3.. 4
.ﬁgrni
$ — 4a.".r-l
23h...“ .
- . .1 at}???
a? $4 . .
x ﬁle-V1»: . om

 

1’
. Itclx {la
“.22”? \vu..’1hl3!.ai.
. 9‘... v1.3... .124... a)

tale: u.

a...
it!!!" ..
Clo‘n 'V“. ."

 

1.

:ktttevsl I.
(:3. 5m”!

v! 53:”; .t-v'afxdh I
iii iv}... ‘3 5 :
\Vv.t‘l‘lvr)}1!§l3
VI

 

.. xv c!

. $1135 0) .
:5: {7‘1
.12.: vvt. ..

E a ‘ a:
; . . i... , 3%.»
A t... 1.....1 . . . .
.. . , 2. .6.
1.... .3 :i

 

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the
dissertation entitled

STUDIES IN THE HYDROGENATION OF BIORENEWABLE
FEEDSTOCKS

presented by

Ketan Pradeep Pimparkar

has been accepted towards fulﬁllment
of the requirements for the

Ph. D. degree in Chemical Engineering

 

 

Major Vrofessor‘s Signature

2/27/09

Date

MSU is an Afﬁnnative ActionEqual Opportunity Employer

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K‘lProlecc8-Pres/ClRC/DaleDue indd

 

STUDIES IN THE HYDROGENATION OF BIORENEWABLE FEEDSTOCKS
By

Ketan Pradeep Pimparkar

A DISSERTATION

Submitted to
Michigan State University
In partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Chemical Engineering

2009

ABSTRACT
STUDIES IN THE HYDROGENATION OF BIORENEWABLE FEEDSTOCKS
By

Ketan Pradeep Pimparkar

Proteins and amino acids are studied as feedstocks for catalytic hydrogenation to
value added chemicals. A literature survey is presented describing the extreme conditions
required for hydrogenation of the carbonyl ﬁmctionality in such feedstocks. Three
approaches tried out include hydrogenation of the peptide bond in dipeptides,
hydrogenation of hydrolyzed proteins (oligopeptides) to polyaminoalcohols, and
hydrogenation of amino acid mixtures. Challenges and interesting results obtained are
discussed. In particular, the aqueous phase hydrogenation of amino acid mixtures was
found to follow Langmuir-Hinshelwood type behaviour, with competitive adsorption
between the amino acids for active sites on the catalyst surface.

Mass transfer resistances often play a limiting role in gas-liquid-solid reactions.
Catalytic hydrogenations are such a class of reactions where high operating pressures and
mechanical stirring are used to overcome these resistances. A novel PDMS based high
pressure mieroreactor system was developed to enable the study of intrinsic kinetics of
hydrogenation reactions under conditions where hydrogen is not the limiting reactant. A
technique to immobilize catalysts within the microreactors was developed and used to
study the aqueous phase hydrogenation of acetone. A kinetic model describes the results

observed and sheds further light on the reaction mechanism.

Copyright by
KETAN PRADEEP PIMPARKAR
2009

Dedicated to my parents

iv

ACKNOWLEDGEMENTS

I would like to express my gratitude to my advisor, Professor Dr. Dennis J. Miller,
for his encouragement, guidance and support during my Ph.D. I would like to thank the
members of my Doctoral Guidance Committee, Professor Christina Chan, Professor
Mark Worden, Professor James E. Jackson and Professor Robert Ofoli for their
objectivity and appreciation of my work.

The generous ﬁnancial support of the US. Department of Energy through the

Consortium for Plant Biotechnology Research, of Procter and Gamble, and of the State of

Michigan Economic Development Corporation through the 21$t Century Jobs Fund

Program is gratefully acknowledged.

I am thankful to Professor G. D. Yadav and Professor Krishnamurthy Jayaraman
for the opportunity to study at Michigan State University.

To members of my lab, I am grateﬁil for their company, advice and help. Lars,
Navin, Aspi, YuQing, Yaoyan, Xi, Kausik, Sanket, Sai, Ambareesh, Alvaro have been
great colleagues.

The constant support and ancouragement from my parents has played a huge role
in making this thesis possible.

To residents of Blandings Castle and my friends, I owe the diversionary humor so

vital to making a tolerable journey enjoyable.

Table of Contents

List of Tables .........................................................................
List of Figures ........................................................................
Key to Symbols .......................................................................

Chapter 1. Introduction
1.1 Background ........................................................
1.2 Signiﬁcance .......................................................

Chapter 2. Hydrogenation of Amides
2.1 Amines - Industrial Uses and Market Size ...................
2.2 Hydrogenation of Amides to Amines - Literature Survey.
2.3 Exploratory Studies in the Reduction of Amides ............
2.4 Protein Hydrolysis and Hydrogenation ......................
2.4.1 Hydrolysis and Reduction Using H3PO4 ,,,,,,,,,,,,,,,,,,,,,,,
2.4.2 Hydrolysis Using HCl Followed by Hydrogenation
in a Parr Reactor .......................................
2.4.3 Enzymatic Hydrolysis using Trypsin and
Subsequent Hydrogenation ..............................
2.4.4 Results .....................................................
2.5 Summary ...........................................................

Chapter 3. Hydrogenation of Amino Acid Mixtures

3.1 Introduction .......................................................

3.2 Hydrogenation of Organic Acids and their Esters —
Literature Survey ..................................................

3.3 Materials and Methods ...........................................

3.4 Results and Discussion ...........................................
3.4.1 Hydrogenation of Individual Amino Acids ............
3.4.2 Hydrogenation of Amino Acid Mixtures ...............
3.4.3 Analysis of Mass Transfer Resistances ................
3.4.4 Kinetic Model .............................................

3.5 Conclusions ........................................................

Chapter 4. PDMS-glass Microreactor
4.1 Introduction .........................................................
4.2 Immobilization of Catalytic Nanoparticles .....................
4.3 PDMS-Glass Microreactors .....................................
4.4 Design of System - Modeling and Preliminary Calculations
4.5 Design of System — Flow loop and System Constraints. .

vi

Page
no.
viii

ix

xiv

~

12
15
16
18

18

19
2O

24
24
25

34
37
37
41
43
45
51

57
57
58
6O
61
63

4.6 Development of the Flow Loop .................................
4.7 Flow Loop: Connections Between the Microreactor and
the HPLC Tubing .................................................
4.8 Catalyst Immobilization ..........................................
4.9 Choice of Substrate ...............................................
4.10 Results and Discussion .........................................
4.11 Attempts at Increasing Catalyst Loading ....................
4.12 Conclusion ......................................................

Chapter 5. PDMS Tubing as Microreactor .......................................
5.1 Development of Immobilization Technique ...................
5.2 Immobilization Procedure .......................................
5.3 Methods ............................................................
5.4 Results and Discussion ..........................................

5.4.1 Results From Tubing 1 ....................................
5.4.1.1 Comparison of Activity of Ru/C and Pd/C. ..
5.4.1.2 Effect of Temperarure ...........................
5.4.1.3 Effect of Catalyst Loading .....................
5.4.1.4 Effect of Acetone Concentration ..............
5.4.1.5 Kinetic Model ....................................

5.4.2 Results From Tubing 2 ..................................
5.4.2.1 Effect of Acetone Concentration .............
5.4.2.2 Kinetic Model ....................................
5.4.2.3 Mass Transfer Analysis .........................

5.5 Conclusion ........................................................

vii

64
66

68
69
70
72
73

75
75
75
80
83
83
83
84
85
86
87
95
96
99
101
105

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Table 7

Table 8

Table 9

Table 10

Table 11

Table 12

Table 13

Table 14

Table 15

List of Tables

Title
General characteristics of various natural feed ingredients .............. 3
Results from studies of transamidation reactions ............................. 15
Solubilities of amino acids in water at 25°C ................................. 36
Kinetic constants for amino acid hydrogenation at 403 K .................. 49

Results of Lactic acid hydrogenation using various supported
nanocatalysts, courtesy Dr. Mukhopadhyay (PG= propylene glycol, Ru
= ruthenium, APTS = amino propyl trimethyl silane, Y = zeolite Y,

M4] = MCM 41, Ru/C = ruthenium supported on
carbon) ....................................................... 59
Summary of results from hydrogenation runs using Pd wire immobilized
in PDMS-glass mieroreactor ............ . ....................................... 71
Charactersitics of catalytic tubing prepared and used in reactions. All
tubing were PDMS with an id. of 800 um ................................... 83
Comparison of the activitiy of Pd/C and Ru/C towards the
hydrogenation of acetone ................................................... 84

Increase in reaction rate with increase in temperature, all data using

Tubing 1 ..................................................................................................... 84
Catalytic effect of tubing — conversion changes with amount of

catalyst .............................................................................. 86
Effect of change in acetone concentration on reaction rate .................. 87
Initial rates as affected by initial concentration of acetone .................. 95

Parameter values giving best ﬁt for the experimental data from Tubing l 96
Parameters giving best ﬁt for data from Tubing 2 ........................... 100

Mass tansfer analysis and observable modulii values for acetone
hydrogenation .................................................................... 104

viii

Figure 1.

Figure 2.

Figure 3.
Figure 4.
Figure 5.
Figure 6.
Figure 7

Figure 8

Figure 9a.
Figure 9b.
Figure 10.

Figure 1 1.

Figure 12.

Figure 13.

Figure 14.

Figure 15.

List of Figures
Caption

Condensation reaction between two amino acids to form a peptide

linkage ......................................................................... 1
The large variety of transformations of Glutamic acid which can yield
useful products. Figure taken from PNNL report5 ....................... 2
Reduction of amides to amines .............................................. 10
Side reactions in the reduction of amides to amines ...................... 10
Examples of polypeptides .................................................... 12
Transamidation reaction of glycylglycine and ethylenediamine ......... l4
Chromatogram showing residence time of ethanolamine .................. 17

Chromatograrns of initial and 5 hour samples ﬁ'om the whey
hydrolysate hydrogenation reaction. There is a considerable increase
in compounds eluting before 5 minutes. A peak in the general region

of ethanolamine is observed in the 5 h sample .............................. 18
Structures of amino acids ...................................................... 25
Hydrogenation of alanine to alanine] ......................................... 25

Schematic representing reduction of organic acid and ester to alcohol 27

Mechanism for hydrogenolysis of methyl formate to methanol as
proposed by Serum and Onsanger ............................................ 28

Substrates used by Varga et al. to study effect of vicinal substituents
on rate of hydrogenation of carboxylic acids ............................... 32

Amino acid hydrogenation on addition of inorganic acid — the
inorganic acid converts the amino acid to its protonated form .......... 33

Generalised amino acid hydrogenation sequence showing aldehyde
and potential enols as proposed by Jere et al. .............................. 34

Composite HPLC Chromatogram of ELSD detector signal showing
individual peaks of compounds studied ................................... 37

ix

Figure 16.

Figure 17.

Figure 18.

Figure 19.

Figure 20.

Figure 21.

Figure 22.

Hydrogenation of serine, alanine and valine at 130°C and 7.0 MPa
hydrogen pressure. (I) - 0.48M serine + 0.52M H3PO4; (O) - 0.46M
alanine + 0.5M H3PO4; (A) - 0.49M valine + 0.5M H3PO4.
Conversions predicted by the model are represented by the lines as:
0.48M serine + 0.52M H3PO4 ; ................ 0.46M alanine + 0.5M
H3P04 ; _ . . _ 0.49M valine + 0.5M H3P04 .......................................

 

Effect of initial concentration on initial hydrogenation rates for the
amino acids. (6) — serine; (A) - valine; (I ) — alanine ....................

Arrhenius plot of rate constants for alanine, serine and valine
hydrogenation. (O) - serine, Ea = 83.1 kJ/mol; (I) - alanine, Ea =
88.5 kJ/mol; (A) - valine, E3 = 95.5 lemol ..............................

Hydrogenation rates of alanine alone and in mixtures with serine and
valine at 130°C and 7.0 MPa hydrogen pressure. (9) - 0.24M alanine
+ 0.56M H3PO4; (A) - 0.22M alanine + 0.13M serine + 0.56M
H3PO4; (I) - 0.23M alanine + 0.24M serine + 0.56M H3PO4; (O) -
0.22M alanine + 0.24M valine + 0.56M H3PO4. Conversions

 

predicted by the model are represented by the lines as: 0.24M
alanine + 0.56M H3PO4; -__- 0.22M alanine + 0.13M serine +
0.56M H3PO4; _ . . _ 0.23M alanine + 0.24M serine + 0.56M

H3PO4; ............ 0.22M alanine + 0.24M valine + 0.56M H3PO4

Hydrogenation of serine alone and in mixtures with alanine and valine
at 130°C and 7.0 MPa hydrogen pressure. (6) - 0.24M serine + 0.56M
H3PO4; (I) - 0.48M serine + 0.52M H3PO4; (0) - 0.24M serine +
0.23M alanine + 0.56M H3PO4; (A) — 0.24M serine + 0.25M valine +
0.56M H3PO4.

Conversions predicted by the model are represented by the lines as:

 

0.24M serine + 0.56M H3PO4; ---- 0.48M serine + 0.52M
H3PO4; _ . . _ 0.24M serine + 0.23M alanine + 0.56M H3PO4;
............ 0.24M serine + 0.25M valine + 0.56M
H3PO4 ....................................................................................................

(l) photolithography is preformed to produce the master, (2) the
PDMS positive is created using the master, (3) remove the
polymerized PDMS from the master, (4) seal the PDMS against a
glass cover slide using plasma bonding to create the microchannels.
Figure taken from reference 74 ..............................................

Schematic showing a) dimensions of microchannels engraved in a
block of PDMS using soft lithography. b) Cross section of a PDMS
microchannel ....................................................................

35

40

41

42

43

60

61

Figure 23.

Figure 24.
Figure 25.
Figure 26.

Figure 27.

Figure 28.

Figure 29.

Figure 30.

Figure 31.

Figure 32.

Figure 33.

Figure 34.

Figure 35a

Figure
35b.

Figure 36

Figure 37

block of PDMS using soft lithography. b) Cross section of a PDMS
microchannel ....................................................................

Dimensions of a glass slide with holes, which is plasma bonded to
PDMS block engraved with microchannels to form a PDMS glass
mieroreactor. (Schematics not to scale) ......................................
Schematic of proposed microreactor-in-pressure-vessel setup ...........

Schematic of modiﬁed ﬂow loop .............................................

Various components of the ﬂow loop ........................................

Close up of the mieroreactor connected to the ﬂow loop and located
inside the Parr reactor ..........................................................

Attempt at connecting the mieroreactor to the ﬂow loop by passing the
tubing through the PDMS side and using epoxy to bond the tubing in
place ..............................................................................
Tubing connections to the mieroreactor through a 10 mm glass slide.

Pd wires placed inside the microchannels in a PDMS block ..............

Hydrogenation of acetone to isopropanol. Standard reaction
conditions: 323K temperature and 700 psi hydrogen ......................

Chromatogram showing RI response and elution times of acetone (23
min) and isopropanol (25.5 min). 5 mM HZSO4 as mobile phase, 0.6
ml/min, 60 °C, Aminex column ...............................................

PDMS tubing with Ru/C catalyst deposited inside the tubing ............

Scanning electron micrograph of a cross section of PDMS tubing with
immobilized catalyst ............................................................

Scanning electron micrograph of a cross section of PDMS tubing with
immobilized catalyst ...........................................................

Scanning electron micrograph of a cross section of PDMS tubing with
immobilized catalyst ..........................................................

Scanning electron micrograph of an edge of a cross section of PDMS
tubing with immobilized catalyst ............................................

Overlaid chromatograms showing decrease in acetone peak area and

xi

62

64

65

65

66

67

68

69

70

70

76

77

78

79

79

81

Figure 38.

Figure 39.

Figure 40.

Figure 41.

Figure 42.

Figure 43.

Figure 44.

increase in isopropanol peak area during the hydrogenation of acetone
using tube based mieroreactor with immobilized Ru/C. Flow rate 0.25
ml/min, temperature 40 °C, 650 psi hydrogen, 6 mg Ru/C in 26 inches
of 800 um tubing ................................................................

Activation energy for acetone hydrogenation using 1 hour data points.

A model for the mieroreactor
setup ...........................................

Predicted versus experimental fractional conversions as predicted by
the second order equation when plugged into the residence time
model .............................................................................

Effect of temperature on the rate of formation of isopropanol. I
denote experimental data at 55 deg C, ---I---- denote predicted data
for 55 deg C; 0 denote experimental data at 40 deg C, ----0 -----
denote predicted data for 40 deg C; A denote experimental data at 27
deg C, ------ A ------ denote predicted data at 27 deg
C ...........................

Effect of length of catalytic tubing on the rate of formation of
isopropanol. I denote. experimental data at 66 cm (6 mg Ru/C), ---I---
- denote predicted data for 66 cm; A denote experimental data for a
tubing length of 33 cm (3 mg Ru/C), ----A ----- denote predicted data
for 33 cm length of tubing; 0 denote experimental data from a repeat
experiment with 66 cm (6 mg Ru/C) length of tubing, ------ o ------
denote predicted data for repeat experiment with 66 cm length of
tubing.

oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo

Effect of initial acetone concentration on the rate of formation of
isopropanol. I denote experimental data at 180 mM initial
concentration of acetone, ---I---- denote predicted data for 180 mM
initial concentration of acetone; A denote experimental data for a
tubing length of 30 mM initial acetone concnetration, ----A -----
denote predicted data for 30 mM starting acetone concnetration; 0
denote experimental data from 5 mM starting acetone conentration , --
---—o ------ denote predicted data for repeat experiment with 5 mM
intial acetone concentration ...................................................

Data from tubing 2, 2mg of Ru/C, at 25°C, in recirculating mode at
0.25 ml/min under 650 psi hydrogen. Each line represents
concentration proﬁle of isopropanol based on a different feed
concentration. ...................................................................

xii

85

88

91

93

94

95

97

Figure 45

Figure 46

Figure 47

Data from tubing 2, 2mg of Ru/C, at 25°C, in recirculating mode at
0.25 ml/min under 650 psi hydrogen. A polynomial ﬁt has been used

to estimate the start of reaction (time = zero). Reaction volume = 7.5
ml .................................................................................

Effect of acetone concentration in feed on reaction rate as seen in data
set 2 with tubing 2. 2mg of Ru/C, at 25°C, in recirculating mode at
0.25 ml/min under 650 psi hydrogen. Reaction solution volume = 7.5
ml .................................................................................

