Maximilian
Ries
ab,
Felix
Nippert
a,
Benjamin
März
cd,
Manuel
Alonso-Orts
e,
Tim
Grieb
ef,
Rudolfo
Hötzel
e,
Pascal
Hille
e,
Pouria
Emtenani
a,
Eser Metin
Akinoglu
b,
Eugen
Speiser
b,
Julian
Plaickner
bg,
Jörg
Schörmann
h,
Matthias
Auf der Maur
i,
Knut
Müller-Caspary
d,
Andreas
Rosenauer
ef,
Norbert
Esser
ab,
Martin
Eickhoff
e and
Markus R.
Wagner
*aj
aTechnische Universität Berlin, Institut für Festkörperphysik, Hardenbergstraße 36, 10623 Berlin, Germany. E-mail: markus.wagner@physik.tu-berlin.de
bLeibniz-Institut für Analytische Wissenschaften – ISAS e.V., Department Interface Analytics, Schwarzschildstraße 8, 12489 Berlin, Germany
cErnst-Ruska-Centre for Microscopy and Spectroscopy with Electrons at Forschungszentrum Jülich, Wilhelm-Johnen-Str., 52425 Jülich, Germany
dDepartment of Chemistry and Center for NanoScience, Ludwig-Maximilians-Universität München, Butenandtstr. 11, 81377 Munich, Germany
eUniversität Bremen, Institut für Festkörperphysik, Otto-Hahn-Allee 1, 28359 Bremen, Germany
fUniversität Bremen, MAPEX Center for Materials and Processes, Bibliothekstr. 1, 28359 Bremen, Germany
gHelmholtz-Zentrum Berlin für Materialien und Energie GmbH, Hahn-Meitner-Platz 1, 14109 Berlin, Germany
hJustus-Liebig-Universität Gießen, I. Physikalisches Institut und Zentrum für Materialforschung (LaMa), Heinrich-Buff-Ring 16, 35392 Gießen, Germany
iUniversity of Rome Tor Vergata, Department of Electronic Engineering, Via del Politecnico 1, 00133 Rome, Italy
jPaul-Drude-Institut für Festkörperelektronik, Leibniz-Institut im Forschungsverbund Berlin e.V., Hausvogteiplatz 5–7, 10117 Berlin, Germany
First published on 10th March 2023
The luminescence of InxGa1−xN nanowires (NWs) is frequently reported with large red-shifts as compared to the theoretical value expected from the average In content. Both compositional fluctuations and radial built-in fields were considered accountable for this effect, depending on the size, structure, composition, and surrounding medium of the NWs. In the present work, the emission properties of InGaN/GaN NWs grown by plasma-assisted molecular beam epitaxy are investigated in a comprehensive study combining ultraviolet-Raman and photoluminescence spectroscopy (PL) on vertical arrays, polarization-dependent PL on bundles of a few NWs, scanning transmission electron microscopy, energy-dispersive X-ray spectroscopy, and calculations of the band profiles. The roles of inhomogeneous In distribution and radial fields in the context of optical emission properties are addressed. The radial built-in fields are found to be modest, with a maximum surface band bending below 350 meV. On the other hand, variations in the local In content have been observed that give rise to potential fluctuations whose impact on the emission properties is shown to prevail over band-bending effects. Two luminescence bands with large positive and moderate negative polarization ratios of ≈+80% and ≤−60%, respectively, were observed. The red-shift in the luminescence is associated with In-rich inclusions in the NWs due to thermodynamic decomposition during growth. The negative polarization anisotropy is suggested to result from spontaneously formed superlattices in the In-rich regions of the NWs. The NWs show a preferred orthogonal absorption due to the dielectric boundary conditions and highlight the extreme sensitivity of these structures towards light polarization.
Owing to the crystal symmetry and peculiarities of the geometry, luminescence emitted from semiconductor NWs exhibits a distinct polarization dependence, which can be of importance for their application in single photon emitters and in imaging and display technologies.4,12 The depolarization ratio ρ compares the intensities of parallel (I∥) and orthogonally (I⊥) polarized emissions with respect to the NW axis:
![]() | (1) |
In the case of group-III nitrides, polarization anisotropy has been found in both photoluminescence (PL) and electroluminescence measurements of pure GaN NWs and those with embedded InGaN nanodiscs (NDs).5,13–16 Interestingly, these anisotropies contradict the expected ratio from the dipole selection rules of excitonic transitions.
