Correlation of Microphotoluminescence Spectroscopy, Scanning

Dec 16, 2013 - Transmission Electron Microscopy, and Atom Probe Tomography on a Single Nano-object Containing an InGaN/GaN Multiquantum Well...
9 downloads 0 Views 4MB Size
Letter pubs.acs.org/NanoLett

Correlation of Microphotoluminescence Spectroscopy, Scanning Transmission Electron Microscopy, and Atom Probe Tomography on a Single Nano-object Containing an InGaN/GaN Multiquantum Well System Lorenzo Rigutti,*,† Ivan Blum,† Deodatta Shinde,† David Hernández-Maldonado,† Williams Lefebvre,† Jonathan Houard,† François Vurpillot,† Angela Vella,† Maria Tchernycheva,‡ Christophe Durand,§ Joel̈ Eymery,§ and Bernard Deconihout† †

Groupe de Physique des Matériaux, UMR CNRS 6634, Normandie University, University of Rouen and INSA Rouen, 76801 St. Etienne du Rouvray, France ‡ Institut d’Electronique Fondamentale, UMR CNRS 8622, University Paris Sud 11, 91405 Orsay, France § CEA/CNRS/Université Joseph Fourier, CEA, INAC, SP2M, 17 rue des Martyrs, 38054 Grenoble Cedex 9, France S Supporting Information *

ABSTRACT: A single nanoscale object containing a set of InGaN/GaN nonpolar multiple-quantum wells has been analyzed by microphotoluminescence spectroscopy (μPL), high-resolution scanning transmission electron microscopy (HR-STEM) and atom probe tomography (APT). The correlated measurements constitute a rich and coherent set of data supporting the interpretation that the observed μPL narrow emission lines, polarized perpendicularly to the crystal c-axis and with energies in the interval 2.9−3.3 eV, are related to exciton states localized in potential minima induced by the irregular 3D In distribution within the quantum well (QW) planes. This novel method opens up interesting perspectives, as it will be possible to apply it on a wide class of quantum confining emitters and nano-objects. KEYWORDS: Atom probe tomography, photoluminescence, transmission electron microscopy, iii-nitride heterostructures, correlative microscopy, single nano-object

W

apex radius lower than 100 nm and through their analysis by means of a position- and time-of-flight-sensitive detector.12 In the last generation of atom probes, the field evaporation is triggered by a femtosecond laser,13 enabling the analysis of semiconducting14 and insulating samples.15−17 These qualities make this technique more and more employed for the analysis of nanostructured materials.18−28 APT has already been correlated with TEM,29,21 but the particular constraints that the technique imposes on the sampleespecially the field emission tip preparation by focused ion beam (FIB)30still represent a major difficulty for its use in correlation with optical spectroscopy. InGaN/GaN quantum wells represent an interesting benchmark for the correlative approach. They have a technological interest as building blocks of visible light-emitting diodes (LEDs) and laser diodes and have already been studied by APT. However, the previous APT studies were either purely structural19−21,31 or based on an approach comparing the

ith the increasing importance of nanostructured and nanoscale functional materials in the domain of photonics and optoelectronics, more and more attention is dedicated to the problem of determining the relationship between structural and optical properties of these systems. In the correlative experimental approaches different experimental techniques are applied on the same nano-object in order to establish a one-to-one correspondence between the optical and the structural properties. A number of correlative studies have been performed in the past few years on single nano-objects of different nature, composition, size, and geometry, providing a deeper and deeper insight into the interplay between their structure and their optical properties. These studies were mostly based on the complementary use of transmission electron microscopy (TEM) correlated with different ex situ1−6 or in situ7−10 optical spectroscopy techniques. Atom probe tomography (APT)11 is an analytical technique allowing for the 3D reconstruction of the elemental composition of a nanoscale sample with subnanometer spatial resolution. This is made possible through the controlled field evaporation of single atoms from a field emission tip with an © 2013 American Chemical Society

