PCCP View Article Online
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
PAPER
Cite this: Phys. Chem. Chem. Phys., 2015, 17, 7060
View Journal | View Issue
Oxidation of the GaAs semiconductor at the Al2O3/GaAs junction Marjukka Tuominen,a Muhammad Yasir,a Jouko Lång,a Johnny Dahl,a Mikhail Kuzmin,a Jaakko Ma¨kela¨,a Marko Punkkinen,a Pekka Laukkanen,*a Kalevi Kokko,a Karina Schulte,b Risto Punkkinen,c Ville-Markus Korpija¨rvi,d Ville Poloja¨rvid and Mircea Guinad Atomic-scale understanding and processing of the oxidation of III–V compound–semiconductor surfaces are essential for developing materials for various devices (e.g., transistors, solar cells, and light emitting diodes). The oxidation-induced defect-rich phases at the interfaces of oxide/III–V junctions significantly affect the electrical performance of devices. In this study, a method to control the GaAs oxidation and interfacial defect density at the prototypical Al2O3/GaAs junction grown via atomic layer deposition (ALD) is demonstrated. Namely, pre-oxidation of GaAs(100) with an In-induced c(8 2) surface reconstruction, leading to a crystalline c(4 2)–O interface oxide before ALD of Al2O3, decreases band-gap defect density at the Al2O3/GaAs interface. Concomitantly, X-ray photoelectron spectroscopy (XPS) from these Al2O3/GaAs interfaces shows that the high oxidation state of Ga (Ga2O3 type) decreases, and the corresponding In2O3 type phase forms when employing the c(4 2)–O interface layer. Detailed synchrotron-radiation XPS of the counterpart c(4 2)–O oxide of InAs(100) has been utilized to elucidate
Received 19th December 2014, Accepted 2nd February 2015
the atomic structure of the useful c(4 2)–O interface layer and its oxidation process. The spectral analysis
DOI: 10.1039/c4cp05972g
between
www.rsc.org/pccp
dation. These results, discussed within the current atomic model of the c(4 2)–O interface, provide insight into the atomic structures of oxide/III–V interfaces and a way to control the semiconductor oxidation.
reveals that three different oxygen sites, five oxidation-induced group-III atomic sites with core-level shifts 0.2 eV and +1.0 eV, and hardly any oxygen-induced changes at the As sites form during the oxi-
Introduction The fundamental phenomenon of the oxidation of semiconductor crystals is one of the most technologically important chemical reactions.1–3 Its exothermic nature causes nearly all device applications to get an oxidized semiconductor surface or interface during the manufacturing of semiconductor-based devices. An oxidized semiconductor part is usually formed during the synthesis of an oxide film(s) on the semiconductor crystal or just due to the contact of material surfaces with air or oxygen-containing environments. The chemical and physical properties of oxidized semiconductor interfaces are relevant, for example, to the functionality of the microprocessor and mobile-phone transistors as well as to the performance of the window and passivation layers of high-efficiency solar cells and a
Department of Physics and Astronomy, University of Turku, FI-20014 Turku, Finland. E-mail: [email protected] b MAX IV Laboratory, Lund University, P. O. Box 118, SE-221 00 Lund, Sweden c Department of Information Technology, University of Turku, FI-20014 Turku, Finland d Optoelectronics Research Centre, Tampere University of Technology, FI-33101 Tampere, Finland
7060 | Phys. Chem. Chem. Phys., 2015, 17, 7060--7066
infrared detectors. This is because electric current flows, at least in part, in the vicinity of oxidized semiconductor interfaces in these devices. The aim toward ever decreasing device dimensions1–7 leads simultaneously to ever increasing significance of the understanding and engineering of the chemical and physical properties of the oxide/semiconductor interfaces on the atomic scale. However, these issues are still hard to predict and characterize as semiconductor oxides usually have an amorphous structure, and the interface layers are thin and lie below the surface between the oxide film and the semiconductor. The junction between an oxide film and a GaAs compound semiconductor is one of the most studied oxide/semiconductor interface systems, because GaAs is the material of various photonic devices and mobile-phone transistors. Concerning further III–V semiconductors, e.g., GaAs, GaInAs, GaSb, InP and InSb, their oxidation and oxide-interface characteristics have become more and more important because after decades of intense research and development,8–23 the III–V metal-oxide–semiconductor fieldeffect transistors (MOSFETs) are now on the roadmap of microelectronics manufacturers. However, the oxide/III–V interfaces are still problematic. III–V surfaces become easily oxidized in an uncontrolled way due to the energetically favored oxidation and
This journal is © the Owner Societies 2015
View Article Online
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
Paper
the practical difficulty to avoid the contact of III–V surfaces with oxygen during manufacturing oxide films on them. The III–V oxidation has been often found to result in a poor-quality interface layer with a high-density of defect states around the III–V band gap. This has led to a reduced electric current and a nonradiative recombination of electrons and holes, for instance. The problems can be summarized as follows: (i) oxidation causes a structural disorder leading to the formation of point defects and local deviation in bonding and (ii) oxygen incorporation into a semiconductor causes a modified electronic structure and produces an oxide phase with unwanted electronic states tailing across the band gap. This is why the success of the future semiconductor technology, in general, depends significantly on the understanding and engineering of the III–V oxidation. Indeed, much work has been done to understand and reduce the harmful oxidation-induced effects. The control techniques include sulphur passivation, depositing silicon interface layer, growth of epitaxial oxides, and removal of surface oxides via self-cleaning effect in atomic layer deposition (ALD).1–2,10,12,24–27 Furthermore, harmful electron states have been found to arise, amongst others, from group-V dimers (pairs); group-V and group-III dangling bonds (missing bonds); and III–Ox or V–Ox phases.2,14,15,18,19,21 Meanwhile, certain forms of semiconductor oxides such as Ga2O and InAsOx have been observed to be nonharmful or even to improve the electrical properties of the oxide/semiconductor interfaces.14–16 In order to develop these device materials for novel and energyefficient applications, it is essential to address the questions as to what kind of oxidation-induced changes cause the harmful defect states around the band gap, and how to process the III–V oxidation effects in a controlled manner. In this paper, a method for decreasing the amount of bandgap defects is demonstrated for the prototypical ALD-grown Al2O3/GaAs interface. Namely, covering the GaAs(100) with an ultrathin indium-induced c(8 2) surface reconstruction, followed by its pre-oxidation resulting in a crystalline c(4 2)–O oxide phase, provides a route to improve oxide/GaAs junctions during device processing. Furthermore, X-ray photoelectron spectroscopy (XPS) of the same Al2O3/GaAs junctions reveals interesting spectral features, which elucidate the role of oxidation states of the semiconductor elements in the formation of harmful interface defects at oxide/semiconductor device structures. To investigate the oxidation-induced features in detail, we have also utilized high-resolution synchrotron-radiation (SR)28–30 XPS of the corresponding c(4 2)–O phase of the oxidized InAs(100) surface, which allows the spectral analysis without complex overlapping of Ga 3d and In 4d core-level emissions. The unveiled spectral changes are important for solving the atomic structure of these interfaces and for controlling the formation of useful oxidationinduced phases.
