- Email: [email protected]
Influence of Au diffusion on the formation of nanodot hexagonal ordered arrays on GaSb substrates by low energy ion sputtering
Available online at www.sciencedirect.com
Solid State Ionics 178 (2007) 1576 – 1584 www.elsevier.com/locate/ssi
Influence of Au diffusion on the formation of nanodot hexagonal ordered arrays on GaSb substrates by low energy ion sputtering J.L. Plaza ⁎, E. Diéguez Laboratorio de Crecimiento de Cristales, Departamento de Física de Materiales, Facultad de Ciencias, Universidad Autónoma de Madrid, 28049, Cantoblanco, Madrid, Spain Received 17 November 2006; received in revised form 17 September 2007; accepted 24 September 2007
Abstract We analyse the effect of Au impurities, diffused onto GaSb substrates, on the formation of nanodot hexagonal arrays created by low energy argon ion sputtering. It is concluded that oblique incidence in rotating configuration most probably delays the formation of the nanodots compared to previously reported normal incidence experiments. The presence of cracks induced by the sputtering process has been observed only in the Audiffused sample. For the same sputtering conditions, the dimensions of the dots in the Au-diffused sample are larger than in the case of the pure sample. However, they keep the same shape in both cases. Energy dispersive X-ray analysis revealed that the sputtering induces a more pronounced Ga enrichment in the case of Au-diffused sample. © 2007 Elsevier B.V. All rights reserved. Keywords: Gallium antimonide; Low energy ion sputtering; Nanodots
1. Introduction In the last few years, much effort has been directed to the development of periodic nanostructures and this has become one of the main goals in the field of nanotechnology. In particular, the formation of controlled nanostructures in semiconducting materials appears to be an important issue in the future development of the next generation of optoelectronic devices. Amongst the available nanostructuring tools, electron beam lithography (EBL) is becoming an essential technique in the fabrication of controlled and periodical nanostructures, however, due to its low throughput the role of this technique is being relegated, at the present, to the fabrication of prototypes in the laboratory environment or as support technique in the industry [1]. These difficulties are stimulating a great amount of effort for developing parallel processing techniques to be employed in the fabrication of nanostructures. Amongst them we can mention, for example, the self-assembly of semiconductor nanocrystals [2] or the Stranski–Krastanow (SK) method for the growth of semiconductor heterostructures and quantum dots [3]. ⁎ Corresponding author. Tel.: +34 91 4974784; fax: +34 91 4978579. E-mail address: [email protected] (J.L. Plaza). 0167-2738/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.ssi.2007.09.011
This technique has been recently used in order to create longrange ordered dot arrays [4–6], and fully ordered 3D quantumdot arrays (so-called 3D quantum-dot crystals) [7]. Such ordered quantum dots show extremely small size dispersions [8], narrow photoluminescence peak line widths [9], fine structure splitting and are even suitable for single photon sources [10]. It has been reported [11–13] that low energy ion sputtering (LEIS) at normal incidence, appears to be a very promising technique for the fabrication of nanostructures and quantum dots. This technique represents an attractive alternative to the SK method due to a much more simple (one small high vacuum chamber and an ion gun) and inexpensive experimental equipment need compared to a sophisticated and expensive MBEbased technique like SK. It has been demonstrated, for example, that hexagonal arrays of uniform semiconductor quantum dots, can be developed by LEIS using Ar+ ions [11,12]. In particular, it has been shown by Facsko et al. [12,13] that dot patterns can be spontaneously formed on GaSb and InSb when surfaces of these materials are sputtered with low energy argon ions at normal incidence. They also demonstrated [12] quantum confinement in quantum dots created by LEIS on GaSb layers on AlSb substrates. An additional advantage of this technique is that, by tuning the sputtering time and the energy of the incident ions, the size
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
