.Xwn . u... W
.5me .. mt.
1-. .c ,

.- u.

 

i .ﬁﬁﬁﬁ
5%?

u .
:- .\l}.>. : 3.
Jung.

 

.12 a.

{8%

11.; z
-. .1
a. in».

 

- , l . _ ; . . . _ _ .
‘43...,_.~.€,.w.¥ A. ., . 4 ‘ 3. ._..
. .. A. ‘ T 12h. at...

 

z
:0

if};

 

LIBRARY
Michigan State
UniverSIty

 

 

 

This is to certify that the
dissertation entitled

ADSORPTION OF BIO-RENEWABLE SUBSTRATES ON

SUPPORTED METAL CATALYST IN WATER

presented by
LARS PEEREBOOM

has been accepted towards fulﬁllment
of the requirements for the

PhD. degree in Chemical Waring;

Major/Professor’s Signature

9/17/07

Date

MSU is an atﬁnnative-ection, equal-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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/07 p:lCIRC/DateDue.indd-p.1

ADSORPTION OF BIO-RENEWABLE SUBSTRATES ON SUPPORTED
METAL CATALYST IN WATER
By

Lars Peereboom

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Chemical Engineering

2007

ABSTRACT

ADSORPTION OF BIO-RENEWABLE SUBSTRATES ON SUPPORTED
METAL CATALYST IN WATER

By

Lars Peereboom

Biomass-based organic substrates are attractive feedstocks for chemicals
production because they are available in quantity and can undergo a variety of
reactions to form useful products. In our efforts to understand the mechanisms of
aqueous transformations of these substrates over supported noble metal catalysts,
we seek to probe interactions between the substrate and catalyst metal and between
the substrate and catalyst support. This dissertation recounts our work to measure
the extent of substrate adsorption on the catalyst surface and on the support based
on difference—Le. by measuring (via HPLC analysis) the initial and ﬁnal
concentrations of substrate in solution. We have developed a novel flow
microreactor system (recirculating batch reactor or RBR) with in-Iine substrate
detection and a high catalyst metal to solution ratio. Polyol and organic acid
interactions with both the unsupported Ru metal catalyst and the carbon support
have been examined, the former in the RBR and the latter in batch systems.

Volumetric gas phase chemisorption of H2 and CO onto unsupported Ru
catalyst gives 4.5 umol Hzlg Ru and 8.7 pmol COlg Ru, respectively. These
quantities are consistent with dissociative H2 chemisorption (H:M = 1). Quantities of
biorenewable substrates adsorbed on a “Hz-free” (evacuated) Ru surface from water

under argon at 25 ’0 are ~0.6 umol sorbitollg Ru, ~0.8 umol/g Ru for CS alcohols

and polyols, and ~4.0 umol/g Ru for C3 organic acids. Degradation products are
seen during adsorption of polyols, especially sorbitol. When the metal surface is pre-
saturated with H; and is under H2, no signiﬁcant adsorbtion of polyols is seen. In
contrast, the adsorption of acids is unaffected by the presence of hydrogen.
Sequential adsorption of glycerol (GO) followed by propylene glycol (PG) shows that
GO inhibits the adsorption of PG, while reversing the order of adsorption shows no
inhibition of GO adsorption by PG.

Single and multiple component physical adsorption on activated carbon
catalyst supports was examined over a broad range of concentrations (0.005 M to
1.5 M) and temperature (298 - 433 K). To obtain adsorption data relevant to glycerol
hydrogenolysis, adsorption of a combination of GO and PG (0.5 M total) was
examined over the same temperature range. An extended Langmuir isotherm model,
using single components’ adsorption parameters, accurately predicted the two-
component adsorption behavior.

Adsorption of combined lactic acid (LA) and propionic acid (PA) on activated
carbon was not predictable with an extended Langmuir model. We therefore
proposed a new model with two modes of adsorption: the ﬁlling of micro-pores, and
adsorption on macro-pore surfaces. When this new model was expanded to multiple
components, it accurately predicted the mulﬁcomponent LA + PA adsorption using
single component adsorption parameters.

Adsorption of biorenewable substrates on catalyst metal and on the catalyst
support has been characterized. Adsorption in both cases can signiﬁcantly inﬂuence

the rate of reaction of these biomass substrates to value-added products.

In memory of Ilse Morgan

ACKNOWLEDGMENT

I would like to thank Dr. Dennis Miller for his guidance and support as my
research advisor through my Ph.D. studies (especially the editing). I also thank
Dr. James E Jackson, Carl T. Lira, Lawrence T. Drzal, and Greg M Swain for
serving on my graduate committee. Special thanks to Lisa Benton (my wife) for
emotional and ﬁnancial support. She put up with me for many years of graduate

school.

TABLE OF CONTENTS

LIST OF FIGURES .............................................................................................. X
LIST OF TABLES ............................................................................................. XVII
1 INTRODUCTION ............................................................................................. 1
1.1 Signiﬁcance ............................................................................................. 2
1.2 Objectives ............................................................................................... 3

2 LITERATURE REVIEW ................................................................................... 5
2.1 Hydrogenolysis of glycerol ...................................................................... 5
2.1.1 Kovacs’s et al. ................................................................................ 5

2.1.2 Lahr and Shanks ............................................................................ 6

2.1.3 Tomishige et al. .............................................................................. 7

2.1.4 Suppes et al .................................................................................... 8

2.1.5 Rosier et al. .................................................................................... 9

2.1.6 Perosa and Tundo ........................................................................ 10

2.2 Aqueous phase adsorption on activated carbon ................................... 11
2.3 Species adsorbed on metal surfaces from liquid phase ........................ 12
2.4 Indirect differential detection ................................................................. 13

3 NOVEL FLOW MICROREACTOR FOR SPECIES ADSORPTION ON METAL
CATALYSTS ................................................................................................. 14
3.1 Design strategy ..................................................................................... 14
3.2 System description ................................................................................ 16
3.3 System evaluation ................................................................................. 18
3.3.1 System mixing .............................................................................. 18

3.3.2 Volume determination of reactor, tank, loop, and system ............. 19

vi

3.3.3 HPLC calibration ........................................................................... 20

3.4 Methods of operation ............................................................................ 23
3.4.1 General description of experiment ................................................ 23
3.4.2 Data processing ............................................................................ 24

CATALYST CHARACTERIZATION .............................................................. 29

4.1 Methods ................................................................................................ 29

4.2 Results .................................................................................................. 30

H2lDzo EXCHANGE ..................................................................................... 34

5.1 Experimental ......................................................................................... 34

5.2 Modeling ............................................................................................... 37
5.2.1 Approach to solutions ................................................................... 39
5.2.2 Parameter optimization ................................................................. 41
5.2.3 Results and conclusions ............................................................... 43

AQUEOUS PHASE ADSORPTION ON RUTHENIUM SPONGE .................. 45

6.1 Sorbitol, glucose and gluconic acid adsorption ..................................... 45

6.2 Glycerol Adsorption ............................................................................... 52
6.2.1 Concentration dependence ........................................................... 52
6.2.2 Hydrogen saturation ..................................................................... 55
6.2.3 Temperature effects on GO adsorption ........................................ 58
6.2.4 Propylene glycol (PG), ethylene glycol (EG), and 1,3-propanediol

(1 ,3-PDO) adsorption .................................................................... 61

6.2.5 Binary adsorption (GO, PG) .......................................................... 64
6.3 lsopropanol and acetone adsorption ..................................................... 67
6.4 n-Propanol and ethanol ......................................................................... 72

vii

6.5 Organic acid adsorption ........................................................................ 74

6.6 Comparisons and conclusions .............................................................. 78
7 AQUEOUS PHASE ADSORPTION ON ACTIVATED CARBON ................... 80
7.1 Experimental methods .......................................................................... 80
7.1.1 Carbon characterization ................................................................ 80
7.1.2 Isothermal adsorption measurements ........................................... 81
7.1.3 Elevated temperature adsorption experiments ............................. 82
7.1.4 Analysis ........................................................................................ 83
7.1.5 Adsorption modeling ..................................................................... 83
7.1.6 Summary of experiments performed ............................................. 85
7.2 Glycerol and propylene glycol ............................................................... 87
7.2.1 Isothermal adsorption of GO+PG on ROX activated carbon ........ 87
7.2.2 Isothermal adsorption of GO+PG on 3310 activated carbon ........ 91
7.2.3 Elevated temperature adsorption .................................................. 92
7.2.4 Two component adsorption studies .............................................. 94
7.2.5 Conclusion .................................................................................. 100
7.3 Ethanol, n—propanol, EG, and 1,3-PDO adsorption ............................. 101
7.4 Organic acids ...................................................................................... 102
7.5 Multi component adsorption studies relevant to hydrogenation of LA and
PA to PG and n-prOH ......................................................................... 107

7.6 Additional notes on adsorption modeling ............................................ 119
8 CONCLUSIONS AND RECOMMENDATIONS ........................................... 122
8.1 Conclusions ........................................................................................ 122
8.2 Recommendations .............................................................................. 124

viii

 

APPENDICES ................................................................................................... 125

Appendix A: Master experiment list ............................................................. 126
Appendix B: Supplemental data .................................................................. 139
References .................................................................................................. 147

LIST OF FIGURES

Figure 3-1: Recirculating reactor system ............................................................ 15
Figure 3-2: Dimensions of 316 stainless steel reactor insert ............................... 16
Figure 3-3: Conﬁguration of in-line sampling system .......................................... 18
Figure 3-4: Pulse attenuation at 6mllmin for recirculating system. ...................... 19

Figure 3-5: Refractive index detector linear response curves for 10 pl injection of
HO-mix and acid-mix standard solutions. .................................................... 21

Figure 3-6: Comparison of auto injector vs. reactor system injector with HO-mix

(Sorbitol, GO, EG, 1,3-PG, MeOH, EthOH, IPA, n-prOH) ............................ 22
Figure 5-1: Corrected and uncorrected data for H2 over D20 with 0.1 g 5% RuIC
(Exp 43B), and H2 over 020 with 0.1 g 5% Pth (Exp 43C) ......................... 35
Figure 5-2: Incorporation of deuterium into the gas phase from (DO)ethanol over
0.1 g 5% Ru/C (10 ml reactor 100 psi H2 Exp 47) ....................................... 36
Figure 5-3: H2 + D2 over D20 with 0.1 g 5% RulC (50 psi each in 70 ml reactor,
Exp 50) ........................................................................................................ 37
Figure 5-4: Proposed reaction model. ................................................................. 38
Figure 5-5: Reversible reaction model ................................................................ 40

Figure 5-6: Model ﬁt to experimental data for 5% Ru/C (Exp 43B). Dotted lines
are predicted values ..................................................................................... 44

Figure 5-7: Model ﬁt to experimental data for 5% Pth (Exp 43C). Dotted lines
are predicted values ..................................................................................... 44

Figure 6-1: Adsorption of sorbitol on 10 g H2-free Ru sponge at 25 °C. Initial
loading of 5 mg (27.5 umol) and 10 mg (55 pmol). ...................................... 46

Figure 6-2: Adsorption of sorbitol at 25 °C on 10 g Ru sponge using stepwise
dosing (Exp 20). Concentration of other species is in sorbitol equivalents. .48

Figure 6—3: Adsorption of glucose at 25 °C on 10 g Ru sponge using stepwise
dosing (Exp 23). Concentration of other species is in glucose equivalents..49

Figure 6-4: Effects of H2 addition on adsorbed glucose at 25 °C on 10 g Ru

sponge aﬂer stepwise dosing (Exp 23) ........................................................ 50
Figure 6-5: Adsorption of gluconic acid at 25 °C on 10 g Ru sponge using
stepwise dosing (Exp 25) ............................................................................. 51
Figure 6-6: Adsorption of glycerol at 25 °C on 10 g Ru sponge comparing
stepwise (Exp 24) and ﬁxed dosing (Exp 26). Left solution
concentration; right quantity adsorbed. ........................................................ 52

Figure 6-7: Adsorption of GO at 25 °C using (A) ﬁxed dosing: maximum initial
concentration vs. total adsorbed, and (8) ﬁnal solution concentration vs.
amount adsorbed. ........................................................................................ 53

Figure 6-8: Adsorption of GO at 25 °C (Exp 67) on Ru sponge prepared with
excess water. For clarity EG and PG concentrations are plotted at 10X ...... 54

Figure 6-9: Adsorption of GO at 25 °C (Exp 68) on Ru sponge prepared by
reduction on the Micromeritics ASAP 2010 instrument. For clarity EG and
PG concentrations are plotted at 10X. ......................................................... 55

Figure 6-10: Comparison of adsorption of GO at 25 °C on H2-saturated (Exp 28)
and H2-free (Exp 39) Ru sponge (10 g). ...................................................... 56

Figure 6-11: Effects of higher GO loading on adsorption at 25 °C on H2-saturated
(Exp 100) and H2-free (Exp 93) Ru sponge (20 g). ...................................... 56

Figure 6-12: Adsorption of G0 on H2-saturated Ru sponge (10 9) under argon at
50 °C (Exp 30). ............................................................................................ 57

Figure 6-13: Temperature dependence of GO adsorption. (50 pmol GO loading,
20 g Ru sponge). Inset shows magniﬁcation of upper region of graph. ....... 58

Figure 6-14: Glycerol decomposition products vs temperature. Multiple
experiments are included at 25 and 60 °C. .................................................. 60

xi

Figure 6-15: Adsorption of G0 (50 pmol loading) at 80 °C on 20 g Ru sponge
(Exp 96) showing the effect of adding H2 and increasing temperature to 200
°C. Quantity GO adsorbed shown on lower graph, solution concentrations on
upper graph. Concentrations of E6 and PG are plotted on right axis. ......... 61

Figure 6-16: Adsorption of PG at 25 °C on 10 g Ru sponge comparing loading
and H2 saturation. ........................................................................................ 62

Figure 6-17: Adsorption of EG (80.6 pmol) on 10 g Ru sponge at 25 °C. ........... 63
Figure 6-18: Adsorption of 1,3—PDO (50 umol) on 20 g Ru sponge at 25 °C. ..... 63

Figure 6-19: Adsorption of PG followed by GO on 10 g Ru sponge at 25°C. A
Exp 42, initial PG loading of 6.6 pmol; and B: Exp 44, initial PG loading of 66
pmol. GO loading of 4.34 umol follows PG loading ...................................... 65

Figure 6-20: Experiment 45, adsorption of GO (6.5 umol) followed by PG (6.5
pmol) at 25 °C on 10 g Ru sponge ............................................................... 66

Figure 6-21: Experiment 49, adsorption of G0 (2.2 pmol) and PG (3.3 pmol) at
25 °C on 10 g Ru sponge ............................................................................. 66

Figure 6-22: Adsorption of IPA (50 umol) at 25 °C on 20 g Ru sponge (Exp 109)
showing effect of addition of H2 .................................................................... 67

Figure 6-23: Adsorption of IPA (50 pmol) at 25 °C on 20 9 H2 saturated Ru
sponge (Exp 111) under 20 psi H2. .............................................................. 68

Figure 6-24: Competitive adsorption of GO and IPA (50 pmol each) at 25 °C on
20 g Ru sponge (Exp 112) showing the effect of adding H2 ......................... 69

Figure 6-25: Adsorption of acetone (50 umol) at 25 °C on 20 g Ru sponge (Exp
115) showing effect of addition of H2. .......................................................... 70

Figure 6-26: Adsorption (50 pmol each) of acetone after GO adsorption at 25 °C
on 20 g Ru sponge (Exp 116) showing the formation of IPA. ...................... 71

Figure 6-27: Comparison of acetone adsorption (50 pmol) at 25 °C on 20 g Ru
sponge (Exp 116, 115) Showing the effects of pre-saturation with GO. Open
symbols are from presaturated experiment .................................................. 72

xii

Figure 6-28: Adsorption (50 pmol loading) of n-propanol (Exp 121) and ethanol
(Exp 114) at 25 °C on 20 g Ru sponge. Acetic acid (AA) and PA are formed.
..................................................................................................................... 73

Figure 6-29: Adsorption of n-prOH (50 umol/system) at 25 °C on 209 Ru sponge
(Exp 121) showing effect of addition of H2 and increase in temperature to 50
°C. Gas C in legend refers to total gaseous carbon. .................................... 74

Figure 6-30: Concentration dependence of LA adsorption at 25 °C on H2 free Ru
surface for different initial loadings. (Exp 54, 56, 57, 58) ............................. 75

Figure 6-31: H2 saturation effects on adsorption of LA at 25 °C on 10 g Ru
sponge metal (Exp 54,55*,56,60*). *denotes experiments on hydrogen
saturated surface. ........................................................................................ 76

Figure 6-32: Predicted vs. experimental pH as function of initial loading (umol). 77
Figure 6-33: Adsorption of equimolar mixture of LA and PA (120 umol/system) at

25 °C on 20 g of H2-free Ru sponge. (Exp 70) ............................................. 78
Figure 7-1: Experimental and predicted a) Langmuir and b) Freundlich adsorption
isotherms for PG and G0 at 25°C on ROX carbon. ..................................... 88
Figure 7-2: Adsorption of PG and GO on ROX carbon on a mass basis. ........... 90

Figure 7-3: Temperature dependence of Langmuir equilibrium constant KA for
GO and PG on ROX (A) and 3310 (B) carbon. ............................................ 93

Figure 7-4: Experimental vs predicted adsorption of PG and G0 on ROX carbon
and 3310 carbon for all data. ....................................................................... 96

Figure 7-5: Experimental and predicted Langmuir isotherms for mixtures of GO
and PG on ROX carbon at 25 °C and 0.05 M, 0.3 M and 0.5 M total species
concentration. Abscissa is fraction of GO in the total species (GO + PG)
present. Lines are the predicted isotherms. ................................................. 97

Figure 7-6: Experimental and predicted Langmuir isotherms for mixture of GO
and PG on ROX carbon at 0.5 M total species concentration and 80 °C, 120

xiii

°C, and 160 °C. Abscissa is fraction of GO in the total species (G0 + PG)
present. Lines are the predicted isotherms. ................................................. 97

Figure 7-7: Extended Langmuir model predictions of a) GO and b) PG pore
concentrations in ROX carbon at 180 °C. Symbols represent trajectory of
1.0 M GO conversion to PG. ........................................................................ 99

Figure 7-8: Experimental and best ﬁt data for isothermal adsorption of n-propanol
(n-prOH), ethanol, 1,3-propanediol (1 ,3-PDO), and ethylene glycol (EG) on
3310 carbon at 25 °C. (A-Langmuir, B-Fruendlich) .................................... 102

Figure 7-9: Experimental and best ﬁt data for lsothennal adsorption of formic,
acetic, and propanoic acid on 3310 carbon at 25 °C. (A-Langmuir, B-
Fruendlich) ................................................................................................. 103

Figure 7-10: Experimental and best ﬁt data for isothermal adsorption of glycolic
and lactic acid on 3310 carbon at 25 °C. (A-Langmuir, B-Fruendlich) ....... 104

Figure 7-11: lsotherrnal adsorption of equimolar mixtures of PA+LA (A) and
PA+PG (B) on 3310 carbon at 25 °C. ........................................................ 106

Figure 7-12: Isothermal Adsorption of equimolar mixtures of PA+PG (A) and
LA+PG (B) on 3310 carbon at 25 °C .......................................................... 106

Figure 7-13 : Low concentration adsorption deviation from Langmuir isotherm at
low concentration for LA, PA, PG, and n-prOH on 3310 carbon at 25°C. .. 108

Figure 7-14: Experimental data and WDL ﬁt for the isothermal adsorption of PA,
LA, n-prOH and, PG on 3310 carbon at 25 °C ........................................... 112

Figure 7-15: Variable temperature adsorption of 1000 mM LA on 3310 carbon.
................................................................................................................... 1 1 3

Figure 7-16: Experimental vs predicted single component variable temperature
adsorption PA, LA, n-prOH and, PG on 3310 carbon from 25-160°C ........ 113

Figure 7-17: Experimental and predicted DWL isotherms for mixture of LA and
PA on 3310 carbon at 0.5 M total species concentration and 25 °C, 80 °C,

xiv

and 150 °C. Abscissa is fraction of LA in the total species (LA + PA) present.
Lines are the predicted isotherms. ............................................................. 114

Figure 7-18: Adsorption isotherm for equimolar mixture of LA and PG on 3310
carbon at 25 °C .......................................................................................... 115

Figure 7-19: Adsorption isotherm for LA and PG on 3310 carbon at 25 °C ...... 116

Figure 7-20: Experimental and predicted DWL isotherms for mixture of LA and n-
prOH on 3310 carbon at 0.5 M total species concentration and 30 °C, 80 °C,
and 150 °C. Abscissa is fraction of n-prOH in the total species (LA + n—prOH)
present. Lines are the predicted isotherms. ............................................... 116

Figure 7-21: Experimental and predicted DWL isotherms for mixture of PA and n-
prOH on 3310 carbon at 0.5 M total species concentration and 30 °C, 80 °C,
and 150 °C. Abscissa is fraction of n-prOH in the total species (PA + n-
prOH) present. Lines are the predicted isotherms. .................................... 117

Figure 7-22: Experimental and predicted DWL isotherms for mixture of PA and
PG on 3310 carbon at 0.5 M total species concentration and 30 °C, 80 °C,
and 150 °C. Abscissa is fraction of PG in the total species (PA + PG)
present. Lines are the predicted isotherms. ............................................... 118

Figure 7-23: DWL predicted adsorptions of PA as percentage of actual adsorbed
on 3310 carbon at initial concentration from 10 mM to 2000 mM. ............. 118

Figure 7-24: Experimental vs predicted adsorption of PA, LA, n-prOH and, PG on
3310 carbon in binary mixtures only. ......................................................... 119

Figure 7-25: Apparent species adsorption in control experiments without carbon
showing the need to correct for vapor fraction of n-prOH. (note data for 150
°C after 14 hours is plotted at 160°C for clarity.) ........................................ 120

Figure B-1: Experimental and predicted Langmuir isotherms for mixture of GO
and PG on 3310 carbon at 0.5 M total species concentration and 80°C,
120°C, and 160°C. Abscissa is fraction of GO in the total species (GO + PG)
present. Lines are the predicted isotherms. ............................................... 139

 

Figure B-2: Formula matrix calculation of number of reactions needed. ........... 139

Figure B-3: Matrix of coefﬁcients for ODE’s ...................................................... 140

LIST OF TABLES

Table 3-1: Retention times and relative response factors on a weight basis. ..... 23

Table 3—2: Calculated values for a hypothetical adsorption experiment on Ru

sponge. ........................................................................................................ 27
Table 41: Physical characterization of Ru sponge. ............................................ 30
Table 4-2: Gas chemisorption on Ru sponge ...................................................... 32
Table 5-1: Assumptions made for H2-D2O exchange model ............................... 38
Table 5-2: Species included in H2, 020 exchange model ................................... 39
Table 5-3: Optimized parameters for experiment 43B and 430 .......................... 43
Table 6-1: Contrasting adsorption properties on Ru sponge of organic acid and

polyol. .......................................................................................................... 79
Table 7-1: Carbon characterization by N2 adsorption at 78 K ............................ 81

Table 7-2: General notation used for modeling adsorption on activated carbon.85

Table 7-3: Summary of experiments preformed during the study of aqueous

phase adsorption on activated carbon. ........................................................ 86
Table 7-4: Langmuir and Freundlich isotherm coefﬁcients for GO and PG on

ROX carbon. ................................................................................................ 89
Table 7-5: Comparison of Langmuir isotherm coefﬁcients for ROX carbon and

3310 carbon. ................................................................................................ 92
Table 7-6: Langmuir isotherm coefﬁcients for GO and PG on ROX carbon ........ 94

Table 7-7: Optimized isothermal adsorption parameters for the DWL model....110
Table 7-8: Optimized isothermal adsorption parameters for the DWL model....111

Table 7-9: Percentage range of optimized parameters whose objective function is
within 0.5% of the “best ﬁt” ......................................................................... 112

Table B-1: Unique rate terms for elementary reactions ..................................... 140

xvii

1 INTRODUCTION

The Miller group at Michigan State University focuses on the evolving
technology of renewable resource reﬁning. We seek to produce commodity
chemicals from readily available bio-based feedstocks. Recent group members
studied itaconic acid production,‘ lactic acid hydrogenation to propylene glycolz,
stereoretentive amino alcohol synthesis3, the processing of sugar alcohols to
value-added polyols" 5, and the hydrogenation of glycerol to propylene glycol.‘3

Our current focus is to understand the mechanism of aqueous
hydrogenolysislhydrogenation of bio—based intermediates (polyols, organic acids,
amino acids, alkyl amines, amides, ...) over noble metals. For example, in
glycerol hydrogenolysis the ﬁrst step in the proposed mechanism (Scheme 1-1)
is the dehydrogenation and adsorption of the substrate as an aldehyde-like
intermediate. Although the exact mechanism of adsorption is not known, we seek
to probe interactions between substrate molecules and the metal catalyst surface
to better understand binding energy, adsorbed structure, and competition for

surface sites.

Scheme 1-1: Proposed sequence of steps for GO hydrogenolysis to EG, PG and

 

LA
OH -H2
Hovk/OHv—T: :—H2 Ho\)\/k
60 GA
retro
[1 aldol
OH ke. t
/O 9' 20C,OH
Glycolcldehyde Ethane-1 ,2-diol
+ Formaldehyde
-H2 11 +H2
OH
bOH . -‘3c
k.e.t.= keto-enol
EG tautomerization

 

1.1 Signiﬁcance

=MHO\i',OH=—‘ ket HO\/DI:K/OH

Trioce encdiol
+H20' —H2O
OH k.e.t. 0
/ T‘ N0
PV enol

‘5 PV
$6 -2H2ﬂ +2H2

Global energy demands keep increasing, fueling development of

alternative renewable energy supplies and bio-based chemical intermediates.

Aqueous phase processing and/or conversion of these bio-based materials is

becoming important because of low environmental impact. Unfortunately, little is

known about aqueous phase processing/conversion of bio-based materials like

glycerol. Many of these processes use metal catalysts which follow the general

mechanism in Scheme 1-2. The ﬁrst step is adsorption of reactant (A) onto the

catalyst metal site (M). Even if this is not the rate determining step it is still

involved in the reaction kinetics. Understanding the mechanism of this adsorption

in aqueous media is important. The other steps are conversion of reactant to

product (B) on the metal and then product desorption from metal

Scheme 1-2
A+ M——* A*-M (+C) -——> B*-M ——> B+ M
For example, the side product of biodiesel formation is glycerol, which has
low fuel value and limited industrial uses. On the other hand, the conversion of
glycerol to propylene glycol is a good value adding process because propylene
glycol has much greater industrial potential as a solvent, heat-transfer ﬂuid, and
as a chemical feed stock. Optimizing the conversion of glycerol to propylene
glycol is thus an important goal, one which would greatly beneﬁt from a detailed

understanding of the reaction mechanism.

1.2 Objectives

Understanding the adsorptive properties on active metals and supports of
glycerol, glycols, organic acids, and related substrates derivable from renewable
resources will give insight into the reactivities of our catalysts. To develop such
an understanding, the following objectives need to be accomplished.

1. Develop a system to measure the adsorption of bio-based substrates

on metal surfaces under elevated pressures and temperatures.

2. Study the adsorption of glycerol, propylene glycol, ethylene glycol and

others individually and in mixtures on ruthenium sponge metal.

3. Study the adsorption of these same species on the carbon support and

develop a model for this adsorption.

 

 

 

4. Compare the reactivity and selectivity between supported and
unsupported ruthenium catalysts for conversion of selected
biorenewable feed stock species, with emphasis on correlating

reactivity with the adsorption results.

 

 

2 LITERATURE REVIEW

2.1 Hydrogenolysis of glycerol

The hydrogenolysis and degradation of polyols has been studied
extensively. The dissertation of Mike Shafer from our group has a large collection
of references and patents.5 Selected patents dealing with hydrogenolysis of
glycerol include Crabtree’s 2005 patent7 that mainly focuses on hydrogenolysis
of sugar alcohols, but also includes glycerol. Their system uses an organic
phosphine and ruthenium or osmium. The use of organic phosphine alone makes
this impractical. Suppes’ 2005 patent“ has a more practical approach and will be
discussed below.

Glycerol hydrogenolysis over ruthenium on carbon (Ru/C) has been
studied by several groups. Kovacs et al. in our group did detailed qualitative
mechanistic studies using deuterium labeling.6 Lahr and Shanks focused on
kinetics under neutral to basic conditionsg' 1° while Tomishige et al. looked at
addition of Amberlyst 15, a strong cationic resin.11 Suppes et al.12 and Rosier et
al.” looked at a range of metal catalysts while Perosa and Tundo used Raney

nickel.“

2.1.1 Kovacs’s et al.

Kovacs et al.6 propose a mechanistic pathway for catalytic hydrogenolysis
of glycerol to propylene glycol, ethylene glycol, and lactic acid over ruthenium on

carbon (Scheme 1-1). The ﬁrst step is dehydrogenation of glycerol to a

glyceraldehyde analog, followed by one of two paths: (1) cleavage to the enediol
of hydroxyacetaldehyde, which, with or without tautomerization, is reduced to
ethylene glycol, or (2) keto-enol tautomerization to the enediol, which then
dehydrates and tautomerizes to pymvaldehyde. Subsequent hydrogenation or
water addition and rearrangement leads to propylene glycol or lactic acid,
respectively. The presence of base promotes enolization and dehydration of the
glyceraldehyde, favoring propylene glycol and lactate formation. Kovacs et al.
substantiated this reaction mechanism by conversion, isotopic labeling, reaction

of proposed intermediates, and computational molecular modeling studies.