Received: September 11, 2013 Revised: December 4, 2013 Published: December 16, 2013 107

dx.doi.org/10.1021/nl4034768 | Nano Lett. 2014, 14, 107−114

Nano Letters

Letter

Figure 1. (a) Scanning electron micrograph of a GaN wire containing an InGaN/GaN multi-QW system. (b) Schematic illustration of the heterostructure geometry, with the quantum well system depicted in yellow; the red dashed and solid lines indicate approximately the region extracted for the correlated analysis and the footprint of the cut-out cylinder, respectively. (c) SEM micrograph of the cut-out cylinder analyzed by μPL and (d) of a field-emission tip analyzed by STEM and APT: (1) Si protection coating; (2) region containing the MQW system; (3) Pt−C welding; (4) supporting tungsten tip. The arrow shows the direction of the wire axis.

1b. The schematics also show the location of the portion of wire extracted for the correlative study. The samples for the correlated experiments have been prepared by FIB. As FIB may produce a significant degradation of the luminescence properties of the regions exposed to the ion beam,30 we applied a two-step method, which is described in detail in the Supporting Information. The extracted portion of microwire, preliminarily coated with 400 nm Si, is first premilled in the form of a cylinder of around 400 nm diameter, with the MQW system at its top, as depicted in Figure 1c. After μPL analysis, it is successively milled down in the form of a field emission tip (Figure 1d) for STEM and APT studies. Microphotoluminescence spectroscopy was performed at T = 4 K. The excitation of the sample was provided by a continuous wave 244 nm laser, focused on a spot with a size of the order of 1 μm at an incident power of around 50 μW. The photoluminescence was analyzed in a 460 mm focal length grating spectrometer, with a spectral resolution roughly equal to 1 nm. Figure 2a reports the spectra collected from two different whole microwires, labeled A and B, illuminated at approximately the same region where the nanoscale cylinder was cut out in wire C. In both cases, the PL signal consists of a continuous band covering the interval 2.8−3.3 eV, indicating a strong degree of inhomogeneity of the 20-fold multi-QW system. No narrow lines can be identified. A significant gain of information is obtained as soon as the probed volume is reduced in the form of a cut-out cylinder. The spectrum from the cut-out cylinder issued from wire C is shown in the bottom part of Figure 2a. Here, the PL band is made of a large number of narrow lines with FWHM of the order of 10 meV, dispersed in the interval 3.0−3.3 eV. Details of some of the narrow lines are illustrated in the plots of Figure 2b. The difference between the spectra of whole wires and of the cutout section can be explained as follows: first, the laser spot size (∼1 μm) is larger than the diameter of the cut-out cylinder;

electroluminescence of a whole device and the atom probe data.32 Only recently, a systematic study by Riley et al.33 compared the correlated STEM and APT data obtained on different sets of InGaN/GaN QWs in nanowire LEDs with the spatially resolved cathodoluminescence (CL) spectra collected on other wires from the same sample. In this work, we present the correlated study of a single nano-object containing a set of InGaN/GaN quantum wells synthesized by metal−organic vapor-phase epitaxy (MOVPE) on the lateral m-plane sidewalls of GaN wires. The nano-object was studied by microphotoluminescence spectroscopy (μPL), scanning transmission electron microscopy (STEM), and APT. To preserve the optical properties upon FIB preparation, a larger sample volume was analyzed by μPL and then refined for the structural studies. These three techniques, which are here applied for the first time on a single nano-object, yield a rich and complementary set of information, which allowed us for the interpretation of the exciton emission mechanisms in this quantum well system. The analyzed multi-QW system has been extracted by FIB milling from self-assembled GaN microwires grown by MOVPE on c-sapphire substrates.34 Previous studies assessed that these wires present a diameter in the range of 0.7−3 μm are oriented along the c-axis and have m-plane lateral facets.35,36 The base of the wires is grown at 1040 °C using trimethylgallium (TMG) and ammonia precursors as well as silane addition to get n+ +-doping and to promote the wire geometry. The silane addition is switched off after about 15 μm length to grow an unintentionally doped GaN part (about 10 μm long) at the top of the wires. The GaN wires are coated at their top with 20 unintentionally doped radial InGaN/GaN quantum wells grown at 750 °C (GaN barriers are grown at 870 °C). The scanning electron microscopy image in Figure 1a shows a typical as-grown wire. The heterostructure scheme in a longitudinal and transversal cross section is illustrated in Figure 108

