Lactic acid conversion to 2,3-pentanedione and acrylic acid over silica-supported sodium nitrate: Reaction optimization and identification of sodium lactate as the active catalyst

Article abstract of DOI:10.1006/jcat.1997.1484

Lactic acid is converted to 2,3-pentanedione, acrylic acid, and other products in vapor-phase reactions over silica-supported sodium lactate formed from sodium nitrate. Multiparameter optimization of reaction conditions using a Box-Benkhen experimental design shows that the highest yield and selectivity to 2,3-pentanedione are achieved at low temperature, elevated pressure, and long contact time, while yield and selectivity to acrylic acid are most favorable at high temperature, low pressure, and short contact time. Post-reaction Fourier transform infrared spectroscopic analyses of the catalyst indicate that sodium nitrate as the initial catalyst material is transformed to sodium lactate at the onset of reaction via proton transfer from lactic acid to nitrate. The resultant nitric acid vaporizes as it is formed, leaving sodium lactate as the sole sodium-bearing species on the catalyst during reaction.

Full text of DOI:10.1006/jcat.1997.1484

JOURNAL OF CATALYSIS 165, 162–171 (1997)
ARTICLE NO. CA971484
Lactic Acid Conversion to 2,3-Pentanedione and Acrylic Acid over
Silica-Supported Sodium Nitrate: Reaction Optimization and
Identification of Sodium Lactate as the Active Catalyst
,1
ꢀ
ꢀ
ꢀ
Douglas C. Wadley, Man S. Tam, Prashant B. Kokitkar,† James E. Jackson,† and Dennis J. Miller
ꢀ
Department of Chemical Engineering and †Department of Chemistry, Michigan State University, East Lansing, Michigan 48824
Received March 14, 1996; revised October 2, 1996; accepted October 7, 1996
The diketone is proposed to form via a Claisen condensa-
tion of two lactate moieties followed by decarboxylation
and dehydration steps.
Lactic acid is converted to 2,3-pentanedione, acrylic acid, and
other products in vapor-phase reactions over silica-supported
sodium lactate formed from sodium nitrate. Multiparameter op-
timization of reaction conditions using a Box–Benkhen experi-
mental design shows that the highest yield and selectivity to 2,3-
pentanedione are achieved at low temperature, elevated pressure,
and long contact time, while yield and selectivity to acrylic acid are
most favorable at high temperature, low pressure, and short contact
time. Post-reaction Fourier transform infrared spectroscopic analy-
ses of the catalyst indicate that sodium nitrate as the initial catalyst
material is transformed to sodium lactate at the onset of reaction
via proton transfer from lactic acid to nitrate. The resultant nitric
acid vaporizes as it is formed, leaving sodium lactate as the sole
c
We have also examined the species present on the surface
of sodium phosphate catalysts via 31P-MAS nuclear mag-
netic resonance (NMR) and post-reaction Fourier trans-
form infrared (FTIR) spectroscopies (8). These analyses in-
dicate that sodium lactate, formed via proton transfer from
lactic acid to phosphate, is a predominant species along with
phosphates on the support surface during reaction. We pro-
pose that both lactate formation and lactic acid conversion
are facilitated by the presence of a liquid phase on the sup-
port surface (8).
In this paper, the reaction of lactic acid to produce 2,3-
pentanedione and acrylic acid is investigated in detail using
sodium nitrate on silica as the initial catalyst material. We
have previously used sodium nitrate for 2,3-pentanedione
formation from lactic acid (7); nitrate has potential pro-
cessing advantages as a low-cost, environmentally benign
material. We present here optimized operating conditions
for the formation of 2,3-pentanedione and acrylic acid over
sodium nitrate, examine the transformation of sodium ni-
trate using previously established post-reaction FTIR ana-
lytical techniques, and provide initial catalyst stability and
coke deposition data for the system. A systematic experi-
mental design is used to arrive at optimum conditions for
conversion to desired products.
sodium-bearing species on the catalyst during reaction.
ꢁ 1997
Academic Press
INTRODUCTION
The production capacity for lactic acid (2-hydroxy-
propanoic acid) from biomass sources has increased
greatly in the U.S. during the past few years (1, 2), driven
primarily by the potential for producing biodegradable
polymers. This growing availability of lactic acid, coupled
with its reactive nature (containing both a hydroxyl and a
carboxyl group), has led us to examine it as a feedstock for
biomass and agriculturally based chemical production.
