Probing Homogeneous Line Broadening in CdSe Nanocrystals Using

Letter pubs.acs.org/NanoLett

Probing Homogeneous Line Broadening in CdSe Nanocrystals Using Multidimensional Electronic Spectroscopy Tobias A. Gellen, Jet Lem, and Daniel B. Turner* Department of Chemistry, New York University, 100 Washington Square East, New York, New York 10003, United States S Supporting Information *

ABSTRACT: The finite spectral line width of an ensemble of CdSe nanocrystals arises from size and shape inhomogeneity and the single-nanocrystal spectrum itself. This line width directly limits the performance of nanocrystal-based devices, yet most optical measurements cannot resolve the underlying contributions. We use two-dimensional electronic spectroscopy (2D ES) to measure the line width of the band-edge exciton of CdSe nanocrystals as a function of radii and surface chemistry. We find that the homogeneous width decreases for increasing nanocrystal radius and that surface chemistry plays a critical role in controlling this line width. To explore the hypothesis that unpassivated trap states serve to broaden the homogeneous line width and to explain its size-dependence, we use 3D ES to identify the spectral signatures of exciton−phonon coupling to optical and acoustic phonons. We find enhanced coupling to optical phonon modes for nanocrystals that lack electron-passivating ligands, suggesting that localized surface charges enhance exciton−phonon coupling via the Fröhlich interaction. Lastly, the data reveal that spectral diffusion contributes negligibly to the homogeneous line width on subnanosecond time scales. KEYWORDS: Semiconductor nanocrystals, homogeneous line broadening, two-dimensional electronic spectroscopy, phasing, electron−phonon coupling, beating maps

T

dependent exciton−phonon coupling.18 As stated by Kambhampati and co-workers, “Any effort to quantify excitonphonon coupling should preferably detect both the optical and acoustic modes”.14 In a recent report, researchers used a technique to probe the spectral information on single nanocrystals in their native environment at fast time scales with short exposure times without user-selection bias and with ensemble-level statistics.3 This study used photoluminescence to examine the singlenanocrystal line width of CdSe nanoparticles as a function of core radius, shell thickness, and composition.19 The authors concluded that the surface chemistry of the nanocrystal has a profound effect on the internal electric fields, coupling the band-edge exciton to the longitudinal optical (LO) phonon and broadening the homogeneous line width. In addition, the authors stated that bare CdSe cores are not appropriate for studies of exciton−phonon coupling because of the significant influence of trapped states. Here, we report a complementary investigation of the homogeneous width of CdSe nanocrystals using two-dimensional electronic spectroscopy (2D ES) as a function of nanocrystal radius (1.4−3.0 nm) and surface chemistry. In addition, we use an extension of this technique, 3D ES, to access “beating maps” of both acoustic and optical modes.

he zero-dimensional semiconductor nanocrystal, colloquially known as a “quantum dot”, is characterized by confinement of the electronic wave functions in all spatial directions. In such cases of full, 3D confinement, the electronic density of states is nonzero only at discrete energy values, leading one to anticipate extremely narrow spectral lines like those of atomic gases.1 However, even single-nanocrystal spectra are significantly broadened by homogeneous sources of line broadening, primarily spectral diffusion, exciton fine structure, and exciton−phonon coupling.2,3 In addition to these homogeneous sources, optical line widths of ensembles of colloidal semiconductor nanocrystals are broadened by size and shape inhomogeneity, an inherent consequence of their “wet” synthesis.4 The convolution of these effects, the ensemble spectral line width, directly limits the performance of nanocrystal-based applications such as gain in lasers, resolution in bioimaging, and speed in quantum information processing.5 In order for nanocrystals to fulfill their technological potential, we must understand the microscopic origin of line broadening. In pursuit of this objective over the past two decades, researchers have extensively studied the line width of CdSe nanocrystals using ensemble time−domain techniques such as pump−probe, photon echo, and hole burning measurements6−14 and single-particle photoluminescence spectroscopy.15−17 These techniques have succeeded in measuring static inhomogeneity, reorganization energy, Stokes shift, and phonon frequencies. These techniques have been less successful at extracting the homogeneous width and quantifying the size© 2017 American Chemical Society

Received: December 6, 2016 Revised: April 17, 2017 Published: April 19, 2017 2809

