You are on page 1of 12

INSTITUTE OF PHYSICS PUBLISHING NANOTECHNOLOGY

Nanotechnology 16 (2005) 2000–2011 doi:10.1088/0957-4484/16/10/004

The use of heat transfer fluids in the


synthesis of high-quality CdSe quantum
dots, core/shell quantum dots, and
quantum rods
Subashini Asokan1 , Karl M Krueger1 , Ammar Alkhawaldeh2,5 ,
Alessandra R Carreon2 , Zuze Mu2,6 , Vicki L Colvin1,2,3 ,
Nikos V Mantzaris2,3,4 and Michael S Wong1,2,3,7
1
Department of Chemistry, Rice University, USA
2
Department of Chemical and Biomolecular Engineering, Rice University, USA
3
Center for Biological and Environmental Nanotechnology, Rice University, USA
4
Department of Bioengineering, Rice University, USA

E-mail: mswong@rice.edu

Received 19 May 2005


Published 3 August 2005
Online at stacks.iop.org/Nano/16/2000
Abstract
Fluorescent semiconductor nanoparticles, or quantum dots, have potential
uses as an optical material, in which the optoelectronic properties can be
tuned precisely by particle size. Advances in chemical synthesis have led to
improvements in size and shape control, cost, and safety. A limiting step in
large-scale production is identified to be the raw materials cost, in which a
common synthesis solvent, octadecene, accounts for most of the materials
cost for a batch of CdSe quantum dots. Thus, less expensive solvents are
needed. In this paper, we identify heat transfer fluids, a class of organic
liquids commonly used in chemical process industries to transport heat
between unit operations, as alternative solvents for quantum dot synthesis.
We specifically show that two heat transfer fluids can be used successfully in
the synthesis of CdSe quantum dots with uniform particle sizes. We show
that the synthesis chemistry for CdSe/CdS core/shell quantum dots and
CdSe quantum rods can also be performed in heat transfer fluids. With the
aid of a population balance model, we interpret the effect of different HT
fluids on QD growth kinetics in terms of solvent effects, i.e., solvent
viscosity, CdSe bulk solubility in the solvent, and surface free energy.
(Some figures in this article are in colour only in the electronic version)

1. Introduction the impending requirement of large quantities of inexpensive


QDs, but published synthesis methods are optimized for
Fluorescent semiconductor nanoparticles (NPs) known as milligram amounts and commercially available QDs are
quantum dots (QDs) have generated scientific and commercial costly8,9,10 . Efforts to develop preparation methods that
interest as a nanomaterial due to their size-dependent are cost-effective, scalable, and environmentally friendly are
optoelectronic properties [1–7] and their potential uses in ongoing, such as replacing the metal precursors with greener
electronics, fluorescence imaging, and optical coding [8–11].
8
The anticipated advances in QD-based technologies point to http://www.evidenttech.com/. Evident Technologies, Inc.; one gram of
CdSe NPs = $6399 (specially quoted in January 2005).
5 Current address: Intel Corporation, Logic Technology Development, USA. 9 http://www.nn-labs.com/. NN-Labs, Inc.; one gram of CdSe NPs = $2100–
6 Current address: University of Houston, USA. $3375 (March 2005).
7 Author to whom any correspondence should be addressed. 10 http://www.qdots.com/. Quantum Dot Corporation.

0957-4484/05/102000+12$30.00 © 2005 IOP Publishing Ltd Printed in the UK 2000


The use of heat transfer fluids in the synthesis of high-quality CdSe QDs, core/shell QDs, and quantum rods

(a) (c)
+ O

O P

(d)

(b)

Figure 1. Molecular structures of (a) TOPO, (b) ODE, and the HT fluids (c) DTA and (d) terphenyl in T66.

Table 1. Organic solvents used in this study.


Synthesis solvent Melting point (◦ C) Boiling point (◦ C) Cost ($/l)
Trioctylphosphine oxide (TOPO) 52 >360 158
Octadecene (ODE) 18 315 23
Dowtherm A (DTA) 12 257 3
Therminol 66 (T66) 6 359 10