Results and discussion Photoluminescence (PL) The comparison of PL spectra (Fig. 1) from the Al2O3-capped GaAs samples, with and without the pre-oxidized c(4 2)–O
This journal is © the Owner Societies 2015
PCCP
Fig. 1 Photoluminescence spectra from the Al2O3/GaAs junctions with and without the pre-oxidized c(4 2)–O layer on GaAs(100). The inset shows low-energy electron diffraction (LEED) from the GaAs(100) surface with the c(4 2)–O reconstruction. The white rectangle visualizes the reciprocal-space unit cell of the oxidized layer.
layer, shows that the PL intensity is higher for the sample which contains the c(4 2)–O interface structure than for the junction without the pre-oxidized phase (i.e., pure Al2O3/GaAs). The PL intensity of the pre-oxidized junction is increased by a factor of 2.2. Because the distinction between the samples is the oxide/ GaAs interface, the PL difference can be justifiably associated with the changes in interface properties. Indeed, it has been previously confirmed that the PL intensity is sensitive to the properties of a thin oxide/III–V interface layer, and higher PL intensity corresponds to lower interface defect density.1,31–34 Therefore, the presented PL results prove that the defect density of the Al2O3/GaAs interface is reduced by using the pre-oxidized c(4 2)–O layer. It is worth noting that the pure Al2O3/GaAs sample, used for the comparison, provides a stringent reference because a well-ordered GaAs starting surface, similar to that described previously,34 was prepared before the ALD growth. Future measurements such as capacitance– voltage characterizations are needed to elucidate further the electrical properties of the interface. However, the above PL results unambiguously reveal a decrease in the interface defects in the energy band gap. The reduced defect density can be connected primarily with an improved crystalline nature of the Al2O3/GaAs interface structure with the c(4 2)–O phase. The presence of indium can also affect the formation of harmful GaAs oxide phases. X-ray photoelectron spectroscopy (XPS) To clarify mechanisms behind the above finding, the Al2O3/ GaAs interfaces were characterized using XPS. This method allows nondestructive probing of thin and buried interface layers when the topmost oxide film is thin enough for the elastic photoemission. The spectral changes and so called corelevel shifts (CLS) unveiled using XPS measurements provide irreplaceable information about the atomic and electronic structures of the materials. Different CLS reflect different
Phys. Chem. Chem. Phys., 2015, 17, 7060--7066 | 7061
View Article Online
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
PCCP
Paper
in the Al2O3/c(4 2)–O/GaAs spectrum leads to the negative values seen in the lower binding energy region of the difference spectra in Fig. 2b. The In 3d emission from the Al2O3/c(4 2)–O/GaAs interface, shown in Fig. 3a, reveals that part of the In atoms have a high oxidation state at the interface since the largest separation in the In 3d components is 1.2 eV, of which oxidation state resembles In2O3.35,36 In contrast, the As 3d spectrum from the Al2O3/c(4 2)–O/GaAs interface in Fig. 3b is very similar to that of Al2O3/GaAs (not shown) and contains only two small shifts 0.36 eV and +0.37 eV, but no apparent signal from the As–As bonding around +0.6 eV or from the high oxidation state of As around +3.5 eV.35,36 Therefore, in combination with the PL finding, it can be concluded that the In2O3 type structure causes a lower density of harmful electronic states compared to Ga2O3 at the interface. The underlying reason can be that the bonding configuration of the In-containing oxides has a better crystalline order than the pure Ga-oxides at the Al2O3/GaAs interface, and/or that the In-oxide structure causes different electronic states in the GaAs gap as compared to the Ga-oxide related states. The results indicate that the deposition of small amount of In on the III–V surface before an oxide growth is helpful for avoiding the formation of the highly oxidized Ga phases that cause harmful electron states around the band gap. Fig. 2 (a) Ga 3d core-level photoelectron spectra measured through the Al2O3 films by XPS from Al2O3/GaAs interfaces. Solid line peaks represent the d5/2 parts of the double peak components and dotted line peaks mark the corresponding d3/2 parts. (b) Difference spectrum: l times Ga 3d of Al2O3/c(4 2)–O/GaAs was subtracted from Ga 3d of Al2O3/GaAs (the bulk peaks were aligned before the subtracting).
atomic bonding environments in the material, so they can be used for controlling and characterizing the properties of device interfaces, such as the oxidation of semiconductor elements during the oxide-film growth. Fig. 2a shows the fitting analysis of the Ga 3d spectra for the both junctions. In addition to the substrate bulk component, the both spectra include a second Ga 3d component labeled as GaOx around +0.5 eV with respect to the bulk peak. This corelevel shift has been identified previously as a Ga2O-type bonding.2,14,15,32,33 A close inspection reveals a spectral difference on the higher binding energy tail, since the Ga 3d emission of the pure Al2O3/GaAs interface includes a third component: GaOy around +1 eV. A similar shift has been previously related to Ga2O3-type bonding.2,14,15,32,33 The GaOy intensity is not high, and in fact, earlier studies have shown how difficult it is to resolve Ga-oxide features in the 3d emission as compared to the surface-sensitive Ga 2p emission.15 In this study the Al2O3 film was too thick to observe Ga 2p with a reasonable signal-to-noise ratio. Thus, an analysis of the difference spectrum between the Ga 3d photoemission from the Al2O3/GaAs junctions with and without the c(4 2)–O interface was performed, as presented in Fig. 2b. By subtracting the Ga 3d emission of the c(4 2)–O interface (multiplied with a factor l between 0.3–1.5) from the emission of the pure Al2O3/GaAs interface, it is seen that the GaOy feature becomes apparent. A small In 4d emission present
7062 | Phys. Chem. Chem. Phys., 2015, 17, 7060--7066
Synchrotron-radiation XPS To investigate the oxidation-induced atomic structure of the c(4 2)–O layer in more detail, we utilized high-resolution SR-XPS of the equivalent InAs(100)c(4 2)–O surface which allows a thorough group-III spectral analysis without the complex emission of the overlapping Ga 3d and In 4d peaks. The As 3d, In 4d, and O 1s spectra measured from InAs(100)c(4 2)–O by SR-XPS are shown in Fig. 4. The possibility of changing the photon energy (i.e., surface vs. bulk sensitivity of the measurement in Fig. 4) was
Fig. 3 (a) In 3d, and (b) As 3d core-level photoelectron spectra measured through the Al2O3/c(4 2)–O/GaAs interface by XPS. Solid line peaks represent the d5/2 parts of the double peak components and dotted line peaks mark the corresponding d3/2 parts.
This journal is © the Owner Societies 2015
View Article Online
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
Paper
PCCP
Fig. 4 Synchrotron-radiation XPS spectra measured at liquid-nitrogen temperature from the c(4 2)–O layer of the oxidized InAs(100): (a) As 3d, (b) In 4d, and (c) O 1s core-level spectra. Solid line peaks represent the O 1s peaks and d5/2 parts of the double peak As 3d and In 4d components, and dotted line peaks mark the corresponding d3/2 parts. The topmost spectra are measured with more bulk-sensitive photon energies than the lower spectra.