Parity plot comparing predicted fractional conversion and
experimental fractional conversions. Experimental data are from
polynomial ﬁts to data ﬁ'om set 2. Predictions are output from
Langmuir-Hinshelwood residence time model ..............................

xiii

98

99

100

Key to Symbols

Species

A = amino acid A

AS] = adsorbed amino acid A

AI'SI = adsorbed intermediate

Aol = amino alcohol from amino acid A

Aol.S1 = adsorbed amino alcohol from amino acid A
B = amino acid B

BS] = adsorbed amino acid B

B"S. = adsorbed intermediate

Bol = amino alcohol from amino acid B

Bol‘Sl = adsorbed amino alcohol ﬁom amino acid B
H2 = molecular hydrogen

H'Sz = adsorbed hydrogen

P = phosphoric acid

S] = amino acid hydrogenation site

82 = hydrogen adsorption site

Variables

77(ij = observable modulus for species j

b = stoiciometric ratio of hydrogen to amino acid

xiv

C]- : concentration of species j, kmol/m3

k}- : surface reaction rate constant for species j, k S / K A , kmol/kgcat's

ks = surface reaction rate constant, kgcatz/kmolz's

K A = adsorption coefﬁcient of protonated amino acid A, m3/kmol
KB = adsorption coefﬁcient of protonated amino acid B, m3/kmol

Kp = phosphoric acid adsorption coefﬁcient, m3/kmol

KH

2 = adsorption coefﬁcient of phosphoric acid m3/kmol

J}; = reaction (consumption) rate of amino acid A, kmongcat's

D

e, j = effective diffusivity of species j, mz/s

XV

Chapter 1. Introduction
1.1 Background

Petroleum based products have become a part of all our daily lives. Petroleum
products extracted from the earth are reﬁned and used in a variety of applications, from
transportation fuels to textiles and medicines. However, petroleum supplies, due to their

non renewable nature, have been falling short of demand]. The resulting search for

alternatives has identiﬁed plant and animal derived chemicals as potential substitutes.
Starch, hemicellulose, cellulose, lignin, oils and proteins are some of the building blocks
found in biomass feedstocks. Starch, oils and proteins are also important for nutritional
purposes. As an example, corn, which is an important food source, can also be used to
produce ethanol, a fuel, via fermentation. Research has therefore focused on using
building blocks found in waste byproducts, such as cellulose, to produce ﬁiels and
chemicals. Byproducts from plant and animal processing - such as soybean meal and
slaughter house waste - are rich in proteins (Table l)2. They are currently used in low
value applications such as ﬁllers in feed for live-stock. Conversion of such waste proteins
and their component amino acids into value added chemical products has been the
primary objective of this work.

Proteins are made up of amino acids linked together by peptide bonds (Figure l).

H
‘ | N
2 \LOH $01.1 “ZN/T $11
- H20
CH3 H3

Glycine Alanine Glycylalanine, a dipeptide

Figure 1 Condensation reaction between two amino acids to form a peptide linkage

Both the breakdown of proteins to amino acids and further conversion of the
amino acids offer challenges. Hydrolytic cleavage of the peptide bonds can be
accomplished using proteolytic enzymes or by using high concentrations of mineral acids
in combination with high temperatures3, yielding individual amino acids or peptide
fragments. Amino acids obtained through such methods or via fermentation can be
subjected to a number of processes such as hydrogenation, condensation, oxidation,

esteriﬁcation to enhance their utility as feedstocks 4’5 (Figure 2).

Polyglutamic acid HOMO"

 

proliﬁc Prolinol Pyroglutamic acid

Figure 2 The large variety of transformations of glutamic acid which can yield useful
products. Figure taken from PNNL report5

Table 1 General Characteristics Of Various Natural Feed lngredients

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Percent
Total digestible
nutrients
‘ Digestable
Dry (.rude Crude Crude , , ,
, ‘ Rummants Swme Protein
Matter protein fat ﬁber .
Rummants
Arm/u meal,
20% 93.1 20 3.5 20 58 50 14
dehydrated
Alfalfa meal,
17% 93 17 3 24 58 32 12
dehydrated
com dimers 93 8 27 9 13 79 92 19
dried solubles '
C om gluten
88 21 2 1O 75 75 19
meal
Cottonseed
meal, solvent, 91.4 41 4 13 71 67 33
40%
Feather meal 93.2 85 2.5 1.5 63 63 70
Fishmeal,
, 93 72 l0 1 73 70 57
herring
Fishmeal,
92 62 10 1 71 62 49
menhaden
F ishmeal,
. 91 65 10 1 73 70 53
Peruvran
Me” and we 92 6 50 8 5 3 68 65 4o
meal, 50% i '
Poultry
byproduct 94 58 14 2. 5 74 74 46
meal
Sorghum grain 88.8 11 3 2 71 78 6
Soybean meal,
89.6 44 0.5 7 78 71 38
solvent 50%
Soybean ma" 89 15 4 1o 62 57 11 5
44% '

 

 

 

Of the various processes available to upgrade amino acids and related
biorenewable substrates, hydrogenation achieves a particularly large value upgrade. This
thesis is divided into three main sections, each giving an account of the work done related
to that topic. The ﬁrst section describes attempts at hydrogenating polypeptides and
proteins. Hydrogenation of amino acid mixtures forms the second part of the thesis, while
in the third part we look at novel reactor conﬁgurations to carry out and study

hydrogenations.

1.2 Signiﬁcance

Conversion of proteins to value added products via hydrogenation requires
reduction of the amide moiety present in proteins. Chapter 2 offers a brief literature
survey of amide hydrogenations and proceeds to describe attempts at reduction of small
molecule amides, oligopeptides and bulk proteins. Amides were found to undergo
transamidation reaction with ethylenediamine as a solvent. This novel reaction opens up a
route to the functionalization of amides, and offers possibilities for the manufacture of
polymers incorporating amino acids. Hydrogenation of bulk protein hydrolyzates under
conditions amenable to hydrolysis through use of inorganic acids led to the formation of
desired product amino alcohols. However, quantitative characterization of the reaction
mixture posed unique challenges and these are highlighted. This is one of the ﬁrst reports
studying conversion of proteins to value added feedstocks for the chemical industry and

summarizes challenges likely to be encountered in ﬁirther development of this route.

Given that amino acid substrates will likely come from complex mixtures arising
from protein hydrolysis, it is important to study how the presence of multiple amino acids
can inﬂuence hydrogenation rates of the individual species. Hence Chapter 3 discusses
the hydrogenation of three amino acids: alanine, serine and valine and their mixtures. The
hydrogenation of amino acid mixtures was found to follow Langmuir Hinshelwood type
behaviour, with competitive adsorption between the amino acids for active sites on the
catalyst surface. This information allows for improved understanding of substrate-catalyst
interactions in such reactions and allows for more robust design of processes employing
such conversions.

The drive towards more environmentally ﬁiendly processes translates to the need
to use benign solvents such as water as widely as possible. In the case of aqueous phase
catalytic hydrogenations, the low solubility of hydrogen in water and the high gas-liquid
mass transfer resistance result in the need for extreme pressures and mechanical agitation
to improve yields. Even so, in a majority of such reactions, hydrogen is found to be the
limiting reactant on account of reduced availibility at the catalyst surface. Therefore,
there was interest in developing an apparatus that allows study of aqueous phase catalytic
hydrogenations under conditions where hydrogen is not the limiting reactant. Chapters 4
and 5 address this problem by detailing the development and use of PDMS-based
microchannel systems to study high pressure hydrogenations. In Chapter 4, a PDMS—
glass mieroreactor is incorporated into a high pressure hydrogenation setup. A novel
technique to immobilize catalysts in PDMS microchannels is reported in Chapter 5, along
with data that allow comparison of catalyst behaviour in batch mode with that in

microchannel ﬂow mode. These studies are an important contribution to the fundamental

understanding of catalyst behaviour in high pressure aqueous phase hydrogenations.
Additionaly, they are one of the few reported implementations of high pressure
hydrogenations in microchannel systems and are well suited to scale up either as
characterization tools or as low volume, low cost manufacturing processes.

To summarize, each of the chapters to follow represents a novel and original
approach to tackle difﬁcult and multidisciplinary problems along the path towards a
sustainable chemical process industry. In each case, substantial progress towards a
solution has been made. This work is thus a signiﬁcant contribution to the interrelated

ﬁelds of aqueous phase catalysis and biorenewable feedstock based processes.

References

1

2

Kerr R. A., Science, 2007, 317, pp.437.

Considine D. M. and Considine G. D., Food and Food Production Encyclopedia, Van
Nostrand Reinhold, New York, 1982.

Fountoulakis M., Lahm H. W., Journal of Chromatography A, 1998, 826 109—134
Varadarajan S., Miller D. J ., Biotechnol. Prog., 1999, 15, 845-854
Werpy T., Petersen G., eds., Top Value Added Chemicals from Biomass Volume I—

Results of Screening for Potential Candidates from Sugars and Synthesis Gas, PNNL
report, August 2004.

Chapter 2. Hydrogenation of Amides
2.1 Amines — Industrial Uses and Market Size

The reduced availability of petroleum based feedstocks has resulted in shortage of
raw materials for a wide variety of chemicals. Polyaminoalcohols, which are widely used
in the cosmetics industry, are among those compounds which have been affected.
Alternative routes for the manufacture of amino alcohols are thus in demand. Amino
acids can be hydrogenated to amino alcohols under relatively mild conditions, and this
offers a possible alternative route. Protein rich-hence amino acid rich-waste is abundantly
available from sources such as crop and animal processing. Examples include soybean
meal (obtained after extraction from soybean oil), distillers grain solubles (obtained after
distillation of ethanol from fermentation broth), slaughterhouse waste, and spoilt meat.
Such protein rich wastes can theoretically be hydrogenated to a mixture of amino
alcohols. Three main approaches for doing so include hydrogenating the proteins intact,
hydrogenation after hydrolysis to oligopeptides, and hydrogenation after hydrolysis to a
mixture of individual amino acids. In the case of the ﬁrst two approaches, reduction of
the amide functionality becomes a necessity. In this section, we will look at attempts to
reduce the amide bond in dipeptides because these are easier to characterize than
oligopeptides.

The total market for amines in the United States for the year 2004 is estimated to
have been worth $1.9 billion, corresponding to a demand of 70 X 106 kg of amines6.
Amines can broadly be classiﬁed into lower aliphatic amines, fatty amines, alkanol

amines, cycloaliphatic amines, and aromatic amines7. Lower aliphatic amines are

derivatives of ammonia with one, two, or all three of the hydrogen atoms replaced by

alkyl groups of ﬁve carbons or less. Examples of lower aliphatic amines include
methylamines, diethylamines ($3.22 per kg), propylamines ($3 to 4 per kg) and
amylamines. Such amines are important intermediates in the manufacture of solvents,
surfactants, agricultural, textile, rubber, medicinal and plastic chemicals. Amines with
higher alkyl groups are known as fatty amines. These are nitrogen derivatives of fatty
acids, oleﬁns, or alcohols prepared from natural sources, fats and oils, or petrochemical
raw materials. Fatty amines are surface active compounds and thus ﬁnd wide application
as surfactants. Alkanol amines (mostly mono, di and tri — ethanolamines; MEA — $1.4
per kg) are chemical intermediates used largely in the manufacture of surfactants and in
gas conditioning applications to remove C02 and H2S impurities from gas streams. Fatty
amines and ethanolamines are the largest aminetypes used in the detergent and cleaner
markets. Cycloaliphatic amines are low cost organic intermediates, and these generally
consist of a cyclic hydrocarbon structural component and an amine ﬁinctional group
external to that ring. Examples include cyclohexylamine, (with a market of several
thousand MT per year, used in water treatment and rubber chemical applications, $0.69
per kg ($0.31/lb)), and derivatives such as 4-methylcyclohexylamine. Aromatic amines
include aniline and its derivatives, diaminotoluenes, diarylamines, methylenedianiline,
and phenylenediamines. They ﬁnd wide use in dyes, antioxidants, explosives, and
pesticides. The varied applications of amines underscore their importance, especially as
specialty chemicals.
2.2 Hydrogenation of Amides to Amines — Literature Survey

The hydrogenation of amides to amines (scheme in Figure 3) was ﬁrst reported by

Adkins and Wojcik in 19348. In dioxane as a reaction medium, with copper- chromium

oxide as a catalyst under 10 to 30 MPa. of Hydrogen at 448 to 523 K , they found that
lauramide and heptamide gave 40 to 70 % yields of corresponding primary amines. They

also achieved conversion of N-laurylpiperidine.

i R2 /
(if . H. _. Rm.

R3

Amide Am inc

Figure 3. Reduction of amides to amines

Side reactions:

 

 

Amide Alcohol Amine

R R
/ 2 /\ / 2
N + R]/\OH ———> R1 N + R3011

L. J

R1
Amme 1511001101 Byproduct amine

Figure 4. Side reactions in the reduction of amides to amines
The chief side reaction is the formation of secondary amines (Figure 4). The production
of alcohols and mixed amines results in a requirement for distillation of the product
mixture to isolate the desired product. This increases energy requirements. The reaction is
. favored by higher temperatures and higher pressures. The use of higher pressures implies
customized equipment, increasing capital costs. The use of hydrogen, a ﬂammable gas, at

high pressures and temperatures poses a health and safety hazard. Better conversion can

10

be achieved by longer reaction times. In a batch setup, this implies the exposure of
already formed product to high temperatures with the inherent risk of degradation. In the
case of ﬂow reactors, higher reaction times imply longer reactors.

Developments towards better efﬁciencies in this process have focused primarily
on catalysts used in the process. Bard et al. in 19659 carried out the catalytic
hydrogenation of N,N-disubstituted amides at 473 to 623 K, under 1.4 to 5.5 MPa of H2
pressure using copper chromium oxide catalyst. Signiﬁcantly, the presence of 2 to 8 wt %
auxiliary dialkyl amine source was found necessary to increase attainable conversion.
The auxiliary secondary amine source (same as that used to prepare the amide) inhibits
the side reaction of conversion to alcohol. Best values reported are 90 % conversion, 81
to 88 % selectivity over 5 to 10 hrs. To utilize the enhanced reaction rates possible with
higher surface area of catalysts, King, in 198410 used copper chromite supported on
zeolite as catalyst. Hydrogen pressures of 1.3 to 3.4 MPa were required, although 72 %
selectivity towards desired amines could be achieved in 1 hr. It was possible to omit the
auxiliary dialkyl amine source. Mixed catalyst system of copper oxide, copper chromite
and manganese oxidel 1, palladium, ruthenium or rhenium supported catalyst and
combinations”, titanium based mixed metal oxide catalysts'3 have also been researched.
The use of a nucleophilic reagent such as sodium methoxide in addition to the copper
chromite catalyst has been found to reduce the pressures required to 0.5 to 2 MPa of
hydrogen”. Bimetallic transition metal catalysts in homogeneous and heterogeneous
forms reduced byproduct formation although typical reaction time required was 16 hrs”.

The use of hydrosilanes as a hydrogen source eliminated the need for pressures.

ll

Transition metal complexes as catalysts at 373 K, and reaction times approaching 16 hrs
under inert atmosphere were required”.

In 2006, in a signiﬁcant step forward, Molengraft et al.'7 reported the aqueous
phase catalytic reduction of the cyclic amide 2,5-piperazinedione to the cyclic amine 2,5—
piperazine under relatively mild reaction conditions of 1800 psi hydrogen pressure at 130
°C over a period of 24 h, using Ru/C catalysts. However, linear amides such as
glycylglycine do not undergo catalytic reduction under the same reaction conditions.

The challenges then are to increase conversion and selectivity in the
hydrogenation of amides, including polypeptides, to amines. Different catalysts, solvents
and reaction conditions are some aspects which can be optimized towards this end. A
reasonable strategy would be to start with the hydrogenation of relatively susceptible
molecules such as glycinamide, then move onto successively difficult molecules such as
diglycine, triglycine and onto other oligopeptides. Glycinamide has been shown to be
susceptible to hydrogenation”.

2.3 Exploratory Studies in the Reduction of Amides

While the ultimate objective was to be able to hydrolyze and reduce proteins, it is
difﬁcult to characterize proteins because of their high molecular weight. Hence,
polypeptides, smaller analogues of proteins, were chosen for study. Some examples of

polypeptides include glycylglycine and glycylglycylglycine as seen in Figure 5 below.

Glycylglycine Glycylglycylglycine

Figure 5. Examples of polypeptides.

l2

In general, polypeptides are soluble in water, but are generally insoluble in any
other solvent. In water, hydrolysis of the peptide bond giving constituent amino acids
occurs far more readily than hydrogenation. Accordingly, our attempts at hydrogenation
of glycylglycine under aqueous condtitions (100 ml solution volume of 0.22 M
glycylglycine with 1 gm dry Ru/C under 6.9 Mpa hydrogen pressure and 130°C with
stirring at 1000 rpm) resulted in complete hydrolysis to glycine and hydrogenation to
ethanolamine, and no aminoethylethanolamine was observed. It was found that
glycylglycine had excellent solubility (approx 20 wt %) in ethylenediamine. Reactions
were therefore carried out using ethylenediamine as a solvent. Very few studies have
reported the use of ethylenediamine as a solvent. Of these, one approach uses a reaction
between fatty acids and amine to form polymerslg. There have been no reported studies
of amide hydrogenations using ethylenediamine as a solvent. Test reactions were carried
out under standard hydrogenation conditions. Typically, 12 gm of substrate glycylglycine
was dissolved in 140 ml ethylenediamine and added to 1 gm of prereduced 5 % Ru/C
(dry basis). Each reaction involved subjecting the reaction mixture to 130°C and 7 Mpa
hydrogen under strirring at 1000 rpm. Samples were withdrawn at regular intervals and
subjected to analysis by nuclear magnetic resonance (N .M.R.) spectroscopy.
Temperatures studied ranged from 100°C to 160°C. Reactions were carried out both with
and without catalyst, and hydrogen pressures were varied from 0.1 to 6.9 Mpa. No
desirable hydrogenation to aminoethylethanolamine was observed in these reactions.

The reactions did result in a transamidation reaction between the amide and
solvent. Glycylglycine (glygly), glycine anhydride, and glycine were tested as substrates.

As a test reaction, glygly was stirred in ethylenediamine at 1000 rpm for 7 hrs under 6.9

13

MPa hydrogen at 433 K, over 1.4 wt % catalyst (Ru on carbon, wet basis) in a Parr
autoclave. NMR analysis showed that the glygly underwent a transamidation reaction
with the ethylenediamine (Figure 6), with all the glygly forming N-(2-

aminoethyl)glycinamide, as seen from H-NMRs and C-13 NMRs.

