Harnessing Organic Ligand Libraries for First-
Principles Inorganic Discovery: Indium Phosphide
Quantum Dot Precursor Design Strategies
The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.
CitationKim, Jeong Yun et al. "Harnessing Organic Ligand Libraries for
First-Principles Inorganic Discovery: Indium Phosphide Quantum
Dot Precursor Design Strategies." Chemistry of Materials 29, 8
(March 2017): 3632-3643 © 2017 American Chemical Society
As Publishedhttp://dx.doi.org/10.1021/acs.chemmater.7b00472
PublisherAmerican Chemical Society (ACS)
VersionAuthor's final manuscript
Citable linkhttps://hdl.handle.net/1721.1/123831
Terms of UseArticle is made available in accordance with the publisher's
policy and may be subject to US copyright law. Please refer to the
publisher's site for terms of use.
1
Harnessing Organic Ligand Libraries for First-
Principles Inorganic Discovery: Indium Phosphide
Quantum Dot Precursor Design Strategies
Jeong Yun Kim
1
, Adam H. Steeves
1
, and Heather J. Kulik
1,
*
1
Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA
02139
ABSTRACT: Indium phosphide quantum dots (QDs) represent promising replacements for more
toxic QDs, but InP QD production lags behind other QD materials due to limited understanding
of how to tune InP QD growth. We carry out a first-principles, computational screen of the
tuning of In carboxylate precursor chemistry to alter the kinetics of elementary steps in InP QD
growth. We employ a large database normally used for discovery of therapeutic drug-like
molecules to discover design rules for these inorganic complexes while maintaining realism (i.e.,
stable, synthetically accessible substituents) and providing diversity in a 210-molecule test set.
We show the In-O bond cleavage energy, which is tuned through ligand functionalization, to be a
useful proxy for In-P bond formation energetics in InP QD synthesis. Energy decomposition
analysis on a 32-molecule subset reveals that lower activation energies correlate to later
transition states, due to stabilization from greater In-P bond formation and more favorable
reaction energetics. Our simulations suggest that altering ligand nucleophilicity tunes the
reaction barrier over a 10 kcal/mol range, providing the conjugate acid’s pK
a
as an experimental
handle to lead to better control of growth conditions and to improve synthesized InP QD quality.
Importantly, these trends hold regardless of phosphorus precursor chemistries and in longer
chain length ligands typically used in synthesis.
1. Introduction
Colloidal quantum dots (QDs) have attracted intense attention for their unique size-
2
dependent electronic and optical properties
1-2
, relevant for applications in photovoltaics,
3-5
light-
emitting diodes,
6-7
and biological imaging.
8-9
Well-known cadmium selenide (CdSe)-based QDs
exhibit controllable sizes
10-11
and light emission throughout the entire visible range
12
, but the high
toxicity of cadmium
13
motivates the search for non-toxic replacements. Indium phosphide (InP)
QDs represent promising alternatives to CdSe-based QDs due to lack of intrinsic toxicity
14-15
and
broader emission profiles.
16-17
Current InP QD synthesis approaches have failed to obtain optimal
InP QD size distribution and high quantum yields comparable to CdSe-based QDs, despite
ongoing experimental efforts to understand the QD formation mechanism
18-23
and tune the growth
process.
24-25
Nucleation models that work for more ionic II-VI compounds fail for III-V QDs
26
.
Further, the need for high temperatures, reactive phosphorus precursors, and long reaction times
for the synthesis of more covalent InP QDs makes control of size distributions challenging.
27
First-principles simulations can provide valuable insight into the chemistry of quantum dots
and their precursors. Simulations have been widely used to understand and improve the growth
mechanism
28-29
of II-VI and IV-VI QDs (e.g., CdSe, PbSe and PbS) or understanding ligand
exchange in coinage metal QDs
30-33
. Although the properties of amorphous
34
, bulk
35-37
, and
confined InP
34, 38-41
have been studied in detail, less work has been carried out to understand their
growth mechanisms. Recently, we used a combined
ab initio
molecular dynamics (AIMD) and
reaction pathway analysis approach to discover the kinetics of early stage growth in InP QDs.
42
Through AIMD, we observed the formation of the earliest stage of InP QDs with an indium rich
surface in excellent structural agreement with recently characterized InP magic sized clusters
19, 43-44
.
We also identified indium carboxylate precursor chemistry (i.e., In-O bond dissociation
energetics) to play an essential role in determining kinetic barriers that necessitate high
temperatures in InP QD synthesis.
