- Email: [email protected]
Atom probe tomography of nitride semiconductors
SMM-11426; No of Pages 7 Scripta Materialia xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Scripta Materialia journal homepage: www.elsevier.com/locate/scriptamat
Viewpoint article
Atom probe tomography of nitride semiconductors L. Rigutti a, B. Bonef b, J. Speck b, F. Tang c, R.A. Oliver c,⁎ a b c
Normandie Univ., UNIROUEN, INSA Rouen, CNRS, Groupe de Physique des Matériaux, 76000 Rouen, France Materials Department, University of California, Santa Barbara, CA 93106, USA Department of Materials Science and Metallurgy, University of Cambridge, 27 Charles Babbage Road, Cambridge CB3 0FS, UK
a r t i c l e
i n f o
Article history: Received 26 October 2016 Accepted 28 December 2016 Available online xxxx Keywords: Gallium nitride Atom probe tomography Light emitting diode High electron mobility transistor Microwire
a b s t r a c t Atom probe tomography (APT) has emerged as a valuable tool in the study of nitride semiconductors, despite the challenges involved in achieving controlled field evaporation. In optoelectronics, it has provided insights into the nanostructure of light emitting diodes, laser diodes and microwires. In electronics, it has allowed insights into impurity doping and alloying effects in transistors. Coupled with direct correlative studies using other techniques and theoretical modelling based on the APT data, the availability of three dimensional compositional information on nitride heterostructures has had (and will continue to have) a profound impact on the design and development of devices. © 2016 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
1. Introduction Optoelectronic and electronic devices based on gallium nitride, most notably light emitting diodes (LEDs) for energy efficient solid state lighting, already enjoy immense commercial success, but the extreme pace of device development has left understanding of the structure of these materials trailing in its wake. This is currently limiting progress on a wide range of devices, particularly green and amber LEDs [1], whose efficiency lags far behind that of their blue counterparts. One question, central to efforts to understand LED performance has endured almost since the field's inception: why are GaN-based light emitting diodes almost unaffected by the presence of defects, primarily dislocations, at densities which would entirely prohibit device operation in other III–V semiconductors [2]? In this context, the concept of localization is key: it is believed that inhomogeneities in the InGaN quantum wells (QWs) which form the active region of LEDs modulate the in plane potential preventing charge carrier diffusion to non-radiative centres at dislocation cores. For several years, these inhomogeneities were thought to be indium rich regions of the InGaN QW (“indium clusters”), which would have a locally reduced bandgap and hence trap carriers [3]. However, transmission electron microscopy (TEM) data, which formed the strongest evidence for the existence of such clusters, was called into question by researchers who noted that the appearance of clustering could be generated in TEM by exposure to the imaging electron beam, even when the QWs initially looked rather uniform [4]. Against this background, a chance meeting between a specialist in nitride materials (RAO) and an atom-probe tomography (APT) expert ⁎ Corresponding author. E-mail address: [email protected] (R.A. Oliver).
(Prof. GDW Smith) resulted in a fruitful collaboration investigating inhomogeneities in InGaN QWs using APT which kick-started widerranging research on APT of nitrides and other wide bandgap materials. Serendipitously, Smith was exploring the capabilities of recently-developed laser-pulsed atom probe systems appropriate to study of materials with comparatively low conductivity, such as wide bandgap semiconductors, where voltage pulsing was likely to prove unsuccessful. Even with laser pulsing, difficulties were anticipated in studying InGaN QWs which were grown on insulating sapphire substrates. Additionally, the use of the focussed ion beam microscope (FIB) to prepare semiconductors for APT was then in its infancy. A little additional consideration might have further discouraged these first attempts to study nitrides by APT: the laser pulsed atom probe system utilised a green laser, and in the bulk form GaN is transparent to green light, so laser heating might seem an unlikely prospect. In fact, APT of nitrides has had a remarkable success rate, confounding such early concerns, and the use of a green laser is surprisingly effective. Whilst indium clustering represented the initial focus of APT studies, APT has now been used to improve understanding of nitride materials in various ways, as summarised in Section 2. Conversely, the nitrides are also offering new opportunities to understand APT and Section 3 will address this development. 2. What APT tells us about nitrides 2.1. Quantum wells and alloy homogeneity As described above, the assessment of alloy homogeneity in blueemitting InGaN/GaN QWs was the first nitride topic to be assessed by APT. Initially, sample preparation proved challenging, with some damage to the QWs observed in the APT [5] and attributed to knock-on
http://dx.doi.org/10.1016/j.scriptamat.2016.12.034 1359-6462/© 2016 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034
2
L. Rigutti et al. / Scripta Materialia xxx (2016) xxx–xxx
damage during the deposition of a protective Pt coating in the FIB. However, an effective FIB-based sample preparation routine rapidly evolved [6] and more recent investigations have confirmed that, provided an appropriate low-voltage cleanup routine is used, there is little evidence for any FIB-induced implantation or damage [7]. Both blue-emitting and green-emitting QWs grown by metal-organic vapour phase epitaxy (MOVPE) on the polar (0001) plane of GaN were successfully imaged, and (as illustrated in Fig. 1) a frequency distribution (FD) analysis provided no evidence for a deviation from randomness in the indium distribution in the plane of the QW [8]. Later, InGaN QWs grown by molecular beam epitaxy (MBE) yielded a similar result, albeit for a limited sampling volume. Prosa et al. [9] extended this analysis to QWs grown on the semi-polar (10-1-1) plane and performed a particularly thorough analysis addressing a variety of different length scales, but still found no evidence of clustering. Further improvements to the analysis methodology have also been introduced by Tang et al. [10] who addressed the impact of the variation in indium content through the thickness of the QW on the FD analysis by developing a modified nearest neighbour analysis in which all of the atoms of a single QW are projected onto a single plane, effectively compressing the reconstructed QW image in the through-thickness direction. The original APT studies of InGaN QWs received extensive scrutiny from within the community of researchers working on InGaN [11], and the ability of APT to detect clusters was called into question. To validate the technique, it was necessary to find an InGaN sample with a genuinely non-random Indium distribution, and this was first achieved by Müller et al. [12] who studied thick, partially relaxed InGaN layers exhibiting a grossly non-random indium distribution, which was effectively imaged and analyzed in APT. In order to study QWs with a nonrandom indium distribution, Bennett et al. [13] used an electron beam to deliberately damage an APT sample in the TEM and showed using an FD analysis that deviations from randomness had thus been induced. Tang et al. [10] later also found evidence for the formation of as-grown indium clusters in a non-polar (11-20)-oriented QW sample. Both these studies validated the ability of APT to detect clustering in QWs, if present. It should also be noted that the statistical randomness of the (0001) QWs grown by MOVPE and assessed in the earliest APT
studies was also independently validated by carefully-designed TEM experiments [14]. Whilst several analyses of InGaN QWs on both the (0001) plane and the (10-1-1) plane have revealed no evidence of indium clustering, despite evidence from photoluminescence (PL) of localisation in the relevant materials [15,16], some APT studies have been used to suggest that In clustering may be present in InGaN QWs gown under other conditions. Unfortunately, for QWs grown on (0001) none of these studies use an effective statistical method to distinguish between random local variations in the indium content and non-random clustering. Jang et al. [17] have however shown that changes to the indium distribution do occur upon raising the pressure used during QW growth. Given that APT suggests that In clusters are not in general the origin of localisation, one must consider what alternative localisation centres may exist. APT data, supported by both TEM and X-ray diffraction, suggests that typical InGaN QWs exhibit fluctuations in thickness of 1 to 2 monolayers on a lateral length scale of only a few nanometres [6,18]. Such thickness fluctuations represent an alternative possible localisation centre and the relevant APT data has been used to model carrier localization using both a continuum [15] and an atomistic approach [19]. Both of these methodologies suggest that the most important localisation centres are in fact random fluctuations in the indium content, although well-width fluctuations sometimes also influence the electron localisation. (Similarly, APT studies of GaN/ AlGaN QWs reveal the AlGaN to be a random alloy and a continuum model, which correlates well to PL data, suggests that random compositional fluctuations can lead to carrier localisation [20]). These studies exemplify the increasing use of APT data to allow the development of physical models of nitride structures which accurately reflect the microstructure. Similarly, APT data has been used to inform models of carrier transport in LEDs and other devices [21,22,23]. 2.2. Nano- and microwires APT has also been applied to the study of III-N nano- and microwires. Nanowires with diameter smaller than ~150 nm can be analyzed without FIB preparation [24,25,26,27]. The shank angle of nanowires being
Fig. 1. Analysis of indium distributions in a blue emitting multiple QW (a) Side-on view of 3D reconstruction of 5 InGaN/GaN QWs, showing 80% of the In atoms and 10% of the Ga atoms (Grid box 45 nm × 45 nm × 60 nm). (b) A representative Frequency Distribution (FD) analysis of the 3rd QW with a 50-atom bin size comparing the observed indium distributions with the binomial distribution for the random case. (c) Plot of the difference between the observed and expected distributions in (b). This provides no evidence of non-random clustering, in keeping with the results of the relevant χ2 analysis which yields a p-value of 0.47.
