
–
OH
(Supplementary Figure 5, entry 1).
NATURE ENERGY
|
www.nature.com/natureenergy
4
SUPPLEMENTARY INFORMATION
DOI: 10.1038/NENERGY.2017.8
©
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
At that point, it was not clear why deoxygenation in those substrates was not observed.
However,
with the knowledge gleaned from the density functional theory studies performed herein, it
becomes clear that the barrier to deoxygenation of these substrates (e.g.,
3O
) is simply too high
to be overcome at the standard
KOSi
temperatures. Indeed,
the energy for displacement of
Me
3
SiO by Me
3
Si is quite high (37.5 kcal/mol)
. Increasing the temperature of the reaction to 165
°C (entry 2) and finally to 200 °C (entry 3) systematically results in increased
conversions
as
expected
and results in high yi
elds of
3a
–
OH
.
However, c
onducting the reaction at
200 °C
for 60
h, 7% of
3a
is obtained, constituting the first observations of deoxygenation of a dibenzofuran
derivative using the
KOSi
system
. This result lends
support to the proposed mechanism and
corre
ponding energetics provided by density
functional theory calculations
and warrants further
exploration into hydrodeoxygenation (HDO) chemistry by
KOSi.
Supplementary
Note
6:
Free energies and transition states for the
KOSi
HDS
of
4,6
-dimethyl
dibenzothioph
ene (3)
.
Refer to Figure 3A in the main text.
The
carbon of the C
–
S bond
of the substrate
(
3
) is proposed
to be attacked by silyl radical through
TS1_S
with a barrier of 15.5 kcal/mol to form
Int1_S
,
which can undergo C
–
S bond
homolytic cleavage through
TS
2_S
(11.8 kcal/mol) to form
Int2_S
. The subsequent silyl radical migration from C to S through
TS3_S
has a high barrier of
31.8 kcal/mol.
The
direct attack onto the heteroatom, S, of the substrate was also considered,
followed by C
–
S homolytic cleavage thr
ough
TS4_S
. This transition state has a lower barrier
than the migration transition state
TS3_S
by 5.4 kcal/mol. Once the carbon radical species,
Int3_S
, is formed, it can abstract an H
atom
from
trimethyl
silane through
TS5_S
to form
silylated
biaryl
-2- thi
ol
Int4_S
. This hydrogen abstraction step is the rate
-determining step with
an overall barrier of 27.2 kcal/mol.
Then
silyl radical attack at the S atom through
TS6_S
forms
biaryl radical and
disilathiane
to complete the desulfurization process. Finally, t
he biaryl radical
abstraction of an H
atom
from
trimethyl
silane (
TS7_S
) is a very facile process to generate the
final biaryl product and regenerate the silyl radical. The overall process is e
xergonic by 32.7
kcal/mol.
In summary, for substrate
4,6-
dimethy
ldibenzo
thiophene (
3
),
silyl radical attack at the
heteroatom S and eventual attack of a second silyl radical at the S of the biaryl silyl thioether to
give desulfurization is a favorable pathway. The calculated transition states discussed above are
visual
ized in Supplementary Figure 6.
NATURE ENERGY
|
www.nature.com/natureenergy
5
SUPPLEMENTARY INFORMATION
DOI: 10.1038/NENERGY.2017.8
©
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Supplementary
Note
7: Free energies and transition states for the
aryl C
–
O
bond cleavage of
4,6
-dimethyldibenzofuran (3O) by the
KOSi
method.
Refer to Figure 3B in the main text.
Different from the 4,6
-dimethyldibenzothiop
hene case
(
3
) ,
the direct attack
of the silyl radical to the heteroatom
of
3O
, breaking
one
C
–
O bond
, is very
unfavorable with a barrier of 43.
3 kcal/mol (
TS4_O
). Instead, silyl radical attack at
carbon atom
of the C
–
O bond
of 4,6
-dimethyldibenzofuran
(
3O
) through
TS1_O
has a barrier of 16.2
kcal/mol
, leading to
Int1_O
, which
undergoes
C
–
O bond homolytic cleavage through
TS2_O
(22.4 kcal/mol) to generate
Int2_O
. The subsequent silyl radical migration from C to O through
TS3_O
requires a
n overall activation
free energy
of
26.7 kcal/mol and is the rate
-determining
step. The carbon radical species,
Int3_O
, is relative stable, and can abstract an H
atom
from
trimethyl
silane through
TS5_O
to form silylated bi
aryl-2- ol,
Int4_O
. The silyl radical
reattack
ing
the O
atom through
TS6_O
is a very unfavorable process
with a barrier of 37.5
kcal/mol
. It is speculated that silylated bi
aryl-2- ol,
Int4_O
, can undergo hydrolysis to form
the
biaryl-2- ol product
3a
–
OH
that is observed
experimentally.
