CO2 Presentation

Preview:

Citation preview

Computational and literature investigation to understand the

mechanism behind the catalyzed hydrogenation of CO2

- Darrell Nelson

Outline I. Background

II. Attacking the Problem

III. Metal Oxides

IV. Mechanism

V. Additional Factors

VI. Future Work

Background

Converting CO2 into useful fuels (i.e. methanol, methane)

(MTO) Methanol to Olefins (i.e. ethylene, propylene)

Create CO to be used as syngas

1$27 billion annual market for ethylene glycol

1. © 2013 Liquid Light Corporation. All rights reserved. | www.llchemical.com

2

35,000 Mt produced per year 27,000 MtCO2

3

Energy Crisis

Over 7 billion people in the world

Depleting energy sources rapidly

Economies of both China and India are growing

Creating scarcity of natural resources

2 problems 1 solution Hydrogenation of CO2 instead of sequestration

Stops emissions and provides very cheap fuel can be used again and again

Attacking the problem

CO2 is a very stable and oxidized form of C Linear molecule No strain Add a high energy electron makes it unstable

Catalyst lifetime

4

How to unravel this complex system?

𝐴+𝐵 𝐴𝐵CAT

How to make the best catalyst?

What is the composition/structure of the catalyst Bulk structure (lattice) Surface composition Subsurface composition “Active Sites”

How to make the best catalyst?

What is the reaction mechanism?

Surface chemistry is dynamic

Active Sites are changing

5

How to make the best catalyst?

4

4

Metal Oxides (Chosen catalyst)

e.g. (Al2O3, SiO2) High surface area Very stable under usual chemical reaction conditions Reduced metal oxides (i.e. CuO, Ce2O) that can change their oxidation

states and have vacancies in their structure upon release/storage of oxygen Diffusion from the bulk

4

Optimization Problem

Keeping the reactivity high involves not changing the catalyst. But, reactivity means that the catalyst is unstable and willing to change.

Using an industrial high-throughput approach to find the maximum between lifetime and reactivity

What is the chemistry? As you can see from previous slides Surface properties of the catalyst are the most

important (e.g. solid acid/base)

CO2 is a Lewis acid

Look at materials that are willing to donate electrons Low valent ion oxides are preferred

Correlation

0 0.5 1 1.5 2 2.5 3 3.50

10

20

30

40

50

60

70

80

Activation vs Conversion

Activation Energy

% C

O2 co

nver

sion

Statistical Analysis

Highest conversion came from Ni and Fe

46 out of 287 Ni, 34 out of 287 Fe

Mechanism (microscopic)

Understanding mechanism sheds light on the “why and how” Create a high performance catalyst based off of its properties and not

through trial and error Adsorption creates carbonate group on M.O. (not MgO)

Why not MgO? What’s special/different? Makes sense to use compounds that have coordinated oxygens because of

their high electronegativity The electronegativity of oxygen activates the binding sites

5

Note on transition states

3H2 + CO2 CH3OH + H2O

6Some elements help facilitate other parts of the reaction (Pt – good at disassocitating H2 , Zn – good at binding, Cu – catalyzes transition states)

6. Ref #2 of J. Graciani, K. Mudiyanselage, F. Xu, a. E. Baber, J. Evans, S. D. Senanayake, D. J. Stacchiola, P. Liu, J. Hrbek, J. F. Sanz, and J. a. Rodriguez, “Highly active copper-ceria and copper-ceria-titania catalysts for methanol synthesis from CO2,” Science (80-. )., vol. 345, no. 6196, pp. 546–550, 2014.

Analyzing hydrogenation of CO2 on ceria

% use ode45 first and then compare to ode15s, this is a 'stiff' problem% (concentrations are changing at different time scales) so I want to% measure the accuracy between the twox0=0; xf=40; %start at time zero and go for 40 seconds%assume 1:1 molar ratio of CO2 and H2 with the number active sites being 7y0=[5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 5]; %last point is active sitesoptions = odeset('RelTol',1e-3,'NonNegative',ones(25,1));

[x,y] = ode15s(@f,[x0 xf],y0,options);

%plot methanol as a function of timefigureplot(x,y(:,19))xlabel('time (s)')ylabel('[CH_3OH]')%plot adsorbed hrdroxylfigureplot(x,y(:,20))xlabel('time (s)')ylabel('[OH*]')%adsorbed waterfigureplot(x,y(:,21))xlabel('time (s)')ylabel('[H_2O*]')%carbon dioxidefigureplot(x,y(:,1))xlabel('time (s)')ylabel('[CO_2]')