Primary lactic acid conversion pathways are shown in
Fig. 1. Dehydration to acrylic acid and cracking to acetalde-
hyde have been known for many years; the former has been
examined as a possible conversion route to acrylate poly-
mers and plastics formation (3–5).We recently discovered
the formation of 2,3-pentanedione from lactic acid over
supported phosphate catalysts in our laboratory (6), and
have examined the activity of a number of different cata-
lyst and support speciesfor 2,3-pentanedione formation (7).
METHODS
Materials. The feed used in all experiments is an aque-
ous solution of 34 wt% L-(-)-lactic acid (Purac, Inc.). This
concentration was chosen, based on previous studies, be-
cause it is high enough to facilitate significant conversion
without leading to plugging of the reactor inlet tube and ex-
cessive catalyst coking. High-purity helium (AGA, 99.99% )
is used as a diluent and to aid in the vaporization of lactic
acid.
1
To whom all correspondence should be addressed at Department of
The catalyst used in all experiments is sodium nitrate
(Aldrich) supported on 80–100 mesh Spherosil XOC-005
Chemical Engineering, A202 Engineering Building, Michigan State Uni-
versity, East Lansing, Michigan 48824. E-mail: millerd@che.msu.edu.
162
0021-9517/97 $25.00
c
Copyright ꢁ 1997 by Academic Press
All rights of reproduction in any form reserved.

LACTIC ACID CONVERSION OVER SODIUM NITRATE
163
der which reactions were carried out are summarized in
Table 1.
Optimization methods. Optimum operating conditions
for formation of acrylic acid and 2,3-pentanedione are
determined by a two-level, three-parameter factorial
design designated as a “Box–Benkhen” design (9). The
Box–Benkhen design provides an efficient method for
determining product yields over a range of defined reaction
conditions (i.e., independent variables) through a limited
number of experiments. Implementation of the design
involves defining “boundary values” (i.e., minima and
maxima) for all reaction parameters, conducting experi-
mental runs at conditions defined by the design, and fitting
the resulting product yields to quadratic model equations
describing their dependence on reaction conditions.
In this study, three reaction conditions were chosen as in-
dependent variables: temperature, contact time, and pres-
sure (x1, x2, x3). The range of these parameters over which
experiments were conducted can be thought of as occupy-
ing a cube, with the center (0, 0, 0) describing the base set
of conditions. Following the Box–Benkhen design, one ex-
periment was conducted at each set of conditions denoted
by the midpoints of the twelve cube edges, and four exper-
iments were conducted at conditions corresponding to the
center point. These experiments were conducted in random
order to avoid any systematic bias in the results. After ev-
ery three experiments, the reactor was shut down, cleaned,
and loaded with fresh catalyst in order to minimize the ef-
fects of coke buildup or catalyst deactivation on the results.
Additional experiments were also conducted to verify the
reliability of the results obtained and extend the study be-
yond the limits of the experimental design.
FIG. 1. Lactic acid conversion pathways.
porous silica (Anspec, Inc.; N2 BET surface area 14 m2/g).
The catalyst wasprepared at a loadingof1.0mmolNaNO3/g
support bywet impregnation and dryingovernight at 100ꢂC.
One batch of catalyst was used for all optimization experi-
ments; 2.0 ꢃ 0.1 g of catalyst was loaded in the reactor in a
given experiment.
Reaction system. The reaction system and product anal-
ysis techniques used in this study are the same as those
described in our earlier studies (6, 7). All reactions are car-
ried out in a vertical, down-flow, fixed-bed integral reac-
tor equipped with a quartz liner. In a typical experiment,
fresh catalyst was loaded into the reactor and heated un-
der helium to the initial reaction temperature. The cata-
lyst was then conditioned by exposing it to an elevated
flow rate (0.5 ml/min) of lactic acid feed solution for about
15 min (without this it took several hours to reach steady
state). The desired feed flow rate for the experiment was
then established (0.05–0.5 ml/min) and steady state prod-
uct collection was begun. Each catalyst was typically tested
under three to five sets of reaction conditions before the
reactor was shut down, cleaned, and reloaded with fresh
catalyst. At a given set of conditions, condensible prod-
ucts and noncondensible products were collected and an-
alyzed separately by gas chromatography. Results of the
analyses were entered into a spreadsheet to calculate prod-
uct yields, selectivities, feed conversion, and an overall
carbon balance for the experiment. Product yield is re-
ported as percentage of the theoretical yield based on lactic
acid fed to the reactor. The experimental conditions un-
In order to identify trends in product yields and optimum
reaction conditions, the experimental yield (Y) of acrylic
acid and of 2,3-pentanedione were fit to a quadratic model
of the form
Y = b0 + b1x1 + b2x2 + b3x3 + b12x1x2 + b13x1x3
+ b23x2x3 + b11x12 + b22x22 + b33x32,
[1]
TABLE 1
Reaction Conditions
Box–Benkhen design values
Parameter
Range
ꢄ1
0
1
Temperature (ꢂC)
Pressure (MPa)
Contact time (s)
Feed composition
(mole fraction)
265–370
0.2–1.0
0.3–5.0
280
0.2
0.5
Lactic acid: 0.08
Water: 0.77
Helium: 0.15
2.0 ꢃ 0.1
315
0.6
2.2
350
1.0
4.0
Catalyst mass (g)

164
WADLEY ET AL.