In summary
, for substrate
4,6-
dimethyldibenzo
furan
(
3O
), the silyl radical attack at O is unfavorable for both the substrate and
the intermediate biaryl silyl ether. The silyl radical preferentially attacks the aryl group of the
substrate and finally undergoes hydrolysis to generate
biaryl
-2- ol
3a
–
OH
instead of the
deoxygenation product
3a
.
The calculated transition states discussed above are visualized in
Supplementary Figure 7.
NATURE ENERGY
|
www.nature.com/natureenergy
6
SUPPLEMENTARY INFORMATION
DOI: 10.1038/NENERGY.2017.8
©
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Supplementary Figures
Supplementary Figure 1
.
Discovery and optimization of the
KOSi
HDS of di
benzothiophene (
1S
).
a
Yields determined
by GC
-FID analysis
using tridecane as a
standard.
NATURE ENERGY
|
www.nature.com/natureenergy
7
SUPPLEMENTARY INFORMATION
DOI: 10.1038/NENERGY.2017.8
©
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Supplementary Figure 2
.
Robustness evaluation of the
KOSi
HDS of 4,6
-dimethyldibenzothiophene (
3
)
and gram
-scale HDS of dibenzothiophe
ne (
1S
).
a
Yields
determined
by GC
-FID analysis
using tridecane as a
standard.
Supplementary Figure 3
.
KOSi
HDS of
4,6-
dimethyl
dibenzothiophene (
3
) under varying conditions.
a
Yields are by GC
-FID analysis
using tridecane as a
standard.
NATURE ENERGY
|
www.nature.com/natureenergy
8
SUPPLEMENTARY INFORMATION
DOI: 10.1038/NENERGY.2017.8
©
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Supplementary Figure 4.
Evaluation of Si
and K
content remaining in
the
feed after
KOSi
HDS.
Supplementary Figure 5
.
Observed deoxygenation of 4,6
-dimethyldibenzofuran (
3O
).
a
Yields are by GC
-FID analysis with
tridecane as a
standard.
NATURE ENERGY
|
www.nature.com/natureenergy
9
SUPPLEMENTARY INFORMATION
DOI: 10.1038/NENERGY.2017.8
©
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Supplementary Figure 6.
Key transition states for
KOSi
HDS of
4,6-
dimethyl
dibenzothiophene (
3
).
Supplementary
Figure
7.
Key transition
states for
reductive C
–
O bond
cleavage
in
4,6-
dimethyl
dibenzofuran
(
3
) by the
KOSi
method.
NATURE ENERGY
|
www.nature.com/natureenergy
10
SUPPLEMENTARY INFORMATION
DOI: 10.1038/NENERGY.2017.8
©
2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Su
pplemen
ta ry
Tables
Supp
lementary Tab
le 1.
Electroni c ene rgies, ent ha lpi es, and
free energies of the
structur es c
alculate
d a
t t he M06-
2X
/6- 311
+G(d,p)
(CPCM
mesitylene
)//B3LYP/ 6- 31G
(d) .
S
t
r
u
c
t
u
r
e
s
Z
P
V
E
T
C
E
T
C
H
T
C
G
E
s
ol
H
s
ol
(
E
s
ol
+
T
C
H
)
G
s
ol
(
E
s
ol
+
T
C
G
)
I
m
a
g
i
n
a
r
y
F
r
e
q
u
e
n
c
y
(
c
m
-
1
)
E
s
ol
(
k
ca
l/
m
o
l)
H
s
ol
(
k
ca
l/
m
o
l)
G
s
ol
(
k
ca
l/
m
o
l)
.