and the density of the dots can be independently controlled. The formation of these ordered nanostructures has been extensively studied and reviewed both for metals [14] and semiconductors [15]. The self-organization of quantum-dot periodic arrays during sputtering has been attributed to the interplay between surface roughening induced by the ion sputtering and smoothing processes of the surface [11]. Gallium antimonide (GaSb) is a relevant semiconductor in the field of optoelectronics and communications (see for example the review from Dutta et al. [16] and references therein). Very recent papers have been published on the formation of nanostructures on this material by using LEIS, such the series of papers above-mentioned by Facsko et al. [11–13], who studied the formation of hexagonal nanostructured arrays and quantum dots on GaSb. We can also mention the analysis carried out by Panning et al. [17], on the p to n-type conversion on GaSb surfaces by using the same technique. However, more research is needed on the LEIS nanostructuring technique on semiconductor materials. For example, to the best of the authors' knowledge, nothing has been published on the effect of diffused dopants on the formation of nanostructures created by LEIS. It is the scope of this paper to analyse the effect of Au impurities, diffused onto GaSb substrates, on the formation of nanodots created by LEIS using Ar+ ions. Dopant diffusion in semiconductors has been traditionally used [see for example the review by Willoughby [18] and references therein] in order to fabricate semiconductor diodes and other devices based on a single piece of semiconductor material with two different kinds of conductivities (p–n). These technological applications drove us to study the effect of LEIS on impurity-diffused GaSb surfaces instead of simply using intentional doping during the GaSb crystal growth process, in which case the whole material would have only one type of conductivity (p or n). It is also the aim of this work to obtain the threshold conditions from which the nanodots are formed both in pure and Au-diffused GaSb. This is the reason why this work is presented as the result of three successive sputtering experiments, analysing the progressive changes in the surface until the nanodots are formed. Due to the surface nature of the LEIS technique, these fundamental studies, which have been developed in crystalline GaSb substrates, can also be applied to crystalline MBE or LPE-grown GaSb layers deposited on latticed-matched substrates for the fabrication of quantum-dot devices. 2. Experimental In order to carry out a systematic study of the effect of LEIS on Au-diffused GaSb substrates, it was considered necessary the analysis of three different samples submitted to LEIS. In order to fulfil this requirement without introducing possible structural differences between the substrates, a single wafer from a Bridgman-grown GaSb ingot grown in our laboratory (Ga, Sb 6N) was initially taken and polished with 1 μm alumina powder. After polishing, the samples were chemically etched in order to remove the native oxides. The formation of these oxides on GaSb surfaces, when exposed to ambient conditions is a wellknown problem [19–22]. The relevant reactions which take
1577
place in the formation of these native oxides are: Sb2O3 + 2GaSb → Ga2O3 + 4Sb [22]. It has been shown by Raman and X-ray diffraction that these native oxides are mostly formed as β-Ga2O3 and Sb2O3 phases with a possible Sb interfacial layer in between the GaSb surface and the native oxide. It has been reported in the literature that HCl-based solutions are the most effective etchant in order to remove native oxides on GaSb. In particular, HCl:CH3COOH (3:1) solutions combined with methanol rinsing reduced the thickness of the residual native oxide layer up to 1.4–1.7 nm [21]. We have used this procedure in order to prepare clean and native oxide-free GaSb surfaces. The wafer was divided into three pieces of the same size in order to prepare the three different samples. The first as-grown sample, S0, was used as a reference and it was not submitted to any further process. The second one, Sctr, served as a control sample and was exposed to the LEIS process without any other previous treatment. On the third sample, SAu, gold was thermally evaporated by using an electron gun. An Edwards auto 306 Turbo was employed for this purpose. The evaporation parameters were: evaporation time = 12 min, electron beam current = 40 mA, base pressure = 3.4 × 10− 5 mbar. The thickness of the evaporated film was 130 nm. Preliminary Au evaporation– diffusion experiments proved this thickness to be enough in order to get a homogeneous Au film and a continuous Au diffusion. In order to diffuse the Au after the evaporation process into the surface of the substrate, an annealing treatment was carried out. The annealing time was 9 h and the annealing temperature was as low as 250 °C in order to avoid high losses of Sb during the annealing. This process