dx.doi.org/10.1021/nl4034768 | Nano Lett. 2014, 14, 107−114

Nano Letters

Letter

Figure 2. (a) Red and green lines: μPL spectra obtained exciting a section of 2 different whole wires, labeled A and B with a focused laser (spot size ∼1 μm). The signal is a large band extending from 2.8 to 3.3 eV; black line: μPL spectrum obtained from a cut-out cylinder FIB-milled from wire C, exhibiting narrow emission lines dispersed in the interval 2.9−3.3 eV. (b) Zoom of the PL spectrum from the cut-out section of wire C in the two spectral regions highlighted by the dashed rectangles in a. (c) Polarization of the integral PL intensity from the cut-out cylinder obtained from wire C. The polarization is approximately perpendicular to the QW planes and to the original microwire axis.

second, the μPL may effectively probe a much larger volume in the case when a whole wire is illuminated, due to carrier diffusion and light scattering inside the wire. The multiple emission lines appearing in the spectrum suggest that the strong degree of inhomogeneity already visible in the spectra from the whole wires is probably accompanied by localization phenomena. Electron−hole pairs may indeed form excitons in correspondence of local potential minima and efficiently recombine in these regions with a well-defined energy, while the energy dispersion can originate from thickness and compositional fluctuations within the same quantum well or from thickness and compositional variations from one well to another.37 The polarization of the integral μPL spectrum from the cutout cylinder is shown in Figure 2c. The light is polarized perpendicularly to the QW planes and to the microwire axis. The polarization ratio is defined as P = (Imax − Imin)/(Imax + Imin), where Imax(min) are the intensities at the angles for which the integral PL intensity is maximum (minimum). For the cutout cylinder from microwire C its value is P = 0.89. The origin of this strong degree of polarization is most likely crystallographic, due to the selection rules for the lowest-energy optical transitions, as the role of the index contrast in the determination of the polarization of this particular system should be relatively low.38,39 On the other hand, it has been previously established that in these microwires the wire axis corresponds to the wurtzite GaN c-axis.35 In InxGa1−3N/GaN quantum wells with low In content (x < 0.25) the lowest energy transition is, as in GaN, the so-called XA excitonthe recombination of an electron bound to a heavy hole: this transition is strictly forbidden for a polarization parallel to the caxis in GaN,40,41 a behavior that is maintained in the case of planar InGaN quantum wells grown on the m-plane.42,43 The observation of the analyzed wire portion by HR-STEM will yield the direct argument for this explanation. After the second milling step, high-angle annular dark field (HAADF) STEM observations were performed on a JEOL ARM 200F microscope equipped with a Schottky field emitter

operating at 200 kV. This instrument is equipped with a spherical aberration Cs-probe corrector correcting the third order spherical aberration. The following parameters were used for the acquisition of HAADF-STEM images. For the image in Figure 3a the probe diameter was set to 0.2 nm, objective aperture semiangle of 30 mrad, and detector half-collection angle ranges between 67 and 250 mrad. For the image in Figure 3c, the probe diameter was set to 0.13 nm, objective aperture semiangle of 22.5 mrad, and detector half-collection angle ranges between 50 and 180 mrad. To avoid electron beam-induced In clustering which could significantly influence the interpretation of the correlated μPL and APT measurements,21,44 only one portion of one quantum well was observed in high resolution to directly visualize the crystallographic orientation and to study the definition of quantum well interfaces. Furthermore, to ascertain if the electron beam had an impact on the In distribution in the quantum wells in the atom probe analysis, the six innermost quantum wells have not been exposed to the electron beam. Figure 3a shows a HAADF micrograph of the upper portion of the field emission tip containing 14 quantum wells. In the image, the InGaN quantum wells are clearly identified as brighter contrast stripes approximately perpendicular to the tip axis. The tilt angle of ∼7° of the QW planes with respect to the tip axis is due to a slight misalignment of the selected wire section with the FIB gun. The HAADF intensity profile extracted from a 20 nm wide line perpendicular to the quantum well planes is reported in Figure 3b. The quantum well thicknesses measured as the FWHM of the HAADF contrast profile shown in Figure 3b are reported in the graph of Figure 4a. Numbering the QWs starting from the upper side (corresponding to the NW surface, so in the opposite order as they were grown) one notices that QWs #2−10 are thinner and somehow better defined than QWs #11−13, while QW #14 is again relatively thin. The HAADF profile also shows that the typical barrier thickness is 23 ± 2 nm. Notice that QW #1 was too close to the tip apex and could not be analyzed. 109