DOI: 10.1021/acs.nanolett.6b05068 Nano Lett. 2017, 17, 2809−2815

Letter

Nano Letters The 2D spectroscopy excels at separating homogeneous and inhomogeneous line widths in spectrally congested systems such as nanocrystal ensembles.20 Peaks in a 2D spectrum are characterized by diagonal and antidiagonal line widths and these values are intimately related to the physical processes of homogeneous and inhomogeneous line broadening. While the diagonal represents a convolution of both homogeneous and inhomogeneous widths, the antidiagonal is only affected by homogeneous broadening.21 In the absence of inhomogeneity, the peak is perfectly circular. In 2011, researchers from the Cundiff group extracted these line width values for GaAs quantum wells by fitting slices of a 2D spectrum with Gaussian inhomogeneity and Lorentzian homogeneity.22 Very recently, the Jonas group demonstrated that the slope of the nodal line separating positive and negative peaks in a 2D spectrum singularly determines bandgap inhomogeneity in PbSe nanocrystals.23 More generally, 2D ES provides a direct, time-resolved probe of the absorption line shape. This four-wave mixing technique requires four electric fields typically provided by femtosecond laser pulses.24,25 The pulse sequences of rephasing and nonrephasing are needed to resolve the true line width.26 Several studies of CdSe nanoparticles using 2D ES27−30 have already appeared, although none have extracted size-dependent properties nor studied the mechanisms underlying homogeneous line broadening. To quantify the size-dependence of the line width, we measured a total of 16 ensembles of nearly monodisperse CdSe semiconductor nanocrystallites using 2D ES. About half of these samples are coated with ligands that passivate hole trap states while the other half has both electron and hole passivating ligands. These two families also differ by crystal structure; the first family (only hole ligands) is zinc-blende while the second (both electron and hole) is wurtzite. These samples were dispersed in hexanes at room temperature and the spectra were collected at a waiting time of 1 ps using 6.8 fs laser pulses. Supporting Information contains the absorption and PL spectra, 2D spectra, and details regarding the fitting procedure. Although many peaks appear in each 2D spectrum, this report primarily focuses on the band-edge exciton peak, denoted |X1⟩ throughout the document. Figure 1 shows the extracted homogeneous line width of the | X1⟩ peak for both families of CdSe nanocrystals. The salient features in Figure 1 are a monotonically decreasing line width as particle radius increases and a vertical shift in energy between the two families of nanocrystals. The remainder of the report describes the use of 2D ES to yield insight about the microscopic origin of these features. The primary contributions to the homogeneous line width are spectral diffusion, exciton fine-structure, and exciton− phonon coupling. Fluctuations in the local electric fields due to migrating charges cause each nanocrystal to experience a unique “instantaneous frequency” that evolves in time.15,31−33 This process, known as spectral diffusion, is typically photoinduced and observed at cryogenic temperature. Previous reports have discounted the contribution of spectral diffusion at room temperature on nanosecond to millisecond time scales.3 To eliminate the possibility of spectral diffusion on a faster time scale, we conducted 2D ES measurements over four decades of waiting time. We monitored the homogeneous line width from 75 fs to 900 ps for one sample from each nanocrystal family, and Figure 2 shows that the line width remains constant on this

Figure 1. Homogeneous line width of the band-edge exciton peak for both families of CdSe nanocrystals suspended in hexanes at room temperature. Error bars arise from 95% confidence intervals. Both crystal structures display a similar functional dependence on nanocrystal radius but differ by a vertical offset due to distinct surface chemistry.

time scale. This finding demonstrates that the homogeneous widths in Figure 1 are not affected by spectral diffusion.

Figure 2. Homogeneous width versus waiting time for zinc-blende (blue, 1.7 nm radius) and wurtzite (red, 3.0 nm radius) nanocrystals. Both traces show that the line width is constant from 75 fs to 900 ps, demonstrating that spectral diffusion does not homogeneously broaden the line width.

The unit cell of a wurtzite crystal is hexagonal and the intrinsic asymmetry of the crystal lattice distorts the p-type hole, splitting the energy levels of the band-edge exciton.34 This fine-structure, which serves to broaden the line shape by 5−8 meV at room temperature,19,35 is absent in zinc-blende nanocrystals that present a cubic unit cell. Because both crystal structures exhibit a similar functional dependence on radius, fine-structure alone cannot explain the common trends observed in Figure 1. By process of elimination, we arrive at exciton−phonon coupling as the underlying source of the sizedependent trends in Figure 1. Next, we consider the sizedependence of the interactions underlying acoustic and optical phonons and then we identify the spectral signatures of exciton−phonon coupling. The most prominent and widely studied phonon in CdSe is the LO phonon, and the magnitude of its coupling to the exciton is given by the Huang−Rhys factor 2810