2.1.2 Lahr and Shanks

Lahr and Shanks studied the hydrogenolysis of glycerol (GO) over 5%
ruthenium on carbon with emphasis on understanding the reaction mechanism.
As a control, they looked at the degradation of propylene glycol (PG) and
ethylene glycol (EG).1o The degradation of both EG and PG were independent of
solution concentration (i.e. zero order), while initial rate data for GO conversion
gave a reaction order of 0.45 to 1. They also reported that the presence of EG
inhibited the conversion of GO, while PG had no effect. Studies of mixtures of EG
and PG suggested that EG forms a stronger bond than PG with the Ru surface,
thus explaining why EG inhibits GO conversion. Based on these results, they
developed a reaction model based on Langmuir—Hinshelwood kinetics where the
polyol adsorbs on Ru, undergoes dehydrogenation to form an aldehyde

intermediate, desorbs into the solution, and then undergoes degradation with the

base. They proposed that the rate limiting step is the dehydrogenation of the
adsorbed polyol.

In 2005, Lahr and Shanks expanded on this system by studying the
effects of sulfur and temperature.9 The degradation rates of PG and EG are
similar when they are reacted individually. The apparent activation energy and
the relative reaction rates were unaffected by the addition of sulfur. The
degradation rate becomes immeasurable at SIRu ratios >0.5. The addition of
sulfur only seems to block the active sites and does not alter the degradation
chemistry. The relative degradation rate of EGzPG was 4:1 in an equimolar
mixture and was unaffected by temperature. This indicates strong competition
between EG and PG on the metal surface.

For the reaction of glycerol over Ru exposed to sulfur, the apparent
activation energy increased while the absolute rate decreased greatly. The
selectivity to PG increased with increase in sulfur concentration on the catalyst.

l‘5. Shanks proposes that there

This effect was also reported by Montassier et. a
are two different pathway to PG; the surface dehydration of glyceraldehyde
becomes more favored than the solution phase dehydration reaction upon

addition of sulfur.

2.1.3 Tomishige et al.

In 2005, Tomishige and coworkers11 tested the hydrogenolysis of GO over
Ru, Pt, Pd, and Rh on carbon without any acid or base at 180 °C and 80 bar for

10 hours. Under these conditions, the RuIC gave 6.3% conversion with 18%

selectivity to PG while the other metals gave conversions of 1% or less but with
higher selectivity.

They studied RuIC with a range of solid acids of which Amberlyst 15 gave
the best conversion, 40% at 140 °C and 80 bar H2. Selectivity to PG was low at
43%. Decreasing the temperature to 120 °C did improve the selectivity to PG
(60%) but conversion was only 33% after 40 hours. Curiously, the use of sulfuric
acid resulted in only 3.2% conversion under the conditions that gave 43% with
Amberlyst 15. They suggest that H2804 interacts with Ru surface sites, perhaps

being reduced to sulﬁde that adsorbs and blocks GO access.

2.1.4 Suppes et al.

Suppes et al.12 focused on the use of copper-chromite as a
hydrogenolysis catalyst because their reactions with Ru, Pd, Pt, Ni, and Cu
catalysts showed that copper-chromite gave 55% conversion and 85% selectivity
after 24 hours at the following conditions: 80% GO in solution; 5% bulk copper-
chromite; 200 °C, and 200 psi H2. They propose a two step mechanism where
glycerol is dehydrated to acetol, and acetol is reduced to PG. They ﬁled a patent8
on a two step process where glycerol is dehydrated under reduced pressure to
form acetol. This acetol is then hydrogenated to PG at the same temperatures
with the addition of H2 (200psi). No mechanistic details were suggested for the
dehydration.

All experiments were done in a batch reactor and only conversion of GO
and selectivity to PG after 24 hours was reported. Solution pH and the possible

presence of lactic, acetic, formic, and other acids was not mentioned. This is

8

somewhat troubling, because they also report that the catalyst loses activity
before the reaction goes to completion. They performed a parametric study to
understand the effects of catalyst reduction temperature, catalyst loading,
reaction temperature, hydrogen pressure, and initial water content. A reduction
temperature of 300 °C gave the best conversion and selectivity. Conversion
increased with catalyst loading, but selectivity had a maximum of 85% at 5%
loading. They speculate that higher loadings result in increases in the
decomposition of the product. The effects of temperature follow the same trend
as the catalyst loading with a maximum at 200 °C. For these cases, it would be
beneﬁcial to have the selectivity and conversion at different times in the reaction.
Increasing hydrogen pressure from 50 to 300 psi resulted in a steady increase in
the conversion and selectivity. Conversion was 65% with a selectivity of 90% at
300 psi, the maximum value they examined. Decreasing the initial water content
increased the conversion but the selectivity went through a maximum of 85% at
20% water. In all cases, they reported that, except for an H2 pressure of 300 psi,
the selectivity decreased when the conversion was >55%. This might be a clear

indication that PG degradation is important, as reported by Shanks et al.10

2.1.5 Rosier et al.

Rosier et al.13 looked at hydrogenolysis of ~20 wt% GO in water over
CuOIZnO, PdIC, RhIC catalysts at 180 °C and 80 bar H2. Their conversions were
generally <30% after 168 hours and PG selectivities <70%. Rth with H2WO4
gave best selectivity to PG. In switching from water to sulfolane as solvent, the

1,3-propanediol (1,3 PDO) was favored over PG by 2:1, but the 1,3—PDO

9

selectivity was only 12%. The observed rate and selectivity changed if the
stainless steel autoclave was lined with Teﬂon. The lined reactor showed
conversions ~twice that of the unlined reactor, and the 1,3-PDOIPG ratio
increased from 0.1 to 0.5. Metal leaching was conﬁrmed by testing the reaction
mixtures. The addition of iron salts decreased the conversion and increased the
1,3-PDOIPG ratio. The addition of copper salts drastically reduced the
conversion while still increasing the 1,3-PDOIPG ratio. They proposed a reaction
scheme that includes several different pathways to PG and 1,3-PDO. The
complexation of the 1,2- or the 1,3-OH groups to the metal and subsequent

hydrogenolysis of the “free” hydroxide is proposed.

2.1.6 Perosa and Tundo

Perosa and Tunclo14 used wet Raney Ni as catalyst and neat glycerol at
high loading (29 Ni to 89 GO) at 190 °C and 10 atm H2, and obtained 63% GO
conversion and selectivity to PG of 77%. The only byproducts that they reported
are CO2 and ethanol, identiﬁed by GC-MS and NMR respectively. CO2 was
identiﬁed by GC-MS, but quantities were based on mass loss. Addition of 2 g
tetrabutylphosphonium bromide or (C6H13)3014H29PCI decreased the conversion
and selectivity. It should be noted that these ionic liquids created a two-phase
system at reaction conditions. The lack of 1,3—propanediol was used as evidence
that it was not formed in the reaction scheme, but the absence of 1,3-PDO could

also result from its rapid degradation as shown by Tomishige".

10

2.2 Aqueous phase adsorption on activated carbon

The literature is ﬁlled with adsorption studies, but very few deal with
aqueous phase adsorption of small molecules on solid adsorbents. Most deal
with adsorption of large bio molecules which are relatively easy to detect and will
not be discussed in this paper. An extremely useful source is
http://adsorption.org/awm/ by Adam W. Marczewski, Ph.D. This site has a great
general overview of adsorption and detailed model descriptions and references.
Several good reviews exist“"18 that cover general solution phase adsorption.

For modeling single species adsorption, several common isotherm models
are used including Langmuir, Freundlich, and Sips (Generalized Freundlich).
Equations for these isotherms are deﬁned by Marczewski and are shown in
Equations 2-1, 2-2, and 2-3.

. q kC
Lan m r =-—'"———— 2-1
9 UI q H kc ( )
- 1/
Fruendlich q = kFC " (2-2)
%
Sips q = i (2.3)
1+kC%

The Langmuir isotherm is based on monolayer coverage of surface sites
that are energetically identical. The Fruendlich isotherm assumes that adsorption
energy increases as the quantity adsorbed increases, but there is no maximum

coverage. The SIPS model combines both Langmuir and Fruendlich isotherms to

11

allow the system to have a maximum coverage and a variable energy of
adsorption.

Several recent studies have examined small molecule adsorption on
activated carbon '9‘” and on polymer supports“ 324“. Of particular interest is
Moon’s28 work on adsorption of sugars on activated carbon. They found that the
adsorption could be modeled using either Fruendlich or SIPS isotherms,
including binary adsorption using single component parameters. The adsorption
of disaccharides was greater than any monosaccharide, with maltose adsorbing

to the largest extent.

2.3 Species adsorbed on metal surfaces from liquid
phase

The literature is ﬁlled with studies of gas adsorption onto metal surfaces,
but these are not generally relevant to our work. Several spectroscopic methods
exist for the study of adsorption from the liquid phase to active metal surfaces.
The most promising is attenuated total internal reﬂection (ATR-IR), which has
been used in our labs but not in this work. A general overview of ATR-IR
methods and references applicable to the overall thrust of research on
hydrogenolysis is provided but will not be discussed?“2 Surface-enhanced
Raman scattering (SERS) is also used to examine metal adsorbate interactions.
Several good reviews exist but are again outside of the scope of this work.“47

X-ray photoelectron spectroscopy (XPS) is extensively used to study

surface composition and has been applied to catalysis but is not as applicable for

12

in situ measurements, particularly liquids, because it is a high vacuum technique.
Several reviews on this method are available.“56

Single crystal surfaces are extensively used in gas phase studies and
have also been applied in liquid phase environments. One particularly interesting
in-situ study stands out.57 Wang et al. used cyclic voltammetry to oxidize and
reduce Ru (001) surface while simultaneously measuring x-ray reﬂectance. They

propose that water is chemisorbed on the metal, possibly as a hydroxide.

2.4 Indirect differential detection

Indirectly detecting quantities adsorbed on solid materials by measuring
differences in liquid-phase concentrations under controlled conditions is one of
the oldest and simplest methods. Batch reactors are widely used, either with in
situ monitoring or periodic sampling. Shallow bed adsorption is practical when
quantities adsorbed are signiﬁcant."3 No literature has been found covering the
adsorption of small organic molecules on bulk sponge metals. Only problems
with adsorption on exposed metal in HPLC system are mentioned, and these do

not look at quantifying the actual species adsorption.

13

3 NOVEL FLOW MICROREACTOR FOR SPECIES
ADSORPTION ON METAL CATALYSTS

3.1 Design strategy

One challenge with measuring species adsorbed on the supported metal
catalyst is that the ratio of support area to metal surface area is large. In the case
of 5% Ru/C, this ratio is >400:1. This large ratio makes it impossible to
distinguish adsorption on the metal from that on the support; for mat reason we
have to look at adsorption on pure metals only. Reactivity, particle size, and
surface area are some of the criteria used in selecting the metal to use. In the
case of thhenium, three forms are commercially available, solid (ingot, bar, foil),
sponge metal, and powder (Ru black). Ru black has the largest surface area
(~10-15 m2/g), unfortunately the particle size is too small to be handled easily
and would result in metal loss or plugged ﬁlters if used in a ﬂow system. This
leaves ruthenium sponge metal which has a low surface area (~0.2 mzlg Ru).
This low surface area means that the quantity of substrate adsorbed is extremely
small (<1mglg Ru). Maximizing the metal to solution volume ratio is therefore
critical; this and the high cost of Ru metal has led us to develop a reactor system
that meets the following design parameters:

1. Low total volume <15 ml

2. Metal capacity up to 20 g

3. Reactor operating temperature range 20 - 200 °C

4. Solid liquid mixing and gas liquid mixing
14

5. Addition of two different gases at pressures up to 400 psi
6. Online sampling without disturbing the system
7. Option to ﬂush system with different liquids
The reactor system shown in Figure 3-1 meets these criteria. The system

consists of an HPLC pump, tubular reactor, inline sampling system, a “tank" (gas

liquid inter-phase), and valving to isolate the reactor.

HPIC vent
systan I 8

 

 

 

 

 

—0[ Cooling coil

 

 

 

 

Oil bath (25 -200 °C)

 

   

 

 

 

  

t:

8

5

Ta

0 .

.3 Er":

:5 Heat

5 exchanger
‘3 (coil)
82

 

 

 

 

 

 

 

Figure 3-1: Recirculating reactor system

A ﬂuidized bed reactor was chosen to allow large solid to liquid ratio with
good mixing. Initially, a simple 10 mm ID column was used. Unfortunately, in this
reactor conﬁguration little axial mixing was observed at 3-6 mllmin. The relatively
large particles and high density of ruthenium sponge metal prevented ﬁuidization.
Calculations showed that at 25 °C the bed should be ﬂuidized at ~2 mllmin while

15

at 100 °C more than 6 mllmin was needed. The general solution to this problem
is a tapered reactor that allows a velocity gradient Internally tapering a small 316
stainless steel reactor or insert proved to be a technical barrier. Machining a
conical insert, on the other hand, was relatively simple, and thus this route was

taken to produce the reactor.

3.2 System description

The circulation pump is a Bio-Rad 1350 HPLC pump with the intake
modiﬁed with a simple t-ﬁtting to decrease the volume and allow operation under
pressure. The output from the pump goes through a home-made pressure
damper consisting of a coil of ﬂattened 1116 SS tubing. The pressure damper is

needed to prevent excess pressure buildup during valve switching.

10 mm ‘216 mm

Figure 3-2: Dimensions of 316 stainless steel reactor insert.

I 100mm I
l

20mm

    

 

The reactor consists of a 10 mm x 100 mm SS HPLC column with 2.0 pm frits at
the ends. The machined conical insert (Figure 3-2) is placed in the column to
create a semi-ﬂuidized bed in the reactor. This was needed for mixing purposes.
A coil of SS tubing was placed before the reactor to act as a heat exchanger to
preheat the liquid entering the reactor. A water cooled heat exchanger is placed

after the reactor to protect the pump and other systems.

16

The “tank” (where the reactor gas-liquid interface is located) is made from
12mm thick-walled Pyrex. lt facilitates dissolution of gases into solution, and acts
as a sight glass, allowing the ﬂuid level in the system to be monitored. Valves
and ﬁttings are added to allow monitoring and manipulation of the head space
gases. It was found that no additional mixing was needed in the tank. An
additional HPLC “sample injector" (Figure 3-1) valve allows the introduction of
substrates to the system.

The in-Iine sampling system consists of two Rheodyne 7010 HPLC valves
conﬁgured as shown in Figure 3-3 so as to allow the sampling loop to be ﬂushed
with water and the sample to be injected on a Bio-Rad HPX87-H column followed
by two detectors; a Perkin-Elmer LC 90 UV detector (210 nm) and a Waters 410
RI detector. It should be noted that the HPLC system can detect dissolved gases
and distinguish between He, H2, and Ar, with the response being directly
proportional to the headspace pressure. The ability to detect dissolved gas could

be used to measure hydrogen consumption on a ﬂow through system.

17

Sample loop

 
 
 
 

From RBR

 

 
 
   
   

HPLC . T0 RBR

 

 

 

 

Load Mode

Waste Sample loop

To column

V

 

HPLC

   
   

mo

 

 

Water

 

 

To column

 

 

 
 

 

 

 

 

Wash Mode

Figure 3-3: Conﬁguration of in-line sampling system

3.3 System evaluation

3.3.1 System mixing

Pulse experiments to evaluate system mixing were performed using lactic

acid with the UV detector placed in—line. A 100 pl pulse of 1.0 M lactic acid was

18

added to the recirculating ﬂuid that was ﬂowing at 6 mllmin. Figure 3-4 shows
that there is sufﬁcient mixing in the system, provided the time scale of sampling
once every 30 minutes. Even if the ﬂow rate is set at 3 mllmin, there should be

sufﬁcient mixing for uniform sampling.

 

0.05
0.045
0.04
0.035
0.03
0.025
0.02
0.015
0.01
0.005
0

UV Adcorbcnce

 

0 2 4 6 8 10
Tune (min)

 

 

 

Figure 3-4: Pulse attenuation at 6mllmin for recirculating system.

3.3.2 Volume determination of reactor, tank, loop, and system

A 10 ml syringe was cut lengthwise and clipped around the tank (112’
Pyrex tube) to help with reading the level in the tank. The reading on this cover
was calibrated by adding ﬁxed volumes of water using a 3 ml syringe and
recording readings. It was found that a reading of 3 corresponded to 2 ml while a
reading of 0 was equivalent to 1 ml in the tank

The volume of the reactor and loop were determined by ﬁrst adding 200
pmol GO to the entire system and letting this recirculate overnight to allow GO to
saturate the catalyst surface. At this point the system was sampled. Then the
recirculating pump was stopped and the tank drained and reﬁlled with 4 ml of

HPLC-grade water. The pump was turned on and recirculated for 2 hours while

19

samples were taken at 30 minute intervals. The reactor was placed out of line

and the dilution process was repeated. The system volume was then calculated

using V=4ml C2 ; this gave a loop volume of 3.82 ml and a reactor volume

I 2

 

of 5.61 ml.

3.3.3 HPLC calibration

HPLC analysis was based on external calibration. Linear response from a
Waters 410 differential Refractive Index (RI) detector was veriﬁed by running
standard solutions containing either alcohols or organic acids at concentrations
from 1 to 10 mg/ml. The standard solution of alcohols is referred to as HO—mix
and contains: sorbitol (SO), glycerol (GO), ethylene glycol (EG), 1,3-propanediol
(1 ,3-PG), methanol (MeOH), ethanol (EthOH), isopropanol (IPA), and n-propanol
(n-prOH), while the standard acid solution referred to as acid-mix contains
glycolic acid (GA), formic acid (FA), acetic acid (AA), and propanoic acid (PA).
The RI detector response curves for each mixture are shown in Figure 3-5; this
ﬁgure demonstrates the linear relationship between concentration and detector
response. Because of this linearity, the response factor (RF) was calculated at
each point for each compound. The RF had an uncertainty of :I:2% and did not
signiﬁcantly change with time, but did show a dependence on ﬂuctuations in ﬂow
rate. Both mixtures were injected using a Waters 717 Autoinjector (10 til/sample)
and also were injected through the reactor system injector to get the correlation
between the two. Figure 3-6 shows a chromatogram of HO-mix (1 mg/ml) run

through both the reactor system injector and the 717 Autoinjector. The peak

20

response ratio between the injectors is 4.7 and was the same for all compounds.
This is slightly higher than the expected ratio of 4.0, because the reactor system
has a 40 pl loop vs. 10 pl injections from the 717 Autoinjector. Examining the
dilution due to sampling of a standard solution in the recirculating reactor, it was
found that the actual injection loop was 43 pl. Therefore the actual amount that

the 717 Autoinjector injects is less than 10 pl.

 

 

~+« Glycolic acid
18 I —x—formic acid
16 . +acetic acid

 

Millions
Millions

 

 

—c— propanoic acid

 

 

 

 

 

 

 

 

 

 

 

 

14
12 4
2% £10 /’
8 8 , //i
s /“
4
2 /
0 Concentrationﬁmglmg 10 0 éonccntraﬁonﬁmglms 10

 

 

 

 

 

Figure 3-5: Refractive index detector linear response curves for 10 pl injection of
HO-mix and acid-rnix standard solutions.

21

 

reactor system

 

717 autoirjectorj

 

 

 

 

0 5 10 15 20 25 30
time (min)

 

 

 

Figure 3-6: Comparison of auto injector vs. reactor system injector with HO—mix
(Sorbitol, GO, EG, 1,3—PG, MeOH, EthOH, IPA, n-prOH)

Standards were made at 1% wlw in water and were run using the 717
Autoinjector (10 pl/rnjection). Relative response factors (RRF), which are
constant over all conditions, were calculated aid are reported relative to sorbitol
on a weight basis ( Table 3-1); the average organic RRF reported was applied to
any unknown peaks. The RRF were used to calculate the RF for each species in
the reactor system. With each adsorption reaction, the initial loop concentration
of the adsorbing species was used to verify the RF for sorbitol by the RRF for

that species.

22

Table 3-1: Retention times and relative response factors on a weight basis.

9291mm! _§_§RTat ° BEE
Oxalic acid 6.55 0.58
Sodium silicate 8.90 0.14
Gluconic acid 9.10 0.92
Glucose 9.22 1.08
Sorbitol 10.43 1.00
Glyceric acid 11.02 0.71
Boric acid 11.39 2.12
Glyceraldehyde 1 1 .55 0.78
Glycolic acid 12.42 0.64
Lactic acid 13.00 0.73
Glycerol 13.59 0.83
Formic acid 14.11 0.36
Acetic acid 15.41 0.46
Ethylene glycol 16.51 0.66
1 ,2-Propanediol 17.55 0.74
Propionic acid 17.94 0.58
1 ,3-Propanediol 18.07 0.71
Methanol 19.57 0.14
1 ,4-Butanediol 21 .97 0.83
Acetone 22.00 0.49
Ethanol 22.62 0.43
lsopropanol 24.50 0.59
n-Propanol 28.30 0.61
1-Butanol 35.50 0.56
average organic peak 0.60

3.4 Methods of operation

3.4.1 General description of experiment

For most experiments, the reactor was charged with 10 g ruthenium
sponge (-100 mesh, Aldrich); this Ru sponge was reused in most experiments.
The system was ﬂushed at 3 mllmin with argon purge and degassed HPLC-
grade water for at least 10 minutes. This was followed by charging the system
with ~12 ml of the same water. The system was set to recirculating mode and
placed under 250 psi H2 to clean and reduce the catalyst metal surface. The

reactor was heated to 200 °C for 2-4 hours. During this time the headspace was

23

exchanged with fresh hydrogen at least three times to remove any methane that
was formed during reduction. The reactor was then either cooled to 25 °C while
recirculating with H2 saturated water to generate the “H2 saturated surface,” or
was ﬂushed at 200 °C with the argon purge and degassed water for at least one
hour, followed by cooling to 25 °C to generate the “H2 free surface”.

With the reactor out of line, the loop was charged with water to 6.1 ml
(~1/3 up the sight glass tank). At this point a sample was taken, followed by the
introduction of the substrate of interest. After 30 min, the loop was sampled and
the reactor placed in-Iine, followed by automated sampling every 30 min until the
experiment was done. UV and RI detector data were collected for 30 minutes
and stored for analysis. The level in the sight glass and the headspace pressure

was recorded at several times during the experiment.

3.4.2 Data processing

The following variables are using in data processing:

Cn Concentration of solute in solution at sample n (pmollml)
Vm, Total system liquid volume (ml)

Va,“ Volume of reactor and heat exchangers (ml)
Vino, Volume of loop part of reactor system (ml)
Vim Volume of liquid in tank (ml)

No Moles added to system (pmol)

N... Moles in solution (pmol)

Nlcakloet Total moles lost due to leaks (pmol)

Nlnj loct Total moles lost due to injections (pmol)
N.“ Total moles adsorbed (pmol)

Pcount Corrected peak area counts

Pcount..." Uncorrected peak area counts
24

RT Retention time (min)

RTavg Average retention time (min)
RF Response factor (counts/(mglml))
RRF Relative response factor

A standard Excel workbook was created and several Visual Basic macros
were written to automate the repetitive tasks involved in processing the large
amount of data generated with each experiment. The amount adsorbed for a
given substrate was calculated using Equation 3-1 and involves the conversion of
raw UV and RI detector data into concentration data, ﬁnding the system volume,

and keeping track of cumulative sampling and leak losses.

 

 

Moles of Initial moles Moles still in Moles lost due
substrate = loaded into - solution - Ito sampling andl (3.1)
adsorbed the system leaks

 

 

 

 

 

 

 

 

 

 

Raw UV and RI detector data from chromatograms of each sample were
integrated using the Interactive Graphics part of Van'an Star Workstation v6.0
Software. For each sample, the retention time and area for each identiﬁed peak
was transferred to a Microsoft Excel worksheet and the areas were corrected for

ﬂuctuations in retention times using Equation 3-2.

RT,g — RT

 

Acorr = A 1+ (34)

3V8

This RT dependent correction of the area was found empirically and is valid for at
least 20% deviations in RT. This corrected area was used to calculate solution

concentrations using the response factors given in Section 3.3.3. If the measured

25

concentration deviated by more than 5% from the predicted value in the initial
sample with the reactor in loop mode, the RF values were sealed for that
experiment to match the measured to the predicted concentration based on the
actual amount of species introduced.

For every identified compound in each sample, the number of moles in
solution was given by C. Vim“ where Via.) = View + Vm + Vim, and C" =
(HPLC counts) RF... In loop mode, Vm was omitted. Only a limited number of
tank volume data points were recorded for each experiment; linear interpolation
was used to find Vim for each sample.

For each sample, the number of moles lost due to sampling for each
identiﬁed compound was determined as the solution conmntration multiplied by
the sample loop volume (43 pl). For leak losses, it was assumed that the
concentration changed linearly between sampling times; therefore, the average
concentration between the current and previous samples multiplied by the
change in volume over this same interval gives the number of moles lost due to
leakage. These two losses were summed and included in computing the

substrate quantities adsorbed from the current sample.

26

Table 3-2: Calculated values for a hypothetical adsorption experiment on Ru

 

 

 

 

 

 

 

 

 

 

 

 

 

 

sponge.

Time Vtotal Cn Neel Ninj lost Nlecli lost chs
(h.) (ml) (mM) ML (pmol) (pmol) ( mol
0.37 11.26 3.404 38.32 0.15 0.00 7.64
0.88 11.24 3.203 35.99 0.28 0.06 9.77
1.39 11.22 3.128 35.09 0.42 0.12 10.47
1.90 11.20 3.049 34.14 0.55 0.18 11.22
2.40 11.18 2.899 32.41 0.67 0.24 12.78
2.91 11.16 2.852 31.83 0.80 0.30 13.18
3.42 11.14 2.771 30.87 0.92 0.35 13.97
3.93 11.12 2.750 30.59 1.03 0.40 14.08
4.43 11.10 2.742 30.44 1.15 0.45 14.05
4.94 11.08 2.715 30.09 1.27 0.51 14.24
5.45 11.06 2.674 29.58 1.38 0.56 14.58
5.45 11.06 2.674 29.58 1.50 0.56 14.46

 

 

 

 

 

 

 

 

New“) = Vmi(i)*Cn(i); Ninjmﬂ) = Ninjlost(i'1) +.043*Cn(i);
Nleak m0): Nleek lost“) 1' (Vtotai(i-1) - Vtorai(i))*(Cn(i-1)+ Cn(i))/2i
Nedc“) = No - Neal“) ' Nlni lost“) ' Nleak lost“)

The total amount adsorbed was based on the number of moles of carbon
in each identiﬁed compound and reported as equivalent number of moles of the

substrate of interest. Table 3-2 shows calculated values for the adsorption of one

27

component on Ru sponge to illustrate the calculations described above. It should
be noted that for cases where products were formed, a negative adsorption value
is reported. This is of course a physical impossibility, but expressing it in this

manner simpliﬁes the calculations.

28

4 CATALYST CHARACTERIZATION

4.1 Methods

For this work, Ru sponge metal (product #267406, Lot# 14021 EB) from
Sigma-Aldrich was used. Catalysts were characterized on a Micromeritics ASAP
2010 adsorption instrument, either in a standard analysis cell or by attaching the
reactor vessel used in the recirculating batch reactor to the ASAP instrument.
Experimental details for nitrogen BET surface area measurements of activated
carbon are described in Section 7.1.1; the same procedure was used for Ru
sponge metal. Attempted N2 BET measurements of catalyst in the reactor failed,
most likely because of ﬂow restrictions through the frit.

General procedures for chemical adsorption experiments consist of
pretreatment, initial adsorption analysis, evacuation, and repeat analysis.
Pretreatment consisted of reducing the sample at 350 °C under a ﬂow of H2 for 2
hours, evacuation for 1.5 hours at 360 °C, cooling to 35 °C under vacuum, and
evacuation at 35 °C for 10 minutes followed by leak testing. All analyses were
run at 35 °C and consisted of dosing known volumes of analysis gas to the
sample chamber and recording the volume added vs. pressure. Following this
initial adsorption, the sample chamber was evacuated at 35 °C for 60 minutes to
remove physically adsorbed analysis gas, and then dosing was repeated as
before. The volume vs. pressure data for each dosing was then plotted and the

linear region (usually P>100 Torr) was extrapolated to P=0 to give the quantity of

29

gas adsorbed. The chemically adsorbed quantity of gas is the difference of
volume adsorbed in the initial and repeat analysis. The surface area is calculated
using the stoichiometry factor (2 for H2, 1 for CO) and the atomic cross-sectional
area of Ru (0.0613 nm2).