helpful to determine carefully the InAs bulk peak (B) positions and the precise oxidation-induced shifts. The As spectra include only small shifts 0.28 eV and +0.20 eV, similar to the aforementioned shifts measured from the Al2O3/ GaAs interfaces, indicating that the c(4 2)–O layer does not contain highly oxidized As or pure As–As bonding. Because there are different As-dimer sites and shifts which can be easily misinterpreted, it might be worth summarizing them: the As–As dimers can be divided between the sites where an As atom also bonds to group-III atoms, and the sites where As atom only bonds to other As atoms. The first type of As dimers exists, for example, at the clean III-As(100)(2 4) surfaces and produces only small shifts between 0.3 eV and +0.1 eV depending on the height of the As site (the clean (2 4) surface itself causes also a +0.3 eV shift).37,38 In contrast, the As dimers without the group-III bonding are formed when the surface structure becomes more As rich, for example, on the clean III-As(100)c(4 4) surface,37,38 and they produce a clear positive shift around +0.6 eV. Arsenic clusters can be also expected to cause the same clear positive shift. Thus, the small As shifts found in our samples mean that the As dimers with group-III bonding can be present. Anyhow, they do not explain the PL difference properly because the As 3d spectra for the both interfaces are very similar; it is unlikely that the enhancement of the PL intensity is associated with the changes occurring in the atomic surroundings of As species. The O 1s spectra in Fig. 4 are composed of three components at 0, +0.41, and +0.91 eV, while the In spectra of InAs(100)c(4 2)–O contain five oxidation-induced components: 0.22, +0.20, +0.46, +0.76, and +1.09 eV. Usually the direct oxygen bonding of a semiconductor element is expected
This journal is © the Owner Societies 2015
to cause a positive shift of +0.5 eV or larger. Thus, the three largest In shifts can arise from different direct In–O bonds. According to the previous studies the +0.46 shift is associated with the In2O type environment, and the +0.76 and +1.09 eV shifts with different In2O3 type structures. The two smaller In shifts are more likely to arise from indirect oxidation effects like In atoms neighboring the new In–O environments. Because the measured surface is crystalline, all these components can be associated with a single unit cell structure of the c(4 2)–O, of which surface area is only 0.74 nm2. It is striking that such a small cell includes so many different In sites. The current atomic model39 (Fig. 5) for InAs(100)c(4 2)–O provides several In shifts between 0.44 eV and +0.24 eV. These theoretical shifts were simulated for the ab initio calculated atomic structure by means of the initial-state model described elsewhere.37 Taking into account of possible experimental and computational errors in the shifts, the calculated and measured negative
Fig. 5
Atomic model suggested previously for the c(4 2)–O structure.39
Phys. Chem. Chem. Phys., 2015, 17, 7060--7066 | 7063
View Article Online
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
PCCP
shifts agree reasonably, but there is a clear deviation in the positive shifts. Furthermore, the calculated O shifts based on the current atomic model and the initial-state picture are in the range of 0.10 eV. On the other hand, the calculated and measured As shifts agree because the calculated ones also lie in the range of 0.10 eV. One possible reason for the differences between the calculated and measured shifts is that the initial-state model, used to simulate the shifts, does not describe properly the photoemission process in the present oxygen-containing system. In other words, the final-state effects cannot be excluded. Although the initial-state model has been previously shown to reproduce well the shifts of various clean and adsorbate containing semiconductor surfaces,37 neglecting the final-state in calculations might lead to the wrong binding energy for the present system because oxygen is quite different from many other adsorbate elements. Incorporating oxygen into the crystal can significantly influence the electronic band structure, and therefore, the final-state effects can play an essential role in the photoemission process. On the other hand, the current c(4 2)–O model in Fig. 5 is well justified by the calculated total energies and by the comparison of the measured and simulated scanning tunneling microscopy images.39 Thus, it is suggested both the reanalysis of the atomic model of c(4 2)–O and the simulation model for the core-level shifts of the oxidized semiconductor surfaces. If the adjustment of the atomic model is considered, the large positive In shifts and changes in the O bonding environment, observed in the experiments, imply that the c(4 2)–O surface oxide has higher oxygen concentration than 0.5 ML of the current model. Also, the In shifts of +0.76 and +1.09 eV suggest that some of the group-III atoms have more than one direct bond with oxygen (i.e., In2O3 type bonding). On the other hand, the negligible As shifts measured indicate that no direct As–O forms, as the current model in Fig. 5 shows. Overall, the total energy range of the In shifts measured from InAs(100)c(4 2)–O is consistent with the XPS results from Al2O3/c(4 2)–O/GaAs described above. This together with the As results indicates that the bonding structure of the c(4 2)–O oxide layer does not change significantly during the Al2O3 growth process, which is therefore assumed to have a negligible effect on the electronic state of the interface. Hence, the presented results suggest that an improved crystalline order at Al2O3/c(4 2)–O/GaAs junction is one reason for the improved PL, and that an increased Ga oxidation (around +1 eV, Ga2O3 type) instead leads to disorder-induced gap states at the pure Al2O3/GaAs interface, decreasing its PL intensity. This is in consistency with the previous electronicstructure calculations of oxidized GaAs and InGaAs surfaces.40
Experimental Al2O3/GaAs samples were synthesized in an Omicron multichamber UHV system. GaAs(100) pieces of size 1.0 cm 0.5 cm were cut from the same n-type (around 1 1018 cm 3 concentration of silicon dopants) wafer supplied by Wafer Technology.
7064 | Phys. Chem. Chem. Phys., 2015, 17, 7060--7066
Paper
The substrate pieces were cleaned by Ar-ion sputtering and post-heating at 550 1C until a well-defined Ga-induced c(8 2) reconstruction was seen on the surface using the low-energy electron diffraction (LEED) method. The In-induced c(8 2) surface reconstruction41 was obtained by depositing one monolayer of In from a Ta-envelope evaporator on cleaned GaAs(100) and post-heating the sample at 500 1C for 20 min in UHV. (The evaporator consists of a thin Ta foil, on the inside of which an indium metal piece is wrapped, and heating current is injected through the Ta foil which contains a small hole.) This GaAs(100)c(8 2)–In surface was oxidized by keeping the substrate temperature at around 470 1C for 5 min at 2 10 7 mbar O2 pressure in order to prepare a pre-oxidized c(4 2)–O surface structure of GaAs(100). A self-made small-scale ALD chamber, connected to the UHV system, was used to grow Al2O3 films (3 nm) at about 250 1C using trimethylaluminum (TMA) and deionized water (a TMA layer was deposited first), before transferring the Al2O3/GaAs junctions via air for ex situ characterizations. A separate XPS (Perkin Elmer) instrument with a monochromator (Al Ka source, photon energy 1486.6 eV) was used. Room-temperature PL spectra were measured using an Accent RPM2000 PL mapping tool equipped with a frequency doubled 532 nm Nd:YAG laser and a GaInAs detector array. SR-XPS measurements of the corresponding c(4 2)–O structure on InAs(100) were performed at the MAX-lab (Sweden) on beamline I311 at liquid nitrogen temperature to improve the energy resolution. The surface was prepared in situ by the method described elsewhere.39 All XPS and SR-XPS spectra were fitted with Voigt-function peaks and Shirley background subtraction by using the Origin program.37,38 A minimum number of components were included in the fittings, taking into account the spin–orbit splitting of Ga 3d, As 3d, In 3d, and In 4d core-levels. The amount of the different components was concluded by analyzing the spectral line shape, for example, shoulders and/or asymmetries appearing in the spectra, and by limiting the maximum peak width that largely depends on the photon source broadening and surface inhomogeneity in this study. The Lorentzian width was 0.15– 0.20 eV for the shallow core-levels of Ga 3d, In 4d and As 3d, and 0.20–0.30 eV for the In 3d and O 1s peaks. The Gaussian width was allowed to vary between 0.50–0.65 eV for the XPS with the monochromatized Al Ka source and 0.30–0.50 eV for the SR-XPS depending on the photon energy. In order to make spectral comparison clear, each spectrum was normalized to its maximum.
Conclusions A method to process the interface properties of the prototypical Al2O3/GaAs(100) junction grown by ALD has been presented. Incorporating the c(4 2)–O interface oxide layer into the Al2O3/GaAs stack has been demonstrated to decrease the density of harmful interface defects around the GaAs band gap. XPS spectra of the Al2O3/GaAs interfaces show that the high oxidation state of Ga (Ga2O3 type) decreases shifting instead to the In oxidation at the c(4 2)–O containing junction. The results
This journal is © the Owner Societies 2015
View Article Online
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
Paper
suggest that the deposition of about one monolayer of indium on III–V surfaces, before an oxide-film growth, is in general helpful to decrease the density of interfacial defect states. SR-XPS results provide a stringent test for the atomic model of c(4 2)–O to understand the oxidation process. It has been found that the current atomic model does not completely reproduce the measured shifts, especially positive ones of In and O above +0.40 eV. Therefore, a refinement of the simulation model of the core-level shifts and of the atomic model is suggested. The presented results contribute to developing the process method and XPS-based control of the semiconductor-oxidation effects on various device applications. In future, it is interesting to study whether similar crystalline oxidized layers can also be synthesized on the other crystal planes, (111) and (110).
Acknowledgements We thank MAX-lab staff for their assistance. This work has been supported by the Finnish Academy of Science and Letters, University of Turku Graduate School, National Doctoral Programme in Nanoscience, National Graduate School of Materials Physics, Academy of Finland (project 259213), and by CALIPSO (Coordinated Access to Lightsources to Promote Standards and Optimization, project 20120035).