% Assume number of available active sites to be 7, therefore they are not% rate limiting. Need 5 sites total for the mechanism.

function dC = f(~,y)% all rate constants are in INVERSE SECONDS%assume them to be for second order processesglobal rateskload myparam% ratesk = [1.95e4 3.23e1% 2.67e4 5.19e4% 6.31e3 3.13e5% 3.11e1 1.78e2% 2.17e3 4.21e3% 1.02e1 3.57e3% 4.24e6 3.82e6% 3.82e-3 1.31e-2% 7.33e-4 5.41e-5% 6.55e-2 8.21e-4% 3.91e4 3.55e2% 5.11e3 8.26e5% 7.63e3 5.19e2% 4.55e-4 1.98e-2% 1.02e1 2.82e3% 3.76e1 5.43e-1% 9.21 2.13e-1% 8.33e1 4.08e2% 5.91e-3 3.41e-5% 7.32e2 2.45e2% 5.89e1 2.87e1];% ratesk=ratesk*1e-5;%rates of depletion -r_1dC(1) = ratesk(1,1)*y(1)*y(25) - ratesk(1,2)*y(2); %multiplying by one species instead of two?dC(2) = ratesk(2,1)*y(2)*y(24) - ratesk(2,2)*y(3)*y(25) - ratesk(1,1)*y(1)*y(25) + ratesk(1,2)*y(2) + ratesk(11,1)*y(2)*y(24) - ratesk(11,2)*y(12)*y(25);dC(3) = ratesk(3,1)*y(3)*y(25) - ratesk(3,2)*y(4)*y(20) + ratesk(4,1)*y(3)*y(24) - ratesk(4,2)*y(5)*y(25)- ratesk(2,1)*y(2)*y(24) + ratesk(2,2)*y(3)*y(25);dC(4) = ratesk(6,1)*y(4) - ratesk(6,2)*y(7)*y(25) - ratesk(3,1)*y(3)*y(25) + ratesk(3,2)*y(4)*y(20) + ratesk(7,1)*y(4)*y(24) - ratesk(7,2)*y(8)*y(25);dC(5) = ratesk(5,1)*y(5) - ratesk(5,2)*y(6)*y(25) - ratesk(4,1)*y(3)*y(24) + ratesk(4,2)*y(5)*y(25);dC(6) = -ratesk(5,1)*y(5) + ratesk(5,2)*y(6)*y(25);dC(7) = -ratesk(6,1)*y(4) + ratesk(6,2)*y(7)*y(25);dC(8) = ratesk(8,1)*y(8)*y(24) - ratesk(8,2)*y(9)*y(25) - ratesk(7,1)*y(4)*y(24) + ratesk(7,2)*y(8)*y(25);dC(9) = -ratesk(8,1)*y(8)*y(24) + ratesk(8,2)*y(9)*y(25) + ratesk(9,1)*y(9)*y(24) - ratesk(9,2)*y(10)*y(25);dC(10) = -ratesk(9,1)*y(9)*y(24) + ratesk(9,2)*y(10)*y(25) + ratesk(10,1)*y(10)*y(24) - ratesk(10,2)*y(11)*y(25);dC(11) = -ratesk(10,1)*y(10)*y(24) + ratesk(10,2)*y(11)*y(25) - ratesk(17,1)*y(17)*y(24) + ratesk(17,2)*y(18)*y(25) + ratesk(18,1)*y(18) - ratesk(18,2)*y(19)*y(25);dC(12) = -ratesk(11,1)*y(2)*y(24) + ratesk(11,2)*y(12)*y(25) + ratesk(12,1)*y(12)*y(24) - ratesk(12,2)*y(13)*y(25);dC(13) = -ratesk(12,1)*y(12)*y(24) + ratesk(12,2)*y(13)*y(25) + ratesk(13,1)*y(13)*y(24) - ratesk(13,2)*y(14)*y(25);dC(14) = -ratesk(13,1)*y(13)*y(24) + ratesk(13,2)*y(14)*y(25) + ratesk(14,1)*y(14)*y(25) - ratesk(14,2)*y(15)*y(20);dC(15) = -ratesk(14,1)*y(14)*y(25) + ratesk(14,2)*y(15)*y(20) + ratesk(15,1)*y(15) - ratesk(15,2)*y(16)*y(25) + ratesk(16,1)*y(15)*y(24) - ratesk(16,2)*y(17)*y(25);dC(16) = -ratesk(15,1)*y(15) + ratesk(15,2)*y(16)*y(25);