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034
L. Rigutti et al. / Scripta Materialia xxx (2016) xxx–xxx
very close to zero has important implications for APT studies focussing on compositional accuracy issues, as described in Section 3. The relationship between structural and functional properties is the focus of studies of larger nanowires [28] or microwires [29,30] containing nanoscale LED structures, prepared by FIB. The InGaN QWs in these systems are grown under significantly different conditions than those in thin films, with possible effects on the incorporation of different species in heterostructures. The APT study reported by Riley et al. [28] on nonpolar and semi-polar QWs in core-shell nanowires identified compositional gradients which could be related to the spatially resolved CL spectra obtained on separate individual nanowire cross sections. In the studies reported by Rigutti et al. [26] and by Mancini et al. [30] on the m-plane InGaN QWs extracted from MOCVD-grown microwires, a PL signal of nanoscale volumes could be extracted despite the damage produced by FIB preparation and was correlated to the structural information obtained later by STEM and APT on approximately the same volume. Fig. 2(a) shows a STEM micrograph of the same field emission tip analyzed by APT and reconstructed in Fig. 2(b). The 3D analysis of
3
the highly non-uniform In distribution found in these QWs correlated well with the PL spectrum (Fig. 2(c)) obtained from the same structure prior to FIB preparation and APT analysis [29]. The accuracy of the correlation could then be improved with the extraction of the PL signal directly from an APT specimen tip and with the consideration of the electronic properties of the stacking faults evidenced by HR-STEM analysis [30], which underlines the importance of direct correlative studies using different techniques, sometimes referred to as “multi-microscopy”, on the same nanoscale structure. 2.3. High electron mobility transistors Nitride based high electron mobility transistors (HEMT) have recently been developed for high frequency and high power electronics. In conventional Ga polar (0001) AlGaN/GaN HEMTs, a two dimensional electron gas (2DEG) forms at the interface between AlGaN and GaN due to a discontinuity in total polarization. Alloy scattering in AlGaN barriers in transistor structures strongly impacts mobility [31]. The addition of
Fig. 2. Correlative μPL, STEM and APT study on a set of m-plane InGaN MQWs extracted from a microwire structure. (a) STEM micrograph of an APT specimen; (b) APT-reconstructed positions of In atoms within the same tip specimen; (c) Correlation of the optical and structural properties of the nanoscale specimen. The blue region highlights the typical In content measured by APT 3D analysis within the quantum wells, whilst the red region corresponds to region in which the peak energies of the μPL spectrum (also shown on the left hand side of the plot) are found (Reprinted with permission from ref. [29] American Chemical Society [29]).
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034
4
L. Rigutti et al. / Scripta Materialia xxx (2016) xxx–xxx
an AlN interlayer, which minimizes the penetration of the electron wavefunction into the alloy, improves mobility for both Ga polar (0001) and N-polar (000-1) orientations [31,32,33,35], the purity of which was examined by APT, showing that whilst typically Ga was unintentionally incorporated into the interlayer, the aluminium site fraction was much higher in samples grown by MBE than those grown by MOVPE [31,32,33]. Interfacial abruptness also profoundly influences transport in HEMTs, and APT has revealed that it is affected by the material's polarity (Ga polar or N polar) [34] and that the two interfaces of the interlayer are not equally diffuse. InAlN has also attracted attention for use as a barrier layer in HEMTs due to the advantage of being able to achieve lattice matching with GaN for an In fraction of 0.18 and its high spontaneous polarization charge. However, the large lattice constant mismatch (13%) and bond strength difference (0.9 eV per bond) between AlN and InN favour phase separation and In segregation in the lateral and vertical directions. This nonuniformity could be source of scattering, leading to mobility degradation in HEMTs. Phase separation resulting in a vertical honeycomb structure in InAlN/GaN heterostructures grown by MBE and MOVPE was first demonstrated by TEM [35,36,37] but projection effects made quantification difficult. APT of In0.18Al0.82N grown by plasma assisted MBE [38] revealed the 3D atomic distribution consisted of columnar cells of 10 nm lateral dimension as shown in Fig. 3(a, b), which, as one might expect, is shown to be statistically non-random by an FD analysis (Fig. 3(c)). At the boundaries of the cells the indium fraction reaches ~0.26, with corresponding depletion of In at the centre of the cells. The elimination of the columnar microstructure was demonstrated using FD analysis of APT data after optimization of the growth conditions [39] and a homogenous In0.17Al0.83N barrier was used in a HEMT which showed promising electrical properties [40]. A second major issue with growth of AlInN in some systems is unintentional incorporation of Ga [41]. APT is particularly powerful in detecting unintentional Ga incorporation where the resulting concentration is low. For example, Tang et al. observed a tail of unintentionally incorporated Ga below 1% in atomic fraction penetrating
into a thick InAlN layer [7] which was attributed to the presence of residual Ga in the growth environment. 2.4. Defects Given the high density of defects found in nitride materials compared to conventional III–V semiconductors and the potential importance of their local compositional environment in understanding their impact on device performance, APT has been relatively sparsely applied to the study of defects in the nitrides, despite very interesting results which have been achieved in group IV semiconductors [42]. Nonetheless, isolated examples exist of studies of point, line, areal and volume defects. Considering impurities as a form of point defect, APT can contribute significantly to the study of dopants in devices. The typical n- and p-type dopants in GaN are Mg and Si respectively, but due to peak overlaps between Si+ and N+ 2 in the mass spectrum, it has thus far been impossible to study Si in APT. However, Mg in p-type layers at concentrations around 1019 cm−3 (~100 ppm) has been successfully addressed [13]. Unintentionally incorporated impurities can also act as dopants, with oxygen being a common unintentional n-dopant. Young et al. [43] studied oxygen incorporation into tunnel junctions grown using a combination of MOVPE and MBE. Surprisingly, the presence of the impurity at the interface between material grown by the two techniques was found to enhance the device performance. Accurate APT quantification of the oxygen content in a very thin layer at this interface was key to understanding this effect. Whilst studies of dopants and unintentional impurities are core strengths of APT, imaging extended defects is more challenging. Dislocations are of particular interest, but some research groups have noted that the success rate in APT imaging of samples known to contain dislocations is far lower than for dislocation free material [44]. The one published data set [45] relevant to dislocations actually shows one side of a V-pit, a structure which forms when the growth of QWs is distorted by the presence of a dislocation forming a pit with the dislocation at its apex and QWs on the sidewalls. The dislocation itself was not part of the data set, and indeed other V-pit analyses [44] suggest that samples
Fig. 3. (a) 2D In and Al atomic map showing the honeycomb microstructure in InAlN. (b) Associated In site fraction map showing the segregation of In into concentrated regions at the boundaries of the cells. (c) Statistical distribution analysis performed with a sampling volume of 100 atoms highlighting the gap between the honeycomb microstructure and a random alloy.