DOI: 10.1021/acs.nanolett.6b05068 Nano Lett. 2017, 17, 2809−2815

Letter

Nano Letters S=

f02 R3

∑ qln

2 1 |νλλ(qln)| qln2 (ℏωLO)2

passivate electron traps, strongly binds to surface Cd atoms.50 TOP passivates the hole traps on the CdSe surface by attaching to uncoordinated Se sites.51 Oleylamine, a fatty amine, also serves as a hole trap passivator in CdSe nanocrystals. In addition, both families draw their source of cadmium from a Ztype ligand, cadmium oleate for wurtzite and cadmium myristate for zinc-blende nanocrystals.52 In summary, the ligands on the surface of wurtzite nanocrystals passivate both electron and hole trap states while the zinc-blende ligands only passivate hole trap states. These differences in surface chemistry are especially pronounced in smaller nanocrystals where the surface-to-volume ratio is substantial. We observe evidence of this effect as the widths in Figure 1 appear to converge for increasing nanocrystal radius. Because larger nanocrystals exhibit increasingly bulklike character, differences in surface chemistry become less relevant for determining the homogeneous width, thus both families should exhibit comparable line widths in this limit. Unpassivated trap states lead to localized surface charges that induce a polarization in the nanocrystal.14,53,54 Because the LO phonon in CdSe couples to the excitons through a Fröhlich interaction, these couplings are highly sensitive to internal electric fields. We hypothesize that unpassivated electron traps enhance coupling to optical phonons in zinc-blende nanocrystals and consequently broaden the homogeneous line width. To assess this possibility, we performed 3D ES to identify the spectral locations where coherent phonons cause peak amplitude changes. These spectral signatures only arise when phonons are coupled to the optical excitation, which is necessary for any phonon to influence the line width. To identify these phonons, we first obtained a one-dimensional vibrational spectrum of the coherent dynamics by integrating the 3D ES in both ω1 and ω3 dimensions around the |X1⟩ peak. Figure 3 contains two representative 1D vibrational spectra and

(1)

where f 0 is the Fröhlich interaction coupling constant, R is the nanocrystal radius, ℏωLO is the energy of the LO phonon, νλλ is the phonon wave function, and the qln are the normal phonon coordinates.36 Because the size dependence of the wave function is weak, the Huang−Rhys factor is proportional to R−3.37 In contrast to the optical phonons, the acoustic modes couple via the deformation and piezoelectric potential.14 The size dependence of these interactions arise from ∇·u, where u is the lattice displacement vector. For piezoelectric coupling, this 1 1 1 1 quantity can be estimated as ∇·u ∝ R 3 ω ∼ R2 . For the case R

of deformation-potential coupling, a similar approach yields ∇· 1 1 1 u ∝R3 R2 R2 ∼ R .37 To assess whether these theoretical dependencies on R are consistent with the trends observed in Figure 1, we fit each homogeneous line width data set to a power law, Γ(R) = bR−n. The vertical intercept, b, accounts for contributions to the line width that are independent of R, such as ligand identity. The exponent, n, controls the size-dependent scaling of the line width. While the fits are in good agreement with the experimental data, shown as dashed curves in Figure 1, the sublinear dependence (n ≈ 0.6) is inconsistent with the theoretical dependences on R as predicted by the Huang−Rhys factor and ∇·u. We hypothesize that the prolate shape of the crysallites, usually with an aspect ratio of about 1.1−1.3, is responsible for these deviations. By violating the assumption of a perfectly radial distribution of charge, the exciton couples to surface modes and l ≠ 0 LO modes.13,38,39 Now we address the distinct vertical offset observed in Figure 1 of about 15 meV difference between the two families of nanocrystals. As described previously, fine-structure serves to broaden the line shape by 5−8 meV in wurtzite nanocrystals. The data show a difference of approximately 15 meV between red and blue traces, but this observation is inconsistent with differences due to crystal structure because the fine structure of the band-edge exciton should increase the wurtzite line width relative to zinc-blende; instead, the wurtzite line width is decreased. To explain the observed disparity, we consider differences in surface chemistry between these samples. The surface of the nanocrystal is capped with ligands that play a pivotal role in altering properties like growth rate, shape, size, electron transfer reactions, and crystal structure.40−45 Unpassivated surface atoms give rise to discrete energy levels that lie within the band gap, capable of trapping optically excited electrons and holes during relaxation. Ligand coordination to the nanocrystal surface is described in terms of the number of electrons that a ligand uses to form bonds with the nanocrystal surface.46 L-type ligands are twoelectron donors such as amines that interact with Lewis-acid surface sites. X-type ligands form a covalent bond by donating one electron. These anionic species such as RCOO− can compensate excess metal charge on the nanocrystal surface.47 Z-type ligands are two-electron acceptors that interact with electron-donating atoms on the nanocrystal surface.48 The measured zinc-blende nanocrystals are capped with oleic acid, a fatty carboxylic acid X-type ligand that acts as a hole trap passivator in CdSe.49 In contrast, the measured wurtzite nanocrystals are passivated with TOP, TOPO, and oleylamine, all of which are L-type ligands. TOPO, which serves to

Figure 3. One-dimensional vibrational spectra of coherent phonon oscillations for zinc-blende (blue, 1.7 nm radius) and wurtzite (red, 1.9 nm radius) nanocrystals. Low-frequency acoustic modes are present in both samples at similar magnitudes but high-energy optical modes are enhanced in the nanocrystals that lack electron passivating ligands.