and safer compounds [12–14] and developing continuous-flow minimum of 2 min until the powder dissolved completely
reactors [15–19]. to form a clear and colourless trioctylphosphine selenide
As the most well-studied QD material, CdSe NPs (TOPSe) solution. The Cd precursor was prepared by
are formed from the rapid decomposition of Cd (e.g., combining 0.102 g of CdO (99.99%, Sigma-Aldrich), 40 ml of
dimethylcadmium [20] and cadmium complexes with a solvent (initial [Cd] = 20 mM), and 4 ml (99.99%, Sigma-
alkylphosphonic acids [12], alkylcarboxylic acids [21, 22], Aldrich) of oleic acid as stabilizer in a three-neck flask (mole
trioctylphosphine [23], and trioctylphosphine oxide [24]) ratio of oleic acid to Cd (oleic acid:Cd) = 50; Cd:Se = 2). A
and Se precursors (e.g., trioctylphosphine selenide and total of four different solvents were used: TOPO, ODE, DTA,
tributylphosphine selenide [25]) in an organic solvent at and T66 (figure 1, table 1). DTA (Dowtherm A, Dow Chemical
elevated temperatures (250–350 ◦ C). Trioctylphosphine oxide Co.) is a eutectic mixture of biphenyl (26.5 wt%) and diphenyl
(TOPO) is most commonly used as the solvent due to its high ether (or biphenyl oxide) (73.5 wt%), and T66 (Therminol 66,
boiling point and chemical stability, and its ability to prevent Solutia Inc.) is a mixture of terphenyl (∼8 mol%, CAS No
particle aggregation via coordination to the NP surface. TOPO [26140-60-3]), partially hydrogenated terphenyls (∼74 mol%,
can be replaced by octadecene (ODE), a non-coordinating [61788-32-7]), and higher polyphenyls/partially hydrogenated
solvent, if another compound such as oleic acid is used as the polyphenyls (∼18 mol%, [68956-74-1]) [26]. Their molecular
surface-binding ligand [22]. ODE provides several advantages structures are shown in figure 1.
over TOPO: its use allows for particle formation kinetics to be Under a constant argon purge, the magnetically stirred
controlled; it is a liquid at room temperature whereas TOPO is Cd–oleic acid-containing reaction mixture was heated slowly
a wax; and it is less expensive than TOPO. (2 ◦ C min−1 ) from room temperature to 250 ◦ C, the injection
Interestingly, ODE accounts for ∼90% of the materials temperature. When the rust-coloured Cd–oleic acid-
cost in preparing one batch of CdSe NPs (∼15 mg of CdSe), containing reaction mixture turned colourless and clear at
using the synthesis method published by Yu and Peng [22]. In 250 ◦ C, the entire aliquot of TOPSe solution was rapidly
this paper, we demonstrate that inexpensive organic liquids injected using a needled syringe (<2 s). The temperature
known as heat transfer fluids (HT fluids) can be used as of the reaction mixture immediately dropped to 240 ◦ C after
solvents for synthesizing QDs. With high boiling points and injection, and was allowed to drop to a steady-state growth
chemical stability at high temperatures, HT fluids are used to temperature of 220 ◦ C. One-millilitre aliquots of reaction
transport heat between unit processes found in the chemical mixture were removed at various times over the course of
process industries, but are generally not used as reaction 10 h (i.e., 10 s, 30 s, 1 min, 2 min, 5 min, 10 min, 15 min,
solvents [26, 27]. 30 min, 1 h and every half-hour afterwards), and deposited
in vials containing 10 ml of ODE (lowering the temperature
to 40 ◦ C in order to quench further particle growth). Final
2. Materials and methods
reaction volumes at the end of 10 h were 12, 12, 15, and 14 ml
2.1. Synthesis of CdSe QDs for TOPO, ODE, DTA, and T66, respectively. Higher injection
and growth temperatures of 300 and 280 ◦ C, respectively, were
A synthesis method of Peng and co-workers [22] was modified used to compare T66 with ODE. The solvent and unreacted
by using lower injection and growth temperatures, allowing precursors were removed through induced aggregation of
a comparison of the organic solvents to be made at constant the QDs with anhydrous methanol (99.8%, Sigma-Aldrich),
conditions. In a typical batch synthesis of CdSe QDs, a centrifugation and washing with methanol, and re-suspension
solution of TOPSe is freshly prepared in a glove bag under in 10 ml of anhydrous hexane (95%, Sigma-Aldrich). The
an argon gas purge. Selenium powder (32 mg, 99.999%, 10 h synthesis experiments for each of the four solvents
Sigma-Aldrich) was added to a vial containing 1.44 ml of were repeated at least three times. There were batch-to-batch
trioctylphosphine (TOP, 90%, Sigma-Aldrich). The vial was variations, as generally known for hot-injection QD synthesis
parafilm-sealed and the suspension was vortex-mixed for a routes [15, 17]. The values reported were somewhat different

2001
S Asokan et al

from the preliminary results reported earlier [28], but the 3. Results and discussion
general trends remained the same.
3.1. Synthesis of CdSe QDs
2.2. Synthesis of CdSe/CdS core/shell QDs The organic solvents used in the generic synthesis of NPs
The above procedure for synthesizing CdSe QDs (section 2.1) are much more varied than those used in the synthesis of
was modified to study the formation of CdSe/CdS core/shell CdSe QDs. Examples include diphenyl ether for PbSe and
QDs. The TOPSe solution was freshly prepared from Se Co NPs [29, 30] and CoPt3 NPs [31]; dioctyl ether for FePt
powder (32 mg) and TOP (1.5 ml). The Cd/oleic acid solution NPs [32] and γ -Fe2 O3 [33]; o-dichlorobenzene for disc-shaped
was prepared by combining CdO (0.102 g), T66 (40 ml, initial Co NPs [34, 35]; ODE for Fe3 O4 NPs [36]; eicosene for mixed-
[Cd] = 20 mM), and oleic acid (4 ml, oleic acid:Cd = 16; phase iron oxide NPs [37]; and squalene for CdSe QDs [15].
Cd:Se = 2). Under a constant argon purge, the magnetically T66 and DTA HT fluids used in this study are solvent mixtures,
stirred Cd–oleic acid-containing reaction mixture was heated and we are not aware of their use in NP synthesis prior to
slowly (2 ◦ C min−1 ) from room temperature to 250 ◦ C, turning this work.
clear at 220 ◦ C. The TOPSe solution was rapidly injected using CdSe QDs were found to form using the DTA and T66
a needled syringe (<2 s), and toluene (5 ml) was then added HT fluids. The particles grow in DTA and T66 with reaction
dropwise to lower the reaction temperature to 150 ◦ C. time, as shown by the red-shifting in the distinctive spectral
The cooled reaction mixture was split into two aliquots. peaks (figure 2). The fine features in the optical spectra
One aliquot was slowly heated to 220 ◦ C, to which additional are similar to those of QDs synthesized in ODE, indicating
Cd (2 ml, 0.04 M) and sulfur precursors (2 ml, 0.04 M) were that particle formation processes occurring in the HT fluids
injected sequentially. The Cd stock solution comprised 0.123 g are also similar to those occurring in ODE. Narrow band-
CdO, 20 ml T66, and 4 ml oleic acid, and the Se stock solution edge PL peaks and no deep trap emissions were observed
comprised 25.6 mg S and 20 ml T66. The reaction temperature from the fluorescence spectra. The QDs formed in the HT
was maintained at 220 ◦ C for 30 min, after which toluene (5 ml) fluids are approximately spherical, resembling those formed
was added dropwise to lower the temperature to 120 ◦ C. The in ODE and TOPO (figure 3). Many faceted NPs can be
other aliquot was used as a control; T66 solvent (4 ml) was observed in the TEM images, which we have found to be
added instead of the Cd and S precursors. a common occurrence and which have been observed by
others [22, 38, 39]. Though TOPO is typically used in
2.3. Synthesis of CdSe rod-shaped QDs conjunction with dimethylcadmium and not with Cd–oleic acid
complex [22], optical data (figures 2(d) and 3(d)) indicated that
The above procedure for synthesizing CdSe QDs (section 2.1) the latter Cd precursor could be used with TOPO to yield CdSe
was modified to study the formation of CdSe quantum QDs, consistent with the results of Qu et al [39].
rods. The Cd/oleic acid solution was prepared by combining CdSe QDs are usually synthesized at higher injection and
CdO (0.102 g), T66 (40 ml, initial [Cd] = 20 mM), and growth temperatures but the injection and growth temperatures
tetradecylphosphonic acid (TDPA, 0.893 g, 98%, Alfa Aesar; (250 and 220 ◦ C) used in this study were limited by the boiling
TDPA:Cd = 2; Cd:Se = 0.62), and heating to 320 ◦ C under
point of DTA. T66 has a boiling point comparable to that of
an Ar purge until the mixture became optically clear. The
TOPO and one that is higher than that of ODE, allowing for
solution was cooled to 300 ◦ C, and the Se precursor (100 mg
higher synthesis reaction temperatures to be used. At injection
Se and 2.4 ml TOP) was injected. The reaction temperature
and growth temperatures of 300 and 280 ◦ C, respectively,
was lowered and kept at 250 ◦ C for 30 min.
QDs formed in T66 were smaller than those formed in ODE
(figures 2(e) and (f)). This difference in growth rates was less
2.4. Characterization apparent at the lower synthesis temperatures.
UV–visible absorbance and fluorescence (or photolumines- The average particle diameters and concentrations were
cence, PL) data were acquired at room temperature using a estimated from the wavelength location and absorbance
Shimadzu UV–visible spectrophotometer (UV-2401PC model) of the first exciton peak maximum, respectively, using
and a SPEX fluorimeter (Fluoromax-3 model), respectively. the correlations published by Yu et al [40]. Clearly,
The PL spectra were collected between 400 and 750 nm using the QD particle sizes increased with synthesis time
an excitation wavelength of 380 nm. The absorbance at 380 nm (figures 4(a) and (c)). At long synthesis times,
of a given QD suspension was kept below 0.05 for quantum larger particle sizes were favoured by the solvents in
yield calculations; a methanol solution of Rhodamine 6G was the following order: TOPO > T66 > ODE ∼ DTA; at
used as the standard (quantum yield = 0.95). Transmission short synthesis times, the solvent order changed slightly:
electron microscopy (TEM) was performed on a JEOL-JEM TOPO > ODE > T66 > DTA. The average particle sizes of
2010 electron microscope operated at 200 kV, using fasTEM QDs produced in DTA, T66, ODE, and TOPO after 1 h were
software. Samples for TEM were prepared by depositing a 2.9, 3.1, 3.1, and 4.0 nm, respectively. These values were
drop of QD suspension dispersed in hexane onto 400-mesh consistently smaller than the particle sizes derived from TEM
copper grids covered with a thin amorphous carbon film, which (in the same order): 3.5, 3.9, 4.5, and 5.2 nm (figure 3). This
was followed by vacuum desiccation. Particle size analysis of could well be an artefact of our TEM image analysis in which
TEM images was carried out using either Image-Pro Plus (Ver- smaller particles are more difficult to resolve visually, leading
sion 5.0, Mediacybernetics) or ImageJ (Version 1.33, NIH) to a bias particle size distribution. In the same order, the
software, on 1000 particles per sample. For rod-shaped QDs, relative standard deviations for the various QDs were 18%,
100 particles were analysed. 17%, 15%, and 13%. Values as small as 5% have been reported