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034
L. Rigutti et al. / Scripta Materialia xxx (2016) xxx–xxx
fracture when the dislocation line intersects the evaporation front. Nonetheless, the analysis showed that QWs on the V-pit sidewalls are both thinner and lower in indium content than those away from the defect, which would result in a locally higher potential, acting to isolate carriers from the defect, in agreement with a model for the robustness of nitride LEDs to dislocations previously proposed by Hangleiter et al. [46]. Rather more success has been achieved in the imaging of basal plane stacking faults (BSFs) [30] and inversion domains (IDs) [47]. In both cases, comparison between APT and TEM data was key to understanding the APT observation, with TEM providing vital information about the crystallographic nature of the defect. For the IDs [47], which were observed in a p-doped AlGaN/GaN superlattice, APT revealed regions of significantly elevated Mg content a few nanometres across, which were correlated with (on a separate region of the same wafer) pyramidal IDs seen in TEM with the same density, distribution and size. APT was able to provide quantitative information about the Mg content, whereas TEM-based compositional analysis failed to detect even the presence of Mg. For the BSFs [30], exactly the same region of the same microwire was imaged first in TEM and then in APT. The BSF intersected QWs on the microwire sidewall but did not observably disturb their compositions. However, the BSF acts as a lower band-gap zinc-blende inclusion in a wurtzite matrix, and this effects the local PL, as was
5
shown in studies performed again on exactly the same material, another example of the multi-microscopy approach highlighted in Section 2.2. 3. What nitrides tell us about APT During APT studies of nitrides it became evident that the technique does not generally yield an accurate measurement of the elemental composition. Due to the relatively high rate of success for the analyses and to the straightforward interpretation of their mass spectra, III-N materials have become a model system for the study of compositional biases in APT, which may occur in a much broader class of compounds. Studies conducted on III-N materials highlighted the need to further explore the field evaporation behavior of compound semiconductors in order to (i) assess whether there is a bias in the compositional measurement, (ii) understand the physical mechanisms inducing any bias and (iii) correct said bias, if possible. The first studies performed on GaN nanowires have shown that the composition measured by APT may change from almost 100% to under 50% of Ga as the impinging laser energy is reduced and, simultaneously, the constant applied voltage is raised. Moreover, the measured Ga atomic fraction correlates well with the charge state ratio Ga2+/Ga+, an indirect indicator of the surface field [25,26,27] according to the post-ionization theory [48]. Fig. 4(a) illustrates further studies of GaN
Fig. 4. (a) Correlation between the Ga atomic fraction measured by APT in GaN on the Ga2+/Ga1+ charge state ratio,. Large open symbols refer to average quantities extracted from datasets containing 105 events at constant detection rate or at constant impinging laser energy, whilst the small are issued by the detector statistics of the charge state ratio (b) Ga2+/Ga1+ and of the Ga atomic fraction (c). (d) Reconstructed 3D positions of the Al atoms in a GaN/AlGaN multi-QW structure. (e) Correlation between the Ga2+/Ga1+ charge state ratio and the measured Al III-site fraction. Large open symbols refer to dataset extracted from single barriers, whilst the small filled symbols refer to the detector statistics extracted from the bottom region indicated in (d).
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034
6
L. Rigutti et al. / Scripta Materialia xxx (2016) xxx–xxx
carried out either by performing separate analyses at constant detection rate or by varying one experimental parameter at once (laser intensity or DC applied bias): the measured Ga atomic fraction correlates to the surface field, which can be estimated through the charge state ratios (Ga2 +/Ga1 +). The detector statistics of the Ga2 +/Ga+ charge state ratio (Fig. 4(b)) show that the surface field is non-uniform across the different regions of the APT specimen apex due to the specimen geometry and crystallography, and the measured atomic fraction of Ga (Fig. 4(c)) depends on the microscopic surface field distribution. Similar dependences on the surface electric field have been found in other wide bandgap binary systems, such as AlN, ZnO and MgO [49,50]. This behavior can be partially explained through preferential evaporation of the metallic element at high field, but the mechanism responsible of the loss of N or O at low field is still under debate. Neutral evaporation or production of neutral N and O upon dissociation of more complex molecules have been proposed [51,52,53,54]. Whilst the above-mentioned binary systems have a known composition, this is not the case for ternary alloys and doped materials. In these technologically important cases, is essential to know under which conditions APT can accurately measure the composition. This problem has been assessed for InyGa1-yN and AlyGa1-yN. Riley et al. [55] and Mancini et al. [49] provided evidence that the In III-site fraction in InGaN is fairly independent of the experimental conditions. However, data on Al0.25Ga0.75N (Fig. 4(d)) showed that the measured Al site fraction deviates from the correct value (y = 0.25) at high field Fig. 4(e). The accurate site fraction of AlGaN may thus be determined performing the measurement at low field or, equivalently, at sufficiently high laser pulse intensity. This, however, could imply a degradation of spatial resolution due to the temperature increase and to the possible modification of tip shape with increasing intensity. To overcome this, a statistical correction scheme was developed, in order to restore the spatial behavior and the amplitude of the alloy fluctuations even in the case of biased measurements. The correction procedure has been tested for Al0.25Ga0.75N by measuring the emission and localization energy of the photons emitted within the alloy, two quantities that depend on the average site fraction and on the alloy fluctuations [56]. 4. Concluding remarks and future outlook This review has illustrated the widespread and successful use of APT in analysing nitride semiconductors. One key theme which has emerged is the important of direct correlative studies, sometimes called “multimicroscopy”, in which a specific nanostructure or region of material is assessed using other nanoscale characterisation techniques, such as TEM or micro-PL, prior to APT studies. This is a challenging but very powerful approach which makes direct links between the structure, composition and properties of a material at the nanoscale, and should see increasing application in the future, utilising a wide range of characterisation techniques, as is increasingly being achieved in other contexts [57]. The ability to directly compare detailed structure and properties, allows for testing and validation of physical models of nanostructures without the uncertainty inherent in comparing different structures within arrays which may exhibit significant statistical variations. The availability of such data sets as an input for modelling raises the possibility of directly inputting APT data into atomistic calculations, although challenges arise because the atoms in the APT data set do not in general sit on crystallographically defined sites (lattice fringe resolution in APT has yet to be achieved in the nitrides) and less than half the atoms present in the original sample are typically present in the data set. The development of protocols to make best use of APT data in developing atomistic models within the limitations it poses will require input from both theoreticians and experimentalists. Other possible future challenges for nitride APT concern improving its effectiveness in assessing impurities and point defects. It would be difficult to use APT to study point defects such as nitrogen vacancies because of both the overall limits of resolution and sampling of