4.2 Results

Surface areas for both physical and chemical adsorption are listed in
Table +1. In the table, “fresh” Ru sponge was taken directly from bottle and
“used” Ru sponge is a collection of recovered catalyst used in a range of
hydrogenolysis experiments with different substrates. Both fresh and used
sponges were from the same lot number. Reactors 1 and 2 are mechanically
equivalent duplicates used in these studies. Results show a BET surface area
(~0.5 mzlg) that does not change upon using the Ru sponge, this area is

consistent with results of prior work by Zhang et al.2

Table 4-1: Physical characterization of Ru sponge.

 

 

 

Fresh Ru Used Ru Fresh Ru Used Ru
Reactor 2 Reactor 1“”
Total (N2 BET) 0561:0006 052710.004 -
surface area (mzlg)
Metal (H2) 0.23 0.12 0.21 0.15(0.13)
surface Area (mzlg)
Metal (CO) 0.21

 

 

 

 

surface area: (mzlg)

 

(" Reactor 1 was charged with 89 used and 29 new Ru sponge. Number in
parentheses is the amount corrected (by weight fraction) using the new Ru
values from Reactor 2.

Unlike the N2 BET, the surface areas found by H2 adsorption (0.23, 0.12

mzlg)» show a ~50% decline in the hydrogen active surface area after using the

30

 

Ru sponge for reaction. The H2 area measured for new Ru sponge in Reactor 2
(0.21 mzlg) is the same within experimental error to that found for new sponge
using a standard analysis cell, conﬁrming that using the reactor does not affect
the H2 adsorption measurements. Reactor 1 was charged with some (2 9) new
and some (8 9) old Ru sponge; the resulting adsorption area of 0.15 mzlg is
expected based on the weight fractions for this sample. The adsorption of CO on
new Ru sponge showed active Ru area of 0.21 m2/g, the same as found for H2.
CO adsorption was not done on reactors for fear of poisoning the metal surface
with CO.

A set of gas adsorption experiments were run with different gases to
compare different molecular adsorption properties (Table 4-2). Interactions of CO
and H2 on Ru surfaces have been studied,“ but our goal was to see if there
might be two different sites, one for CO and one for H. If there are two
independent sites, than there should be little dependence on the order of
adsorption.

The presence of CO on the surface does not seem to block the adsorption
of additional H2 (Exp A6), while the presence of H2 partially blocks the adsorption
of CO (Exp A3). Blocking of the CO by hydrogen was reported by Peebles”.

Subsequent addition of H2 (Exp A4) showed no signiﬁcant adsorption.

31

Table 4-2: Gas chemisorption on Ru sponge.

 

Volume
Adsorbed: Surface
Exp# Sample reduction Ema/98TH area:(mZLgL umolljg_
A1 N2 BET no 0.1146 0.4980
A2 H2 initial yes 0.1020 0.3350 4.55
A3 C0 after H2 no 0.1 100 0.1820 4.91
A4 H2 after H2 and CO no 0.0190 0.0640 0.85
A5 CO with reduction yes 0.1940 0.3200 8.66
A6 H2 after CO no 0.1040 0.3440 4.64
A7 N2 with reduction yes 0.0090 0.0149 0.40
A8 He with reduction yes 0.0005 0.0008 0.02
A9 Methanol attempt yes (~0.05) -—
A10 H2 after methanol no 0.0034 0.01 12 0.15
A1 1 Methane yes 0.0306 0.0504 1.36
A12 Methane after H2 yes 0.0016 0.0026 0.07
A13 H2 after Methane yes 0.0450 0.1483 2.01

Methane (Exp A11) adsorbed in a smaller amount than H2; this could be a
result of the larger size of methane, or it could be that methane adsorbs with full
dissociation. It should be noted that methane did not adsorb on an H2 saturated
surface (Exp A12). The adsorption of H2 (2.0 pmollg) after methane (Exp A13)
show that there are still sites available. If methane adsorbed with full dissociation,
little H2 adsorption would be expected, while if it were adsorbed as a methyl
radical (H30) or Methylene (H2C:), then there should be sites for H2.

To summarize the characterization of Ru sponge, fresh sponge has the

same BET area as used Ru but the fresh sponge H2 area is twice that of used

32

sponge. CO adsorption does not affect the adsorption of H2, but H2 does affect
CO adsorption.

Several analyses were run to evaluate the possibility of measuring H2
areas after aqueous phase reduction. It was found that a hydrogen-saturated
surface ﬂushed with He (to remove water) and then evacuated at 25 °C gave
inconsistent results. The adsorbed hydrogen came off the surface during the He
ﬂush and evacuation. Measuring gas phase adsorption after liquid phase

reduction is not possible.

33

5 Hle20 EXCHANGE

It has been observed that when H2 gas and liquid D20 are placed in a
reactor over ruthenium catalyst, D and H atoms are exchanged. This was a
problem when we were trying to determine the pathways for aqueous
hydrogenolysis of polyols using H-D exchange, as under normal experimental
conditions (100-150 °C, 500—1500 psi) the H-D ratios equilibrated on a time scale
faster (<30 min.) than our reactions, resulting in a statistical distribution based on
the overall H-D pool. At 25 °C and 10-200 psi, this exchange happens over
several hours. At this time scale, it is possible to perform experiments to

elucidate kinetic parameters for this reaction.

5.1 Experimental

Two different sets of experiments were run. In the ﬁrst set (Exp #43a-c), a
10 ml pressure vessel was charged with 2 ml of D2O, 0.1 g dry weight of catalyst,
and hydrogen gas (125 psi). The head space was continually sampled through a
capillary tube (0.05 mm ID. x 5 m) into a mass spectrometer. The second set of
experiments consisted of H-D exchange using a 70 ml Parr reactor equipped with
a magnetic stirring bar, gas connections, and sampling port. This reactor was
charged with 10ml D2O, 0.1 g dry catalyst or 1.0 g Ru sponge, H2 (50 psi), and D2
(50 psi). The head space was continually sampled through a capillary tube (0.05
mm ID. x 5 m) into a mass spectrometer. A pressure drop of ~25 psi was

observed during each experiment.

An Ametek Dycor M100M Quadrupole Mass Spectrometer was used for
the analysis of the headspace. A HD gas mixture was made by charging a 40 ml
pressure vessel with ~O.5 g 5 wt% Ru/C, H2 (150psi), and D2 (150psi). After
equilibrating this mixture for several days, the H2:HD:D2 ratios were assumed to
be 1:2:1. Instmment response was examined with H2, D2, HD, methane, and Ar.
The response ratios were inconsistent and seemed to have some dependence

on the degassing and tuning of the instrument.

 

         

Uncorrected Corrected
1 .7 1 , ,. a-
0.9 Ru/C 5% 0.9 Ru/C 5%
0.0 0.0
0.7 0.7
0.6 0.6
g 0.5 g 0.5
c: 0.4 l: 0.4
0.3 0.3
02 0.2

0.1 0.1
0
0 500 1 000 0 500 1 000

Tlmc (mln) 11m (mln)

 

 

 

   

0 500 1000 500
Tine (min) Tine (mln)

 

 

 

 

 

Figure 5-1: Corrected and uncorrected data for H2 over D20 with 0.1 9 5% RulC
(Exp 43B), and H2 over D20 with 0.1 9 5% PVC (Exp 43C)

35

After consultation with a technical expert from the mass spectrometer
manufacturer, It was concluded that this inconsistency is expected because this
instrument is designed and optimized for gases other then hydrogen. Thus no
further attempts were made to calibrate and optimize the system. The H-D test
mixture was analyzed either before or after each experiment to obtain the relative
response factors (RF). The assumption was made that the relative responses
were constant with respect to pressure even with a 50 psi pressure drop.
Therefore only the ratios of H2, HD, and D2 are plotted. Comparison of raw and
corrected data for Experiments 43B and 430 are shown in Figure 5-1. It can be
seen that applying the RF correction has little effect on Experiment 43C, while

there were signiﬁcant changes in Experiment 438.

 

 

 

 

 

 

 

Time (min)

 

Figure 5-2: Incorporation of deuterium into the gas phase from (DO)ethanol over
0.1 g 5% Ru/C (10 ml reactor 100 psi H2 Exp 47)

It can be seen in Figure 5-1 that, over a Pt catalyst surface, deuterium is
rapidly incorporated into the gas phase, while over a Ru surface the process is
slower, consistent with water strongly dissociating on the Pt surface while little

water dissociates on a Ru surface. Dissociation of a O-H bond in the liquid phase

36

is further conﬁrmed by deuterium incorporation into the gas phase when
(DO)ethanol is placed over 5% RuIC under H2 gas (Exp 47, Figure 5-2). This
H-O dissociation is slow, while H2 and D2 adsorption\desorption is rapid as seen
in Experiment 50 (Figure 5-3). Looking at the data in Figure 5-3, it appears that
deuterium incorporation in the gas phase is a lot slower than seen in Exp 438
(Figure 5-1), but this is deceptive because the headspace volume differs by a
factor ~10, while the catalyst loadings were the same. The absolute rates should

be the same; therefore the apparent rate should be slower with a larger

headspace.

 

 

1
0.9 - -----------------------------------------------------------------------------
0.8 « -----------------------------------------------------------------------------
0.7 « -----------------------------------------------------------------------------

g 0.6 i

E 0.5 -

  
 

 

 

O 200 400 600 800 1000
Time (nin)

 

 

 

Figure 5-3: H2 + D2 over D20 with 0.1 g 5% RuIC (50 psi each in 70 ml reactor,
Exp 50)

5.2 Modeling

The general approach to this system involved a Langmuir-type adsorption

with a pseudo steady state approximation (PSSA). In general, the Langmuir

37

model looks at adsorption at equilibrium; we wanted to look at the kinetics of

adsorption and desorption individually. The model equations are given as follows.

k1
XYIaq) + 2Mls) ‘_kz— XMIS) + YMIS) combinations ofXandY
k3 X=H Y=H
XYO(|) 'I' 2M(s) :ZiLT XM(s) + YOMM §;3 :;3
4 X=D Y=D
k5
H20“) + 020“) ? HDO(|) + DHO(|)

Figure 54: Proposed reaction model.
The kinetic model proposed is shown in Figure 5-4. Several simplifying

assumptions were made and are listed in Table 5-1. Assumption #1 is somewhat
unrealistic, because H or D is involved in all the reactions and it is generally
accepted that isotopes should have some effect. Assumption #9 is also
somewhat suspect. We have some evidence that there might be two different
sites in aqueous phase reactions, one for hydrogen and one for oxygenated
species such as water and organics. Some published calculations by Neurock'31

also suggest a two site model.

Table 5-1: Assumptions made for H2-D2O exchange model

There is no isotope effect.

Dissolved gas concentrations are in equilibrium with gas phase.
Only dissolved gases react.

There are no transport limitations.

Dissolved gases can only react with free sites.

@QF‘SPNT‘

Liquid species can react with free sites and with other liquid
species.
7. Liquid species are in equilibrium with other liquid species.

38

8. Migration on the metal surface is not considered
9. There is only one type of site on the metal for H-, D-, and 0-
adsorption.

5.2.1 Approach to solutions

All species involved in this system are shown in Table 5-2. Note that HD,
DH and HDO, DHO, which physically are identical pairs, are treated as separate
species; this was done to not have to take into account the probability of H or a D

reacting.

Table 5-2: Species included in H2, D20 exchange model

Dissolved Gases: H2 H0 D2
Species on Metal: HM DM HOM DOM M (free sites)
Liquid species: H2O HDO DHO D20

To determine the minimum number of reversible independent reactions
needed to describe the isotope exchange, a procedure described by J. M. Haile62
was used. In Haile’s method, the formula matrix is generated and the matrix rank
is then determined. Then the number of independent reversible reactions needed
(Rn) is the number of compounds minus the rank of the formula matrix. Figure
B-2 shows formula matrix and calculation to give Rn=9. The 9 reactions are

shown in Figure 5-5.

39

H2(aq) + 2M(s) :2 2HM(s)

 

 

 

 

 

Hem) + 2M(s)‘_:12_ HM(S) + ems)
DH(aq) + 2M(s)‘_:21__ DM(s) + HM(s)
02(aq) + 2M(s) :12 2DM(s)

 

 

 

H20(|) + 2M(s)
HDO(|) + 2M(s)

HM(S) + HOM(s)
HM(s) + DOM(s)

k4

 

 

k4

 

DHO(|) + 2M(s) DM(s) +HOM(s)

DM(5) + DOM(5)

 

k4

 

020“) + 2W8)

 

k4

H200) + DzO(|):_—_::: HDO(|) + DHO(|)

Figure 5-5: Reversible reaction model

From these 9 reversible reactions, we get 18 elementary reactions, that
are all either 2"6 or 3’d order. These reactions will have a rate (velocity) term of
the form r=k[A][B] for 2"d order and r=k[A][B]2 for 3rd order. It should be noted that
only availability of free metal sites (M) and not their spatial orientation has been
included in the model. Because of this, dissociative adsorptions are third order
reactions. These terms, when calculated for all 18 reactions, result in 17 unique
rate terms shown in Table 84 To get the corresponding 13 ODE’s, one creates

a matrix with the coexistence for each term as need to account for the species

balance. This matrix is shown in Figure B-3. The vector of ODE’s % = 1; a;

can be used as a function block for matlab’s ODE solver. The reason for using
this matrix system is that the rate terms are calculated only once in the cycle. If
the terms are added as terms in each individual ODE, they would be recalculated

each time they are used. This method signiﬁcantly speeds up the calculations.

40

The ODE solver used from MATLAB was ODE15S. Initially ODE45, a 4th
order Runge-Kutta method, was used, but this problem is too stiff for that solver
and tended to diverge.

The initial conditions of the system were calculated from experimental
data and were kept constant throughout this modeling exercise. All variables are
listed at the top of the Matlab code. Parameters that were adjusted to observe
the system behavior were: k1, Kr, Kg, Kl. The reference rate constant was R1, the
forward rate of the “gas” side reactions. Kr is equal to k1lk3, the ratio between
the “gas“ side and “liquid “side forward reactions rates, while Kg = k1lk2 and Kl
=k3/k4, the equilibrium constants for the “gas“ and “liquid” side reactions,
respectively. These values were adjusted to get the model response to be close
to experimental. The k1 value of ~1000 (Lzlmolzls) is within a few orders of
magnitude of expected reactions with a surface. Generally, adsorption and
desorption reactions are rapid and are not the rate limiting steps in catalytic
systems. The value of k5 was set at 1000 (L/molls); this was large enough that
the proton exchange in the water was close to instantaneous. The value of k5
was varied by a factor of 100 in both directions with no noticeable effect on the

model predictions.

5.2.2 Parameter optimization

For the experimental data, the raw data for Experiment 43b (5% RulC)
and Exp 43c (5% Pth) were used. These data were not corrected for the
differences in response factors. The only measurable quantities were the

compositions in the gas phase. Dissolved gasses are assumed to be in

41

equilibrium with the head space gas and related by Henry’s law. For simplicity,
gas phase concentrations were assumed to be the same as dissolved gas
concentrations. The actual dissolved gas concentrations in the liquid are lower by
a factor ~1000, which if used would raise the corresponding k values.

Initial fitting was done by manually changing k1, Kr, K9, and KI to
approximate the results shown experimentally. The ﬁnal optimization was done
using a random walk approach. For each cycle, one parameter was chose at
random and increased or decreased by a random percentage, limited such that
no change was greater than 10%. The ODE’s were then integrated over the
same total time as the experiment, followed by calculation of the overall error

using Equation 5-1.
Error=2((C:2‘p —c;:'.>2 +(Cff; 42st ace; -C‘3I.>2 +(C3‘é’ -C3'2 2 (54)
0

If this error was lower than the previous cycle, the change was kept; it not, the
old values were kept and the cycle was repeated. This is a “brute force” method
but it is much simpler to implement than some of the gradient methods, and also
it tends to avoid local minima if the changes are large. The possibility of an
evolutionary model, in which all parameters are changed at once followed by
some selection rules as to the diredion of the next change, was also examined.
This method sounds like it might be a better way of doing the optimization if there
are a lot of variables.

The error calculation is simply the summing of all the squares of the
individual errors. The one challenge is that ODE15S uses variable time steps so

it is necessary to either interpolate the model point to match the experimental
42

time or the reverse. It was easier to interpolate the experimental data than the
model. The experimental data has >5000 points, while the model generally

generates 100-200 points depending on the steepness of the curves.

5.2.3 Results and conclusions

Figures 5-7 and 5-8 show the experimental results and the calculated
results. The model predicts the shape of the curves nicely. The optimized
parameters are shown in Table 53. With ruthenium, the gas side reactions are
much faster than liquid side, as evident by the large value of Kr. When a
molecule of deuterium (D2) makes it to the surface, is rapidly scrambled with the
gas phase. In the case of platinum, the liquid and gas exchange rates are almost
the same but the liquid side equilibrium lies a lot more towards the surface bound
side. This is consistent with water readily dissociating upon adsorbing on Pt as

described in the literature”65

Table 5-3: Optimized parameters for experiment 43B and 43C

Exp 43B k1=50586; Kr=129.54; Kg=6.9848; KI=8.6846; k5=10000;
Exp 43C k1=31093; Kr=2.0048; Kg=5.3232; KI=94.379; k5=10000

43

 

5 0.8

 

0 200 400 600 800 1000
Time (min.)

 

 

 

Figure 5-6: Model ﬁt to experimental data for 5% RuIC (Exp 43B).
Dotted lines are predicted values.

 

 

0 200 400 600 800 1000
Time (min.)

 

 

 

Figure 5-7: Model ﬁt to experimental data for 5% Pth (Exp 43C).
Dotted lines are predicted values.

6 AQUEOUS PHASE ADSORPTION ON RUTHENIUM
SPONGE

All experiments were performed in the recirculating batch reactor (RBR)
using previously described experimental procedures (Section 3.4.1). A summary
of all experimental conditions can be found in the master experiment list
(Appendix A). Data processing and calculations are explained in Section 3.4.2.
Unless noted differently, in the ﬁgures used to display data generated with the
RBR, the left graph displays solution concentration while the right displays
quantity adsorbed.

6.1 Sorbitol, glucose and gluconic acid adsorption

Previous work in our lab has looked at transformations of sorbitol (SO) to
either propylene glycol (PG)5 or isosorbide“. It has been proposed that the ﬁrst
step in conversion of SO to PG involves the dehydrogenation of $0 at the C1
position to form an “aldehyde” intermediate. To get some mechanistic insight into
this ﬁrst step, information on the adsorption of $0 on Ru metal would be helpful.
Initial adsorption work on ruthenium sponge was done by Schafer. He looked at
$0 adsorption on hydrogen-saturated metal in a 300 ml Parr reactor with 20 g
Ru sponge. No signiﬁcant adsorption was seen until the temperature was raised
above 80 °C. Above 100 °C, ~5-10 mg SOIg Ru was unaccounted for and was
assumed to be adsorbed. This is signiﬁcantly higher than the 0.2-0.5 mg SO/g
Ru predicted for the measured metal surface area of ~0.2 mzlg-sponge and an

area of ~30 A2 per sorbitol molecule. The appearance of activation with
45

temperature is possibly due to the desorption of hydrogen from the metal surface
at ~80 °C, as mentioned in the characterization of the catalyst. The apparent
over-adsorption of $0 is probably due to the degradation of sorbitol into a large
range of products that were not analyzable by HPLC or were below the detection
limits in Shafer‘s system.

The RBR was used to look at sorbitol, glucose, and gluconic acid
adsorption at 25 °C on Ru sponge metal. Preliminary experiments with sorbitol
(SO) showed that on a hydrogen-saturated Ru surface there was no signiﬁcant
adsorption, while on a hydrogen-free surface sorbitol adsorbed and degraded

into a large range of products.

 

 

 

 

 

 

 

 

 

 

 

 

 

+ Total Material (10mg) —e— Total Material (5mg)
+Somitol (10mg) —Q—Sorbitol (5m)
+Side Products (10mg) +SideProducts (5mg)
6
............................................................. ﬁ
5 ‘ n.9._~ __o~0.5 5...? _. H
E _
54
i
a
T!
'2 L
1 __ -__
...L7 -Aiiﬁiiiiiiiéﬁivéﬁi‘igAiiiviiiiiiidA
o_;_.‘..A.-a““ I I I
0 5 10 15 20
Time(l'louls)

 

 

 

Figure 6-1: Adsorption of sorbitol on 10 g H2-free Ru sponge at 25 °C. Initial
loading of 5 mg (27.5 pmol) and 10 mg (55 pmol).

Figure 6-1 shows that at a high SO concentration (2-5 mM) the metal is
saturated in a relatively short time (1 -2 hours), while at lower concentrations
adsorption is slower as seen in stepwise additions (Figure 6-2). The presence of
a large range of products makes it hard to give an accurate accounting of the
amount adsorbed. Using an average response factor and average molecular
weight (100 glmol), each of these experiments show an adsorption of ~0.1 mglg
Ru (0.55 pmollg). It should be noted that no glucose is seen in adsorption
experiments with sorbitol. Control experiments were run to see if this apparent
amount of adsorption was really due to the Ru metal and not a result of
calculation error or system behavior. First, an empty column (Exp 11) was used
but this had a signiﬁcantly larger volume than a Ru sponge ﬁlled column. To keep
the volume constant with minimal increase in surface area, 150 pm glass beads
(Exp 21) were used. To give a larger surface, a third experiment (Exp22) used
Controlled-Pore Glass ® (product# CPG03000D, lot#11D011) with a total surface

area of ~15 m2. All three of these experiments showed no adsorption of SO.

47

 

 

o
on

p

on

—e— [others]" —a— [$0] —I— adsorbed

o
N

p

N

 

 

O

G

T!
O
G

O

U!

I'
O
U'I

 

 

 

 

 

Concentration (mM)
O
.5
I
l
O
.5

 

. .0 . . .
00
Adsorptlon (pmollg Ru)

0 O
N (.0
O
N

‘IA’c;

 

 

4n
A _§'r 4
o v v

 

 

O
—L
f I
n I
0 I
i
0
q!
n
0
r
b
a
(.
J
V
‘\
ID
(-
-)
(.
.\
1
ﬂ
0
A

1
O

 

 

 

Figure 6-2: Adsorption of sorbitol at 25 °C on 10 g Ru sponge using stepwise
dosing (Exp 20). Concentration of other species is in sorbitol
equivalents.

When sorbitol was adsorbed in the presence of 5 mM H2804 (Exp 14),
only one main product was formed at 10.8 minutes retention time on the HPLC.
In the presence of acid, it is well known that sorbitol can cyclize by dehydration to
form 1,4-anhydro-D-glucitol (sorbitan) and 143,6 dianhydro-D-glucitol
(isosorbide).67 These species were identiﬁed by comparing their retention times
with those determined by Dozeman‘37 on the same Bio-Rad column. The
formation of sorbitan at 25 °C with dilute acid only in the presences of Ru sponge

metal gives some evidence of the activation of the C1 position.

48

 

 

0.50 _ g 1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

+iothetsr /M
0.45- , [mucosa] 0.9
0,40 . +[Gluconic acid] 0.8
c ‘ A
A 0.35 — +IS°"’“°“ 0.7 E
E +Adsorbed ,R . g
~0w~ oea
5 \ a
g 0'25 \ I 0.5 E
a
8 0.15 - .‘ _. 0.3 v
N ;: it I “
0.10 9 ' I“ . .55 4/“”"' '- 0.2
’ L t Y r-K'r'iLiv'LK
0.05 - l -_ ’3‘“ 0.1
0.“) ‘V‘ _ \‘ 7‘", - '- jam-:1: I. m A .. T "' I T T T T 0
0 5 10 15 20 25 30 35 40 4s

 

 

Time (hours)

 

Figure 6-3: Adsorption of glucose at 25 °C on 10 g Ru sponge using stepwise
dosing (Exp 23). Concentration of other species is in glucose
equivalents.

Adsorption of glucose at 25 °C on Ru sponge using stepwise dosing (Exp
23) is shown in Figure 6-3. The total adsorbed material is 0.88 pmollg Ru, which
is similar to glycerol under the same conditions but more than sorbitol (0.55
pmollg Ru). The only products detected to any extent are sorbitol and gluconic
acid. Both these products are formed to the same extent, suggesting a
Cannizzaro-type disproportionation. It should be noted that in the presence of air,
gluconic acid is the main product. After stopping the stepwise addition of
glucose, the system was left to equilibrate overnight, sampling was started, and
then 500 psi H2 was added to the headspace. Figure 6-4 shows that the
concentration of gluconic acid jumped up followed by a slow conversion to

sorbitol. This is a real change, because gluconic acid is measured by UV

49

detection and not refractive index (RI) detection. Because of glucose and
gluconic acid peak overlap in RI, the glucose concentration is subject to error.

This shows that some form of adsorbed acid can be removed by adding H2.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.35
——o— [othersl'
0.30
<———Added 500 psi H2 +[G'"°°s°l
g 0.25 n —9—[GIUCOII'C acid]
E, +[Sorbitol]
g 0.20 -
E
c 0.15 —
8 .
C
8 0.10 /
O 05 ‘ W
jw
Dom ﬁ f T I I I
62 64 66 68 70 72 74 76 78
Time (hours)

 

 

 

Figure 64: Effects of H2 addition on adsorbed glucose at 25 °C on 10 g Ru
sponge after stepwise dosing (Exp 23)

Adsorption of gluconic acid on a hydrogen-free surface shows no
formation of side products. Figure 6-5 shows the adsorption of gluconic acid at
25 °C on Ru sponge (Exp 25); the leveling off and slight decrease seen between
10-23 hours is mainly due to a small leak in the system that was not accounted
for in the calculations; the actual amount adsorbed should be ~0.8 mg (0.42
pmollg Ru). The addition of H2 shows a slight increase the solution concentration
of gluconic acid. This increase and the increase seen in Exp 23 (Figure 6-4) are
taken as evidence that some gluconic acid adsorption is reversible. It should be

noted that the observed increase is with in the uncertainty of the data.

50

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.7 — 0.7
[Sorbitol] Added 500 psi H2 ——p~
0.6 + ﬁlm 0.6
—c—[Gluconic acid] I

A 0.5 ~— 0.5 ‘5
.. s
“g- 0.4 0.4 g
’5' i... ‘5

0.3 — A 4:.» l ‘. 0 3 °
g r T “E
8 ' 3

0.2 ' 0.2 <

0.1 0.1

0,0 - “AC-Z1eccecccecceuccccccceccecccceecccccccc‘ excescttsw 0

0 5 10 15 20 25 30
Time (hours)

 

 

Figure 6-5: Adsorption of gluconic acid at 25 °C on 10 g Ru sponge using
stepwise dosing (Exp 25)

The stepwise addition of sorbitol and glucose both show a lag time before
products appear. This lag seems to be related to the oxidation state of the
substrate: it is 5 hours for sorbitol and 10 hours for glucose. Comparing the total
material adsorbed shows that glucose (0.88 pmollg Ru) is signiﬁcantly higher
than sorbitol (0.55 pmollg Ru) and gluconic acid (0.42 pmollg Ru). Unfortunately,
because of product formation and the large molecule size, this system is too
complex to draw meaningful mechanistic conclusions. Therefore, we changed to

a smaller substrate, glycerol, as discussed below.

51

6.2 Glycerol Adsorption

The adsorption of glycerol was studied by itself and in combination with
propylene glycol. The effects of concentration, hydrogen saturation, and
temperature were looked at with the hope of gaining mechanistic insight into the
adsorption process. Experiments were done to look for evidence that the

adsorption is reversible.

6.2.1 Concentration dependence

Initial glycerol (GO) adsorption studies were performed at 25 °C using
stepwise addition (Exp 24) and a ﬁxed dose of 4.3 pmol (Exp 26) of GO. Figure
6-6 cleariy shows that in both experiments adsorption plateaus at ~0.3 pmollg Ru

even though concentrations are 2.6 and 0.05 mM respectively.

 

   

3.5 0.40
3.0 A 0.35
A “=5 0 30
E 2.5 g -
5 2.0 E 0.25
E 5 0.20
g 1.5 5
o E 0.15
= 1.0
8 '8 0.10
0.5 < 0.05
0 exp 24 A exp 26
0.0 0.00
0 10 20 30 40 0 10 20 30 40
Time (hours) Time (hours)

 

 

 

 

 

Figure 6-6: Adsorption of glycerol at 25 °C on 10 g Ru sponge comparing
stepwise (Exp 24) and ﬁxed dosing (Exp 26).
Left solution concentration; right quantity adsorbed.

52

This result and the results from sorbitol adsorption mentioned before led us to

assume that the total quantity adsorbed was independent of concentration. Most

subsequent experiments were run based on this assumption, but it was only after

examining all experiments at 25 °C that this assumption was validated.

 

Adsorption (pmollg Ru)

A

 

.-*.-‘
more

99.69
N‘O

0.0

 

 

 

 

 

 

 

 

 

e
e
e
e
0
“it
0.0 2.0 4.0

Initial Concentration (mM)

6.0

 

 

 

dd

Adsorption (pmollg Ru)
o o p o c

'o

'cnbe'oio

'ro

 

 

 

Q

 

 

 

.5
r

 

 

C

 

 

0.0

1.0 2.0 3.0
Final Concentration (mM)

4.0

 

 

Figure 6-7: Adsorption of GO at 25 °C using (A) ﬁxed dosing: maximum initial
concentration vs. total adsorbed, and (B) ﬁnal solution concentration vs.
amount adsorbed.