Two main mechanisms have been identified to be of relevance for the polarization anisotropy in quasi-one-dimensional sample geometries: (i) quantum-mechanical confinement and (ii) dielectric confinement.2 The latter leads to a preponderating emission of light with polarization along the nanowire axis and a positive ρ. The contribution of these effects is mainly based on the diameter of the respective NW and the dielectric contrast between the NW and the surrounding medium. As a rule of thumb, once the actual diameter of the NW exceeds the exciton Bohr radius, the first effect drastically weakens. For InGaN, the exciton Bohr radius is commonly expected to be around 3–10 nm, depending on the In content.17 For NWs of larger diameter, the second, purely classical effect dominates, until the NW diameter becomes large enough to be comparable to bulk material.2 Depolarization ratios of GaN NWs with embedded InGaN QWs5,18–23 and quantum dots (QDs) have been frequently reported,12,24–26 but not for InGaN NWs. Especially in structures containing QDs in a wire, i.e., embedded regions that act as QDs, the emission properties can differ strongly due to non- or semi-polar side-facets and the emission typically shows polarization perpendicular to the NW axis.24,25
This study reports the composition, influence of radial electric fields, and polarization anisotropy of luminescence light emitted from InxGa1−xN NWs grown on top of GaN NWs in a self-assembled process. The PL energy differs significantly from what is expected for the average In concentration as determined by X-ray diffraction (XRD) analysis. The origin of this red-shift is discussed with respect to compositional fluctuations and band-bending effects. The depolarization ratio reaches values as high as +90% at approximately 2 eV. An additional higher energy luminescence band is observed that exhibits a negative ρ. The origin of this band is addressed, and experimentally observed higher intensities for excitation with polarization perpendicular to the NW axis are discussed.
The following discussion is divided into four sections: (i) PL and Raman measurements of the as-grown vertically aligned NW arrays, (ii) PL experiments on bundles of a few NWs, (iii) scanning transmission electron microscopy (STEM) and energy-dispersive X-ray spectroscopy (EDX) analysis, and (iv) calculations of band bending.
![]() | ||
Fig. 1 Photoluminescence spectra of the as-grown GaN and InGaN/GaN NW ensembles recorded at 300 K. The spectra are offset for clarity. |
On the same two samples, ultraviolet (UV)-Raman spectroscopy was performed. The first-order modes observed for the as-grown GaN NWs in Fig. 2(a) are the E2(high) at 567 cm−1 and the longitudinal optical (LO) phonon A1(LO) at 718 cm−1; see also Table 1. The E2(high) frequencies are close to the literature value of strain-free GaN (568 cm−1).28 However, the A1(LO) shifts significantly from the expected position at 734 cm−1.28 The shift of the NW Raman modes to lower wavenumbers compared to bulk data is a result of phonon confinement,29–31 and of excitation above the fundamental bandgap energy.32 The NW sidewalls form boundaries that confine the phonons and lead to uncertainty in the associated wavevector. The larger shift of A1(LO) originates from the steeper slope in the phonon dispersion, whereas the E2(high) dispersion is rather flat.33
UV-Raman spectra of the as-grown InGaN/GaN NWs in Fig. 2(b) are characterized by an E2(high) peak at 552 cm−1 with low-frequency shoulders at 536 cm−1 and an A1(LO) structure consisting of two peaks at 710 cm−1 and 671 cm−1, respectively. Concerning the as-grown InGaN/GaN NWs, the excitation wavelength has strong implications on the scattering volume: photons with energy below the InGaN bandgap penetrate deep into the NW material, probing both the InGaN and GaN parts, while photons with 266 nm are mainly absorbed in the InGaN part.
The frequency shift of the ternary InxGa1−xN NW E2(high) at 552 cm−1 yields an In distribution centered at around x ≈ 22%.34 Considering the enlarged diameter of the InxGa1−xN wire segment, the frequency shift due to phonon confinement can be neglected, so the A1(LO) phonon frequency of 710 cm−1 translates directly to a value of x ≈ 16%.34 The differing values obtained from the E2(high) and A1(LO) phonon modes highlight the inhomogeneity of the material and the complexity to relate a Raman shift to an In content.