the technique (applicable to all materials), and the difficulty in measuring reasonable stoichiometries in the nitrides in all samples (see Section 3). One interesting challenge for future APT studies would be the quantification of hydrogen incorporation, particularly in MOVPE-grown p-type GaN where hydrogen is known to act to compensate the Mg acceptors, reducing conductivity. For n-type doping, questions remain about the role of defects in dopant segregation, but these can only effectively be addressed if the peak overlap between Si + and N 22 + is addressed, perhaps by studying samples grown using a source of the Si29 or Si30 isotopes, or by using high mass resolution instruments to address the low concentration of naturally occurring Si29 and Si30 . Addressing these issues will contribute to the usefulness of APT in analysing full device structures. Such studies will in the future increasingly address complete devices including the metal and/or dielectric layers and their interface to the nitride semiconductor. Overall, the increasing availability of APT and it's great strengths in providing nanoscale three dimensional compositional information with sufficient sensitivity to address dopants and impurities will lead further increases in its application in the nitrides and other wide bandgap materials. Acknowledgements This work was supported in part by the European Research Council under the European Community's Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement no 279361 (MACONS). References [1] R.A. Oliver, Mat. Sci. Technol. 32 (2015) 737. [2] R.A. Oliver, S.E. Bennett, T. Zhu, D.J. Beesley, M.J. Kappers, D.W. Saxey, A. Cerezo, C.J. Humphreys, J. Phys. D 43 (2010) 354003. [3] Y. Narukawa, Y. Kawakami, M. Funato, S. Fujita, S. Fujita, S. Nakamura, Appl. Phys. Lett. 70 (1997) 981. [4] T.M. Smeeton, M.J. Kappers, J.S. Barnard, M.E. Vickers, C.J. Humphreys, Appl. Phys. Lett. 83 (2003) 5419. [5] M.J. Galtrey, R.A. Oliver, M.J. Kappers, C. McAleese, D. Zhu, C.J. Humphreys, P.H. Clifton, D. Larson, A. Cerezo, Appl. Phys. Lett. 92 (2008) 41904. [6] M.J. Galtrey, R.A. Oliver, M.J. Kappers, C.J. Humphreys, D.J. Stokes, P.H. Clifton, A. Cerezo, Springer Proc. Phys. 120 (2007) 161. [7] F. Tang, M.P. Moody, T.L. Martin, P.A.J. Bagot, M.J. Kappers, R.A. Oliver, Microsc. Microanal. 21 (2015) 544. [8] M.J. Galtrey, R.A. Oliver, M.J. Kappers, C.J. Humphreys, P.H. Clifton, D. Larson, D.W. Saxey, A. Cerezo, J. Appl. Phys. 104 (2008) 13524. [9] T.J. Prosa, P.H. Clifton, H. Zhong, A. Tyagi, R. Shivaraman, S.P. DenBaars, S. Nakamura, J.S. Speck, Appl. Phys. Lett. 98 (2011) 191903. [10] F. Tang, T. Zhu, F. Oehler, W.Y. Fu, J.T. Griffiths, F.C.-P. Massabuau, M.J. Kappers, T.L. Martin, P.A.J. Bagot, M.P. Moody, R.A. Oliver, Appl. Phys. Lett. 106 (2015) 72104. [11] C. Kisielowski, T.P. Bartel, Appl. Phys. Lett. 91 (2007) 176101. [12] M. Müller, G.D.W. Smith, B. Gault, C.R.M. Grovenor, Acta. Mater. 60 (2012) 4277. [13] S.E. Bennett, D.W. Saxey, M.J. Kappers, J.S. Barnard, C.J. Humphreys, G.D.W. Smith, R.A. Oliver, Appl. Phys. Lett. 99 (2011) 21906. [14] K.H. Baloch, A.C. Johnston-Peck, K. Kisslinger, E.A. Stach, S. Gradečak, Appl. Phys. Lett. 102 (2013) 191910. [15] D. Watson-Parris, M.J. Godfrey, P. Dawson, R.A. Oliver, M.J. Galtrey, M.J. Kappers, C.J. Humphreys, Phys. Rev. B 83 (2011) 115321. [16] P. Dawson, S. Schulz, R.A. Oliver, M.J. Kappers, C.J. Humphreys, J. Appl. Phys. 119 (2016) 181505. [17] D.H. Jang, G.H. Gu, B.H. Lee, C.G. Park, Ultramicroscopy 127 (2013) 114. [18] G.H. Gu, D.H. Jang, K.B. Nam, C.G. Park, Microsc. Microanal. 19 (2013) 99. [19] S. Schulz, D.P. Tanner, E.P. O'Reilly, M.A. Caro, T.L. Martin, P.A.J. Bagot, M.P. Moody, F. Tang, J.T. Griffiths, F. Oehler, M.J. Kappers, R.A. Oliver, C.J. Humphreys, D. Sutherland, M.J. Davies, P. Dawson, Phys. Rev. B 92 (2015) 235419. [20] L. Rigutti, L. Mancini, W. Lefebvre, J. Houard, D. Hernàndez-Maldonado, E. Di Russo, E. Giraud, R. Butté, J. Carlin, N. Grandjean, Semicond. Sci. Tech. 31 (2016) 095009. [21] Y.-R. Wu, R. Shivaraman, K.-C. Wang, J.S. Speck, Appl. Phys. Lett. 101 (2012) 083505. [22] D.A. Browne, B. Mazumder, Y.-R. Wu, J.S. Speck, J. Appl. Phys. 117 (2015) 185703. [23] D.N. Nath, Z.C. Yang, C.Y. Lee, P.S. Park, Y.R. Wu, S. Rajan, Appl. Phys. Lett. 103 (2013) 022102. [24] N.A. Sanford, P.T. Blanchard, M. Brubaker, K.A. Bertness, A. Roshko, J. Schlager, R. Kirchhofer, D. Diercks, B. Gorman, Phys. Status Solidi (c) 11 (2014) 608. [25] R. Agrawal, R.A. Bernal, D. Isheim, H.D. Espinosa, J. Phys. Chem. C 115 (2011) 17688–17694. [26] J.R. Riley, R.A. Bernal, Q. Li, H.D. Espinosa, G.T. Wang, L.J. Lauhon, ACS Nano 6 (2012) 3898–3906. [27] D.R. Diercks, B.P. Gorman, R. Kirchhofer, N. Sanford, K. Bertness, M. Brubaker, J. Appl. Phys. 114 (2013) 184903.
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034
L. Rigutti et al. / Scripta Materialia xxx (2016) xxx–xxx [28] J.R. Riley, S. Padalkar, Q. Li, P. Lu, D.D. Koleske, J.J. Wierer, G.T. Wang, L.J. Lauhon, Nano Lett. 13 (2013) 4317–4325. [29] L. Rigutti, I. Blum, D. Shinde, D. Hernandez-Maldonado, W. Lefebvre, J. Houard, F.O. Vurpillot, A. Vella, M. Tchernycheva, C. Durand, Nano Lett. 14 (2013) 107–114. [30] L. Mancini, D. Hernández-Maldonado, W. Lefebvre, J. Houard, I. Blum, F. Vurpillot, J. Eymery, C. Durand, M. Tchernycheva, L. Rigutti, Appl. Phys. Lett. 108 (2016) 042102. [31] M.N. Fireman, D.A. Browne, B. Mazumder, J.S. Speck, U.K. Mishra, Appl. Phys. Lett. 106 (2015) 202106. [32] B. Mazumder, S.W. Kaun, J. Lu, S. Keller, U.K. Mishra, J.S. Speck, Appl. Phys. Lett. 102 (2013) 111603. [33] S.W. Kaun, B. Mazumder, M.N. Fireman, E.C.H. Kyle, U.K. Mishra, J.S. Speck, Semicond. Sci. Tech. 30 (2015) 055010. [34] B. Mazumder, M.H. Wong, C.A. Hurni, J.Y. Zhang, U.K. Mishra, J.S. Speck, Appl. Phys. Lett. 101 (2012) 091601. [35] D.S. Katzer, D.F. Storm, S.C. Binari, B.V. Shanabrook, A. Torabi, L. Zhou, D.J. Smith, J. Vac, Sci. Technol. B 23 (2005) 1204. [36] S. Kret, A. Wolska, M.T. Klepka, A. Letrouit, F. Ivaldi, A. Szczepanska, N. Grandjean, J. Phys. Conf. Ser. 326 (2011) 012013. [37] L. Zhou, D.J. Smith, M.R. McCartney, D.S. Katzer, D.F. Storm, Appl. Phys. Lett. 90 (2007) 081917. [38] S. Choi, F. Wu, R. Shivaraman, E.C. Young, J.S. Speck, Appl. Phys. Lett. 100 (2012) 232102. [39] E. Ahmadi, R. Shivaraman, F. Wu, S. Wienecke, S.W. Kaun, S. Keller, U.K. Mishra, Appl. Phys. Lett. 104 (2014) 072107. [40] S.W. Kaun, E. Ahmadi, B. Mazumder, F. Wu, E.C.H. Kyle, P.G. Burke, J.S. Speck, Semicond. Sci. Tech. 29 (2014) 045011. [41] A. Redondo-Cubero, K. Lorenz, R. Gago, N. Franco, M.A. di Forte Poisson, E. Alves, E. Munos, J. Phys. D: Appl. Phys. 43 (2010) 055406.
7
[42] D. Blavette, S. Duguay, J. Appl. Phys. 119 (2016) 181502. [43] E.C. Young, B.P. Yonkee, F. Wu, S.H. Oh, S.P. DenBaars, S. Nakamura, J.S. Speck, Appl. Phys. Express 9 (2016) 022102. [44] S. Bennett, Nitride Semiconductors Studied by Atom Probe Tomography and Correlative TechniquesPhD Thesis University of Cambridge, 2010. [45] S. Tomiya, Y. Kanitani, S. Tanaka, T. Ohkubo, K. Hono, Appl. Phys. Lett. 98 (2011) 181904. [46] A. Hangleiter, F. Hitzel, C. Netzel, D. Fuhrmann, U. Rossow, G. Ade, P. Hinze, Phys. Rev. Lett. 95 (2005) 127402. [47] S.E. Bennett, R.M. Ulfig, P.H. Clifton, M.J. Kappers, J.S. Barnard, C.J. Humphreys, R.A. Oliver, Ultramicroscopy 111 (2011) 207. [48] D.R. Kingham, Surf. Sci. 116 (1982) 273. [49] L. Mancini, N. Amirifar, D. Shinde, I. Blum, M. Gilbert, A. Vella, F.O. Vurpillot, W. Lefebvre, R. Lardé, E. Talbot, J. Phys. Chem. C 118 (2014) 24136. [50] A. Devaraj, R. Colby, W.P. Hess, D.E. Perea, S. Thevuthasan, J. Phys. Chem. Lett. 4 (2013) 993. [51] B. Gault, D.W. Saxey, M.W. Ashton, S.B. Sinnott, A.N. Chiaramonti, M.P. Moody, D.K. Schreiber, New J. Phys. 18 (2016) 033031. [52] M. Karahka, Y. Xia, H. Kreuzer, Appl. Phys. Lett. 107 (2015) 062105. [53] Y. Xia, M. Karahka, H. Kreuzer, J. Appl. Phys. 118 (2015) 025901. [54] D. Saxey, Ultramicroscopy 111 (2011) 473. [55] J.R. Riley, T. Detchprohm, C. Wetzel, L.J. Lauhon, Appl. Phys. Lett. 104 (2014) 152102. [56] L. Rigutti, L. Mancini, D. Hernández-Maldonado, W. Lefebvre, E. Giraud, R. Butté, J. Carlin, N. Grandjean, D. Blavette, F. Vurpillot, J. Appl. Phys. 119 (2016) 105704. [57] T.J. Puchtler, A. Woolf, T. Zhu, D. Gachet, E.L. Hu, R.A. Oliver, ACS Photonics 2 (2014) 137.