was performed under 300 cm3 (cubic centimeter per minute) of nitrogen flux. The metal layers were removed after the annealing process by using wet chemical etching with HCl:CH3COOH (3:1). This acid does not damage the GaSb surfaces. The etching time was 2 h immersing the sample in the solution at 60 °C in a hot plate inside a laminar flow fume-cupboard. Subsequently, the sample was rinsed in acetone and methanol. This process ensures a native oxide layer thickness below 2 nm as mentioned above. Finally the sputtering process was carried out on the pure and on the Au-diffused GaSb sample by using Ar+ ions. In order to study the progressive effect of the LEIS on the surface of the samples, three different processes were developed with different Ar ion energies and different exposure times. For the sake of clarity, the following notation will be used throughout the text: Sctr1, Sctr2, Sctr3, and SAu1, SAu2 and SAu3 will denote the pure GaSb and Au-diffused samples after the first, Table 1 Sample notation and sputtering parameters Name of Description Sputtering the sample process
Accelerating Sputtering Ion flux voltage (V) time (s) (cm− 2 s− 1)
S0 Sctr1 SAu1 Sctr2 SAu2 Sctr3 SAu3
– 500 500 600 600 600 600
Pure GaSb Pure GaSb Au-diffused Pure GaSb Au-diffused Pure GaSb Au-diffused
No sputtered First First Second Second Third Third
– 200 200 300 300 1800 1800
– 6 × 1015 6 × 1015 1.5 × 1016 1.5 × 1016 1.5 × 1016 1.5 × 1016
1578
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
second and third sputtering processes respectively. The details of the sputtering parameters of all the samples are given in Table 1. As in between the separate sputtering processes, some measurements have been carried out in ambient conditions, the samples were chemically etched in order to keep the native oxide layer to a minimum thickness (1.4–1.7 nm) as previously explained. Therefore at this level we do not expect any important effect induced by the native oxide on each subsequent sputtering process. This later statement is supported by the necessary condition of surface amorphization as reported by Facsko et al. [11–13]. This amorphization is created after the native oxide layer removal. For the case of 1.4–1.7 nm thick oxide layer, this process would be fast. In order to promote a faster nanodot formation after the first sputtering, it was decided to simultaneously increase in the next (second) sputtering process, the three sputtering parameters (accelerating voltage, sputtering time and ion flux). However for third sputtering process only the sputtering time was increased i) in order keep the regime at low energy (as the accelerating voltage, set at the second sputtering, was considered high enough) and ii) an ion flux of 1016 cm− 2 s− 1 is considered by many authors [1–3,23– 25] as the upper limit in which the conditions of the Bradley– Harper model are fulfilled [24]. The Bradley–Harper model explains the ripples and nanostructures formed during LEIS as an instability of fluctuation in the height function, which results from the competition between the roughening (induced by the sputtering which depends on the local curvature) and the smoothing (induced by the surface diffusion) of the surface. In all the cases, the incidence angle was 45° and the target (samples) rotated at 15 rpm. Several researchers have successfully obtained hexagonal arrangement of regular dots by using this set-up (oblique incidence and rotation) [12,22–24]. We decided to perform the successive sputtering processes in the same samples instead of using three different ones sputtered with three different times in order to get a closer value to the actual unknown minimum sputtering time required for the dot
formation. This has been achieved by considering the preliminary results obtained after each sputtering. The system used in our sputtering experiments was a Millatron Ion Miller from Commonwealth Scientific Corp. 3. Results and discussion 3.1. Analysis of the samples after the first sputtering process In order to compare the results, both Sctr1 and SAu1 samples were simultaneously sputtered. The parameters of the first sputtering process are shown in Table 1. They were chosen to be initially the same as those used by Facsko et al. [2] for the sputtering of amorphous and crystalline GaSb surfaces. However, there were two main differences in our experimental set-up compared to Facsko's. In the first place, in our experiment the ion flux was set to 6 × 1015 cm− 2 s− 1. This value is about 40% lower than that used by Facsko which, was 1 × 1016 cm− 2 s− 1 [2]. The second difference concerns to the geometrical arrangement of our sputtering processes. While Facsko