2002
The use of heat transfer fluids in the synthesis of high-quality CdSe QDs, core/shell QDs, and quantum rods

(a) (b) (c)

Absorbance/PL Intensity (AU)

400 500 600 700 400 500 600 700 400 500 600 700
Wavelength (nm) Wavelength (nm) Wavelength (nm)

(d) (e) (f)


Absorbance/PL Intensity (AU)

Increasing
time

400 500 600 700 400 500 600 700 400 500 600 700
Wavelength (nm) Wavelength (nm) Wavelength (nm)

Figure 2. UV–visible absorbance and fluorescence spectra of CdSe QDs synthesized in (a) DTA, (b) T66, (c) ODE, and (d) TOPO, as a
function of time (injection at 250 ◦ C and growth at 220 ◦ C). Each set of spectra was collected at 10 s, 1 min, 5 min, 30 min, 1 h, 3 h, and 5 h
after injection of TOPSe solution. Corresponding spectra for QDs synthesized in (e) T66 and (f) ODE at higher temperatures (injection
at 300 ◦ C and growth at 280 ◦ C). Each set of spectra was collected at 5 min, 15 min, 30 min, 1 h, and 2 h after injection of TOPSe solution.

for CdSe QDs, and the size distributions were attributed to emission peak (data not shown). This decrease is attributed to a
uncertainty in TEM measurements to within one lattice plane focusing of the particle size distributions, arising from the rapid
(∼0.18 nm) [20]. Our wider size distributions could, too, be growth of all particles in the presence of high concentrations
an artefact of TEM analysis. The particle concentrations were of Cd and Se monomers [1, 38, 41, 42]. We do not make
comparable among the different samples, allowing for proper the distinction between ‘precursor’ (e.g., CdO:oleic acid and
comparison of the particle sizes at similar reaction conversions TOPSe complexes) and ‘monomer’ (reactive intermediate
(figures 4(b) and (d)). TOPO led to the formation of larger formed from the precursor, which leads to nucleation or adds
particles compared to the other solvents, suggesting that TOPO to a growing particle). We observed an increase in FWHM
(with its ability to coordinate to the CdSe particle surface that after ∼2 h of synthesis time in DTA and T66, marking the
the other solvents lack) can increase the reactivity of the Cd onset of particle size distribution broadening or defocusing.
precursor. This is consistent with previously observed Cd– Defocusing occurred after 1 and 1.5 h using TOPO and
TOPO complex formation [24]. ODE solvents, respectively (data not shown). To achieve
CdSe QDs synthesized at short times in the different the same small particle size (<3 nm), longer times were
solvents were compared at a common size (table 2). The full needed with HT fluids than with TOPO and ODE (figure 4(a)).
width at half-maximum (FWHM) indicates the sharpness of The difference in growth kinetics is clearly observed, with
the PL emission, and indirectly, the particle size distribution. the results indicating that particle growth can be slowed
Thus, TOPO- and ODE-derived QDs have lower fluorescence down and that smaller particles with equal or higher quality
(and particle size distribution) quality than those synthesized (e.g., comparable UV–visible absorbance spectra, comparable
using HT fluids. This trend may be correlated with the particle size distributions, and narrower PL peaks) can be
synthesis time. The FWHM values for QDs prepared in all four synthesized using HT fluids (figure 5).
solvents tend to decrease with synthesis time in the first hour, For all solvents, the quantum yields of CdSe QDs
indicating that the longer synthesis times help to sharpen the fluctuated significantly over the course of 10 h (data not

2003
S Asokan et al

(a) (b)

(c) (d)

Figure 3. TEM images of CdSe QDs synthesized for 1 h in (a) DTA, (b) T66, (c) ODE, and (d) TOPO solvents (injection at 250 ◦ C and
growth at 220 ◦ C). Scale bars: 20 nm.