where x1, x2, and x3 are again the three independent vari-
ables temperature, contact time, and pressure. The coeffi-
cients (bi , bi j ) in the model equation were determined by
regression analysis using the Linest function in Microsoft
Excel 5.0. Contour plots were then produced from Eq. [1]
as a function of process conditions to illustrate trends and
identify optimum conditions for each of the two desired
products.
FTIR spectroscopic analyses. Post-reaction Fourier
transform infrared (FTIR) spectroscopy was used to inves-
tigate the nature of stable species present on the catalyst
and to gain insight into the role of the nitrate catalyst and
the mechanism of product formation. These studies were
performed using equipment and methods similar to those
previously described in (8). Sodium nitrate solution (100 ꢀl,
either 5 mmol/liter or 50 mmol/liter) was deposited on both
sides of an IR-transparent silicon disk using a micropipet.
Water was then removed at ambient temperature under
vacuum in a desiccator, leaving a thin layer of NaNO3 on
the disk as a model catalyst-support system. The loaded
disk was then placed in the IR sample preparation appara-
tus (8) and exposed to lactic acid vapors for (in most cases)
10 min at a specified temperature. Following exposure to
lactic acid, the disk was transferred immediately to a sam-
ple holder and placed in a Nicolet IR/42 spectrophotome-
ter wherein the transmission FTIR spectrum was obtained.
Subtraction of background and silicon disk spectra gives a
spectrum representative of the material on the catalyst sur-
face following reaction at the specified temperature. The
entire procedure, starting with the clean silicon disk, was
repeated for each temperature studied. This technique fa-
cilitates rapid and reliable collection of post-reaction IR
spectra; the method gives essentially the same spectra as
those taken by diffuse reflectance IR spectroscopy of actual
supported catalysts (8) and thus is a valid analytical tool.
FIG. 2. Contour plot of 2,3-pentanedione yield from regression anal-
ysis of experimental data at 0.6 MPa.
for the fits of Eq. (1) to acrylic acid and 2,3-pentanedione
yields are 0.81 and 0.83, respectively; the confidence level
of the fit, determined by a standard F-test (10), is 90% for
2,3-pentanedione and 98% for acrylic acid.
Because maximum product yields were predicted to
occur at conditions along the edge of the design ma-
trix, additional experiments were conducted outside the
experimental design range to determine if the above results
represented true optima. Results of these experiments are
given in Table 3 and are described below. Table 3 also in-
cludes results of repeated experiments used to verify prod-
uct yields.
Yield of 2,3-pentanedione drops as temperature is
decreased from 280 to 265ꢂC (Exp. 28b) and as pressure
RESULTS
Optimization of reaction conditions. Complete results
of experiments conducted within the Box–Benkhen design
are given in Table 2 in the order they were carried out, with
each number (e.g., 18) signifying a particular charge of cata-
lyst to the reactor. Contour plots of 2,3-pentanedione and
acrylic acid yields versus temperature and contact time at
each pressure studied were generated following regression
analysis of the results; representative plots are presented in
Fig. 2 for 2,3-pentanedione at 0.6 MPa pressure and in Fig. 3
for acrylic acid at 0.2 MPa. The contour surfaces are gen-
erally in the shape of saddles; the highest 2,3-pentanedione
yield predicted within the experimental design matrix is
18% at 280ꢂC, 0.6 MPa, and 4 s contact time, and the high-
est acrylicacid yield predicted is18% at 350ꢂC, 0.2MPa, and
0.5 s contact time. Plots at other pressures are qualitatively
FIG. 3. Contour plot of acrylic acid yield from regression analysis of
similar but yields are lower. The regression coefficients (r2) experimental data at 0.2 MPa.