Notes and references 1 Materials Fundamentals of Gate Dielectrics, ed. A. A. Demkov and A. Navrotsky, Springer, Netherlands, 2005. 2 Fundamentals of III–V Semiconductor MOSFETs, ed. S. Oktyabrsky and P. D. Ye, Springer, New York, 2010. 3 Silicon Nanocrystals: Fundamentals, Synthesis and Applications, ed. L. Pavesi and R. Turan, Wiley-VCH, Verlag, 2010. 4 K. Adlkofer, E. F. Duijs, F. Findeis, M. Bichler, A. Zrenner, E. Sackmann, G. Abstreiter and M. Tanaka, Phys. Chem. Chem. Phys., 2002, 4, 785. 5 S.-Y. Chen, C.-Y. Wang, A. C. Ford, J.-C. Chou, Y.-C. Wang, F.-Y. Wang, J. C. Ho, H.-C. Wang, A. Javey, J.-Y. Gan, L.-J. Chen and Y.-L. Chueh, Phys. Chem. Chem. Phys., 2013, 15, 2654. 6 N. B. Vandana, J. Gope, R. Singh, J. Panigrahi, S. Tyagi, P. Pathi, S. K. Srivastava, C. M. S. Rauthan and P. K. Singh, Phys. Chem. Chem. Phys., 2014, 16, 21804. 7 A. Sills and M. Califano, Phys. Chem. Chem. Phys., 2015, 17, 2573. 8 H. Hasegawa, L. He, H. Ohno, T. Sawada, T. Haga, Y. Abe and H. Takahashi, J. Vac. Sci. Technol., B: Microelectron. Process. Phenom., 1987, 5, 1097. 9 M. Passlack, M. Hong and J. P. Mannaerts, Appl. Phys. Lett., 1996, 68, 1099. 10 M. Hong, J. Kwo, A. R. Kortan, J. P. Mannaerts and A. Sergent, Science, 1999, 283, 1897. 11 M. J. Hale, S. I. Yi, J. Z. Sexton, A. C. Kummel and M. Passlack, J. Chem. Phys., 2003, 119, 6719.
This journal is © the Owner Societies 2015
PCCP
12 Y. Liang, J. Kulik, T. C. Eschrich, R. Droopad, Z. Yu and P. Maniar, Appl. Phys. Lett., 2004, 85, 1217. 13 T. Ashley, L. Buckle, S. Datta, M. T. Emeny, D. G. Hayes, K. P. Hilton, R. Jefferies, T. Martin, T. J. Phillips, D. J. Wallis, P. J. Wilding and R. Chau, Electron. Lett., 2007, 43, 777. 14 C. L. Hinkle, M. Milojevic, B. Brennan, A. M. Sonnet, F. S. Aguirre-Tostado, G. J. Hughes, E. M. Vogel and R. M. Wallace, Appl. Phys. Lett., 2009, 94, 162101. 15 C. L. Hinkle, M. Milojevic, E. M. Vogel and R. M. Wallace, Appl. Phys. Lett., 2009, 95, 151905. 16 H. Ko, K. Takei, R. Kapadia, S. Chuang, H. Fang, P. W. Leu, K. Ganapathi, E. Plis, H. S. Kim, S.-Y. Chen, M. Madsen, A. C. Ford, Y.-L. Chueh, S. Krishna, S. Salahuddin and A. Javey, Nature, 2010, 468, 286. 17 J. A. Del Alamo, Nature, 2011, 479, 317. 18 M. Scarrozza, G. Pourtois, M. Houssa, M. Heyns and A. Stesmans, Phys. Rev. B: Condens. Matter Mater. Phys., 2012, 85, 195307. 19 J. Wu, E. Lind, R. Timm, M. Hjort, A. Mikkelsen and L.-E. Wernersson, Appl. Phys. Lett., 2012, 100, 132905. 20 M. L. Huang, Y. H. Chang, T. D. Lin, H. Y. Lin, Y. T. Liu, T. W. Pi, M. Hong and J. Kwo, Appl. Phys. Lett., 2012, 101, 212101. 21 L. Lin and J. Robertson, J. Vac. Sci. Technol., B: Nanotechnol. Microelectron.: Mater., Process., Meas., Phenom., 2012, 30, 04E101. 22 X. Wang, L. Dong, J. Zhang, Y. Liu, P. D. Ye and R. G. Gordon, Nano Lett., 2013, 13, 594. 23 J. Ahn and P. C. McIntyre, Appl. Phys. Lett., 2013, 103, 251602. 24 H. Hasegawa, M. Akazawa, A. Domanowska and B. Adamowicz, Appl. Surf. Sci., 2010, 256, 5698. 25 J. L. Davis, Surf. Sci., 1964, 2, 33. 26 W. Wang, C. Gong, B. Shan, R. M. Wallace and K. Cho, Appl. Phys. Lett., 2011, 98, 232113. 27 P. D. Ye, G. D. Wilk, B. Yang, J. Kwo, S. N. G. Chu, S. Nakahara, H.-J. L. Gossmann, J. P. Mannaerts, M. Hong, K. K. Ng and J. Bude, Appl. Phys. Lett., 2003, 83, 180. 28 R. I. G. Uhrberg, H. M. Zhang and T. Balasubramanian, Phys. Rev. Lett., 2000, 85, 1036. 29 H. W. Yeom, H. Hamamatsu, T. Ohta and R. I. G. Uhrberg, Phys. Rev. B: Condens. Matter Mater. Phys., 1999, 59, 10413. 30 C. J. Nelin, F. Uhl, V. Staemmler, P. S. Bagus, Y. Fujimori, M. Sterrer, H. Kuhlenbeck and H.-J. Freund, Phys. Chem. Chem. Phys., 2014, 16, 21953. 31 M. Passlack, M. Hong, E. F. Schubert, J. R. Kwo, J. P. Mannaerts, S. N. G. Chu, N. Moriya and F. A. Thiel, Appl. Phys. Lett., 1995, 66, 625. ¨rvi, J. Salmi, A. Schramm, A. Tukiainen, M. Guina, 32 V. Poloja ¨yrynen and J. Pakarinen, E. Arola, J. Lång, I. J. Va P. Laukkanen, Appl. Phys. Lett., 2010, 97, 111109. ¨rvi, J. Salmi, P. Laukkanen and M. Guina, 33 J. Dahl, V. Poloja Appl. Phys. Lett., 2011, 99, 102105. 34 M. Yasir, J. Dahl, M. Kuzmin, J. Lång, M. Tuominen, M. P. J. ¨rvi, Punkkinen, P. Laukkanen, K. Kokko, V.-M. Korpija ¨rvi and M. Guina, Appl. Phys. Lett., 2013, 103, 191601. V. Poloja 35 F. S. Aquirre-Tostado, M. Milojevic, C. L. Hinkle, E. M. Vogel, R. M. Wallace, S. McDonnell and G. J. Hughes, Appl. Phys. Lett., 2008, 92, 171906.
Phys. Chem. Chem. Phys., 2015, 17, 7060--7066 | 7065
View Article Online
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
PCCP
36 B. Brennan and G. Hughes, J. Appl. Phys., 2010, 108, 053516. 37 M. P. J. Punkkinen, P. Laukkanen, K. Kokko, M. Ropo, ¨yrynen, H. P. Komsa, T. T. M. Ahola-Tuomi, I. J. Va ´r and Rantala, M. Pessa, M. Kuzmin, L. Vitos, J. Kolla B. Johansson, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 76, 115334. ¨la ¨, M. Ahola, S. Mattila, 38 P. Laukkanen, M. Kuzmin, R. E. Pera ¨yrynen, J. Sadowski, J. Konttinen, T. Jouhti, C. Peng, I. J. Va M. Saarinen and M. Pessa, Phys. Rev. B: Condens. Matter Mater. Phys., 2005, 72, 045321.
7066 | Phys. Chem. Chem. Phys., 2015, 17, 7060--7066
Paper
39 M. P. J. Punkkinen, P. Laukkanen, J. Lång, M. Kuzmin, M. Tuominen, V. Tuominen, J. Dahl, M. Pessa, M. Guina, ¨yrynen and L. Vitos, K. Kokko, J. Sadowski, B. Johansson, I. J. Va Phys. Rev. B: Condens. Matter Mater. Phys., 2011, 83, 195329. 40 M. Scarrozza, G. Pourtois, M. Houssa, M. Caymax, A. Stesmans, M. Meuris and M. M. Heyns, Microelectron. Eng., 2009, 86, 1747. 41 J. J. K. Lång, M. P. J. Punkkinen, P. Laukkanen, M. Kuzmin, ¨yrynen, K. Kokko, V. Tuominen, M. Pessa, M. Guina, I. J. Va B. Johansson and L. Vitos, Phys. Rev. B: Condens. Matter Mater. Phys., 2010, 81, 245305.