Table 2. A comparison of ∼2.7 nm CdSe QDs synthesized at the presence of Cd and S precursor at 220 ◦ C, which resulted
220 ◦ C using the different organic solvents. in the red-shifting of the absorbance and PL spectra. This
Particle red-shift is similar to earlier observations made by Alivisatos
Synthesis diameter FWHM of PL Quantum Synthesis and co-workers [44]. The QDs grew in size to 4.17 nm (peak
solvent (nm) peak (nm) yield (%) time (s) maximum = 590 nm), indicating a shell thickness of 0.65 nm
TOPO 2.65 33 15.9 10 or ∼2 monolayers of CdS (thickness of a monolayer shell
ODE 2.64 34 11.7 60 of CdS ∼ 0.35 nm [44]). The formation of a CdS shell
DTA 2.69 24 3.3 900 corresponded to the observed increase in quantum yield, a
T66 2.69 27 8.4 300 hallmark of core/shell QDs [44–46]. For comparison, the
same CdSe starting material was heated in the absence of
the Cd and S precursors. The optical spectra of the resultant
shown). This observation contrasts with that for the QDs QDs indicated particle growth, due to the presence of free Cd
prepared by Talapin et al using a different method, in which and Se precursors which remained from the original synthesis
quantum yields dropped with increasing QD particle size or (figure 6(c)). These particles grew to the same size as
synthesis time [43]. No extra care was taken to exclude oxygen the core/shell QDs but had a lower quantum yield, further
and moisture during the washing process or the spectroscopy indicating CdSe/CdS core/shell QD formation in HT fluids
measurements, which probably contributed to the observed (figure 6(b)).
quantum yield fluctuations. Quantum yield differences could Rod-shaped CdSe QDs, or quantum rods (QRs), were also
also result from surface structure differences, as several found to be successfully prepared in HT fluids. Researchers
research groups have concluded that photoluminescence is have demonstrated that shape anisotropy could be introduced
very sensitive to surface structure and surface defects [21, 42]. through the use of alkylphosphonic acids [47–49] or different
Cd–ligand complexes (ligand = naphthenic acid [50] or
3.2. Synthesis of core/shell and rod-shaped CdSe QDs alkylphosphonic acid [51]) in TOPO and ODE solvents. Here,
QRs were prepared using TDPA and T66 and the average width
Core/shell QD synthesis using HT fluids was successfully and length were measured to be 5.0 and 15.7 nm, respectively
carried out (figure 6). A suspension of CdSe QDs (peak (figures 7(a) and (b)). The absorbance and fluorescence spectra
maximum = 541 nm → diameter = 2.87 nm) was heated in were similar to previous reports of optical properties of CdSe

2004
The use of heat transfer fluids in the synthesis of high-quality CdSe QDs, core/shell QDs, and quantum rods

5.0 1.6E-04
(a) (b)
4.5 TOPO
1.4E-04
ODE
4.0
QD Particle Diameter (nm)

1.2E-04 DTA

QD Concentration (M)
3.5 T66
3.0 1.0E-04

2.5 8.0E-05
2.0
6.0E-05
1.5 TOPO
ODE 4.0E-05
1.0
DTA
0.5 2.0E-05
T66
0.0 0.0E+00
0.001 0.01 0.1 1 10 0.001 0.01 0.1 1 10
Time (hr) Time (hr)
5.0 1.6E-04
(c) (d)
4.5 TOPO
1.4E-04
4.0 ODE
QD Particle Diameter (nm)

1.2E-04 DTA
3.5 QD Concentration (M)
T66
3.0 1.0E-04

2.5 8.0E-05
2.0
6.0E-05
1.5 TOPO
ODE 4.0E-05
1.0
DTA
0.5 2.0E-05
T66
0.0 0.0E+00
0 2 4 6 8 10 0 2 4 6 8 10
Time (hr) Time (hr)

Figure 4. (a) Average particle diameter and (b) particle concentration plotted against synthesis time (log scale), calculated from UV–visible
absorbance data for CdSe QDs synthesized in the four different organic solvents (injection at 250 ◦ C and growth at 220 ◦ C). ((c), (d)) The
same data plotted on a linear scale.

QRs (figure 7(c)) [47, 52]. The QRs exhibited a Stokes shift of the growth of all QDs in the synthesis medium in the absence
∼30 nm (=660−629 nm) or ∼93 meV, which is as expected of Ostwald ripening. The particle size distributions became
larger than that for CdSe QDs. For example, QDs prepared narrower in this stage, on the basis of the FWHM decrease
in T66 for 30 min (figure 2(c)) have a Stokes shift of ∼13 nm (not shown). In the second stage (between 0.5 and 3 h), the
(=563 − 550 nm) or ∼52 meV. The quantum yield of the QRs particle concentration decreased by ∼67% for all solvents,
was measured to be 0.8%. while the average diameter continued to increase at a smaller
rate than in the first stage. Finally, in the third stage, the particle
3.3. Computational analysis of the solvent effect on CdSe QD concentration and average diameter–time profiles became flat.
growth kinetics Using Se instead of Cd as the basis for calculation since it was
the limiting reagent (Cd:Se = 2), we calculated the percentage
The particles were significantly larger using TOPO than the of TOPSe that was converted into nanoparticulate form. On the
other solvents, and the particles synthesized in T66, DTA, and basis of the average CdSe QD particle size and concentration
ODE were close in size. The differences among the diameter values at the end of 10 h, nearly all the Se reacted to form
size curves indicated that particle growth depends on the nature QDs in TOPO (∼99%), whereas 33%, 34%, and 27% of Se
of the solvent, e.g., molecular structure or physical property. reacted to form QDs in DTA, T66, and ODE, respectively.
As discussed earlier, TOPO appears to increase the reactivity Steady-state average particle sizes were reached even though
of CdO–oleic acid precursor to form the reactive monomer TOPSe precursor was in abundance, suggesting the CdSe QDs
needed for nucleation and growth, and it can be reasoned that reached concentration equilibrium with the remaining TOPSe
T66, DTA, and ODE do not increase the Cd precursor reactivity in the latter solvents. We note that the durations of each of the
to the same extent. three stages were virtually identical for all four solvents, and
We observed that the particle concentration and particle larger particle growth rates were favoured by the solvents in
average diameter–time profiles exhibited three stages over the the following order: TOPO > T66 > ODE ∼ DTA.
course of 10 h (figure 4). To a first approximation, particle In order to obtain insight into these experimental findings
concentrations remained constant while the average diameter we utilized the population balance model recently developed
increased in the first stage (<0.5 h). This can be attributed to by one of us to describe the dynamics of QD growth [53].