S
i
M
e
3
0.110424
0.117885
0.118829
0.079221
-
409.158357
-
409.039528
-
409.079136
—
—
—
—
3
0.217085
0.229713
0.230657
0.178435
-
938.839221
-
938.608564
-
938.660786
—
0.0
0.0
0.0
I
n
t
1_S
0.328288
0.349215
0.350159
0.279833
-
1348.014038
-
1347.663879
-
1347.734205
—
-
10.3
-
9.9
3.6
I
n
t
2_S
0.330154
0.351135
0.352080
0.280549
-
1348.012083
-
1347.660003
-
1347.731534
—
-
9.1
-
7.5
5.3
I
n
t
3_S
0.328484
0.350246
0.351190
0.276754
-
1347.992661
-
1347.641471
-
1347.715907
—
3.1
4.2
15.1
I
n
t
4_S
0.341170
0.363113
0.364057
0.289086
-
1348.674759
-
1348.310702
-
1348.385673
—
-
17.4
-
14.2
-
4.1
I
n
t
5_S
0.224008
0.236741
0.237686
0.181958
-
541.146777
-
540.909091
-
540.964819
—
-
24.5
-
21.6
-
12.8
3a
0.237022
0.249704
0.250648
0.196045
-
541.831749
-
541.581101
-
541.635704
—
-
46.9
-
41.7
-
32.7
T
S
1_S
0.327127
0.348360
0.349304
0.276936
-
1347.992210
-
1347.642906
-
1347.715274
-
337.565
3.4
3.3
15.5
T
S
2_S
0.328544
0.348977
0.349922
0.280571
-
1348.001678
-
1347.651756
-
1347.721107
-
120.53
-
2.6
-
2.3
11.8
T
S
3_S
0.328904
0.349338
0.350282
0.280941
-
1347.970110
-
1347.619828
-
1347.689169
-
202.426
17.2
17.7
31.8
T
S
4_S
0.327483
0.348823
0.349767
0.276719
-
1347.974625
-
1347.624858
-
1347.697906
-
246.532
14.4
14.6
26.4
T
S
5_S
0.445642
0.476409
0.477353
0.381812
-
1757.796673
-
1757.319320
-
1757.414861
-
837.468
5.4
5.2
27.2
T
S
6_S
0.451190
0.482019
0.482964
0.386782
-
1757.827069
-
1757.344105
-
1757.440287
-
270.192
-
13.6
-
10.3
11.3
T
S
7_S
0.341541
0.363019
0.363963
0.288859
-
950.953069
-
950.589106
-
950.664210
-
837.588
-
23.6
-
21.9
-
0.9
3O
0.220027
0.232261
0.233205
0.181464
-
615.861082
-
615.627877
-
615.679618
—
0.0
0.0
0.0
I
n
t
1_O
0.331020
0.351420
0.352365
0.283120
-
1025.031234
-
1024.678869
-
1024.748114
—
-
7.4
-
7.2
6.7
I
n
t
2_O
0.331018
0.352039
0.352984
0.2811
7
6
-
1025.038304
-
1024.685320
-
1024.757128
—
-
11.8
-
11.2
1.0
I
n
t
3_O
0.330414
0.351891
0.352836
0.278016
-
1025.041895
-
1024.689059
-
1024.763879
—
-
14.1
-
13.6
-
3.2
I
n
t
4_O
0.343489
0.364900
0.365844
0.292144
-
1025.726940
-
1025.361096
-
1025.434796
—
-
36.5
-
33.7
-
23.1
T
S
1_O
0.330084
0.350780
0.351724
0.280156
-
1025.013163
-
1024.661439
-
1024.733007
-
339.657
3.9
3.7
16.2
T
S
2_O
0.330092
0.350194
0.351138
0.282888
-
1025.005890
-
1024.654752
-
1024.723002
-
254.07
8.5
7.9
22.4
T
S
3_O
0.330244
0.350341
0.351285
0.2823
0
6
-
1024.998477
-
1024.647192
-
1024.716171
-
183.413
13.2
12.7
26.7
T
S
4_O
0.327728
0.348515
0.349459
0.278658
-
1024.968347
-
1024.618888
-
1024.689689
-
642.444
32.1
30.4
43.3
T
S
5_O
0.447660
0.478053
0.478997
0.384030
-
1434.848709
-
1434.369712
-
1434.464679
-
703.013
-
13.5
-
14.3
7.8
T
S
6_O
0.452636
0.482724
0.483668
0.391004
-
1434.845160
-
1434.361492
-
1434.454156
-
765.371
-
11.3
-
9.1
14.4
.
S
i
E
t
3
0.197772
0.209068
0.210012
0.159653
-
527.056709
-
526.846697
-
526.897056
—
—
—
—
T
S
1_
S
_
E
t
0.414615
0.439549
0.440493
0.359844
-
1465.890237
-
1465.449744
-
1465.530393
-
346.000
3.6
3.5
17.2
T
S
4_
S
_
E
t
0.414891
0.440035
0.440980
0.358859
-
1465.875045
-
1465.434065
-
1465.516186
-
253.873
13.1
13.3
26.1
ZPVE
= zero
-p oint
vibrational energy;
TC
E
= therma
l corr
ection
to energy;
TC
H
= thermal corr
ection
to
enthalpy;
TC
G
= thermal corr
ect
ion
to Gibbs free energy.
NATURE ENERGY
|
www.nature.com/natureenergy
11
SUPPLEMENTARY INFORMATION
DOI: 10.1038/NENERGY.2017.8