Figure 6-7 shows maximum initial concentration vs. total adsorbed (A)

and also the ﬁnal solution concentration vs. amount adsorbed (B) after ~18

hours. Maximum initial concentration is the concentration in the loop before the

reactor is put inline. It should also be noted that this data has both 10 and 20

gram loadings of ruthenium sponge while the total volume of the system is about

the same (11-13 ml). As long as there is enough GO in the system, the

adsorption peaks at ~0.8 :I: 0.2 pmol/g Ru. The uncertainty is based on four

nearly identical experiments at 7.5 mM initial concentration. The quantity

adsorbed, 0.8 pmollg Ru, is only about 10% of the quantity of CO adsorbed via

chemisorption (~9 pmol COlg Ru). Yet the value is close to that predicted (1 -1 .2

53

pmollg Ru) by assuming end-on GO adsorption with a 30 A2 per molecule
surface area. Most experiments were run for at least 18 hours and adsorption
curves seem to level off after ~10—20 hours. This behavior was also observed in
stepwise addition of GO (Exp 24). It does not appear that exposure of H2-free Ru
surface to water for an extended time changes the adsorption behavior, but even
small amounts of polyol decomposition could deactivate the surface.

Some noted exceptions to the above behavior are Experiment 67 and 68
(Figure 6-8 and Figure 6-9) which showed that the initial 5 pmol GO added
adsorbed completely in the ﬁrst 15 hours; the subsequent addition of 5 more
pmol GO still adsorbed readily. Reduction conditions were not the same as for
most experiments. In Experiment 67, the metal was washed with degassed water

for 2 days (over the weekend); this was ~2-3 L of water that might have oxidized

 

some metal.
1.2 0.12
-3 1.0 0.1
a E . z
2
=3? 0.8 008-; g
E - 0
a”: 06 0065 n
it ,2
6
0.4 0.04 F <9
8 E 8 "i
< c
3 0.2 0.02

 

O

-1 9 19 29 39 49 59 69
Time (hours)

 

 

 

Figure 6-8: Adsorption of GO at 25 °C (Exp 67) on Ru sponge prepared with
excess water. For clarity EG and PG concentrations are plotted at 10X.

54

 

 

 

 

 

 

 

 

 

 

 

 

 

   

 

 

 

 

 

 

 

1.2 0.12
-e—GO -e—adsorbed A EG a PG |
‘2“ 0.1
E 5 ,, ..
980 I owe:
3 h g V
E .2 0
3 c 0. 0% E 9'
o “a
E 5
1 c 0
g g 0. 0.04 8 m
< i:
8 0. ‘A A A . ‘Ir A A" I 0.02
A AA ﬂ m®-0%D ;I..
A B ACPndPu
o I T I 0
-1 4 9 14 19 24 29 34 39
Time (hours)

 

Figure 6-9: Adsorption of GO at 25 °C (Exp 68) on Ru sponge prepared by
reduction on the Micromeritics ASAP 2010 instrument. For clarity EG
and PG concentrations are plotted at 10X.

For Experiment 68, the Ru sponge was dry reduced on the ASAP instrument and
ﬂushed only with 10-20 ml degassed water. It is interesting to note that EG and
PG did not form initially in Exp 68, while in Exp 67 and most other experiments
these products were formed almost from the start.

6.2.2 Hydrogen saturation

Generation of the “H2 saturated” and “H2 free“ surfaces is described in
Section 3.4.1. Discussion of the nature of the Ru surface is presented in Chapter
4 and 5. Flgure 640 shows adsorption of GO on both H2-saturated (Exp 28) and
H2-free (Exp 39) surfaces at 25 °C. The H2-saturated metal shows limited
adsorption of ~0.05-O.1 pmol GOlg Ru, while ~0.3 pmol GOIg Ru is adsorbed on

hydrogen free metal surface with the formation of trace amounts of EG and PG.

55

 

 

 

 

 

 

 

 

 

 

 

0‘40 oI-I25at. eH2free I
S 0.35 c
g 0.30 ‘W
g 0.25 . °“°°°e
E 0.20 4V.
§ 0.15 ’27
C 0.10 L
8 005 M-

0.00 l l

0 5 10 15
Time (hours)

 

 

 

Adsorption (mol/g Ru)

9.0
88

0.25

'8

0.15

9.0.0
888

 

 

 

 

 

 

 

 

4
g o
0.0 9°
n

 

 

o 9
.0 u.°o° ° °oo° -
WV I1 °

 

 

0

Time (hours)

 

 

Figure 6-10: Comparison of adsorption of G0 at 25 °C on H2-saturated (Exp 28)
and H2-free (Exp 39) Ru sponge (10 9).

After some minor repairs and the addition of 109 more Ru sponge into the

reactor, the effect of H2 saturation was rechecked (Figure 6-11). Total GO added

was 50 pmol (7.5 mM max) for each reaction. Unlike Exp 28 and 29, where the

headspace gas was H2, both Experiments 93 and 100 used Ar in the headspace.

 

5.0
4.5
A 4.0
E 3.5
‘a' 3.0

_o

E 2.5

= 20

E 1.5

8 1.0
0.5
0.0

ol-I25at.

eH2ere

0 5 10
Time (hours)

 

 

15

 

 

Adsorption (pmollg Ru)

.099
(nude

99.09
NOD-#01

.09
0—3

cHZIrBe

 

5 10
Time (hours)

15

 

 

Figure 641: Effects of higher GO loading on adsorption at 25 °C on H2-
saturated (Exp 100) and H2-free (Exp 93) Ru sponge (20 g).

GO adsorption (0.3 umol/g Ru) was higher than expected in Exp 100; this
might be due to the higher initial concentration of GO or it might just be that the
headspace gas composition is critical. Unfortunately, no direct comparison of the
effects of the headspace gas during the adsorption run was done, but Figure
6-12 shows adsorption of GO (4.5 pmol loading) at 50 °C under Ar on a H2-
saturated surface (Exp 30). The observed adsorption of GO (~0.3 umol/g Ru) is
comparable to GO adsorption seen on a Hz-free surface at the same loading
(Exp 39), except that the adsorption rate is a lot slower. The higher adsorption
under Ar could be explained by desorption of H2 from the Ru surface, freeing up
sites for GO to adsorb. At 25 °C, the rate of H2 desorption is practically zero while

at 50 °C it is signiﬁcant.

 

300"”)

Adsorbed (umollg Ru)
Concentrationfor
99999999999
O‘Nu-hUIGNOCOO

 

-2024681012141618202224
Time(hours)

 

 

 

Figure 6-12: Adsorption of G0 on Hz-saturated Ru sponge (10 9) under argon at
50 °C (Exp 30).

The effects of adding hydrogen to the headspace after GO adsorption on
a Hz-free metal surface will be discussed in Section 6.2.3. Heating of the H2-
saturated system to 100 °C (end of Exp 28 not shown) resulted in rapid

decomposition of glycerol with the formation of methane, PG, EG, and trace

57

amounts of CO. It should be mentioned that at 25 °C the addition of H2 (250 psi)

appears to stop any further decomposition reactions.

6.2.3 Temperature effects on GO adsorption

Experiments were run at 25, 40, 60, and 80 °C under identical conditions
(50 pmol GO, 20 g Ru sponge, and standard metal preparation). Gas analysis
was not performed on all experiments; those that had no actual analysis done but

were corrected with assumed methane production are labeled 'exp#*".

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

"° W
0e
.0 p o
.l ’9 e

0.9 “‘6“ ,,

08 _ t. O ‘ ‘

07 - .. 51.31123?"
% 0.6 " AP""I‘~ £0 1‘ A mc( ) N
E 0'5“ ”I" M9 8°ch
35 :H' 939 6
c #10,, 'o' memes)
O 0.4 4 .,, 0.9 ”' .
3 0.3 ’ ’- 800mm) .,-' ammonia)
'0 .
“ 02 g ' ..

- 3 0.8 - . ~ . g ‘ . .5
0.1 1‘] . “ - ‘ can AAA memo“)
' r 250(9‘l93')
0.0 U 0'72: . 5' ' 1b 115 25 25
' k
'0.1 I I l f I l u
0 5 1o 15 20 25 30 35

Thne(houn9

Figure 643: Temperature dependence of GO adsorption. (50 pmol GO loading,
20 g Ru sponge). Inset shows magniﬁcation of upper region of graph.

Figure 6-13 shows that adsorption does not significantly change at different
temperatures and is ~0.8 - 1.0 pmollg Ru. Multiple experiments at the same

temperature were added to the graph to show that experimental variations have

58

a greater effect than temperature. Looking at the products formed as a function
of temperature (Figure 6-14), it can clearly be seen that GO decomposition goes

through a maximum at 60 °C.

Scheme 6—1

OH

HO OH
+ Ru
H

H
i

 

OH

HOVKfO 4' H2(adﬁ) H2

H ds

 

Fragmentation,
retroaldol,
C-Cclavage

i

 

c1-c3 #2:. Methane

Scheme 6-1 can be used to explain the observed behavior. As the temperature
increases, so does the fragmentation of GO, but hydrogen is needed to react
with these fragments to remove them from the surface. Surface bound H2 is
produced by the adsorption of GO, but can come off into solution at elevated
temperatures. This H2 desorption starts to be signiﬁcant at 60-80 °C. This release

of H2 has the effect of oxidizing the metal surface.

59

 

20 30 4O 50 60 70 80 90
temperature (degc)

 

 

 

 

Figure 6-14: Glycerol decomposition products vs temperature. Multiple
experiments are included at 25 and 60 °C.

As shown previously, hydrogen saturation inhibits the adsorption of
glycerol at 25 °C. Adding H2 to the headspace at 25 ’0 only stopped the
adsorption of GO, but at 80 °C (Exp 96) the result is the conversion of adsorbed
GO to methane without adsorption of additional GO (Figure 6-15). It should be
noted that adding H2 and increasing the temperature to 200 °C results in
quantitatively converting all carbon species to methane.

Some additional temperature effect experiments were run (Exp 30, 31,
32). Experiment 32 is run stepwise at 80 °C and includes the addition of E6 and
PG. Data from these experiments were unreliable due to fact that changes in

temperature were made may days after the start of the experiment.

60

 

9.0
88

i
E . E.
5 Added 250 pSI H2 g
‘E 0.20 .°
.o E
E 0.15 g
E
g o1o§
8 2
0.05 :5

 

 

 

Added 250 psi

Adsorption (pmollg Ru)

 

0 5 10 15 20 25
Time (hours)

 

 

 

Figure 6-15: Adsorption of GO (50 pmol loading) at 80 °C on 20 g Ru sponge
(Exp 96) showing the effect of adding H2 and increasing temperature to
200 °C. Quantity GO adsorbed shown on lower graph, solution
concentrations on upper graph. Concentrations of EG and PG are
plotted on right axis.

6.2.4 Propylene glycol (PG), ethylene glycol (EG), and 1,3-
propanediol (1 ,3-PDO) adsorption
Propylene glycol adsorption experiments were run at 25 °C with 6.5 and
66 pmol PG loading. Figure 6-16 shows PG adsorption of ~0.18 umol/g Ru at
low loading (Exp 42) and ~0.75 pmol/g Ru at high loading (Exp 41). The amount

61

of PG adsorbed at the high loading is similar to that found for glycerol. This would
indicate that the surface is saturated. Comparing adsorption at lower

concentrations suggests that PG binds less strongly than GO.

 

 

 

00>:
OO
)
99
NO

 

 

 

 

 

 

 

 

“99‘.
00
‘r
’1‘
mollgRu
0.0
(no
i
d

 

 

 

 

 

 

Concentration (mM)
2 o

 

 

 

 

 

 

 

 

 

+66 umol exp 41 :
+6.5 umol exp 42 g 0.3 f
, +66umoll‘l2sat.exp51 3 02 J}
1.0 2 0.1 W
0.0 I A AA: A;#‘ AA‘ ATAAA 0.0 T' T 1 T
0 5 10 15 20 O 5 10 15 20

Time (hours) Time (hours)

 

 

 

 

 

Figure 6-16: Adsorption of PG at 25 °C on 10 g Ru sponge comparing loading
and H2 saturation.

As with GO, the pro-saturation of Ru with H2 inhibited the adsorption of PG
(Figure 646, Exp 51). This ﬁnding is consistent with the concept that an initial
dehydrogenation step as proposed for glycerol also applies to propylene glycol.
The cyclic nature of the noise in Experiment 51 probably stems from the HPLC
system, but no satisfactory cause was ever found. Lactic acid and an unidentiﬁed
peak were the main products and were only seen in the high concentration run
with no hydrogen (Exp 41). The area response (RI) for these two peaks is almost

the same, suggesting that they might be related.

62

 

16.0
14.0
12.0
10.0

6.0
4.0
2.0
0.0

Concentration (mM)
on
O

 

+exp 52 —.— exp 53

5 10 15 20

Time (hours)

 

 

 

 

Time (hours)

 

 

Figure 6-17: Adsorption of E6 (80.6 pmol) on 10 g Ru sponge at 25 °C.

Ethylene glycol was run only at 81 pmol EG loading (Exp 52, 53) and showed an

extent of adsorption of ~0.8 - 0.9 pmollg Ru (Figure 6-17), again similar to GO.

H2 saturation effects were not checked.

 

4.0
3.5
E 3.0

E
= 2.5
O

'5 2.0

§ 1.5

E

3 1.0
0.5
0.0

 

A1,3 FED O 3—l-PA?

XAA CIFA

 

5

10 15
Time (hours)

20

 

 

Adsorption (umol/g Ru)

0.8

0.6

0.4

0.2

  

0.0

 

—0.2
0 5 10 15
Time (hours)

 

Figure 6-18: Adsorption of 1,3—PDO (50 pmol) on 20 g Ru sponge at 25 °C.

Adsorption of 1,3—Propanediol (1 ,3-PDO) on 20 g Ru sponge at 25 °C

(Exp 113) gave an extent of adsorption of ~0.8 umol/g Ru (Figure 6-18). This is

similar to other polyols. It should be noted that 1,3-PDO showed a range of

63

decomposition products including ethanol, formic acid (FA), acetic acid (AA), and

3-hydroxypropanoic acid (3-HPA).

6.2.5 Binary adsorption (GO, PG)

To examine the difference in binding energies of glycerol (GO) and
propylene glycol (PG) on Ru, experiments were run with both species
simultaneously. In one run, PG was added ﬁrst followed by GO, in another run,
GO followed by PG, and in third run both were added at the same time. These
experiments were all run at 25 °C with 10 g Ru sponge.

Initial adsorption of PG after 16-20 hours was ~0.18 pmol PGIg Ru for a
loading of 6.6 pmol (Exp 42). At a higher loading of 66 pmol (Exp 44), this
increased to ~06 umol PG/g Ru. In each experiment, 4.34 pmol of GO was then
added and allowed to adsorb for 20 hours (Figure 6-19). In both cases, GO
adsorbed after PG in the amount of ~0.2 (Exp 42) and ~0.1 (Exp 41) pmollg Ru,
respectively. The total amount adsorbed is similar to the amount GO adsorbed

at similar loadings (Exp 39 and 93).

 

A

5

Adsorption (umol!

 

A

0.8

0.7 .

0.6
0.5
0.4
0.3
0.2

0.1 ~

 

—e— G0
+ total

 

 

 

 

 

 

 

 

 

0.0 #—

 

 

10

20 30 40

Time (hours)

 

 

.0
co

 

 

+GO

‘ —o—total

 

 

 

9 i
.o .o
G N

 

99
our

 

 

 

 

.0
u

 

 

 

Adsorption (pmol! Ru

9
N
F—h

 

pp
O-l

 

 

Time (hours)

 

Figure 6-19: Adsorption of PG followed by GO on 10 g Ru sponge at 25°C.
A: Exp 42, initial PG loading of 6.6 pmol; and B: Exp 44, initial PG
loading of 66 pmol. GO loading of 4.34 umol follows PG loading.

umol) followed by PG (6.6 pmol), there is no adsorption of PG. The amount of
GO adsorbed was ~0.27 pmollg Ru before the addition of PG and ~0.32 pmollg
Ru at the end of the experiments. It is clear that adsorbed GO inhibits the

adsorption of PG, while adsorption of PG leaves sites available for GO to adsorb

on.

When the order is reversed (Figure 6-20) with initial addition of GO (6.5

65

 

 

 

E
.5.
i
8
E
8
0 10 20
Time (hours)

30

 

40

 

 

Adsorption (umol/g Ru)

0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
-0.05

 

0 10 20 30 40
Time (hours)

 

Figure 6-20: Experiment 45, adsorption of G0 (6.5 pmol) followed by PG (6.5
pmol) at 25 °C on 10 g Ru sponge.

The simultaneous adsorption of GO and PG (2.2 GO, 3.3 PG umol) is

shown in Figure 6-21. These results show behavior that is consistent with a G0

preference over PG for Ru and an irreversible interaction with the Ru.

 

0.6
0.5
0.4

Concentration (mM)
9 o
N 0)

9.0
°—L

0 10 20

Time (hours)

 

 

 

 

Adsorption (pmollg Ru)

0.30
0.25
0.20
0.15
0.10
0.05
0.00

 

0 10 20 30
Time (hours)

 

 

Figure 6-21: Experiment 49, adsorption of G0 (2.2 pmol) and PG (3.3 pmol) at

25 °C on 10 g Ru sponge.

 

6.3 lsopropanol and acetone adsorption

Glycerol and PG are C3 species with multiple hydroxyl groups; each
shows adsorption on Ru sponge metal. It is logical to next look at C3 compounds
with one hydroxyl group, namely isopropanol (IPA), n-propanol (n~prOH), and
ethanol. Figure 6-22 shows the adsorption of isopropanol (IPA) on 20 g Ru
sponge at 25 °C. It should be noted that a negative adsorption value seen in
ﬁgures means that a compound is formed. It can be seen in Figure 6-22 that
acetone is formed. This is expected if IPA undergoes dehydrogenation and if
acetone is not strongly adsorbed. The addition of hydrogen at 20 hours converts

all the acetone formed to IPA

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

4, 0.6
0 A2-propanol eacetone omet 05 ‘2-propmol eacetone oAds. IPA
3.5 A A .
3 04
A _ z -
E 3.0 N g 0.3
g 2.5 g 02
£20 A” | x 5 0.1
3 1.5 250psil-l2 ‘E. 0-0
c -0.1
3 1.0 8
2 -0.2
0-5 W -0.3
0.0 r . m -0.4
0 10 20 0 1O 20
Time (hours) Time (hours)

 

 

 

 

 

Figure 6-22: Adsorption of IPA (50 pmol) at 25 °C on 20 g Ru sponge (Exp 109)
showing effect of addition of H2.

It would be expected that if IPA were adsorbed on a H2-saturated surface, we
would see no acetone. Figure 6-23 show that this is the case. Also, no additional
products are seen. It is uncertain why a small amount of IPA adsorbs, but clearly

no dehydrogenation occurs when the surface is ﬁrst saturated with hydrogen.

67

 

.5. 3.0

0 10
Time (hours)

 

 

20

 

 

sorption (pmollg Ru)

Ad

 

10
Time (hours)

20

 

 

Figure 6-23: Adsorption of IPA (50 pmol) at 25 °C on 20 9 H2 saturated Ru
sponge (Exp 1 11) under 20 psi H2.

Figure 6-22 shows a peak that was tentatively labeled as methanol, but

this was an impurity that was there at the start of the experiment. It is curious that

it disappeared when H2 was introduced, which suggests that it might have been a

dissolved gas or a carbonyl species. If it was a carbonyl, we would expect to see

a peak in the UV, but none was seen. Unfortunately, no W peak (210 nm) was

seen for acetone either at these low concentrations, so we cannot rule out the

possibility that this peak is an unsaturated compound. For now, the “methanol”

peak is assumed to be an inert molecule of some sort and is ignored.

68

 

 

2.0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

A IPA e Acetone
.. . eds. GO | . ads. IPA
a I
or
a
E.
c
f
8
'8
<
0 1o 20 30 «1
Time (hours)

 

 

Figure 6-24: Competitive adsorption of GO and IPA (50 pmol each) at 25 °C on
20 g Ru sponge (Exp 112) showing the effect of adding H2.

During the adsorption of IPA, acetone is formed in the ﬁrst 5 hours and then its
concentration stays constant When IPA is adsorbed in the presence of GO
(Figure 6-24), the maximum amount of acetone (~0.5 pmollg Ru) is formed in
about the same time frame as in the adsorption of IPA alone. In the presence of
GO, the acetone that was formed slowly converts back to IPA Presumably the
hydrogen liberated from adsorbing GO reduces the acetone back to IPA.

The next logical experiment was the adsorption of acetone. Figure 6-25
show adsorption of acetone at 25 °C on 20 g of H2-free Ru sponge. It clearly

shows that IPA is formed.

69

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Time (hours)

A2-propanol e acetone
3.5
E 3.0
.5. 2 5 1‘ Added
5 ' '- 250 psi Hz W
5 1.5 OM
:
3 1.0 . I
0.5 .
0.0 It . . and
o 10 20

 

 

0 10

20

 

Time (hours)

 

 

Figure 6-25: Adsorption of acetone (50 pmol) at 25 °C on 20 g Ru sponge (Exp
115) showing effect of addition of H2.

The observation of acetone hydrogenation with no H2 available can possibly be

explained by the dissociation of water on the Ru surface to supply the H2.

Another possibility is that when an acetone molecule adsorbs, it dissociates

completely (Scheme 62) to liberate six H-M. This is sufﬁcient hydrogen to

convert three molecules of acetone to IPA So if the hydrogen comes only from

acetone, then the amount of IPA formed cannot exceed three-fourths that of

amount of acetone adsorbed.
Scheme 62

CO C H

+ M 114 + 2 114 + 6 I
O /
H3C CH3 \ f
H
+ 2 3'4 H3C CH3

70

 

It might be possible to take the carbon all the way to CO2 by a water shift

type reaction (Scheme 62) which could generate 10 more H’s. But there is no

evidence of CO2 liberation and the ~% ratio of ~0.5 umol IPA formed to ~0.8

umol acetone adsorbed per gram of Ru sponge might suggests that the

hydrogen comes from the acetone. The subsequent addition of H2 results in the

rapid conversion of all acetone to IPA.

 

 

5.0

4.5
.. 4.0
E 3.5
g 3.0
25
8 20
c 1.5
3 10

0.5

0.0

 

 

 

 

 

 

 

A IPA e acetone - GO
/ Added acetone
'/ 50 (mol [system
I

 

 

«W

 

 

 

 

 

 

 

45

A
-

50 55
Time (hours)

 

 

Adsorption (pmollg Ru)

2.0

1.5 .
1.0 -

0.5
0.0
-0.5
-1.0
-1.5

 

 

 

 

 

 

 

 

 

 

 

 

 

 

AIPA eacetone
eadsGO oads.Acetone
O
O
s
0
4“; e A n P ’ . . .
A Addedecetone
———-‘"" ‘ ‘ 50mpllsysbm
‘L‘T——
45 50 55 60
Time(hours)

 

 

 

Figure 6-26: Adsorption (50 pmol each) of acetone after GO adsorption at 25 °C
on 20 g Ru sponge (Exp 116) showing the formation of IPA.

It is possible to convert more acetone to IPA if GO is adsorbed on the Ru

surface ﬁrst. Doing this should make more H2 available. Figure 6-26 shows

adsorption of acetone after the saturation of Ru sponge with G0. The quantity of

IPA produced is >1.0 pmol after 7 hours; this is twice as much IPA as is

produced from acetone by itself. Figure 6-27 shows the direct comparison of

these two cases. Because the initial quantities of acetone are not the same, we

can only reach the qualitative conclusion that more IPA is produced from acetone

71

in the presence of some hydrogen donating species such as GO than from

acetone alone.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5.0
45 _ AZ-propanol eacetone
4'0 . acetone' alPA‘
€35 f
.5 3.0 . ’ .
2.5 9., - .
O.
2.0 ’34-: . f
8 :73.‘. ..Oee
1: 1.5 ‘ —
8 1° “' ‘ AAALLA
. ‘ALA‘
05‘1‘;t=“‘
0.0
O 5 10
Time (hours)

 

 

 

Figure 6-27: Comparison of acetone adsorption (50 pmol) at 25 °C on 20 g Ru
sponge (Exp 116, 115) Showing the effects of pre-saturation with G0.
Open symbols are from presaturated experiment.

The total quantity adsorbed in IPA/acetone system is ~0.2 pmollg Ru. This
is independent of hydrogen saturation or which substrate is used ﬁrst. It is
interesting that this is about the same amount as seen in GO adsorption on a
hydrogen-saturated Ru surface. It is possible that there are other adsorption

mechanisms at work.

6.4 n-Propanol and ethanol

The interaction of two primary alcohols, n-propanol (n-prOH) and ethanol,
with H2-free Ru surface was looked at. Both ethanol and 1-propanol adsorbed to
he same extent, as seen in Figure 6-28. Both produced the corresponding acids.
There was production of methane in both cases but more appeared to form from

ethanol than from n-propanol.

72

 

0 5

10
Time (hours)

 

15

 

”I

dsorption (pmollg Ru)
.0 .0 P .0 9 .-‘
O N -A O) on O

A
9‘: .b
A N

10
Time (hours)

15

 

2‘)

 

 

4.0
3.5
E 3.0

a 2.0

a

0 1.5

3 1.0
0.5

0.0

 

0 ethanol

0 10

20
Time (hours)

 

30

40

 

 

PP.“
cacao

0.4

.0
0

Adsorption (pmollg Ru)
0
is

as
.r-io

eads.

10

A AA —Methane

20
“me (hours)

 

30

40

 

 

Figure 6-28: Adsorption (50 pmol loading) of n-propanol (Exp 121) and ethanol

temperature was increased to 50 °C after 18 hours with the following three
results: conversion of n-prOH to PA, increased production of methane, and an

increase in total quantity n-prOH adsorbed. It might be possible that at higher

(Exp 114) at 25 °C on 20 g Ru sponge. Acetic acid (AA) and PA are

formed.

During Experiment 121, adsorption of n-prOH (Figure 6-29), the

temperatures water acts as an oxidant of the Ru, releasing hydrogen and

providing oxygen for oxidation of n-prOH to PA. It is also possible that the

73

adsorbed intermediate “propionaldehyde’ undergoes disproportionation to form
n-prOH and PA

It has been observed that upon addition of H2 the solution concentration of
the acids produced drops sharply followed by a steady conversion to the
corresponding alcohol. This sharp drop suggests adsorption of PA on the newly
reduced surface. It is unlikely that PA reduction takes place, because reduction
of the acids is very slow at 25 °C. This same behavior was also seen in the

adsorption of gluconic acid.

 

5.0
45 e PA A n—prOH Ges
.. 4.0
E 3.5
.5 3.0
2.5
2.0
I: 1.5
8 1.0
0.5
0.0

 

 

 

 

 

 

 

 

 

 

 

Adsorptlon (pmollg Ru)

 

 

 

 

 

 

 

0 20 40 60
Time (hours) Time (hours)

 

 

 

 

 

 

Figure 6-29: Adsorption of n-prOH (50 umol/system) at 25 °C on 209 Ru sponge
(Exp 121) showing effect of addition of H2 and increase in temperature
to 50 °C. Gas C in legend refers to total gaseous carbon.

6.5 Organic acid adsorption

Adsorption of LA was conducted at 25 °C on both H2-free and H2-
saturated Ru sponge metal. Initial loadings ranged from 5—190 umol. Adsorption

of equimolar mixtures of LA and PA were also examined.

74

Figure 6—30 shows concentration dependent adsorption of LA at 25 °C on
10 g of H2-free Ru sponge surface. In each case the adsorption levels off after
~4-6 hours. There were no products seen. The amount of LA adsorbed seems
directly proportional to the initial loading. At 5 pmol loading, the adsorption is
~0.15 umol/g Ru, while at 50 umol/system the adsorption is 1.3 umol/g Ru. At

190 pmol, the adsorption is 4.3 umol/g Ru, which is several times the adsorption

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

seen with polyols.
16.0
A 5umol s 50umol A 5umol e 50umol
14.0 . 93umol - 1soumor 5-° ‘ - 93umol . 190umol
E 12") “AW” . I: 4.0 ..""”
E 10.0 g 30 e
5 ' e
E 8.0 T. c {M
0 o
8 6-0 4 K...- B, 2.0 «——-t
= I
8 4'0 3 W
1.0 ~
2.0 “ 0
0.0 W, 0.0 M ,
0 5 10 15 20l 0 5 10 15 20
Time (hours) Time (hours)

 

 

Figure 6-30: Concentration dependence of LA adsorption at 25 °C on H2 free Ru
surface for different initial loadings. (Exp 54, 56, 57, 58)

The gas phase was not analyzed, but if there was production of signiﬁcant
amounts of methane it would have shown up in the HPLC analysis; no methane
was detected. If we compare the maximum amount of LA adsorbed (4.3 pmollg
Ru) to the gas phase H2 and CO adsorption (~45 and 9 pmollg Ru), LA can
occupy about two metal sites at the highest concentration examined. Control

runs of LA with no Ru present showed no adsorption, but in the presence of air in

75

 

a control run lactic acid was oxidized to pymvic acid, presumably by the tube
walls.