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034
Contents lists available at ScienceDirect
Scripta Materialia journal homepage: www.elsevier.com/locate/scriptamat
Viewpoint article
Atom probe tomography of nitride semiconductors L. Rigutti a, B. Bonef b, J. Speck b, F. Tang c, R.A. Oliver c,⁎ a b c
Normandie Univ., UNIROUEN, INSA Rouen, CNRS, Groupe de Physique des Matériaux, 76000 Rouen, France Materials Department, University of California, Santa Barbara, CA 93106, USA Department of Materials Science and Metallurgy, University of Cambridge, 27 Charles Babbage Road, Cambridge CB3 0FS, UK
a r t i c l e
i n f o
Article history: Received 26 October 2016 Accepted 28 December 2016 Available online xxxx Keywords: Gallium nitride Atom probe tomography Light emitting diode High electron mobility transistor Microwire
a b s t r a c t Atom probe tomography (APT) has emerged as a valuable tool in the study of nitride semiconductors, despite the challenges involved in achieving controlled field evaporation. In optoelectronics, it has provided insights into the nanostructure of light emitting diodes, laser diodes and microwires. In electronics, it has allowed insights into impurity doping and alloying effects in transistors. Coupled with direct correlative studies using other techniques and theoretical modelling based on the APT data, the availability of three dimensional compositional information on nitride heterostructures has had (and will continue to have) a profound impact on the design and development of devices. © 2016 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
1. Introduction Optoelectronic and electronic devices based on gallium nitride, most notably light emitting diodes (LEDs) for energy efficient solid state lighting, already enjoy immense commercial success, but the extreme pace of device development has left understanding of the structure of these materials trailing in its wake. This is currently limiting progress on a wide range of devices, particularly green and amber LEDs [1], whose efficiency lags far behind that of their blue counterparts. One question, central to efforts to understand LED performance has endured almost since the field's inception: why are GaN-based light emitting diodes almost unaffected by the presence of defects, primarily dislocations, at densities which would entirely prohibit device operation in other III–V semiconductors [2]? In this context, the concept of localization is key: it is believed that inhomogeneities in the InGaN quantum wells (QWs) which form the active region of LEDs modulate the in plane potential preventing charge carrier diffusion to non-radiative centres at dislocation cores. For several years, these inhomogeneities were thought to be indium rich regions of the InGaN QW (“indium clusters”), which would have a locally reduced bandgap and hence trap carriers [3]. However, transmission electron microscopy (TEM) data, which formed the strongest evidence for the existence of such clusters, was called into question by researchers who noted that the appearance of clustering could be generated in TEM by exposure to the imaging electron beam, even when the QWs initially looked rather uniform [4]. Against this background, a chance meeting between a specialist in nitride materials (RAO) and an atom-probe tomography (APT) expert ⁎ Corresponding author. E-mail address: [email protected] (R.A. Oliver).
(Prof. GDW Smith) resulted in a fruitful collaboration investigating inhomogeneities in InGaN QWs using APT which kick-started widerranging research on APT of nitrides and other wide bandgap materials. Serendipitously, Smith was exploring the capabilities of recently-developed laser-pulsed atom probe systems appropriate to study of materials with comparatively low conductivity, such as wide bandgap semiconductors, where voltage pulsing was likely to prove unsuccessful. Even with laser pulsing, difficulties were anticipated in studying InGaN QWs which were grown on insulating sapphire substrates. Additionally, the use of the focussed ion beam microscope (FIB) to prepare semiconductors for APT was then in its infancy. A little additional consideration might have further discouraged these first attempts to study nitrides by APT: the laser pulsed atom probe system utilised a green laser, and in the bulk form GaN is transparent to green light, so laser heating might seem an unlikely prospect. In fact, APT of nitrides has had a remarkable success rate, confounding such early concerns, and the use of a green laser is surprisingly effective. Whilst indium clustering represented the initial focus of APT studies, APT has now been used to improve understanding of nitride materials in various ways, as summarised in Section 2. Conversely, the nitrides are also offering new opportunities to understand APT and Section 3 will address this development. 2. What APT tells us about nitrides 2.1. Quantum wells and alloy homogeneity As described above, the assessment of alloy homogeneity in blueemitting InGaN/GaN QWs was the first nitride topic to be assessed by APT. Initially, sample preparation proved challenging, with some damage to the QWs observed in the APT [5] and attributed to knock-on
http://dx.doi.org/10.1016/j.scriptamat.2016.12.034 1359-6462/© 2016 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034
2
L. Rigutti et al. / Scripta Materialia xxx (2016) xxx–xxx
damage during the deposition of a protective Pt coating in the FIB. However, an effective FIB-based sample preparation routine rapidly evolved [6] and more recent investigations have confirmed that, provided an appropriate low-voltage cleanup routine is used, there is little evidence for any FIB-induced implantation or damage [7]. Both blue-emitting and green-emitting QWs grown by metal-organic vapour phase epitaxy (MOVPE) on the polar (0001) plane of GaN were successfully imaged, and (as illustrated in Fig. 1) a frequency distribution (FD) analysis provided no evidence for a deviation from randomness in the indium distribution in the plane of the QW [8]. Later, InGaN QWs grown by molecular beam epitaxy (MBE) yielded a similar result, albeit for a limited sampling volume. Prosa et al. [9] extended this analysis to QWs grown on the semi-polar (10-1-1) plane and performed a particularly thorough analysis addressing a variety of different length scales, but still found no evidence of clustering. Further improvements to the analysis methodology have also been introduced by Tang et al. [10] who addressed the impact of the variation in indium content through the thickness of the QW on the FD analysis by developing a modified nearest neighbour analysis in which all of the atoms of a single QW are projected onto a single plane, effectively compressing the reconstructed QW image in the through-thickness direction. The original APT studies of InGaN QWs received extensive scrutiny from within the community of researchers working on InGaN [11], and the ability of APT to detect clusters was called into question. To validate the technique, it was necessary to find an InGaN sample with a genuinely non-random Indium distribution, and this was first achieved by Müller et al. [12] who studied thick, partially relaxed InGaN layers exhibiting a grossly non-random indium distribution, which was effectively imaged and analyzed in APT. In order to study QWs with a nonrandom indium distribution, Bennett et al. [13] used an electron beam to deliberately damage an APT sample in the TEM and showed using an FD analysis that deviations from randomness had thus been induced. Tang et al. [10] later also found evidence for the formation of as-grown indium clusters in a non-polar (11-20)-oriented QW sample. Both these studies validated the ability of APT to detect clustering in QWs, if present. It should also be noted that the statistical randomness of the (0001) QWs grown by MOVPE and assessed in the earliest APT
studies was also independently validated by carefully-designed TEM experiments [14]. Whilst several analyses of InGaN QWs on both the (0001) plane and the (10-1-1) plane have revealed no evidence of indium clustering, despite evidence from photoluminescence (PL) of localisation in the relevant materials [15,16], some APT studies have been used to suggest that In clustering may be present in InGaN QWs gown under other conditions. Unfortunately, for QWs grown on (0001) none of these studies use an effective statistical method to distinguish between random local variations in the indium content and non-random clustering. Jang et al. [17] have however shown that changes to the indium distribution do occur upon raising the pressure used during QW growth. Given that APT suggests that In clusters are not in general the origin of localisation, one must consider what alternative localisation centres may exist. APT data, supported by both TEM and X-ray diffraction, suggests that typical InGaN QWs exhibit fluctuations in thickness of 1 to 2 monolayers on a lateral length scale of only a few nanometres [6,18]. Such thickness fluctuations represent an alternative possible localisation centre and the relevant APT data has been used to model carrier localization using both a continuum [15] and an atomistic approach [19]. Both of these methodologies suggest that the most important localisation centres are in fact random fluctuations in the indium content, although well-width fluctuations sometimes also influence the electron localisation. (Similarly, APT studies of GaN/ AlGaN QWs reveal the AlGaN to be a random alloy and a continuum model, which correlates well to PL data, suggests that random compositional fluctuations can lead to carrier localisation [20]). These studies exemplify the increasing use of APT data to allow the development of physical models of nitride structures which accurately reflect the microstructure. Similarly, APT data has been used to inform models of carrier transport in LEDs and other devices [21,22,23]. 2.2. Nano- and microwires APT has also been applied to the study of III-N nano- and microwires. Nanowires with diameter smaller than ~150 nm can be analyzed without FIB preparation [24,25,26,27]. The shank angle of nanowires being
Fig. 1. Analysis of indium distributions in a blue emitting multiple QW (a) Side-on view of 3D reconstruction of 5 InGaN/GaN QWs, showing 80% of the In atoms and 10% of the Ga atoms (Grid box 45 nm × 45 nm × 60 nm). (b) A representative Frequency Distribution (FD) analysis of the 3rd QW with a 50-atom bin size comparing the observed indium distributions with the binomial distribution for the random case. (c) Plot of the difference between the observed and expected distributions in (b). This provides no evidence of non-random clustering, in keeping with the results of the relevant χ2 analysis which yields a p-value of 0.47.