2005
S Asokan et al

(a) (b)

Absorbance (AU)

PL Intensity(AU)
350 400 450 500 550 600 650 700 350 400 450 500 550 600 650 700

(c) (d)

350 400 450 500 550 600 650 700 350 400 450 500 550 600 650 700
Wavelength (nm)

Figure 5. UV–visible absorbance and fluorescence spectra of ∼2.7 nm CdSe QDs synthesized in (a) DTA, (b) T66, (c) ODE, and (d) TOPO
(injection at 250 ◦ C and growth at 220 ◦ C).

properties especially if these values are not known (e.g., surface


free energy and monomer solubility in a given solvent).
We briefly present the basic mathematical formulation
(see [53] for details). The time evolution of the particle
Quantum
Absorbance/PL intensity (AU)

yields size distribution is given by the following first-order partial


differential equation:
(c) 4.3% ∂ N (r, t) ∂
+ [ṙ (r, Cb )N (r, t)] = 0. (1)
∂t ∂r
N (r, t) dr denotes the number of particles which have radii
between r and r + dr at time t; ṙ (r, Cb ) represents the single-
(b) 11.2%
particle growth rate, which depends on the particle radius r ,
and the bulk monomer concentration Cb . Here, the monomer
refers to TOPSe, and the equation assumes that particles do
not interact or aggregate.
(a) To derive the single-particle growth rate ṙ (r, Cb ) first-
6.2%
order kinetics (with a rate constant ks ) for the surface
400 450 500 550 600 650 700
reaction between the monomer and the particle was assumed.
Figure 6. UV–visible absorbance and fluorescence spectra of CdSe Moreover, the monomer concentration at equilibrium (Ce ) with
QDs (a) before heating, (b) after heating at 220 ◦ C for 30 min in the CdSe particles with the particle radius r was expressed as a
presence of Cd and S precursors, and (c) after heating at the same function of the CdSe molar volume Vm (32.995 mol cm−3 ),
conditions but without the shell precursors.
the surface free energy γ , the temperature T , and the bulk
CdSe solubility (C∞ ) using the Gibbs–Thompson equation
(Ce = C∞ exp[ 2γ Vm
r RT
]). Finally, monomer molecules diffuse
One key advantage offered by this deterministic framework from the bulk to the surface of the particles through a diffusion
over stochastic models (such as the Monte Carlo simulations layer. This is taken to be a pseudo-steady-state process as
of Talapin et al [54]) is the use of dimensionless parameters. was also assumed by Talapin et al [54] and Sugimoto [55].
This allows one to easily analyse the effects of key parameters The thickness of the diffusion layer was assumed to be much
on the asymptotic and transient NP growth behaviour without bigger than the particle radius as was also argued by Peng
having to specify the various synthesis parameters and physical and Peng [49]. These assumptions allow the derivation of the

2006
The use of heat transfer fluids in the synthesis of high-quality CdSe QDs, core/shell QDs, and quantum rods

(a)

(b)
25 35

30
20
25
15
20
count

count
10 15

10
5
5

0 0
3 4 5 6 7 8 10 12 14 16 18 20 22 24
width (nm) length (nm)

(c)
Absorbance (AU)

PL Intensity (AU)

400 450 500 550 600 650 700 750 800


Wavelength (nm)

Figure 7. (a) TEM image, (b) size histograms, and (c) optical spectra of CdSe QRs synthesized in T66 for 30 min (injection at 300 ◦ C and
growth at 250 ◦ C). Scale bar: 50 nm.

following closed-form expression that relates the growth rate To simplify the mathematical formulation and identify
law with the particle radius and the monomer concentration in key systemic parameters, we non-dimensionalized the model
the bulk: (equations (1)–(3)) with respect to both radius and time, which
dr Cb − C∞ exp[ 2γ Vm
] yielded two dimensionless parameters: χ and Da. Parameter
ṙ (r, Cb ) ≡ = Vm ks k r
r RT
. (2) χ largely controls transient behaviour and is defined as
dt 1+ D s

Lastly, mass conservation requires that the total monomer S0


χ= 4 π
 rmax (4)
amount existing either as free monomer or as monomer S0 + 3 Vm C∞ 0 r 3 N0 (r ) dr
incorporated in the NPs remains constant and equal to the total
initial monomer amount in either form. Thus, for constant where S0 = Cb,0C−C∞

is the initial supersaturation. Ranging
reactor volume, we have from 0 to 1, parameter χ quantifies the magnitude of the initial
 rmax
4π 1 monomer concentration in the form of free excess monomer
Cb,0 + r 3 N0 (r ) dr = Cb (t)
3 Vm 0 (monomer concentration above bulk CdSe solubility) relative
 rmax
4π 1 to the total monomer concentration that exists in the form of
+ r 3 N (r, t) dr (3)
3 Vm 0 both free excess monomer and particles.
where Cb,0 is the initial monomer concentration in the bulk and The dimensionless Damköhler number Da quantifies the
N0 (r ) is the initial particle size distribution. magnitude of surface reaction of monomer relative to monomer

2007
S Asokan et al

diffusion, and is given as 1.5


(a)
ks 2γ Vm
Da = · . (5) 1.4
D RT
The system is surface reaction-limited for Da = 0, and is 1.3

<r>/<ro>
monomer diffusion-limited for Da → ∞. Mantzaris [53]
showed that the model was in excellent agreement with the 1.2
analytical solutions for the asymptotic, normalized particle size
distribution in the diffusion—(Da → ∞) and reaction-limited 1.1
(Da = 0) cases, as derived by Lifshitz and Slyozov [56] and
Wagner [57], respectively. We used this computational model 1
to elucidate the salient features of the QD growth process 0 1 2 3 4 5
through comparison with our experimental data. dimensionless time
The particle size distribution evolution behaviour is fully 5
1 10
characterized by the initial conditions S0 and N0 (r ) and the
(b)
two dimensionless parameters Da and χ. To illustrate the

particle concentration
4
typical behaviour predicted by the population balance model, 9 10
we chose the following dimensionless parameter values as
nominal: Da = 4, χ = 0.6, S0 = 10, and N0 (r ) = 8 10
4