The effects of H2 saturation on adsorption of LA at 25 °C on 10 g of Ru
sponge is shown in Figure 6-31. Unlike the adsorption of polyols, the adsorption
of LA is not limited by H2 saturation. This is not unexpected, because there is no
need to dehydrogenate LA on the surface in order to adsorb a molecule of LA.
Clearly the mode of adsorption is different thanis seen for polyols.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

8.0 3.0
8e ~°M~ .
“6m °‘ F ’
2 - a e w
E = 2.0 e ‘9
E 5.0 g 000 . 50m
0 4 0 5 1 5 .Q 0 5umol‘
E ' e Sumo! g ' o e 93umol
g 3 o o 5umol' b
- . 93umol ‘5 1,0 o 92umor
5 20 e 92urnol' 3 °
1 o 3 0.5 a
.
0.0 dbﬁl l r I" 0.0 A“ 1 T T
—5 O 5 10 15 20 -5 0 5 10 15 20
Time (hours) Time (hours)

 

 

 

 

 

Figure 6-31: H2 saturation effects on adsorption of LA at 25 °C on 10 g Ru
sponge metal (Exp 54,55*,56,60*). *denotes experiments on hydrogen
saturated surface.

The pH was checked after LA adsorption for Exp 54, 57, and 58 and
appeared to be higher than would be expected based on the LA concentration.
Figure 6-32 shows measured and calculated pH vs initial concentration.
Calculated pH is based on simple weak acid dissolving in water. The model
curve is based on the assumption that when a molecule of LA adsorbs, it also

generates one molecule of base.

76

 

.m 9:
O O

P
:3

pH - Adsorption (pmollg Ru)
N 9’
O O

..L
O

.°
0
o L

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

wt
{\2 . .
\Q """"""""""""""""" ‘ “
‘hr
0 HAW”)
. . . .pH (model)

pHaAonly)

_. . pH measured

 

 

 

 

 

 

100

120

140 160 180 200

 

lnkial Loading (movsystem)

 

Figure 6-32: Predicted vs. experimental pH as function of initial loading (pmol).

Only one experiment was run with the adsorption of PA and that was the

competitive adsorption of LA and PA (120 pmol loading each, Figure 6—33). PA

seems to have a slightly lower adsorption (2.5 pmollg Ru) than LA (3.0 pmollg

Ru). These quantities are the same as found for LA by itself at similar loading. if

we look at the total loading of 240 pmol, we see an adsorption of 5.5 pmollg Ru.

This is what would be expected if adsorption is directly proportional to initial

concentration, so it is apparent that LA and PA are not competing for surface

sites in adsorption and that the surface is not saturated under these conditions.

77

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

10.0 35

9.0
A 8.0 A 3 0 .e “uéee 6 “7

O
E 7.0 u 25 —_._-L¢P°°o,; :5 Ana
7'." 6.0 g 20.8)
o 20 v
E 5.0 5 o
c 15

g 4.0 g -
5 3-0 3 1.0

2.0 2

1 0 0.5

' 0 LA 0 PA L ’ LA 0 PA
0.0 0.0 . . 5
-5 0 5 10 15 20 0 5 10 15 20
Time (hours) Time (hours)

 

 

Figure 6-33: Adsorption of equimolar mixture of LA and PA (120 pmol/system) at
25 °C on 20 g of H2-free Ru sponge. (Exp 70)

6.6 Comparisons and conclusions

The differences in adsorption properties of the various organic substrates
on Ru sponge metal are contrasted in Table 6—1. The total quantities adsorbed
for acids is ~4 pmol/g Ru, while that of all polyols, except sorbitol and
isopropanol, is ~0.8 pmollg Ru. These results suggest that there two different
methods of adsorption. The molecular size does not vary signiﬁcantly between
LA and C3 polyols, ruling out simple sten'c effects as an explanation for the
differences observed.

For polyol adsorption, a minimum of one terminal hydroxyl group is
needed; this is supported by the lack of adsorption in the case of IPA. The
inhibition of adsorption of polyols by hydrogen pre-saturation of the Ru surface is

evidence supporting dehydrogenation as the ﬁrst step hydrogenolysis of polyols.

78

Table 6-1: Contrasting adsorption properties on Ru sponge of organic acid and

 

polyol.

Organic Acids C3 polyols
Adsorption ~4 pmollg Ru Adsorption ~0.8 pmollg Ru
No H2 inﬂuence Strong H2 inﬂuence
Consistent with H2 adsorption Consistent with ~30Azl molecule

(2 Ru atoms! acid molecule) (~10 atoms/molecule)
No degradation Dissociative with degradation

 

79

7 AQUEOUS PHASE ADSORPTION ON ACTIVATED
CARBON

7.1 Experimental methods

7.1 .1 Carbon characterization

The carbons used in this study were a 0.8 mm extrudate activated carbon
(ROX 0.8, lot #520020, Norit Americas, Marshall, TX) and a powder activated
carbon (designation 3310, lot #28850, Johnson Matthey, Sevierville, TN). Both of
these carbons have served as support materials for catalysts we have used in
GO hydrogenolysis, and lactic acid hydrogenation studies. The total surface area
of each carbon was characterized by BET nitrogen physisorption at 78 K over a
relative pressure of PlPo from 0.0 to 0.2 in a Micromeritics ASAP 2010
(Micromeritics Instrument, Norcross, GA). Surface area was calculated from the
BET equation; micropore volume was determined using the t-plot method and
total pore volume was characterized as volume adsorbed at the maximum
relative pressure of 0.99. A summary of the carbon characterization results is

given in Table 7-1.

80

Table 7-1: Carbon characterization by N2 adsorption at 78 K

 

 

Carbon Type 3310 ROX
BET Surface Area (m’lg) 715.6 833.6
Micropore Area (m’lg) 374.8 585.8
Total Pore Volume (ems/g) 0.654 0.536
Micropore Volume (cm’lg) 0.173 0.272

 

7.1.2 lsotherrnal adsorption measurements

The quantity of material adsorbed onto activated carbon in this study was
determined by the difference in initial and ﬁnal species concentration in solution,
which were measured prior to and following exposure to activated carbon
catalyst supports, respectively.

CA0 _ CA) . Vsolutr'on
m

CAS=(

 

(7-1 )

carbon
Isothermal adsorption experiments at 25°C were performed using 8.5 mL
glass vials with Teﬂon-lined plastic lids. The vials were initially washed in HPLC-
grade water, air-dried, and weighed in preparation for experiments. Carbon was
weighed and added to the vials based on the total concentration of the solute(s)
to be studied (0.1g carbon 5 0.1 M < 0.59 carbon 5 0.750 M < 1.09 carbon 5 2.0
M) in order to maintain at least a 15% change in solution concentration before
and after adsorption. A quantity of solution of known concentration was then
added to the vials to give approximately 7mL total solution + carbon in the vial.

The ﬁnal vial weight was then recorded for analysis following reaction. The vials

81

were capped and rotated end-over-end on a rotator overnight to ensure thorough
mixing and equilibration. Upon removal from the rotator, the vials were either
centrifuged or left standing for approximately 30 minutes to allow the suspended
carbon in the sample to settle. One milliliter samples were then taken and

analyzed via HPLC using the method described below.

7.1.3 Elevated temperature adsorption experiments

Adsorption measurements were performed at elevated temperatures
(40°C - 160°C) using a Parr 5000 multireactor system (Parr Instrument
Company, Moline, IL). This system has six 75mL stainless steel reactors with
internal stirring, independent temperature control, and continuous pressure
monitoring. For these experiments, the reactors were cleaned and air-dried, and
than carbon was added to the reactors according to the solution concentration to
be examined: one gram of carbon was used for 0.05 M and 0.2 M concentrations
of GO or PG, and four grams were used for a 0.5 M concentration. A Teﬂon stir-
bar and 60mL of solution were added to the reactor. All starting, intermediate,
and ﬁnal weights were recorded. The desired temperature was set and the
reactor contents were held for at least two hours at the desired temperature, after
which a liquid-phase sample was taken. (The two-hour equilibration time was
veriﬁed by room temperature experiments in vials for different time periods from
0.5 to 24 hr.) Sampling consisted of removing 1.5mL of waste to clean the
sampling line, followed by removal of 1.0 mL as a sample. Typically, multiple

temperatures between 40 and 160°C were examined in each experiment. All

82

waste aliquots and samples were collected and weighed at the conclusion of the
experiment to check for mass loss via system leaks. Samples were analyzed

using the HPLC method described below.

7.1.4 Analysis

All samples from adsorption experiments were analyzed using a high
pressure liquid chromatography system consisting of a Waters 717 Plus
autosampler (Waters Corporation, Milford, MA), a Perkin Elmer pump (Perkin
Elmer, Wellesley, MA), a Waters 410 differential refractometer, and a Perkin
Elmer LC 90 UV spectrophotometric detector. The system used a Bio-Rad
HPX87H column (Bio-Rad Laboratories, Hercules, CA) with 5mM sulfuric acid
solution as the mobile phase. The column was operated isocratically at 50°C
and a mobile phase ﬂow rate of 0.6 mllmin.

A sample injection size of ﬁve microliters was used for samples above 0.1
M concentration; ten microliter injections were used for samples below or at 0.1
M. Samples with a concentration above 0.5 M were diluted by a factor of four to
maintain an injected concentration between 0.1 M and 0.5 M. A three-point
calibration curve was used to determine response factors - no internal standards

were thus used in determining species concentrations.

7.1.5 Adsorption modeling

The remainder of this chapter will discuss modeling of single component

and multi-component adsorption; therefore the general method for ﬁtting

83

experimental data to Langmuir and Freundlich adsorption models is discussed in

this section. Unless noted, the notations used for modeling are give in Table 7-2.
All data from single component vial adsorption experiments at 25°C were

modeled using both the Freundlich and Langmuir isotherms to obtain adsorption

constants at room temperature.

 

Freundlich: CA3 = K). C A" (7-2)
Langmuir' CAS = KA CA CTA (7-3)
' l + K A C A

2(Cisi - CZEW

Goodness ofFit: X2 = "1 n (74)

 

For the Freundlich isotherm, the coefﬁcients K; and n were found by a
least-squares ﬁtting of experimental data as CA8 vs CA with predicted CAs using
solver in Microsoft Excel. Goodness of ﬁt (x2) is the average residuals squared.
The coefﬁcients for the Langmuir isotherm (KA, Cm) were found using the same
procedure. Our published adsorption of GO and PG on activated carbon uses the
more traditional method, where the Freundlich isotherm coefﬁcients, K; and n
were found by a least-squares linear regression of experimental data as In(CA.)
vs. ln(CA) to give slope n and intercept In(Kp). Goodness of ﬁt (R2) is given by
Microsoft Excel. For the Langmuir isotherm, a plot of experimental data as
(CA/CA.) vs. CA gives a slope 1ICTA and an intercept (1/KACTA). The traditional
method was more cumbersome to use and also tended to weigh small quantities

adsorbed more heavily.

Table 7-2: General notation used for modeling adsorption on activated mrbon
CA Observed solution concentration of species A in equilibrium with carbon
(M)
CA0 Initial solution concentration of species A prior to adsorption (M)
CA3 Concentration of solute A in activated carbon adsorbent (mol/kg carbon)
Cm Maximum concentration of A in activated carbon adsorbent (mol/kg carbon)
AHA Heat of adsorption (kJ/mol)
Ka Van’t Hoff constant (M")
KA Langmuir adsorption model equilibrium constant for adsorption of
solute A (M“)
K).- Constantin Freundlich isotherm model (mol/M"lkg)
mm" mass of carbon in solution (lg)
n Exponent in Freundlich isotherm model equation
R Ideal gas constant (8.31 JImol/K)
T Temperature (K)
VWVolume of solution (liter)
Subscript A,B Denotes component, either glycerol or propylene glycol

7.1.6 Summary of experiments performed

In our lab, the catalyst used for most hydrogenation/hydrogenolysis work
is 5% RuIC (PMC 3310). For this reason, the majority of adsorption experiments
are on 3310 carbon, the support used on the RulC catalyst obtained directly from
the manufacturer. The substrates were selected to match with Chen’s‘58
hydrogenation experiments and include organic acids, primary alcohols, diols,
and glycerol. Examination of an additional carbon, Norit ROX, was included to
help with our mechanistic investigations of glycerol hydrogenolysis; on that
support only GO and PG were studied. Table 7-3 summarizes adsorption

experiments covered in this work.

85

Table 7-3: Summary of experiments preformed during the study of aqueous
phase adsorption on activated carbon.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Compound 3310 carbon ROX carbon Reduced Catalyst
(5%RuIC 3310)

F orrnic acid (FA) I

Acetic acid (AA) I

Glycolic Acid (GA) I

Propanoic acid (PA) IN V

Lactic acid (LA) IN V

Ethanol (EtOH) I

n-propanol (n-prOH) IN V

Ethylene glycol (EG) I,V

Propylene glycol (PG) IN IN

1,3-propane diol (1 ,3—PD) I

Glycerol (GO) IN IN

LA&PA EI,xV

LA&PG El, xV

LA&n-prOH El, xV

PA&PG XV

PA&n-prOH EI,xV

GO&PG xl,xV xl,xV

EG&PG El

 

 

 

=isothermal adsorption; El= equimolar isothermal adsorption; xl=variable mole fraction
isothermal adsorption; V=variable temperature adsorption; EV= equimolar variable temperature
adsorption; xV=variable mole fraction variable temperature adsorption.

Single component adsorptions on 3310 carbon were done with organic
acids, primary alcohols, diols, and glycerol. Adsorption of equimolar mixtures of
LA&PA, LA&PG, PA&PG, GO&PG, and EG&PG were examined. Variable
temperature (40°C - 160°C) adsorption measurements of LA, PA, n-PROP, EG,
PG, GO, LA&PA, and GO&PG were done on 3310 carbon.

Control experiments were carried out in vials or in the reactor without
addition of carbon. No degradation or loss of PG or GO was observed at any
temperature. Experiments in which total vial or reactor weight after experiment
differed from initial weight by more than 1 2%, indicative of a leaking vessel, were

discarded.

7.2 Glycerol and propylene glycol

7.2.1 lsothennal adsorption of GO+PG on ROX activated carbon

Adsorption experiments were conducted at room temperature (25°C) in
glass vials for substrate solution concentrations ranging from 0.01 M to 2.0 M.
Figure 7-1 compares the experimental quantities adsorbed with those predicted
by the Freundlich and Langmuir isotherm models for individual adsorption of GO
and PG on ROX activated carbon. A summary of the calculated parameters at
room temperature for each adsorption isotherm is given in Table 7-4. As can be
seen, the Langmuir isotherm gives the best ﬁt of GO adsorption over the entire

concentration range studied.

87

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2.5 1.5
/-
52.0 ‘5 A /
a A‘ ‘
0
w 'l-". "I
5...
§ - ”x ‘ PG
2 0.5 _ I, " —— Imgmuir PG
1!, I I GO
0.0 . . ' 1 , ---------- Imgmuir GO
A 500 1000 1500 2000 4“ 0 100 zoo—1500
Solution concentration (ml) A Solution concentration (ml)
2.5 1.5
A A 1.3 ‘ p
0 2.0 - /¢ 6‘
g / g 1.1 ‘ /
g AL --“. O 1/ I r
5 ' 7 ‘g’ 0.7 / _.'.' .»--"""
§ Freundich PG § ' y; . PG
3 0'5 o co 3 0-3 ’- Fremdich PG
A
.......... Fremdich so 0.1 Tr————' » I GO
0.0 I I I I I """""" FMdiCh GO
B 500 1000 1500 2000: ‘0-1 0 1 ﬂ
Solution concentration (mM) B Solution concentration (mM)

 

 

Figure 7-1: Experimental and predicted a) Langmuir and b) Freundlich

adsorption isotherms for PG and GO at 25°C on ROX carbon.

For PG, the Langmuir model gives the best ﬁt below 0.75 M, but the Freundlich

model better describes the data at concentrations above 0.75 M. The observed

isotherm for PG suggests that the quantity adsorbed reaches a plateau at the

Langmuir maximum of Cu ~1.64 mol/kg at about 0.75 M in solution and then

adsorbs by another mechanism at higher concentration. In Section 7.5 a new

adsorption model is proposed that takes in to account two different modes of

adsorption, micropore ﬁlling, and surface adsorption. In micropore ﬁlling the

88

adsorption energy is signiﬁcantly higher than that found for surface adsorption.
Micropore ﬁlling dominates at low concentrations, while at high concentrations
both modes contribute to the observed adsorption. Filling of the micropores has a
relatively constant energy of adsorption and is best represented by the Langmuir
isotherm. In the F ruendlich adsorption isotherm, adsorption energies are
dependent on coverage, making it better able to predict adsorption at higher

concentrations where both adsorption modes are active.

Table 7-4: Langmuir and F reundlich isotherm coefﬁcients for GO and PG on

 

 

ROX carbon.
60 PG GO PG
Langmuir Freudlich
CTA(mol/kg) 1.77 1.64 K; 0.105 0.202
KA.293K(M") 4.755-03 1.63E-02 n 0.379 0.323
R2 Fit 0.983 0.997 R2 Fit 0.946 0.916

Glycerol shows no tendency to adsorb beyond its Langmuir maximum of Cu =
1.77 mol/kg. While the reported value of the Langmuir Cm for GO is ~8% larger
than that for PG, we do not believe that the difference in the two values is
signiﬁcant. Because GO and PG have nearly identical molar volumes of 73.0
mL/mol, the volume of species adsorbed according to the Langmuir model is
likewise similar. It is signiﬁcant, however, that the molar quantity of PG adsorbed
is greater than that of GO at any concentration below the Langmuir maximum;
this is reﬂected by the fact that the calculated adsorption equilibrium constant for

PG is approximately three times that for GO. Interestingly, the mass of GO and

89

PG adsorbed at any concentration is nearly identical (Figure 7-2); we assume

this is coincidental but worth noting.

 

180

Adsorption (mglg C)
essaseéee

APG I GO

 

0 500 1000 1500 2000
Solution concentration (mM)

 

 

 

Figure 7-2: Adsorption of PG and G0 on ROX carbon on a mass basis.
Using the measured ROX carbon micropore volume of 0.27 cmalg, the

micropore concentrations of PG and GO corresponding to Cu at room
temperature can be calculated to be 6.07 M and 6.55 M, respectively. This is
44% and 48%, respectively, of the pure component molar concentration of 13.7
M for both PG (1.04 glcma) and G0 (1.26 glcm3). Assuming that the density of
PG/water and GOIwater varies linearly with mass concentration, the remaining
micropore volume is ﬁlled with water (the only other species present) at a
concentration of 29.0 M for GO and 31.0 M for PG. This approximate 5:1 molar
ratio of water to GO or PG is similar to the result reported by Parsons et al. 69 in
their analysis of glycerol aqueous solutions - they report that all water is
associated with glycerol and solution behavior changes signiﬁcantly at a glycerol
mole fraction of ~0.15 (5.6:1 molar ratio). Thus it wpears that glycerol retains
water associated by hydrogen bonding when adsorbed into the carbon. On the

other hand, the continued adsorption of PG beyond the Langmuir model value of
90

Cu may be because PG is less hydrophilic than PG and thus sheds solvating
water to reach a higher concentration in the pore as PG solution concentration is
increased above 1.0 M. A second paper by Parsons and Koga 7° on PG-water
solutions supports this interpretation of lower PG hydrophilicity in solution. The
less hydrophilic nature of PG also accounts for its larger adsorption equilibrium
constant relative to GO; PG favors residence in the carbon pores rather than in

bulk solution.

7.2.2 lsothennal adsorption of GO+PG on 3310 activated carbon

Isothermal adsorption on the Johnson-Matthey 3310 powder activated
carbon, as with the ROX carbon, follows the Langmuir isotherm model closely for
PG and GO solution concentrations up to 0.75 M. A summary and comparison of
the results for the ROX and 3310 carbons ﬁt to the Langmuir isotherm are found
in Table 7-5. As can be seen, the values of CT for GO and PG are similar on the
3310 carbon, in agreement with the results found for the ROX carbon. If the
3310 carbon micropore volume of 0.17 cm3lg is used, the concentrations‘of
species in the micropores at the maximum Langmuir concentration Cu are 6.35
M and 7.0 M for GO and PG, respectively. These values are very close to the
values of 6.5 M and 6.1 M obtained on the ROX carbon, indicating that
adsorption is strongly associated with the micropore volume of each carbon.
Further, if the additional PG adsorption (up to 2.0 mol/kg) on the ROX carbon is
considered, the micropore concentration for PG on the ROX carbon becomes
7.35 M, very close to the PG value of 7.0 M for the 3310 carbon. Thus the

absolute and relative quantities of PG and GO adsorbed in the micropores of the

91

two carbons are similar; this is supported by the ratio of adsorption constants
(erlKeo) at 298 K (Table 7-5), which are 3.6 and 3.4 for the 3310 carbon and
the ROX carbon, respectively. However, the afﬁnity of PG and GO for the 3310
carbon is greater than that for the ROX carbon, as the ratio (3310/ROX) of
adsorption constants (KA) at 25°C for GO and PG is 1.32 and 1.39, respectively.
The greater adsorption afﬁnity for the 3310 carbon may result from differences in
surface functionality or in micropore structure, considering the two carbons come

from different sources.

Table 7-5: Comparison of Langmuir isotherm coefﬁcients for ROX carbon and

 

 

3310 carbon.
Carbon Type CTA (MOI/k9) KAZG c ("4)
G0 ROX 1.77 4.75E-03
3310 1.10 6.25E-03
PG ROX 1.64 1.53E-02
3310 1 .22 2.28E-02

 

7.2.3 Elevated temperature adsorption

Quantities of GO and PG adsorbed onto ROX and 3310 carbon were
determined at concentrations from 0.1 to 0.5 M and temperatures from 313 to
433 K. At each temperature, the equilibrium constant (Kit) for the Langmuir model
was calculated using CTA determined from isothermal adsorption experiments at

298 K. The equilibrium constant is assumed to vary with temperature based on

92

the Van’t Hoff equation, as the Langmuir model assumes all adsorption sites

have equal binding energies and thus the heat of adsorption (AH) is constant.
1n(1<,,)= —%*%+ln(1€o) (7-5)

For each species, the plot of In (KA) vs. 1fT (absolute) from experiments
was ﬁtted with a straight line via least squares linear regression to give the heat
of adsorption (AH) from the slope (-AH/R) and the preexponential constant (K0)
from the intercept |n(Ko). The plots for GO and PG on ROX and 3310 are seen in
Figure 7-3 and the calculated heats of adsorption and preexponential factors are

given in Table 7-6.

 

 

 

 

 

 

 

 

 

 

 

 

 

9 9
8 . = -1913.7x + 11.771 8 _ y = -2373.3x + 12,516
7- . . R2=0-804 7- - . _ 52:0.8031
6 _ I I I 6 — ‘
A A I
g 5 - I g 5 ‘ I
r: 4 « ‘E 4 - A
"' 3 _ I GO y = ~1765.9x +1004 -. 3 _ . GO ‘ A
f 1 A PG R2 = 03593 2 « 3 PG y = -2079.8x + 10.561
0 ‘ ——Lrnear 1 - —|_jnear R2 = 0.7011
. . o r ,
0.002 0.0025 0.003 0.0035 0 002 0_0025 (1003 0.0035
A 1” (K) B 1IT (K)

 

Figure 7-3: Temperature dependence of Langmuir equilibrium constant Kit for
GO and PG on ROX (A) and 3310(8) carbon.

It is seen that the data follow the Langmuir model well over the measured
temperature range of 298 K - 433 K for both carbons. The heats of adsorption of
GO and PG are both approximately 15 kJ/mol (3.6 kcal/mol) on ROX carbon,
while on 3310 carbon G0 is 14 kJ/mol and PG is some what higher at 22 kJ/mol.
This is indicative of similar van der Waal’s type forces and weak hydrogen

bonding responsible for adsorption of each species.

93

Table 7-6: Langmuir isotherm coefﬁcients for GO and PG on ROX carbon

 

ROX 3310
GO PG GO PG

 

 

Langmuir
CTA (mol/kg) 1.77 1.64 1.10 1.22
KA295K (M’ 1) 4.75E-03 1.63E-02 6.25E-03 2.28E-02
AH (kJ/mol) -15.91 -14.68 -14.02 -22.32
Ko(M“) 7.73E-06 4.36E-05 1.95E-05 4.36E-06
R2 Fit

 

 

7.2.4 Two component adsorption studies

Aqueous solutions of PG and GO at total species concentrations ranging
from 0.05 to 0.5 M and species fractions of PG and GO ranging from 0 to 1.0
were conducted at room temperature on ROX carbon and at elevated
temperatures on both ROX and 3310 carbon. The extended Langmuir model
(Eq. (7-6) and (7-7)) has been applied as the model of choice for mixed solute
adsorption; the denominator of the extended Langmuir model accounts for
competitive adsorption of the two species (A=GO; B=PG) into the activated

carbon micropores.

 

= KACACTA (76)
AS 1+KACA+KBCB
KBCBCTB (7_7)

 

”3 = 1+ KACA + KBCB

All coefﬁcients (Cu, Cm, AHA, AHB, KM, K30) required in the extended
Langmuir model were taken from the single component results reported in Table
7-6.

Figure 7-5 shows the predicted dual component adsorption isothen'ns on
ROX carbon at 25 °C and total initial concentrations of 0.05 M, 0.3 M, and 0.5 M.
The abscissa in Figure 7-5 is the fraction of glycerol in the total solute present
(GO/(GO+PG)) - not mole fraction in solution. The data are reported in this
fashion to show comparison of model with experiment and accentuate the
different adsorption behavior of GO and PG. It is seen that the extended
Langmuir model accurately predicts the adsorption of propylene glycol and
glycerol over the studied concentration range. The preference for PG adsorption
is clear; in fact, at equimolar concentrations the quantity of PG adsorbed is
approximately three times that of GO, in accordance with the ratio of equilibrium
constants determined in single component experiments. The adsorption on 3310
shows similar behavior as ROX only with higher adsorption quantities. Predicted
and experimental isotherms for GO and PG on 3310 can be found in Appendix B
(Figure B-1). It should be noted that predicted GO values are higher then

experimentally found but the experimental data has a lot of scatter.

95

 

ROX Carbon

1.60
1.40
1.20
1.00

predicted adsorption (mmol/g
Cl

 

9.0.0.09
88888

p
8

0.50 1.00 1.50 2.00
experimental adsorption (mmollg C)

 

 

1.40
Carbon 3310

g 1.20

2

g 1.00

C

3 030

go

1, 0.60

a

U 0 40

g ' 6 ° x=y

E, 0.20 0 GO

 

APG

0.00
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80

experimenhl adsorption (mmollg C)

 

 

 

Figure 7-4: Experimental vs predicted adsorption of PG and GO on ROX carbon
and 3310 carbon for all data.

Performance of the extended Langmuir model for GO and PG on ROX
carbon and 3310 is shown in Figure 7-4, which compares experimental to
predicted values. All data are included (single and dual component isothermal

and variable temperature data)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1.6
14 APG IGO‘ 250‘;
° 60m"
1.2
‘6
310
E09
30.6 I\‘
0.4 Tr i
- _____
0.2 \.\
-. I‘
doc-4 . . \
0.0 02 0.4 0.6 0.8 1.0
390

 

 

 

 

 

 

 

 

 

 

 

1.6
14 APG IGOI 25°C
' 300mm
1.2 4
£10 A \
E06 mil—4
0 I,
0.4 T
_ .d'
0.2 9
0.02". . . T
0.0 0.2 0.4 0.6 0.9 1.0
no

 

 

APG I Go 2593
500mm“

 

 

 

 

 

 

 

 

 

0.00-"’2 . . .
0.0 02 0.4 0.6 0.8 1.0

 

 

 

yeo

 

Figure 7-5: Experimental and predicted Langmuir isotherms for mixtures of GO
and PG on ROX carbon at 25 °C and 0.05 M, 0.3 M and 0.5 M total
species concentration. Abscissa is fraction of GO in the total species
(G0 + PG) present. Lines are the predicted isotherms.

Figure 7-8 shows the predicted isotherms at a total initial concentration of 0.5 M

and temperatures of 80 °C, 120 °C, and 160 °C. The predicted values again

show good agreement with experimental data, and validate the assumption of

constant heat of adsorption over the range of concentrations and temperatures

 

 

 

 

 

 

 

 

 

 

 

 

 

 

investigated.
1.6 1.6 1-6
14 APG IGOl 80°C 14 APG IGOI 120°C. 14 APG 'GOF180‘64
' 500nm ' WM . m
1.2 1.2 - 12 .
a10 ‘ §10A 310
= . ‘ . = .
9 l
0.8 \ . E08 \ E03 X
30.6 ‘ './' 30.6 \f ‘£06 A
I g .r
o , o ,. o \
0.4 .X 0.4 ’ .5 0.4 \{
0.2 _1 0.2- .. ' 02 i
”I" - —"" .... .'
000'". . . . 0.00-‘2' . . . 0.0-4L. . .
0.0 0.2 0.4 0.6 0.3 1.0 0.0 0.2 0.4 0.6 0.9 1.0 0.0 0.2 0.4 0.6 0.3 1.0
too we yeo

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 7-8: Experimental and predicted Langmuir isotherms for mixture of GO
and PG on ROX carbon at 0.5 M total species concentration and 80 °C,
120 °C, and 160 °C. Abscissa is fraction of GO in the total species (G0
+ PG) present. Lines are the predicted isotherms.