LACTIC ACID CONVERSION OVER SODIUM NITRATE
165

166
WADLEY ET AL.
is increased from 0.6 to 0.8 MPa (Exps. 24b, 28c). Increas- acrylic acid yield of about 14% of theoretical, very little
ing contact time from 4 to 5 s at 265ꢂC enhances 2,3- coking of the catalyst or conversion of acrylic acid, either
pentanedione yield (Exp. 27c, 27d); unfortunately, we were fed alone or in combination with lactic acid, acetaldehyde,
unable to lengthen contact time beyond 5 s at 265ꢂC in our or 2,3-pentanedione, took place over the range 280–350ꢂC
reactor system as a consequence of bed height and mini- and contact times to 5 sec. We thus conclude that secondary
mum liquid feed rate restrictions. We thus believe 280ꢂC reaction of acrylic acid is important only at high acrylic acid
and 0.6 MPa to be optimal conditions, but higher 2,3- formation rates and product concentrations, and may be a
pentanedione yields may be achievable at residence times significant source of coking at these conditions.
longer than 4–5 s at 280ꢂC.
Application of kinetic model. In our previous work (6),
we developed a simple kinetic model and calculated rate
For acrylic acid formation, increasing temperature to
370ꢂC significantly increases acrylic acid yield at short resi-
constants for lactic acid conversion in our integral fixed-
dence times (Exps. 27a, 27b, 28a); longer residence times at
bed catalytic reactor. The molar balances for the three ma-
370ꢂC (Exp. 26a) lead to a decline in acrylic acid yield as a
jor species are written as follows, with concentrations of
result of an apparent secondary reaction to propanoic acid.
lactic acid, 2,3-pentanedione, and acrylic acid denoted by
CL, CP, and CA respectively, and residence time by ꢁ.
We did not examine temperatures above 370ꢂC, because
we cannot achieve residence times of less than 0.3 s in our
reactor and we expect that secondary reactions will reduce
acrylic acid yields at 0.3 s residence time and temperatures
above 370ꢂC. Optimum conditions for acrylic acid forma-
tion, based on these additional experiments, are 370ꢂC, 0.3–
0.5 s contact time, and low pressure (0.1–0.4 Mpa). It is pos-
sible, however, that other combinations of temperature and
short residence time may lead to higher acrylic acid yields
than those achieved here.
dCL
= ꢄ k1CL ꢄ 2k2CL2 ꢄ k3CL
= k2CL2 ꢄ k5CP
[2]
[3]
[4]
dꢁ
dCP
dꢁ
dCA
= k3CL ꢄ k4CA
dꢁ
The data in Table 3 were combined with those in Table 2
and a second regression analysis was performed to develop
contour plots based on all experiments performed. The con-
tour plots obtained were essentially the same as those re-
sulting from regression of the formal Box–Benkhen design
only (Figs. 2 and 3) and are thus not reported.
The balancesinclude a first-order rate expression for forma-
tion of acrylic acid, a first-order rate expression for forma-
tion of acetaldehyde plus propanoic acid, hydroxyacetone,
and “other” products together, and a second-order rate ex-
pression for formation of 2,3-pentanedione. The model also
includes first-order decomposition pathways for acrylic acid
and 2,3-pentanedione. The model was applied to the results
of this study (Tables 2 and 3) by simultaneously varying the
preexponential factors (ki,o) and activation energies (Ei )
of rate constants k1 through k5 (in the standard Arrhenius
form) for the five reaction steps using Excel 5.0 to minimize
the sum-of-squares residuals of species’ exit concentrations
and lactic acid conversion for all experimental runs. The re-
sulting rate constants for each reaction are given in Table 4.
Turnover frequencies (TOF) for each reaction step are also
given in Table 4 at the centerpoint conditions (0, 0, 0) and
at 50% lactic acid conversion in the reactor. The TOF cal-
culation is based on a catalyst loading of 1.0 mmol/g; we
have previously shown that mass transport limitations do
not affect reaction rates at these reaction conditions (6).
Secondary reactions of acrylic acid and 2,3-pentanedione.
To further investigate secondary reaction pathways
leading to the observed maxima in acrylic acid and
2,3-pentanedione yields with contact time, we conducted
a series of experiments at 0.6 MPa using acrylic acid and
2,3-pentanedione as feed materials.
