Growth of Magneto-optically Active (Zn,Mn)Se Nanowires

Jul 31, 2009 - Growth of Magneto-optically Active. (Zn,Mn)Se Nanowires. B. J. Cooley,† T. E. Clark,† B. Z. Liu,† C. M. Eichfeld,† E. C. Dickey...
0 downloads 0 Views 3MB Size
NANO LETTERS

Growth of Magneto-optically Active (Zn,Mn)Se Nanowires

2009 Vol. 9, No. 9 3142-3146

B. J. Cooley,† T. E. Clark,† B. Z. Liu,† C. M. Eichfeld,† E. C. Dickey,† S. E. Mohney,† S. A. Crooker,‡ and N. Samarth*,† Center for Nanoscale Science and Materials Research Institute, The PennsylVania State UniVersity, UniVersity Park, PennsylVania 16802, and National High Magnetic Field Laboratory, Los Alamos National Laboratory, Los Alamos, New Mexico 87545 Received April 21, 2009; Revised Manuscript Received June 10, 2009

ABSTRACT We describe the growth of Zn1-xMnxSe nanowires in ultrahigh vacuum seeded by Au nanodroplets. Electron microscopy reveals the formation of single-crystal c-axis wurtzite nanowires (typically 1-3 µm long) with Mn concentrations up to x ≈ 0.6, accompanied by a dense horizontal undergrowth of shorter, crooked nanowires. Magnetophotoluminescence measurements show evidence for sp-d exchange effects in a reduced symmetry environment. We find that the optical emission is surprisingly dominated by the undergrowth of crooked nanowires.

Contemporary interest in semiconductor spintronics1 provides a strong motivation for studying magnetic semiconductor quantum structures that incorporate local moments (e.g., d electrons in transition-metal atoms) in bulk geometries as well as in reduced dimensional structures that exploit confinement effects.2 The sp-d exchange interaction between extended band states and local moments leads to an enhanced Zeeman splitting (typically at cryogenic temperatures), resulting in highly spin polarized electronic states. Extensive magneto-optical studies have explored consequences of the exchange-enhanced Zeeman splitting in bulk crystals,3 twodimensional quantum well structures, superlattices,2 and, more recently, zero-dimensional quantum dots.4-6 However, similar experiments aimed at the one-dimensional (1D) nanowire (NW) regime remain a challenge; for instance, spectroscopic evidence for sp-d exchange has been found in lithographically patterned magnetic semiconductor NWs (down to about 25 nm widths),7 but such “top down” fabrication methods are unable to easily access the dimensions necessary for 1D quantum confinement (NW widths e 15 nm). More recent efforts to fabricate 1D magnetic semiconductors have exploited self-organized growth of NWs via the vapor-liquid-solid (VLS) mechanism8-10 and colloidal techniques,11 often producing 1D NWs with excellent crystallinity. However, there is scant evidence in these NWs for sp-d exchange-induced phenomena (such as enhanced Zeeman splitting or carrier-mediated magnetic ordering). Here, we describe the growth of quasi-1D Zn1-xMnxSe NWs * To whom correspondence should be addressed. E-mail: nsamarth@ psu.edu. † The Pennsylvania State University. ‡ Los Alamos National Laboratory. 10.1021/nl901272q CCC: $40.75 Published on Web 07/31/2009

 2009 American Chemical Society

with diameters as small as d ≈ 10 nm and with Mn compositions as high as x ≈ 0.6. The growth of these relatively straight NWs is accompanied by the formation of a nanostructured undergrowth that largely consists of crooked (and more horizontal) NWs whose dimensions are also in the 1D regime. Magnetophotoluminescence (magneto-PL) studies of as-grown samples show Zeeman shifts and circular polarization characteristics that are consistent with emission from highly spin-polarized electronic states in a nanostructured environment and therefore a strong sp-d exchange interaction. However, control measurements reveal that the emission (and thus the magneto-optical activity) is surprisingly dominated by the undergrowth of crooked NWs. We grew the NWs in an EPI620 molecular beam epitaxy chamber under ultrahigh vacuum conditions, using Si (111) and GaAs (001) substrates coated with gold as the seed particle material. We first etched the Si substrates with a 10:1 buffered oxide etch for 10 s to remove the native oxide layer and then used thermal evaporation to deposit a gold film of thickness 0.5-3 nm. The epiready GaAs substrates were not etched prior to gold deposition. After gold deposition, we annealed the substrates for 1 min at 500 °C in a rapid thermal annealer. In the growth chamber, the substrates were either directly heated to the growth temperature or first heated to 500 °C for approximately 5 min before the temperature was lowered to the growth temperature. (The two protocols yielded similar results.) All of the samples described in this paper were grown at a substrate temperature of 370 °C under Zn-rich conditions; the beam equivalent pressure for both Zn and Se was ∼2 × 10-7 Torr. When