O
O
NHz -1120
HzN H 2N
Glycylglycine Ethylenediamine 2—Aminoethylglycinamide

Figure 6. Transamidation reaction of glygly and ethylenediamine

Subsequently, a similar reaction was - observed with glycine anhydride as
substrate, under similar conditions. The absence of catalyst and hydrogen pressure also
resulted in the same product with glygly. Thus, the transamidation reaction does not
require catalyst or hydrogenation pressure. A control reaction was tried out with glycine
as substrate, no catalyst, no hydrogen pressure. This, too resulted in amidation, albeit
only at temperatures greater than 413 K. The results of the transamidation runs are
summarized in Table 2.

Transamidation may offer novel routes to the synthesis of amines. Additionaly,
the observed reaction opens up a route to potentially make polymers starting from amino

acids or other organic acids.

14

Table 2. Results from studies of transamidation reactions.

 

 

 

 

 

 

 

 

 

Sr
H T -
Substrate Solvent 2 Catalyst Temperature Time mus
pressure mdam"
no
2.33 gm
12 gm 140 ml 6.9
l _ _ . Ru on 433 K 7hrs Yes
glycylglycrne ethylenediamine MPa
C (wet)
1.3 2.3 m
gm 140 ml 6.9 g
2 glycrne . Ru on 433 K 7 hrs Yes
. Ethylenedramine MPa
anhydride C (wet)
None,
12.09 gm 120 ml but No
3 . 433 .K 7 hrs Yes
glyclyglycine ethylenediamine spar 86d catalyst
with H2
None. 2 hrs 373 K,
4 5.58 gm 120 ml but ‘No 2 hrs 393 K, Y
- es
glycine ethylenediamine sparged catalyst 2 hrs 413 K,
with H2 7 hrs 433 K

 

 

 

 

 

 

2.4. Protein Hydrolysis and Hydrogenation

Given the ready availability of cheap protein feedstocks from sources such as
slaughter house waste, boven serum albumin (BSA) and commercially available whey
hydrolyzate were chosen as substrates to study their potential utility as sources of
chemical feedstocks. The approach taken was to hydrolyze the proteins or bulk
hydrolyzates using either mineral acid such as H3PO4 or enzyme such as trypsin and then
hydrogenate the hydrolyzate obtained.

Protein hydrolysis is a well studied subject, pursued primarily to discern the
structure of proteins”. The standard procedure for breaking down a protein into its

components involves dissolving the protein into a solution of 6 N HCl and heating the

l5

same at 110°C for 24 hours. This results in a solution of the constituent amino acids
which can then be analyzed to determine protein structure. Use of inorganic acids
frequently results in the breakdown to the monomeric amino acids. Control of the
reaction to achieve oligopeptides is difﬁcult to achieve. Enzymatic hydrolysis, on the
other hand, cleaves the protein at speciﬁc peptide bonds, such as alanine-serine, and thus
results in a relatively more well-defmed hydrolyzate solution. Both approaches were tried

out, the objective being to obtain a solution of oligomeric amino alcohols.

2.4.1 Hydrolysis and Reduction Using H3PO4

Hydrogenation of amino acids to amino alcohols can be carried out using an
acidic solution. The fact that proteins can be hydrolyzed in an acidic solution is thus
advantageous, since a single pot reaction can theoretically be carried out. The protein
BSA (Bovine Serum Albumin) was chosen as a substrate. This protein was subjected to
hydrolysis and hydrogenation conditions in a Parr reactor. The resultant reaction mixture
was analyzed using HPLC.

One gram of whey hydrolyzate was added to 100 ml water with vigorous stirring.
30 ml (85 wt %) H3PO4 was added to this suspension. The resulting suspension was

transferred to a feed vessel. The feed vessel was then connected to a Parr reactor
containing prereduced catalyst, 2.25 gm Ru on C (pre-reduced at 2.1 MPa hydrogen
pressure, 523 K, for 10 h) and the feed was transferred to the reactor. The reaction
conditions were 6.9 MPa of hydrogen pressure at a temperature of 428 K accompanied by
stirring at 1000 rpm. The reaction mixture was sampled at 0 hr and 3 hrs. Samples were
ﬁltered over 0.22 micron Millex GV ﬁlters to remove catalyst particles. Reaction was

stopped at the end of 4 hours, and samples were stored at temperatures below 277 K. The

16

signiﬁcant quantity of phosphoric acid in the reaction mixture tends to leach nickel from
the stainless steel reactor vessel. Hence, 20 ml of reaction sample at end of 4 hrs. was
reﬂuxed with CaCO3 overnight to remove nickel. The resultant solution was ﬁltered, the

residue washed with water again, and the water was evaporated ﬁom the pooled solution,
to give a white residue.

HPLC analysis of the reaction mixture (Figure 7 and 8) showed peaks
corresponding to those of amino alcohols. Thus, some amount of the intended reaction
products were obtained. However, analysis of high molecular weight peptide mixtures
poses unique challenges, and these are discussed further towards the end of this chapter.

d:\dc\hplc disc\8thfeb\ethanolamine 2-8-05 1_38_32 pm.run

 

 

 

 

 

 

l l l I l l T l l 1

Minutes

Figure 7. Chromatogram showing residence time of ethanolamine.

l7

      
     

l 9-nov-05-
0hr
ethanol

 

 

aminel
i; '77 VT v v '—"’Tf v —T_T- ~‘ﬁ—‘l
2.5 5.0 7.5 Minutes 100

Figure 8. Chromatograms of initial and 5 hour samples from the whey hydrolysate
hydrogenation reaction. There is a considerable increase in compounds eluting before 5
minutes. A peak in the general region of ethanolamine is observed in the 5 h sample.

2.4.2 Hydrolysis Using HCl Followed by Hydrogenation in a Parr Reactor
Hydrochloric acid is a stronger hydrolyzing agent than H3PO4. However, the
corrosive nature of HCl prevents a one pot reaction from being carried out in the stainless
steel Parr reactor. Instead, hydrolysis was carried out in a glass round bottom ﬂask using
6 M HCl and the solution, after neutralization with NaOH, was hydrogenated. Apart from
this separation of the hydrolysis step, the procedure was similar to that mentioned in the
previous section. However, due to concerns related to degradation of the HPLC column,

no analysis of samples from the reaction was attempted.

2.4.3 Enzymatic Hydrolysis using T rypsin and Subsequent Hydrogenation
Enzymatic hydrolysis of substrates enables mild reaction conditions (room
temperature) and thus requires less energy input. Although enzymes cost more than

mineral acids, enzymes can be immobilized onto supports, allowing their separation,

recovery and reuse. Experiments were therefore carried out using trypsin as a hydrolytic
enzyme, with the objective of cleaving the protein BSA (Bovine Serum Albumin) into
low molecular weigh fragments”. The resulting solution was subjected to hydrogenation
in a Parr reactor, using a stoichiometric quantity of H3PO4.

Trizma—HCl (0.2586 gm) was dissolved in 40 ml water and 1 gm bovine albumin
added to this. A solution of the enzyme was prepared by adding 0.01 gm trypsin to a 1
mM HCl solution. The 2 solutions were mixed, and kept at 313 K along with occasional
shaking of the ﬂask. At the end of 2 hours, a solution of 0.0053 gm trypsin inhibitor in 2
ml water was added to stop the hydrolysis. 10 ml of this solution was kept aside for
analysis. The remaining approx 30 ml was hydrogenated as follows: 30 ml of the
enzymatic hydrolyzate was diluted to 100 ml, 0.8 ml 85 wt % phosphoric acid was added

and this was fed to the Parr reactor containing catalyst prereduced at 523 K, 2.1 MPa Hz

for 6 hrs. Subsequent hydrogenation conditions were 473 K, 6.9 MPa H2, 1000 rpm for
6.5 hrs.
Gel electrophoretic analysis of the hydrolyzate before hydrogenation showed that

the protein had not been hydrolyzed.

2.4.4 Results

The starting protein material had a high molecular weight (66000 daltons).
Analysis of the protein hydrolysis was attempted using gel electrophoresis. Gel
electrophoresis has the drawback that conventional techniques do not allow for
determination of protein ﬁ'agments with molecular weights less than 1000 daltons.

Furthermore, this technique is not quantitative in nature. HPLC analysis of the reaction

19

mixture (hydrolysis and hydrogenation using H3PO4) (Figures 7 and 8) showed peaks
corresponding to those of amino alcohols. Thus, some amount of the intended reaction
products were obtained. However, accurate quantitative determination of the extent of
hydrolysis and subsequent hydrogenation proved difficult.

To summarize, protein hydrolysis and subsequent hydrogenation do yield amino
alcohols. However, quantitative analysis of the reaction mixture is difﬁcult to achieve.
Progress in this particular area will require close cooperation with experts in the analysis

of high molecular weight biomolecules.

2.5 Summary

Attempts at hydrogenation of low molecular weight amides using
ethylenediamine as solvent led to the discovery of a transamidation reaction between the
substrate and solvent, but did not result in any reduction of the substrate. The
transamidation of amino acids offers routes to the ﬁrnctionalization of amines and can
also allow the incroporation of amino acids into polymers.

The hydrolysis and hydrogenation of oligopeptides and bulk protein hydrolyzates
was attempted using inorganic acids and enzymes. In a single step reaction using
phosphoric acid and under hydrogenation conditions, commercially available whey
hydrolysate was found to yield ethanolamine. This is one of the ﬁrst attempts at

conversion of bulk proteins to value added chemical products.

20

Acknowledgements

This work was supported by a grant from the Consortium for Plant Biotechnology
Research (US. DOE DE-FG36-02G012026) under grant G012026 and by Procter and
Gamble. The NMR analysis of the transamidation reaction samples was performed by Dr.

Mikhail Redkovich. Whey hydrolysate was received as a gift from AMCO.

2l

References

10

ll

12

l3

l4

14

15

16

17

18

l9

19

20

US Amines Industry Serves Diverse Markets in Competitive Environment, Report
by the Freedonia Group, 2000

Encyclopedia of Chemical Technology, vol. 2, Wiley, New York, 1991
H. Adkins. and B. Wojcik, J. Am. Chem. Soc, vol. 56, pg. 247, 1934

N.M.L. Bard, Hopkins, Vertnik L., Fisher R., McCaleb K., US Patent 3190922, to
General Mills Corporation, June 22, 1965

R. King, US Patent 448998, to The Procter and Gamble Company, May 15, 1984

C. Forquy and R. Brouard, US patent 5075505, to CECA SA, France, December
24th 1991

I. Dobson, EP0286280 Al, to BP Chemicals Limited, October 12th 1988
J. Barrault and M. Seffen, US Patent 4,935,546, CECA SA, June 19th 1990

M. Nepras, R Bernhardt. and C. Sporer, US Patent 5840985, Stepan Company,
January 29th 1998

M. Nepras, R Bernhardt. and C. Sporer, US Patent 5840985, Stepan Company,
January 29th 1998

C. Hirosawa, N. Wakasa and T. Fuchikami, Tetrahedron Lett., 1996, Vol. 37, No.
37, pp. 6749-6752

M. Igarashi, T. Fuchikami, Tetrahedron Lett., 2001, Vol. 42, No. 42, pp. 1945-1947,
A. Molengraft, M. S. Thesis, Michigan State University, East Lansing, MI, 2006
Unpublished results, Miller Jackson group, Michigan State University

J. Cowan, L. B. Falkenberg, H. M. Teeter and P. S. Skell US. Patent 2450940 ,
1948.

J. Cowan, L. B. Falkenberg, H. M. Teeter and P. S. Skell US. Patent 2450940 ,
1948.

M. Fountoulakis and I-I-W. Lahm, J. of Chromatogr. A, 1998, 826, pp. 109—134

22

21 T. Peters and R. Feldhoff, Biochemistry, 1975, Vol. 14, No. I 5, pp. 3384-3391

23

Chapter 3. Hydrogenation of Amino Acid Mixtures
3.1 Introduction

Catalytic hydrogenation of individual amino acids has been studied previously in
22-25 2627 . . . .
our laboratory and elsewhere . As compared to traditional methods of ammo acrd

reductions, viz. metal hydride catalyzed reductions, these methods use less hazardous and
less expensive chemicals, produce no inorganic byproducts, eliminate the need for
intermediate processing steps such as esteriﬁcation, and allow the use of water as a
reaction medium. Given that amino acid substrates will likely come from complex
mixtures arising from protein hydrolysis, it is important to study how the presence of
multiple amino acids can inﬂuence hydrogenation rates of the individual species. Hence
we report here the results of a study of the hydrogenation of three amino acids: alanine,
serine and valine and their mixtures (structures given in Figure 9a, representative reaction

24 25 - -
‘ rn Wthh

given in Figure 9b). The work is an extension of the prior work of Jere et al.
a rigorous kinetic model detailing active species, hydrogen pressure, and catalyst loading
was developed. The goal of this work is to quantify reaction rates of the amino acids
alone and in mixtures, to understand the competition between substrates and interactions
between the substrates and catalyst, and to shed light onto the reaction mechanism. The

understanding gained from this model system will aid in design of hydrogenation

processes for more complex biorenewable feedstocks.

24

 

Serine Alanine Valine

Figure 9a. Structures of amino acids

 

Zwitterion Protonated Alcohol
form

Figure 9b. Hydrogenation of alanine to alaninol

3.2 Hydrogenation of Organic Acids and their Esters - Literature Survey

Amino alcohols are used as chiral building blocks for a wide variety of specialty
chemicals. The amino alcohols can be obtained by reduction of amino acids.
Traditionally, such reductions have been carried out using metal hydrides such as sodium
borohydride (NaBH4) or lithium aluminum hydride (LiAlH4). However, metal hydrides
are hazardous, they need to be used in stoichiometric quantities, and they cannot be easily
separated and reused. Heterogeneous catalytic hydrogenations offer the advantage of easy

catalyst separation and reuse and thus allow the design of continuous processes.

25

Organic acids are not readily amenable to heterogeneous catalytic hydrogenations.
Some of the earliest reports of catalytic hydrogenation of carboxylic acids to alcohols
used copper catalysts at 573-673 K , 19.3 MPa hydrogen pressure”, copper chromium
catalystszg’29 at 573 K and 24-35 MPa, copper cadmium catalysts30 at 553 K and 11.7
MPa, copper-cobalt-nickel promoted catalysts31 at 393-573 K and 36-386 MPa, ferrous
metals modiﬁed with various non-ferrous chromites32 at 473-673 K and 10-20 MPa,
ruthenium dioxide33 at 423 K and 48-56 MPa, rhenium black34 at 423-473 K and 16-29
MPa. The hydrogenation of esters (Figure 10) in general requires milder conditions than
those for the corresponding acid. Laszier35 reported conversion of substrates such as n-

butyl acetate over a mixed oxide catalyst containing copper, zinc and manganese, at 673

K and 20.7 MPa.

Subsequently, researchers have tried to reduce the severity of temperature and
pressure conditions required for hydrogenation. These attempts have largely targeted

catalyst optimization and solvent selection.

26

H High H 2 pressure R1

R2 ’ H
0 High temperature R2

Organic acid Ca‘a‘YS‘ Alcohol

R1 High H 2 pressure 1

o —, H + R3OH
\ High tern rature R2
R2 R3 Pe »
Catalyst
0
Ester Alcohol

Figure 10. Schematic representing reduction of organic acid (R3=H) and ester (R3=alkyl)
to alcohol

One of the earliest systems involving hydrogenation of esters studied was that of
methyl formate to yield methanol. In the two step process of Christiansen36, methanol
was reacted with carbon monoxide to yield methyl formate which was subsequently
hydrogenated to give net yield of one mole of methanol. Copper oxide reduced to copper
was used as catalyst. Evans37 studied the hydrogenation of methyl formate using copper
chromite as catalyst, achieving 90% selectivity to methanol at atmospheric pressure and
373-503 K . The mechanism of methyl formate hydrogenolysis, as per Scrum and
Onsanger38, involves the formation of a hemiacetal intermediate, which is also the rate
limiting step. The hemiacetal decomposes rapidly to formaldehyde and methanol, and

the formaldehyde is hydrogenated to methanol (Figure 11).

27

*
H
W + H2* H
H*
“2* l

’1‘ *
HCH20H( )+ CH3OH() .— + CH3OH(*)

:1:

H

Figure 11. Mechanism for hydrogenolysis of methyl formate to methanol as
proposed by Serum and Onsanger38.

Using in-situ IR spectroscopy and silica supported copper as catalysts, Trimm et

31.39 suggested that the formyl group is bound to the copper during hydrogenation.

Copper in the form of Raney Copper has been used to effect the hydrogenation of
ethyl acetate (500 K, 4.8 MPa)“, yielding ethanol. Formation of acetaldehyde, which is
in equilibrium with ethanol, may limit conversion. Opportunities exist for innovative
separation processes. Ethanol can also be prepared by hydrogenation of methyl acetate,

with ethanol and ethyl acetate as intermediates, using a copper catalyst“.

Studying ﬂuorinated and perﬂuorinated esters, Yan42 et al. proposed that the
hydrogenolysis of esters proceeds through a hemiacetal intermediate. They based this
assumption on the trace quantities of hemiacetals and or or aldehydes observed at the end
of the reaction. Larger quantities of intermediates, particularly the hemiacetal, occurred in
the case of the hydrogenolysis of ﬂuorinated esters, since ﬂuorinated hemiacetals are

known to be stable. Additionally, they pointed out that the acetates were likely to

28

dissociatively adsorb to the catalytic surface via the acyl fragment”. Mechanistic studies

have also been performed using isotopic labeling of methyl formate”.

Adkins and coworkers44 reported the low temperature (298-423 K) hydrogenation
of esters to alcohols using modiﬁed raney nickel catalysts. In doing so, they improved on
the work by Mozingo and Folkers45’46 involving hydrogenation of malonates,
acetoacetates, and benzoates using a high ratio of copper-chromium oxide to ester. By
reducing the temperature and time of digestion of the raney nickel, they were able to
achieve enhancement in catalytic activity towards hydrogenation of esters. With copper
chromite, a high ratio of catalyst/substrate (1.5 gm of catalyst per gm of substrate) was
sufﬁcient for diethyl succinate, glutarate, adipate and methyl laurate - these were
hydrogenated alter 4 to 13 hrs at 423 K, even though they did not have any activating
group. Raney nickel was active towards hydrogenation, yet all attempts to achieve
selectivity w.r.t. phenol groups (avoid reduction of phenol group) did not succeed —
cyclohexyl substituted alcohols were obtained as products. Amide linkages in a
carboethoxy pyrrolidone and a hippurate were not hydrogenated while the ester group
was being converted to a carbinol group, i.e. cyclic amide reduction was not observed.
Beta substituted amino ester, ethyl beta piperidinopropionate underwent cleavage as the
chief reaction with only a low yield of the desired piperidino alcohol. With alpha, beta
hydroxy esters, hydrogenolysis was a problem. Difﬁculty of hydrogenating substituted
malonates increased as the size of substituent increased. Adkins and coworkers also

. . . . . . 47
reported increases rn reaction rates wrth increase in temperature and hydrogen pressure .