Over fresh nitrate catalyst or over catalyst previously
used for lactic acid conversion, no significant reaction of a
5 wt% solution of 2,3-pentanedione, either alone or in com-
bination with similar quantities of acetaldehyde or acrylic
acid, was observed over the range 280–350ꢂC and contact
times up to 5 sec. When co-fed with lactic acid over the
used catalyst at 350ꢂC and 5 s contact time, however, less
2,3-pentanedione exited the reactor than was expected as
the sum of the quantity formed via lactic acid conversion
Extended reaction. Stability of the nitrate catalyst for
plus the quantity (5 wt% ) added to the feed. We thus con- acrylic acid and 2,3-pentanedione formation was tested by
clude that secondary reaction of 2,3-pentanedione does conducting an extended run for a 24 hr period at 350ꢂC,
take place in the presence of lactic acid and at high tem- 0.6 Mpa, and 1.0 s contact time. These conditions were cho-
peratures.
sen to give reasonably high yields of both acrylic acid and
For acrylic acid, a feed of 17 wt% acid in water initially 2,3-pentanedione. Average yield for acrylic acid was 24% ,
pumped into the reactor led to coking and eventual plug- for 2,3-pentanedione it was 14% , for acetaldehyde 9% , and
ging of the catalyst bed, indicating that acrylic acid decom- for propanoic acid 4% ; the average lactic acid conversion
poses or polymerizes over the catalyst. When the feed con- was about 88% . Product yields and conversion were es-
centration was reduced to 4 wt% , which corresponds to an sentially constant over the 24 hr experiment, indicating

LACTIC ACID CONVERSION OVER SODIUM NITRATE
167
TABLE 4
Rate Constants from Kinetic Model
Activation
energy
(kJ/mol)
Reaction
order
Preexponential factor
TOFa
(sꢄ1
Reaction
(ki,o
)
)
Lactic acid → Acetaldehyde + Others
Lactic acid → 2,3-Pentanedione
Lactic acid → Acrylic Acid
1
2
1
1
1
9.7 ꢅ 109 sꢄ1
115
110
137
132
138
5.7 ꢅ 10ꢄ3
1.1 ꢅ 10ꢄ3
1.2 ꢅ 10ꢄ3
1.9 ꢅ 10ꢄ4
1.1 ꢅ 10ꢄ4
6.5 ꢅ 1010 liter molꢄ1 sꢄ1
1.9 ꢅ 1011 sꢄ1
Acrylic acid → Products
7.2 ꢅ 1010 sꢄ1
2,3-Pentanedione → Products
2.0 ꢅ 1011 sꢄ1
a Turnover frequency determined at centerpoint conditions (315ꢂC, 0.6 MPa) at 50% lactic acid conver-
sion in reactor.
that the catalyst is stable and remains active at these would lead to a higher carbon content.) Carbon recovery
conditions.
in products exiting the reactor was measured by averaging
each individual run weighted to the quantity of feed input
during that run. It is seen that solids deposition in most
experiments accounts for most of the carbon absent from
the recovered products. It is interesting that the quantity
of coke deposited during the extended (24 hr) run is only
slightly higher than the average quantity deposited during
the typical experiment of 4–5 hr duration, suggesting that
coke deposition slows over extended reaction times.
Coke deposition. Weight gain of the catalyst and reac-
tor tube was measured upon reactor disassembly at the end
of each experiment. Coke deposits were consistently ob-
served on the reactor wallsabove the catalyst, in the greatest
amount at the top of the catalyst bed, and to a lesser degree
through the length of the bed. We attribute this coke de-
position to feed lactic acid and its oligomers, both of which
have low volatility and are thus prone to decomposition
prior to vaporization.
FTIR analyses. Results of post-reaction transmission
FTIR analyses are given in Figs. 4–7: only the portion of the
IR spectrum from 700 to 2000 cmꢄ1 is reported, as the most
important spectral information is contained within this re-
gion. To verify thermal stability of the nitrate catalyst over
the temperature range of interest, FTIR spectra of sodium
nitrate deposited on the silicon disk (100 ꢀl of 5 mmol/liter
solution) and then exposed only to water vapor at differ-
ent temperatures were collected and are reported in Fig. 4.
Key features of the sodium nitrate spectrum include sharp
peaks at 836 and 1788 cmꢄ1, a broad, strong peak centered
at 1355 cmꢄ1, and a weaker band at 2430 cmꢄ1 (not shown).
These absorption bands are in agreement with those re-
ported in the literature for sodium nitrate (11–14). Figure 4
clearly shows that sodium nitrate is thermally stable over
the range of temperatures studied.