Supporting Information contains another set from two distinct samples. The highest-amplitude features of the spectra appear at 7, 21, 170, 185, and 214 cm−1 for zinc-blende. These wavenumbers correspond to the ellipsoidal (l = 2) acoustic phonon, longitudinal acoustic (LA) phonon, the transverse optical (TO) phonon, the “surface” optical (SO) phonon, and the LO phonon, respectively.9,11,12,14,55−58 We find enhanced coupling to optical phonons in zinc-blende 1D vibrational spectra as compared to those of wurtzite. The SO and TO phonons are largely absent in the wurtzite 1D spectrum, and 2811

DOI: 10.1021/acs.nanolett.6b05068 Nano Lett. 2017, 17, 2809−2815

Letter

Nano Letters

mode. Figures 4 and (5) reveal significant coupling between the |X2⟩ diagonal peak and the LO phonon. In contrast, the SO phonon exhibits negligible coupling to |X2⟩ but substantial coupling to |X3⟩. The common trend between the 3D ES shown in the bottom two panels of Figure 4 are significant amplitude modulation of the below-diagonal cross peak between |X1⟩ and a higher lying exciton. The amplitude of the above-diagonal cross peak between these excitons is modulated significantly less strongly, implying that optical phonons play a minor role in downhill energy relaxation. In both Figures 4 and 5, the acoustic phonons at 7 and 21 cm−1 have the highest amplitude in the above-diagonal region of the 3D ES. The presence of above-diagonal peaks indicates that these modes couple higher-energy excitons to lower-energy excitons, implying that these phonons mediate downhill energyrelaxation processes. The 2D spectra reveal one additional finding regarding the coupling of exciton states. Cross peaks are expected between | X1⟩ and |X2⟩ because these two excitons share a common conduction band, the 1Se. Cross peaks are not expected between |X3⟩ and either of the other two excitons because |X3⟩ arises from p-like electron and hole states whereas |X2⟩ and |X1⟩ arise from s-like electron and hole states. Yet, in all of the 2D spectra there are cross peaks among |X3⟩ and the other two excitons. Whereas the upper-diagonal cross peak can arise from conventional downhill electronic relaxation, the lower-diagonal cannot. Specifically, ground-state bleach and stimulated emission pathways can result in either type of cross peak, but they require coupling between |X3⟩ and the lower energy exciton. We include all 35 possible double-sided Feynman pathways in the Supporting Information. Each diagram represents one of the possible pathways for generating signal under the experimental conditions. These diagrams show that cross peaks cannot arise without coupling. One microscopic mechanism that could lead to these unexpected cross peaks involves Coulombic interactions that induce a many-body correlation among the charged particles. Alternatively, spin− orbit coupling mixes the s and p states, yielding excitonic basis states that must be described as linear combinations of the single-particle basis states. These explanations require further study by experiments and calculations. We have examined the two-dimensional electronic spectra of CdSe nanocrystals across a range of sizes. Fits to the 2D spectra revealed that the homogeneous line width of the band-edge exciton decreases monotonically as nanocrystal radius increases and the combination of TOP, TOPO, and oleylamine narrows the homogeneous line width by 15 meV relative to oleic acid. This disparity confirms that a combination of hole and electron trap passivating ligands (TOP, TOPO, oleylamine) narrows the homogeneous line width as compared to the exclusive use of hole trap ligands (oleic acid). This finding suggests that fully passivated nanocrystals are useful for optoelectronic devices that benefit from narrow homogeneous line widths. In addition, our results demonstrate that the small contribution to the homogeneous line width from fine-structure is significantly overshadowed by face passivation. We had anticipated that a comparison between zinc-blende and wurtzite structures would reveal the subtle fine-structure contribution to the homogeneous width, but the data show that any attempt at quantifying the fine-structure must control for surface chemistry. Our attempt at exchanging the ligands of the oleate-capped zincblende nanocrystals with pyridine was unsuccessful due to the formation of metal oleates, as shown in FT−IR spectra

the LO phonon peak is relatively suppressed. In contrast to the optical phonons, acoustic phonons are similarly coupled to |X1⟩ in wurtzite and zinc-blende nanocrystals. These findings support the hypothesis that localized surface charges enhance coupling to optical phonons in zinc-blende nanocrystals due to the absence of electron trap passivation. We attempted to quantify the size-dependent exciton−phonon coupling strength by comparing the amplitude of the LO phonon to its first overtone, following previous resonance Raman measurements.59 Although we observed the overtone using a resonance Raman microscope, we were unable to reproduce this signal using 3D ES, suggesting that the these techniques have slightly different selection rules. To study other effects of these phonons, we inspect slices of the 3D ES, which are often called “beating maps”. A 3D ES slice shows the features of the 2D ES that oscillate at a given frequency, and they have been extracted experimentally for GaAs quantum wells60 and potassium vapors61 and theoretically for CdSe nanocrystals.62 The 2D ES shown in the upper-left panel of Figure 4 reveals |X1⟩, |X2⟩, and |X3⟩ diagonal peaks, as

Figure 4. The 2D and 3D ES of 2.9 nm radius zinc-blende nanocrystals. (Top, left) The 2D spectrum at a waiting time of 1 ps. (Top, right) Slice of 3D spectrum at LA phonon frequency. (Bottom, left) Slice of 3D spectrum at SO phonon frequency. (Bottom, right) Slice of 3D spectrum at LO phonon frequency.

