Nuclear Spectroscopy in Single Quantum Dots: Nanoscopic Raman Scattering and Nuclear Magnetic Resonance

See allHide authors and affiliations

Science  04 Jul 1997:
Vol. 277, Issue 5322, pp. 85-88
DOI: 10.1126/science.277.5322.85


Resonant Raman and nuclear magnetic resonance spectroscopies from single gallium arsenide quantum dots are demonstrated. The nuclei were probed through changes in the optical spectra of the quantum dot exciton arising from exciton-nuclear interactions. This approach allowed the application of optical spectroscopy with its extremely high sensitivity and selectivity. The experiments had a lateral spatial resolution of about 10 nanometers and probe a volume that was five orders of magnitude smaller than that of previous semiconductor nuclear spectroscopic studies.

Semiconductors are typically modeled in terms of almost independent electronic and nuclear systems with coupling through relatively weak electron-nuclear interactions. The electronic system can be studied at optical frequencies through sensitive optical spectroscopies with tunable lasers and highly sensitive detectors. Recently there has been significant progress in the spectroscopic study of very small semiconductor samples, culminating with the detailed study of individual quantum dots (QDs), which are solid-state crystalline structures so small that their electronic wave function is completely localized and their energy spectrum is fully quantized (1-6). It has been shown that the optical linewidths of single QDs are orders of magnitude narrower than those observed in ensemble measurements. In fact, the linewidths can approach the natural linewidths expected from radiative lifetimes (3). With such narrow lines in the optical spectra, single QD spectroscopy has led, for example, to the direct observation of fine structure (4), hyperfine shifts (5), and spectral wandering (6). Single QD spectroscopy has followed the earlier examples of atomic and, especially, single molecule spectroscopies (7). To date, however, only the electronic spectra of single QDs have been measured. Here we report spectroscopic measurements of the nuclear system in a single QD. The capability to do nuclear spectroscopy on a scale of 10 nm in single QDs could provide sensitive measurements of local strain and chemical composition, issues of great importance for semiconductor nanostructures.

Here we will consider the bound electron-hole pair known as an exciton. An exciton in a QD interacts with the nuclei through either an electric or a magnetic field. In the case of the electric-field–mediated interaction, the dipolar electric field of the exciton couples to longitudinal optical vibrations of the nuclei (LO phonons) through the Frohlich interaction (8). By measuring the QD emission intensity under appropriate optical resonance conditions, the optical phonon Raman spectrum of individual QDs can be measured. In the case of magnetic-field–mediated interactions, the exciton spin polarizes the nuclear spins within the QD through the contact hyperfine interaction (9). In return, the nuclear spins act back on the exciton and shift the exciton's energy (Overhauser shift). By measuring the magnitude of the Overhauser shift under nuclear magnetic resonance (NMR) conditions, it is possible to measure the NMR spectra of the nuclei within individual QDs.

We have studied semiconductor samples grown by molecular beam epitaxy containing layers of GaAs sandwiched between layers of AlxGa1− xAs. The GaAs behaves like a quantum well for excitons. In thin GaAs layers (<100 nm), the energy resulting from spatial confinement is large, leading to significant changes in excitonic properties. Furthermore, because of atomic fluctuations at the interfaces, the excitons are also laterally confined (Fig. 1A). A one-monolayer change in the thickness of a 5-nm GaAs quantum well changes the exciton energy by about 10 meV, which is sufficient to completely localize the exciton at low temperature and provides an excellent model QD system for spectroscopic study (1-5). The NMR results were taken from a (001) GaAs/Al0.3Ga0.7As sample with a GaAs well width of 15 monolayers and Al0.3Ga0.7As barrier widths of 25 nm. Each monolayer includes an As and Ga layer and is 0.28 nm thick. This structure produced strong photoluminescence (PL) with narrow spectral linewidths. For the Raman scattering experiment, it was necessary to reduce the PL in order to measure the resonant Raman signal. For this purpose, we used thin AlAs barriers, which increased the tunneling rate out of the quantum well and reduced the PL intensity. A 19-monolayer GaAs quantum well sample with 5-nm barriers surrounded by 60-nm GaAs buffer layers was found to work well. Pure AlAs barriers were used to avoid AlxGa1− xAs alloy modes and simplify the phonon spectrum. In both samples, the lateral sizes of the QDs formed by the interface fluctuations are on the order of 10 to 100 nm (2, 4).