The quantity of coke deposited and its contribution to
the overall carbon balance is reported in Table 5. For the
purpose of conducting the carbon balance, we assume the
coke deposited contains 50% carbon by weight. (This is
reasonable and conservative; polylactic acid contains 50%
carbon by weight and further decomposition of the polymer
TABLE 5
Summary of Carbon Deposition and Carbon Recovery
Carbon
depositeda
over
experiment
(g)
Average
carbon recoveryb
in products
Carbon depositedc
over experiment
over experiment
(percent carbon fed) (percent carbon fed)
Experiment
Spectra for sodium nitrate exposed to lactic acid vapors
at different temperatures were collected at two different
nitrate loadings on the disk: 100 ꢀl of 5 mmol/liter solution
(Fig. 5) and 100 ꢀl of 50 mmol/liter solution (Fig. 6) Figure 5
shows that no nitrate is observable on the disk at low load-
ings at any of the temperatures studied. Sodium lactate is
observed at all temperatures, as evidenced by a strong peak
18
20
21
22
26
27
28
1.54
1.05
1.98
1.76
1.87
0.84
1.30
1.81
77
67
77
62
82
69
85
86
16
15
33
25
26
7
14
3
Extended run
at 1599 cmꢄ1 (C O stretch) and several weaker bands at
==
a
1050–1424 cmꢄ1 (15, 16). Some lactic acid is also present, ꢄas1
Absolute quantity of solids deposited; catalyst weight 2.0 ꢃ 0.1 g in
each experiment.
==
seen by its strong C O stretching frequency at 1730 cm
b
Based on product recovery from each run weighted to total feed
(15, 16). Reference spectra for lactate species are not given
here, but have been collected by us (8) and agree with the
literature spectra (15, 16).
volume during run.
c Solids deposited on catalyst and reactor walls assumed to contain 50%
carbon by weight.

168
WADLEY ET AL.
2,3-pentanedione and acetaldehyde plus other species for-
mation are somewhat (ꢆ30 kJ/mol) higher than those calcu-
lated for phosphates. Turnover frequencies calculated here
are based on the assumption that all catalyst is participat-
ing in the reaction, even though sufficient sodium lactate
is present to form 20–30 monolayers on the silica surface.
The IR results clearly show that all nitrate initially present
is converted to lactate, and our postulate that the catalyst
is present as a molten phase on the catalyst surface (see
below) is consistent with all catalyst being taken as active.
The magnitude of the turnover frequencies are low; TOF
for product decomposition pathways are of the order of
15–20% of the corresponding product formation pathways
at the conditions specified.
The kinetic model predicts product yields and conver-
sions reasonably well over the entire range of reaction
conditions, but the experimental results at 1.0 MPa do
not match model predictions as well as those at lower
pressures. First, the experimentally observed yields of 2,3-
pentanedione increase significantly from 0.2 to 0.6 MPa
total pressure, but little further increase is observed as
pressure is raised to 1.0 MPa. The kinetic model predicts
that 2,3-pentanedione formation should increase with total
FIG. 4. FTIR spectra of species present on sodium nitrate-loaded sil-
icon disk (100 ꢀl of 0.005 M NaNO3 solution) following exposure to water
vapor at different temperatures.
Figure 6 gives IR spectra for the silicon disk carrying 10
times the quantity of nitrate as used for spectra reported in
Fig. 5. In Fig. 6, sodium nitrate is seen to remain on the disk
at temperatures up to 300ꢂC following a 10-min exposure
to lactic acid. Above 300ꢂC, nitrate disappears and sodium
lactate is observed in large quantities with little lactic acid
present. These results with more nitrate on the support disk
suggest that the consumption of nitrate is kinetically lim-
ited on the disk surface. Exposure of the disk at 300ꢂC for
longer time periods shows that the nitrate peak disappears
(Fig. 7), further indication of the kinetic limitations of pro-
ton transfer.
DISCUSSION
The kinetic model (Table 4) predicts the favorability
of acrylic acid formation at high temperatures, short con-
tact times, and low pressures, and the favorability of 2,3-
pentanedione at low temperature, long contact times, and
elevated pressure. The activation energy calculated for
acrylic acid formation is the same as the calculated pre-
FIG. 5. FTIR spectra of species present on sodium nitrate-loaded sil-
icon disk (100 ꢀl of 0.005 M NaNO3 solution) following exposure to lactic
viously for phosphate salts (6); the activation energies for acid at different temperatures.

LACTIC ACID CONVERSION OVER SODIUM NITRATE
169
thus additional acetaldehyde is formed at higher pressures
by decomposition of unvaporized lactic acid which has en-
tered the reactor. For 2,3-pentanedione, yield does not in-
crease as pressure is increased from 0.6 to 1.0 MPa because
a greater fraction of the entering lactic acid decomposes
prior to exposure to the catalyst at 1.0 MPa and is thus
unavailable for diketone formation.
