Data-driven discovery and optimisation of organometallic catalystsDr Natalie Fey

Dial-a-Molecule Annual Meeting

3rd July 2019, York

Natalie.Fey@Bristol.ac.uk

Opportunities from Data

▪ Select “best” synthetic route– Better use of resources/time.

– Break “cultural attachments” to routes.

– Find new approaches.

– When to stop.

▪ Broaden our toolkit, explore new chemistry.

▪ Deepen our understanding (mechanism, catalyst/substrate matching).

▪ Combine experiment and computation to work towards prediction.

Chem. Central J. 2015, 9, 38

Property Databases

Ligand Knowledge Bases▪ Ligands used to tune complex

properties.

▪ Characterise ligands (size, electronics).

▪ Map ligand properties.

▪ Derive structure-property relationships.

▪ Are two dimensions sufficient?

▪ Transferability to other ligand types.

80 100 120 140 160 180 200

Cone Angle (°)

2060

2080

2100

TE

P (

cm

-1)

CHCH2

R

Ar

Hal

OR

NR2

PF3

PCl3

P(C6F

5)3

P(O-o-tBu-C6H

4)3

P(o-tolyl)3

PtBu3PCy

3

PiPr3

P(NMe2)3

PBz3

PPh3

PEt3

PMe3

PH3

P(OEt)3

P(OMe)3

P(OPh)3

electron

withdrawing

electron

donating

small large

Tolman, Chem. Rev. 1977, 77, 313–348.

For other examples: QALE, Cundari, Rothenberg.

See Coord. Chem. Rev. 2009, 253, 704-722.

LKB-P▪ 28 descriptors, quite highly correlated.

▪ Projection methods:– Reduce number of variables (dimensions) by transformation of original

variables.

– Optimally represent distances between objects.

▪ Principal Component Analysis– Describe variation of data in terms of uncorrelated variables.

– Derive from linear combinations of original variables.

– Maximum variance in first PC, decreasing order of importance, orthogonal.

PC1 = a1,11 + a1,22 + … + a1,nn

PC2 = a2,12 + a2,22 + … + a2,nn

BP86/6-31G*, LACV3P on Au, Pd, Pt

PA3, PA2B, A =RCH2=CH2

SiMe3

ArOR/OArNR2/NAr2

HalPABC

PI3

PBr3

PCl3

PF3

PtBu3

PMe3

P(CF3)3

PMe2(CF3)

PMe(CF3)2

P(C6F5)3

PPh3

P(OMe)3

P(OPh)3

Organometallics, 2010, 29, 6245;

Chem. Eur. J. 2006, 12, 291Dr Jesús Jover, with JNH, AGO, GCLJ

28 descriptors, 348 L, 65%

LKB-P Map

Other ligands

LKB-PP

Organometallics, 2012, 31, 5302;

Organometallics 2008, 27, 1372

LKB-PPscreen

Dalton Trans. 2013, 42, 172

(i) Me

(ii) CF3

(iii) tBu

(iv) Ph

(v) C6F5

(vi) o-tol

(vii) p-tol

(viii) p-CF3-Ph

(ix) OMe

(x) OPh

(xi) NMe2

A2P~PA2, A=N

EW

S

Other ligands

LKB-PP

Organometallics, 2012, 31, 5302;

Organometallics 2008, 27, 1372

LKB-PPscreen

Dalton Trans. 2013, 42, 172

LKB-C

Dalton Trans. 2009, 8183

(i) Me

(ii) CF3

(iii) tBu

(iv) Ph

(v) C6F5

(vi) o-tol

(vii) p-tol

(viii) p-CF3-Ph

(ix) OMe

(x) OPh

(xi) NMe2

A2P~PA2, A=N

EW

S

Bidentate Ligand Space

with Dr Charlotte Willans, Dr Bao Nguyen, Prof. Andrei

Malkov, Dr Jonathan Moseley, Dr Simon Tyler

224 ligands, 52%

C-, N-, O-donor ligands only

146 ligands, 57%

Publication in preparation

Dr Gareth Owen-Smith, Dr Jesús Jover, with

Jeremy Harvey, Guy Orpen, Guy Lloyd-Jones

Experiment

Dr Gareth Owen-Smith, Dr Jesús Jover, with

Jeremy Harvey, Guy Orpen, Guy Lloyd-Jones

Experiment

Global Models:

-15 -10 -5 0 5

PC1

-11

-6

-1

4

PC

2

LM5: Q(B fragm.), Q(Pd fragm.), Q(Pt

fragm.), BE(Au), BE(Pd), P-Pt, P-A(B),

P-A(Pt) [8], R2 = 0.687, CV PE = 0.83

-4.00

-3.00

-2.00

-1.00

0.00

1.00

2.00

3.00

-8.00 -6.00 -4.00 -2.00 0.00

LM

5 R

es

idu

als

LM5 Predicted log(initial rate)

Dr Gareth Owen-Smith, Dr Jesús Jover, with

Jeremy Harvey, Guy Orpen, Guy Lloyd-Jones

LASSO (1000 random samples)

Dr Ben Swallow (JGI Seedcorn Funding)

Beyond Ligands…

Alkynes & Vinylidenes

with Dr Jason Lynam,

Organometallics, 2014, 33, 1751

Tautomer Preference Ev-a(Ru)

Chem. Commun. 2015, 51, 9702

64 pairs, 56 %

with Prof. Paula Diaconescu

Redox-Switchable Catalysts

64 pairs, 56 %

Chem. Commun., 2019, 55, 7021-7024

MechanismM

M

A

A

B

M

A

B

M

AB

AB

M

A

B1

B2

AB1 M

B2

+

Unwanted

Desired

Bristol Reactivity in Catalysis Knowledgebase (BRiCK)

▪ Target: New applications for known catalysts.

▪ Systematic study of catalytic cycles.– Phase 1: Substrate effects on barriers.

– Phase 2: Catalysts modification (metal, ligands).

– Phase 3: Test experimentally.

– Repeat…

Derek Durand, with Prof. Paul Pringle, Dr Jason Lynam,

Dr Ruth Webster, Dr Alex Cresswell,

coming soon: Dr Thibault Cantat, Dr Jordi Carbo

Derek Durand

Jaguar 8.5, B3PW91-D3, 6-31G(d), 298 K, PCM (methanol)Derek Durand

Substrate (U) ∆∆G Oxidative Addition (kcal mol-1)Ethene 9.23Acetylene 4.90Carbon Monoxide 8.71

Substrate (U) ∆∆G Reductive Elimination (kcal mol-1)Ethene 5.29Acetylene 5.12

Carbon Monoxide 3.93 Derek Durand

Conclusions▪ Databases of descriptors (ligands, substrates, catalysts).

– Maps.

– Models (experimental and computational responses).

– Screening – design and optimisation.

▪ Mechanism.– Reasonable accuracy.

– Multiple catalysts.

– Selectivity and change in mechanism remain challenging.

▪ Now moving on to mapping reactivity – BRiCK, with experimental testing.

▪ Extreme value prediction.

We like data!

Acknowledgements▪ Derek Durand, Claire McMullin, Jesús

Jover, Gareth Owen-Smith, Ben Swallow, Mairi Haddow, Keshan Bolaky, Ben Rawe, Tim King

▪ Paul Pringle, Jeremy Harvey, Guy Orpen, Guy Lloyd-Jones

▪ Jason Lynam, Charlotte Willans, Bao Nguyen, Andrei Malkov, Jonathan Moseley, Simon Tyler, Paula Diaconescu, Ruth Webster, Alex Cresswell

▪ Bristol Centre for Computational Chemistry

(CLM)

(JJM)

University of York pump-priming fund, JGI Seedcorn Funding

bris.ac.uk/chemistry/facilities/basf

Physical Insights into Mechanistic Processes in Organometallic Chemistry

Faraday Discussion

2– 4 September 2019York, UK http://rsc.li/organometallic-fd2019

Upcoming deadlines

Early bird registration – 15 July 2019

Standard registration – 05 August 2019