standard sonication procedure. Low-resolution measurements used a Philips EM420T TEM with a 120 kV accelerating voltage, while high-resolution imaging and scanning TEM (STEM) work employed a JEOL EM-2010F field emission TEM/STEM running at a 200 kV accelerating voltage. TEM reveals that the NWs are largely single crystal, though some show a number of stacking faults. Diffraction pattern analysis shows that the NWs form in the wurtzite (hexagonal) crystal structure, with the growth direction along [001] (Figure 1c and d). We studied the composition of the NWs using X-ray energy dispersive spectroscopy (EDS) on the JEOL 2010F in the STEM mode. Since we found the NWs to be sensitive to beam damage, the beam was rastered across the wire or a segment of the NW, rather than holding the beam in a single spot.

Figure 1. (a) Plan-view FESEM image of as-grown NWs on a Si(111) substrate. (b) TEM image of single Mn-doped NW. The wire diameter is ∼12 nm. (c) High-resolution TEM image of a single NW showing lattice fringes. (d) Electron diffraction pattern from the same wire as that in (c), revealing a wurtzite crystal structure of the wire. The NW axis is along the [001] direction. (e) Plan-view FESEM image of a postsonicated sample in which NWs were grown on a Si(111) substrate. The image shows a dense network of short, kinked NWs not removed by the sonication process.

incorporating Mn, we varied the Mn flux from ∼1 × 10-9 to ∼5 × 10-9 Torr. Growth times varied from 30 min to 2 h. After growth, we carried out initial morphological characterization using a LEO 1530 field emission scanning electron microscope (FESEM). The NWs grow at an angle to the substrate plane, in a random array. On the GaAs substrates, we sometimes observe a partial ordering of the NWs along the [010] directions, when viewed from above. No such ordering is seen on the Si substrates (Figure 1a). We observe a variety of NW sizes, ranging from ∼10 to 100 nm in diameter. We also find variation in the lengths of the NWs, with the longest wires typically having lengths of ∼1-3 µm. The maximum NW length, however, is not linearly related to growth time; for instance, with a 30 min growth, the longest NWs are typically ∼1-1.5 µm long, while for a 2 h growth, the maximum length is only ∼3 µm. This suggests that the surface diffusion of adatoms plays a key role during NW formation in the high-vacuum conditions used in MBE growth. This speculation is consistent with our observation that many NWs have a slight taper from the base to the tip. The straight NWs were characterized by transmission electron microscopy (TEM); these measurements were carried out after releasing sonicated NWs onto grids using a Nano Lett., Vol. 9, No. 9, 2009

We performed EDS measurements on several NWs from each sample, using a Zn1-xMnxSe epilayer with x ) 0.1 as a calibration standard. Over the range of Mn cell temperatures used (761 °C e TMn e 806 °C), we obtained Zn1-xMnxSe NWs with 0.02 e x e 0.60. We note that Zn1-xMnxSe epilayers grown with similar Mn beam fluxes show significantly smaller Mn composition; for instance, the calibration Zn1-xMnxSe epilayer grown using TMn ) 796 °C yielded a Mn composition of x ) 0.1, while NWs grown at an even lower flux (TMn ) 786 °C) have Mn compositions in the range of 0.15 e x e 0.35 (Figure 2a). As shown in Figure 2a, although we observe a statistical trend of increasing Mn incorporation with increasing Mn cell temperature, we find a significant variation in the Mn composition between individual NWs in a given growth. In addition, we often observe variations in the Mn composition along the axis of an individual NW; for instance, Figures 2c-e shows the EDS spectra from three different sections of a given NW (Figure 2b), clearly indicating variations in the Mn composition. This does not appear to be simply a thickness effect due to the tapering of the wires; we took spectra at multiple points on the calibration sample, and there was no trend with increasing thickness. Further, while some wires show a decrease in Mn composition from base to tip, others are consistent throughout or even have a maximum somewhere other than the base. In order to further confirm the Mn composition, electron energy loss spectroscopy (EELS) was performed simultaneously with the EDS measurements on selected wires. While precise quantification of the EELS spectra was not practical, the rough quantification that we performed was consistent with the EDS measurements. In addition to the straight NWs that grow at an angle to the substrate plane, we also observe a substantial nanostructured undergrowth of material, largely consisting of relatively short, very kinked horizontal NWs (see the FESEM image in Figure 1e). These NWs adhere quite strongly to the substrate and are difficult to remove during sonication. Thus far, we have been unable to carry out any systematic structural and compositional analysis of the undergrowth. However, as we point out later in the manuscript, our magneto-optical measurements strongly suggest that these 3143