The low-frequency peak of the A1(LO) structure at 671 cm−1 suggests the occurrence of InxGa1−xN with a higher In content x, which is also indicated by a low-frequency shoulder of the E2(high) at 532 cm−1. Both peak positions yield x ≈ 40%.34 Surface optical modes, providing another possible explanation for the low-frequency A1(LO) peak, should be located significantly above the observed mode at 671 cm−1, which is not the case.35–37
In the following, the polarization anisotropy of two different bundles with fewer than 10 NWs will be compared. The respective PL spectra are shown in Fig. 4 together with atomic force microscopy (AFM) images. The excitation power was approximately 1 mW and corresponds to 75% in Fig. 3. The depolarization ratio is mainly determined by the emission properties. Although the polarization anisotropy is stronger in perpendicular absorption and less prominent in the parallel case, as summarized in Table 2, the sign of the ratio is determined by the emission rather than the absorption. The difference in the depolarization ratio for perpendicular or parallel excitation, i.e., absorption, can be ascribed to the dielectric boundary conditions and is extremely sensitive to the diameter, refractive indices, and wavelength.2
The InGaN-related part of both bundles exhibits two peaks. They have a dominating emission at around 2.0 eV, band (I), with a high mean depolarization ratio of up to 80% and a second emission band (II) at around 2.2–2.6 eV with a moderate negative
below −60%, the latter being hardly noticeable in bundle B. The mean depolarization ratio is the average value from the two excitation polarizations. It is important to note that the excitation energy of 3.4 eV determines the absorption anisotropy, while the respective PL energy yields the emission anisotropy, i.e., 2.0–2.6 eV. The structures of the two bundles are different. From the AFM insets, one can see that bundle B consists of at least two wires that are coalesced around the center. The bundle A in Fig. 4(a) has the same height, but the structure is less clear. The elevated part presumably at the bottom segment indicates additional NWs in this bundle.
In most reported cases on polar NWs, the luminescence is strongly polarized parallel to the NW axis, i.e., ρ ≫ 0.5,12,22 In this case, this is true for band (I), but the depolarization ratio of the higher energy band (II) is negative. Chen et al. ascribed the reduced depolarization ratio for thicker, i.e., coalesced GaN NWs to defect-related emission from localized excitons, rather than from free exciton emission.15 This reflects well the geometry of the NWs studied in this publication. However, it only explains a reduced from one wire to another, but not the change in sign.
A possible origin is the existence of QD-like formations in the NWs, e.g., confined In-rich regions. It is well known from the literature that the so-called QDs in a wire structures can exhibit large negative polarization anisotropy.12,24,25 This possible assignment is supported by Raman measurements, indicating localized regions with a high In content that most likely give rise to the low-frequency peak of the A1(LO) structure, cf. Fig. 2.
The implication from Fig. 5(b) is that the measured In concentration decreases with increasing distance from the apex. These observations are supported by the results from Raman measurements, revealing a high In content and a low In-content phase, respectively. However, the concentration gradient seen by EDX is largely due to the fact that the width of the GaN shell increases towards the base of the NWs, which leads to an effective reduction of the In content measured with EDX.
Fig. 5(c) shows an ADF-STEM image of the upper part of the NWs in Fig. 5(a). It reveals an inner structure of the In-containing inclusion consisting of brighter and darker layers with a periodicity of about 2.2 nm. The high contrast in ADF-STEM images suggests a higher In content based on two mechanisms: (i) the contrast strongly depends on the material's atomic number (Z-contrast) and (ii) static disorder affects electron scattering significantly. An EDX intensity map from the region inside the black rectangle in Fig. 5(c) is shown in Fig. 5(d) with the intensities of the In–L and Ga–K lines color-coded. From the profile in Fig. 5(e) it is discernible that bright layers in the ADF image contain more indium compared to the darker layers. Due to a broadening of the STEM beam in the rather thick specimen and dynamical scattering effects, assigning an absolute composition to the EDX signal is a challenge that lies beyond the scope of the current study. However, it can be concluded that the In content fluctuates across the layers periodically, complementary to the Ga-content. The two characteristic features marked by arrows in Fig. 5(c) and Fig. 5(d) are stacking faults. The STEM analysis proves the presence of a superlattice that is not related to defects, but defects were observed especially in the upper half of the wires. In general, the NWs comprise the GaN base and the InGaN part, starting at the InGaN/GaN interface. This part is surrounded by a very thin GaN shell. The superlattice can be situated in between two InGaN parts or directly starting at the interface. In some cases, the top part is missing; however, this can be due as well to the dispersion of the nanowires via ultrasonication.
Fig. 6 depicts the calculated profiles of the lowest conduction and the top-most valence band edges along the diameter in single freestanding InGaN/GaN NWs with a nominal In concentration of 30%. Inclined sidewalls with increasing diameter towards the top in combination with the growth direction (000) lead to a discontinuity of the normal component of the polarization field across the side surface, inducing an equivalent negative surface charge. Bulk doping with a density above 1 × 1017 cm−3 compensates for the polarization-induced surface charge, which becomes apparent from the realignment of the Fermi energy, as shown in Fig. 6(a).