97

The adsorption data clearly show that PG and GO concentrations in the carbon
pore structure are signiﬁcame different than those in bulk solution. This
obviously affects the interpretation and kinetic modeling of reactions taking place
over activated carbon catalysts. Figure 7-7 shows the extended Langmuir model
predictions of GO and PG pore concentrations, based on the measured
micropore volume, in ROX activated carbon at a temperature of 180°C for
different bulk GO and PG concentrations. GO pore concentration is only
modestly different than GO solution concentration at this temperature, while PG
pore concentration is approximately ﬁve-fold larger than its corresponding

solution concentration.

98

 

 

 

 

 

 

 

 

 

 

 

 

 

1.2
—0M
5 1.0- ——0.1M .t
: —"02M ”*”
O 0.84 . /’ ‘/‘
E ----0.4M / .4" .1
’/n” eeee
g 0.6~ —0.7M ’/"’ """""
8 ’6' ’.-x‘.
g 0.4~ /,’ ‘,."‘
8. /.”,’."'
g 0.24 (4“
0.0 t
A 00 02 0.4 08 08 10
GO solution concentration (M)
2 01M
2 13° N ‘ —--0.2M
5 15‘ *— --- 0.4M
1.42 _______________ ——07M
1.2~ """" X ------ . Series1
g 1 ~~n
00.82 "'“""---—_..
—-—--—_-+___—--.
508- _________
00.4« _—"'-—"—-—-X—_.
F 0.2. )L
o I 1 T I
0.0 0.2 0.4 0.8 0.8 1.0
B 60 solution concentration (M)

 

 

 

Figure 7-7: Extended Langmuir model predictions of a) GO and b) PG pore
concentrations in ROX carbon at 180 °C. Symbols represent trajectory
of 1.0 M GO conversion to PG.

The presence of PG lowers GO pore concentration signiﬁcantly, while the
presence of GO only affects PG pore concentration slightly. The “x” symbols in
Figure 7-7 represent an idealized (100% selective) reaction trajectory of GO
hydrogenolysis to PG starting with 1.0 M GO solution. During reaction, GO pore
concentration declines more rapidly than does solution concentration -
converting 70% of GO to PG, which lowers solution GO concentration from 1.0 M

99

to 0.3 M, reduces the corresponding GO pore concentration from 1.15 M to 0.19
M, a six-fold decrease. At the same time, PG concentration increases from zero
to 0.7 M in solution, and to nearly 1.8 M in the pore. These differences deviate ’
from bulk solution concentrations to a signiﬁcant enough extent that they must be
accounted for in accurate kinetic modeling of GO hydrogenolysis over activated
carbon catalysts. Moreover, the local concentrations may depend on the
particular carbon support in use, so that any comparisons between carbon
catalysts must include comparison of the adsorption properties of reactants and

products.

7.2.5 Conclusion

The adsorption of propylene glycol and glycerol over two activated carbon
supports has been examined over a broad range of concentrations (0.005 M to
1.5 M) and temperature (298 - 433 K). Adsorption of individual species is well-
represented by the Langmuir isotherm, and simultaneous adsorption of the two
species is described by an extended Langmuir model that uses the single—
component adsorption constants and accounts for competitive adsorption
between the two species. The overall quantities of PG and GO adsorbed are
comparable and are closely related to the carbon micropore volume. Propylene
glycol exhibits an enhanced afﬁnity for adsorption relative to glycerol, with its
equilibrium constant approximately 3.5 times that for glycerol. This enhanced
adsorption is attributed to the less hydrophilic nature of PG, which favors its

presence in the carbon micropores vs. the bulk solution, and also accounts for

100

the lower extent of solvation of PG via hydrogen bonding. The adsorption model
shows that pore concentrations in liquid phase reactions using activated carbon
catalysts can be signiﬁcantly different than the corresponding concentrations in

the surrounding solution.

7.3 Ethanol, n-propanol, EG, and 1,3-PDO adsorption

Adsorption of ethanol, n—propanol (n—prOH), ethylene glycol (EG), and 1,3-
propanediol (1 ,3-PDO) are compared in Figure 7-8 and include the Langmuir
and Fruendlich predicted adsorption. It should be noted that both models were ﬁt
over the entire concentration range. The Fruendlich model predicts the observed
adsorption better for all species at concentration >200 mM. The predicted
concentration needed to adsorb 1 mmol of substrate per gram carbon is 734 mM
for EG (0.017), 213 mM for 1,3-PDO (0.091 ), 162 mM for ethanol (0.49), and 34
mM for n-prOH (1.78), this is consistent with the order of the K0‘" value given in
parenthesis. Kow is the distribution coefﬁcient between octanol and water as
deﬁned in the literature.71

On 3310 carbon both the Langmuir and Fruendlich adsorption models
perform worse than on ROX carbon, which could be due to the large micropore
volume found in ROX. In Section 7.5, a new adsorption model is proposed that
gives better performance for organic acids, PG, and n—prOH. It is expected that it
would also perform better for ethanol, EG, and 1,3-PDO but it was not tested on

these compounds.

101

 

 

 

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3-0 2.0 I 2"
e e 1.8 I A 7 -
532.5 / ————————————— 81 6 .I
2 " . ° u ' [.3
.520 1’9. [Av If", 31.‘ ’
E ’ ”. ___. E1 2 2'
E ........... I --------- E - .' /-
:1.5 - " :1.0 7 ,/ A
8 n 80.3 s / f- """ II
W / O/ .-
3” ‘ ‘ e0.6 , , a?"
20 5 $0.4 zit-L ./ i
- ‘en-prOH cEthanol ‘02 ﬂ 1' cn-prOH cEthanol
I1,3-PDO AEG ' r I1,3—PDO AEG
0.0 t t r . 0.0 i . r
A 0 500 1000 1500 2000. A 0 50 100 150 200
Solution concentration (mM) Solution concentration (mM)
3.5 2.0
, . .. - ’ 1.8 J . .
£25 .7 ’ / £14
0 , 4 ,e ’ O
r v ........
: ,' /n --------- r :10 / /4,_»
01.5 --f——O'7H / Po 3 - 4 """""""" Q.
a , A g - :v ________________
31.0 i / en—prOH 30-6 - cn-prOH
‘2 c Ethanol 20.4 c Ethanol
0'5 I1,3'Pm o 2 I 1,3'PDO
EG ' EG
0.0 . T . ‘ - 0.0 . ‘ .
B 0 500 1000 1500 2000 B 0 50 100 150 200
Solution concentration (mM) Solution concentration (mM)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 7-8: Experimental and best ﬁt data for isothermal adsorption of n-
propanol (n-prOH), ethanol, 1,3-propanediol (1 ,3-PDO), and ethylene
glycol (EG) on 3310 carbon at 25 °C. (A—Langmuir, B-Fruendlich)

7.4 Organic acids

The effect of chain length was examined in the adsorption of formic acid

(FA), acetic acid (AA), and propanoic acid (PA). Figure 7-9 compares the

experimental quantities adsorbed with those predicted by the Freundlich and

102

Langmuir isotherm models for individual adsorption of FA, AA, and PA on 3310

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

carbon.
A 3.50
5
.e 3.00
8
a 2.50
6
E 2.00
5 1.50
8
a 1.00
E-
8 0.50
2 000 eFA IAA APA ' cFA IAA APA
A 0 500 1000 1500 2000 A 0 50 100 150 200
Solution concentration (mM) Solution concentration (mM)
‘5 3.50
g 3.00
a 2.50 .
6
E 2.00 —
5 1.50
c
i 1.00
3 0.50 3 '
g 000 leFA IAA APA 3 ' eFA IAA APA
0 500 1000 1500 2000 B 0 50 100 150 200
3 Solution concentration (mM) Solution concentration (mM)

 

 

 

 

 

Figure 7-9: Experimental and best ﬁt data for lsothen'nal adsorption of formic,
acetic, and propanoic acid on 3310 carbon at 25 °C. (A—Langmuir, B—
Fruendlich)

As can be seen, the F reundlich isotherm gives a better ﬁt for all three
acids over the entire concentration range studied. Examining the adsorption at
high concentration, we see that AA and PA adsorb ~3 mmol/g C while the FA is
closer to 2.2 mmol/9 C. The last two data points for the FA do not ﬁt the
Freundlich model. Including these points in ﬁtting the Freundlich model results in
decreasing adsorption (~2 mmol/g C) while overestimating the adsorption at low

103

concentrations. Looking at the predicted concentration needed to adsorb 1 mmol

of acid, PA (16 mM) is adsorbed much more strongly than AA (72 mM) which is

stronger then FA (176mM). This behavior could simply be explained by the

increased organic character from FA<AA<PA. A similar trend can be seen in the

octanol-water distribution coefﬁcient (Km) for each acid; 0.58, 0.84, 1.39.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2.00 1.40
A 1.” A
0 o . , .
«E: 1.40 E 1-00 ,1'
E133 g 0.80 '.
g 0:80 8 0.80 15%
3' 0.60 E 0.40
8 0,40 8 - LA ————Fit
< 020 2 0.20 I A GA ---------- Fit
0.00 0.1!) T .
A 0 500 1000 1500 2000 A 0 100 200 300
Solution concentration (mM) Solution concentration (ml!)
1.40
‘7 2.00
o G
E 2 1.20
'o A ‘6 1.00 I ,.
g g 0.80 [#7"
g1 00 S 0.60-
0.40
3 0.50 3
1: -———I-”rt 1: 0,2) I LA ————Fit «
< .......... Fit < . GA .......... Fit
0.00 i I 0.00 T .
B 0 500 1000 1500 2000 0 100 200 300
Solution concentration (mM) Solution concentration (mM)

 

 

 

 

 

 

 

   

 

 

 

 

 

 

 

 

 

 

Figure 7-10: Experimental and best ﬁt data for isothermal adsorption of glycolic
and lactic acid on 3310 carbon at 25 °C. (A-Langmuir, B-Fruendlich)

Glycolic acid and lactic acid are compared in Figure 7-10 and show

almost identical adsorption behavior. It should be noted that LA data only goes to

800 mM; this was done to prevent any dimer adsorption. Again, the Fruendlich

104

model nicely predicts the observed adsorption of ~2 mmol/g C. The adsorption of
GA appears lower than LA, but this is well within experimental error. The
predicted concentration needed to adsorb 1 mmol of acid is 190 mM for GA and
142 mM for LA; this puts it in the same range as FA (176 mM). LA appears to
adsorb to a slightly higher extent than FA, which is counter to the behavior
predicted from the Kow values LA(O.49) and FA(0.58). GA has the lowest Kw,
value of 0.33, and is also the least adsorbed.

Figure 741 shows the adsorption of equimolar mixtures of PA+LA and
PA+PG on 3310 carbon at 25 °C. The Fruendlich isotherm prediction of
adsorbed pure PA is shown for reference. Solution concentrations and the
amounts adsorbed were each summed for PA and LA and plotted as PA+LA. PA
adsorption in the mixture is less than the pure component adsorption, indicating
that LA competes with PA. When we look at the mixture of PA+PG, this is not the
case. This result suggests that PG does not compete with PA and is excluded
from the carbon. It should be noted that at higher total concentration (>500 mM)
the adsorbed amount of PA appears less than that for pure PA, but this is merely
the result of deﬁning total solution concentration as the sum of the PA and PG

concentrations.

105

 

 

 

fruendlich M1
---------- freundlich Mid:

 

 

 

Adsorption (mmollg c

N
o

 

.5
0|

 

 

'o

1

.0
OI

.0
o

    
 

."
.-
-

 

 

100 200 300 400
Salmon concentration (mM)

500

 

 

 

 

 

 

 

 

 

 

 

 

 

A PA I PG
3.5—fnnndIchPA(1) o PA+PG
a 3.0 /
'5 2.5 / 3
E A‘ 0
g 2.0 ~ &
3 1.5
I f
3 1.0
so... _
0.0 ' L' ' . i
0 500 1000 1500 2000
SohltlonconcentnﬂoMmM)

 

 

Figure 7-1 1: Isothermal adsorption of equimolar mixtures of PA+LA (A) and
PA+PG (B) on 3310 carbon at 25 °C.

Adsorption of equimolar mixtures of PA+PG and LA+PG is shown in Figure 7-12.

PG appears not to inhibit adsorption of LA to any signiﬁcant extent, while LA

inhibits the adsorption of PG. The temperature dependence of organic acid

adsorption was only examined for LA and PA and will be discussed in the multi-

component modeling section.

 

 

 

 

 

 

 

 

 

 

 

 

 

    
 

 

 

 

 

 

 

 

 

HA I PG A LA I PG
2.5—truandIich PA(1) o PA+PG 2,5 ———fruend|ch LA(1) o LA+PG
. 2.0 f + 2.0
E, 1.5 E 1.5 /
O
1.0 - i 1.0 .
8 l
P 0.5 0.5 —
2 I . g ' '
I I
0.0 1!! . . o.o . . . .
0 200 400 0 100 200 300 400 500
Solution concentration (mM) Solution communion (mM)

 

 

 

 

 

Figure 7-12: lsothen'nal Adsorption of equimolar mixtures of PA+PG (A) and
LA+PG (B) on 3310 carbon at 25 °C.

106

7.5 Multi component adsorption studies relevant to

hydrogenation of LA and PA to PG and n-prOH

As previously mentioned, a goal of this work is to have a comprehensive
model for predicting pore concentrations of reactive species from single
component data to be used in kinetic models for hydrogenation reactions of LA
and PA to PG and n-prOH. Single compound isothermal and variable
temperature data were collected for each species and variable temperature data
were collected for combinations of LA + PA, LA + n-prOH, LA + PG(limited), PA +
PG and, PA + n-prOH. Data for these experiments will be presented and
discussed together with the modeling of this system.

Initial attempts to use the extended Langmuir model failed to predict the
competitive nature of the adsorption, even though the Langmuir model gave a
reasonable ﬁt for single compound adsorptions. It was thought that large
differences in the maximum Langmuir concentration Cm for PA (2.3mmollg) and
LA (1 .6mmollg) were overestimating the effect of LA, so we adjusted the

extended Langmuir model with “weighting factors” 0A3 and 03A. The adsorption is

then deﬁned by Equation 7-8 and 7-9, where a” = E12 and a3, = J—

 

 

TA aAB
K C C
AS = A A TA (7-8)
1+KACA +aABKBCB
KBCBCTB
(7'9)

BS =1+KACA+aBAKBCB

107

This model is referred as Weighted Extended Langmuir or WEL and is
also applied to PG-GO system, where it showed no signiﬁcant changes
(aAewaaAM) and therefore was not used in that system. This correction did not
signiﬁcantly improve our model’s performance in the LA-PA system.

A closer look at single compound data showed that at low concentrations
(<100 mM), the adsorption of LA, PA, and n—prOH is much stronger (Figure
7-13).

700

5004 wlw” _- '
---o---lPGl

200

o
r '0
0' ‘-°'
O
O '
' " -' +[PA1
U ,.
A
' f
o '0

 

 

 

 

 

 

 

1

 

 

 

 

 

 

 

. —-—rLA1
100‘ . "mu-[won
o I l T I .....--[a
o 200 400 600 300 1000 o 20 40 so so 100
Ce(mmolIL) CarinmoIII-l

 

 

 

 

 

Figure 7-13 : Low concentration adsorption deviation from Langmuir isotherm at
low concentration for LA, PA, PG, and n—prOH on 3310 carbon at 25°C.

We therefore propose that there are two different modes of adsorption: the
ﬁlling of micro-pores, and adsorption on macro-pore surfaces. In the ﬁlling of the
micro-pores, interactions of the substrate with the walls are much more
pronounced compared with planar surface interactions. Each adsorption mode is
assumed to follow the weighted extended Langmuir model, and the adsorption is
simply the sum of both. This adsorption model will be referred to as Dual

Weighted Langmuir (DWL). This approach of using two part adsorption modeling

108

is not new, some of the literaturezz' 72 73 shows models where part of the
adsorption is competitive and some is independent.

The generalized form of the DWL adsorption isotherm for n competing
species is shown in Eq. 7-10. The labeling of constants and concentrations
follows the same approach used earlier in this work, where the ﬁrst subscript
refers to the compound (except numbers are used in place of letters (i.e. A and B

are now 1 and 2), and the last subscript designates the adsorption mode.

C _ leKnlCn + CntZKnZCn (7_10)

n: — nt nt
1+—‘—ZC.~..K..C.- 1+CLZC..2K.2C.

ml i=1 m2 i=1

 

 

_ leKnlCn + CntZKnZCn
"3 1+ KMC" 1+ anCn

 

(741)

For single component adsorption, Equation 7-10 simpliﬁes to Equation 7-
11. The adsorption now depends of four parameters: Cm,Km and Cman. Cm is
limited by the micropore volume (Vmp), and molar volume (V...) of the species
adsorbing cannot exceed Vmme. can is the maximum adsorption on macro-pore
surfaces and using a Langmuir model should theoretically be limited to a
monolayer coverage. This restriction was not applied because of the difﬁculty of
calculating a monolayer coverage and also the possibility that there is multilayer
adsorption.

All single component isothermal adsorption data on 3310 carbon and ROX

carbon were ﬁtted with the Dual Weighted Langmuir model. Optimization was

109

done using Microsoft Excel Solver to minimize the sum of the square of the
residuals. It should be noted that in some cases there was no stable solution and
it was necessary to assign a ﬁxed value to C11: (maximum adsorption second

adsorption mode). Optimized parameters can be found in Table 7-7

Table 7-7: Optimized isothermal adsorption parameters for the DWL model.

 

PA LA n-prOH PG
Km (M1) 1.225+00 1.225-01 1.20501 5.45502
Cm (M) 8.00E-01 8.39E-01 1.18E+00 1.005+00
an (M1) 5.59503 3.44504 2395-03 1025-04
Cnt2 (M) 2.25 4.00 2.08 4.00

 

Extending the DWL model to variable temperature data requires that the

temperature dependence of the equilibrium constants Km and Km be known.

_ﬂi)

K :Koe[ R T (7'12)

Equation 7-12 shows that K(T) depends on two constants, Ko and AHIR (H
for short). For single component data, this means that we are trying to optimize
six parameters (Cm, Kn1o,Hm,Cm2, Km, an ), which is too many degrees of
freedom. Langmuir adsorption isotherms have an implied assumption that C... is
independent of temperature; therefore Cm and Can were found using isothermal
data. The same least square ﬁtting method describe previously was used to ﬁnd
Km, Hm, Km, an, with the additional constraint that the calculated K‘s at 298 °K
must equal those found in the isothennal ﬁtting. (Table 7-7) The optimized values
are showing in Table 7-8. Generally, the optimization converged on a single

110

solution, except for n-propanol. With n-propanol, the initial guess for these
parameters did have some effect on the ﬁnal optimized values found using Excel

Solver.

Table 7-8: Optimized isothermal adsorption parameters for the DWL model.

 

 

 

 

PA LA n-prOl-l PG
Kn10(M-1) 2.46E-05 2335-05 1255-04 1.18E-05
1572111 (K) -3.225+03 -2.55E+03 -2.075+03 -2.525+03
ano (MT) 1.91E-05 8.87E-06 8885-05 1345-08
Agnz (K) -1.895+03 -1.095+03 -1.125+03 -1.435+03

 

 

 

 

 

 

A simple program was written that would randomly change one of the initial
values for Kn1o,Hn1,Kn20, an by a random amount not exceeding 150%. The
system was then solved using Microsoft Solver and the ﬁnal parameters saved.
These ﬁnal parameters were used as the input for the next cycle. The program
was set to run overnight but stopped after 13944 cycles when there was a
combination that did not converge. The value of the objective function (sum of
the square of residuals) was examined and it was found that 82% of all solutions
were within 0.5% of the lowest value found. The range of solutions that were
within 0.5% of the optimum value are listed in Table 7-9 and varied by more than
t7%. These results emphasize the importance of checking the stability of the
solutions found using numerical optimization and making sure a global minimum

has been found.

111

Table 7-9: Percentage range of optimized parameters whose objective function

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

    
  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

is within 0.5% of the “best ﬁt”.
[ H.- l muss-l ”g L '"(ngl m l
7.79% 5.90% -3.74% -1 .48% 0.49%
-7.51% -5.69% 7.04% 2.79% 0.50%
0.00% 0.00% 0.00% 0.00% 0.00%
6.63% -5.02% 7.09% 2.81% 0.44%
7.28% 5.51% -7.25% -2.88% 0.49%
3.50 3.00
O 3'00 e O 2.50 /
9 .4/—_— 9
'5 2.50 6 .0. 2.00 d
E ’ E
E 2.00 - E
: 2' 1.50 ~
g 1.50 - €-
1.00
18:! 1.00 i 3 A n-prOH
< 0.50 0 0 PA 412 0-50
—dual LM n-prOl-l
0.00 0 . _. “aim PA 0.00 . . .
0 500 1000 1500 2&0 0 500 1000 1500
Concentration (mmollL) Concentration (mmollL)
2.50
8 6
a a 2.00 4—
E E e
g E, 1.50
c r:
° ° 100 -
a u
s E
0.50
3 I LA 3 . PG
__ dual LM PG
“‘3' LM LA 0.00 . . 5
0 500 1000 0 500 1000 1500 2000
Concentration (mmollL) Concentration (mmollL)

 

 

 

 

 

 

Figure 7-14: Experimental data and WDL ﬁt for the isothermal adsorption of PA,
LA, n-prOH and, PG on 3310 carbon at 25 °C

An Excel sheet that expands Equation 7-10 and 7-12 to model the
competitive adsorption of up to 4 compounds has also been developed.
lsotherrnal adsorption data for LA, PA, PG and, n-prOH on 3310 carbon at 25 °C

112

is shown in Figure 7-14 and includes the DWL ﬁt The model accurately predicts

the experimental data.

 

 

0 20 40 60 80 100 1 20 140 1 8
Tempenhue (°C)

 

 

 

Figure 7-15: Variable temperature adsorption of 1000 mM LA on 3310 carbon.

 

   

 

 

 

 

 

A 3.5 .4 3.5
o o
:3 3.0 i 3.0
~$2.5 E 2.5
722.0 g 2.0
3 1.5 3 1.5
$1.0 g 1_0
3 0.5 3 0,5 A
n 0.0 I 0.0 —X-y
0.0 1.0 2.0 3.0 4.0 o_o 1_o 2,0 3.0 4.0
experlmental adsorption (mmollg C) experimental adsorption (mmollg C)
8- 3.5
=3 3.0
E
g 2.5

£20
3 1'3
g0.5
1goo

 

0.0 1.0 2.0 3.0 4.0 0.0 0.5 1.0 1.5 2.0
experlmental adsorption (mmovg C) experimental adsorptlon (mmollg C)

 

 

 

 

 

Figure 7-16: Experimental vs predicted single component variable temperature
adsorption PA, LA, n-prOH and, PG on 3310 carbon from 25-160°C

113

The variable temperature data were obtained as described above.
Adsorption data for a typical single variable temperature experiment are shown in
Figure 7-15. The parity plots are shown in Figure 7-16 for single compound
system; the DWL reasonably predicts the adsorption on 3310 carbon over a wide
temperature range. Figure 7-17 shows the predicted isotherms for LA+PA at a
total initial concentration of 0.5 M and temperatures of 25 °C, 80 °C, and 150 °C.
The predicted values again show good agreement with experimental data, and
validate the assumption of constant heat of adsorption over the range of

concentrations and temperatures investigated. PA is preferentially adsorbed on

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3310 carbon.
COMMItZS‘C COMMItOO‘C CWIHGO'C
250 250 250
_ > e PA —PA" 6 6
0
g 200 A LA ------- LA 9 200 o 200
5 150 . A S 150 E 1.50
‘5' \ ,. ‘5' 5
g 1 00 5: i 100 i 1.00
0'
050 .9 "\ g 0.50 3 050
3 . ..... e 3
0.00 m """" . 0.00 ' 000
0.00 0.50 1.00 0.00 0.50 1.00 0.00 0.50 1.00
Fraclon LA Fruclen LA Fraction LA

 

 

 

 

 

 

Figure 7-17: Experimental and predicted DWL isotherms for mixture of LA and
PA on 3310 carbon at 0.5 M total species concentration and 25 °C, 80
°C, and 150 °C. Abscissa is fraction of LA in the total species (LA + PA)
present. Lines are the predicted isotherms.

Figure 7-18 shows the predicted adsorption isotherms for an equimolar
mixture of LA+PG at total initial concentrations from 10 to 500 mM and
temperatures of 25 °C to 150 °C. It can be seen that the DWL model overpredicts
the adsorption of PG relative to LA If we compare the low concentration ﬁt of

DWL for LA and PG, it can be seen in Figure 7-19 that LA is underpredicted

114

while PG is slightly overpredicted. This imperfect ﬁtting may be due to the

difﬁculties of obtaining reliable adsorption data at very low concentrations.

Gathering more data to improve the model is possible, but likely not worth it

because experimentally it was found that PG generally does not inhibit the

hydrogenation of LA to a signiﬁcant extent, so it would be simpler to leave out PG

and any non-adsorbed species in the ﬁnal model.

 

Adsorption (mmol/g 0)
.° .° .0 .O .0 :-*
8 8 8 8 8

8

 

1.40

E?

 

 

 

 

 

 

 

 

 

 

 

0.0

0

100.0 200.0 300.0 400.0 500.0 600.0
Concentration (mmollL)

 

 

Figure 7-18: Adsorption isotherm for equimolar mixture of LA and PG on 3310
carbon at 25 °C

115

 

CI
0 .o
8 8

..9.
8888

p
_I
O

adsorbed pmollg
O O O

l
I

 

p
8

0.0 200.0 400.0 630.0 Kilo 1000.0 0.0 10.0 20.0 30.0 40.0 500
Solution concentration (mM) Solution concentration (mM)

 

 

 

 

 

Figure 7-19: Adsorption isotherm for LA and PG on 3310 carbon at 25 °C
Figure 7-20 shows the predicted isotherms for LA+n—prOH at a total initial

concentration of 0.5 M and temperatures of 30 °C, 80 °C, and 150 °C. The
predicted values for LA show good agreement with experimental data while the
n-prOH is underpredicted at low temperatures. This deviation is probably due to

the difﬁculties encountered in ﬁtting the n-prOH data as described later in this

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

chapter.
comm-1m Combhetlonetco‘c Combhetlonettdo'c
2.5 - O M u. 4) 2.5 o u w 2.5 . LA LA.
8 <| ‘ W ....... J 8 . ....... l 6 ‘ W ....... W
€20 . 2.0 ""'°" ""0" 2.0
. .
£15 , ‘ ‘_.. ' E15 ‘ E15
.r' "A
£10 ¥ .+" 1.00 ‘... " E10 \ ﬂ .. ‘
A . 1» -' '
80.5 \ 0.5 \ 0.5 . "°
2 2 ‘
0.0 ' 0.0 ' . 0.0 '
00 05 10 0.0 0.5 10 00 05 1o
Blane-plat Freshen-pron Fauna-pron

 

 

 

 

 

 

 

Figure 7-20: Experimental and predicted DWL isotherms for mixture of LA and n-
prOH on 3310 carbon at 0.5 M total species concentration and 30 °C,
80 °C, and 150 °C. Abscissa is fraction of n-prOH in the total species
(LA + n-prOH) present. Lines are the predicted isotherms.

Figure 7-21 shows the predicted isotherms for PA+n-prOH at a total initial

concentration of 0.5 M and temperatures of 30 °C, 80 °C, and 150 °C. The
1 16

predicted values for PA show good agreement with experimental data while the
n-prOH is underpredicted at low temperatures. This difﬁculty of n-prOH
adsorption is similar to what was seen in LA + n-prOH and will be discussed later
in this chapter. Experimental data show that PA and n-prOH compete for the
same sites and adsorb in very similar quantities. Therefore, n-prOH should show

inhibition on the reaction rate of hydrogenation of PA.

 

Mdﬂ'c comm-woe W11“?
1) e T! ——-m' A i e F! -——Fl' —
‘ A n-prOt ------- mum—ll

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

‘0
I
I
.-
l

Adeerptlenonmellgq
3
r 1
/
WMWC)
. . 3 .
4
./
e
l
WMMC)
O

 

 

 

 

 

 

e I
I c
I I
..o' _p'
A.

 

 

 

1 00 t (A 00 ' .
0.00 0.5) 100 0m 05) 1. 0.00 0.50 1.00

 

 

 

 

 

 

 

Figure 7-21: Experimental and predicted DWL isotherms for mixture of PA and
n—prOH on 3310 carbon at 0.5 M total species concentration and 30
°C, 80 °C, and 150 °C. Abscissa is fraction of n-prOH in the total
species (PA + n-prOH) present. Lines are the predicted isotherms.