dC(17) = -ratesk(16,1)*y(15)*y(24) + ratesk(16,2)*y(17)*y(25) + ratesk(17,1)*y(17)*y(24) - ratesk(17,2)*y(18)*y(25);dC(18) = -ratesk(10,1)*y(10)*y(24) + ratesk(10,2)*y(11)*y(25) - ratesk(17,1)*y(17)*y(24) + ratesk(17,2)*y(18)*y(25) + ratesk(18,1)*y(18) - ratesk(18,2)*y(19)*y(25);dC(19) = -ratesk(18,1)*y(18) + ratesk(18,2)*y(19)*y(25);dC(20) = -ratesk(3,1)*y(3)*y(25) + ratesk(3,2)*y(4)*y(20) - ratesk(14,1)*y(14)*y(25) + ratesk(14,2)*y(15)*y(20) + ratesk(19,1)*y(24)*y(20) - ratesk(19,2)*y(21)*y(25);dC(21) = -ratesk(19,1)*y(24)*y(20) + ratesk(19,2)*y(21)*y(25) + ratesk(20,1)*y(21) - ratesk(20,2)*y(22)*y(25);dC(22) = -ratesk(20,1)*y(21) + ratesk(20,2)*y(22)*y(25);dC(23) = ratesk(21,1)*y(23)*y(24)*y(24) - ratesk(21,2)*y(25)*y(25);dC(24) = ratesk(2,1)*y(2)*y(24) - ratesk(2,2)*y(3)*y(25) + ratesk(4,1)*y(3)*y(24) - ratesk(4,2)*y(5)*y(25) + ratesk(7,1)*y(4)*y(24) - ratesk(7,2)*y(8)*y(25) + ratesk(8,1)*y(8)*y(24) - ratesk(8,2)*y(9)*y(25)... + ratesk(9,1)*y(9)*y(24) - ratesk(9,2)*y(10)*y(25) + ratesk(10,1)*y(10)*y(24) - ratesk(10,2)*y(11)*y(25) + ratesk(11,1)*y(2)*y(24) - ratesk(11,2)*y(12)*y(25) + ratesk(12,1)*y(12)*y(24) - ratesk(12,2)*y(13)*y(25)... + ratesk(13,1)*y(13)*y(24) - ratesk(13,2)*y(14)*y(25) + ratesk(16,1)*y(15)*y(24) - ratesk(16,2)*y(17)*y(25) + ratesk(17,1)*y(17)*y(24) - ratesk(17,2)*y(18)*y(25) + ratesk(19,1)*y(24)*y(20) - ratesk(19,2)*y(21)*y(25)... -2*ratesk(21,1)*y(23)*y(24)*y(24) + 2*ratesk(21,2)*y(25)*y(25);dC(25) = dC(1) - ratesk(2,1)*y(2)*y(24) + ratesk(2,2)*y(3)*y(25) + ratesk(3,1)*y(3)*y(25) - ratesk(3,2)*y(4)*y(20) - ratesk(4,1)*y(3)*y(24) + ratesk(4,2)*y(5)*y(25) - ratesk(5,1)*y(5) + ratesk(5,2)*y(6)*y(25)... - ratesk(7,1)*y(4)*y(24) + ratesk(7,2)*y(8)*y(25) - ratesk(8,1)*y(8)*y(24) + ratesk(8,2)*y(9)*y(25) - ratesk(9,1)*y(9)*y(24) + ratesk(9,2)*y(10)*y(25) - ratesk(10,1)*y(10)*y(24) + ratesk(10,2)*y(11)*y(25)... - ratesk(11,1)*y(2)*y(24) + ratesk(11,2)*y(12)*y(25) - ratesk(12,1)*y(12)*y(24) + ratesk(12,2)*y(13)*y(25) - ratesk(13,1)*y(13)*y(24) + ratesk(13,2)*y(14)*y(25) + ratesk(14,1)*y(14)*y(25) - ratesk(14,2)*y(15)*y(20)... - ratesk(15,1)*y(15) + ratesk(15,2)*y(16)*y(25)- ratesk(16,1)*y(15)*y(24) + ratesk(16,2)*y(17)*y(25) - ratesk(17,1)*y(17)*y(24) + ratesk(17,2)*y(18)*y(25) - ratesk(18,1)*y(18) + ratesk(18,2)*y(19)*y(25)... - ratesk(19,1)*y(24)*y(20) + ratesk(19,2)*y(21)*y(25) - ratesk(20,1)*y(21) + ratesk(20,2)*y(22)*y(25) + 2*ratesk(21,1)*y(23)*y(24)*y(24) - 2*ratesk(21,2)*y(25)*y(25);