This journal is © the Owner Societies 2015
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
PAPER
Cite this: Phys. Chem. Chem. Phys., 2015, 17, 7060
View Journal | View Issue
Oxidation of the GaAs semiconductor at the Al2O3/GaAs junction Marjukka Tuominen,a Muhammad Yasir,a Jouko Lång,a Johnny Dahl,a Mikhail Kuzmin,a Jaakko Ma¨kela¨,a Marko Punkkinen,a Pekka Laukkanen,*a Kalevi Kokko,a Karina Schulte,b Risto Punkkinen,c Ville-Markus Korpija¨rvi,d Ville Poloja¨rvid and Mircea Guinad Atomic-scale understanding and processing of the oxidation of III–V compound–semiconductor surfaces are essential for developing materials for various devices (e.g., transistors, solar cells, and light emitting diodes). The oxidation-induced defect-rich phases at the interfaces of oxide/III–V junctions significantly affect the electrical performance of devices. In this study, a method to control the GaAs oxidation and interfacial defect density at the prototypical Al2O3/GaAs junction grown via atomic layer deposition (ALD) is demonstrated. Namely, pre-oxidation of GaAs(100) with an In-induced c(8 2) surface reconstruction, leading to a crystalline c(4 2)–O interface oxide before ALD of Al2O3, decreases band-gap defect density at the Al2O3/GaAs interface. Concomitantly, X-ray photoelectron spectroscopy (XPS) from these Al2O3/GaAs interfaces shows that the high oxidation state of Ga (Ga2O3 type) decreases, and the corresponding In2O3 type phase forms when employing the c(4 2)–O interface layer. Detailed synchrotron-radiation XPS of the counterpart c(4 2)–O oxide of InAs(100) has been utilized to elucidate
Received 19th December 2014, Accepted 2nd February 2015
the atomic structure of the useful c(4 2)–O interface layer and its oxidation process. The spectral analysis
DOI: 10.1039/c4cp05972g
between
www.rsc.org/pccp
dation. These results, discussed within the current atomic model of the c(4 2)–O interface, provide insight into the atomic structures of oxide/III–V interfaces and a way to control the semiconductor oxidation.
reveals that three different oxygen sites, five oxidation-induced group-III atomic sites with core-level shifts 0.2 eV and +1.0 eV, and hardly any oxygen-induced changes at the As sites form during the oxi-
Introduction The fundamental phenomenon of the oxidation of semiconductor crystals is one of the most technologically important chemical reactions.1–3 Its exothermic nature causes nearly all device applications to get an oxidized semiconductor surface or interface during the manufacturing of semiconductor-based devices. An oxidized semiconductor part is usually formed during the synthesis of an oxide film(s) on the semiconductor crystal or just due to the contact of material surfaces with air or oxygen-containing environments. The chemical and physical properties of oxidized semiconductor interfaces are relevant, for example, to the functionality of the microprocessor and mobile-phone transistors as well as to the performance of the window and passivation layers of high-efficiency solar cells and a
Department of Physics and Astronomy, University of Turku, FI-20014 Turku, Finland. E-mail: [email protected] b MAX IV Laboratory, Lund University, P. O. Box 118, SE-221 00 Lund, Sweden c Department of Information Technology, University of Turku, FI-20014 Turku, Finland d Optoelectronics Research Centre, Tampere University of Technology, FI-33101 Tampere, Finland
7060 | Phys. Chem. Chem. Phys., 2015, 17, 7060--7066
infrared detectors. This is because electric current flows, at least in part, in the vicinity of oxidized semiconductor interfaces in these devices. The aim toward ever decreasing device dimensions1–7 leads simultaneously to ever increasing significance of the understanding and engineering of the chemical and physical properties of the oxide/semiconductor interfaces on the atomic scale. However, these issues are still hard to predict and characterize as semiconductor oxides usually have an amorphous structure, and the interface layers are thin and lie below the surface between the oxide film and the semiconductor. The junction between an oxide film and a GaAs compound semiconductor is one of the most studied oxide/semiconductor interface systems, because GaAs is the material of various photonic devices and mobile-phone transistors. Concerning further III–V semiconductors, e.g., GaAs, GaInAs, GaSb, InP and InSb, their oxidation and oxide-interface characteristics have become more and more important because after decades of intense research and development,8–23 the III–V metal-oxide–semiconductor fieldeffect transistors (MOSFETs) are now on the roadmap of microelectronics manufacturers. However, the oxide/III–V interfaces are still problematic. III–V surfaces become easily oxidized in an uncontrolled way due to the energetically favored oxidation and
This journal is © the Owner Societies 2015
View Article Online
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
Paper
the practical difficulty to avoid the contact of III–V surfaces with oxygen during manufacturing oxide films on them. The III–V oxidation has been often found to result in a poor-quality interface layer with a high-density of defect states around the III–V band gap. This has led to a reduced electric current and a nonradiative recombination of electrons and holes, for instance. The problems can be summarized as follows: (i) oxidation causes a structural disorder leading to the formation of point defects and local deviation in bonding and (ii) oxygen incorporation into a semiconductor causes a modified electronic structure and produces an oxide phase with unwanted electronic states tailing across the band gap. This is why the success of the future semiconductor technology, in general, depends significantly on the understanding and engineering of the III–V oxidation. Indeed, much work has been done to understand and reduce the harmful oxidation-induced effects. The control techniques include sulphur passivation, depositing silicon interface layer, growth of epitaxial oxides, and removal of surface oxides via self-cleaning effect in atomic layer deposition (ALD).1–2,10,12,24–27 Furthermore, harmful electron states have been found to arise, amongst others, from group-V dimers (pairs); group-V and group-III dangling bonds (missing bonds); and III–Ox or V–Ox phases.2,14,15,18,19,21 Meanwhile, certain forms of semiconductor oxides such as Ga2O and InAsOx have been observed to be nonharmful or even to improve the electrical properties of the oxide/semiconductor interfaces.14–16 In order to develop these device materials for novel and energyefficient applications, it is essential to address the questions as to what kind of oxidation-induced changes cause the harmful defect states around the band gap, and how to process the III–V oxidation effects in a controlled manner. In this paper, a method for decreasing the amount of bandgap defects is demonstrated for the prototypical ALD-grown Al2O3/GaAs interface. Namely, covering the GaAs(100) with an ultrathin indium-induced c(8 2) surface reconstruction, followed by its pre-oxidation resulting in a crystalline c(4 2)–O oxide phase, provides a route to improve oxide/GaAs junctions during device processing. Furthermore, X-ray photoelectron spectroscopy (XPS) of the same Al2O3/GaAs junctions reveals interesting spectral features, which elucidate the role of oxidation states of the semiconductor elements in the formation of harmful interface defects at oxide/semiconductor device structures. To investigate the oxidation-induced features in detail, we have also utilized high-resolution synchrotron-radiation (SR)28–30 XPS of the corresponding c(4 2)–O phase of the oxidized InAs(100) surface, which allows the spectral analysis without complex overlapping of Ga 3d and In 4d core-level emissions. The unveiled spectral changes are important for solving the atomic structure of these interfaces and for controlling the formation of useful oxidationinduced phases.