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034
L. Rigutti et al. / Scripta Materialia xxx (2016) xxx–xxx
very close to zero has important implications for APT studies focussing on compositional accuracy issues, as described in Section 3. The relationship between structural and functional properties is the focus of studies of larger nanowires [28] or microwires [29,30] containing nanoscale LED structures, prepared by FIB. The InGaN QWs in these systems are grown under significantly different conditions than those in thin films, with possible effects on the incorporation of different species in heterostructures. The APT study reported by Riley et al. [28] on nonpolar and semi-polar QWs in core-shell nanowires identified compositional gradients which could be related to the spatially resolved CL spectra obtained on separate individual nanowire cross sections. In the studies reported by Rigutti et al. [26] and by Mancini et al. [30] on the m-plane InGaN QWs extracted from MOCVD-grown microwires, a PL signal of nanoscale volumes could be extracted despite the damage produced by FIB preparation and was correlated to the structural information obtained later by STEM and APT on approximately the same volume. Fig. 2(a) shows a STEM micrograph of the same field emission tip analyzed by APT and reconstructed in Fig. 2(b). The 3D analysis of
3
the highly non-uniform In distribution found in these QWs correlated well with the PL spectrum (Fig. 2(c)) obtained from the same structure prior to FIB preparation and APT analysis [29]. The accuracy of the correlation could then be improved with the extraction of the PL signal directly from an APT specimen tip and with the consideration of the electronic properties of the stacking faults evidenced by HR-STEM analysis [30], which underlines the importance of direct correlative studies using different techniques, sometimes referred to as “multi-microscopy”, on the same nanoscale structure. 2.3. High electron mobility transistors Nitride based high electron mobility transistors (HEMT) have recently been developed for high frequency and high power electronics. In conventional Ga polar (0001) AlGaN/GaN HEMTs, a two dimensional electron gas (2DEG) forms at the interface between AlGaN and GaN due to a discontinuity in total polarization. Alloy scattering in AlGaN barriers in transistor structures strongly impacts mobility [31]. The addition of
Fig. 2. Correlative μPL, STEM and APT study on a set of m-plane InGaN MQWs extracted from a microwire structure. (a) STEM micrograph of an APT specimen; (b) APT-reconstructed positions of In atoms within the same tip specimen; (c) Correlation of the optical and structural properties of the nanoscale specimen. The blue region highlights the typical In content measured by APT 3D analysis within the quantum wells, whilst the red region corresponds to region in which the peak energies of the μPL spectrum (also shown on the left hand side of the plot) are found (Reprinted with permission from ref. [29] American Chemical Society [29]).
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034
4
L. Rigutti et al. / Scripta Materialia xxx (2016) xxx–xxx
an AlN interlayer, which minimizes the penetration of the electron wavefunction into the alloy, improves mobility for both Ga polar (0001) and N-polar (000-1) orientations [31,32,33,35], the purity of which was examined by APT, showing that whilst typically Ga was unintentionally incorporated into the interlayer, the aluminium site fraction was much higher in samples grown by MBE than those grown by MOVPE [31,32,33]. Interfacial abruptness also profoundly influences transport in HEMTs, and APT has revealed that it is affected by the material's polarity (Ga polar or N polar) [34] and that the two interfaces of the interlayer are not equally diffuse. InAlN has also attracted attention for use as a barrier layer in HEMTs due to the advantage of being able to achieve lattice matching with GaN for an In fraction of 0.18 and its high spontaneous polarization charge. However, the large lattice constant mismatch (13%) and bond strength difference (0.9 eV per bond) between AlN and InN favour phase separation and In segregation in the lateral and vertical directions. This nonuniformity could be source of scattering, leading to mobility degradation in HEMTs. Phase separation resulting in a vertical honeycomb structure in InAlN/GaN heterostructures grown by MBE and MOVPE was first demonstrated by TEM [35,36,37] but projection effects made quantification difficult. APT of In0.18Al0.82N grown by plasma assisted MBE [38] revealed the 3D atomic distribution consisted of columnar cells of 10 nm lateral dimension as shown in Fig. 3(a, b), which, as one might expect, is shown to be statistically non-random by an FD analysis (Fig. 3(c)). At the boundaries of the cells the indium fraction reaches ~0.26, with corresponding depletion of In at the centre of the cells. The elimination of the columnar microstructure was demonstrated using FD analysis of APT data after optimization of the growth conditions [39] and a homogenous In0.17Al0.83N barrier was used in a HEMT which showed promising electrical properties [40]. A second major issue with growth of AlInN in some systems is unintentional incorporation of Ga [41]. APT is particularly powerful in detecting unintentional Ga incorporation where the resulting concentration is low. For example, Tang et al. observed a tail of unintentionally incorporated Ga below 1% in atomic fraction penetrating
into a thick InAlN layer [7] which was attributed to the presence of residual Ga in the growth environment. 2.4. Defects Given the high density of defects found in nitride materials compared to conventional III–V semiconductors and the potential importance of their local compositional environment in understanding their impact on device performance, APT has been relatively sparsely applied to the study of defects in the nitrides, despite very interesting results which have been achieved in group IV semiconductors [42]. Nonetheless, isolated examples exist of studies of point, line, areal and volume defects. Considering impurities as a form of point defect, APT can contribute significantly to the study of dopants in devices. The typical n- and p-type dopants in GaN are Mg and Si respectively, but due to peak overlaps between Si+ and N+ 2 in the mass spectrum, it has thus far been impossible to study Si in APT. However, Mg in p-type layers at concentrations around 1019 cm−3 (~100 ppm) has been successfully addressed [13]. Unintentionally incorporated impurities can also act as dopants, with oxygen being a common unintentional n-dopant. Young et al. [43] studied oxygen incorporation into tunnel junctions grown using a combination of MOVPE and MBE. Surprisingly, the presence of the impurity at the interface between material grown by the two techniques was found to enhance the device performance. Accurate APT quantification of the oxygen content in a very thin layer at this interface was key to understanding this effect. Whilst studies of dopants and unintentional impurities are core strengths of APT, imaging extended defects is more challenging. Dislocations are of particular interest, but some research groups have noted that the success rate in APT imaging of samples known to contain dislocations is far lower than for dislocation free material [44]. The one published data set [45] relevant to dislocations actually shows one side of a V-pit, a structure which forms when the growth of QWs is distorted by the presence of a dislocation forming a pit with the dislocation at its apex and QWs on the sidewalls. The dislocation itself was not part of the data set, and indeed other V-pit analyses [44] suggest that samples
Fig. 3. (a) 2D In and Al atomic map showing the honeycomb microstructure in InAlN. (b) Associated In site fraction map showing the segregation of In into concentrated regions at the boundaries of the cells. (c) Statistical distribution analysis performed with a sampling volume of 100 atoms highlighting the gap between the honeycomb microstructure and a random alloy.