dC = dC'; %to return a column vector

% dC(7) = ratesk(7,1)*y(4)*y(24) - ratesk(7,2)*y(8)*y(25);% dC(8) = ratesk(8,1)*y(8)*y(24) - ratesk(8,2)*y(9)*y(25);% dC(9) = ratesk(9,1)*y(9)*y(24) - ratesk(9,2)*y(10)*y(25);% dC(10) = ratesk(10,1)*y(10)*y(24) - ratesk(10,2)*y(11)*y(25);% dC(11) = ratesk(11,1)*y(2)*y(24) - ratesk(11,2)*y(12)*y(25);% dC(12) = ratesk(12,1)*y(12)*y(24) - ratesk(12,2)*y(13)*y(25);% dC(13) = ratesk(13,1)*y(13)*y(24) - ratesk(13,2)*y(14)*y(25);% dC(14) = ratesk(14,1)*y(14)*y(25) - ratesk(14,2)*y(15)*y(20);% dC(15) = ratesk(15,1)*y(15) - ratesk(15,2)*y(16)*y(25);% dC(16) = ratesk(16,1)*y(15)*y(24) - ratesk(16,2)*y(17)*y(25);% dC(17) = ratesk(17,1)*y(17)*y(24) - ratesk(17,2)*y(18)*y(25);% dC(18) = ratesk(18,1)*y(18) - ratesk(18,2)*y(19)*y(25);% dC(19) = ratesk(19,1)*y(24)*y(20) - ratesk(19,2)*y(21)*y(25);% dC(20) = ratesk(20,1)*y(21) - ratesk(20,2)*y(22)*y(25);% dC(21) = ratesk(21,1)*y(23)*y(24)*y(24) - ratesk(21,2)*y(25)*y(25); %make sure this rate form is correct

EndEnd

function dC = f(~,y)% all rate constants are in INVERSE SECONDSglobal rateskload myparam% ratesk = [1.95e4 3.23e1% 2.67e4 5.19e4% 6.31e3 3.13e5% 3.11e1 1.78e2% 2.17e3 4.21e3% 1.02e1 3.57e3% 4.24e6 3.82e6% 3.82e-3 1.31e-2% 7.33e-4 5.41e-5% 6.55e-2 8.21e-4% 3.91e4 3.55e2% 5.11e3 8.26e5% 7.63e3 5.19e2% 4.55e-4 1.98e-2% 1.02e1 2.82e3% 3.76e1 5.43e-1% 9.21 2.13e-1% 8.33e1 4.08e2% 5.91e-3 3.41e-5% 7.32e2 2.45e2% 5.89e1 2.87e1];

Additional Factors

Treatments (a priori) Incipient wetness method, reverse microemulsion (RM), coprecipitation

and impregnation

Dopants and deposits

Varying temperature and pressure a priori and/or in situ Thermodynamic and kinetic effects

Future Work Comparing catalysts from database Mapping and understanding the change in electron density as CO2 is adsorbed onto

surface and converted

Plot activity, activation energy and adsorption energy vs d-band center

Machine Learning

Understand Transition State Theory and recalculate k’s

Future Work

Machine Learning Supervised Learning

Using gradient descent to maximize specified characteristics (θ0, θ1 are independent variables)

Machine Learning Classification

Where 0 is negative and 1 is positive

Decision Boundary

Machine Learning

Questions?

Works Cited1. © 2013 Liquid Light Corporation. All rights reserved. | www.llchemical.com

2. UNEP, Introduction to Climate Change http://www.grida.no/climate/vital/05.htm

3. Global greenhouse gas emissions (2012), International Energy Agency.4. Dr. John T. Gleaves – received on December 2, 2015

5. Lo, C.; Cheng, Zhuo. EECE Washington University6. Ref #2 of J. Graciani, K. Mudiyanselage, F. Xu, a. E. Baber, J. Evans, S. D. Senanayake, D. J. Stacchiola, P. Liu, J. Hrbek, J. F. Sanz, and J. a. Rodriguez, “Highly active copper-ceria and copper-ceria-titania catalysts for methanol synthesis from CO2,” Science (80-. )., vol. 345, no. 6196, pp. 546–550, 2014.

Recommended