Lactic acid decomposition prior to vaporization is also
responsible for most of the coke deposited during reaction,
as evidenced by the majority of coke on the reactor walls
above and directly on the top of the catalyst bed. As seen in
Table 5, carbon balances are close for the most part when
coking is taken into account (the two right-hand columns
add to 100% at closure), especially recognizing that lac-
tic acid analysis by gas chromatography is only accurate to
within ꢃ14% (6). Even for those experiments where car-
bon balance closure is still incomplete, we have shown pre-
viously that product distributions depend only slightly on
carbon recoveries (6, 7) and thus have confidence that prod-
uct yields reported are a true representation of the reaction
system. It is unfortunate that the contribution of coking to
the carbon balance and the overall reaction system cannot
be treated any further than was reported in Table 5: because
FIG. 6. FTIR spectra of species present on sodium nitrate-loaded sil-
icon disk (100 ꢀl of 0.050 M NaNO3 solution) following exposure to lactic
acid at different temperatures.
pressure, as expected for a second-order reaction, although
the predicted increase is small at high temperatures and
long contact times because of competing reaction pathways
and depletion of reactant.
Second, trends in acetaldehyde yield deviate from trends
predicted by the model in several sets of experiments (19b,
21a and 19a, 20a) where only total pressure is changed from
0.2 to 1.0 MPa. In these experiments the yield of acrylic
acid remains essentially constant (as expected for a first-
order dominated reaction), but the yield of acetaldehyde
(also formed via a first-order pathway) increases signifi-
cantly and corresponding CO formation increases as well.
It thus appears that lactic acid decomposes to acetaldehyde
via a pathway not present in the kinetic model.
We attribute these differences in experimental and pre-
dicted yields at higher pressures to the increased difficulty
in vaporizing lactic acid as it enters the reactor at high pres-
sure. Lactic acid vaporizes completely only at reduced pres-
sure and is known to dimerize (and thus become essentially
nonvolatile) along with vaporizing during heating (17, 18).
Further, in a previousstudy(19) we have observed acetalde-
FIG. 7. FTIR spectra of species present on sodium nitrate-loaded sil-
icon disk (100 ꢀl of 0.050 M NaNO3 solution) following exposure to lactic
hyde formation in condensed-phase pyrolysis of lactic acid; acid at 300ꢂC for different lengths of time.

170
WADLEY ET AL.
coke deposition takes place for the most part above the
catalyst bed, there is no way to incorporate it as a reaction
into the ideal, plug-flow reactor model used to generate
the kinetic constants. The same holds true for acetaldehyde
formed via condensed lactic acid decomposition at higher
pressures.
Finally, the extended run shows that the catalyst is stable
for formation of 2,3-pentanedione and acrylic acid even
upon substantial coke deposition over the course of the
experiment. Apparently there is not enough coke deposited
in the catalyst bed to substantially deactivate the sodium
lactate over the time scale of the experiment conducted. It
may be that the reactions occur on the coke as well as on
the silica support, but based on earlier studies with carbon
supports (7), where we observed little 2,3-pentanedione or
acrylic acid formation, we do not believe that this is the case.
FTIR experiments. Post-reaction FTIR transmission
spectroscopic analyses of model catalyst surfaces show that
sodium nitrate is consumed upon exposure to lactic acid.
The absence of nitrate and presence of sodium lactate at
high temperatures indicates that lactic acid transfers a pro-
ton to sodium nitrate to form nitric acid and sodium lactate:
ꢄ→
NaNO3 + HC3H5O3 ꢇꢄ HNO3 + NaC3H5O3 [5]
Nitric acid has a normal boiling point of 80ꢂC; it also forms
a binary azeotrope with water that boils at 120ꢂC (at at-
mospheric pressure) and thus vaporizes as it is formed on
the support. The proton transfer reaction can therefore
proceed to completion, despite the unfavorable pKa dif-
ference between nitric acid (pKa = ꢄ1.4) and lactic acid
(pKa = 3.1). The dominant, stable species on the catalyst
surface is sodium lactate, which must play a primary role in
2,3-pentanedione and acrylic acid formation.