Results and discussion Photoluminescence (PL) The comparison of PL spectra (Fig. 1) from the Al2O3-capped GaAs samples, with and without the pre-oxidized c(4 2)–O
This journal is © the Owner Societies 2015
PCCP
Fig. 1 Photoluminescence spectra from the Al2O3/GaAs junctions with and without the pre-oxidized c(4 2)–O layer on GaAs(100). The inset shows low-energy electron diffraction (LEED) from the GaAs(100) surface with the c(4 2)–O reconstruction. The white rectangle visualizes the reciprocal-space unit cell of the oxidized layer.
layer, shows that the PL intensity is higher for the sample which contains the c(4 2)–O interface structure than for the junction without the pre-oxidized phase (i.e., pure Al2O3/GaAs). The PL intensity of the pre-oxidized junction is increased by a factor of 2.2. Because the distinction between the samples is the oxide/ GaAs interface, the PL difference can be justifiably associated with the changes in interface properties. Indeed, it has been previously confirmed that the PL intensity is sensitive to the properties of a thin oxide/III–V interface layer, and higher PL intensity corresponds to lower interface defect density.1,31–34 Therefore, the presented PL results prove that the defect density of the Al2O3/GaAs interface is reduced by using the pre-oxidized c(4 2)–O layer. It is worth noting that the pure Al2O3/GaAs sample, used for the comparison, provides a stringent reference because a well-ordered GaAs starting surface, similar to that described previously,34 was prepared before the ALD growth. Future measurements such as capacitance– voltage characterizations are needed to elucidate further the electrical properties of the interface. However, the above PL results unambiguously reveal a decrease in the interface defects in the energy band gap. The reduced defect density can be connected primarily with an improved crystalline nature of the Al2O3/GaAs interface structure with the c(4 2)–O phase. The presence of indium can also affect the formation of harmful GaAs oxide phases. X-ray photoelectron spectroscopy (XPS) To clarify mechanisms behind the above finding, the Al2O3/ GaAs interfaces were characterized using XPS. This method allows nondestructive probing of thin and buried interface layers when the topmost oxide film is thin enough for the elastic photoemission. The spectral changes and so called corelevel shifts (CLS) unveiled using XPS measurements provide irreplaceable information about the atomic and electronic structures of the materials. Different CLS reflect different
Phys. Chem. Chem. Phys., 2015, 17, 7060--7066 | 7061
View Article Online
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
PCCP
Paper
in the Al2O3/c(4 2)–O/GaAs spectrum leads to the negative values seen in the lower binding energy region of the difference spectra in Fig. 2b. The In 3d emission from the Al2O3/c(4 2)–O/GaAs interface, shown in Fig. 3a, reveals that part of the In atoms have a high oxidation state at the interface since the largest separation in the In 3d components is 1.2 eV, of which oxidation state resembles In2O3.35,36 In contrast, the As 3d spectrum from the Al2O3/c(4 2)–O/GaAs interface in Fig. 3b is very similar to that of Al2O3/GaAs (not shown) and contains only two small shifts 0.36 eV and +0.37 eV, but no apparent signal from the As–As bonding around +0.6 eV or from the high oxidation state of As around +3.5 eV.35,36 Therefore, in combination with the PL finding, it can be concluded that the In2O3 type structure causes a lower density of harmful electronic states compared to Ga2O3 at the interface. The underlying reason can be that the bonding configuration of the In-containing oxides has a better crystalline order than the pure Ga-oxides at the Al2O3/GaAs interface, and/or that the In-oxide structure causes different electronic states in the GaAs gap as compared to the Ga-oxide related states. The results indicate that the deposition of small amount of In on the III–V surface before an oxide growth is helpful for avoiding the formation of the highly oxidized Ga phases that cause harmful electron states around the band gap. Fig. 2 (a) Ga 3d core-level photoelectron spectra measured through the Al2O3 films by XPS from Al2O3/GaAs interfaces. Solid line peaks represent the d5/2 parts of the double peak components and dotted line peaks mark the corresponding d3/2 parts. (b) Difference spectrum: l times Ga 3d of Al2O3/c(4 2)–O/GaAs was subtracted from Ga 3d of Al2O3/GaAs (the bulk peaks were aligned before the subtracting).
atomic bonding environments in the material, so they can be used for controlling and characterizing the properties of device interfaces, such as the oxidation of semiconductor elements during the oxide-film growth. Fig. 2a shows the fitting analysis of the Ga 3d spectra for the both junctions. In addition to the substrate bulk component, the both spectra include a second Ga 3d component labeled as GaOx around +0.5 eV with respect to the bulk peak. This corelevel shift has been identified previously as a Ga2O-type bonding.2,14,15,32,33 A close inspection reveals a spectral difference on the higher binding energy tail, since the Ga 3d emission of the pure Al2O3/GaAs interface includes a third component: GaOy around +1 eV. A similar shift has been previously related to Ga2O3-type bonding.2,14,15,32,33 The GaOy intensity is not high, and in fact, earlier studies have shown how difficult it is to resolve Ga-oxide features in the 3d emission as compared to the surface-sensitive Ga 2p emission.15 In this study the Al2O3 film was too thick to observe Ga 2p with a reasonable signal-to-noise ratio. Thus, an analysis of the difference spectrum between the Ga 3d photoemission from the Al2O3/GaAs junctions with and without the c(4 2)–O interface was performed, as presented in Fig. 2b. By subtracting the Ga 3d emission of the c(4 2)–O interface (multiplied with a factor l between 0.3–1.5) from the emission of the pure Al2O3/GaAs interface, it is seen that the GaOy feature becomes apparent. A small In 4d emission present
7062 | Phys. Chem. Chem. Phys., 2015, 17, 7060--7066
Synchrotron-radiation XPS To investigate the oxidation-induced atomic structure of the c(4 2)–O layer in more detail, we utilized high-resolution SR-XPS of the equivalent InAs(100)c(4 2)–O surface which allows a thorough group-III spectral analysis without the complex emission of the overlapping Ga 3d and In 4d peaks. The As 3d, In 4d, and O 1s spectra measured from InAs(100)c(4 2)–O by SR-XPS are shown in Fig. 4. The possibility of changing the photon energy (i.e., surface vs. bulk sensitivity of the measurement in Fig. 4) was
Fig. 3 (a) In 3d, and (b) As 3d core-level photoelectron spectra measured through the Al2O3/c(4 2)–O/GaAs interface by XPS. Solid line peaks represent the d5/2 parts of the double peak components and dotted line peaks mark the corresponding d3/2 parts.
This journal is © the Owner Societies 2015
View Article Online
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
Paper
PCCP
Fig. 4 Synchrotron-radiation XPS spectra measured at liquid-nitrogen temperature from the c(4 2)–O layer of the oxidized InAs(100): (a) As 3d, (b) In 4d, and (c) O 1s core-level spectra. Solid line peaks represent the O 1s peaks and d5/2 parts of the double peak As 3d and In 4d components, and dotted line peaks mark the corresponding d3/2 parts. The topmost spectra are measured with more bulk-sensitive photon energies than the lower spectra.