Despite this essential role of the indium precursor, focus has remained on the highly reactive
nature of common phosphorus precursors (e.g., tris(trimethylsilyl)-phosphine, (P(SiMe
3
)
3
)
employed for the growth of InP QDs, as they are believed to be the main cause of the poor size
QD distribution, in analogy to other classes of QDs.
14, 23
Indeed, experimental effort has been
3
aimed at identifying phosphorus precursors with slower conversion rate to prevent fast depletion
at high temperatures
24, 45-47
, but this approach has proven ineffective in improving QD size
distributions.
26
The challenges encountered in adjustment of group-V precursor chemistry to
enhance the quality of QDs
26
and earlier successes with improving yield through indium
precursor tuning
19, 48-49
motivate the reconsideration of altering indium precursor chemistry to tune
In-P bond formation. Six-coordinate indium carboxylates
50-52
are widely employed in InP QD
synthesis. Although carboxylate chain length is known to be important to synthetic outcomes
14, 48
and the form of the indium complex (i.e., phosphonate instead of carboxylate) has been shown to
produce altered properties
19
, subtle variation in ligand chemistry has not been widely pursued as a
mechanism to tune early stage growth kinetics.
In this work, we introduce a first-principles computational discovery approach to identify
design strategies for the chemistry of indium carboxylate precursors to tune early-stage InP QD
growth kinetics. We repurpose large libraries
53
and molecular similarity metrics, normally
employed in organic molecule design, in an inorganic discovery toolkit
54
to identify new indium
precursor candidates. A similar repurposing approach has very recently been demonstrated
55
in
experimental screening of transition metal catalysts. Through our computational approach, we
identify relationships between indium carboxylate chemistry and In-P bond formation energetics
that can help guide alternative recipes for InP QD synthesis. In Section 2, we summarize the
Computational Details used in our study. In the Results and Discussion (sec. 3), we present the
approach and results of our screen, a detailed analysis of the source of changes in energetics, and
provide strategies for implementing suggested design elements in experimental synthesis.
Finally, we summarize our Conclusions in section 4.
2. Computational Details
Density functional theory (DFT) single point energies and geometry optimizations were
carried out with the B3LYP
56-58
hybrid exchange-correlation functional with the VWN1-RPA form
for the LDA VWN
59
component of LYP
56
correlation, as implemented in TeraChem
60-61
. The
composite LACVP* basis set was used, which consists of the Los Alamos effective core
4
potential (LANL2DZ)
62
for indium atoms and the 6-31G* basis set for the remaining atoms.
Larger basis sets yielded comparable results on representative cases (Supporting Information
Table S1). Geometry optimizations used the L-BFGS algorithm in Cartesian coordinates with the
DL-FIND
63
interface to TeraChem. Default thresholds were used, which correspond to 4.5 × 10
-4
hartree/bohr for the maximum gradient and 1.0 × 10
-6
hartree for the change in total energy
between steps. All reported energies are obtained in the gas phase, as tests with implicit solvent
using the dielectric constant of solvents used in experimental conditions (ε=2) yielded
comparable results (Supporting Information Table S2). Initial structures for geometry
optimizations in the high-throughput screen were generated in both bidentate and monodentate
structures using a template (In(Ac)
3
) structure (Supporting Information Figure S1).
Using In-P bond formation pathways in the reaction of indium acetate precursors with
phosphine we obtained previously,
42
we generated initial guesses for intermediates and transition
states of modified precursors with the molSimplify toolkit.
54
We preserved the central core
reacting In, P, H, and surrounding carboxylate atoms, which acted as a template for the
placement of functionalized ligands. This structure building approach is the same as outlined for
our previous efforts in screening functionalized ferrocenium complexes.
64
Constrained
optimizations, with the central core atoms fixed, were carried out on these structures to obtain a
good initial guesses of transition states for subsequent transition state searches. For larger
phosphorus precursors (P(XH
3
)
3
, where X = Ge or Si), we first performed a nudged elastic band
(NEB) calculation
65-66
for P-X bond dissociation in TeraChem. The highest energy image from
NEB calculation was extracted for further refinement.
Transition states were obtained with partitioned rational function optimization (P-RFO)
67
at the B3LYP/LACVP* level of theory using Q-Chem 4.2
68
. All transition states were
characterized with vibrational frequency analysis to confirm a single imaginary frequency
corresponding to P-H bond dissociation. For intercomplex In-P bond formation, a second
imaginary frequency with zero intensity and a value of < 30i cm
-1
was typically detected, likely
due to soft modes in the weakly bound intercomplex structures. Activation energies (
E
a
) for the