29

Hydrogenation of dibasic esters such as diethyl oxalate has been studied38’48‘49.

Catalysts used include Copper-Zinc oxide, silica supported copper and copper chromite.

Segel50 studied hydrogenation of amino acid esters, revealing that copper
chromite can give excellent yields of optically active amino alcohols, without the
necessity of using large amounts of catalyst. Dioxane was used as a reaction medium.
Esteriﬁcation was carried out using the amino acid, ethanol and sulfuric acid. L-leucine
ethyl ester was hydrogenated in 3hrs at 423 K, 1.38 MPa hydrogen pressure, giving 85%

yield.

Broadbent et al.34 studied the hydrogenation of organic acids using rhenium black
synthesized in situ from rhenium heptoxide. They hydrogenated a wide variety of organic
acids, such as acetic acid, capric acid, lauric acid, succinic acid to the corresponding
alcohols. Solvent effects played an important role — while the hydrogenations run on the
anhydrous acids always resulted in some by-product ester formation, those run in water

as solvent gave markedly reduced ester formation, or in most cases no ester by-product at

all.

Studer5| et al. reported room temperature hydrogenation of chiral (it-amino esters
and (Jr-hydroxy esters using a bimetallic nishimura catalyst and found no racemisation.
The mixed Pt/Rh oxide, is generally active for ring hydrogenation. However, the authors
chanced upon reduction of the carbonyl group during reduction of 2-hydroxy—4—phenyl
butyrate. A relatively high loading of the catalyst was necessary — 10% (w/w) — lower
catalyst loadings resulted in reactions but a variety of byproducts (dimmers and

oligomers) were formed. Only the esters of (it-hydroxy and a-amino acids were converted

30

— non functionalized esters, acids or amides failed to react. a-functionalised carboxylic
acid derivatives could be selectively hydrogenated in the presence of another ester group,
the non activated ester ﬁlnction remains unreduced but formed a lactam. Reactions with
B-amino acid esters were slower than with the (ll-amino esters. Higher catalyst loadings
were necessary for (ll-hydroxy esters than for (ll-amino esters. The catalyst had a different
appearance in the low temperature experiments (ﬁnely divided powder), as compared to.
high temperature experiments (aggregates). Reuse of the catalyst was not possible. The
addition of acid led to a signiﬁcantly reduced reaction rate. The authors propose a
multifunctional active centre consisting of a basic and multi-metallic sites. They theorize
that the a-NHZ, a-NHR, or a-OH groups act as efficient anchoring groups by interaction
with a basic site situated very close to the metallic centers leading to the strongly
adsorbed and activated centre, with implications for the reactivities of 01 vs. [3 substituted

CStCI‘S.

Varga52 et al. studied the effect of different electron withdrawing and hydrogen

bonding or sterically hindering vicinal substituents on the carbonyl group of carboxylic
acids (Figure 12). Both hydrogen bonding and electron withdrawing nature of the vicinal
substituents was found to activate the substrate towards hydrogenation. The compounds
studied included lactic acid (LA), glycolic acid (GA), 2-methoxypropanoic acid (MPA),
methoxyacetic acid (MA), isobutyric acid (IBA), propanoic acid (PA), 2—chloropropanoic
acid (2CPA), ethyl lactate (EL), 2-acetoxypropanoic acid (2APA). The order of reactivity
towards hydrogenation at the carboxyl group was: PA < 2MPA < IBA < MA < GA < LA
< EL. 2-acetoxypropanoic acid underwent hydrolysis to lactic acid and so the effect of

the acetoxy group at the 2 position could not be studied. l-chloropropanoic acid

31

converted to propanoic acid and did not undergo further conversion. They also propose a

reaction mechanism proceeding through a hemiacetal, aldehyde and ﬁnally resulting in

the alcohol.

OH 3 i

H OH OH
0
WiC acid (LA) Glycolic acid (GA) 2-Methoxypropanoic acid (M PA)
CH3
H ”3 I I
Methoxyacetic acid (MA) lsobutyric acid (13 A) Prop anoic acid (PA)
Ac
Cl OH
H 0V OH
O O
2-Chlorop ropanoic acid (2CPA) Ethyl Lactate (EL) 2-Acetoxy prop anoic acid (2APA)

Figure 12. Substrates used by Varga et al. to study effect of vicinal substituents on rate of
hydrogenation of carboxylic acids.

Zhang et al.53 studied the aqueous phase hydrogenation of lactic acid to propylene

glycol. They achieved 95% conversion at 373 to 443 K, hydrogen pressures of 7 to 14
MPa. Selectivity to propylene glycol was 90%. Potassium, calcium lactate salts cannot be
directly converted to PG, however, addition of sulfuric acid converts the salts to calcium,
potassium sulfates and allows hydrogenation. Lactic acid is expected to be available

cheaply ($0.85/ lb), hence this route can compete with traditional route for manufacture of

32

PG. Traditional routes to propylene glycol involve hydroperoxidation or the chlorohydrin
process. Pervious work by Adkins44 (25 MPa, high catalyst loading), Broadbent34
(hydrogenation of free lactic acid using unsupported rhenium black as catalyst, 8h, 423
K, 27 MPa), Ford, Camahan33 (Pressures > 48 MPa for direct hydrogenation of
carboxylic acids) relied on high hydrogen pressures. In comparison, the work by Zhang et
al. offers higher reaction rates, better selectivities under milder conditions. At
temperatures greater than 443 K, byproducts such as propanol and ethanol were formed.

As stated earlier, the salts need to be acidulated prior to hydrogenation.

54,55 55,57

Patents by Antons et al. and work by Jere et al. revealed that the use of

inorganic acid allowed the conversion of amino acids to the corresponding amino
alcohols at relatively mild conditions of temperature and pressure (373 K, 6.9 MPa

hydrogen), with stereoretention (Figure 13, 14) .

Jere also found rapid H-D exchange at the C2 position without loss of
stereochemistry and without any reaction in the hydrogenation of alanine. In addition,

stereoretentivity was observed at temperatures as high as 383 K.

 

Zwitterion Protonated Alcohol
form

Figure 13. Amino acid hydrogenation on addition of inorganic acid — the inorganic acid
converts the amino acid to its protonated form.

33

  

H NH3

   

/

H H
R 100°C R 150°C R
ﬂ‘OH I /“‘t()H ’ ¢
+ + 3‘
NH3

NH3 H

    

+
NH3

Figure 14. Generalized amino acid hydrogenation sequence showing aldehyde and
potential enols as proposed by Jere et al. 0

Recent work by Chen et al.58 has shown that competition for catalyst active sites

exists between lactic acid and propionic acid during the course of their hydrogenation. In
their case, addition of a second acid signiﬁcantly decreased the conversion rate of the
ﬁrst acid. A Langmuir - Hinshelwood type mechanism was able to predict the reaction
kinetics. Lahr and Shanks59 have also looked at competitive adsorption during

hydrogenation with glycerol, ethylene glycol and propylene glycol.
3.3 Materials and Methods

Aqueous phase catalytic hydrogenation of amino acids was carried out in a
Teﬂon-lined 300ml Parr high pressure reactor (Model 4560). Alanine, serine, valine, and

HPLC—grade water were used as purchased from Sigma-Aldrich. The catalyst was a pre-

34

reduced 5% ruthenium on carbon (Johnson Matthey - Tennesee, lot no. 20892), obtained

as a slurry containing approximately 50 wt% water.

To begin a reaction, 1.0 g of 5 wt% Ru/C catalyst (dry basis) was added to the dry
reactor. The reactor was ﬂushed to remove air, and then the catalyst was reduced under
1.5 MPa hydrogen and 200°C overnight. The reactor was then cooled to room
temperature, a solution of amino acid substrate in 100 ml water was added to the reactor
under low hydrogen pressure, and the stirring rate was ﬁxed at 1000 rpm. Because
hydrogenation of amino acids requires that they be present in the protonated form, a
slight molar excess (10—20%) of phosphoric acid (relative to the amino acid) was added
to the feed solution in each reaction prior to reaction. When the reactor temperature
stabilized at the desired value (363-403 K), the reactor was pressurized with hydrogen
gas (7 MPa) to initiate the reaction (t=0). Samples (1 to 2 ml) were withdrawn at regular
intervals through a sample loop, ﬁltered using Millipore Millex GV syringe ﬁlters,
diluted in water to ~0.005M, and analyzed on a Waters HPLC with ultraviolet (UV) and
Evaporative Light Scattering (ELS) detectors using a reverse phase ion exchange column
from Silec Technologies”. The HPLC method employed an acetonitrile/triﬂuoroacetic
acid (TFA)/water gradient as follows: the binary gradient shifted linearly ﬁom a 0.05%
TFA - 10% ACN solution (mobile phase A) in water to a 0.5% TFA - 10% ACN solution
in water (mobile phase B) over 12 minutes, and then the latter mobile phase continued for
an additional 18 minutes. A ten minute equilibration time on mobile phase A was
included in each run before the next sample was injected. Five-point calibration curves

were obtained for each compound at concentrations from 0.0005M to 0.005M. The

35

solubilities of the substrates in water at 28°C are listed in Table 3. The concentrations

used for all experiments were well below these maximum values.

 

 

 

 

 

Table 3. Solubilities of amino acids in water at 25°C 60’“
Solubility in water Solubility
Amino acid
(g/100g, 25 T) (M)
Alanine 16.6 19
Valine 8.85 0.8
Serine 50.23 4. 8

 

 

 

The calibration curves for the ELS detector are non-linear at lower concentration
ranges, but were reproducible in repeated calibrations. The combined ELSD
chromatogram for all three amino acids and their alcohols is given in Figure 15 . Only the
alaninol and valine peaks show signiﬁcant overlap — these species concentrations were
determined by integrating the UV peak for valine (alaninol has no UV signal) to
determine valine concentration, then subtracting the corresponding valine area from the
combined ELSD peak to give an alaninol peak area and thus the alaninol concentration.
We attempted a direct deconvolution of the valine — alaninol overlapping peaks, but this
procedure yielded. signiﬁcantly less reliable results compared to the UV/ELSD
measurements. Concentration changes over the course of the reaction were thus obtained
from the HPLC data and were used to calculate conversion, product yields, selectivities,
and overall material balances. Fractional conversion was calculated relative to the

concentration of amino acid in the solution added to the reactor — in some instances

36

 

(Figures 16, 19, 20), the conversion of the reaction sample taken upon reaching steady
state conditions (t = 0) is less than zero, because of either uncertainty in the analysis or

evaporation of water in the reactor that results in artiﬁcially high concentrations. In no

case was the conversion less than -4%.

   
 

 

 

 

:rinol
serine
S‘
E
g. alanine
2 V valinol
:33 A alaninol
x
m
o
o.
2 *3 '4 ‘5 '6 '7 ‘
Time (min)

Figure 15. Composite HPLC chromatogram of ELSD detector signal showing
individual peaks of compounds studied.

3.4 Results and Discussion

3.4.1 Hydrogenation of Individual Amino Acids

Serine, alanine and valine were hydrogenated at 130°C and 7.0 MPa H2 over

concentration ranges from 0.12 M to 0.5 M with 5 to 20 per cent excess H3PO4. As can

37

be seen in Figure 16 for a 0.46 M feed solution of the amino acid with 0.5M H3PO4,

serine and alanine react signiﬁcantly faster than valine. The same trend was observed
when the acids were reacted at an initial concentration of 0.22 M, with serine and alanine

conversion nearly complete by the end of 3.5 hrs.

100 _,__-__-....-_w,._._,.. I.“ m - - 4,

 

 

 

Conversion (%)

 

 

 

 

Figure 16. Hydrogenation of serine, alanine and valine at 130°C and 7.0 MPa hydrogen
pressure. (I) - 0.48M serine + 0.52M H3PO4; (O) - 0.46M alanine + 0.5M H3PO4; (A) -
0.49M valine + 0.5M H3PO4. Conversions predicted by the model are represented by the
lines as: 0.48M serine + 0.52M H3PO4 ; ................. 0.46M alanine + 0.5M H3PO4 ;
_ . . ._ 0.49M valine + 0.5M H3PO4

 

For alanine and valine, carbon molar balances closed to 96% :t 4%, and both
alanine and valine showed greater than 90% selectivity to their corresponding amino

alcohols. In contrast, liquid phase carbon molar recovery in serine hydrogenation was

38

only 70%. A control reaction (130°C, 7.0 MPa H2) starting with serinol and H3PO4
resulted in 26% serinol degradation over 5 hr. At these temperatures, methane and ethane
were observed as byproducts in GC-MS of the vapor phase. Thus, serinol degradation
leads to lower than anticipated selectivity to serinol and lower mass balance closure in

serine hydrogenation than with the other amino acids.

Concentration proﬁles versus time were plotted for each reaction and ﬁtted to a
second-order polynomial. Differentiation of the resulting curves allowed calculation of
rates at various times over the course of reaction. Initial rates were thus calculated by
evaluating the slope of the curve at time zero. By incorporating the acid concentration
and the quantity and properties of the catalyst, reaction rates on a catalyst mass basis
(kmol acid/kg catalyst/s) or on a reactor ﬂuid volume basis (kmol acid/m3 solution/s)
were determined. These initial rate data were used to estimate the activation energy and
the dependence of rates on acid feed concentrations. This initial rate evaluation.

minimized possible complications from side reactions and catalyst deactivation.

Initial hydrogenation rates at different feed concentrations for serine, alanine, and
valine (approximately 0.15, 0.22, and 0.5 M) with otherwise identical reaction conditions
are reported in Figure 17. The reaction rates for serine and alanine exhibit a non-linear
dependence on initial concentration, and thus the hydrogenation reactions are not ﬁrst
order with respect to the amino acid concentrations. This observation is consistent with

the results obtained by Jere et al.,56 who reported a Langmuir-Hinshelwood rate

expression describing the reaction kinetics of alanine hydrogenation. Valine reaction
rates were zero order in valine concentration over feed concentration ranges 0.02 M to

0.5 M.
39

Activation energies for amino acid hydrogenations were estimated by
plotting the natural logarithm of the initial rate versus reciprocal absolute temperature at
constant initial concentration and catalyst loading (Figure 18). The activation energies
reported in the caption are similar for the three amino acids; the value for alanine (81.5

kJ/mol) is in reasonable agreement with the value reported by Jere et 31.5657 (86.4 kJ/mol)

in earlier work.

 

 

 

 

 

 

0.025

’5 0.02 , , i , r.
g l
.3
3 1"
g. 0.015 - . - , , 0
~<
'3 0
E.
3 - - {a 4
to ..
b
.7!
1‘."
5 0.005 ,

g I

0 Y j 77 ‘T T
O 0.1 0.2 0.3 0.4 0.5 0.6

Initial concentration (M)

Figure 17. Effect of initial concentration on initial hydrogenation rates for the amino
acids. (0) -— serine; (A) — valine; (I ) — alanine.

40

 

-10

-11 _

-12 4

.— ———-—- ‘—————__.._ m-

     
 

y = -10000x + 12.4
y = -10648x +14

ln(dCAldt) (mollgmcatls)
63

 

 

 

 

-14

-15 4 = -11484x + 14.6

-16 r l l 7

0.00245 0.0025 0.00255 0.0026 0.00265 0.0027
1IT(1IK)

Figure 18. Arrhenius plot of rate constants for alanine, serine and valine hydrogenation.
(O) - serine, E3 = 83.1 kJ/mol; (I) - alanine, Ea = 88.5 kJ/mol; (A) - valine, E3 = 95.5
kJ/mol.

3.4.2 Hydrogenation of Amino Acid Mixtures

Recent work by Chen et a158 illustrates competition between lactic acid and
propionic acid for catalyst active sites during hydrogenation. In that work, addition of a
second acid signiﬁcantly decreased the conversion rate of the ﬁrst acid. A similar effect
is observed in the case of hydrogenation of amino acid mixtures. As seen in Figure 19,
addition of serine or valine to the feed at a ﬁxed alanine concentration slows alanine
hydrogenation increasingly as the concentration of the competing amino acid increases.
A similar effect can be observed when alanine or valine is added to serine hydrogenation

(Figure 20).

41

  
 

 

 

120 1~-—-— _- — - —- -——-—----— ~ ~ .,_,,_L,__ ‘7 -— we ~~~~-e~ * ~i~ﬁa
100 1 .
. i
o A i
1: 3° ‘ ‘
2‘.
c
.2
e 60 *
O
>
C
8
o 40 i . - ' ‘.
c - a
'E ' 0 . - ‘ ' 1
2 . - I ’ 1
< 20 _. ’ ‘ I . _
o i
'20 l T r 1
0 1 2 3 4 5 6
Time (hr)

Figure 19. Hydrogenation rates of alanine alone and in mixtures with serine and valine
at 130°C and 7.0 MPa hydrogen pressure. (9) - 0.24M alanine + 0.56M H3PO4; (A) -
0.22M alanine + 0.13M serine + 0.56M H3PO4; (I) - 0.23M alanine + 0.24M serine +
0.56M H3PO4; (o) - 0.22M alanine + 0.24M valine + 0.56M H3PO4. Conversions
predicted by the model are represented by the lines as: 0.24M alanine + 0.56M
H3PO4; - - __ 0.22M alanine + 0.13M serine + 0.56M H3PO4; _ . . _ 0.23M alanine +
0.24M serine + 0.56M H3PO4; ............ 0.22M alanine + 0.24M valine + 0.56M H3PO4

 

120 7— ,- ,_ ~- _, ,.

 

 

100' . ‘7/_ g g ,,,A . Q
/ //7 z ’3
sear 9 /’—"' .
c /' (‘. ‘ I
.g 60 / / ' / ‘ _
3 / ’I/ ’ ° . — ‘ ' A
E ./ ,'O . ‘ ‘ A
0 4o / /’ ‘ v ‘A
E I/" ’ i' l
,7, 20 o// — ‘ '
(’ v‘ A
0
'201 r I l
0 1 2 3 4 5 6

Time (hr)

Figure 20. Hydrogenation of serine alone and in mixtures with alanine and valine at
130°C and 7.0 MPa hydrogen pressure. (6) - 0.24M serine + 0.56M H3PO4; (I) - 0.48M
serine + 0.52M H3PO4; (0) - 0.24M serine + 0.23M alanine + 0.56M H3PO4; (A) -
0.24M serine + 0.25M valine + 0.56M H3PO4.