Gaussian distribution. As shown in figure 8(a) (solid line) the


average particle size exhibits a fast increase before levelling 4
off at ∼135% of its initial value, during which time the 7 10

particle concentration is unchanged (figure 8(b)). Above


the dimensionless time τ (=t/( RT2γ k s C∞
)) of ∼2, particle 6 10
4

0 1 2 3 4 5
concentration decreases and particle size increases due to
dimensionless time
Ostwald ripening (larger particles grow at the expense of
smaller ones). The particle size distribution exhibits focusing– 1.6 (c)
defocusing behaviour, in which the minimum standard
deviation is located at τ = 0.3, before Ostwald ripening 1.4
occurs (figure 8(c)). The model predictions for the dynamics
of the average radius and QD concentration are in qualitative
sd/sdo

1.2
agreement with the experimental data for the first two stages
of QD growth (compare with figure 4). The steady-state 1
third stage is not predicted by the model, however, as
the model indicates further particle growth and decreasing 0.8
QD concentration (beyond τ = 5; simulation data not
shown). To explain this discrepancy, we suggest that another 0.6
0 1 2 3 4 5
particle formation mechanism besides growth and Ostwald
dimensionless time
ripening is responsible for stabilizing the QD particle size and
concentration at long synthesis times. Nevertheless, the model Figure 8. Simulation results for the dynamics of: (a) the average
appears to capture the dynamics well for the first two stages of particle radius normalized by the initial average particle radius,
the process. (b) the total number of particles, and (c) the standard deviation of
the particle size distribution divided by the initial standard deviation.
With this observation in mind, we then used the
Solid lines: nominal case (Da = 4, χ = 0.6, S0 = 10). Dotted
computational model to interpret the experimentally observed lines: (1/2) × nominal viscosity (Da = 2, χ = 0.6, S0 = 10).
differences in the QD growth kinetics for T66 and DTA Dashed lines: (1/2) × nominal viscosity, (1/2) × nominal
solvents. To establish a basis for our comparison, we assumed solubility, and 1.5 × nominal surface free energy (Da = 3,
the base case (solid lines, figure 8) to represent particle χ = 0.318, S0 = 21).
formation in T66 (figure 4). The quantifiable difference
between T66 and DTA was in the solvent viscosity µs , and
thus we were interested in understanding its effect on the With an estimated radius of 1 nm for TOPSe, we calculated
growth kinetics. Viscosity is related to molecular precursor DT66 = 5×10−10 m2 s−1 and DDTA = 1×10−9 m2 s−1 . Thus,
diffusivity in the solvent via the Stokes–Einstein equation the Da number for DTA is half that of T66.
D = 6π µkBsrTmono (where kB is the Boltzmann constant and rmono The dotted lines in figure 8 show the QD growth kinetics
is the precursor radius), which indicates that the monomer obtained for half the Da number of the nominal case. The
species diffuses more slowly in more viscous solvents than average radius reaches a plateau value close to that of the
in less viscous ones. We found the viscosities of T66 and nominal case, reaches this plateau slightly faster, and persists
DTA to be 0.72 and 0.35 cP at 220 ◦ C using an HT fluid for a shorter period of time. However, this particle growth
physical properties program (Therminol Reference Disk11 ). behaviour does not match the experimental results for DTA
(figures 4(a) and (c)) as the average particle size in DTA is
11 http://www.therminol.com/pages/. Solutia Inc. clearly smaller than that of T66. The model predicts also that

2008
The use of heat transfer fluids in the synthesis of high-quality CdSe QDs, core/shell QDs, and quantum rods

1.25 4. Conclusions

1.2 HT fluids are shown as viable alternatives to the solvents


currently used in the synthesis of high-quality CdSe QDs,
1.15 which will help to eliminate cost as one barrier to large-scale
production of QDs and nanoparticles of other compositions.
FWHM / FWHMo

1.1 CdSe quantum rods and CdSe/CdS core/shell QDs can also be
prepared in the HT fluids. QDs with smaller diameters can be
1.05 prepared more controllably using T66 and DTA compared to
TOPO or octadecene, indicating that QD synthesis chemistry in
1 these fluids can be optimized to yield higher-quality materials
in the smaller nanometre size range. Only TOPO leads to
0.95
near complete conversion of TOPSe into CdSe QDs. Detailed
simulation studies strongly suggest that the experimentally
0.9
observed differences in the QD growth kinetics in different HT
fluids are the combined result of solvent-dependent differences
0.85
0 0.5 1 1.5 2 in viscosity, CdSe bulk solubility, and surface free energy.
time (hr) Further study is needed to reconcile the disagreement between
Figure 9. Normalized FWHM (by initial FWHM value)–time experimental data and the growth model during the last
profiles for CdSe QD growth (injection at 250 ◦ C and growth stage of QD growth. The use of different HT fluids could
at 220 ◦ C) in T66 (solid line–open circles, lower curve) and DTA lead to predictive modulation of the QD growth kinetics, if
(dashed line–filled circles, upper curve). CdSe solubility and surface free energy values at elevated
temperatures are known.