well as cross peaks among these excitons. Figure 5 reveals the same analysis on a distinct sample and at a lower-frequency

Figure 5. The 2D and 3D ES of 2.6 nm radius zinc-blende nanocrystals. (Left) The 2D spectrum at a waiting time of 1 ps. (Right) Slice of 3D spectrum at ellipsoidal acoustic phonon frequency. 2812

DOI: 10.1021/acs.nanolett.6b05068 Nano Lett. 2017, 17, 2809−2815

Letter

Nano Letters ϕ(ω3) = c0 + c1ω3 + c 2ω32 + ... + ckω3k

included in Supporting Information, with electrostatic binding. To displace a surface carboxylate, it must be neutralized and the surface charge must be balanced; however, displacing the carboxylate is not possible with pyridine. Any future attempt at extracting the fine-structure can overcome this obstacle by coating the particles with ZnS or CdS shells.19 To explore the effects of a polarization induced by coupling to trapped electron states, we used 3D ES to study “beating maps” of the following five prominent phonon modes in CdSe: the ellipsoidal (l = 2) acoustic phonon, the longitudinal acoustic (LA) phonon, the transverse optical (TO) phonon, the “surface” optical (SO) phonon, and the LO phonon. The 3D spectra revealed that nanocrystals lacking electron-trap passivating ligands exhibit enhanced coupling to optical phonons, as predicted by the underlying Fröhlich interaction. In addition, we observe a stronger modulation of abovediagonal features in 3D slices at acoustic frequencies whereas below-diagonal features are amplified in slices at optical frequencies. Lastly, we discounted the contributions of spectral diffusion to the homogeneous line width on subnanosecond time scales by demonstrating a persistent line width from 75 fs to 900 ps. Experimental Methods. Femtosecond Laser System. We have previously described the laser and spectrometer used to acquire transient absorption and 2D spectra.63,64 Briefly, a Ti:sapphire laser system yields near-IR pulses that pump a home-built noncollinear optical parametric amplifier. This output propagates through chirped mirrors and a pulse shaper, which compress the pulse to 6.8 fs as measured by the transient-grating frequency resolved optical gating method.65 The pulse spectrum was identical to that used in a recent publication.66 To suppress noise that would otherwise contaminate the third-order signal, we incorporate balanced detection with dual-beam chopping.66 We collected each 2D electronic spectrum by scanning time interval τ1 to 70 fs in 1 fs steps. We performed both rephasing and nonrephasing scans, and then computed their sum, the total correlation spectrum.26 We collected 3D spectra by scanning and Fourier transforming the waiting time interval, τ2, to 12 ps in 25 fs steps. To resolve the phonons, we fit the τ2 evolution of the 2D magnitude spectrum to a biexponential decay for each (ω1, ω3) grid point independently. We subtracted these decay traces to yield a purely oscillatory residual. Fourier transformation of this residual yielded a spectrum of coherent lattice oscillations. The power in the pump beam for transient absorption yielded on average one photon absorbed per two nanocrystals, ⟨N⟩ = 0.5. The total power of the three excitation beams used in the 2D measurements also yielded ⟨N⟩ = 0.5. This power is low enough to prevent significant biexcitonic contributions.14,56 The optical density was below 0.3. The detected 2D spectrum does not accurately reflect the absolute phase of the signal because of (1) our inability to determine absolute zero timing between the input pulses (τ1 = τ2 = 0), (2) error in setting τ3 = 0, (3) chirp on the laser pulses, and (4) phase error introduced by slight optical imbalances in the interferometer arms.26,67 To correct for these intrinsic issues, we “phase” the 2D correlation spectrum in the ω3 dimension. According to the projection-slice theorem for 2D Fourier transforms, the projection of the 2D correlation spectrum at a particular τ2 delay is equal to the pump−probe signal measured at the same delay. We achieve this equality by multiplying the detected 2D spectrum by exp(iϕ), where

(2)

We determine coefficients ci by minimizing a fitness function equal to the sum of the squared difference between the normalized projection of the 2D correlation signal onto the ω3 axis and the normalized pump−probe spectrum. In general, we find that k = 3 is necessary for a residual sum of squares below 1%, and we routinely achieve a residual below 0.1%. Supporting Information contains all 2D spectra and phasing results used in Figure 1. In spectrally congested systems involving multiple transitions with overlapping features, it can be challenging to extract the line width associated with a particular exciton. To report the homogeneous line width, we fit the 2D correlation spectrum to the sum of n Gaussian functions. For example, the region within the dashed rectangle in the left panel of Figure 5 contains the following four prominent peaks: |X1⟩ and |X2⟩ diagonal peaks and cross peaks between |X1⟩ and |X2⟩. The reconstructed 2D spectrum consists of four 2D, 45°-rotated Gaussian functions, one for each of the peaks in the 2D correlation spectrum. For the ith Gaussian Pi(ω3 , ω1) = A exp(− C+(ω3 − a3)2 + 2C −(ω3 − a3)(ω1 − a1) − C+(ω1 − a1)2 )