dx.doi.org/10.1021/nl4034768 | Nano Lett. 2014, 14, 107−114

Nano Letters

Letter

Information, provide the evidence of the presence of stacking faults (SFs) in correspondence of the bright stripes propagating perpendicularly to the quantum wells. The presence of SFs has already been assessed in similar microwires grown by this method.35 These defects can form at heterostructure interfaces due to plastic relaxation of strain accumulation, a phenomenon also known in the growth of planar quantum well systems on the GaN nonpolar m-plane.45,46 The high-resolution HAADF image of QW #3 is presented in Figure 3c. The quantum well is the brighter zone close to the top of the figure. The two quantum well interfaces are not equivalent. The interface on the core side is much more abrupt than the one on the surface side. This phenomenon is related to the growth technique, in which the quantum wells are deposited at a lower temperature than the barriers.35 However, it is not possible to discriminate whether the observed irregularity of the upper interface results from the roughness of a sharp interface or rather from a compositional gradient within the well. In the image of Figure 3c the atomic columns are visible and indicate that the crystal containing the quantum well system is hexagonal GaN, observed along the [1120̅ ] zone axissee Figure 3d for a schematics of the atomic column arrangement. Furthermore, the polar c-axis of the crystal is aligned with the wire axis, and the quantum wells lie on the nonpolar m-planes. These observations confirm thus that the polarization of the emitted light is in agreement with the selection rule for the XAtype exciton. The field emission tip from wire C has been eventually analyzed by APT in a CAMECA laser-assisted wide-angle tomographic atom probe (LAWATAP). The tip was cooled down to a temperature of 20 K. The laser illuminated the tip at a wavelength of 343 nm with pulses of 400 fs temporal width and energy in the range of 1.0−1.8 nJ, focused on a ∼80 μm wide spot (corresponding to 5−10 μJ/cm2, uniformly distributed on the tip lateral surface) and at a repetition frequency of 100 kHz. The ion detection rate was kept in the interval 0.005−0.01 ions/pulse by adjusting the DC voltage applied to the tip in the interval 6.2−10.2 kV. These parameters, especially the relatively low laser energy used, are adopted in order to ensure a correct measurement of the composition.27,47,48 The details concerning the species detected in the mass spectrum are given in the Supporting Information. The 3D reconstruction of the analyzed volume has been obtained by applying the cone angle/initial tip radius protocol,49,50 supported by the available STEM image of a large portion of the tip. The distribution of individual In atoms is presented in Figure 5. For visualization purposes, a further animation of the 3D reconstruction is reported as Supporting Information [S-3D movie]. In Figure 5a the reconstructed volume is oriented as observed along the [112̅0] direction, that is, the same as in the STEM image of Figure 3a. In Figure 5b the volume is oriented along the [0001] zone axis, while in part c it is oriented in another perspective, in order to show the nonuniform In distribution within the different QW planes. The 3D reconstruction clearly shows 16 InGaN quantum wells as approximately flat discs containing a high In density and a Ga depletion with respect to the barrier regions. QWs #2−4 appear slightly distorted, and in the same region the measured interwell distance is slightly lower than in the lower part of the analyzed volume. This is due to the irregular shape of the tip apex, which had a different cone angle and orientation with respect to the lower part of the tip, as it can be observed in