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034
L. Rigutti et al. / Scripta Materialia xxx (2016) xxx–xxx
fracture when the dislocation line intersects the evaporation front. Nonetheless, the analysis showed that QWs on the V-pit sidewalls are both thinner and lower in indium content than those away from the defect, which would result in a locally higher potential, acting to isolate carriers from the defect, in agreement with a model for the robustness of nitride LEDs to dislocations previously proposed by Hangleiter et al. [46]. Rather more success has been achieved in the imaging of basal plane stacking faults (BSFs) [30] and inversion domains (IDs) [47]. In both cases, comparison between APT and TEM data was key to understanding the APT observation, with TEM providing vital information about the crystallographic nature of the defect. For the IDs [47], which were observed in a p-doped AlGaN/GaN superlattice, APT revealed regions of significantly elevated Mg content a few nanometres across, which were correlated with (on a separate region of the same wafer) pyramidal IDs seen in TEM with the same density, distribution and size. APT was able to provide quantitative information about the Mg content, whereas TEM-based compositional analysis failed to detect even the presence of Mg. For the BSFs [30], exactly the same region of the same microwire was imaged first in TEM and then in APT. The BSF intersected QWs on the microwire sidewall but did not observably disturb their compositions. However, the BSF acts as a lower band-gap zinc-blende inclusion in a wurtzite matrix, and this effects the local PL, as was
5
shown in studies performed again on exactly the same material, another example of the multi-microscopy approach highlighted in Section 2.2. 3. What nitrides tell us about APT During APT studies of nitrides it became evident that the technique does not generally yield an accurate measurement of the elemental composition. Due to the relatively high rate of success for the analyses and to the straightforward interpretation of their mass spectra, III-N materials have become a model system for the study of compositional biases in APT, which may occur in a much broader class of compounds. Studies conducted on III-N materials highlighted the need to further explore the field evaporation behavior of compound semiconductors in order to (i) assess whether there is a bias in the compositional measurement, (ii) understand the physical mechanisms inducing any bias and (iii) correct said bias, if possible. The first studies performed on GaN nanowires have shown that the composition measured by APT may change from almost 100% to under 50% of Ga as the impinging laser energy is reduced and, simultaneously, the constant applied voltage is raised. Moreover, the measured Ga atomic fraction correlates well with the charge state ratio Ga2+/Ga+, an indirect indicator of the surface field [25,26,27] according to the post-ionization theory [48]. Fig. 4(a) illustrates further studies of GaN
Fig. 4. (a) Correlation between the Ga atomic fraction measured by APT in GaN on the Ga2+/Ga1+ charge state ratio,. Large open symbols refer to average quantities extracted from datasets containing 105 events at constant detection rate or at constant impinging laser energy, whilst the small are issued by the detector statistics of the charge state ratio (b) Ga2+/Ga1+ and of the Ga atomic fraction (c). (d) Reconstructed 3D positions of the Al atoms in a GaN/AlGaN multi-QW structure. (e) Correlation between the Ga2+/Ga1+ charge state ratio and the measured Al III-site fraction. Large open symbols refer to dataset extracted from single barriers, whilst the small filled symbols refer to the detector statistics extracted from the bottom region indicated in (d).
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034
6
L. Rigutti et al. / Scripta Materialia xxx (2016) xxx–xxx
carried out either by performing separate analyses at constant detection rate or by varying one experimental parameter at once (laser intensity or DC applied bias): the measured Ga atomic fraction correlates to the surface field, which can be estimated through the charge state ratios (Ga2 +/Ga1 +). The detector statistics of the Ga2 +/Ga+ charge state ratio (Fig. 4(b)) show that the surface field is non-uniform across the different regions of the APT specimen apex due to the specimen geometry and crystallography, and the measured atomic fraction of Ga (Fig. 4(c)) depends on the microscopic surface field distribution. Similar dependences on the surface electric field have been found in other wide bandgap binary systems, such as AlN, ZnO and MgO [49,50]. This behavior can be partially explained through preferential evaporation of the metallic element at high field, but the mechanism responsible of the loss of N or O at low field is still under debate. Neutral evaporation or production of neutral N and O upon dissociation of more complex molecules have been proposed [51,52,53,54]. Whilst the above-mentioned binary systems have a known composition, this is not the case for ternary alloys and doped materials. In these technologically important cases, is essential to know under which conditions APT can accurately measure the composition. This problem has been assessed for InyGa1-yN and AlyGa1-yN. Riley et al. [55] and Mancini et al. [49] provided evidence that the In III-site fraction in InGaN is fairly independent of the experimental conditions. However, data on Al0.25Ga0.75N (Fig. 4(d)) showed that the measured Al site fraction deviates from the correct value (y = 0.25) at high field Fig. 4(e). The accurate site fraction of AlGaN may thus be determined performing the measurement at low field or, equivalently, at sufficiently high laser pulse intensity. This, however, could imply a degradation of spatial resolution due to the temperature increase and to the possible modification of tip shape with increasing intensity. To overcome this, a statistical correction scheme was developed, in order to restore the spatial behavior and the amplitude of the alloy fluctuations even in the case of biased measurements. The correction procedure has been tested for Al0.25Ga0.75N by measuring the emission and localization energy of the photons emitted within the alloy, two quantities that depend on the average site fraction and on the alloy fluctuations [56]. 4. Concluding remarks and future outlook This review has illustrated the widespread and successful use of APT in analysing nitride semiconductors. One key theme which has emerged is the important of direct correlative studies, sometimes called “multimicroscopy”, in which a specific nanostructure or region of material is assessed using other nanoscale characterisation techniques, such as TEM or micro-PL, prior to APT studies. This is a challenging but very powerful approach which makes direct links between the structure, composition and properties of a material at the nanoscale, and should see increasing application in the future, utilising a wide range of characterisation techniques, as is increasingly being achieved in other contexts [57]. The ability to directly compare detailed structure and properties, allows for testing and validation of physical models of nanostructures without the uncertainty inherent in comparing different structures within arrays which may exhibit significant statistical variations. The availability of such data sets as an input for modelling raises the possibility of directly inputting APT data into atomistic calculations, although challenges arise because the atoms in the APT data set do not in general sit on crystallographically defined sites (lattice fringe resolution in APT has yet to be achieved in the nitrides) and less than half the atoms present in the original sample are typically present in the data set. The development of protocols to make best use of APT data in developing atomistic models within the limitations it poses will require input from both theoreticians and experimentalists. Other possible future challenges for nitride APT concern improving its effectiveness in assessing impurities and point defects. It would be difficult to use APT to study point defects such as nitrogen vacancies because of both the overall limits of resolution and sampling of