Conversions predicted by the model are represented by the lines as: _,-- L 0.24M serine +

0.56M H3PO4; ____ 0.48M serine + 0.52M H3PO4; _ . . _ 0.24M serine + 0.23M
alanine + 0.56M H3PO4; ............ 0.24M serine + 0.25M valine + 0.56M H3PO4

3.4.3 Analysis of Mass Transfer Resistances

For the data to represent intrinsic kinetics of the reaction, the contents of the
reactor must be well mixed and resistance to mass transfer should not limit the
hydrogenation rate. Mass transfer in the reactor consists of the following steps: 1)
hydrogen transport from the gas phase to the liquid phase across the gas-liquid interface;
2) hydrogen and alanine transport from the bulk liquid phase to the catalyst particle
surface across the solid-liquid interface; and 3) hydrogen and alanine diffusion through
the porous catalyst support to the metal surface sites. In addition, a prerequisite for

43

unhindered mass transport is that all catalyst in the reactor is suspended in solution. The
Zwietering correlationb2 gives an estimate of the minimum stirring rate necessary for
complete catalyst suspension. For this system, with 1 gm of catalyst in 100 ml of dilute
aqueous solution, the minimum stirring speed required for complete catalyst suspension
is 500 rpm; a stirring rate of 1000 rpm was therefore used in all experiments to ensure all

catalyst was suspended.

The limiting reactant inside the catalyst particles is estimated by calculating the
ratio of the observable modulus of hydrogen to that of alanine,63 where the liquid-phase
effective diffusivities of alanine and hydrogen are estimated using the Wilke-Chang
equation.64 For all reactions, the ratio of observable moduli was much greater than 1,
indicating that hydrogen is the limiting reactant. Applying the Weisz-Prater65 criterion to

the observable modulus for hydrogen shows that the effect of mass transport inside the

catalyst pores is negligible, as 77(sz 2 for H2 is <10'3 for all reaction conditions.

WH2 2 C A De,A

l
TWP/12 b CH2 D6112

 

(1)

The Yagi and Yoshida66 and Boon-Long"7 correlations were used to estimate gas-liquid
(G-L) and solid-liquid (S-L) interfacial mass-transfer coefﬁcients for H2 and alanine,

respectively. The maximum mass transfer rates, estimated as the product of the mass
transfer coefﬁcient and bulk liquid-phase concentration, were then calculated and found

to be at least two orders of magnitude larger than the observed reaction rates, indicating

44

that the mass-transport resistances across phase boundaries were negligible under the

experimental conditions studied.

In summary, from the analysis of various mass transfer processes we conclude
mass transfer resistances do not limit amino acid conversions over the range of

temperatures and pressures studied in this work.

3.4.4 Kinetic Model

In a prior publication,24 we showed that alanine hydrogenation over ruthenium on
carbon occurs by a two-site Langmuir-Hinshelwood (L-H) kinetic mechanism in which
protonated amino acids adsorb on one type of surface catalytic site (81) and hydrogen
dissociatively adsorbs on a second type of site (S2). The ﬁnal rate equation is given in
Eq. (2) below. This two-site catalyst-substrate interaction is supported by the molecular
modeling results of Neurock and coworkers68 for hydrogenation of acetic acid. over
palladium. They show that the adsorbed acid is initially bound via the carboxylate ~2 A
above catalyst surface palladium atoms while the dissociated hydrogen atoms69 nestle in

the 3-fold coordinated interstices between surface metal atoms.

We have ﬁt alanine hydrogenation data, presented in Figure 3 and determined at
additional combinations of reaction conditions, to our previously developed kinetic

model24 given in Eq. (2).

KACA+ KH2CH2
(2)
(1+ KACA+ "'KPCP)(1+\/K11_12CH2)2

 

—rA=ks

 

45

The model was applied with constants from the prior work,24 except that the

preexponential factor of the forward rate constant (ks) in the model was adjusted

downward by 40% from the original value24 to represent the somewhat lower intrinsic
activity of the particular lot of Ru catalyst used in this study. Equation (2) describes the
current alanine hydrogenation data with <15% average error, a level of error similar to

that found in our prior study.

Although Eq. (2) involving protonated alanine describes the data well, it is
somewhat cumbersome to implement because the protonated alanine concentration must
be calculated at each value of bulk concentration via a rather complicated charge balance
involving alanine and phosphoric acid in all of their ionic forms. For multiple amino acid
substrates, this charge balance is further complicated by multiple species and thus
constitutes a signiﬁcant barrier to the practical application of the model. Therefore, we
have chosen here to develop an alternate model for amino acid hydrogenation involving
total amino acid concentration. This simpliﬁed model is more amenable to mixed
substrates and is still a valid description of hydrogenation, because for the rather narrow
range of reaction conditions investigated here the fraction of total amino acid in
protonated form only ranges from about 40% to 60% over the course of reaction. Thus,
protonated acid concentration is for the most part proportional to the total amino acid

concentration, and a viable, practical model can therefore be implemented.

The two-site Langmuir-Hinshelwood model, in which molecular hydrogen and
amino acid adsorb on the catalyst surface and subsequently react to form amino alcohols,

can be represented as follows.

46

KA

 

 

 

 

 

 

 

 

A + 81 A81
KB
B + S1 B8]
KH
H2 + 2 82 2 H82
ks =1:
A.S1+ 2H.82 —* A.S1+ 282
*
A8] + 2 H82 Aol.S1 + 2 82
kg :1:
B.81+ 2H.82 —'> B.81+ 282
*
8.81 + 2 H82 = B0181 + 2 82
801.81 ‘———" 801 + 81 fast
A0181 Aol + 81 fast

 

 

(33)

fast (3b)

(30)

slow (3d)
fast (36)
slow (31)
fast (3g)
(3h)
(3i)

where A and B are the amino acids adsorbing onto active sites of type 81, A*Sl and

B*Sl are adsorbed intermediates, and A01 and B0] are the amino alcohol products.

Derivation of the kinetic rate equation from the above mechanism follows the
standard procedure for L-H kinetics, and results in the same form as Eq. (2) except with
total amino acid concentration instead of the protonated species. In addition, in this work

all reactions were run at 7.0 MPa, so the hydrogen pressure term is incorporated into the

composite rate constant k'.

47

KHZCHZ

 

(4)

 

kIZkS(1+\/K1r,{2CHz)2

The rate form for hydrogenation of a single amino acid therefore reduces to:

kAKACA

 

 

(1+KACA +KPCP)
For the case of two amino acids A and B, the rate form is
k' K C
—"A = A A A (6)
(1+KACA +KBCB +KpCp)

Values of the three composite reaction rate constants, k'serine, k'alanine, and
k'vaune, and the adsorption constants Kserinea Kalanines Kvaline and KH,P0, in the kinetic

model were determined by ﬁrst estimating values for the constants and then numerically
integrating the rate expressions to obtain predicted concentration versus time proﬁles for
each reaction. Optimum rate constant values were then determined by varying the rate

constants systematically to minimize the error deﬁned in Eq. (7) below.

CA,pred ': (3A , exp

 

 

Z

n samples C A , exp

Error (%) = x100% (7)

 

"samples

48

The error minimization was carried out over the entire dataset containing both
individual and mixed amino acid hydrogenations. The resulting optimized rate constants

are listed in Table 4.

Table 4. Kinetic constants for amino acid hydrogenation at 403 K.

 

 

 

 

 

 

Surface reaction rate Ad . ﬁ‘ . K
Compound constant k,” sorptlonl‘zole rcrent ,
(kmol/(kgcat.hr)) ( )
Serine 0.018 56
Alanine 0.016 24
Valine 0.0015 124
H3PO4 19

 

 

 

Parity plots of experimental vs. predicted concentrations for each amino acids
over all experiments gave correlation coefﬁcients (R2 values) of 0.98 for serine, 0.97 for
alanine, and 0.99 for valine. The model prediction of amino acid concentration vs. time in
reaction is given as lines in Figures l4, l7, and 18. For the optimized rate constants, the
calculated average error between experimental and predicted amino acid concentrations
was 14% for amino acid conversions up to 60%, and 38% for amino acid conversions up
to 90%. Calculating error on a percentage basis as in Eq. (7) magniﬁes differences at low
concentration, and since there is more uncertainty present in data at low concentrations
because of handling and analytical uncertainties, the signiﬁcantly higher error at high

conversions is not unexpected. Fortunately, the model’s uncertainty at low

49

 

concentrations, although high on a percentage basis, is small on the absolute
concentration scale. Thus, the model predicts single and mixed amino acid

hydrogenation data reasonably well over the DJ" range of conditions examined.

The L-H rate expression provides insight into the relative surface concentrations
of hydrogen and amino acid on the catalyst, as according to the L-H model the
denominator terms (Eq. 5) give the relative concentrations of vacant, hydrogen-occupied,
and amino acid-occupied surface sites. For example, at 0.22 M alanine concentration
with 0.3 M H3PO4, alanine occupies 39% of the active sites, 52% of the active sites are
occupied by phosphoric acid, and 9% of the active sites are unoccupied. If a second
amino acid (serine) is added at a concentration of 0.22 M and the H3PO4 concentration is
increased to 0.5 M, the distribution of site occupation is 15% alanine, 46% serine, 35%
H3PO4, and 4% unoccupied. Thus, the second amino acid reduces the number of active
sites occupied by the original amino acid, resulting in a decrease in conversion rate for

the latter.

The adsorption afﬁnity of the amino acids on the catalyst, as reﬂected in the
values of adsorption constants in Table 4, is valine > serine > alanine. The strong
adsorption of valine is evidenced by its essentially zero order behavior at higher
concentrations. The reason why valine adsorbs more strongly is not fully understood, but
valine is a larger molecule and more “organic” in character, so it may prefer to adsorb on
the surface as opposed to remaining in the aqueous phase. Further study will be required

to understand ﬁtlly the adsorption afﬁnities of the different amino acids.

50

3.5 Conclusions

The hydrogenation of serine, alanine and valine and their mixtures in aqueous
solutions has been examined. The presence of more than one substrate affects
hydrogenation rates through competitive adsorption. Of the three amino acids, serine and
alanine exhibit comparable reactivity, while valine is hydrogenated at a considerably
slower rate. A Langmuir—Hinshelwood kinetic model with surface reaction as the rate
limiting step describes the behavior of both single and mixed amino acid hydrogenation
using a single set of parameter values. Valine has the largest adsorption constant,
suggesting that strong valine adsorption is responsible for valine excluding other amino
acids from active sites and for the observed zero-order valine behavior, both leading to

reduced reactivities.

Acknowledgements

The support of the Consortium for Plant Biotechnology Research (CPBR, US.

DOE DEFG36- 02G012026) under grant G012026 is grateﬁllly acknowledged. We also

thank Procter and Gamble for their support of this work.

Nomenclature

Species
A = amino acid A

A8. = adsorbed amino acid A

51

A’s. = adsorbed intermediate

Aol = amino alcohol from amino acid A

A0181 = adsorbed amino alcohol from amino acid A
B = amino acid B

B'S. = adsorbed amino acid B

Bi'S. = adsorbed intermediate

80] = amino alcohol from amino acid B

8018. = adsorbed amino alcohol from amino acid B
H2 = molecular hydrogen

H'SZ = adsorbed hydrogen

P = phosphoric acid

81 = amino acid hydrogenation site

82 = hydrogen adsorption site

Variables

77(ij = observable modulus for species j

b = stoiciometric ratio of hydrogen to amino acid

C]- = concentration of species j, kmol/m3

k; = surface reaction rate constant for species j, k S / K A , kmol/kgcat's
k3 = surface reaction rate constant, kgcatZ/kmolz's

K A = adsorption coefﬁcient of protonated amino acid A, m3/kmol

K 3 = adsorption coefﬁcient of protonated amino acid B, m3/kmol

K p = phosphoric acid adsorption coefﬁcient, m3/kmol

K H2 = adsorption coefﬁcient of phosphoric acid m3/kmol

-rA = reaction (consumption) rate of amino acid A, kmo l/kgcat's

52

D

e,j

= effective diffusivity of species j, mz/s

53

References

22 Jere, F. Catalytic Hydrogenation of Amino Acids and Polyhydroxyalkanoates. Ph.D.
Dissertation, Michigan State University, East Lansing, MI, 2003.

23 Molengraﬁ, A. Catalytic hydrogenation of 2,5-piperazinedione to piperazine. MS.
Thesis, Michigan State University, East Lansing, MI, 2006.

24 Jere, E; Jackson, J.; Miller, D. Kinetics of the Aqueous-Phase Hydrogenation of L-
Alanine to L-Alaninol. Ind. Eng. Chem. Res. 2004, 43, 3297.

25 Jere, F.; Miller, D.; Jackson, J.. Stereoretentive C-H Bond Activation in the Aqueous
Phase Catalytic Hydrogenation of Amino Acids to Amino Alcohols. Org. Lett. 2003,
5(4), 527.

26 Antons, 8.; Beitzke, B. Process for preparing optically active amino alcohols.
US. Patent 5,536,879, 1996.

27 Antons, 8.; Tilling , A.; Wolters E. Method for producing optically active amino
alcohols. US. Patent 6,310,254, 2001.

28 Schrauth W., . Schenck O, and Stickdom K., Ben, 1931, 64, 1314

29 Guver A., Bieler.A. and Jabere. K., Helv. Chim.Acta,, 1947, 30, 39; Sandoz'Ltd.,
German Patent 711,180,Aug. 21, 1941 [Chem Abstr., 1940, 37, 3767 ]

30 Schuckmann G., U. 8. Patent 2,332,834, Oct. 26,1944; Bohme-Fellchemie-
Ges.mb.H., French Patent 819,255,0ct. 13, 1937 [Chem Abstr., 1938, 32, 2143]

31 Schmidt 0., U. 8. Patents 2,322,095-9, June 15, 1943; Schiller G., U. 8. Patents
2,121,367-8, June 21, 1938.

32 Arnold H. R. and Lazier W. A., U. 8. Patent 2,116,552, May 10, 1938.

33 Ford T. A., U. 8. Patent 2,607,807, Aug. 19, 1952; Carnahan J. E., Ford T. A.,
Gresham W. A., Grigsby W. E., and Hager G. F ., J. Org. Chem, 1955, 77, 3766.

34 Broadbent H. 8., Graham C. C., Bartley W. J., and Johnson J. H., J. Org. Chem,
1959, 24, 1847.

35 Lazier W. A., U. 8. Patent 2079414, 1937
36 Christiansen J. A., US. Patent 1302011, 1919

37 Evans J. W., Casey P. 8., Vainwright M. 8., Trimm D. L., Cant N., Appl. Catal. A:
General 1983, 7, 31-41

38 Sorum P.A., Onsanger O.T., Proc. Int. Congr. Catal., 1984, 8, 233-244
54

39 Monti D. M., Cant N. W., Trimm D. L. and Mainwright M. 8., J. Catal., 1986, 100,
17-27

40 Evans J.W., Mainwright M. S., Cant N. W. and Trimm D. L., J. Catal. 1984, 88, 203-
213

41 Claus P., Lucas M., Lucke B., App]. Catal. A: General, 1991, 79, 1-18

42 Yan T. Y., Albright L. F., Case L. C., Ind. Eng. Chem. Res. Prod. Res. Dev. 1965, 4,
101-110

43 Turek T., Trimm D. L., Cant N. W., Catal. Rev. Sci. Eng, 1994, 36, 645
44 Adkins H., Billica H. R., J. Am. Chem. Soc., 1948, 70, 3121-3125

45 Mozingo and Folkers, J. Am. Chem. Soc., 1948, 70, 227

46 Mozingo and Folkers, J. Am. Chem. Soc., 1948, 70, 229.

47 Adkins H., Billica H. R., J. Am. Chem. Soc., 1948, 70, 695-698

48 Thomas D. J., Vehrli J. T., Mainwright M. 8., Trimm D. L App]. Catal A: General
1992, 86, 101-114.

49 Turek T., Trimm D. L,. Black D. 8., Cant. N. V., Appl. Catal A; General 1994. 116.
137-150.

50 Segel E., J. Am. Chem. Soc., 1952, 74, 1096

51 Studer M., Bruckhardt 8., Blaser HU., Adv. Synth. Catal., 2001, No.8, 343,

52 Varga N., M. S. Thesis, Department of Chemistry, Michigan State University, 2002
53 Zhang Z. G., Jackson J.E., and Miller D. J., Appl. Catal. A-Gen., 2001, 219, 89.

54 Antons 8., US. patent 5,536,879

55 Antons 8., Tilling A., Wolters E., WO9938838, 1999.

56 Jere E, Jackson J. E., Miller D. J., Ind. Eng. Chem. Res, 2004, 43, 3297.

57 Jere F., Miller D. J., Jackson J. E., Org. Lett. 2003, 5(4), 527.

58 Chen Y., Miller D. J., Jackson J. E., Ind. Eng. Chem. Res, 2007,46, 3334.

59 Lahr D. G., Shanks B. H., Ind. Eng. Chem. Res. 2003, 42, 5467-5472

60 Dunn, M. 8.; Ross, F. J.; Read L. S. The Solubility of the Amino Acids in Water. J.
Biol, Chem. 1933, 103 (2), 579.

61 Dalton, J. B.; Schmidt, C. L. A. The solubilities of certain amino acids and related
compounds in water, the densities of their solutions at twenty-ﬁve degrees, and the

55

calculated heats of solutions and partial molal volumes. II, J. Biol, Chem. 1935, 109 (I),
241.

62 Zwietering, T. N. Suspending of solids in liquid by agitators, Chem. Eng. Sci. 1958, 8,
244-253.

63 Ramachandran, P.; Chaudhari, R; T hree-Phase Catalytic Reactors; Gordon and
Breach Science Publishers: Philadelphia, PA, 1983.

64 Wilke, C.; Chang, P. Correlation of diffusion coefficients in dilute solution. AIChE J.
1955, 1, 264-270.

65 Weisz, P.B.; Prater, C. D. Interpretation of measurements in experimental catalysis.
Adv. Catal. 1954, 6, 143-196.

66 Yagi, H.; Yoshida, F. Gas absorption by Newtonian and non-Newtonian ﬂuids in
sparged agitated vessels. Ind. Eng. Chem. Process Des. Dev. 1975, I4 (4), 488-493.