![]() | ||
Fig. 6 Calculations of the valence (solid lines) and conduction band energies (broken lines) for different doping concentrations of 1 × 1016–1 × 1019 cm−3 in (a) tapered (left) and (b) straight (right) In0.3Ga0.7N/GaN NWs. The density of the surface states was assumed to be 6 × 1014 cm−2 at energies consistent with Segev et al.40 The NWs are assumed to be in air without surrounding electric fields. The insets show where the band profiles are extracted: 800 nm from the GaN base (blue), in the middle of the InGaN part (green). The intersection is visualized by a red plane. |
The strain at the GaN–InGaN interface relaxes after <50 nm in both directions, thus piezoelectric effects exist only close to the interface. Consequently, the radial field close to the interface is slightly stronger compared to the center of the InGaN part. For low and moderate doping levels, the radial field is very small. Based on typical values for ne ≈ 0.1–1.0 × 1018 cm−3,39,41,42 the radial electric field induces a band bending of less than 200 meV (for In0.1Ga0.9N/GaN NWs less than 350 meV), i.e., it can not explain the observed red-shift of the luminescence, which enforces the suggested existence of In-rich regions.
1D 8-band k·p simulation results of superlattices with alternating layers of 40%/60%-InGaN and 30%/50%-InGaN as shown in Fig. 7 underline the effect of strain on the emission characteristics of such structures. Both emission energy and dominant polarization depend crucially on the strain state of the nanowire due to the strain-induced band shifts. Fully strained to a surrounding GaN matrix, the orthogonal polarization dominates over the parallel case and the photoluminescence is strongly blue-shifted compared to the relaxed case (here 40%-InGaN). Furthermore, the emission characteristics approach those of the bulk case (yellow and green shaded areas) when the relaxation increases.
ADF-STEM imaging and EDX map data acquisition were conducted at 200 kV using a Hitachi HF5000 field emission STEM equipped with a dual Ultim® Max silicon drift detector system from Oxford Instruments. An area of 50.1 nm × 26.1 nm was scanned at a step size of 57 pm. The data set was 8 × 8 binned before peak deconvolution and display. The line scan was reconstructed from the original map data with applied 4 × 4 binning and a smoothing factor of 5.
Raman and PL spectra in the UV region (266 nm, Nd:YAG) were recorded in a back-scattering configuration using a Horiba T64000 spectrometer in single grating mode with a thermoelectrically cooled Synapse charge-coupled device. A UV-optimized 2400 lines per mm holographic grating with a spectral resolution of 4(2) cm−1 at 266 nm served as a dispersing element. UV-Raman spectra were collected with an LMU-40X-UVB, NA = 0.5 objective from Thorlabs. The spectra were calibrated to the Raman bands of β-Ga2O3 and atmospheric N2 and O2.
PL measurements at 355 nm were recorded with an 1800 lines per mm grating. The spectral resolution was around 0.5 meV at 3.492 eV. A Zeiss LD EC Epiplan-Neofluar ×100 (NA = 0.75) objective was used to focus light on the sample with a spot-size below 500 nm confirmed by confocal lateral laser scanning over a NW. The polarization of the excitation was controlled with a λ/2-plate behind the laser. The polarization dependence of the beamsplitter was taken into account. The emission polarization was rotated using a Fresnel rhomb to match the preferred configuration of the grating. A Rochon prism served as an analyzer in front of the spectrometer. The choice of these polarizing elements ensured stable polarization conditions over the entire spectral range. Measurements were performed with polarization parallel or perpendicular to both incident laser and emitted PL light with respect to the NW axis.
The band profile in In0.3Ga0.7N–GaN NWs was calculated by solving the 3-dimensional nonlinear Poisson equation, discretized with a standard finite-element scheme, for different values of bulk n-type doping. No compositional fluctuations have been considered. Electron and hole densities were calculated based on the standard bulk expressions using strain-corrected 8-band bulk k·p to calculate band edge energies.46 The strain was calculated using linear elasticity under natural (zero force) boundary conditions.47 For the spontaneous and piezoelectric polarization, the nonlinear model described by Prodhomme et al. was applied.48 In the Poisson equation, surface states with a density of 6 × 1014 cm−2 were used as boundary conditions with energy levels according to Segev et al.,40 and assuming zero electric fields outside the NW. For the calculation of the spectra shown in Fig. 7, we have taken into account that the emission from a dipole oriented orthogonally to the nanowire is partly suppressed.2 Assuming the case of a small nanowire and an optical relative permittivity of εr ≈ 6.2, we reduced the orthogonal polarization by a factor of 4/(1 + εr)2 ≈ 0.077.
This journal is © The Royal Society of Chemistry 2023 |