Figure 1

(A) The QD's lateral localization within the GaAs/AlAs quantum well. (B) The resonant Raman process: The laser excites an exciton at energyE exc, which can emit an optical phonon and reemit a photon at the QD energy. (C) The NMR experiment: Exciting the sample with circularly polarized laser light in the presence of a static magnetic field B ext pumps the nuclear spins into a high state of polarization. The resonant transverse magnetic field B RF destroys the nuclear polarization. The effect is measured through the magnitude of the Overhauser shift in the QD PL spectrum.

The samples were excited by a Ti:sapphire laser with a power of about 50 W cm−2, and the emitted light was detected in a backscattering geometry through a triple-grating spectrometer by a charge-coupled device detector with a resolution down to 30 μeV. Individual QDs were studied by exciting and detecting light through apertures ranging in diameter from 25 down to 0.2 μm in a metal mask patterned directly on the sample with electron-beam lithographic techniques (2). In this way, it is possible to spectroscopically study individual QDs with high sensitivity and spatial resolution (down into the optical near-field regime) (2-5). The sample was kept at a temperature of 6 K. For the NMR measurement, the sample was inserted into a longitudinal magnetic field of 1 T aligned normal to the quantum well. A four-turn Helmholtz coil provided a transverse radio-frequency magnetic field with an amplitude of about 0.7 G. The Raman experiments were performed with the polarization of the incident and detected light parallel to the same (110) crystal axis, whereas in the NMR experiments, the polarization of the incident and detected light was circular and linear, respectively.

We first demonstrate the Raman spectroscopy of a single QD (Fig.2). The PL spectrum from the GaAs/AlAs sample with an aperture size of 25 μm (Fig. 2A) provides an ensemble measurement of QDs, whereas the PL spectrum from a 0.8-μm aperture (Fig. 2C) shows a single QD PL line at E 0(ground-state energy). An excitation spectrum in which the intensity of the QD PL line at energy E 0 is plotted as a function of the laser energy (Fig. 2C) exhibits sharp resonances within about 10 meV of the QD PL line corresponding to direct absorption into QD states followed by relaxation into the luminescing state by emission of acoustic phonons (2). At higher energies (Fig. 2D), starting about one transverse optical (TO) phonon energy above the luminescing state, there is additional sharp structure, although about an order of magnitude weaker than the lower energy excitation resonances. This higher energy structure arises from absorption into two-dimensional (2D) exciton states, followed by a transition to the luminescing QD state accompanied by the emission of an optical phonon (Fig. 1B). The structure between the energies labeledE 0 + E TO andE 0 + E LO represents the optical phonon spectrum modulated by the QD exciton-phonon interaction. This strong enhancement corresponds to resonant Raman scattering from a single QD. The resonance linewidths are extremely sharp, allowing the selective probing of individual QDs through individual phonons. There is also a peak at higher energies (E 1 +E LO) that corresponds to a two-step process in which the 2D exciton makes a transition into the excited state of the QD at E 1 accompanied by the emission of an LO phonon, and then makes a transition to the QD luminescing state by emitting an acoustic phonon.