the QD concentration drops significantly faster for the less


Acknowledgments
viscous solvent (figure 8(b)), which contrasts with the slightly
higher QD concentrations in DTA than in T66 (figures 4(b)
We gratefully acknowledge Solutia Inc. and Dow Chemical
and (d)).
Co. for providing the HT fluids, Dr Satish Mohapatra
These discrepancies between model and experimental
(Dynalene) for helpful discussions, Dr Pradeep Rai (Rice) for
results indicate that solvent viscosity differences cannot fully
solvent viscosity measurements, and NSF (EEC-0118007) for
explain the experimentally observed patterns. It is thus
financial support.
reasonable to consider other possibilities resulting from the
use of different solvents. Specifically, solvents could have
different bulk CdSe solubilities C∞ . Also, it is reasonable References
to consider that the QD surface is not completely coated by
oleate ligands and that the free energy of the QD surface γ [1] Rogach A L, Talapin D V, Shevchenko E V, Kornowski A,
Haase M and Weller H 2002 Organization of matter on
could be modified by interactions with solvent molecules at the
different size scales: monodisperse nanocrystals and their
solid–liquid interface. So, we suggest that the experimentally superstructures Adv. Funct. Mater. 12 653–64
observed differences in the QD growth kinetics for the different [2] Liz-Marzan L M et al 2001 New aspects of nanocrystal
HT fluids result from the combined differences in their research Mater. Res. Soc. Bull. 26 981–4
corresponding µs , C∞ , and γ values. Through exhaustive [3] Trindade T, O’Brien P and Pickett N L 2001 Nanocrystalline
numerical simulations, we found that a lower viscosity, smaller semiconductors: synthesis, properties, and perspectives
Chem. Mater. 13 3843–58
solubility, and higher surface free energy led to a lower plateau [4] Grieve K, Mulvaney P and Grieser F 2000 Synthesis and
in the average particle radius and higher particle concentration electronic properties of semiconductor
(figures 8(a) and (b)), qualitatively matching the experimental nanoparticles/quantum dots Curr. Opin. Colloid Interface
profiles for DTA much better (figure 4). Sci. 5 168–72
This model predicts a less pronounced focusing stage [5] Efros A L and Rosen M 2000 The electronic structure of
semiconductor nanocrystals Annu. Rev. Mater. Sci. 30
and a broader particle size distribution relative to the base 475–521
case (figure 8(c)). The FWHM data associated with figure 4 [6] Alivisatos A P 1996 Perspectives on the physical chemistry of
were low quality unfortunately and are not presented; so we semiconductor nanocrystals J. Phys. Chem. 100 13226–39
re-synthesized CdSe QDs in T66 and DTA, and carefully [7] Weller H 1993 Colloidal semiconductor
measured the FWHM values. Due to batch-to-batch variations, q-particles—chemistry in the transition region between
solid-state and molecules Angew. Chem. Int. Edn Engl. 32
the actual particle sizes and concentrations did not exactly
41–53
correspond to those shown in figure 4 and defocusing occurred [8] Ridley B A, Nivi B and Jacobson J M 1999 All-inorganic field
at ∼5 min for these samples instead of the ∼2 h indicated effect transistors fabricated by printing Science 286 746–9
for figure 4. Still, we found that the experimentally observed [9] Dubertret B, Skourides P, Norris D J, Noireaux V,
FWHM values matched the predicted particle size distribution Brivanlou A H and Libchaber A 2002 In vivo imaging of
trends (figure 9). Specifically, with T66 as the base case (solid quantum dots encapsulated in phospholipid micelles
Science 298 1759–62
lines, figures 8 and 9), the FWHM data for the DTA case [10] Han M Y, Gao X H, Su J Z and Nie S 2001
(dashed line, figure 9(c)) qualitatively matched the standard Quantum-dot-tagged microbeads for multiplexed optical
deviation curve predicted for DTA (dashed line, figure 8(c)). coding of biomolecules Nat. Biotechnol. 19 631–5