Figure 7-22 shows the predicted isotherms for PA+PG at a total initial
concentration of 0.5 M and temperatures of 30 °C, 80 °C, and 150 °C. PA greatly
inhibits the adsorption of PG, and therefore PA is adsorbs much more strongly
than PG. The predicted values for PA show good agreement with experimental
values except at high temperatures. This deviation is not due to competition from
PG but due to a bias in the DWL model, which tends to overpredict adsorption at
high temperatures while underpredicting the adsorption at low temperatures.
Figure 7-23 clearly shows this bias. Attempts were made to eliminate this

observed bias by adjusting the parameters but these attempts resulted in a

117

signiﬁcantly worse ﬁt. Lacking a logical basis to eliminate data points that

resulted in this bias, the best ﬁt was used and is reported.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

W‘ﬂrc mast)? mauve
-250 1» e PA —n\ 25° e m —m 1m
0 2m \‘ ‘ PG ------- EU" 6 2m ‘ R3 ....... R3. 6 1.0

0 1m
2150 A €1.50 \ E100
" 080

s \\ .- = \ s
100 \Q' 1.00 v\. 3' gm
050 ‘5 ..' 0.50 g 04)
.4.. 5 ,,,,, 4 ..... A ..... V 0.20
000 - 0.00 ------- 4 0.00

 

 

 

 

 

 

000 0.50 1.00 000 0.50 1.00 000

 

 

 

 

 

 

 

 

Figure 7-22: Experimental and predicted DWL isotherms for mixture of PA and
PG on 3310 carbon at 0.5 M total species concentration and 30 °C, 80
°C, and 150 °C. Abscissa is fraction of PG in the total species (PA +
PG) present. Lines are the predicted isotherms.

 

 

140%
Z 130%

12011.
8
I 110%
h
; 1m
g 90%
2 8091.
e.
a! 70%

m

0 20 40 60 80 100 120 140 160
Temper‘ure (‘6)

 

 

 

Figure 7-23: DWL predicted adsorptions of PA as percentage of actual adsorbed
on 3310 mrbon at initial concentration from 10 mM to 2000 mM.

The DWL model’s performance in dual compound systems is best
summarized by looking at the parity plots in Figure 7-24. LA and PA show
linearity to X = Y while n-prOH is generally underpredicted. PG shows a
signiﬁcant amount of scatter. The cluster of points between 0.6 and 1.0 mmol/9 C
are from one experiment with a 4:1 PG to PA ratio.

1 18

 

 

23.0 —— ------- --
U
5 2.5
5
~— 20
i
a 1.5
’6
i 1.0
.0 0.5 . PA
8 _FY
3 0.0
0.0 1.0 2.0 3.0

experinental adsorption (mmollg C)

 

 

 

0.0 0.5 1.0 1.5 2.
experimental adsorption (m mollg C)

 

 

 

I LA

 

0.0 0.5 1.0
experlmertal adsorption (mmovg C)

 

 

 

    

0.9
0.8

0.6

predlcted adeorptlon (mmollg C)

4
3
2 ePG
1
0

 

0.0 0.5 1.11
experimental Idsorptlon (mmollg C)

4—1

 

Figure 7-24: Experimental vs predicted adsorption of PA, LA, n—prOH and, PG
on 3310 wrbon in binary mixtures only.

7.6 Additional notes on adsorption modeling

n-Propanol data was inconsistent with model. The question arises whether

n-prOH behaves differently or the experimental data were ﬂawed some how.

Control experiments were run to validate the variable temperature experimental

design. Experiments were performed as described earlier, except that no carbon

was added. Even though no carbon was added, a carbon loading of 1 gram was

119

assumed to maintain consistency in the calculations. Figure 7-25 shows single
component measured adsorption of 100mM n-prOH, LA and, PA without any

carbon over a temperature range 25 - 160 °C.

 

Adsorption (mmol/g C)

 

 

 

 

Figure 7-25: Apparent species adsorption in control experiments without carbon
showing the need to correct for vapor fraction of n—prOH. (note data for
150 °C after 14 hours is plotted at 160°C for clarity.)

PA and LA show no signiﬁcant deviation from zero as expected. For n-prOH
there is a signiﬁcant deviation at high temperatures due to high vapor pressure of
n-prOH. To correct for the fraction of n-propanol in the vapor phase, partial
pressure (yP), vapor volume and temperature were used to calculate the quantity
(mmols) of n—propanol in the vapor phase. Partial pressure was calculated using

modiﬁed Raoult’s law( yP = xyP”t ) and the following assumptions:

1. Gas phase acts as ideal gas.

120

2. PS"t for n-prOH can be described by Antoine’s Equation

3. Solution density is ~1 glml

4. y (gamma) is a linear function of x and T with the form y = DT + Ex + F
Antoine coefﬁcients (A=7.77374, B=1518.16, C =213.076) where taken from
Yaws' Handbook of Antoine Coefﬁcients“. Aspen properties was used to
calculate gamma’s over range 05x5.02, 25$Ts160. Least square ﬁtting was used
to ﬁnd D, E and F (D=0.009439, E=-196.09, F=16.955). This correction was

applied to all n-prOH data at temperatures greater then 80 °C.

121

8 CONCLUSIONS AND RECOMMENDATIONS

8.1 Conclusions

A study of substrate-catalyst interactions has been completed; both metal

and carbon support adsorb reacting species and inﬂuence apparent reactivity.

Methods and a recirculating microreactor apparatus for measuring

adsorption on metal surfaces from the aqueous phase have been developed.

The key features are a large metal to solution volume ratio, automated sampling,

an operating range of 25-200 °C, and headspace gas control.

Adsorption of organic acids and polyols on Ru sponge metal from water

was characterized.

The extent of lactic acid is ~4 pmol/g Ru and is independent of
hydrogen saturation of the metal surface.

C3 polyols with terminal hydroxyl groups adsorb irreversibly through a
dehydrogenation step to an “aldehyde like” species followed by
degradation. This adsorption is inhibited by presaturating the metal
surface with hydrogen.

Adsorption of isopropanol shows little adsorption on either a Hz-free or
Hz-saturated Ru surface. On the Hz-free surface formation of acetone,
a dehydrogenation product, is seen.

Primary alcohols, when adsorbed on a Hz-free Ru surface, show the
formation of acids, an oxidation product.

122

Single component and multicomponent competitive adsorption on activated

carbon at a range of concentrations and temperatures was characterized.

For glycerol and propylene glycol, adsorption of individual species is
well-represented by the Langmuir isotherm, and simultaneous
adsorption of the two species is described by an extended Langmuir
model that uses the single-component adsorption constants and
accounts for competitive adsorption between the two species.
Propanoic acid (PA) and lactic acid (LA) adsorption is competitive and
PA adsorbs three times more strongly than LA.

A two part adsorption model consisting of adsorption into the micro
pore and adsorption on the surface of the macropores was developed.
The model accurately predicts the individual species adsorption and
when the individual adsorption parameters are used, competitive
adsorption was accurately predicted with the noted exception of n-
propanol.

Incorporation of the competitive adsorption model into our kinetic
model for the hydrogenation of LA and PA was done by Chen“. The
combined model performed reasonable well when using reaction rate
constants found on unsupported Ru catalyst and the pore

concentrations predicted based on competitive adsorption on carbon.

H-D exchange on H20-D20-H2—Dz over metal catalyst was characterized. The

dissociation of water is the limiting step on 5% RulC but not on 5% PVC. The

experimental data can accurately ﬁt the kinetic model.

123

8.2 Recommendations

There are several areas that warrant further investigation to elucidate the
mechanisms of adsorption that were observed in this work.
0 Simpler systems with substrates containing no carbon atoms or only
one carbon atom should be looked at in more detail.

a. A thorough study of the interactions, in liquid phase, between
water and the metal surface focusing on the actual oxidation
state for the metal.

b. Characterization of interaction between Ru surfaces and 01
species (ie. methane, methanol, formaldehyde, formic acid, and
carbonate). This should include hydrogen saturation affects and
temperature dependence.

0 The reversibility of lactic acid adsorption on Ru sponge metal should
be examined.

. Adsorption behaviors on different metal surfaces should be
investigated. One option is adsorption on Raney nickel, which is a

known active hydrogenation catalyst.

124

APPENDICES

125

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

E888
8.28 cc 5:5 52.988 c: 388.8 8 mv or on - . 2 83 8? $820 8328 a
8:88 8838 888$ 5.888
2 8258 882a 95 8588 use? 2: fa: 8 f be 5m 2 88.. 832 8828 8:83 m
888$ E858
8988 >8 222 mv 2.. mm «2888 ed 3. 2 83 82.3 3820 88:3 m
:52 2 858
.2358 2.23 888838 88:3 98 Bow
2:833 .8862 .o 50:88 85. 8:88
060m 28 use 826 Eur 8:58 Boa
2:820 .cozaEEano 8 2 8888 3.8 888$ 8858
29:. .88; 9a 88288 no; 38: of. mv 2.. mm 28:2» ed :1 2 83 8? 8820 83:3 N.
28 08am 8 8:58 .2388 888$ E888
new new 8.58:3 8288 SEES. 7. mv 8: mm 28:28 he :1 2 83 2: 8820 8823 a
888$ .5288
83.88 SEER. T. mv 81 mm 2888 No :1 2 83 8.. .898 8:98 a
888$ 5288
04¢: 53 8.888 =8 28.56 am .2558 88:8
3.331 new 58:33 :8. «I0 3.. .88 43 .u 83 83. 8E 8:833. 83 8xm
n. ...
«502050 8238 5:22.350

.8: 282.898 .8885. u< 52.8%?

 

126

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

888$ E088
mv o... .8 222.8 a 3.. 2 83 2:2 _2.28 8
888$ E088
mv 0: 8 28:2» o 3... 2 83 2:2 5.5.8 Nu.
888$ E088
8.8.800 0E8 .508 2. 800 8.5., mv 0... mu 20088 a :1 2 83 quw 8.88m ad
vow”...
_2Em
8802.
885E888 0E8 mEoﬁo8 8.. 3...... :05 888$ E088
93 8 8:88 swE 88.. E808 v.8 mv 0: mm 2088 o ..m 2 83 8.3 88.8 8: :8 0:...
828.8888?! F _oEsoF «own...
880 8; 8:88 8E 0E: .05: 22.9.8 .2Em
8.888 8808 08 8 8.882 05 00803
:. $.82 8...; 5w"... s.Ev :. 828.8 m. :05 888$ E088
Om 8.5. 8.... E0Et8x0 :8 EaE 8. 8...... mv 0... mm 2888 a :m 2 83 8.9 8.8.8 8388 «a
8.80. E088 381 5.; .8088 ..0. 8838 ”A
goEzn. T 2 8888 08E
_oEzoP U080 :05 00288 goEsodz 888$ E088
889. 3 8.. £88 .0. .28 .oanF 0028 mv 0: mu 20808 ho am 2 83 8.3 8.88m 82.88 «H
E088
8228 o: 5.; 8.8.888 008098 8 mv 0... mu . . 2 83 2: 8.8.8 88:8 .3.

 

 

 

8.28. a. 538.8

127

 

22 8:82 E228 a 3 832.2 95.9.

 

 

 

 

 

 

 

 

 

 

 

 

  

 

ho 5:988 a on 2 2:08 22:. .o:_::2 $9.82 956 9:3"
89. 2 E88 0n. :5 0m :5 8::2 m. 00 on 2.. cm. 22:58 or am. aEE -953 .88an 3822: 5
8:988
8268 .o. a 8268 £29m m2: 0.. on"... :< $382 95.6
825 08:8 8.228 N... a :o 5.5.82 cm N... om .28 NT. or am. .8820 vo\wm\volq|m
828.2. 95.0 .22 Bow
3. 9.22662 .8888 9:06 ow. 0.. mm E2628 or 9:2 uEvz 2:820 33:8 N
uEmd .22 REES
om. 9.083.? ow a: mm 22.8% o: 2:2 uEv... .2820 voatmo «M
3. 9.2:: 7 2:88 953 .22
2::oEm .38 :2 2.88 :8 28 2:83 cm 2.. mm 22:28 or 9:2 9:? 883.0 3: :8 mm
8:988
o: .8268 9N<E¢Nz g2. v.38 :22) 288 aENd .22
25 um 3% 22. 8:228 95: m. 8.5 om 2.. mm :8; me $20 9:2 men: .228 3828'“
22> mason aEmd .22
8:988 o: :8 5:82 o: ow 0... mm 82> m2 850 2:2 asz .258 3398
392.2, 95.0 .22
828:: .22 9:7. on 2.. mm 22258 or 3”: 2:2 aENz

 

 

 

 

 

 

 

 

 

 

 

 

6:80.

  

 

n< 5.2.8.3.,

 

128

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

898:2 E288
« on .2. mm 23:8 2 3m 283 95 00 E
895an E296
8988 .0635... on :< mm E2629 or :m 2 goo. 25.6 00 No.
8282 E296
aEvdz 3888 ..< on :< mm 22829 or am. 2 :84 aEvd 00 um
$92.2 E23». J
8988 00 9:3: 23% on 2 SN 22:59 2 a: 233 8; om. H
0. on
«a 82s
828: on EQE .223 8382 E28?
5230: :5 23m 20 2a 20:23.8 om :< pm. 22:29 or am. 2 80.. as: 00 ﬂ
qla
mm
00 on": 2.. 8908.. anm
325 82.8 3.233 f a 8 8.5.83 o... 8 .23 f 2 :m 2.98.3862 %
.w>o. 5:288
.820 o: 22 5:02.: «a 9.32: m. On.
:5 82 ___.m 2 258 0m 23 8.52 8. oo
98".. 2 89:2 8.a 225 on. :8 89.83 aENd aEmN
h. .8552 8:228:00 5.; 8383, om 2.. mm. 82% 2 HT :62 -953 .2820 36:8. N»

 

 

 

 

 

 

.828. "< 5232,

129

 

 

 

 

 

 

 

.8886 :2: so. on.
m. ow .0 232a 8288 2.. .. 8288 9:3
on. 2. 09.2.0 2. ... 32.3mm. 0n. 3 5:63 $9.03. Sci» 00 on.
.25? 8288 23 22.. 8.6a 8; 00 cm 2 mm 2222» 2 2... 2 83 25... =2: 00 um
00
.223... 39.83 2. new 00 $9.82 82m? 953 00
825 52 323-9? 0. 3283 2.. cm 2 mm 28:2... 2 3. 2 83 9. can 22.. on. «2
Nn. new 01.0 5.5.2.2 v.92 m2: 22..
2:28 .2. 2.? .8 8. mites“...
a 25 f Ea 08 .em 5.; 832 .22 €va to 2.. of. .2 2 06 f mam
an. .0 5.38.2
2.. .6 832.201... 8.62.2 3; 22..
Eu 2.... .228 .2, am? .8 2: 222282
a 8° N... 2.... own. _EN 5.; .232 .22 2.3 to 2... 0.3. _E 2 Own. f a.
o... 3 Nn. .o 22.2.22 2. 225 22.. .8 cop 922282
a 23 N... 25 own. .2... 2.; .282 .22 2va 2.2. .2 2 06 f «um
00
25....
.oEamN... 89.83 2. new $9.82 Ema? on. 00
00 88a 22.. .25? on. 8288 22.. cm 2 mm 222% 9 3. 2 83 new... 22.. 0“. «v.
.0839 2 2.25 25 oo 2 23.28.22 89.82 Emu?
8.322.. .2.» 25 8288 an. .853... om .< 8.3 222% 2 am 2 0.8.. 92... on. 3

 

 

 

 

 

 

 

 

 

 

 

 

 

6.2.8. 2 2.2.8.3

 

130

 

 

 

 

 

 

6:522... oEm... 22. 2.. ... 8.2. 89.82 E226
2.2.8 a 8; 2.2... .3288 on. .22.. T 2. N... om 28 a... 2 2. 2 88.. 2:3 0.. 3.8.: H
62.30.. ES» 5 comm wﬁtoﬁao. some
.88. 2... 0.3. 2.3 2.; 06 .26 Na + N... .58 :2. ..o 0.322» .... 2. .88 N0.... 3.8.2 2.
00
25...
.2506 ...5 22; 5.5.85 0n. 2...; 35% E06? on.
...a 82.2 .22. o. T .28. 8288 00 cm .< mm .2822» 2 2.. 2 88.. 2:8. 8+0... %
c205. c. am 025
0.05 02.005 .2 cc 98 Ea... .552... E08
.2. 88 .252. 3 .2 0. 8m .... 8262
22; E222 005% am 9 05 Nn. 05 N.... .0 09.0% 92.585. 06
58 .8 on 23 06 ....2 2.; .208. .22 .53 am .2 2. Nn. + N... 2.
02.22 225
an. 2. ﬁoEE Sn 0.... 5 55:22 22... 2.2..
.5 2.9.2.28 .25 2.? .8 22 s 222208. 2.2.500
8.. N... 2.... 322.300 ....» 2.2.82 ....2 .58. 2.. 0.2. .... 2 f 5.
5.5.8 E2.
0 .0 5.8528... 0.. on 2 mEoom 22..
gm 2.. 22.. N0 2... o... 2 8.2228
32m a 2. 2 «Emma 0.05 .No 05 ...... 5 22.1205. 06
..28 .8 on 2.... 03 .22 2.; .282 .22 2.2 2.2. .... 2. Nat... %

 

 

 

 

 

 

 

 

 

 

 

 

 

.228. .< 22.2.2.

 

131

 

 

 

  

 

 

  

 

 

 

   

 

 

 

 

 

 

 

 

 

    

 

 

  

 

 

 

800000. E2200 0.0...
00.0 00.... I0 00.00.... 000.0000 5 0E3. 3 .< mu 0.00020 0. ..m 2 000... .0E000 0.80... 00.8.00
00800.8 00... .000... 0. 0.... 0.0... 800000. E20»... 0.0..
00.... 000 .0 I0 0.... 000.0000 «1. _oEzmT. n. .< mu 0000008 or :m 0. 000.. .0503 0.80.. moNoRo
000030 00.. N... 0 0...... 00... .0320 00E000
0.... .2. S .o .0050. T .o 02.282. 00080., 0.0.2.0 0.2.
.08. ... 0.2. 22...... .o 8.20.2 20%. 2 m... 00 ..00 N... 2 an. 2 08.. 00.0.0 0.00.. 00.5.3
.005... .020 0020.20 f
0... .05 000.0 2.0. 000 .0... 00.8008 0E00
00 2 0E000 0.0.. .r ...0 2 0000 ._ >2. .2.
0.0 .... .0000 000.. 0>0.. 0.00.... ...0E000x0
0.0.. 00.5.00 .00... 00.000000... 00.0 0.0...
.0... 0.00 0.203 2 02.0>000 0.00 0.80.. o ..< mm 200 000.. aEmd 0.80.. 00.3.5 |mmn
000.0000 «1. .0 _0E0N. ....
000 0.00 0.250 .0 00000.>0 2. 00... 0.0...
0.... 0. 00.... 0E=_00 00...... ...... 000. 0... 0. 0.00 800000. E228 0.0.4.
0.20.... 2 52.8 2 020000 0.8 2.8.. 00 ... 0L0 0.02.20 2 2 08.. 00.0.0 2.00.. 8.8.3 «I...
E08»... 0... 0. 02.0.0080: 2 0:0 00 >0E
0.... .00 02.0.0000 .000... 0 020 0.0 0m .0 800000. E290
E00 02.00830 02.880 00 .2030. on .< 00 0.02.20 2 =0 2 08.. 2.5.0 00 00.00.... 2
300000. E220
009.0000 0m .2000... on .< mm 0.00008 9. 0m 0. 000.. 900.... 0m QUE—.5 Mm

 

 

50.2.00. .< ...—0:022

132

 

 

 

 

 

 

 

    

 

 

 

 

  

 

 

 

 

 

 

 

 

 

35.25 2 o_aE_m 8: 9.3g 23.5

:ano EEomr 05 :o 28 282 he woman $9.83 204.

5.; Eoetwaxo coach 26: a ma; $2... an? 263 moNoBot via
.29: E2. mxmma :08 was of. $5
96m acoE_w 628m >3 m_ 35 5E 9mm 8
32a 8.x xwoa < Ema? m5 2 E 9:58.
22:2 E2228 “6 E285 05 02a Boa Ia

m_ 805. .830: m E: 05 E 3 2 a: E0; 8225 + E298 22

In 2: So .58 co:&omum ESE—.9.» oz on .< mm 235% or 2 83 _oEzm 0:03
on 96 E:

_oEsom 3258 05 :05 E222“. ma; mEF E296 Eo<

.28 .5267. a 5 So 3983 <._ _oEsoN... on .< mm 2355 or 2 33 _oanm 283

22

oE: m>o 3:an E8537. c: on .< mm 2355, or 28 goo. _oEsom 283

E893 Eo<

I”: mm; In 2t .8288 <._ BEEN... 9 f mm 43. f 2 2 83 BESS 263

.on. 53% BEEN 0n. +

ho BE: 5 EB 8988 .8329: ho _oEam on .2 mm 22.5% or 2 83 _oEzoN 6533.: 38:3

«55598

25 .95 «ton _ S _oEzvve o>ao E293 Eo<

uca mEoBoa E8 .8; EoEtwaxo £5. 2. .< mu 23:99 9. 9. goo; _oEsooP 283 3323

 

 

 

 

 

.6.E8v u< 5.88.2

 

133

 

 

 

 

 

 

 

A8823. 888$ Ewﬁﬁ
82. 52.5 ..o 3.8. 2% Base... .2. cm 2 mm 285% S a... 2 83 on 00 8.38 3
2a.. 0E8 o5 Sena 898%. E06? (new...
3 <4 .oano new 8988 (n. _oanNz m? .2 mm 9.353 or am 2 goo. Sour (n+5 3:28 3.
£38.. 320
can 8: 23 can 056 c. :m coup—mam... 5.832 ho 69.8w.
2? :23 63 a c. 5.5.83 on 2 mEEoE «2 mm Cu .md... am 33 _oeau... 00 3.3.3 ma
@3522 8: on 2:8 5.632
«oz. :05 .6267. >3 .oca 95E o 333 5.8.62 693%. Emu?
0E: ...oao 39.83 m.oE:m $5 265 3 .< mm to or am 2 goo. _oEsoim 00 38:8 ad
m>wu
.298
:3;
883
$3 .33 20E m 338 .55 895%. E2m>m
0E: some 3983 m_oE:m «a... 26% o. ..< mm 235% or :m 0.. goo. _oE:m+m 00 3.2.3 H
2.3 5.55598 .0 $.38 .620 o: 895%. «00 29...
=25 to: .2. Eu 2238 E2 5 mﬁEoﬁ am 0.83 33.93 ad
Boa .o 5.5.33
326:... E... ..ano noon mama oEom 39.0%.
:8 83 So: coco 0: 3 96 o..a._EE_m :m «2.2. 3328 mm

 

 

 

 

 

 

 

 

 

 

 

 

 

.628. 2 528.3

 

134

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

g £83 8: Sn 2:2..8 3.88.60
a 2.2: 2988 2:8 9:955 22> 80:83 232?.
2:: 5.522 on so: 8:62... um: 0.5.... on .< oo 23:2"... on :m 2 goo; on 00 83:3 mm
on 2 .23: no; >3 as: $5 $9.83 2226
3% =5... .26 was» 253 2:28.32 8 2 nu 2.856 cm 3“. 2 33 8 an. 8523 a
80:82 2826
.023...” >8. 5:283 on 2 on 235% cu :m 2 33 on on. 8823 Nm
f 5.3
95am; :8: 528 __a .6 2382 «26%. 69.82 222m
83 . 23 man 9.053... 8:283 00 om .< on 2356 on 3. 2 30.. on 00 025:3 a
o. on E 3233 320
__a 02982 A«I 9.62. 0320:. 32 o: o9 89.82 229m
2 9.22.5 :23 9.0636 3283 00 om 2 cc 2356 on :m 2 goo; on 00 82.23 a
.323 __o 0. com 328$ 222$
“a 2382 2:2.» on? 9823.0 8283 cu .< on. 2856 om am 2 33 on 00 83:3 wlw
.0633... 2 E25 3:983 0.
oo 2 33205 ... :2? .0237. 0. mm 5 828$ Eovém
8988 00 .22 $3... 8m 5.; 5: Em. ca 2 8.3 2855 om am 2 83 o... 00 SEE Ma
3283 00 _oEanN 2.5 :83 €9.82 22%.»
:2 «ma. 20cm .23, t 5.3 3:83 2:0 on .< mu .2.; or am 2 33 on 00 8628 um
8:272: 820 on new 69.82 222$
3883 Sun 58 3933 00 3:59.. on 2.4: mm 2353 or am 2 goo; on 00 09330 a

 

 

.6203 2 52.8.2

 

135

 

0.

 

 

 

 

 

 

 

 

oo 3 8 595 $982.83 En 825.228 898%. 82m?
28 8o... moon £9: .o>o 8:88 359:8 on .< SN «2853 om :m o. 83 on 00 8: :3 add
.88: 9.8. 88. 8.3. 895%. Emu...»
Eoﬁxm «Soot .2: 2: .8. Eco 022. zoo om 2 8 28:2.» om 3. o. ooo.. cm. 00 8.8.3 M4
. 39.5%. E296
=23: .8 v.59. u .23 8.3.... on .... mm 285% on :m 2 ooo... on 00 8.8.8 8H
o.o 8 2:23 895%.
8.883 8:88.. =2 8.. 5.6.3.8 o: ow ..< .8 «285% cu :m on 00 8.8.8 add
895%. 8293
...:o 5.88.. on om. o. ooo.. on 8.8.8 3|.
8:8 398%. 8296
8.5.88 00 .88: 5.3 5296 95.85 om .... .8 2853 on :m o. ooo.. on 00 8.3.8 mg
5.58 :. no.8. mo; 8:on 35
5896 8:88: .2.; 23 .o: cow E2. 688...... Eoﬁxm
ooEomoo 6..oE:n.o-~.ovm_oE= We 8:65 on E .8 285% on :m 2 ooo.. on 00 8NF.8 mad
E285 $9.8m. 628m
9353...... .25.... 28 8:938 on 2 .8 235% om am 2 83 on 00 8.3.8 3|.
3m 8a :0
... ”.5 205 00 96539653 8989.
cops 5.3 :83 .88, .3 3.52.8 A«... 8.238 898%. E06?
co 8.88 B 832.2 8.268 .3 8.88.5 om .< .8 f on :m o. 83 on 00 8.8.8 add

 

 

 

 

 

 

 

 

 

 

 

 

 

.628. 3 5.2.3.3.

 

136

 

 

 

 

 

 

 

8.5.88 965358.?
83 82.. new (a. o. 852.8 8208 2: A858. 5288
5.5.8 5 5....> :5 ooo. 5 .28 Co. 5.. g 8 .< .8 2858 8 am 2 goo. on 888,. 3.8.8 a.
28 c. 8858 58
a 5.3 N.... .o 53.8 on. 5 858 5.5.88
21.98.. .2.. .38 85.82% .3
8E 83 055 28 ozooo 2:8 .o 5:52 .858. E298
2.. 5.3 8988 2o; a..oE:...o... 8 .< .8 28:8 8 :m o. ooo. on .893 8.2.8 «Hm
5: o... .o 25.8
2: c. .8. o 83 22:. .9555 8.88
.o $98... 88:85 8.8 .288 82.. A858. .588
82.. 2o: 8988 Coma; .o 9553. F... 8 ..< .8 2858 8 am. o. 53 on can; 8.3.8 3|.
8288 00
2... o: N... 8.28 59. 8:5. 83 8208 A858. .58...." 5...
0&8 .<n._ o: .3 8288 00 .o 9.25.... 8 .< .8 2858 8 3m. o. 53 58 on o5 00 8.8.8 a
8.238 .858. 8088
8988 96:586.. <n= 8 N... .8 N... ow :m o. 53 on <n= 8.8.8 3
829:8 S 55 8.5. .588 955.86... .858. E288
5.5.88 8208 o. 852.8 (a. 8 .< .8 2858 on om. o. goo. on (a. 8.8.3 3
8288 .858. 808..»
9.023.? 2.... 828.. 828. <n._ on .< .8 28:8 om am 2 83 om <n__ 8.2.3 8“

 

 

 

 

 

 

 

 

 

 

 

 

 

.828. .< 5239..

 

137

 

 

 

 

 

 

 

69.8». 52%»
3908» 8:59.»... 8 ..< 8 28:2» 8 3m 0. goo. on 105-: 8.3.5. «NH.
888». £29»
3288 .258... 8 .< 8 28:2» 8 am 2 goo. on :05... 8. 3.3 a.
$9.8». ESQ.»
932 no.5. 62%» 8 ..< 8 28:2» 8 3... 2 80.. on 10.6.: 8.8. 2 j
69.8». 899.»
»8 5.; :2 82 08.88 8:58... 8 ._< 8 88:2» 8 :m o. 80.. on 00 8.8.3 a
a
.3; 2:» .0: 8.9.8» 08...» new 0.8... 89.8». ESQ.»
8; 0...“... on... 88.68» “...... 9.0.886... 8 ..< .8 28:8» 8 3m. 0. goo. on ".1... 8.3.8 NAIF
.283 828» 85 288 .092
a o. 5: 2 8.32.8 .8 Eu .3 £88 8: 688». E29.» 8208
Eu 80.8» 2.... 8:58... o. 8958» 00 8 .< .8 28:2» 8 am 2 80.. 58 on a»... 00 8.8.8 88

 

 

 

 

 

 

 

 

 

 

 

 

 

.958. .< ...—Ea}.