Figure 2

GaAs/AlAs emission spectra demonstrating single QD Raman spectroscopy. (A) An ensemble PL measurement through a 25-μm aperture, showing emission at 1.656 and 1.668 eV from the 19- and 18-monolayer parts of the quantum well, respectively. (B) Ensemble PL excitation spectrum (25 μm). (C and D) PL and PL-excitation spectra, respectively, obtained through a 0.8-μm aperture. The PL spectrum shows a single-QD PL line at E 0. The excitation spectrum shows sharp electronic resonances within the first 10 meV and Raman resonances starting about 34 meV above the luminescing state. (E) The top spectrum is a high-resolution LO-phonon spectrum from a QD, obtained from the intensity of the QD emission as a function of the energy difference between the QD (E QD) and the laser (E exc) as the laser is scanned. The bottom (ensemble) spectrum is the resonant Raman spectrum obtained in the conventional way from the 25-μm aperture.

The Raman spectrum from individual QDs can be acquired by plotting the intensity of that QD PL line as a function of the difference between the exciting and detected light energy (E excE QD) (Fig. 2E) for the LOn phonon region. The structure in the ensemble spectrum (10) (bottom trace in Fig. 2E) is similar to previous studies (11) and arises from confined and interface LO phonons due to the vertical confinement as discussed previously (8, 12). For example, LOnarises from a confined phonon with an envelope function that hasn − 1 nodes, and it goes to zero in the AlAs barriers. The LOn phonons in single QD spectra (Fig. 2E) occur at the same energies as those measured in the ensemble spectrum with little fluctuation and with similar linewidths. This correspondence implies that the LO phonons are not confined in the lateral dimensions by the interface fluctuations, in sharp contrast to the excitons. Nevertheless, because the phonons are resonantly excited through the wave function of a single localized exciton, a localized phonon wave packet is excited, and only the nuclei within the QD are being probed. Interestingly, although the energies are the same, the relative intensities of the spectral features vary significantly. For example, the TO2 phonon intensity relative to that of LO2 becomes significantly stronger than the ensemble average in some cases (Fig. 2, B and D). This result likely arises from changes in the exciton-phonon scattering strength as the lateral shape and size of the QD varies, although we cannot rule out the possibility that the apertures lead to a z component of the electric field of the light that breaks the selection rules in some cases.

To probe the nuclear spins through magnetic resonance, it is necessary to apply a static longitudinal magnetic field (Fig. 1C). In this external field, the exciton spectral lines split into doublets as a result of the Zeeman interaction with the external field (13). However, there is also a contribution to the Zeeman splitting from the nuclear spin (5, 9). This additional contribution can become strong under optical pumping with circularly polarized light in a sufficiently strong external field (B ext > 0.2 T) (14). In the QDs, the nuclear spins can be optically pumped to polarization values in excess of 70% through the hyperfine interaction (5). This nuclear polarization acts back on the exciton and strongly enhances or reduces the Zeeman splitting, depending on the orientation of the polarization of the exciting light. This change in the Zeeman splitting, or Overhauser shift, can be described in terms of an internal effective magnetic field. We have measured internal effective fields as high as 1.5 T (5). The measurement of the Overhauser shift in the QD exciton's PL line provides a direct measure of the strength of the hyperfine interaction and the value of the nuclear polarization.

The nuclei can be strongly depolarized by applying a transverse magnetic field at radio frequencies (RF) that are resonant with the nuclear Zeeman splittings. The PL was first excited with σ+ and σ polarized light (15) and with an applied longitudinal magnetic field of 1 T, but without the transverse resonant RF field (solid curves in Fig.3A). With the σ+) excitation, the Zeeman splitting was increased (decreased) by 90 μeV because of the optically pumped effective internal field. In a second measurement, the RF frequency was continuously swept over the Ga and As nuclear resonances with a sweep rate of 5 Hz as the PL was being measured (dashed curves in Fig. 3A). In this case, the nuclear contribution to the Zeeman splitting was strongly reduced, leading us to conclude that most of the nuclei are depolarized by the resonant RF field. This result verifies the nuclear origin of the enhanced Zeeman splitting and at the same time demonstrates the effectiveness of the depolarizing NMR field.