used normal incidence, our sputtering experiments were all under rotation (15 rpm) and oblique incidence (45°). Both configurations lead to stable hexagonally ordered dot patterns [2,11–13]. The mass spectrum obtained from the secondary ions corresponding to this first sputtering process is shown in the left-hand side of Fig. 1. The mass spectra corresponding to the second and third successive LEIS processes are also shown and will be referred below. Only the relevant element Au is shown in Fig. 1 as Ga and Sb signals remained constant in the bulk material throughout the process and gave no further information. The presence of Au in sample SAu1 several nm deep into the GaSb surface shown by the Secondary Ion Mass Spectrometry ( SIMS) spectrum is a confirmation of the actual formation of Au diffusion layer. However the Au concentration decays monotonically with depth. This is a common behaviour of impurity diffusion in semiconductors [18]. We can also extract
Fig. 1. Secondary ion mass spectrum of the Au impurity obtained during the consecutive three sputtering processes from the Au-diffused sample (SAu1, SAu2, SAu3). The arrows indicate the extent of the first, second and third sputtering process respectively.
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
1579
information about the thickness of the impurity layer created during the diffusion process, according to the point where the signal reaches a flat shape. This point indicates that the diffusion layer has been almost completely removed from the surface by the erosion effect of the sputtering process. The thickness thus obtained is about 12 nm for Au. The formation of nanodot hexagonal patterns on the surface of the sample was not observed after this first LEIS process. This is not in agreement with the observed formation of dot hexagonal patterns reported by Facsko et al. [2] for the same values of the ion energy and sputtering time. However, the differences in our set-up (lower ion flux and oblique incidence with rotation), compared to Facsko's most probably delay the formation of this kind of nanostructures. 3.2. Analysis of the samples after the second sputtering process With the aim of studying further effects induced by LEIS, a further second sputtering process on samples Sctr1 and SAu1 (thus becoming samples Sctr2 and SAu2) was then carried out. As before, samples Sctr2 and SAu2 were simultaneously sputtered. In order to activate the formation of hexagonal dot patterns, the accelerating voltage, ion flux and the exposure time were further increased as shown in Table 1. The secondary ion mass spectrum for Au corresponding to this second sputtering process is also presented in Fig. 1 (part of the spectrum limited by the arrow named as “2nd”) . In this case, it can be observed that the Au concentration decreases monotonically versus depth but also indicates that the presence of this element is still detectable even at nearly 10.5 nm in depth. Several optical images were taken, in order to study in a preliminary and qualitative way the effect of this second LEIS process on pure and Au-diffused GaSb surfaces. Fig. 2. shows the optical images taken from the reference sample S0 without sputtering and from the two samples Sctr2 and SAu2 after the second sputtering process (Fig. 2B–C). It must be pointed out that, as all the samples come from the same wafer, Fig. 2A is a general reference of the surface morphology of Sctr2 and SAu2 before the sputtering. Prior to any sputtering process, polishing scratches were intentionally left on the surface of the samples (see Fig. 2A), as an optical reference of the erosion effects. The sputtered surface of Sctr2 presented in Fig. 2B shows a lower density of polishing scratches compared to the reference sample shown in Fig. 2B. This is due to the erosion of the surface induced by the impinging Ar ions. However, the presence of cracks is not observed in this sample. This is not applicable to the SAu2 sample where the presence of a high density of defects, in the form of cracks, is observed after the second sputtering as shown in Fig. 2C. This points to the fact that the formation of cracks is greatly enhanced by the presence of impurities on the GaSb surface. This is probably related to the lattice distortion generated by the impurities, evolving into cracks during the sputtering. The Au impurities diffuse most probably via interstitial sites into the GaSb [18]. An externally applied stress field is also supplied by the subsequent ion bombardment. Following the theory of Nowick and Heller [26] the response of a crystal containing
Fig. 2. Optical images from (A) S0, (B) Sctr2 and (C) SAu2 after the second sputtering.