The proton transfer in Eq. [5] agrees with our findings
over phosphate salts (8), in which lactic acid transferred
a proton to trisodium phosphate to form sodium lactate
and disodium phosphate, which subsequently condensed
to tetrasodium pyrophosphate. Based on the nitrate results
presented here, in which higher yields of 2,3-pentanedione
and acrylic acid are achieved than with the phosphate salt,
and that only sodium lactate is observable, we conclude that
the phosphate played no role of consequence in the forma-
tion of desirable products. In fact, silica-supported sodium
lactate itself is as effective a catalyst as sodium nitrate for
lactic acid conversion to 2,3-pentanedione. Thus, the alkali
salt catalysts active for lactic acid conversions only serve
as sources of alkali metal cations which form lactate salts
upon exposure to lactic acid. The conversion of lactic acid
to acrylic acid and 2,3-pentanedione takes place upon in-
teraction of lactic acid with sodium lactate.
FIG. 8. FTIR spectra of species present on sodium nitrate-loaded
silicon disk (100 ꢀl of 0.005 M NaNO3 solution) following exposure to
propanoic acid at different temperatures.
weaker propanoic acid (pKa = 4.9) was passed over a
sodium nitrate-loaded silicon disk. The FTIR spectra as a
function of temperature are given in Fig. 8; formation of
sodium propanoate along with the disappearance of nitrate
is clearly seen at 300ꢂC and above.
We are continuing investigations of the mechanism of
2,3-pentanedione and acrylic acid formation over alkali
metal salt catalysts. We now know that proton transfer to
the salt from lactic acid occurs on the support surface; Figs. 4
and 5showthat thistransfer isnot necessarilyrapid and may
take substantial exposure time at reaction temperature to
reach completion. We postulate the presence of a liquid
phase on the support that expedites proton transfer as well
as condensation to 2,3-pentanedione and dehydration to
acrylic acid.
ACKNOWLEDGMENTS
The support of USDA Grant 93-37500-9585, under the National Re-
search Initiative, is greatly appreciated. Support of the Crop and Food
Bioprocessing Center by the State of Michigan Research Excellence Fund
The accessibility of the proton transfer reaction (de-
spite the unfavorable pKa difference between nitric and
lactic acid) is confirmed by an experiment in which the is also gratefully acknowledged.

LACTIC ACID CONVERSION OVER SODIUM NITRATE
171
REFERENCES
10. Dixon, W. J., and Massey, F. J., Jr., “Introduction to Statistical Analy-
sis,” McGraw–Hill, New York, 1957.
1. Lipinsky, E. S., and Sinclair, R. G., Chem. Eng. Prog. 82(8), 25 (1986). 11. Vratny, F., Appl. Spectrosc. 13(3), 59 (1959).
2. Keeler, R., Res. Dev., 52 (February 1991). 12. Ferraro, J. R., J. Molec. Spectrosc. 4, 99 (1960).
3. Fisher, C. H., and Filachione, E. M., U.S. Department of Agriculture, 13. Williamson, K., Li, P., and Devlin, J. P., J. Chem. Phys. 48, 3891
Bur. Agr. and Ind. Chem. AIC-279 (1950).
4. Holman, R. E., U.S. Patent 2,859,240 (1958).
5. Sawicki, R. A., U.S. Patent 4,729,978 (1988).
6. Gunter, G. C., Jackson, J. E., and Miller, D. J., J. Catal. 148, 252
(1994).
7. Gunter, G. C., Langford, R. H., Jackson, J. E., and Miller, D. J., Ind.
Eng. Chem. Res. 34, 974 (1995).
8. Gunter, G. C., Tam, M. S., Craciun, R., Jackson, J. E., and Miller, D.
J., J. Catal., 164, 207 (1996).
(1968).
14. Nakagawa, I., and Walter, J. L., J. Chem. Phys. 51, 1389 (1969).
15. Cassanas, G., Morssli, M., Fabregue, E., and Bardet, L., J. Raman
Spectrosc. 22, 409 (1991).
16. Simons, W. M. (Ed.), “The Infrared Spectra Handbook of Inorganic
Compounds,” Sadtler Research Laboratories, Philadelphia, PA, 1984.
17. Nielsen, J. I., and Veibel, M., Acta Polytech. Scand. Chem. Meatall. Ser.
63 (1967).
18. Mok, W., Antal, M. J., Jr., and Jones, M., J. Org. Chem. 54, 4596
(1989).
19. Gunter, G. C., “Catalytic Upgrading of Lactic Acid over Supported
Salt Catalysts,” Ph.D. dissertation. Michigan State University, 1994.
9. Box, G. E. P., Hunter, W. G., and Hunter, W. S., “Statistics for Ex-
perimenters: An Introduction to Design, Data Analysis, and Model
Building,” Wiley, New York, 1978.