Figure 3. (a) Circular-polarization resolved PL spectrum of an asgrown Zn1-xMnxSe NW sample at +7, ∼0, and -7 T, revealing a substantial energy shift and polarization effects. Data taken at T ) 4 K. (b) Circular-polarization resolved PL spectrum an of as-grown ZnSe NW sample at +7, ∼0, and -7 T, showing no discernible magneto-optical effects. Data taken at T ) 4 K. (c) Circularpolarization resolved PL spectrum of a control Zn1-xMnxSe epilayer sample at +7, ∼0, and -7 T, showing complete circular polarization of the emission. Data taken at T ) 4 K.

Figure 2. (a) Mn percentage of individual wires from samples grown with three different Mn cell temperatures. Distinct symbols correspond to different NWs. (b) TEM image of a NW grown with a Mn cell temperature of 806 °C. (c) EDS spectra from the indicated sections of the NW in (b) showing variations in the Mn KR1 line relative to the KR1 and Kβ1 lines of Zn and Se.

NWs also contain Mn, but perhaps with a very different composition compared with the straight NWs. Having established the crystal structure and composition of our straight NW samples, we address an important issue: does the incorporation of Mn result in an observable sp-d exchange interaction? In principle, magneto-PL spectroscopy can provide the most direct answer to this question by revealing the exchange-enhanced Zeeman splitting of ground-state excitons. This Zeeman splitting is empirically given by a “modified” Brillouin function,3 ∆Es ) gµBB +(∆E)satB5/2[(5µBB)/kB(T + T0)], where BS(x) is a Brillouin function for S ) 5/2 and T0 is an effective temperature that accounts for the antiferromagnetic d-d exchange. In 2D and 3D, the decay of the lowest-energy excitons involves an electric-dipole-allowed transition between the Zeeman split conduction band and heavy hole (hh) band states; the resulting PL signal is circularly polarized and shows a strong red shift with increasing magnetic field. In the II-VI magnetic semiconductors, the Zeeman splitting of the excitonic states is large (∼10 meV/T at low temperatures), while spin-flip scattering is rapid (tens of picoseconds) compared to typical radiative 3144

recombination times (nanoseconds); thus, the large Zeeman red shift of the PL is accompanied by complete circular polarization. In 1D and 0D nanostructures, both the polarization characteristics and the red shift are expected to be strongly modified by the reduced symmetry which affects both the electronic structure and dielectric boundary conditions.12 We searched for sp-d exchange effects by studying the magneto-PL from as-grown NW samples. These measurements were performed in an Oxford Spectramag liquid helium magneto-optical cryostat, with excitation provided by a 405 nm diode laser. We used the Faraday geometry, with the magnetic field and light propagation collinear to each other and normal to the sample surface. We sent the PL signal through a circular analyzer, collecting spectra at a fixed setting of the circular analyzer as we swept the magnetic field B between +7 and -7 T; measurements at positive (negative) fields yielded the σ+ (σ-) components of the emitted light. As shown in Figure 3a, the emission spectrum from the as-grown Zn1-xMnxSe NWs is dominated by a disorder broadened peak near the band edge at ∼2.75 eV, most likely due to recombination of bound excitons; we also observe a broad-band emission centered at around 2.1 eV that is consistent with emission from intraionic 4T1 f 6 A1 Mn2+ transitions. The PL from as-grown Zn1-xMnxSe NW samples shows a substantial Zeeman shift, accompanied by a modest degree of circular polarization of ∼30%. As expected, control measurements on as-grown ZnSe NW samples show neither the strong magneto-optical effects nor the Mn-related emission (Figure 3b). Additionally, as Nano Lett., Vol. 9, No. 9, 2009

polarization is much smaller than the 100% circular polarization of the PL from the epilayer sample (not shown).