point defects to an externally applied stress field takes the form of “stress induced ordering” or a preferential alignment of the defects. This alignment is observed in Fig. 4 where the cracks, understood as an accumulation of point defects, are more or less aligned along a certain preferential direction. It seems that the formation of this kind of defects depends on a threshold sputtering time, as they have not been observed during the shorter first sputtering process. A surface analysis has also been carried out by using atomic force microscopy, (AFM). This was done in order to determine, the morphology of the sputtered surfaces in more detail, and in order to establish whether or not there were some other unnoticed features on the surface induced by the sputtering that could not be identified by the optical microscope. The images
1580
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
resulting from this AFM analysis are presented in Fig. 3 for S0, Sctr2 and SAu2 (Fig. 3A,B,C respectively). In the case of S0 (Fig. 3A) the polishing marks can be clearly observed. However, in the case of Sctr2 and SAu2 (Fig. 3B–C) the polishing marks are smeared out as a result of the LEIS process and appear to be present in a lower density compared to S0 (Fig. 3A). This is due to the surface erosion caused by the Ar ion beam, which induces the partial removal of the polishing marks. Therefore the LEIS process effectively reduces the roughness of the samples. This effect can be quantified by a proper analysis of the AFM pictures as shown below. The three-dimensional morphology of one of the cracks appearing on the surface of SAu2 is also shown in detail in Fig. 3C. In order to get some information about the dimensions of this crack, a line profile shown in Fig. 3D, was obtained along the line drawn on the AFM image (Fig. 3C). From this profile, it can be easily observed that this crack is around 1000 nm wide by 170 nm deep. These AFM results and a further ultrahigh resolution SEM analysis showed that, as in the first sputtering process, the formation of hexagonal arrays of nanodot structures has not been attained yet. Not even after rising the sputtering time and the ion flux to the level used by Facsko et al. for the formation of hexagonally ordered nanodots. As we previously mentioned, it is likely that the differences in the sputtering configuration (45° oblique incidence and sample rotation versus normal incidence
used by Facsko et al.) might be the cause of this delay in the formation of the nanodots, even for similar ion flux and longer sputtering times. 3.3. Analysis of the samples after the third sputtering process Because of the absence of “dot-like” structures after the second sputtering, a third sputtering process was simultaneously carried out on the samples Sctr2 and SAu2 (thus becoming Sctr3 and SAu3). In order to establish some comparisons with the data above (first and second sputtering processes) it was decided to keep the same values of accelerating voltage and ion flux. Only the sputtering time was increased in this third sputtering process up to 30 min, six times longer than in the previous case (see Table 1). With this longer sputtering time, it was expected to induce the formation of nanodot hexagonal arrays. The corresponding mass spectrum from the third sputtering process is shown in Fig. 1 running along the range indicated by the arrow labelled as “3rd”. We proceed to the analysis of the effects induced by this third sputtering process as before. First, we performed an optical microscopy analysis. Fig. 4 shows the corresponding optical images of the sputtered surfaces. Fig. 4A presents the sputtered surface of Sctr3 where the polishing scratches have almost disappeared due to the erosive effect of the sputtering process.
Fig. 3. AFM images from (A) S0, (B) Sctr2 and (C) SAu2. (D) shows the line profile along the line shown in (C).
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
1581
Two typical AFM images of the surface of Sctr3 and SAu3 are shown in Fig. 6A and B respectively. In disagreement with the SEM results, some differences between the nanostructures formed onto Sctr3 and SAu3 are observed. A first look to these pictures shows that the shape of the dots in Sctr3 appears to be slightly different compared with that of the dots formed on SAu3, which reveals itself as a smoother surface. In addition, both the lateral dimensions and height of the dots in the SAu3 are slightly larger than in Sctr3. The AFM analysis enabled us to determine the 3d structure of the cracks observed in SAu3. One of these cracks is shown in Fig. 6C. In this case and due to the big difference in the scales, the nanodots are not observed. However, from the SEM analysis (Fig. 5C) it has been shown that such structure exists even inside the defects. Obviously, the surface close to these defects shows an irregular topography. This is illustrated by the line (shown in Fig. 6C) profile in Fig. 6D. From this profile it is observed that the crack depth is about 0.3 μm, even deeper than
Fig. 4. Optical images from (A) Sctr3 and (B) SAu3 after the third sputtering.