helpful to determine carefully the InAs bulk peak (B) positions and the precise oxidation-induced shifts. The As spectra include only small shifts 0.28 eV and +0.20 eV, similar to the aforementioned shifts measured from the Al2O3/ GaAs interfaces, indicating that the c(4 2)–O layer does not contain highly oxidized As or pure As–As bonding. Because there are different As-dimer sites and shifts which can be easily misinterpreted, it might be worth summarizing them: the As–As dimers can be divided between the sites where an As atom also bonds to group-III atoms, and the sites where As atom only bonds to other As atoms. The first type of As dimers exists, for example, at the clean III-As(100)(2 4) surfaces and produces only small shifts between 0.3 eV and +0.1 eV depending on the height of the As site (the clean (2 4) surface itself causes also a +0.3 eV shift).37,38 In contrast, the As dimers without the group-III bonding are formed when the surface structure becomes more As rich, for example, on the clean III-As(100)c(4 4) surface,37,38 and they produce a clear positive shift around +0.6 eV. Arsenic clusters can be also expected to cause the same clear positive shift. Thus, the small As shifts found in our samples mean that the As dimers with group-III bonding can be present. Anyhow, they do not explain the PL difference properly because the As 3d spectra for the both interfaces are very similar; it is unlikely that the enhancement of the PL intensity is associated with the changes occurring in the atomic surroundings of As species. The O 1s spectra in Fig. 4 are composed of three components at 0, +0.41, and +0.91 eV, while the In spectra of InAs(100)c(4 2)–O contain five oxidation-induced components: 0.22, +0.20, +0.46, +0.76, and +1.09 eV. Usually the direct oxygen bonding of a semiconductor element is expected
This journal is © the Owner Societies 2015
to cause a positive shift of +0.5 eV or larger. Thus, the three largest In shifts can arise from different direct In–O bonds. According to the previous studies the +0.46 shift is associated with the In2O type environment, and the +0.76 and +1.09 eV shifts with different In2O3 type structures. The two smaller In shifts are more likely to arise from indirect oxidation effects like In atoms neighboring the new In–O environments. Because the measured surface is crystalline, all these components can be associated with a single unit cell structure of the c(4 2)–O, of which surface area is only 0.74 nm2. It is striking that such a small cell includes so many different In sites. The current atomic model39 (Fig. 5) for InAs(100)c(4 2)–O provides several In shifts between 0.44 eV and +0.24 eV. These theoretical shifts were simulated for the ab initio calculated atomic structure by means of the initial-state model described elsewhere.37 Taking into account of possible experimental and computational errors in the shifts, the calculated and measured negative
Fig. 5
Atomic model suggested previously for the c(4 2)–O structure.39
Phys. Chem. Chem. Phys., 2015, 17, 7060--7066 | 7063
View Article Online
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
PCCP
shifts agree reasonably, but there is a clear deviation in the positive shifts. Furthermore, the calculated O shifts based on the current atomic model and the initial-state picture are in the range of 0.10 eV. On the other hand, the calculated and measured As shifts agree because the calculated ones also lie in the range of 0.10 eV. One possible reason for the differences between the calculated and measured shifts is that the initial-state model, used to simulate the shifts, does not describe properly the photoemission process in the present oxygen-containing system. In other words, the final-state effects cannot be excluded. Although the initial-state model has been previously shown to reproduce well the shifts of various clean and adsorbate containing semiconductor surfaces,37 neglecting the final-state in calculations might lead to the wrong binding energy for the present system because oxygen is quite different from many other adsorbate elements. Incorporating oxygen into the crystal can significantly influence the electronic band structure, and therefore, the final-state effects can play an essential role in the photoemission process. On the other hand, the current c(4 2)–O model in Fig. 5 is well justified by the calculated total energies and by the comparison of the measured and simulated scanning tunneling microscopy images.39 Thus, it is suggested both the reanalysis of the atomic model of c(4 2)–O and the simulation model for the core-level shifts of the oxidized semiconductor surfaces. If the adjustment of the atomic model is considered, the large positive In shifts and changes in the O bonding environment, observed in the experiments, imply that the c(4 2)–O surface oxide has higher oxygen concentration than 0.5 ML of the current model. Also, the In shifts of +0.76 and +1.09 eV suggest that some of the group-III atoms have more than one direct bond with oxygen (i.e., In2O3 type bonding). On the other hand, the negligible As shifts measured indicate that no direct As–O forms, as the current model in Fig. 5 shows. Overall, the total energy range of the In shifts measured from InAs(100)c(4 2)–O is consistent with the XPS results from Al2O3/c(4 2)–O/GaAs described above. This together with the As results indicates that the bonding structure of the c(4 2)–O oxide layer does not change significantly during the Al2O3 growth process, which is therefore assumed to have a negligible effect on the electronic state of the interface. Hence, the presented results suggest that an improved crystalline order at Al2O3/c(4 2)–O/GaAs junction is one reason for the improved PL, and that an increased Ga oxidation (around +1 eV, Ga2O3 type) instead leads to disorder-induced gap states at the pure Al2O3/GaAs interface, decreasing its PL intensity. This is in consistency with the previous electronicstructure calculations of oxidized GaAs and InGaAs surfaces.40
Experimental Al2O3/GaAs samples were synthesized in an Omicron multichamber UHV system. GaAs(100) pieces of size 1.0 cm 0.5 cm were cut from the same n-type (around 1 1018 cm 3 concentration of silicon dopants) wafer supplied by Wafer Technology.
7064 | Phys. Chem. Chem. Phys., 2015, 17, 7060--7066
Paper
The substrate pieces were cleaned by Ar-ion sputtering and post-heating at 550 1C until a well-defined Ga-induced c(8 2) reconstruction was seen on the surface using the low-energy electron diffraction (LEED) method. The In-induced c(8 2) surface reconstruction41 was obtained by depositing one monolayer of In from a Ta-envelope evaporator on cleaned GaAs(100) and post-heating the sample at 500 1C for 20 min in UHV. (The evaporator consists of a thin Ta foil, on the inside of which an indium metal piece is wrapped, and heating current is injected through the Ta foil which contains a small hole.) This GaAs(100)c(8 2)–In surface was oxidized by keeping the substrate temperature at around 470 1C for 5 min at 2 10 7 mbar O2 pressure in order to prepare a pre-oxidized c(4 2)–O surface structure of GaAs(100). A self-made small-scale ALD chamber, connected to the UHV system, was used to grow Al2O3 films (3 nm) at about 250 1C using trimethylaluminum (TMA) and deionized water (a TMA layer was deposited first), before transferring the Al2O3/GaAs junctions via air for ex situ characterizations. A separate XPS (Perkin Elmer) instrument with a monochromator (Al Ka source, photon energy 1486.6 eV) was used. Room-temperature PL spectra were measured using an Accent RPM2000 PL mapping tool equipped with a frequency doubled 532 nm Nd:YAG laser and a GaInAs detector array. SR-XPS measurements of the corresponding c(4 2)–O structure on InAs(100) were performed at the MAX-lab (Sweden) on beamline I311 at liquid nitrogen temperature to improve the energy resolution. The surface was prepared in situ by the method described elsewhere.39 All XPS and SR-XPS spectra were fitted with Voigt-function peaks and Shirley background subtraction by using the Origin program.37,38 A minimum number of components were included in the fittings, taking into account the spin–orbit splitting of Ga 3d, As 3d, In 3d, and In 4d core-levels. The amount of the different components was concluded by analyzing the spectral line shape, for example, shoulders and/or asymmetries appearing in the spectra, and by limiting the maximum peak width that largely depends on the photon source broadening and surface inhomogeneity in this study. The Lorentzian width was 0.15– 0.20 eV for the shallow core-levels of Ga 3d, In 4d and As 3d, and 0.20–0.30 eV for the In 3d and O 1s peaks. The Gaussian width was allowed to vary between 0.50–0.65 eV for the XPS with the monochromatized Al Ka source and 0.30–0.50 eV for the SR-XPS depending on the photon energy. In order to make spectral comparison clear, each spectrum was normalized to its maximum.
Conclusions A method to process the interface properties of the prototypical Al2O3/GaAs(100) junction grown by ALD has been presented. Incorporating the c(4 2)–O interface oxide layer into the Al2O3/GaAs stack has been demonstrated to decrease the density of harmful interface defects around the GaAs band gap. XPS spectra of the Al2O3/GaAs interfaces show that the high oxidation state of Ga (Ga2O3 type) decreases shifting instead to the In oxidation at the c(4 2)–O containing junction. The results
This journal is © the Owner Societies 2015
View Article Online
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
Paper
suggest that the deposition of about one monolayer of indium on III–V surfaces, before an oxide-film growth, is in general helpful to decrease the density of interfacial defect states. SR-XPS results provide a stringent test for the atomic model of c(4 2)–O to understand the oxidation process. It has been found that the current atomic model does not completely reproduce the measured shifts, especially positive ones of In and O above +0.40 eV. Therefore, a refinement of the simulation model of the core-level shifts and of the atomic model is suggested. The presented results contribute to developing the process method and XPS-based control of the semiconductor-oxidation effects on various device applications. In future, it is interesting to study whether similar crystalline oxidized layers can also be synthesized on the other crystal planes, (111) and (110).
Acknowledgements We thank MAX-lab staff for their assistance. This work has been supported by the Finnish Academy of Science and Letters, University of Turku Graduate School, National Doctoral Programme in Nanoscience, National Graduate School of Materials Physics, Academy of Finland (project 259213), and by CALIPSO (Coordinated Access to Lightsources to Promote Standards and Optimization, project 20120035).