 

138

Appendix 8: Supplemental data

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1.6 1.6 1.6
14 I60 APG! we 14 I60 AFC-Til 120°C 14 I6!) APG 1.0-c
' GOOmM ' 600ml! ' mm
12 1.2 - 1.2 -
31.0 €10 £10
£00 ‘ \ g'01; A £08
:05 A A :06 ‘ .
0 ‘ x’. o N g
0.4 x- 0.4 ‘
0.2 1,1" 0.2
00%-"5’" . . r - 0.0% .' .' 9 .
0.0 02 0.4 0.6 0.8 1.0 0.0 02 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.d
you 3100 yea

 

 

 

 

Figure 34: Experimental and predicted Langmuir isotherms for mixture of GO
and PG on 3310 carbon at 0.5 M total species concentration and 80°C,
120°C, and 160°C. Abscissa is fraction of GO in the total species (GO +
PG) present Lines are the predicted isotherms.

 

 

 

 

 

 

H2 Hi) DH Dz M-H Mi M M-OH M-OD H20 DHO H00 020
H 2 1 1 0 1 0 O 1 0 0 0 0 0
D 0 1 1 2 0 1 0 O 1 0 O 0 O
O O O 0 0 0 0 0 1 1 1 1 1 1
M 0 O 0 O 1 1 1 1 1 0 O 0 0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Rn= #cp —Rank(FM)= 13-4 =9

Figure B-2: Formula matrix calculation of number of reactions needed.

139

 

Figure B-3: Matrix of coefﬁcients for ODE’s

Table B-1: Unique rate terms for elementary reactions

k2y5y5

r1r=

k1y7y7y1
k1y7y7y2
k1y7y7y3
k1y7y7y4
k3y5y8
k3y5y9
k3y6y8

r1f=

k2y5y6

r2r

r2r=

r2f=

r3r=

r3f=

WW

nhw=

n¢f=

k4y7y7y10
k4y7y7y11

r5r=

r5f=

r6r=

r6f=

k4y7y7y12

r7r=

r1r=

k4y7y7y13
k5y11y12

r8r=

k3Y5Y9

r8f=

r9r=

k5y10y13

r9f=

140

Matlab Code
H2D2mg§tm

1 clear all

2 %hold on .

3 global Mt Dt Nt NI N9 N5 Ht XX reactor_V

4 optz=0

5 %this section deﬁnes initial conditions the ﬁxed parameters

6 DZO_i=0; H20_i=10;%ml

7 H2=0; D2=180%psi

8 grams_Ru=13;

9 reactor_V=23;

10 Gas_V=reactor_V-D20_i-H20_i-grams_Ru/3;

11 % calculated invariants

12 Ng=(H2+DZ)/14.5*Gas_V/22400; Ns=0.4l22400*2*grams_Ru;
13 Nl=(H20__i+D20_i)/18*2; S_i=Ns*1; Nt=Ns+NI+Ng;

14 DI=D20_i/18*2; Ds=D|INI*Ns; Dg=D2l(D2+H2)*Ng; Dt=Dg+Ds+Dl;
15 %initial conditions

16 Yo(1:3)=[(H2)I14.5*Gas_V/22400 ,0,0];

17 Yo(4:13)=[(Ng-Yo(1 )),0,0,S_i*.9999,0,0,(Nl-Dl),0,0,DI];

18 %changeable parameters

19 % k1 =4093; Kr=2.0048; Kg=5.3232; KI=94.379; k5=10000;

20 % k1 =50586; Kr=129.54; Kg=6.9848; KI=8.6846; k5=1000000;
21 % k1 =90938; Kr=308.28; Kg=13.365; Kl=1.5545; k5=1000000;
22 k1 =50586; Kr=129.54; Kg=6.9848; KI=8.6846; k5=10000;

23 % k1 =31093; Kr=2.0048; Kg=5.3232; KI=94.379; k5=10000
24 k1 =5000586; Kr=1000; Kg=1; Kl=.1; k5=1000000;

25 k=[k1,Kr,Kg,Kl,k5];

26 %solve ODE's from 0 to 1000

27 options = odeset( 'RelTol',.00000001,'MaxStep' ,10);

28 [t,Y] = ode15s(@(t,Y) larsf(t,Y,k),[.000010 1000],Yo,options);
29 %calculate fractional info

141

30 calfractions;

31 % plot results (data was really move back to excel and ploted from there!)
32 subplot(3,1 ,1 );plot (t,g);

33 subplot(3,1 ,2);plot (t, (Y(:,1 1 )I2+Y(:,12)12+Y(:,13))l((D20_i+H20_i)l500));
34 subplot(3,1,3);plot (t,HD);

35 % %
36 % % % optimization atempts
37 % % %
38 if ( optz > 0 )

39 getdata %get experiment 1 data

 

 

40 Va: zeros(5,1);temp=1 ; temp2=1 ;sheet=0;
41 k=[k1,Kr,Kg,Kl,k5];

42 R=objectivef(k,tdata1 ,ydata1,Yo);Ro=R+10;
43 while R>.01;

44 dp=1;

45 a = randint(1,1,[1,4]); %randomly pick one to change
46 KTemp=k(a);

47 k(a)=k(a)+(0.5-rand)*k(a)*dp"2;

48 %solve ODE's from 0 to 1000

49 R=objectivef(k,tdata1,ydata1 ,Yo);

50 if (R < R0) ; %if improvement then keep it and save the conditions
51 KR(temp,:)=[R,k];

52 temp=temp+1;

53 R0=R

54 else %restore original value and continue
55 KF(temp2,:)=[R,k];

56 k(a)=KTemp;

57 temp2=temp2+1;

58 end

59 if temp2>1000

60 sheet=sheet+1

142

61 xlswrite( 'H20modeltempdata.xls', KF ,sheet)
62 temp2=1

63 end

64 Va(a)=Va(a)+1; [R0,a,temp2]

65 end

66 end

67

13M

1 function dy = larsf(t,Y,k)

2 global Mt Dt Nt NI Ng Ns Ht XX reactor_V

3 %calculate k values

4 y=Y;

5 y(1 :4)=exp(-U2000)*Y(1 :4);

6 k1 =k(1 );k2=k(1 )Ik(3);k3=k(1 )Ik(2);k4=k3lk(4);k5=k(5);

7 %k1=0;k2=0;k3=0;k4=0;k5=k(5);

8 % clear

9 dydy = zeros(13,1);

10 % calculate

11 V=reactor_V/1000;% vol in liters

12 %calculate each reaction rate pates

13 k1y7y7=k1‘y(7)'Y(7); k4y7y7=k4‘Y(7)‘Y(7) ;

14 >0<=lk2*y(5)*Y(5)N; k2'v(5)*Y(6)N; k2‘y(6)"Y(6)N;

1 5 k1 y7y7*y(1 )NN; k1 y7y7'y(2)NN; k1 y7y7‘y(3)NN;

16 k1 y7y7‘y(4)NN; k4y7y7‘y(10)NN; k4y7y7‘y(1 1 )NN;

1 7 k4y7y7*y(12)NN; k4y7y7‘y(1 3)NN; k3‘y(5)*y(8)N;

18 k3‘y(5)*Y(9)N; k3‘y(6)‘Y(8)N; k3*Y(6)'3/(9)N;

19 k5“y(10)‘y(13)N; k5l4*(y(11)+y(12))"2N];%try using average of the HDO and
HOD

20 % diﬁne the coat for all the ODE

21A=[100-10000000000000;...

143

220100-1000000000000;...
2301000-100000000000;...
24001000-10000000000;...
25-2-2021101100-1-10000;...
260-2-20112001100-1-100;...
27242-2-2-2-2-2-2-2-2222200;...
2800000001010-10-1000;...
29000000001010-10-100;...
300000000-10001000-11;...
31 00000000-10001 001 -1
32000000000-1 0001 01 -1
330000000000-10001-11];
34 %calculate all the ODE's
35dy=A’)0(;

36

37

calfra i ns.m

1 % calculate fractions for each spieces

2 g(:,1:3)= [Y(:,1)/Ng, (Y(:,2)+Y(:,3))lNg, Y(:,4)lNg];

3 l(:,1:3)= [Y(:,10)/Nl, (Y(:,11)+Y(:,12))INI, Y(:,13)INI];

4 s(:,1:5)= [Y(:,5:6)le, Y(:,7)/Ns, (Y(:,8:9))INs];

5 % calculate free site fraction and M-X M-OX fractions

6 ss(:,1:3)= [Y(:,7),(Y(:,5)+Y(:,6))/Ns, (Y(:,8)+Y(:,9))le ];

7 % calculate HID ration in G,S,L

8 HD(:,1:3)= [(Y(:,1)+Y(:,2)I2+Y(:,3)I2)/Ng, (Y(:,5)+Y(:,8)).l(Ns-Y(:,7)),
(Y(: , 1 0)+Y(: ,1 1 )I2+Y(:,12)/2)INI];

144

H202modelfunctions.m

1 %this is a script -— getdate.m

2 %get experimental data

3 tdata1=xlsread('f:\che 801 \data.xls', 'data2', 'time2');
4 ydata1=xlsread('f:\che 801\data.xls', 'data2','HDF2');
5 %
6 function dy = larsf(t,y, k)

7 global Mt Dt Nt Nl Ng Ns Ht XX reactor_V

8 %calculate k values

9 k1 =k(1 );k2=k(1 )Ik(3);k3=k(1)Ik(2);k4=k3lk(4);k5=k(5);
10 % clear

11 dydy = zeros(13,1);

12 % calculate

13 V=reactor_V/1000;% vol in liters

14 %calculate each reaction rate pates

15 k1y7y7=k1‘y(7)*>'(7); k4y7y7=k4*y(7)"Y(7) ;

16 xx=lk2‘y(5)*Y(5)N; k2*y(5)"‘y(6)N; k2‘v(6)‘y(6)N;
17 k1 y7y7*y(1)NN; k1 y7y7*y(2)NN; k1 y7y7‘y(3)NN;
18 k1 y7y7‘y(4)NN; k4y7y7*y(10); k4y7y7*y(1 1 )NN;

 

19 k4y7y7*y(12)NN; k4y7y7‘y(13)NN; k3‘y(5)*y(8)N;

20 k3*Y(5)*y(9)N; k3‘Y(6)‘Y(8)N; k3*y(6)‘Y(9)N;
21 k5*y(10)*y(13)N; k5‘y(11)*y(12)/V];

22 % diﬁne the coeffor all the ODE
23A=[100-10000000000000;...
240100-1000000000000;...
2501000-100000000000;...
26001000-10000000000;...
27-2-2021101100-1-10000;...
280-2-20112001100-1-100;...
29242-2-2-2-2-2-2-2-2222200;...
3000000001010-10-1000;...

145

31000000001010-10-100;...
320000000-10001000-11;...
3300000000-10001001-1;...
34000000000-1000101-1;...
350000000000-10001-11];
36 %calculate all the ODE's
37dy=A*)0(;

38%

 

39 function RR = objectivef(k,tdata,ydata,Yo);

40 global Mt Dt Nt NI Ng Ns Ht XX reactor_V

41 %solve ODE's from 0 to 1000

42 options = odeset( 'RelTol',.00000001,'MaxStep' .10);
43 [t,Y] = ode15$(@(t,Y) larsf(t,Y,k),[0 995],Yo,options);
44 %calculate fractional info

45 calfractions;

46 % extrapolate Experimental data to calculated data
47 Yexp = interp1(tdata,ydata,t, 'pchip');

48 %calculate least squares

49 R=(g-Yexp)."2;

50 %add all least square values

51 RR=sum(sum(R, 'double'), 'double');

146

10.

References

B. Hogle, D. S., K Nagarajan, J.E. Jackson, D.J. Miller Formation and recovery
of itaconic acid from aqueous solutions of citraconic acid and succinic acid.
Industrial and Engineering Chemistry Research 2002, 41, (9), 2069-2073.

Z. Zhang, J. E. J., D.J. Miller, Aqueous-phase hydrogenation of lactic acid to
propylene glycol. Applied Catalysis A: General 2001, 219, 89-98.

Jere, F. T.; Miller, D. J.; Jackson, J. E., Stereoretentive C-H bond activation in
the aqueous phase catalytic hydrogenation of amino acids to amino alcohols.
Organic Letters 2003, 5 (4), 527-530.

Chopade, S. P.; Miller, D. J.; Jackson, J. E.; Werpy, T. A.; Frye, J. G., Jr.;
Zacher, A. H. Catalysts and process for hydrogenolysis of sugar alcohols to
polyols. 2001-US536,2001066499, 2001, 2001.

Shafer, M. W. Catalytic processing of monosaccharides and sugar alcohols to
value-added products. Michigan State University, East Lansing, 2004.

Kovacs, D. 6.; Jackson, J. E.; Miller, D. J., Mechanistic investigation of glycerol
hydrogenolysis. Abstracts of Papers of the American Chemical Society 2001 ,
221, U177-U178.

Crabtree, S. P.; Tyers, D. V. Process for the catalytic hydrogenolysis of sugar
feedstocks into glycols and glycerol. 2004-GB4391,2005051874, 2004, 2005.

Suppes, G. J.; Sutterlin, W. R.; Dasari, M. A. Method of producing lower alcohols
from glycerol. 2005-U89901,2005095536, 20050324., 2005.

Lahr, D. G.; Shanks, B. H., Effect of sulfur and temperature on ruthenium-
catalyzed glycerol hydrogenolysis to glycols. Journal of Catalysis 2005, 232, (2),
386-394.

Lahr, D. G.; Shanks, B. H., lGnetic analysis of the hydrogenolysis of lower
polyhydric alcohols: Glycerol to glycols. Industrial & Engineering Chemistry
Research 2003, 42, (22), 5467-5472.

147

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

Kusunoki, Y.; Miyazawa, T.; Kunimori, K; Tomishige, K, Highly active metal-
acid bifunctional catalyst system for hydrogenolysis of glycerol under mild
reaction conditions. Catalysis Communications 2005, 6, (10), 645-649.

Dasari, M. A.; Kiatsimkul, P. P.; Sutterlin, W. R.; Suppes, G. J., Low-pressure
hydrogenolysis of glycerol to propylene glycol. Applied Catalysis a-General
2005, 281, (1-2), 225-231.

Chaminand, J.; Djakovitch, L.; Gallezot, P.; Marion, P.; Pinel, C.; Rosier, C.,
Glycerol hydrogenolysis on heterogeneous mtalysts. Green Chemistry 2004, 6,
(8), 359-361.

Perosa, A.; Tundo, P., Selective Hydrogenolysis of Glycerol with Raney Nickel.
Industrial & Engineering Chemistry Research 2005, 44, (23), 8535-8537.

Montassier, C.; Menezo, J. C.; Hoang, L. C.; Renaud, C.; Barbier, J., Aqueous
polyol conversions on ruthenium and on sulfur-modified ruthenium. Journal of
Molecular Catalysis 1991, 70, (1 ), 99-110.

Hahn, R.; Tscheliessnig, A; Zochling, A.; Jungbauer, A., Shallow bed
adsorption: Theoretical background and applications. Chemical Engineering &
Technology 2005, 28, (1 1), 1241 -1 251.

Alduri, B., A Review in Equilibrium in Single and Multicomponent Liquid
Adsorption Systems. Reviews in Chemical Engineering 1995, 11, (2), 101-143.

Comelissen, G.; Gustafsson, 0.; Bucheli, T. D.; Jonker, M. T. 0.; Koelmans, A.
A.; Van Noort, P. C. M., Extensive sorption of organic compounds to black
carbon, coal, and kerogen in sediments and soils: Mechanisms and

consequences for distribution, bioaccumulation, and biodegradation.
Environmental Science & Technology 2005, 39, (18), 6881 -6895.

Smith, E. H., Evaluation of Multicomponent Adsorption Equilibria for Organic
Mixtures onto Activated Carbon. Water Research 1991, 25, (2), 125-134.

Xing, B. S.; Pignatello, J. J.; Gigliotti, 8., Competitive sorption between atrazine
and other organic compounds in soils and model sorbents. Environmental
Science & Technology 1996, 30, (8), 2432-2440.

148

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

Li, F. S.; Yuasa, A.; Ebie, K; Azuma, Y.; Hagishita, T.; Matsui, Y., Factors
affecting the adsorption capacity of dissolved organic matter onto activated
carbon: modified isotherm analysis. Water Research 2002, 36, (18), 4592-4604.

Antonjuk, A. A; Redkovsky, N. N.; Cari, P. S.; Seidel, A.; Marutovsky, R. M.,
Determination of lntraparticle lﬁnetic-Parameters for Adsorption of Binary-
Solutions on Activated Carbon. Chemical Engineering Science 1991, 46, (4),
1 035-1039.

Leitao, A.; Serrao, R., Adsorption of phenolic compounds from water on
activated carbon: Prediction of multicomponent equilibrium isotherms using
single-component data. Adsorption-Joumal of the lntemational Adsorption
Society 2005, 1 1, (2), 167-179.

Abburi, K, Adsorption of phenol and p—chlorophenol from their single and
bisolute aqueous solutions on Amberlite XAD-16 resin. Journal of Hazardous
Materials 2003, 1 05, (1 -3), 143-1 56.

Yaacoubi, A.; Mazet, M.; Dusart, 0., Competitive Effect in Bi-Solute Adsorption
onto Activated Carbon - Dss, Alcohols and Phenols as Solutes. Water Research
1991, 25, (8), 929-937.

Lu, Q. L.; Sorial, G. A, Adsorption of phenolics on activated carbon - impact of
pore size and molecular oxygen. Chemosphere 2004, 55, (5), 671 -679.

Dekock, F. P.; Vandeventer, J. S. J., Stochastic—Model for Equilibrium Adsorption
onto Activated Carbon. Chemical Engineering Journal and the Biochemical
Engineering Journal 1 995, 59, (3), 205-220.

Lee, J. W.; Kwon, T. 0.; Moon, I. 8., Adsorption of mono saccharides,
disaccharides, and maltooligosaccharides on activated carbon for separation of
maltopentaose. Carbon 2004, 42, (2), 371 -380.

Liu, X.; Pinto, N. 6., Ideal adsorbed phase model for adsorption of phenolic
compounds on activated carbon. Carbon 1997, 35, (9), 1387-1397.

Kumar, A.; Kumar, S.; Kumar, S., Adsorption of resorcinol and catechol on
granular activated carbon: Equilibrium and kinetics. Carbon 2003, 41, (15), 3015-
3025.

149

31.

32.

33.

35.

37.

38.

39.

Lu, Q. L.; Sorial, G. A, The role of adsorbent pore size distribution in
multicomponent adsorption on activated carbon. Carbon 2004. 42, (15), 3133-
3142.

Braeken, L.; Boussu, K; Van der Bruggen, 8.; Vandecasteele, C., Modeling of
the adsorption of organic compounds on polymeric nanoﬁltration membranes in
solutions containing two compounds. Chemphyschem 2005, 6, (8), 1606-1612.

Frimmel, F. H.; Assenmacher, M.; Kumke, M. U.; Specht, C.; Abbt-Braun, G.;
Grabe, 6., Removal of hydrophilic compounds from water with organic polymers
- Part II: Adsorption behavior of industrial wastewater. Chemical Engineering and
Processing 2002, 41, (9), 731 -736.

Matsui, Y.; Yuasa, A; Ariga, K, Removal of a synthetic organic chemical by
PAC-UF systems - l: Theory and modeling. Water Research 2001, 35, (2), 455-
463.

Bonalumi, N.; Vargas, A; Ferri, D.; Burgi, T.; Mallat, T.; Baiker, A, Competition
at chiral metal surfaces: Fundamental aspects of the inversion of
enantioselectivity in hydrogenations on platinum. Journal of the American
Chemical Society 2005, 127, (23), 8467-8477.

Vargas, A; Ferri, D.; Baiker, A, DFT and ATR-IR insight into the conformational
ﬂexibility of cinchonidine adsorbed on platinum: Proton exchange with metal.
Journal of Catalysis 2005, 236, (1 ), 1-8.

Keresszegi, C.; Ferri, D.; Mallat, T.; Baiker, A, Unraveling the surface reactions
during liquid-phase oxidation of benzyl alcohol on Pd/Al203: an in situ ATR-IR
study. Journal of Physical Chemistry B 2005, 109, (2), 958-967.

Hamminga, G. M.; Mul, G.; Moulijn, J. A, Real-time in situ ATR-FI’IR analysis of
the liquid phase hydrogenation of gamma-butyrolactone over Cu-ZnO catalysts:
A mechanistic study by varying lactone ring size. Chemical Engineering Science
2004, 59, (22-23), 5479-5485.

Lefevre, G., In situ F ourier-transform infrared spectroscopy studies of inorganic
ions adsorption on metal oxides and hydroxides. Advances in Colloid and
Interface Science 2004, 107, (2-3), 109-123.

150

40.

41.

42.

43.

45.

47.

48.

49.

Mul, G.; Hamminga, G. M.; Moulijn, J. A, Operando ATR-FI' IR analysis of liquid-
phase catalytic reactions: can heterogeneous catalysts be observed? Vibrah’onal
Spectroscopy 2004, 34, (1 ), 109-121.

Hind, A. R.; Bhargava, S. K; McKinnon, A, At the solid/liquid interface:
FTIRIATR - the tool of choice. Advances in Colloid and Interface Science 2001,
93, (1-3), 91-114.

Ninness, B. J.; Bousﬁeld, D. W.; Tripp, C. P., In situ infrared technique for
studying adsorption onto particulate silica surfaces from aqueous solutions.
Applied Spectroscopy 2001, 55, (6), 655-662.

Lazorenko-Manevich, R. M., Adatom hypothesis as a predominant mechanism of
surface enhanced Raman scattering: A review of experimental argumentation.
Russian Journal of Electrochemistry 2005, 41 , (8), 799-831.

Tian, Z 0.; Ron, 8, Adsorption and reaction at electrochemical interfaces as
probed by surface-enhanced Raman spectroscopy. Annual Review of Physical
Chemistry 2004, 55, 197-229.

Brolo, A. G.; Irish, D. E.; Smith, 8. D., Applications of surface enhanced Raman
scattering to the study of metal-adsorbate interactions. Journal of Molecular
Structure 1997, 405, (1 ), 29-44.

Urban, M. W., Surface and Interface Vibrational Spectroscopy - Relevance to
Adhesion. Journal of Adhesion Science and Technology 1993, 7, (1 ), 1-47.

Otto, A, Surface-Enhanced Raman-Scattering of Adsorbates. Journal of Raman
Spectroscopy 1991, 22, (12), 743-752.

Guo, Q. L.; Wang, E. G., Geometric and electronic structure of Cu on corundum
(0001) surfaces. Science and Technology of Advanced Materials 2005, 6, (7),
795-798.

Reniers, F.; Tewell, C., New improvements in energy and spatial (x, y, 2)
resolution in AES and XPS applications. Journal of Electron Spectroscopy and
Related Phenomena 2005, 142, (1 ), 1-25.

151

50.

51.

52.

53.

55.

57.

58.

59.

Duwez, A S., Exploiting electron spectroscopies to probe the structure and
organization of self-assembled monolayers: a review. Journal of Electron
Spectroscopy and Related Phenomena 2004, 134, (2-3), 97-138.

Marcus, P.; Galtayries, A, ESCA and metal surfaces - Alloys, corrosion and
surface treatments. Vide-Science Technique Et Applications 2003, 58, (308),
230-+.

Sun, C. H.; Berg, J. C., A review of the different techniques for solid surface
acid-base characterization. Advances in Colloid and Interface Science 2003,
105, 151-175.

Pollini, l.; Mosser, A; Pariebas, J. C., Electronic, spectroscopic and elastic
properties of eariy transition metal compounds. Physics Reports-Review Section
of Physics Letters 2001, 355, (1 ), 1-72.

Zemek, J., Electron spectroscopy of solid surfaces. Acta Physica Slovaca 2000,
50, (5), 577-589.

Ronning, C.; Felderrnann, H.; Merk, R.; Hofsass, H.; Reinke, P.; Thiele, J. U.,
Carbon nitride deposited using energetic species: A review on XPS studies.
Physical Review B 1998, 58, (4), 2207-2215.

Weitzsacker, C. L.; Xie, M.; Drzal, L. T., Using XPS to investigate ﬁber matrix
chemical interactions in carbon-ﬁber-reinforced composites. Surface and
Interface Analysis 1997, 25, (2), 53-63.

Wang, J. X.; Marinkovic, N. S.; Zajonz, H.; Ocko, B. M.; Adzic, R. R., In situ X-
ray reﬂectivity and voltammetry study of Ru(0001) surface oxidation in electrolyte
solutions. Journal of Physical Chemistry B 2001 , 105, (14), 2809-2814.

Ciobica, l. M.; Kleyn, A. W.; Van Santen, R. A, Adsorption and coadsorption of
CO and H on ruthenium surfaces. Journal of Physical Chemistry B 2003, 107,
(1 ), 164-172.

Riedmuller, B.; Ciobica, I. M.; Papageorgopoulos, D. C.; Frechard, F.; Berenbak,
B.; Kleyn, A. W.; van Santen, R. A, CO adsorption on hydrogen saturated
Ru(0001). Journal of Chemical Physics 2001, 115, (11), 5244-5251.

152

60.

61.

62.

63.

65.

67.

68.

69.

70.

Peebles, D. E.; Schreifels, J. A; White, J. M., The Interaction of Coadsorbed
Hydrogen and Carbon-Monoxide on Ru(001). Surface Science 1982, 116, (1 ),
1 1 7-1 34.

Pallassana, V.; Neurock, M., Reaction paths in the hydrogenolysis of acetic acid
to ethanol over Pd(111), Re(0001), and PdRe alloys. Journal of Catalysis 2002,
209, (2), 289-305.

Haile, J. M., Molecular dynamics simulation : elementary methods. Wiley: New
York, 1992; p xvii, 489 p.

Jacob, T.; Goddard, W. A, Agostic interactions and dissociation in the ﬁrst layer
of water on Pt(111). Journal of the American Chemical Society 2004, 126, (30),
9360-9368.

Ogasawara, H.; Brena, B.; Nordlund, D.; Nyberg, M.; Pelmenschikov, A;
Pettersson, L. G. M.; Nilsson, A, Structure and bonding of water on Pt(111).
Physical Review Letters 2002, 89, (27).

Ito, M.; Nakamura, M., Hydration processes of electrolyte anions and cations on
Pt(111), Ir(111), Ru(001) and Au(111) surfaces: Coadsorption of water
molecules with electrolyte ions. Faraday Discussions 2002, 121, 71 -84.

Thor Dozeman, G. J., TLE Catalytic conversion of D-glucitol and isosorbide
using zeolite.

Dozeman, G. J., Catalytic conversion of D-glucitol and isosorbide using zeolite
catalysts. In 1989.

Chen, Y. Aqueous-Phase Catalytic Hydrogenation of Organic Acids and Theit
Mixtures. Michigan State University, East Lansing, 2007.

Parsons, M. T.; Westh, P.; Davies, J. V.; Trandum, C.; To, E. C. H.; Chiang, W.
M.; Yee, E. G. M.; Koga, Y., A thermodynamic study of 1-propanol-glyceroI-H20
at 25 degrees C: Effect of glycerol on molecular organization of H20. Journal of
Solution Chemistry 2001 , 30, (1 1 ), 1007-1028.

Parsons, M. T.; Koga, Y., Interactions in 1-propanoI-(1,2-and 1,3-)propanediol-
H20: The effect of hydrophobic vs hydrophilic moiety on the molecular

153

71.

72.

73.

74.

organization of H20. Journal of Physical Chemistry B 2002, 106, (28), 7090-
7095.

Hansch, C.; Hoekman, 0; Leo, A; Zhang, L. T.; Li, P., The Expanding Role of
Quantitative Stmcture-Activity-Relationships (Qsar) in Toxicology. Toxicology
Letters 1995, 79, (1-3), 45-53.

Sorial, G. A; Suidan, M. T.; Vidic, R. D.; Maloney, S. W., Competitive Adsorption
of Phenols on Gac .1. Adsorption Equilibrium. Journal of Environmental
Engineering-Asce 1 993, 1 19, (6), 1026-1043.

McKay, G.; Alduri, B., Multicomponent Dye Adsorption onto Carbon Using a
Solid Diffusion Mass-Transfer Model. Industrial & Engineering Chemistry
Research 1991, 30, (2), 385-395.

Yaws, C. L. N., Prasad, K; Gabbula, Chaitanya . Yaws' Handbook of Antoine
Coefﬁcients for Vapor Pressure (Electronic Edition). Knovel.

154

 

Illllllllllll