2009
S Asokan et al

[11] Klimov V I, Mikhailovsky A A, Xu S, Malko A, [32] Sun S H, Murray C B, Weller D, Folks L and Moser A 2000
Hollingsworth J A, Leatherdale C A, Eisler H J and Monodisperse FePt nanoparticles and ferromagnetic FePt
Bawendi M G 2000 Optical gain and stimulated emission in nanocrystal superlattices Science 287 1989–92
nanocrystal quantum dots Science 290 314–7 [33] Hyeon T, Lee S S, Park J, Chung Y and Bin Na H 2001
[12] Peng Z A and Peng X G 2001 Formation of high-quality CdTe, Synthesis of highly crystalline and monodisperse
CdSe and CdS nanocrystals using CdO as precursor J. Am. maghemite nanocrystallites without a size-selection process
Chem. Soc. 123 183–4 J. Am. Chem. Soc. 123 12798–801
[13] Peng X G 2002 Green chemical approaches toward [34] Puntes V F, Krishnan K M and Alivisatos A P 2001 Colloidal
high-quality semiconductor nanocrystals Chem.—Eur. J. 8 nanocrystal shape and size control: the case of cobalt
335–9 Science 291 2115–7
[14] Malik M A, Revaprasadu N and O’Brien P 2001 Air-stable [35] Puntes V F, Zanchet D, Erdonmez C K and Alivisatos A P
single-source precursors for the synthesis of chalcogenide 2002 Synthesis of hcp-Co nanodisks J. Am. Chem. Soc. 124
semiconductor nanoparticles Chem. Mater. 13 913–20 12874–80
[15] Yen B K H, Stott N E, Jensen K F and Bawendi M G 2003 A [36] Jana N R, Chen Y F and Peng X G 2004 Size- and
continuous-flow microcapillary reactor for the preparation shape-controlled magnetic (Cr, Mn, Fe, Co, Ni) oxide
of a size series of CdSe nanocrystals Adv. Mater. 15 nanocrystals via a simple and general approach Chem.
1858–62 Mater. 16 3931–5
[16] Edel J B, Fortt R, deMello J C and deMello A J 2002 [37] Park J, An K J, Hwang Y S, Park J G, Noh H J, Kim J Y,
Microfluidic routes to the controlled production of Park J H, Hwang N M and Hyeon T 2004 Ultra-large-scale
nanoparticles Chem. Commun. 1136–7 syntheses of monodisperse nanocrystals Nat. Mater. 3
[17] Chan E M, Mathies R A and Alivisatos A P 2003 891–5
Size-controlled growth of CdSe nanocrystals in microfluidic [38] Peng X G, Wickham J and Alivisatos A P 1998 Kinetics of
reactors Nano Lett. 3 199–201 II–VI and III–V colloidal semiconductor nanocrystal
[18] Kawa M, Morii H, Ioku A, Saita S and Okuyama K 2003 growth: ‘Focusing’ of size distributions J. Am. Chem. Soc.
Large-scale production of CdSe nanocrystal by a 120 5343–4
continuous flow reactor J. Nanopart. Res. 5 81–5 [39] Qu L H, Peng Z A and Peng X G 2001 Alternative routes
[19] Shestopalov I, Tice J D and Ismagilov R F 2004 Multi-step toward high quality CdSe nanocrystals Nano Lett. 1
synthesis of nanoparticles performed on millisecond time 333–7
scale in a microfluidic droplet-based system Lab Chip 4 [40] Yu W W, Qu L H, Guo W Z and Peng X G 2003 Experimental
316–21 determination of the extinction coefficient of CdTe, CdSe,
[20] Murray C B, Norris D J and Bawendi M G 1993 Synthesis and and CdS nanocrystals Chem. Mater. 15 2854–60
characterization of nearly monodisperse CdE [41] Bullen C R and Mulvaney P 2004 Nucleation and growth
(E = S, Se, Te) semiconductor nanocrystallites J. Am. kinetics of CdSe nanocrystals in octadecene Nano Lett. 4
Chem. Soc. 115 8706–15 2303–7
[21] Qu L H and Peng X G 2002 Control of photoluminescence [42] Talapin D V, Rogach A L, Shevchenko E V, Kornowski A,
properties of CdSe nanocrystals in growth J. Am. Chem. Haase M and Weller H 2002 Dynamic distribution of
Soc. 124 2049–55 growth rates within the ensembles of colloidal II–VI and
[22] Yu M W and Peng X G 2002 Formation of high-quality CdS III–V semiconductor nanocrystals as a factor governing
and other II–VI semiconductor nanocrystals in their photoluminescence efficiency J. Am. Chem. Soc. 124
noncoordinating solvents: tunable reactivity of monomers 5782–90
Angew. Chem. Int. Edn Engl. 41 2368–71 [43] Talapin D V, Rogach A L, Kornowski A, Haase M and
[23] Mekis I, Talapin D V, Kornowski A, Haase M and Weller H 2001 Highly luminescent monodisperse CdSe and
Weller H 2003 One-pot synthesis of highly luminescent CdSe/ZnS nanocrystals synthesized in a
CdSe/CdS core–shell nanocrystals via organometallic and hexadecylamine–trioctylphosphine oxide–trioctylphospine
‘greener’ chemical approaches J. Phys. Chem. B 107 mixture Nano Lett. 1 207–11
7454–62 [44] Peng X, Schlamp M C, Kadavanich A V and
[24] Wong M S and Stucky G D 2002 The facile synthesis of Alivisatos A P 1997 Epitaxial growth of highly luminescent
nanocrystalline semiconductor quantum dots Mater. Res. CdSe/CdS core/shell nanocrystals with photostability and
Soc. Symp. Proc. 676 Y2.3–6 electronic accessibility J. Am. Chem. Soc. 119 7019–29
[25] Katari J E B, Colvin V L and Alivisatos A P 1994 X-ray [45] Hines M A and Guyot-Sionnest P 1996 Synthesis and
photoelectron spectroscopy of CdSe nanocrystals with characterization of strongly luminescing ZnS-capped CdSe
applications to studies of the nanocrystal surface J. Phys. nanocrystals J. Phys. Chem. 100 468–71
Chem. 98 4109–17 [46] Dabbousi B O, Rodriguez-Veijo J, Mikulec F V, Heine J R,
[26] Singh J 1985 Heat Transfer Fluids and Systems for Process Mattoussi H, Ober R, Jensen K F and Bawendi M G 1997
and Energy Applications (New York: Dekker) (CdSe)ZnS core–shell quantum dots: synthesis and
[27] Green R L, Larsen A H and Pauls A C 1989 The heat-transfer characterization of a size series of highly luminescent
fluid spectrum Chem. Eng. 2 90–8 nanocrystallites J. Phys. Chem. B 101 9463–75
[28] Asokan S, Carreon A R, Mu Z, Krueger K M, Alkhawaldeh A, [47] Peng X G, Manna L, Yang W D, Wickham J, Scher E,
Colvin V L and Wong M S 2005 Synthesis of high-quality Kadavanich A and Alivisatos A P 2000 Shape control of
CdSe nanocrystals in heat transfer fluids Nanobiophotonics CdSe nanocrystals Nature 404 59–61
and Biomedical Applications II ed A N Cartwright and [48] Manna L, Scher E C and Alivisatos A P 2000 Synthesis of
M Osinski (Bellingham, WA: SPIE) pp 60–7 soluble and processable rod-, arrow-, teardrop-, and
[29] Murray C B, Sun S, Gaschler W, Doyle H, Betley T A and tetrapod-shaped CdSe nanocrystals J. Am. Chem. Soc. 122
Kagan C R 2001 Colloidal synthesis of nanocrystals and 12700–6
nanocrystals superlattices IBM J. Res. Dev. 45 47–56 [49] Peng Z A and Peng X G 2001 Mechanisms of the shape
[30] Sun S H and Murray C B 1999 Synthesis of monodisperse evolution of CdSe nanocrystals J. Am. Chem. Soc. 123
cobalt nanocrystals and their assembly into magnetic 1389–95
superlattices (invited) J. Appl. Phys. 85 4325–30 [50] Nann T and Riegler J 2002 Monodisperse CdSe nanorods at
[31] Shevchenko E V, Talapin D V, Rogach A L, Kornowski A, low temperatures Chem.—Eur. J. 8 4791–5
Haase M and Weller H 2002 Colloidal synthesis and [51] Peng Z A and Peng X G 2002 Nearly monodisperse and
self-assembly of CoPt3 nanocrystals J. Am. Chem. Soc. 124 shape-controlled CdSe nanocrystals via alternative routes:
11480–5 nucleation and growth J. Am. Chem. Soc. 124 3343–53

2010
The use of heat transfer fluids in the synthesis of high-quality CdSe QDs, core/shell QDs, and quantum rods

[52] Hu J T, Li L S, Yang W D, Manna L, Wang L W and [55] Sugimoto T 1987 Preparation of monodispersed colloidal
Alivisatos A P 2001 Linearly polarized emission from particles Adv. Colloid Interface Sci. 28 65–108
colloidal semiconductor quantum rods Science 292 2060–3
[56] Lifshitz I M and Slyozov V V 1961 The kinetics of
[53] Mantzaris N V 2005 Liquid-phase synthesis of nanoparticles:
particle size distribution dynamics and control Chem. Eng. precipitation from supersaturated solid solutions J. Phys.
Sci. 60 4749–70 Chem. Solids 19 35–50
[54] Talapin D V, Rogach A L, Haase M and Weller H 2001 [57] Wagner C V 1961 Theorie der Alterung von Niederschlägen
Evolution of an ensemble of nanoparticles in a colloidal durch Umlösen (Ostwald-Reifung) Z. Electrochem. 65
solution: theoretical study J. Phys. Chem. B 105 12278–85 581–91

2011