Notes and references 1 Materials Fundamentals of Gate Dielectrics, ed. A. A. Demkov and A. Navrotsky, Springer, Netherlands, 2005. 2 Fundamentals of III–V Semiconductor MOSFETs, ed. S. Oktyabrsky and P. D. Ye, Springer, New York, 2010. 3 Silicon Nanocrystals: Fundamentals, Synthesis and Applications, ed. L. Pavesi and R. Turan, Wiley-VCH, Verlag, 2010. 4 K. Adlkofer, E. F. Duijs, F. Findeis, M. Bichler, A. Zrenner, E. Sackmann, G. Abstreiter and M. Tanaka, Phys. Chem. Chem. Phys., 2002, 4, 785. 5 S.-Y. Chen, C.-Y. Wang, A. C. Ford, J.-C. Chou, Y.-C. Wang, F.-Y. Wang, J. C. Ho, H.-C. Wang, A. Javey, J.-Y. Gan, L.-J. Chen and Y.-L. Chueh, Phys. Chem. Chem. Phys., 2013, 15, 2654. 6 N. B. Vandana, J. Gope, R. Singh, J. Panigrahi, S. Tyagi, P. Pathi, S. K. Srivastava, C. M. S. Rauthan and P. K. Singh, Phys. Chem. Chem. Phys., 2014, 16, 21804. 7 A. Sills and M. Califano, Phys. Chem. Chem. Phys., 2015, 17, 2573. 8 H. Hasegawa, L. He, H. Ohno, T. Sawada, T. Haga, Y. Abe and H. Takahashi, J. Vac. Sci. Technol., B: Microelectron. Process. Phenom., 1987, 5, 1097. 9 M. Passlack, M. Hong and J. P. Mannaerts, Appl. Phys. Lett., 1996, 68, 1099. 10 M. Hong, J. Kwo, A. R. Kortan, J. P. Mannaerts and A. Sergent, Science, 1999, 283, 1897. 11 M. J. Hale, S. I. Yi, J. Z. Sexton, A. C. Kummel and M. Passlack, J. Chem. Phys., 2003, 119, 6719.
This journal is © the Owner Societies 2015
PCCP
12 Y. Liang, J. Kulik, T. C. Eschrich, R. Droopad, Z. Yu and P. Maniar, Appl. Phys. Lett., 2004, 85, 1217. 13 T. Ashley, L. Buckle, S. Datta, M. T. Emeny, D. G. Hayes, K. P. Hilton, R. Jefferies, T. Martin, T. J. Phillips, D. J. Wallis, P. J. Wilding and R. Chau, Electron. Lett., 2007, 43, 777. 14 C. L. Hinkle, M. Milojevic, B. Brennan, A. M. Sonnet, F. S. Aguirre-Tostado, G. J. Hughes, E. M. Vogel and R. M. Wallace, Appl. Phys. Lett., 2009, 94, 162101. 15 C. L. Hinkle, M. Milojevic, E. M. Vogel and R. M. Wallace, Appl. Phys. Lett., 2009, 95, 151905. 16 H. Ko, K. Takei, R. Kapadia, S. Chuang, H. Fang, P. W. Leu, K. Ganapathi, E. Plis, H. S. Kim, S.-Y. Chen, M. Madsen, A. C. Ford, Y.-L. Chueh, S. Krishna, S. Salahuddin and A. Javey, Nature, 2010, 468, 286. 17 J. A. Del Alamo, Nature, 2011, 479, 317. 18 M. Scarrozza, G. Pourtois, M. Houssa, M. Heyns and A. Stesmans, Phys. Rev. B: Condens. Matter Mater. Phys., 2012, 85, 195307. 19 J. Wu, E. Lind, R. Timm, M. Hjort, A. Mikkelsen and L.-E. Wernersson, Appl. Phys. Lett., 2012, 100, 132905. 20 M. L. Huang, Y. H. Chang, T. D. Lin, H. Y. Lin, Y. T. Liu, T. W. Pi, M. Hong and J. Kwo, Appl. Phys. Lett., 2012, 101, 212101. 21 L. Lin and J. Robertson, J. Vac. Sci. Technol., B: Nanotechnol. Microelectron.: Mater., Process., Meas., Phenom., 2012, 30, 04E101. 22 X. Wang, L. Dong, J. Zhang, Y. Liu, P. D. Ye and R. G. Gordon, Nano Lett., 2013, 13, 594. 23 J. Ahn and P. C. McIntyre, Appl. Phys. Lett., 2013, 103, 251602. 24 H. Hasegawa, M. Akazawa, A. Domanowska and B. Adamowicz, Appl. Surf. Sci., 2010, 256, 5698. 25 J. L. Davis, Surf. Sci., 1964, 2, 33. 26 W. Wang, C. Gong, B. Shan, R. M. Wallace and K. Cho, Appl. Phys. Lett., 2011, 98, 232113. 27 P. D. Ye, G. D. Wilk, B. Yang, J. Kwo, S. N. G. Chu, S. Nakahara, H.-J. L. Gossmann, J. P. Mannaerts, M. Hong, K. K. Ng and J. Bude, Appl. Phys. Lett., 2003, 83, 180. 28 R. I. G. Uhrberg, H. M. Zhang and T. Balasubramanian, Phys. Rev. Lett., 2000, 85, 1036. 29 H. W. Yeom, H. Hamamatsu, T. Ohta and R. I. G. Uhrberg, Phys. Rev. B: Condens. Matter Mater. Phys., 1999, 59, 10413. 30 C. J. Nelin, F. Uhl, V. Staemmler, P. S. Bagus, Y. Fujimori, M. Sterrer, H. Kuhlenbeck and H.-J. Freund, Phys. Chem. Chem. Phys., 2014, 16, 21953. 31 M. Passlack, M. Hong, E. F. Schubert, J. R. Kwo, J. P. Mannaerts, S. N. G. Chu, N. Moriya and F. A. Thiel, Appl. Phys. Lett., 1995, 66, 625. ¨rvi, J. Salmi, A. Schramm, A. Tukiainen, M. Guina, 32 V. Poloja ¨yrynen and J. Pakarinen, E. Arola, J. Lång, I. J. Va P. Laukkanen, Appl. Phys. Lett., 2010, 97, 111109. ¨rvi, J. Salmi, P. Laukkanen and M. Guina, 33 J. Dahl, V. Poloja Appl. Phys. Lett., 2011, 99, 102105. 34 M. Yasir, J. Dahl, M. Kuzmin, J. Lång, M. Tuominen, M. P. J. ¨rvi, Punkkinen, P. Laukkanen, K. Kokko, V.-M. Korpija ¨rvi and M. Guina, Appl. Phys. Lett., 2013, 103, 191601. V. Poloja 35 F. S. Aquirre-Tostado, M. Milojevic, C. L. Hinkle, E. M. Vogel, R. M. Wallace, S. McDonnell and G. J. Hughes, Appl. Phys. Lett., 2008, 92, 171906.
Phys. Chem. Chem. Phys., 2015, 17, 7060--7066 | 7065
View Article Online
Published on 09 February 2015. Downloaded by University of Sheffield on 14/09/2015 19:07:00.
PCCP
36 B. Brennan and G. Hughes, J. Appl. Phys., 2010, 108, 053516. 37 M. P. J. Punkkinen, P. Laukkanen, K. Kokko, M. Ropo, ¨yrynen, H. P. Komsa, T. T. M. Ahola-Tuomi, I. J. Va ´r and Rantala, M. Pessa, M. Kuzmin, L. Vitos, J. Kolla B. Johansson, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 76, 115334. ¨la ¨, M. Ahola, S. Mattila, 38 P. Laukkanen, M. Kuzmin, R. E. Pera ¨yrynen, J. Sadowski, J. Konttinen, T. Jouhti, C. Peng, I. J. Va M. Saarinen and M. Pessa, Phys. Rev. B: Condens. Matter Mater. Phys., 2005, 72, 045321.
7066 | Phys. Chem. Chem. Phys., 2015, 17, 7060--7066
Paper
39 M. P. J. Punkkinen, P. Laukkanen, J. Lång, M. Kuzmin, M. Tuominen, V. Tuominen, J. Dahl, M. Pessa, M. Guina, ¨yrynen and L. Vitos, K. Kokko, J. Sadowski, B. Johansson, I. J. Va Phys. Rev. B: Condens. Matter Mater. Phys., 2011, 83, 195329. 40 M. Scarrozza, G. Pourtois, M. Houssa, M. Caymax, A. Stesmans, M. Meuris and M. M. Heyns, Microelectron. Eng., 2009, 86, 1747. 41 J. J. K. Lång, M. P. J. Punkkinen, P. Laukkanen, M. Kuzmin, ¨yrynen, K. Kokko, V. Tuominen, M. Pessa, M. Guina, I. J. Va B. Johansson and L. Vitos, Phys. Rev. B: Condens. Matter Mater. Phys., 2010, 81, 245305.
This journal is © the Owner Societies 2015