67 Boon-Long, 8.; Laguerie, C.; Couderc, J. Mass transfer from suspended solids to a
liquid in agitated vessels. Chem. Eng. Sci. 1978, 33 (7), 813—819.

68 Pallassana, V.; Neurock, M. Reaction paths in the hydrogenolysis of acetic acid to
ethanol over Pd(1 1 l), Re(0001), and PdRe alloys. J.‘ Catal. 2002, 209 (2), 289-305.

69 Singh, U. K.; Vannice, M. A. Kinetics of liquid phase hydrogenation reactions over
supported metal catalysts A review. Appl. Catal., A: Gen, 2001, 213 (I), 1.

56

Chapter 4. PDMS-glass Microreactor

4.1 Introduction

Work presented in the following sections arose from the need to address the

following problems and related opportunities:

Aqueous phase catalytic hydrogenations involve considerable gas liquid mass
transfer resistance. This, coupled with low solubilities of hydrogen in the reaction
mixture (0.02 M at 350 psi and 27 °C; 0.056 M at 1050 psi and 27 °C), results in
the need for high pressures and mechanical stirring to ensure availability of
hydrogen at the active site of the catalyst. In spite of this, a large number of
heterogenous hydrogenations involve hydrogen as the limiting reactant.
Microchannel reactors, on the other hand, present a very small cross sectional
path length and hence very low resistance to the difﬁlsion of hydrogen to the
active site. Thus, high pressure hydrogenations, if carried out in a microchannel
reactor with immobilized catalyst, would result in conditions where hydrogen is
not the limiting reactant. This would allow study of the intrinsic kinetics of
aqueous phase hydrogenation, which is otherwise difficult to examine.

Nanoparticulate catalysts are unstable and tend to agglomerate over long
durations. This tendency to agglomerate can be aggravated by high temperatures
and other conditions commonly found in heterogeneous catalytic reactions. One
strategy to study nanoparticle behaviour as catalysts, while preventing
agglomeration, is to anchor the nanoparticles on porous supports. Another
strategy can be to pass reaction mixture over a monolayer of catalytic
nanoparticles — this would allow characterization of the activity of nanoparticles

towards hydrogenation in the absence of any interactions by the support.

57

To implement the above approaches, a microchannel reactor would have to be
immersed in a hydrogen rich environment. One face of the microchannels could be a
catalytic membrane permeable to hydrogen, such as a palladium membrane. The
hydrogen could thus diffuse through the membrane into the microchannels, while the
palladium simultaneously catalyzes the reaction. The lack of mechanical stability of most
microchannel reactors to pressure differentials (between inside and outside of the
mieroreactor) greater than 50 psi is an inherent limitation of such an approach. A
different solution would involve presaturating the reaction mixture passing through the
microchannels with hydrogen gas at high pressure, thus allowing equal pressures inside
and outside the mieroreactor and eliminating pressure gradients and related problems of
mechanical integrity. It is this latter approach which is presented in the following

sections.

4.2 Immobilization of Catalytic Nanoparticles

Exploratory work in the Miller Jackson lab (Dr. Kausik Mukhopadhyay)
demonstrated that ruthenium nanoparticles supported on MCM-41 were active towards
the catalytic reduction of lactic acid and glycinem, as shown in Table 5 below. On a per
mole ruthenium basis, the nanocatalysts exhibit an order of magnitude increase in
catalytic activity towards hydrogenation as compared to carbon-supported ruthenium

catalysts.

58

Table 5. Results of Lactic acid hydrogenation using various supported nanocatalysts,
courtesy Dr. Mukhopadhyay (PG= propylene glycol, Ru = ruthenium, APT 8 = amino
propyl trimethyl silane, Y = zeolite Y, M41 = MCM 41, Ru/C = ruthenium supported on
carbon)

 

Catalysts Temp. Press. Substrate Speed Time Conv. Selectivity Catalyst

 

 

 

(K) (MPa) ““147 (rpm) (hours (%) (%) (g)
m \

Ru-APTS- 403 6.9 50 1000 48 10 52 (PG) 0.10
Y

Ru-APTS— 403 6.9 50 1000 48 11 52 (PG) 0.10
Y

Ru-APTS- 403 6.9 50 1000 48 30 98.5 (PG) 0.10
M41

 

Ru-M4l-l 403 6.9 50 1000 24 32 99.3 (PG) 0.10

 

Ru/C 403 6.9 50 1000 48 51.5 99.2(PG) 0.50

 

 

 

 

 

 

 

 

 

 

 

However, to understand the behaviour of nanocatalysts, it is important to
characterize their activity in the absence of interactions with the support, and to
immobilize them in an appropriate manner. Work by Kobayashi et 31.71 has shown that
hydrogenation catalysts can be immobilized onto the glass surface of a microchannel
reactor. They achieved this by functionalizing the silica groups on the glass surface using
an amine, and cross linking microencapsulated palladium onto the functionalized glass
surface. A few workers have coated glass with titania-palladium”. Based on prevalent

literature, it was decided to pursue a strategy based on coating the glass face with titania
and then immobilizing ruthenium nanoparticles onto the coated glass. Once a monolayer
of catalyst had been immobilized onto the glass surface of the mieroreactor, the
mieroreactor would be placed in a high pressure reactor to enable liquid saturated with

hydrogen gas at high pressures to ﬂow past the catalyst.

59

4.3 PDMS-Glass Microreactors

PDMS (Poly DiMethyl Siloxane) based microreactors are usually formed by soft
lithography”. This technique starts by spin coating 3 layer of photoresist up to 200nm
thick onto a support. This photoresist is then exposed to UV light through a pattern
printed on a transparency. The photoresist cures on exposure to UV light. The rest of the
photoresist can be dissolved away, leaving a mould which can be used to cast multiple
PDMS microchannels. The PDMS prepolymer is poured onto the mould and cured at 70
°C for 2 hours. On hardening, the PDMS microchannels slab is ready to be plasma

bonded to glass to form a PDMS glass mieroreactor (Figures 21,22 and 23).

LIGHT
[ * * + * ' 35111111111111
—7Jf‘ Pllotoresist

1V

2V

1 ..__ PDMS

3 I

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PDMS
W
4V ,
2-..-..— Microchannel
mm m FT
' " “ Glass

Figure 21. (l) photolithography is preformed to produce the master, (2) the PDMS
positive is created using the master, (3) remove the polymerized PDMS ﬁom the master,
(4) seal the PDMS against a glass cover slide using plasma bonding to create the
microchannels. Figure taken from reference 74.

60

4.4 Design of System - Modeling and Preliminary Calculations
To determine feasibility, the proposed system was modeled as a differential batch
reactor, i.e. essentially a 10 ml batch reactor with a certain amount of catalyst loading.

Preliminary calculations for catalyst loading were as follows:

 

“—450 pm I
r——a

300 um

Figure 22. Schematic showing a) dimensions of microchannels engraved in a block of
PDMS using soft lithography. b) Cross section of a PDMS microchannel.

61

5 mm

    

%

Figure 23. Dimensions of a glass slide with holes, which is plasma bonded to PDMS
block engraved with microchannels to form a PDMS glass mieroreactor. (Schematics not
to scale).

For mieroreactor dimensions 25 cm length X 600 um channel width X 50 um
channel depth, the surface area of the bottom face works out to be 1.5 X 10‘4 m2. For
ruthenium nanoparticles of 3.5 nm diameter, this area allows approximately 1.1 X 1013
nanoparticles as a monolayer on the bottom face. Based on a ruthenium bulk density of
12.03 gin/cm}, 4 micrograms of the ruthenium nanoparticles could thus be incorporated
as a monolayer of nanoparticles onto the bottom face of the mieroreactor.

For 3 micrograms of Ru nanoparticles (of diameter 3.5 nm), the total surface area
is 4.2 X 10‘4 m2. Hydrogenation of alanine was chosen as a model reaction, the reaction
rate based on active surface area of catalyst being 0.004 moV(m2Ru'hr), for Ru/C catalyst,
as seen from an experimental study of hydrogenation of amino acid mixtures and
conﬁrmed from previous work by Frank Jere in our laboratory“. For such a reactor with

the above assumptions,
dC
V dA = r A RAcat
t
Where V = volume of the reactor, liters = l X 10'2

62

C A0 = concentration of substrate amino acid A, mol/lit = 0.005
~rA = reaction rate for substrate A, mo l/(mzcat'hr) = 0.004

Amt = total catalyst surface area possible on bottom face of microchannel with a

monolayer of3.5 nm particles, m2 = 4 X 104

Using these values, dCA/dt = 1.83 X 104 mol/(lit'hr) and dXA/dt = 0.04 hr'l. Thus,

the system would give 4% conversion of the starting material per hour. Given a method
to immobilize catalyst on the bottom face of the microchannel, the required catalyst

loading could be achieved, and thus, the proposed system appears to be feasible.

4.5 Design of System - Flow Loop and System Constraints

The proposal divided the goal into two major sub tasks, viz. to establish a reactor
conﬁguration and demonstrate test reaction within the assembled setup.
1) Establish reactor conﬁguration

High hydrogen pressures are necessary for hydrogenation but the PDMS
microchannel reactor being used could only withstand pressures of up to 0.5 MPa. Hence,
a novel reactor conﬁguration involving equally high hydrogen pressures both within and
outside the mieroreactor was proposed. This involved placing a mieroreactor in a high
pressure Parr reactor, with pressure in the Parr reactor adjusted so that there was equal

hydrogen pressure both within and outside the mieroreactor, as shown in Figure 24.

63

 

Hydrogen line . .9 7*»

 

 

C
8
9 “c’
U =
>
I
Hydrogen
source
Hold up
chamber

 

Pressure vessel

Figure 24. Schematic of proposed microreactor-in-pressure-vessel setup

2) Test reactions

The PDMS mieroreactor swells in organic solvents, and has limited temperature
stability75 (373 K). Hence, test reactions should ideally be low temperature and low

hydrogen pressure, such as hydrogenation of acetone.

4.6 Development of the Flow Loop

During the development of the mieroreactor setup, changes were made to the
design. It was decided that the holdup reservoir could be incorporated within the Parr
reactor itself. Initially, the holdup volume at the bottom of the Parr reactor was chosen as
the reservoir. Further modiﬁcations were made by inserting a 7-ml vial placed in the Parr

reactor was used as a reservoir (Figure 25). This reduced the length of the ﬂow loop,

64

enabled better control over temperature and pressure in the various components of the

setup, and reduced the volume of the reservoir. The setup is shown in Figures 26 and 27.

 

 

I I 1 Hydrogen feed to

7 Pressure vessel

HPLC pump

 

 

Microreactor

   

7 ml Glass
vial

 

 

300 ml Parr pressure vessel

Figure 25. Schematic of modiﬁed ﬂow loop

HPLC
pump

 

Figure 26. Various components of the ﬂow loop

65

All tubing connections were made using standard 1/16” O.D. HPLC tubing
(initially steel, later replaced by polymeric tubing). One sixteenth inch 8.8. 316 steel
ferrules were used to seal the tubing as it entered and left the high pressure Parr reactor.
The vial was open to the contents of the Parr reactor, thus by pressurizing the Parr
reactor, the contents of the vial and associated tubing were also pressurized. This ensured

that the pressures within the mieroreactor and that outside the mieroreactor were equal.

To pump

    

Microreactor

Vial as reservoir

Figure 27. Close up of the mieroreactor connected to the ﬂow loop and located inside the Parr reactor

4.7 Flow Loop: Connections between the Microreactor and the HPLC Tubing

The diameter of the microchannels in the PDMS-glass mieroreactor was of the
order of 50 pm. This is considerably smaller than the diameter of standard HPLC tubing.
Also, the ﬂow through the microchannels is likely to result in a pressure drop of the order

66

of 0.1 to 0.2 MPa. Hence, to prevent leakage, connections at the inlet and the outlet of
the mieroreactor needed to be mechanically strong. There is a considerable amount of
literature dealing with connections between the micro and macro parts of a microchannel
system“, but the high pressures involved in the current setup, especially at startup, led us
to develop alternatives. We began by passing the HPLC tubing through the PDMS block
of the PDMS-glass microreactor (Figure 28), such that the tubing would be open to the
ends of the microchannels. The tubing was glued by using epoxy or PDMS itself. Various
types of tubing (polyethylene, 8.8.316, and polyimide, all of 1/16” O.D.) were tried.
However, the soft nature of the PDMS meant that these connections were prone to

leakage once ﬂow was started.

 

Figure 28. Attempt at connecting the mieroreactor to the ﬂow loop by passing the tubing
through the PDMS side and using epoxy to bond the tubing in place.

The next approach therefore relied on passing the tubing through the glass side of
the PDMS-glass mieroreactor. Holes were drilled or sand blasted into the glass before
plasma bonding. The holes through thin glass slides (thickness 1 mm) afforded too little
surface area for epoxy binding to the tubing. Consequently, glass slides with a thickness
of 10 mm were tried (Figure 29). With these slides, a strong and leak proof connection

Could be obtained between the inlet / outlet tubing and the mieroreactor. However, the

67

uneven nature of their surface resulted in an uneven plasma bond between the glass and
the PDMS, leading to leakage at the PDMS-glass interface. Slides with a thickness of 5

mm were better suited to the task at hand and were used in further studies.

Figure 29. Tubing connections forth-e mieroreactor through a 10 mm glass slide.

Other alternatives tested included slides with 1 mm thickness having glass tubes
glued onto them to act as ports for the tubing, but the ports themselves would come apart

after 36 hours.

4.8 Catalyst Immobilization

Initially tests were performed to ensure that the setup was leak proof. Various
approaches towards immobilization of nanoparticles were being tried out. These included
using APTS (Amino Propyl Trimethyl Silane) as an anchoring agent between the glass
and the ruthenium nanoparticles, and using biotinylated palladium to enhance bonding to
the glass slide. Immobilization of bulk metals such as palladium or platinum within the

microchannels was suggested as a practical and elementary approach for demonstrating

68

reactivity within the proposed setup. Although this approach would give catalytic surface
area two orders of magnitude smaller than the immobilized nanoparticles originally
planned, it would nevertheless validate the concept. Hence, palladium microwires of 50
micron diameter were placed in the microchannels before irreversible bonding to the

glass, as shown in Figure 30.

 

 

 

 

 

Figure 30. Pd wires placed inside the microchannels in a PDMS block.

4.9 Choice of Substrate

The substrate chosen had to be reactive enough to demonstrate signiﬁcant
conversion to the hydrogenation product at low temperatures, viz. 50 deg C.
Hydrogenation of acetone to isopropanol proceeds relatively fast under such mild
reaction conditions. Additionally, the existing analytical setup was capable of analyzing
acetone and isopropanol at the low concentrations being considered. Hence
hydrogenation of acetone was selected as a test reaction. A schematic of the reaction is

shown in Figure 31 and a chromatogram is shown in Figure 32.

69

H2
’

H CH3 Ru/C catalyst H3 CH3

 

Acetone Isopropanol

Figure 31. Hydrogenation of acetone to isopropanol. Standard reaction conditions: 323K
temperature and 700 psi hydrogen,

 

 

 

 

[. _ I, L,-, ____ “I,
1
: 11 l
l: 1 l I
I 10mM acetone 1 1
! j l
1 1 i 10mM isopropanol
1 l l
1 l
l l i
l l
i i /
l 1 .. 1' l '
1 ' ‘T T “"T'_ WW“ WI ”'ﬁbﬂwﬁ- "PM 7' $.- ‘12—”:m‘wﬁi'2’f‘r
i— l H__'_‘ _ _—— 1_— —— —‘ W —' _l ‘h —_" "h— _ “"‘7— .._ " T“ ' 'b" T" _l
15 20 25 30 Minutes

Figure 32. Chromatogram showing RI response and elution times of acetone (23 min) and
isopropanol (25 .5 min). 5 mM H2804 as mobile phase, 0.6 m1/min, 60 °C, Aminex
column (Bio-Rad HPX-87H ion exclusion column)

4.10 Results and Discussion

Five ml of 5mM acetone in water was subjected to hydrogenation in the
mieroreactor setup. A typical run involved passing the reaction mixture at 1 ml/min
through the ﬂow loop for 20 hours. The temperature of the system was varied between 22
and 55 °C between runs, while hydrogen pressure was kept constant at 350 psi or higher.
Five mM glycerol was used as an internal standard on account of its negligible volatility

under reaction conditions. This in conjunction with control runs without any catalyst gave

70

an estimate of the amount of acetone lost due to evaporation during each run. Table 6

summarizes results ﬁom the runs.

Table 6. Summary of results from hydrogenation runs using Pd wire immobilized in
PDMS-glass mieroreactor

 

 

 

 

 

 

 

. . . Yield to IPA
Feed reaction conditions Catalyst +/_ 4%
5 ml of 6mM acetone 20 °C, 350 psi H2, . o
+ internal standard 22 h at 1 ml/min 25 cm OfSO 11m Pd WEB 3'5 A)
5 ml of 6mM acetone 22 °C, 350 psi H2, 0
+ internal standard 22 h at 1 mein no catalyst 0'75 A)
mM 55 °C, 725 si H
5'5 ml Of8'8 p 2 25 cm of50 11m Pd wire 6 %
acetone + mternal std. 22 h at 1 ml/min
mM 55 °C, 725 si H
5'5 ml 0.f8'8 p 2 25 cm of50 11m Pd wire 36 %
acetone + internal std. 92 h at 1 ml/min
55 °C, no H
5.5 ml of8.8mM 2 25 em of50 11m Pd wire No IPA
acetone + internal std. 21.5 h at 1 ml/min observed

 

 

 

 

 

To compare the reactivity seen with Pd wire with that seen in the case of Ru/C

catalysts, a calculation for the Turn Over Frequency (TOF) was performed as follows:

From amino acid studies, 100 ml of 0.22 M alanine solution reacts completely in 5 hours.

Hence 0.022 moles of amino acid react in 5 hours.
Catalyst used was 1 gm of 5 % Ru/C, i.e. 0.05 gm or Ru. With a metal dispersion of 10%,

the active metal weight is 0.005 gm or 5 X 10'5 moles.
Thus the turnover frequency for amino acid hydrogenation

71

 

 

 

mol alanine 0.022mol.alanine 1 1hr
T0 = x

x
mol Ru X sec 5hr 5 x10_5 moles. Ru 3600s

In the case of the Pd wire in the mieroreactor, the calculations are as follows:
Number of moles of acetone reacted: 8.4 X 10'7 moles in 22 hours.