Figure 3

(A) Zeeman splitting in a single QD PL spectral line centered at 1.6286 eV in a longitudinal magnetic field of B ext = 1 T excited with either σ+ or σ polarized light with (dashed lines) and without (solid lines) a transverse RF magnetic field continuously scanning through all the Ga and As nuclear magnetic resonances. (B and C) NMR spectra from another QD, showing the 75As and 69Ga resonances, respectively, at B ext = 1 T.

To demonstrate the NMR spectroscopy, the Zeeman splitting was measured as a function of the frequency of the transverse field as the RF was stepped through each of the nuclear resonances. In this way, single-QD75As and 69Ga NMR spectra were recorded (Fig.3, B and C, respectively). The relative strengths of the two NMR spectra can be understood in terms of the abundances of the nuclear isotopes and the nuclear gyromagnetic ratios (9). The resonant frequencies vary by tens of kilohertz between QDs, and some spectra are significantly broadened (up to a factor of 10). These effects illustrate the potential of NMR as a local probe. The capability of measuring changes in the Overhauser energy shift provides a direct and quantitative measure of the nuclear spin polarization that is not available if only the electronic polarization is monitored as in previous macroscopic studies using optical NMR (16).

The experiments described above were performed on single QDs that range in size from 3 to 6 nm in the vertical dimension and from about 10 to 100 nm in the lateral dimension. Such QDs consist of as few as 104 nuclei. These experiments are possible because of the high sensitivity of optical spectroscopy and because of the good selectivity possible with high-resolution detection and resonant excitation when the spectral lines are sharp. Nevertheless, these measurements remain challenging, and there are several conditions that we found to be important for success. In order to measure the resonant Raman enhancement, it was necessary that the Raman intensities be at least comparable to the PL intensities. In the Raman sample, this condition was satisfied by using thin barriers to reduce the PL intensity. However, in other samples, such as the GaAs/Al0.3Ga0.7As sample used for the NMR experiment, it was not possible to detect sufficient intensity enhancements on top of much stronger PL intensities, and we could not do Raman spectroscopy. In contrast to the Raman spectroscopy, the NMR was measured through changes in the PL energy. For this reason, the experiment benefits from a strongly luminescing sample. In both cases, the narrow spectral lines allowed us to resolve individual QDs, but in the NMR, because we were measuring changes in the Overhauser shifts, it was especially important to have very narrow PL spectral lines. Our spectral resolution (in other words, how well we can pick the resonance energy) is given roughly by the PL linewidth divided by the signal to noise ratio. Currently, this is about 5 μeV, and we can measure changes in nuclear polarization down to about 6%.

Previously, the most sensitive resonant Raman and optical NMR experiments performed on semiconductor nanostructures involved at least 109 nuclei (11, 16). The current experiments represent an increase in sensitivity of five orders of magnitude. Recently, there has also been extensive effort in micro-Raman spectroscopy, including work at helium temperature (17), and there are pioneering efforts in optical-near-field Raman imaging at room temperature with lateral spatial resolutions down to 250 nm (18). It is important to note that, although the aperture sizes in our experiments were restricted to 200 nm or larger, the actual resolution of our experiments is much better because of our use of resonance techniques. By resonating with a single QD, we are probing the nuclei within that QD, and thus, our lateral spatial resolution is about 10 nm, an order of magnitude better than previous optical-near-field Raman spectroscopy. In some cases it should be possible to apply such resonance techniques to optical near-field imaging to enhance both sensitivity and selectivity.

There is no obvious reason why the experiments presented here could not be extended to smaller sizes or to other material systems. The present experiments demonstrate that if optical spectroscopy on single quantum units can be done, then nuclear spectroscopy on the same scale is possible.

  • * To whom correspondence should be addressed. E-mail: gammon{at}

  • Present address: National Institute of Standards and Technology, Gaithersburg, MD 20899, USA.


View Abstract

Stay Connected to Science

Navigate This Article