However, in the case of SAu3, apart from the removal of polishing marks, there is a noticeable increase of the crack density compared to the previous sputtering processes (Fig. 2C). This fact is clearly shown in Fig. 4B. In order to identify possible sputtering-formed nanostructures, we have also performed an UltraHigh Resolution Secondary Electron Microscopy (UHRSEM) analysis in combination with AFM. Some typical UHRSEM images are presented in Fig. 5. As we expected after increasing the sputtering time, the formation of dot arrays at the nanometre scale (about 50–80 nm in diameter) is observed similar to those reported by Facsko et al. [2–13]. However the hexagonal symmetry is slightly distorted. This distortion could be due to the different sputtering conditions (oblique incidence plus rotation) compared to the normal incidence used by Facsko. Fig. 5A shows an SEM image obtained from the Sctr3 surface after the third LEIS process. A similar image, obtained from the SAu3 is shown in Fig. 5B. It is worth noting that, apparently, the structure of the surface regarding to the geometry, arrangement and size of the nanodots, is nearly the same in both samples. However some differences between both samples have been found. We are referring to the above-mentioned formation of cracks in the diffused samples. The structure of these cracks, has also been studied in detail by SEM. An example of two of these cracks, found on the surface of SAu3 after the third sputtering process is shown in Fig. 5C being approximately 4 μm long and 1 μm wide. In order to complete the information given by the SEM, an AFM analysis in Tapping Mode has also been developed on the surface of the two samples after the third sputtering process.
Fig. 5. Ultra-high resolution SEM images from (A) Sctr3 and (B) SAu3 after the third sputtering. (C) Detail of a crack found in SAu3.
1582
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
Fig. 6. AFM images from (A) Sctr3 and (B) SAu3. (C) Detail of a crack found in SAu3. (D) shows the line profile along the line shown in (C).
the cracks produced during the second sputtering. This would indicate that the size and depth of the cracks increases with the sputtering time. A quantitative statistical analysis from AFM and SEM images (Figs. 5 and 6) of the nanodots in Scrt3 and SAu3 has been developed. This analysis has been carried out by using the Grain Analysis Module of the SPIP® (Scanning Probe Image Processor) software. Three main parameters characterizing the lateral dimensions and shape of the nanostructures have been analysed, namely, the size (mean lateral dimension, from AFM, SEM), area (from AFM, SEM) and aspect ratio (mean lateral dimension/height, from AFM). It must be pointed out that due to the irregular form of the dots (see Figs. 5 and 6) the assumption that area = π · (size/2)2 cannot be made for calculating the area. The SPIP® program takes into account the proper area of each dot and performs an statistical analysis based on these results. For the sake of brevity and considering that SEM analysis showed no differences between the samples Sctr3 and SAu3, the SEM-related statistical graphs are not presented here. However, in Fig. 7A–D we do present the column graphs from the AFM analysis. The nominal values of the different parameters are given in Fig. 8 as a compact summary of both AFM and SEM results. From this graph it is concluded that the statistical size, and area distributions obtained from SEM are nearly the same in both Sctr3 and SAu3. Therefore, from the SEM analysis, this
would mean that the dopant diffusion does not seem to have any effect on the lateral dimensions of the nanostructures created by the sputtering process. However, from the AFM results we can conclude that, the effect of the Au impurities diffused onto the GaSb surface tends to increase the size and height of the dots as the mean dot heights recorded for Sctr3 and SAu3 were 10 nm and 16 nm respectively. Although the size of the dots could be slightly over-estimated by the AFM, this over-estimation is always a constant parameter related to the tip dimensions (tip convolution) and therefore, the difference between pure GaSb and Au-diffused dot sizes, which is the important information we were looking for, is not affected. The same aspect ratio for pure and Au-diffused samples means that, although the height and the size of the dots increased in the presence of Au, their general shape remained the same. The reason why AFM and SEM results disagree is most probably related to an SEM dot size under-estimation due to the lack of secondary electrons reaching the SEM detector. This fact is attributed to the valleys in between the dots, which act as obstacles between the point of the surface where the secondary electrons are emitted from, and the detector. This results in darker regions between the dots with minimum or no difference in the contrast between the images of the two samples, thus giving no apparent difference in the dot size. Therefore the actual mean size of the dots must be somewhere in between (close to 50 nm) the size measured by the AFM and that obtained by SEM.