the technique (applicable to all materials), and the difficulty in measuring reasonable stoichiometries in the nitrides in all samples (see Section 3). One interesting challenge for future APT studies would be the quantification of hydrogen incorporation, particularly in MOVPE-grown p-type GaN where hydrogen is known to act to compensate the Mg acceptors, reducing conductivity. For n-type doping, questions remain about the role of defects in dopant segregation, but these can only effectively be addressed if the peak overlap between Si + and N 22 + is addressed, perhaps by studying samples grown using a source of the Si29 or Si30 isotopes, or by using high mass resolution instruments to address the low concentration of naturally occurring Si29 and Si30 . Addressing these issues will contribute to the usefulness of APT in analysing full device structures. Such studies will in the future increasingly address complete devices including the metal and/or dielectric layers and their interface to the nitride semiconductor. Overall, the increasing availability of APT and it's great strengths in providing nanoscale three dimensional compositional information with sufficient sensitivity to address dopants and impurities will lead further increases in its application in the nitrides and other wide bandgap materials. Acknowledgements This work was supported in part by the European Research Council under the European Community's Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement no 279361 (MACONS). References [1] R.A. Oliver, Mat. Sci. Technol. 32 (2015) 737. [2] R.A. Oliver, S.E. Bennett, T. Zhu, D.J. Beesley, M.J. Kappers, D.W. Saxey, A. Cerezo, C.J. Humphreys, J. Phys. D 43 (2010) 354003. [3] Y. Narukawa, Y. Kawakami, M. Funato, S. Fujita, S. Fujita, S. Nakamura, Appl. Phys. Lett. 70 (1997) 981. [4] T.M. Smeeton, M.J. Kappers, J.S. Barnard, M.E. Vickers, C.J. Humphreys, Appl. Phys. Lett. 83 (2003) 5419. [5] M.J. Galtrey, R.A. Oliver, M.J. Kappers, C. McAleese, D. Zhu, C.J. Humphreys, P.H. Clifton, D. Larson, A. Cerezo, Appl. Phys. Lett. 92 (2008) 41904. [6] M.J. Galtrey, R.A. Oliver, M.J. Kappers, C.J. Humphreys, D.J. Stokes, P.H. Clifton, A. Cerezo, Springer Proc. Phys. 120 (2007) 161. [7] F. Tang, M.P. Moody, T.L. Martin, P.A.J. Bagot, M.J. Kappers, R.A. Oliver, Microsc. Microanal. 21 (2015) 544. [8] M.J. Galtrey, R.A. Oliver, M.J. Kappers, C.J. Humphreys, P.H. Clifton, D. Larson, D.W. Saxey, A. Cerezo, J. Appl. Phys. 104 (2008) 13524. [9] T.J. Prosa, P.H. Clifton, H. Zhong, A. Tyagi, R. Shivaraman, S.P. DenBaars, S. Nakamura, J.S. Speck, Appl. Phys. Lett. 98 (2011) 191903. [10] F. Tang, T. Zhu, F. Oehler, W.Y. Fu, J.T. Griffiths, F.C.-P. Massabuau, M.J. Kappers, T.L. Martin, P.A.J. Bagot, M.P. Moody, R.A. Oliver, Appl. Phys. Lett. 106 (2015) 72104. [11] C. Kisielowski, T.P. Bartel, Appl. Phys. Lett. 91 (2007) 176101. [12] M. Müller, G.D.W. Smith, B. Gault, C.R.M. Grovenor, Acta. Mater. 60 (2012) 4277. [13] S.E. Bennett, D.W. Saxey, M.J. Kappers, J.S. Barnard, C.J. Humphreys, G.D.W. Smith, R.A. Oliver, Appl. Phys. Lett. 99 (2011) 21906. [14] K.H. Baloch, A.C. Johnston-Peck, K. Kisslinger, E.A. Stach, S. Gradečak, Appl. Phys. Lett. 102 (2013) 191910. [15] D. Watson-Parris, M.J. Godfrey, P. Dawson, R.A. Oliver, M.J. Galtrey, M.J. Kappers, C.J. Humphreys, Phys. Rev. B 83 (2011) 115321. [16] P. Dawson, S. Schulz, R.A. Oliver, M.J. Kappers, C.J. Humphreys, J. Appl. Phys. 119 (2016) 181505. [17] D.H. Jang, G.H. Gu, B.H. Lee, C.G. Park, Ultramicroscopy 127 (2013) 114. [18] G.H. Gu, D.H. Jang, K.B. Nam, C.G. Park, Microsc. Microanal. 19 (2013) 99. [19] S. Schulz, D.P. Tanner, E.P. O'Reilly, M.A. Caro, T.L. Martin, P.A.J. Bagot, M.P. Moody, F. Tang, J.T. Griffiths, F. Oehler, M.J. Kappers, R.A. Oliver, C.J. Humphreys, D. Sutherland, M.J. Davies, P. Dawson, Phys. Rev. B 92 (2015) 235419. [20] L. Rigutti, L. Mancini, W. Lefebvre, J. Houard, D. Hernàndez-Maldonado, E. Di Russo, E. Giraud, R. Butté, J. Carlin, N. Grandjean, Semicond. Sci. Tech. 31 (2016) 095009. [21] Y.-R. Wu, R. Shivaraman, K.-C. Wang, J.S. Speck, Appl. Phys. Lett. 101 (2012) 083505. [22] D.A. Browne, B. Mazumder, Y.-R. Wu, J.S. Speck, J. Appl. Phys. 117 (2015) 185703. [23] D.N. Nath, Z.C. Yang, C.Y. Lee, P.S. Park, Y.R. Wu, S. Rajan, Appl. Phys. Lett. 103 (2013) 022102. [24] N.A. Sanford, P.T. Blanchard, M. Brubaker, K.A. Bertness, A. Roshko, J. Schlager, R. Kirchhofer, D. Diercks, B. Gorman, Phys. Status Solidi (c) 11 (2014) 608. [25] R. Agrawal, R.A. Bernal, D. Isheim, H.D. Espinosa, J. Phys. Chem. C 115 (2011) 17688–17694. [26] J.R. Riley, R.A. Bernal, Q. Li, H.D. Espinosa, G.T. Wang, L.J. Lauhon, ACS Nano 6 (2012) 3898–3906. [27] D.R. Diercks, B.P. Gorman, R. Kirchhofer, N. Sanford, K. Bertness, M. Brubaker, J. Appl. Phys. 114 (2013) 184903.
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034
L. Rigutti et al. / Scripta Materialia xxx (2016) xxx–xxx [28] J.R. Riley, S. Padalkar, Q. Li, P. Lu, D.D. Koleske, J.J. Wierer, G.T. Wang, L.J. Lauhon, Nano Lett. 13 (2013) 4317–4325. [29] L. Rigutti, I. Blum, D. Shinde, D. Hernandez-Maldonado, W. Lefebvre, J. Houard, F.O. Vurpillot, A. Vella, M. Tchernycheva, C. Durand, Nano Lett. 14 (2013) 107–114. [30] L. Mancini, D. Hernández-Maldonado, W. Lefebvre, J. Houard, I. Blum, F. Vurpillot, J. Eymery, C. Durand, M. Tchernycheva, L. Rigutti, Appl. Phys. Lett. 108 (2016) 042102. [31] M.N. Fireman, D.A. Browne, B. Mazumder, J.S. Speck, U.K. Mishra, Appl. Phys. Lett. 106 (2015) 202106. [32] B. Mazumder, S.W. Kaun, J. Lu, S. Keller, U.K. Mishra, J.S. Speck, Appl. Phys. Lett. 102 (2013) 111603. [33] S.W. Kaun, B. Mazumder, M.N. Fireman, E.C.H. Kyle, U.K. Mishra, J.S. Speck, Semicond. Sci. Tech. 30 (2015) 055010. [34] B. Mazumder, M.H. Wong, C.A. Hurni, J.Y. Zhang, U.K. Mishra, J.S. Speck, Appl. Phys. Lett. 101 (2012) 091601. [35] D.S. Katzer, D.F. Storm, S.C. Binari, B.V. Shanabrook, A. Torabi, L. Zhou, D.J. Smith, J. Vac, Sci. Technol. B 23 (2005) 1204. [36] S. Kret, A. Wolska, M.T. Klepka, A. Letrouit, F. Ivaldi, A. Szczepanska, N. Grandjean, J. Phys. Conf. Ser. 326 (2011) 012013. [37] L. Zhou, D.J. Smith, M.R. McCartney, D.S. Katzer, D.F. Storm, Appl. Phys. Lett. 90 (2007) 081917. [38] S. Choi, F. Wu, R. Shivaraman, E.C. Young, J.S. Speck, Appl. Phys. Lett. 100 (2012) 232102. [39] E. Ahmadi, R. Shivaraman, F. Wu, S. Wienecke, S.W. Kaun, S. Keller, U.K. Mishra, Appl. Phys. Lett. 104 (2014) 072107. [40] S.W. Kaun, E. Ahmadi, B. Mazumder, F. Wu, E.C.H. Kyle, P.G. Burke, J.S. Speck, Semicond. Sci. Tech. 29 (2014) 045011. [41] A. Redondo-Cubero, K. Lorenz, R. Gago, N. Franco, M.A. di Forte Poisson, E. Alves, E. Munos, J. Phys. D: Appl. Phys. 43 (2010) 055406.
7
[42] D. Blavette, S. Duguay, J. Appl. Phys. 119 (2016) 181502. [43] E.C. Young, B.P. Yonkee, F. Wu, S.H. Oh, S.P. DenBaars, S. Nakamura, J.S. Speck, Appl. Phys. Express 9 (2016) 022102. [44] S. Bennett, Nitride Semiconductors Studied by Atom Probe Tomography and Correlative TechniquesPhD Thesis University of Cambridge, 2010. [45] S. Tomiya, Y. Kanitani, S. Tanaka, T. Ohkubo, K. Hono, Appl. Phys. Lett. 98 (2011) 181904. [46] A. Hangleiter, F. Hitzel, C. Netzel, D. Fuhrmann, U. Rossow, G. Ade, P. Hinze, Phys. Rev. Lett. 95 (2005) 127402. [47] S.E. Bennett, R.M. Ulfig, P.H. Clifton, M.J. Kappers, J.S. Barnard, C.J. Humphreys, R.A. Oliver, Ultramicroscopy 111 (2011) 207. [48] D.R. Kingham, Surf. Sci. 116 (1982) 273. [49] L. Mancini, N. Amirifar, D. Shinde, I. Blum, M. Gilbert, A. Vella, F.O. Vurpillot, W. Lefebvre, R. Lardé, E. Talbot, J. Phys. Chem. C 118 (2014) 24136. [50] A. Devaraj, R. Colby, W.P. Hess, D.E. Perea, S. Thevuthasan, J. Phys. Chem. Lett. 4 (2013) 993. [51] B. Gault, D.W. Saxey, M.W. Ashton, S.B. Sinnott, A.N. Chiaramonti, M.P. Moody, D.K. Schreiber, New J. Phys. 18 (2016) 033031. [52] M. Karahka, Y. Xia, H. Kreuzer, Appl. Phys. Lett. 107 (2015) 062105. [53] Y. Xia, M. Karahka, H. Kreuzer, J. Appl. Phys. 118 (2015) 025901. [54] D. Saxey, Ultramicroscopy 111 (2011) 473. [55] J.R. Riley, T. Detchprohm, C. Wetzel, L.J. Lauhon, Appl. Phys. Lett. 104 (2014) 152102. [56] L. Rigutti, L. Mancini, D. Hernández-Maldonado, W. Lefebvre, E. Giraud, R. Butté, J. Carlin, N. Grandjean, D. Blavette, F. Vurpillot, J. Appl. Phys. 119 (2016) 105704. [57] T.J. Puchtler, A. Woolf, T. Zhu, D. Gachet, E.L. Hu, R.A. Oliver, ACS Photonics 2 (2014) 137.
Please cite this article as: L. Rigutti, et al., Atom probe tomography of nitride semiconductors, Scripta Materialia (2016), http://dx.doi.org/10.1016/ j.scriptamat.2016.12.034