propylene yield data, reported as a function
of time for different catalysts, were analyzed
assuming first-order deactivation kinetics, as
is usually done for catalytic reactions on sup-
ported metal NPs ( 9 ). This analysis allowed us
to calculate the catalyst deactivation coefficient,
kd(see supplementary text for derivation),
for different catalysts. Data in table S2
show the calculated values of the deactivation
coefficient.
To quantify the inherent kinetic rates for
the same catalysts, the initial forward rates of
propylene formation were obtained by analyz-
ing the data reported in the literature. Most
data are reported for an integral reactor design
(i.e., high propane conversion), so we performed
an integral reactor analysis to obtain an analytical
expression for the initial reaction rates (see sup-
plementary text for derivation). This analytical
expression was used to calculate the initial
forward rate of propylene production for various
catalysts (table S2). The product of the initial
forward reaction rate and the inverse of the
deactivation coefficient (proportional to catalyst
stability) provided a measure of the overall
catalyst productivity. Data in Fig. 2B show the
productivities of different catalysts operated
under a range of conditions. The data show
that the Pt 1 Sn 1 /SiO 2 catalyst reported here dem-
onstrated the highest productivity for propylene
production of all tested catalysts. For example,
Pt 1 Sn 1 /SiO 2 in the absence of hydrogen was
more productive than the commercial mimic
Pt-Sn/g-Al 2 O 3 by two orders of magnitude.
The overall catalyst performance of the Pt 1 Sn 1 /
SiO 2 catalysts was also studied at different
WHSVs (i.e., including those away from the
equilibrium conversion). Catalyst productivity
remained high, between 28 and 40 moles per
gram of catalyst, for different WHSVs (fig. S4
and table S2).
The data above show that the Pt 1 Sn 1 /SiO 2
catalysts exhibited very high productivity and
stability compared with other catalysts. Because
the dehydrogenation reaction is operated under
carbon-rich reducing conditions, it is inevitable
that, over time, catalyst deactivation due to
deposition of solid carbon would compromise
the performance of the catalysts. To investigate
whether this Pt 1 Sn 1 /SiO 2 catalyst can be regen-
erated, we intentionally increased the rate of
catalyst deactivation by increasing the propane
flow rate (i.e., propane WHSV) and diluting the
propane stream (which increases the propane
equilibrium conversion). These conditions push
the catalysts away from the equilibrium conver-
sion and lead to a more rapid, forced carbon-
induced deactivation. The data in fig. S5 show
that, under these conditions, the propane con-
version goes from ~41 to 39% in 120 min. Under
identical conditions, the other tested catalysts
deactivate very rapidly to very low conversion,
as illustrated for the commercial mimic. A
transmission electron microscopy (TEM) image
of the spent catalyst, after 120 min of PDH
reaction, is shown in fig. S6. The image shows
that the Pt-Sn NPs are well dispersed, with an
average particle diameter of ~1.6 ± 0.67 nm,which is consistent with the particle size in
fresh samples. Raman spectroscopy (fig. S7) on
the spent catalyst confirmed that the slight loss
in catalyst activity was due to coking. We at-
tempted to regenerate the Pt 1 Sn 1 /SiO 2 catalyst
by oxidation, using 1% O 2 as a mild oxidant at
500°C followed by reduction at 600°C (see
supplementary materials for details of the
regeneration procedure). The regeneration
studies showed that Pt 1 Sn 1 /SiO 2 catalysts can
be partially regenerated. For example, we mea-
sured ~97% of the initial activity and >99% of the
initial selectivity after three PDH/regeneration
cycles (fig. S5).
We hypothesized that the performance and
stability of the Pt 1 Sn 1 /SiO 2 catalyst was derived
from favorable interactions among the support
(SiO 2 ), promoter (Sn), and active metal (Pt),
which allowed for a high degree of atomic mix-
ingofPtandSnwithinverysmallintermetallic
PtSn NPs. Mixing of Pt and Sn atoms in NPs can
be compromised when these materials are in-
troduced on other oxide supports such as Al 2 O 3 ,
where Pt and Sn atoms segregate, leading to the
formation of Pt NPs and an Sn-oxide phase ( 54 ).
We studied the structure of the catalyst and its
precursors by using a combination of x-ray dif-
fraction (XRD), ultraviolet-visible (UV-Vis) spec-
troscopy, x-ray photoelectron spectroscopy (XPS),
diffuse reflectance infrared Fourier transform
spectroscopy (DRIFTS), TEM, and x-ray absorp-
tion spectroscopy [extended x-ray absorption
fine structure (EXAFS) and x-ray absorption
near-edge structure (XANES)].SCIENCEsciencemag.org 9JULY2021•VOL 373 ISSUE 6551 219
Fig. 2. Comparison of Pt 1 Sn 1 /SiO 2 with previously reported PDH
catalysts.(A) Conversion-selectivity plots for different PDH catalysts.
Numbers correspond to row numbers in table S1. (B) Catalyst productivity
(product of the initial reaction rate and the inverse of the deactivation
coefficient) for various PDH catalysts studied in the literature. Numbers
in brackets correspond to row numbers in table S2. Two data points
from this work are for two different conditions (with and without dilution
in the feed).RESEARCH | REPORTS