(3)

where

C± =

1⎛ 1 1 ⎞ ⎜ 2 ± 2⎟ 4 ⎝ σ1 σ3 ⎠

(4)

There are a total of five parameters for each Gaussian: { a3, a1, σ3, σ1, A }. The subscripts refer to excitation (1) and emission (3) frequency axes. The parameter σ defines the widths, A defines the amplitude, and a defines the intercepts. We determine these values by minimizing a fitness function equal to the sum of the squared difference between the 2D correlation spectrum and the reconstructed 2D spectrum composed of the sum of n Gaussians. In most cases the reconstructed 2D spectrum recovers more than 99.9% of the measured 2D spectrum. Supporting Information contains examples of reconstructed spectra and additional details of the fitting procedure. Sample Preparation and Characterization. We synthesized samples of zinc-blende CdSe nanocrystals from 1.4−2.9 nm in radius following the protocol of Cao et al.68 This approach is a noninjection synthesis, based on controlling the thermodynamics and kinetics in the nanocrystal nucleation stage. In a typical synthesis, SeO2 and cadmium myristate are added to a three-neck flask with 1-octadecene. The mixture is heated with stirring to 240 °C. After 3 min, oleic acid is added dropwise. The reaction mixture is cooled to room temperature. Larger nanocrystals are synthesized by maintaining the mixture at the reaction temperature for 20−45 min. Nanocrystals are washed with acetone and suspended in hexanes. We synthesized wurtzite CdSe nanocrystals from 1.6−3.0 nm in radius according to a hot-injection synthesis.69 Briefly, cadmium oleate and trioctylphosphine (TOP) oxide are added to a three-neck flask with 1-octadecene. At 300 °C, oleylamine and TOP/Se are quickly injected. For larger nanocrystals, a mixture of cadmium oleate and TOP/Se is reinjected. After 3 min, the solution is cooled to room temperature. Nanocrystals are washed with ethanol and suspended in hexanes. To characterize these samples, we measured the linear absorption and fluorescence with a Cary 100 Bio UV/vis and a 2813