Surface area of25 cm of 50 um Pd wire is 8 X 10" m2
Radius of a palladium atom is 1.69 X 10'10 m

Cross sectional area of a Palladium atom is 8.97 X 10'20 m
No ofatoms in area 8.7 X 10'3

No of moles of Pd on surface of Pd wire is 1.5 X 10'10 moles

Thus the turnover frequency for acetone hydrogenation is 0.07 moles acetone/mole Pd/s

’7

A:

8.4X10-7mol.acetone)< 1 1hr

10 x =0.07
22hr 1.5x10’ molesPd 3600?

TOP:

 

 

From the above calculations, it can be seen that the turn over frequencies are of
the same order of magnitude in both cases, even though these are different catalyst
metals. 1

It should be noted that the result obtained in the absence of any catalyst, line 2 in
Table 6, shows that the hydrogenation of acetone to isopropanol can proceed, albeit to a
small extent (1%), in the absence of any catalyst. One possible source of this reactivity
can be the metal from the stainless steel tubing used in the ﬂow loop. To minimize this
possibility, all steel tubing was replaced by polymeric PEEK tubing. However, it was not
possible to remove metal from the ﬂow loop entirely since the pump heads and valves
utilize stainless steel as primary material of construction. To conﬁrm that the catalyst was
indeed playing a signiﬁcant role, it was decided to increase the catalyst loading.

4.11 Attempts at Increasing Catalyst Loading
To ensure that the conversion to IPA observed in the experiments was catalytic

and not due to background reactivity alone, attempts were made to increase the catalyst

72

loading inside the mieroreactor. These included mixing Pd/C with the PDMS prepolymer
to test if any Pd/C would be exposed in the surface of the microchannels. Both this
approach and introduction of Pd precursors such as PdCl3 into the PDMS before curing

resulted in negligible amounts of Pd on the surface of the microchannels as measured by

XPS.

4.12 Conclusion

A PDMS-glass mieroreactor was incorporated into a 300 ml high pressure
hydrogenation reactor. A ﬂow loop was developed which ensured equal pressure inside
and outside the mieroreactor, overcoming limitations related to mechanical stability of
the mieroreactor. As a result, liquid saturated with hydrogen at high pressures could be
passed through the microchannels. This eliminates gas liquid mass transfer resistance,
and enables the study of intrinsic reaction kinetics under conditions where hydrogen is
not the limiting reactant. Preliminary results were obtained using immobilized palladium
wires. However, increase in the catalyst loading or active catalytic surface area is
necessary to demonstrate sufficient enhancement in reaction rates. This can be achieved
by immobilizing nanoparticulate catalysts in the microchannels, an approach which is
being pursued by our colleagues, or by developing a different immobilization technique

altogether as demonstrated in Chapter 5.

73

References
70 Mukhopadhyay K., Jackson, J. E., Miller D. J., unpublished results.

71 Kobayashi J., Mori Y., Okamoto K, Akiyama R., Ueno M., Kitamori T., and
Kobayashi 8., Science, 28 May 2004, 304, 1305-1308.

72 Carturan, G.; Faechin, G.; Navazio, G.; Gottardi, V.; Cocco, G. Ultrastruct. Process.
Ceram, Glasses, Compos., [Proc. Int. Conf.] (1984), Meeting Date 1983, 197-206
Publisher: Wiley, New York, N. Y, editors Hench, Larry L.; Ulrich, Donald R.

73 McDonald J., Whitesides G., Accounts of Chemical Research, vol 35, no 7, pg 491
(2002)

74 Zwolinski A. M., Ph. D. Thesis, The Florida State University College of Engineering,
Spring 2004

75 Lee J. N., Park C., and Whitesides G., Anal. Chem, vol 75, pp. 6544-6554 (2003)

76 Bhagat A. A. 8., Jothimuthu P., Pais A., and Papautsky 1., J. Micromech. Microeng.,
17, 42—49 (2007)

74

Chapter 5. PDMS Tubing as Microreactor

Because of the challenges encountered in immobilizing catalyst in and making
secure plumbing connections to the PDMS-glass mieroreactor, it was decided to use
PDMS tubing itself as the mieroreactor. This approach greatly reduced the time required

in testing a prototype, since it eliminated the time lost due to leaking microreactors.

5.1 Development of Immobilization Technique

It was decided that slurry deposition was the most straightforward approach
to immobilization of the catalyst in the PDMS microchannels. However, solvents such as
water do not wet the surface of the microchannels and this did not result in any
immobilization. A search through the literature revealed that there were very few solvents
capable of interacting with the microchannel material on time scales fast enough to allow
evaporative deposition. A screening of various solvents led us to amines as suitable
solvents for the catalyst slurry. Additionally, catalyst, once deposited by the evaporating
solvent front in the microchannels, was fairly strongly immobilized — repeated washes

with solvents did not result in any signiﬁcant leaching of the catalyst.

5.2 Immobilization Procedure

Silicone (essentially PDMS) tubing of 0.03 inch I.D., 0.065 inch CD. was
procured from VWR Scientiﬁc Products (Catalog number 60985—706). A piece of tubing
of known length (ranging from 30 cm to 70 cm) was weighed. A slurry of 80 mg Ru/C
catalyst in 5 ml diisopropylamine was drawn into the tubing, 1 ml at a time, such that at a

time the slurry occupied not more than 5 cm of the tubing. The drawn slurry was allowed

75

to stand in the tubing for 1 minute to allow the solvent to evaporate. The evaporating
solvent front deposited a layer of catalyst particles on the tubing walls. This procedure
was repeated, with a different region of the tubing (of length 2 cm) being targeted for
deposition each time, until most of the tubing had catalyst deposited in its interior. After
all the solvent had evaporated, the tubing was weighed. Subsequently the tube was
ﬂushed with water to remove any undeposited catalyst. Drying and weighing again
complete the procedure. The resultant catalyst loadings ranged ﬁ‘om 3 to 8 mg of Ru/C
deposited per foot of tubing (Figure 33). '

After 50 hours in 10 runs we saw little or no leaching of the catalyst or decrease
in catalytic activity. The above immobilization technique thus allows easy
immobilization of a catalyst of choice in the microchannel tubular reactors that are cheap
and readily available. It is thus potentially a signiﬁcant contribution to the ﬁeld of

catalytic mieroreactor technology.

 

Figure 33. PDMS tubing with Ru/C catalyst deposited inside the tubing

76

Scanning Electron Micrographs of a cross section of the tubing showed Ru/C
catalyst uniformly deposited along the length of the tubing. Catalyst particles ranged in
size from 2 to 50 microns. From the SEM images (Figures 34 to 36), the retention of
activity and the constant weight of the tubing after a number of experiments, it can be

concluded that the catalyst is ﬁrmly anchored to the tubing.

500 um

 

Figure 34. Scanning electron micrograph of a cross section of PDMS tubing with
immobilized catalyst

77

20 pm

 

Figure 35a. Scanning electron micrograph of a cross section of PDMS tubing with
immobilized catalyst

78

 

 

Figure 35b. Scanning electron micrograph of a cross section of PDMS tubing with
immobilized catalyst

79

 

 

Figure 36. Scanning electron micrograph of a edge of a cross section of PDMS tubing
with immobilized catalyst

5.3 Methods

Reactions were performed by connecting the catalytic tubing described above to
the high pressure hydrogenation setup described in the earlier chapter. The catalytic
tubing replaced the PDMS glass mieroreactor used in that setup. As before, 7.5 ml of the
reaction mixture was recirculated through the reactor using an external HPLC pump at
0.25 ml/min for the time of reaction, generally 1 hour. Hydrogen pressure was maintained
at 650 psi and the temperature was maintained at the desired level using the heater
integral to the 300 ml Parr reactor.

Analysis of 200 111 samples drawn at regular intervals was performed using an

HPLC setup equipped with refractive index (R.I.) detector. Data from the UV. detector

80

was collected but not used. The column used was an Aminex column (Bio-Rad HPX-87H
ion exclusion column), maintained at 60°C in the HPLC oven. The mobile phase was 5
mM H2804 as mobile phase ﬂowing at 0.6 ml/min. Evaporation of acetone was
signiﬁcant in reactions at high temperatures (40, 55°C) and reactions involving high
acetone feed concentrations (180 mM). Peak areas so obtained were used to calculate the
concentrations of the various species using a 3 point calibration curve. A representative
chromatogram is shown in Figure 37. Isopropanol was the primary product from the
hydrogenation reactions as shown in Figure 37. Upto 1 to 2% of the acetone reacted to
form by products such as ethanol. A few peaks appeared intermittently and were most

likely intermediates; no attempt was made to identify these minor byproducts.

 

initial 15min

. 30min
1 hr
2hr
acetone isopropanol

1““!

A All),

 

 

 

 

 

l l l I l l 1

10 15 20 25 30 35
Minutes

Figure 37 . Overlaid chromatograms showing decrease in acetone peak area and increase
in isopropanol peak area during the hydrogenation of acetone using tube based

81

mieroreactor with immobilized Ru/C. Flow rate 0.25 ml/min, temperature 40 °C, 650 psi
hydrogen, 6 mg Ru/C in 26 inches of 800 um tubing.

Evaporation of acetone was signiﬁcant in reactions at high temperatures (40,
55°C) and reactions involving high acetone feed concentrations (180 mM). Evaporation
of acetone often resulted in a lower carbon balance for the acetone and artiﬁcally high
conversion values. Observed isopropanol concentrations were therefore used to back
calculate acetone concentrations. Because isopropanol is the only signiﬁcant product
observed in batch reactions where there is no evaporation and the mass balance tallies to
100%, this methodology is valid.

There is an induction period in in the tubing reactor at 25°C that the yield
of IPA in reactions at 25°C is zero at 15 minutes. This is likely because the catalyst was
not prereduced prior to reaction. The setup could not, in its present fonrl, prereduce the
catalyst. Modiﬁcation are possible to enable the same, however, they were not made in
the present case, and therefore a suitable zero time point was assumed based on a
polynomial ﬁt of the reaction data. The fact that the reaction rate at 25°C is low enough
for the data points to be linear is advantageous. This polynomial ﬁt was also used to
compare the predictions from the model at suitable time intervals and obtain the
parameters giving the best ﬁt. Where this approach has been used, primarily with data
from tubing 2, the points from polynomial ﬁt are shown alongwith the original reaction

data.

82

5.4 Results and Discussion

Three microchannel tubular reactors with catalyst immobilized in them were used
to study the hydrogenation of acetone. Three separate pieces of tubing were prepared and
studied in this manner. These three reactors are detailed in Table 7.

Table 7. Charactersitics of catalytic tubing prepared and used in reactions. All
tubing were PDMS with an id. of 800 um.

 

 

 

 

Length Catalyst loading
Name Catalyst
cm mg
Tubing l Ru/C 66 6 :1: 2
Tubing 2 Ru/C 45 2 :1: 1
Pd/C tubing Pd/C 53 10 i 4

 

 

 

 

 

 

5.4.1 Results from Tubing 1

Tubing l was used to compare activity of Ru/C with Pd/C, and to study the effects
of temperature, catalyst loading, and acetone concentration. These are detailed in the
sections to follow.
5.4.1.1 Comparison of Activity of Pd/C with Ru/C

Ru/C was found to have considerably higher activity for converting acetone to
isopropanol as compared to Pd/C, as seen from Table 8 below. The material balances are

low in this case because of acetone evaporation at relatively high temperature (5 5°C).

83

Table 8. Comparison of the activitiy of Pd/C and Ru/C towards the hydrogenation

 

 

 

of acetone
Amount of
catalyst; Time, Yield to % Carbon
Feed length of temperature, IPA conversion balance
catalytic pressure +/- 4% of acetone %
tubing
5 mg of
7-5 ml 10/ Pd/C o
5mM ° Zhatsiﬂc’ﬁo 942% 55% 53%
acetone 53 cm psr 2
Tubing Pd/C
6 mg of
7-5 ml 50/ Ru/C o
5mM ° 'hatSiHC’ﬁo 662.4% 87% 77%
acetone 66 cm psr 2
Tubing l

 

 

 

 

 

 

 

 

5.4.1.2 Effect of Temperature

The following table (Table 9) illustrates the increase in reaction rate with

increase in temperature

Table 9. Increase in reaction rate with increase in temperature, all data using
Tubing 1

 

 

 

 

 

 

 

 

 

Yield to % Carbon
Feed Time IPA, conversion balance
+/- 4% of acetone %
27°C 625 '
acetone 2
- 40°C 650 ’
7.5 ml 5 mM H PS1 1 hour 22 i 4% 35 % 88 %
acetone 2
55°C 600 '
7.5 ml 5 mM H PS1 1 hour 71¢ 4% 90 % 78%
acetone 2

 

 

 

84

The initial rates were calculated using the one hour data points (concentration
change after 1 hour divided by time). These data give an activation energy value of 91
kJ/mol (Figure 38). This value of the activation energy is on the higher side of values
reported in the literature77 of 75 kJ/mol but is close to our ﬁndings from studies of
organic and amino acid hydrogenations. It should be noted in Table 9 that carbon

balances again are worse at higher temperatures because of acetone evaporation.

 

y = ~10918x + 24.9
'7 " R2 = 0.99

In rats (kmol/(m3.s))
do

-11 .

 

 

-12 F l I T T I
0.003 0.00305 0.0031 0.00315 0.0032 0.00325 0.0033 0.00335

1rr, K"

 

Figure 38. Activation energy for acetone hydrogenation using 1 hour data points

5.4.1.3 Effect of Catalyst Loading

To conﬁrm that the reaction was catalytic (e.g. proportional to catalyst loading
and free of mass transfer limitations), the tubing was cut in half and one section was
placed back into the reactor system. This should reduce the amount of catalyst, and thus
the conversion, by 50%. On reconnecting the tubing in its entirety, we expected
conversion to increase back to its original value. The values in Table 10 conﬁrm that the

reaction is indeed catalytic.

85

Table 10. Catalytic effect of tubing length — conversion changes with amount of catalyst.
All data using Tubing l. '

 

 

 

 

 

 

 

 

 

 

Length of Time te erature Yield to % Carbon
Feed catalytic ’ renspure ’ IPA conversion balance
tubing p S +/- 4% of acetone %
7.5mlSmM 66cm lhat40 C,650ps1 22% 35% 84%
acetone Hz
7.5 ml 5mM 30 cm 1 h at 40 C, 650 psr 9 % l6 % 92 %
acetone H2
7.5mlSmM 66cm lhat40 C,675psr 24% 35% 90%
acetone H2

 

5.4.1.4 Effect of Acetone Concentration

At 650 psi of hydrogen pressure, the concentration of hydrogen in water is

approximately 30mM. To conﬁrm that hydrogen was not a limiting reactant at 650 psi

and 5mM acetone, reactions were run at higher concentrations of acetone, 30mM and 180

mM, while keeping the hydrogen pressure at 650 psi. As seen in Table 11, it is observed

that the concentration of isopropanol formed after 1 hour at 40 °C increases with an

increase in the acetone concentration even though the hydrogen concentration is kept

constant at 650 psi. This conﬁrms that hydrogen is not the limiting reactant under the

reaction conditions.

86

 

Table 11. Effect of change in acetone concentration on reaction rate, all data using

 

 

 

 

 

 

 

 

 

 

 

 

Tubing]
Time, H2 Cone. of Yield % Carbon
Feed temperature, Concn IPA to IPA conversion balance
pressure formed +/- 4% of acetone %
7.5 ml
1 h at 40 °C, 30 o o o
5mM 650 psi H2 mM lmM 22 /o 35 /o 84 /o
acetone
7.5m]
1 h at 40 °C, 30 o o o
30mM 640 psi H2 mM 2.6 mM 8 /o 20 /o 88 /o
acetone
7.5m] 180 0
mM ”at“? C’ 30 l3mM 8% 20% 88%
640 ps1 H2 mM
acetone
5.4.1.5 Kinetic Model

the governing equation

where it is assumed that the reaction kinetics can be describe by an nth order dependence

. n
on the concentration of acetone, kCa . However, an accurate model of the system

requires that the different components of the system be included along with equations

that describe their behaviour (Figure 39).

87

(1)

A simple model for the system would be a differential batch reactor with

 

v = 0.25 mein

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

A
V dc,
Sample I d = -k CA
port I T
I Microreactor
1 with catalyst
F: “—4? l 1 a—j Volume
V = 0.3ml
l R
' l
HPLC
' l
pump CA1 i
‘ ch CA I
i A1 2 I
Dotted line I *4 DO I
represents highl Reservoir vial l
33:36 Pa" l volume 7.5 ml 700 ps. I
boundary ‘— ________ hldrggegl

VT= total volume = 9 ml

7:111: lml

 

v — 0.25 ml/min

= 4min

Figure 39. A model for the mieroreactor setup.

88

Accordingly, the reservoir vial containing the liquid can be modelled as a stirred tank,

with the mass balance

Mass in — mass out — diseppeeranee—by—reae—tien = accumulation

dC
vCAz — vC A1 = VCST —d;“—‘ (2)

V
With residence time in the stirred tank denoted by 1' CS = ﬂ , the above equation
v

becomes

 

dCAl : (CA2 ‘CA1)

3
dt TCS ( )

The microreactor with catalyst deposited within it can be modeled as a plug ﬂow reactor.

For such a reactor,

dCAl
dTpF

 

= —ka11 (4)

with the RHS again being based on the assumption that the reaction kinetics can be

describe by an nth order dependence on the concentration of acetone, kCa". Equations (1)

89

and (4) together can be solved to describe the variation of concentration over time in the
reservoir vial. These predicted concentration proﬁles can be compared with experimental
values obtained from a set of experiments.

For a zero and ﬁrst order reaction behaviour, analytical solutions exist. In the case

/ _ \
7.-. o
t

6X
CAlt = CA0 p Q

of a ﬁrst order reaction,

 

(5)

 

 

For a zero order reaction,

kTPFR t
TCS

 

CA0 — CAlt = (6)

However these solutions described above do not describe the experimental data

accurately. In the case of second order reaction kinetics, i.e.

dC
’42 = —kC/2,2 (7)
‘17 PF

 

an analytical solution of the plug ﬂow (above) equation is possible yielding

—l

 

CA2 = +kTpF (8)

Al

For reaction behaviour of this sort, plugging in the above equation into equation (3) for

the stirred tank results in the following equation:

90

—l

1
—-—+kTpF _CA0
dCAi = CA0

dt 1' CS

 

(9)

which is not amenable to an analytical solution. Numerical integration allows calculation
of the predicted concentration proﬁle; however, this model does not give a good

description of the data either, as shown in Figure 40 below.