Figure 4. (a) Zeeman shift of the near-band-edge PL peak of the three NW samples and the epilayer whose spectra are shown in Figure 3a and c. Data taken at T ) 4 K. The solid lines show modified Brillouin function fits to the data. (b) Temperature variation of the Zeeman shift of the near-band-edge PL peak of the NW sample grown with a Mn cell temperature of 786 °C. (c) Polarization of the NW PL plotted as a function of the Zeeman shift. Data taken at T ) 4 K.

anticipated, control measurements on a bulk Zn1-xMnxSe epilayer with x ) 0.1 (Figure 3c) show a large Zeeman shift as well as complete circular polarization of the PL, with only the σ+ component of the circular polarization observable at even modest magnetic fields. The Zeeman shift of the PL from the control epilayer is well-described by a modified Brillouin function (Figure 4a). The Zeeman shift of the near-band-edge NW emission also roughly follows this modified Brillouin function (Figure 4a), clearly indicating the presence of sp-d exchange. We note however that the fits are not perfect, showing deviations particularly at low fields. The temperature dependence of the Zeeman shifts in the NW emission also behaves qualitatively as expected, decreasing with increasing temperature (Figure 4b). However, the low-field deviation from the modified Brillouin function becomes more pronounced at higher temperatures. Finally, we note that the degree of circular polarization of the NW emission saturates at a value that is considerably smaller than 100%, so that both σ+ and σ- components are always observed in a given magnetic field, despite the very large Zeeman shift. This behavior of the NW polarization is summarized in Figure 4c which plots the circular polarization P versus the Zeeman shift ∆E. (We define the polarization as P ) (Iσ+ - Iσ-)/(Iσ+ + Iσ-), where Iσ+ and Iσ- are the peak intensities of the respective circular polarization resolved components.) Note that the NW Nano Lett., Vol. 9, No. 9, 2009

The reduced circular polarization of the NW PL, despite a large Zeeman shift, strongly suggests that the PL signal originates from sp-d interactions in a reduced symmetry environment. In 3D or 2D systems in the Faraday geometry, symmetry dictates that pure angular momentum eigenstates are naturally quantized in the direction of the magnetic field (also the observation direction). Optical transitions then occur via electric dipole selection rules with a change of angular momentum of (1; hence, strong circular polarization is expected. In contrast, the shape and orientation of nanostructures, as well as noncubic crystal structures such as wurtzite, can impose other directions along which the exciton angular momentum is most naturally quantized. These axes may not lie along the magnetic field, and therefore, circular polarization resolved PL in the Faraday geometry does not probe specific spin states but rather admixtures, giving mixed polarization and P < 100% even in large magnetic fields. In addition to the effects of reduced symmetry and spatial confinement on electronic wave functions, the dielectric boundary conditions imposed by a nanostructured geometry affects the nature of the excitonic states by modifying the Coulomb interaction. For instance, the combination of all of these effects accounts well for the strong linear polarization of PL from nanorods.12 Although our magneto-PL measurements are consistent with emission from a nanostructured environment, we cannot exclusively identify it as originating from the straight NWs. To identify the source of the observed optical emission, we carried out magneto-PL measurements on samples before and after sonication; as mentioned earlier, SEM measurements showed that the sonication process removes almost all of the straight NWs, leaving behind a much higher density of crooked horizontal nanowires (Figure 1e). Surprisingly, we found that the PL spectrum in such sonicated samples showed almost an identical PL spectrum and magneto-PL behavior as the as-grown samples. We attribute the PL to efficient exciton localization in the nanostructured undergrowth, where the typical confinement length scales are similar to those in the straight NWs (∼10-20 nm). The unusual polarization characteristics in the magneto-PL would be equally consistent with the reduced symmetry of this inhomogeneous nanostructured undergrowth as well as with confinement in straight NWs. In addition, inhomogeneous Mn speciation is likely enhanced in the undergrowth (for instance, a preference for Mn incorporation at NW surfaces); this could influence the coupling with the electronic band structure, thus modifying the selection rules for optical transitions and the resultant circular polarization. We also note that the saturation value of the Zeeman shift is uncorrelated with the Mn composition (Figure 4a), unlike the behavior of 3D and 2D Zn1-xMnxSe systems, where the exchange-enhanced Zeeman splitting is a strong and nonmonotonic function of Mn composition. This reinforces our conclusion that the PL is dominated by the undergrowth and not the straight NWs. We speculate that the nanostructured undergrowth must have a relatively constant Mn content, independent of the incident Mn flux 3145