Figure 3. (a) STEM-HAADF micrograph of a portion of the fieldemission tip from wire C. In this portion of the tip, 14 quantum wells can be identified as the bright stripes approximately perpendicular to the tip axis. The bright stripes propagating perpendicularly to the quantum well planes are induced by structural defects (stacking faults and dislocations) formed upon plastic relaxation of the strained heterostructure interfaces. (b) HAADF contrast profile extracted from the red dashed line in a. (c) High-resolution image from QW#3, showing the details of the quantum well interface and the atomic columns observed along the [112̅0] zone axis of the GaN wurtzite crystal. (d) Schematic representation of the crystal structure as observed in c.

Figure 4. (a) Measurement of quantum well thickness taking into account the FWHM of the STEM HAADF contrast profile (red empty squares) and from the APT 1D InN fraction profiles, and as the FWHM of the InN-rich region (black empty triangles). (b) Maximum InN fraction obtained from the 1D APT composition profiles.

Figure 3a also shows several brighter contrast stripes, highlighted by arrows, perpendicular to the quantum well planes. These stripes have been analyzed in more detail in another tip extracted from a wire labeled D. These observations, reported in Figure S2 of the Supporting 110

dx.doi.org/10.1021/nl4034768 | Nano Lett. 2014, 14, 107−114

Nano Letters

Letter

Figure 5. (a, b, c) 3D reconstruction of the In atom distribution in the analyzed sample from wire C, evidencing the presence of 16 quantum wells. The reconstructed volume in part a is oriented along the [112̅0] zone axis and can therefore be compared directly with the STEM image in Figure 4a. Part b is oriented along the [0001] zone axis, while part c is tilted in order to show the nonuniform In distribution within the QW planes. The red and blue arrows point to the extremities of two separate In-rich regions which propagate through the QW system. (d) 2D-mapping of the in-plane fraction of InN calculated through 8 nm thick boxes containing QWs #4, #12, and #17. Notice that the InN fraction appearing on the scale is underestimated as the size of the box exceeds the quantum well thickness. (e) InN fraction maps calculated within the 1 nm thick slices of QW #12 shown in the upper part of the figure.

of the box with the local QW interface. The quantum well thicknesses and peak InN fractions obtained from the 1D composition profile are reported in the graph of Figure 4a and b, respectively. The thickness values obtained as the FWHM of the InN-rich regions are in reasonable agreement with but systematically lower than those calculated from the STEM HAADF profile. This is due to the fact that the measurement based on the HAADF contrast is based on a projection over the whole tip thickness, which may induce an overestimation of the measurement because of long-range roughness. For the same reason, the thickness of the QWs #11 to #13 can be measured more accurately based on APT data than by STEM. A further and decisive insight in the relationship between radiative transitions and quantum well structural properties can be achieved by the analysis of the 3D distribution of InN within the QW planes. The 3D reconstruction in Figure 5c and the concentration maps for QWs #4, #12, and #17 in Figure 5d evidence that the In distribution is not uniform, especially in the analyzed volume of the QWs #12 and #17. In particular, the reconstruction and the concentration maps show that there are several stripe-like In-rich regions, approximately directed along the c-axis. These In-rich regions, highlighted by the blue and red arrows in Figure 5c, propagate through the multi-QW system, perpendicularly to the QW planes. Due to the fact that the multi-QW system is slightly tilted with respect to the axis of the tip, these regions progressively shift out of the analyzed volume. This also shows that these features are not related to the tip geometry and exclude that they could be ascribed to quantification artifacts or laser-induced In segregation. The presence of these In-rich regions could be related either to the specific MOVPE growth method, realized with two different temperatures for the deposition of quantum wells and barriers, or, more likely, to the presence of stacking faults (SFs).51,52 It has been indeed observed in planar c-plane InGaN/GaN QW systems that under SF-rich conditions large thickness

the STEM micrograph of Figure 3a. The STEM measurement, however, allows for the correction of this artifact when plotting 1D composition profiles. The very small concentration (