0.9 ~- 4 —~——— ”eve 2.- W ﬁve? 77777— ——.1
0.8 . :1
0.7 ~ 3
: y = 0.6957x
0.6 a o R2 = «0.3595
.
. l

Predlctod Fractlonal Conversions

 

 

 

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Experimental fractional convors ions

Figure 40. Predicted versus experimental fractional conversions as predicted by
the second order equation when plugged into the residence time model.

Most studies of acetone hydrogenation conclude that the hydrogenation of acetone
follows Langmuir-Hinshelwood type kinetics, with acetone and hydrogen adsorbing on

different types of active sites on the catalyst surface and then reacting to form the product

91

which then desorbs. Accordingly, the reaction kinetics can be described by an equation of

the form

_ k'CA KHZCHZ
1+KACA J1+KH2CH2

 

(7)

 

where k’ denotes the reaction rate constant and KA and KHz denotes the adsorption

coefﬁcients for acetone and hydrogen on the catalyst surface respectively. If the
concentration of hydrogen is maintained constant through the set of experiments, then the
hydrogen related terms remain constant and can be incroporated into a composite reaction

rate constant k as denoted in the expression below:

 

 

k C A
— rA = —— (8)
1+ KACA

When used in the PFR, the equation is

dCAZ __ ICC/12 (9)

dTpF 1+ KACAZ
which on integration yields

__1n __ .___.. C — C = ’Z’ (10
k (CA0) k ( A2 A0) PF )

Because the solution is implicit in C A2, it is more convenient to solve equation (9)

numerically to obtain the concentration at the exit of the PFR and use it in equation (3)

for the stirred tank. These 2 equations allow prediction of the concentration proﬁles using

92

a given set of parameter values, viz. k and KA. The comparison of the predicted and

experimental data for three different temperatures is shown in Figure 41 below:

 

 

 

 

 

 

0.006 -22
0.005 .
i
c 0.004 —
2
g I I
5 0.003 ~ ' '
O
2 0.002 1
2
0001 ~ ,
5'. ‘ . . - - ‘ ‘
,c ' A
o I" .—d I ‘ i r i I 1 _ .7
0 1 2 3 4 5 6
Time (hours)

Figure 41 . Effect of temperature on the rate of formation of isopropanol. I denote
experimental data at 55 deg C, ---I--- denote predicted data for 55 deg C; 0 denote
experimental data at 40 deg C, ----e---- denote predicted data for 40 deg C; A denote
experimental data at 27 deg C, -----A ----- denote predicted data at 27 deg C. All

reactions were run in the recirculating mode at 0.25 ml/min and 650 psi Hg with a
solution volume of 7.5 ml.

To demonstrate that the reaction being observed was indeed catalytic in nature, the
amount of catalyst was varied by cutting the length of tubing in half and then

reconnecting the two ends. The LH model is able to predict behaviour for this set of

reactions as shown in Figure 42 below.

93

 

 

 

 

0.004 1
0.0035 « .
.-:- l
0.003 « ,..-1.' l
g ...l .
v .:-- I i
I
E 0.0025 ..:-' . l
”a .:-' I I
C ,z. a. i
2 0.002« .- u l
on II. "V .
ﬂ 0. ‘t
2'- I:- a“
3 0.00151 .9 u‘
g . .- ‘.““ ‘
O ’l a ‘
o 0001-l ,!' a‘ﬂ
' . 0:. a“ l
m?" t“
A
3
0.0005 :'“.A‘
e,“
0.I r 7 i 1 1 T T
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (hours)

Figure 42. Effect of length of catalytic tubing on the rate of formation of isopropanol. I
denote experimental data at 66 cm (6 mg Ru/C), ---I---- denote predicted data for 66 cm;
A denote experimental data for a tubing length of 33 cm (3 mg Ru/C), ----A--- denote
predicted data for 33 cm length of tubing; 0 denote experimental data from a repeat
experiment with 66 cm (6 mg Ru/C) length of tubing, ----- e---- denote predicted data
for repeat experiment with 66 cm length of tubing. All reactions were run in the
recirculating mode at 0.25 ml/min, 40 deg C and 650 psi Hz with a solution volume of 7.5
ml.

In the case of Langmuir-Hinshelwood type kinetics, in the higher concentration ranges,
the reaction is likely to be zero order in the reactant concentration. Accordingly, the
initial rate should not change signiﬁcantly upon increasing the intial concentration of

acetone ﬁom 30 mM to 180 mM. However, we see a four fold increase in the intial rate

(Table 12).

94

Table 12 . Initial rates as affected by initial concentration of acetone.

 

 

 

 

Initial rate if the
Initial concn of acetone Experimental initial rate
. reaction is ﬁrst order
(mM) moll(ht.hr) _
moll(ht.hr)
5 0.001 0.001
30 0.004 0.006
180 0.019 0.036

 

 

 

 

 

This behaviour is reﬂected in the predictions of the model for the 180 mM run.

002%-
(MH8~ I
00u3~
0004- .
00“:-
0.01 l
(umai I

00m5« I ,.ar‘1
0.004- -3..-“ i

,I
I , ~ ,- a.
I ‘I .,veeeeo
‘5‘ ,..c-H.

6
0mm— ,1! ,,,.-+«0585~ e
10!.+*‘°'
O'L‘ r

l
0 0.5 1 1.5 2 2.5 3 3.5 4
Time (hours)

 

 

IPA Concentration (M)

 

I T 1' I T

Figure 43. Effect of initial acetone concentration on the rate of formation of isopropanol.
I denote experimental data at 180 mM initial concentration of acetone, ---I--- denote
predicted data for 180 mM initial concentration of acetone; A denote experimental data
for a tubing length of 30 mM initial acetone concnetration, ----A----- denote predicted
data for 30 mM starting acetone concnetration; 0 denote experimental data from 5 mM
starting acetone conentration , ------ e ------ denote predicted data for repeat experiment
with 5 mM intial acetone concentration. All reactions were run in the recirculating mode
at 0.25 mein, 40 deg C and 650 psi Hz with a solution volume of 7.5 ml.

95

The parameters obtained from least squares regression which give the best ﬁt for the
model are listed in Table 13 below:

Table 13. Parameter values giving best ﬁt for the experimental data from Tubing].

 

 

 

 

Parameter Units Value
k298 m3 solution/ (gmcat. s) 2.25 x 10‘8
KA m3/kmol 138
AB kJ/mol 90

 

 

 

 

 

5.4.2 Results From Tubing 2

In order to study the effect of acetone concentration at a different catalyst loading,
a fresh piece of tubing was prepared. This piece of 800 um id. tubing was approximately
45 cm long and had 2 mg of Ru/C immobilized in it. Reactions were run at 25 deg C and

650 psi hydrogen at 0.25 ml/min.

5.4.2.1. Effect of acetone concentration
Reactions were starting with feed acetone concentrations of 3, 5, 9, 30 and 180

mM. The results of these runs are summarized in Figure 44.

96

0.02 r -

 

   
  
     

 

 

 

 

0.018 « ,
180mM acetone in feed
5 0.016 l
C
.2 0.014 4
ﬂ
.5;
5 0'012 l 30mM acetone in feed
8
0
g 0.008 “
5
°' 0 006
g ' l 10mM acetone in feed
3 0.004 . ’,’-A
0 002 , ’’’’’’’’’ 5mM acetone in feed
0 a ___.: 7.13.3559": ”5”": 11.51;- f ‘ ﬂ EnlMacetone in feed!
0 0.2 0.4 0.6 0.8 1 1.2

Time (hours)

Figure 44. Data from tubing 2, 2mg of Ru/C, at 25°C, in recirculating mode at 0.25
ml/min under 650 psi hydrogen Each line represents concentration proﬁle of isopropanol
based on a different feed concentration.

As can be seen in Figure 44, an increase in the feed acetone concentration results
in an increase in the net yield of isopropanol after 1 hour. Figure 45 shows the conversion

vs. time plots normalized to a zero time of reaction; this accounts for the induction period

observed at the low reaction temperature.

97

0.02 ”r

 

    
   
 

 

 

 

 

 

 

 

 

180mM acetone in
0.018 1 feed
2: 0.016
1:
.2 0.014
a
E
"’ 0.012 .
§ 30mM acetone In feed
3 0.01 «
o
g 0.008 '1
ﬂ
3‘ 0 006 «
g. 10mM acetone in feed
3 0.004 r
0.002 ~ 5mM acetone inf ed
0 , r . 3mM acetone in feed

 

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time (hours)

Figure 45. Data from tubing 2, 2mg of Ru/C, at 25°C, in recirculating mode at 0.25
ml/min under 650 psi hydrogen. A polynomial ﬁt has been used to estimate the start of
reaction (time = zero). Reaction volume = 7.5 ml.

Figure 46 shows the effect of change in acetone concentration of feed on the intial

rate of formation of isopropanol.

98

 

 

 

0.04 i ~—ﬁ ,7, , I , ,. w “__., i -¥ 7 i if “i ,7
0.035 j
0
0.03 1
8
=. 0.025
‘-
5 i
E: 0.02 ~ l
'6 .
2 ~ 0 E
o 0.015 1
1’ !
0.01 1 1,
0.005 . o 1
O. ;
0 r t r I r r r T T H
0 20 40 60 80 100 1 20 140 160 180 200

Initial acetone concentration (mM)

Figure 46. Effect of acetone concentration in feed on reaction rate as seen in data set 2
with tubing 2. 2mg of Ru/C, at 25°C, in recirculating mode at 0.25 ml/min under 650 psi
hydrogen. Reaction solution volume = 7.5 ml.

5.4.2.2 Kinetic Model

The data set from polynomial ﬁts for Tubing #2 was tested for ﬁt using the
Langmuir-Hinshelwood residence time model used for data set 1 (Equations 3 and 9). As
can be seen from Figure 47 below, the model predicts experimental data with an average

error of i 2%.

99

60 1 . , i I , ~ - - -—-

 

 

y=0.9838x
R2=0.9587
50 4
O
2 l ’ .
a
L
2 . °
5 e
O 30 Q 0
u . °
° —1
E 20 . .
e
e
10
o I j T T I
0 10 20 30 40 50

Experimental conversions

Figure 47. Parity plot comparing predicted and experimental conversions. Experimental
data are ﬁom polynomial ﬁts to data from set 2. Predictions are output from Langmuir-

Hinshelwood residence time model.

The parameters giving optimum ﬁt are listed in Table 14 .

Table 14. Parameters giving best ﬁt for data from Tubing 2

 

 

 

Parameter Units Value
k2.8 m3 solution/ (gmcat. s) 1.32 x 10“
KA m3/kmol l 8

 

 

 

 

 

100

5.4.2.3 Mass Transfer Analysis

The Weisz-Prater criterion78 was applied to calculate the observable modulus for
acetone and hydrogen inside the catalyst pores and thus gain an estimate of the role
played by mass transfer inside the pores of the catalyst particles. Sample calculations for
the observable modulus for acetone are shown below.

The observable moduls is calculated as per

__ — RGL2
— DeCB <1”

 

77¢2

For a reaction with 180 mM acetone at 25 deg C and 650 psi hydrogen with the
recirculating system, the values are as follows:

0 Rate of reaction

 

 

 

— RG =0.007 m°les"e“°’°d x 0.0075101101111101: x __L— x 1
litsolution 0.01gmcat 0.5X3600SCC
-RG = 2.91 x 10‘6 __EELJ 1,2 8““
gmcat x sec cm3cat
"RG = 3_5()x10'6 __3ﬂ01__
cm catxsec

0 The diffusivities of hydrogen and acetone in water were calculated using the Wilke-

Chang79 correlation:

T(V’H20 MH20 )0.5
0.6

 

—8
D acetone—H 20 Z 7'4 X 10
I” V acetone

where

101

T = absolute temperature [°K] = 298 K

(”H20 2 an association parameter for the solvent water = 2.26
M H20 = molecular weight of water = 18 g/mole

,u = viscosity of water in centipoises = 0.6552 cP
V acetone = molar volume of acetone = 75 cm3/gm-mole

This gives a value of diffusivity of acetone in water as

2 .
d
D _ 21.69 x 10‘5 ML
acetone H 20 sec
The effective difﬁisivity De takes into account the porosity of the catalyst

particles

_ 2
De — Dacetone—H20 X 8
where s = porosity of the catalyst z 0.6

_5 cm3ﬂuid

cmcat X SCC

Hence, for acetone in water, De =5.63 X 10

0 Concentration of acetone in water

mol

cm3 ﬂuid

C B = 0.00018
0 The characteristic length L of the spherical catalyst particle is Dp/6 where Dp is the

diameter of the particles, in this case 150 um. Thus

L = 0.0025 cm cat

Using the above values,

77¢zacetone = 0-022

102

I For hydrogen in water, the value of the diffusivity changes accordingly:

 

.5
D =74X10_8 T(WH20MH20)O
”VH2
where
V H2 = molar volume of hydrogen = 20.82 cm3/gm-mole

The value for diffusivity of hydrogen in water works out to

5 cm2 ﬂuid

DH2_H20 = 2.63 x 10 sec

The effective diffusivity takes into account the porosity of the catalyst particles
D - D x 82

e " acetone—1120
where 8 = porosity of the catalyst : 0.6

43 cm3 ﬂuid
cmcat X sec

 

Hence, for hydrogen in water, De :87] X 10

For the values

mol

cm 3601 X SCC

 

- —RG = 2.45 x 10*6

I L = 0.0025 cm cat

mol

- CB = 0.00003 3
cm ﬂuid

the value of the modulus works out to be

17,021.72 2 0.06

103

cogcowoham 0:88.44 o8 32.3 5232 0553030 98 Ewen—mic. Sumner—1 mmﬂz .2 033.

 

    

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

22.... B 33%. SEE $8 $8; SE m.~ 5E on SE on «H m 0 a8 8 2 9. SE 02 .E m... 5
Swan 5.0 ad unwed ed wommea $8.. $9. SE SN EE 2 SE on E S o moo ow SE n _E 2 S
2:: :22. 20
comm?” to $8.0 mod 8.33 $8 3». SE 5 :.E on SE on E 9 o moo mu SE 02 a 2 m
8mg... Rd F 386 «No mommod $3, $2 SE md :.E on SE on as mam 0 out mm SE a _E mg m
.528 .8 9:2 8mm: 33 3; 25m 52 8 22 8 as o... o 8e 8 2 8 :2 2: _e 2 a
35.. seam—
ouaao>m swag: coco
8&8; one 5 End 8; nomad So: $3 SE m.m 5E ow SE on E N 0 mon mu SE a _E mun. o
mo-m8.m em... md 88o Ed mommud Smm $9 SE «.9 E F SE on ma u 0 men mm SE 02 a E m
women? who md «wand and momma; $8 .53 SE or n V SE on E u 0 non mm SE 8 _E mg q
guano mg F 23.0 who commune .32 $5 SE 9m F. S SE on E a 0 men mm SE 9 _E mg m
8&8.“ 2.0 F .5de Ed oommuN $8? 3% SE 3 : S SE on E N 0 m8 mm SE m _E ms N
oomvvd «Nd E 33.0 mad oommné $3 $8 SE vé c S SE on E N 0 not mm SE n _E ms F
a 3.5 o 3|: a
.3 E N... Nec 2088 Ne: .wﬁonEoSBE 8: S u... e\. BEE E: :88 N... 0.580. SEEBEE 0:8 oE=_o> o: ...w
BnEovnoE mm. <n= 83m. 82QO
0.22:..1 VS E 5:00 3238
.xoanm
3.:voE 0.3-E030 «9.3:. Ea: 3.3.3.

104

A summary of such calculations for the 2nd set of reactions is shown in Table 15 above.

It can be seen that the reaction rates lie in a regime where diffusion within the
pores reduces the observed reaction rate by 25% as compared to the intrinsic reaction
rate. Higher acetone concentrations and lower reaction temperatures can reduce

diffusional limitations.

5.5 Conclusion
An immobilization technique to deposit catalyst in microchannels was
developed. The technique was used in conjunction with a low volume high pressure
recirculating system to demonstrate catalytic hydrogenation of acetone. The effect of
various parameters on the reaction rate was studied, and it was demonstrated that the
reaction is catalytic in nature and that hydrogen is not the limiting reactant under standard
reaction conditions. Comparison with an analogous batch setup demonstrates that the
catalyst exhibits similar activities in both the batch and continuous modes of operation.
The setup can be easily modiﬁed to act as a one pass ﬂow through system and thus has
potential application as a high throughput catalyst screening system A small amount of
catalyst, of the order of a few milligrams can be used to study catalytic reactions, which
can be very advantageous when screening catalysts.
Some of the advantages in using a PDMS tube as a microreactor instead of a
PDMS-glass mieroreactor are mentioned below. Connections between the tubing and
flow loop are considerably simpler and more robust as compared to connections between

a composite microreactor and the ﬂow loop. This saves a lot of development time as a

105

large proportion of prototypes failed due to leakage at the interface with the ﬂow loop.
Because glass and PDMS have different coefficients of thermal expansion, rapid heating
of a glass-PDMS mieroreactor can cause strain at the PDMS-glass bond. This can cause
distortion of the ﬂow path and in some cases breakage of the microreactor. In the case of
PDMS tubing, there are no such problems and the tubing can be heated to temperatures as
high as 200 °C. It is also possible to introduce more catalytic tubing into the ﬂow path
and thus check for the effect of increase in catalyst loading without having to discard the

original mieroreactor, as would be the case with the PDMS-glass microreactor.

Nomenclature

TCS = Space time in stirred tank = 1920 (m3 reactor . sec)/m3 solution
TPF- Space time in catalytic tubing (PFR) = 72 (m3 reactor. sec)/m3 solution
C A1 = concentration of acetone in the stirred tank at any time t, kmol/m3

C A2 = Concentration of acetone at the exit of catalytic tubing at any time t, kmo l/m3
k = Composite reaction rate constant, 8'1
K A = adsorption coefficient for acetone, m3/kmol

v = flow rate of reaction mixture, m3/s = 4.2 X 10'9 m3/s

VCST = Volume of stirred tank = 8 X 101’ m3

106

References

77. R.Alcala, J. Greeley, M. Mavrikakis, and J. A. Dumesic, J. Chem. Phys, 2002, 116,
20, 8973-8980

78. P. Weisz, C. Prater, Adv. Catal. 1954, 6, 143-196.

79. C. Wilke, P. Chang, AIChE J. 1955, 1, 264-270

107