during sample growth, even though the incorporation of Mn in the straight NWs clearly increases with the Mn flux (Figure 2a). Finally, we note that similar control measurements on pre- and postsonicated ZnSe NW samples also show that the near-band-edge PL is dominated by the nanostructured undergrowth rather than by straight NWs, serving as a strong caution to earlier interpretations of optical emission from as-grown ZnSe NWs.13 In summary, we have demonstrated the growth of Zn1-xMnxSe NWs in an ultrahigh vacuum environment, resulting in single-crystal wurtzite NWs with large Mn composition (x ∼ 0.6) and with diameters approaching the 1D regime (∼10 nm). The growth of these NWs is accompanied by the simultaneous formation of a dense network of crooked, horizontal NWs that surprisingly appear to serve as an efficient region for exciton localization and thus dominate the PL signal from as-grown samples. MagnetoPL measurements clearly reveal the presence of exchangeenhanced Zeeman splitting of the electronic states. The functional form of the temperature and magnetic field dependence of the Zeeman shift, as well as the reduced degree of circular polarization, strongly suggests modifications due to a nanostructured environment. A more systematic understanding of the underlying physics will require magnetooptical measurements on single NWs. Acknowledgment. This work is supported by the Penn State Center for Nanoscale Science (funded by NSF under Grant No. DMR-0820404). This work was performed in part

3146

at the Penn State Nanofabrication Facility, a member of the NSF National Nanofabrication Users Network. Work at the National High Magnetic Field Laboratory is supported by NSF, DOE, and the State of Florida. The authors acknowledge the contributions of Tina Lin during the early stages of this project. References (1) Awschalom, D. D.; Flatté, M. E. Nat. Phys. 2007, 3, 153–159. (2) Awschalom, D. D.; Samarth, N. J. Magn. Magn. Mater. 1999, 200, 130–147. (3) Furdyna, J. K. J. Appl. Phys. 1988, 64, R29–R64. (4) Leger, Y.; Besombes, L.; Fernandez-Rossier, J.; Maingault, L.; Mariette, H. Phys. ReV. Lett. 2006, 97, 107401. (5) Beaulac, R.; Archer, P. I.; Liu, X.; Lee, S.; Salley, G. M.; Dobrowolska, M.; Furdyna, J. K.; Gamelin, D. R. Nano Lett. 2008, 8, 1197–1201. (6) Bussian, D. A.; Crooker, S. A.; Yin, M.; Brynda, M.; Efros, Al. I.; Klimov, V. I. Nat. Mater. 2009, 8, 35–40. (7) Ray, O.; Sirenko, A. A.; Berry, J. J.; Samarth, N.; Gupta, J. A.; Malajovich, I.; Awschalom, D. D. Appl. Phys. Lett. 2000, 76, 1167– 1169. (8) Radovanovic, P.; Barrelet, C.; Gradecak, S.; Qian, F.; Lieber, C. Nano Lett. 2005, 5, 1407–1411. (9) Choi, H.-J.; Seong, H.-K.; Kim, U. NANO 2008, 3, 1–19. (10) Zaleszczyk, W.; Janik, E.; Presz, A.; Dluiewski, P.; Kret, S.; Szuszkiewicz, W.; Morhange, J.-F.; Dynowska, E.; Kirmse, H.; Neumann, W.; Petroutchik, A.; Baczewski, L. T.; Karczewski, G.; Wojtowicz, T. Nano Lett. 2008, 8, 4061–4065. (11) Chin, P. T. K.; Stouwdam, J. W.; Janssen, R. A. J. Nano Lett. 2009, 9, 745–750. (12) Shabaev, A.; Efros, A. Nano Lett. 2004, 4, 1821–1825. (13) Saxena, A.; Yang, S.; Philipose, U.; Ruda, H. E. J. Appl. Phys. 2008, 103, 053109.

NL901272Q

Nano Lett., Vol. 9, No. 9, 2009