DOI: 10.1021/acs.nanolett.6b05068 Nano Lett. 2017, 17, 2809−2815

Letter

Nano Letters

(15) Empedocles, S. A.; Bawendi, M. G. J. Phys. Chem. B 1999, 103, 1826−1830. (16) Moerner, W. E.; Orrit, M. Science 1999, 283, 1670−1676. (17) Moerner, W. E.; Fromm, D. P. Rev. Sci. Instrum. 2003, 74, 3597−3619. (18) Kelley, A. M. J. Phys. Chem. Lett. 2010, 1, 1296−1300. (19) Cui, J.; Beyler, A. P.; Coropceanu, I.; Cleary, L.; Avila, T. R.; Chen, Y.; Cordero, J. M.; Heathcote, S. L.; Harris, D. K.; Chen, O.; Cao, J.; Bawendi, M. G. Nano Lett. 2016, 16, 289−296. (20) Jonas, D. M. Annu. Rev. Phys. Chem. 2003, 54, 425−463. (21) Hamm, P.; Zanni, M. T. Concepts and Methods of 2D Infrared Spectroscopy; Cambridge University Press, 2011. (22) Bristow, A. D.; Zhang, T.; Siemens, M. E.; Cundiff, S. T.; Mirin, R. P. J. Phys. Chem. B 2011, 115, 5365−5371. (23) Park, S. D.; Baranov, D.; Ryu, J.; Cho, B.; Halder, A.; Seifert, S.; Vajda, S.; Jonas, D. M. Nano Lett. 2017, 17, 762−771. (24) Turner, D. B.; Arpin, P. C.; McClure, S. D.; Ulness, D. J.; Scholes, G. D. Nat. Commun. 2013, 4, 2298. (25) Brańczyk, A. M.; Turner, D. B.; Scholes, G. D. Ann. Phys. (Berlin, Ger.) 2014, 526, 31−49. (26) Khalil, M.; Demirdöven, N.; Tokmakoff, A. Phys. Rev. Lett. 2003, 90, 047401. (27) Cassette, E.; Pensack, R. D.; Mahler, B.; Scholes, G. D. Nat. Commun. 2015, 6, 6086. (28) Lenngren, N.; Abdellah, M. A.; Zheng, K.; Al-Marri, M. J.; Zigmantas, D.; Zidek, K.; Pullerits, T. Phys. Chem. Chem. Phys. 2016, 18, 26199−26204. (29) Caram, J. R.; Zheng, H.; Dahlberg, P. D.; Rolczynski, B. S.; Griffin, G. B.; Fidler, A. F.; Dolzhnikov, D. S.; Talapin, D. V.; Engel, G. S. J. Phys. Chem. Lett. 2014, 5, 196−204. (30) Caram, J. R.; Zheng, H.; Dahlberg, P. D.; Rolczynski, B. S.; Griffin, G. B.; Dolzhnikov, D. S.; Talapin, D. V.; Engel, G. S. J. Chem. Phys. 2014, 140, 084701. (31) Roberts, S. T.; Loparo, J. J.; Tokmakoff, A. J. Chem. Phys. 2006, 125, 084502. (32) Empedocles, S. A.; Bawendi, M. G. Science 1997, 278, 2114− 2117. (33) Empedocles, S. A.; Neuhauser, R.; Shimizu, K.; Bawendi, M. G. Adv. Mater. 1999, 11, 1243−1256. (34) Wong, C. Y.; Scholes, G. D. J. Lumin. 2011, 131, 366−374. (35) Moreels, I.; Raino, G.; Gomes, R.; Hens, Z.; Stoferie, T.; Mahrt, R. F. ACS Nano 2011, 5, 8033−8039. (36) Nomura, S.; Kobayashi, T. Phys. Rev. B: Condens. Matter Mater. Phys. 1992, 45, 1305−1316. (37) Takagahara, T. Phys. Rev. Lett. 1993, 71, 3577−3580. (38) Klein, M. C.; Hache, F.; Ricard, D.; Flytzanis, C. Phys. Rev. B: Condens. Matter Mater. Phys. 1990, 42, 11123−11132. (39) Fedorov, A. V.; Baranov, A. V.; Inoue, K. Phys. Rev. B: Condens. Matter Mater. Phys. 1997, 56, 7491−7502. (40) Wang, W.; Banerjee, S.; Jia, S.; Steigerwald, M. L.; Herman, I. P. Chem. Mater. 2007, 19, 2573−2580. (41) Li, Z.; Peng, X. J. Am. Chem. Soc. 2011, 133, 6578−6586. (42) Gao, J.; Johnson, J. C. ACS Nano 2012, 6, 3292−3303. (43) Garcia-Rodriguez, R.; Liu, H. J. Am. Chem. Soc. 2014, 136, 1968−1975. (44) Yu, W. W.; Wang, Y. A.; Peng, X. Chem. Mater. 2003, 15, 4300− 4308. (45) Hines, D. A.; Kamat, P. V. J. Phys. Chem. C 2013, 117, 14418− 14426. (46) Green, M. J. Organomet. Chem. 1995, 500, 127−148. (47) Kim, D.; Kim, D.-H.; Lee, J.-H.; Grossman, J. C. Phys. Rev. Lett. 2013, 110, 196802. (48) Houtepen, A. J.; Hens, Z.; Owen, J. S.; Infante, I. Chem. Mater. 2017, 29, 752−761. (49) Grandhi, G. K.; Arunkamar, M.; Viswanatha, R. J. Phys. Chem. C 2016, 120, 19785−19795. (50) Rafeletos, G.; Nørager, S.; O’Brien, P. J. Mater. Chem. 2001, 11, 2542−2544.

Horiba PTI QM400, respectively. We determined the average radius in a given batch by an empirical fitting function as described by Peng et al.70 Supporting Information contains linear absorption and photoluminescence spectra for every sample used in Figure 1.

■

ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.nanolett.6b05068. Linear absorption and photoluminescence spectra, phased 2D ES and pump−probe, Stokes shift, TEM images and size-distributions, and FT−IR spectra (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Daniel B. Turner: 0000-0002-3148-5317 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The MRSEC Program of the National Science Foundation partially supported this work under Award Number DMR1420073. This work was also supported by NSF CAREER Grant CHE-1552235. We thank Yoichi Kobayashi, Elsa Cassette, and Benoit Mahler for advising the synthesis of nanocrystal samples, NYULMC DART Microscopy Laboratory for the consultation, and Kristen Dancel for her assistance with TEM work.

■

REFERENCES

(1) Klimov, V. I. Nanocrystal Quantum Dots, 2nd ed.; Laser and Optical Science and Technology; CRC Press, 2010. (2) Knowles, K. E.; Frederick, M. T.; Tice, D. B.; Morris-Cohen, A. J.; Weiss, E. A. J. Phys. Chem. Lett. 2012, 3, 18−26. (3) Cui, J.; Beyler, A. P.; Marshall, L. F.; Chen, O.; Harris, D. K.; Wanger, D. D.; Brokmann, X.; Bawendi, M. G. Nat. Chem. 2013, 5, 602−606. (4) Peng, X.; Wickham, J.; Alivisatos, A. P. J. Am. Chem. Soc. 1998, 120, 5343−5344. (5) Arzberger, M.; Amann, M.-C. Phys. Rev. B: Condens. Matter Mater. Phys. 2000, 62, 11029−11037. (6) Schoenlein, R. W.; Peteanu, L. A.; Mathies, R. A.; Shank, C. V. Science 1991, 254, 412−415. (7) Takemoto, K.; Hyun, B.-R.; Masumoto, Y. J. Lumin. 2000, 87− 89, 485−487. (8) Woggon, U.; Gaponenko, S.; Langbein, W.; Uhrig, A.; Klingshirn, C. Phys. Rev. B: Condens. Matter Mater. Phys. 1993, 47, 3684−3689. (9) Cerullo, G.; De Silvestri, S.; Banin, U. Phys. Rev. B: Condens. Matter Mater. Phys. 1999, 60, 1928−1932. (10) Salvador, M. R.; Hines, M. A.; Scholes, G. D. J. Chem. Phys. 2003, 118, 9380−9388. (11) Salvador, M. R.; Graham, M. W.; Scholes, G. D. J. Chem. Phys. 2006, 125, 184709. (12) McKimmie, L. J.; Lincoln, C. N.; Jasieniak, J.; Smith, T. A. J. Phys. Chem. C 2010, 114, 82−88. (13) Murray, C. B.; Norris, D. J.; Bawendi, M. G. J. Am. Chem. Soc. 1993, 115, 8706−8715. (14) Sagar, D. M.; Cooney, R. R.; Sewall, S. L.; Kambhampati, P. J. Phys. Chem. C 2008, 112, 9124−9127. 2814

DOI: 10.1021/acs.nanolett.6b05068 Nano Lett. 2017, 17, 2809−2815

Letter

Nano Letters (51) Grandhi, G. K.; Tomar, R.; Viswanatha, R. ACS Nano 2012, 6, 9751−9763. (52) Fritzinger, B.; Capek, R. K.; Lambert, K.; Martins, J. C.; Hens, Z. J. Am. Chem. Soc. 2010, 132, 10195−10201. (53) Krauss, T. D.; Wise, F. W. Phys. Rev. B: Condens. Matter Mater. Phys. 1997, 55, 9860−9865. (54) Wise, F. W. Acc. Chem. Res. 2000, 33, 773−780. (55) Landolt, H. H.; Börnstein, R. Group III Condensed Matter; Springer: Verlag, 2012; Vol. 44E. (56) Sagar, D. M.; Cooney, R. R.; Sewall, S. L.; Dias, E. A.; Barsan, M. M.; Butler, I. S.; Kambhampati, P. Phys. Rev. B: Condens. Matter Mater. Phys. 2008, 77, 235321. (57) Dzhagan, V. M.; Valakh, M. Y.; Milekhin, A. G.; Yeryukov, N. A.; Zahn, D. R.; Cassette, E.; Pons, T.; Dubertret, B. J. Phys. Chem. C 2013, 117, 18225−18233. (58) Sciacca, M. D.; Mayur, A. J.; Oh, E.; Ramdas, A. K.; Rodriguez, S.; Furdyna, J. K.; Melloch, M. R.; Beetz, C. P.; Yoo, W. S. Phys. Rev. B: Condens. Matter Mater. Phys. 1995, 51, 7744−7752. (59) Lin, C.; Gong, K.; Kelley, D. F.; Kelley, A. M. J. Phys. Chem. C 2015, 119, 7491−7498. (60) Turner, D. B.; Stone, K. W.; Gundogdu, K.; Nelson, K. A. J. Chem. Phys. 2009, 131, 144510. (61) Li, H.; Bristow, A. D.; Siemens, M. E.; Moody, G.; Cundiff, S. T. Nat. Commun. 2013, 4, 1390. (62) Seibt, J.; Hansen, T.; Pullerits, T. J. Phys. Chem. B 2013, 117, 11124−11133. (63) Brazard, J.; Bizimana, L. A.; Turner, D. B. Rev. Sci. Instrum. 2015, 86, 053106. (64) Bizimana, L. A.; Brazard, J.; Carbery, W. P.; Gellen, T.; Turner, D. B. J. Chem. Phys. 2015, 143, 164203. (65) Trebino, R.; DeLong, K. W.; Fittinghoff, D. A.; Sweetser, J. N.; Kumbugel, M. A.; Richman, B. A.; Kane, D. J. Rev. Sci. Instrum. 1997, 68, 3277−3296. (66) Gellen, T.; Bizimana, L. A.; Carbery, W. P.; Breen, I.; Turner, D. B. J. Chem. Phys. 2016, 145, 064201. (67) Hybl, J. D.; Albrecht Ferro, A.; Jonas, D. M. J. Chem. Phys. 2001, 115, 6606−6622. (68) Chen, O.; Chen, X.; Yang, Y.; Lynch, J.; Wu, H.; Zhuang, J.; Cao, Y. C. Angew. Chem., Int. Ed. 2008, 47, 8638−8641. (69) Mahler, B.; Lequeux, N.; Dubertret, B. J. Am. Chem. Soc. 2010, 132, 953−959. (70) Yu, W. W.; Qu, L.; Guo, W.; Peng, X. Chem. Mater. 2003, 15, 2854−2860.

2815

DOI: 10.1021/acs.nanolett.6b05068 Nano Lett. 2017, 17, 2809−2815