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
1583
Fig. 7. Statistical column graphs obtained from the AFM images shown in Fig. 6B–C. (A) Size, (B) area (C) aspect ratio (diameter/height) and (D) height versus counts which is the number of counted features with a certain value of the parameter (size, aspect ratio… etc.).
However, at this point it is necessary to make some remarks: It has been previously mentioned that, according to the SIMS analysis, by the time the nanodot formation starts (range 10– 40 nm approx. of surface material has been removed), the diffusion layer (only 12 nm deep) has almost been completely removed by the sputtering process. Considering this fact we should not, in principle, expect any major effect induced by the diffused Au impurities on the structure of the nanodots. However, as it has been previously shown by SEM and AFM, the Au impurities induced defects on the lattice appearing as cracks on the surface, which are still present even when the diffusion layer has been removed. Therefore, it can be assumed that the same kind of effects, which induced these cracks could also be the responsible for altering the structure of the dots, increasing their dimensions. However, it is expected that the “penetration depth” of the impurities diffused on the surface of GaSb samples is a key factor in order to induce further changes in the structure of the nanodots. This suggests the possibility to diffuse smaller elements or the use of non-equilibrium techniques, like ionic implantation, in order to increase the thickness of the diffusion layer and to find out whether or not there is any additional effect on the formation of nanodots induced by the presence of impurities. Such experiments are currently under way. In order to quantitatively compare the effects of the consecutive sputtering processes, a roughness analysis from the AFM measurements after the second (Fig. 3) and third (Fig. 6) sputtering process, was carried out by using the roughness
analysis module of the SPIP® software. This analysis gives the average (roughness average) obtained from the roughness values in different regions of the analysed area. The mean roughness basically gives the mean height (in nm) encountered in a surface which is not completely flat (rough surface) and the precise mathematical expression, according to international standards, can be found in the SPIP® software. The results are shown in Table 2. Due to the short sputtering time during the first process, no major difference in the roughness average was detected between the no-sputtered sample and after the first sputtering process. Nearly 50% reduction of the roughness
Fig. 8. Summarized parameters obtained from the statistical analysis from SEM (Fig. 5) and AFM (Figs. 6, 7) images using the SPIP® software.
1584
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
Table 2 Roughness average analysis from AFM analysis from pure and Au-diffused GaSb samples after the second and third sputtering Sputtering process
Roughness average (nm)
Roughness average (nm)
Pure GaSb
Au-diffused GaSb
2nd Sputt 3rd Sputt
10.4 (Sctr2) 5.37 (Sctr3)
15.2 (SAu2) 7.46 (SAu3)
arrangement of the dots is distorted, especially in the case of Au-diffused samples. The sputtering leads to smoother surfaces with smaller roughness averages. The dot structures appear when the surfaces have been smoothed up to values of the surface average in the range of 3–8 nm. From the compositional point Au-diffused samples showed more pronounced gallium enrichment. Acknowledgement
average can be noticed with the sputtering time (from the second to the third sputtering) for the two samples. A roughness analysis of the pure GaSb reference sample without any previous sputtering treatment gave a roughness average of 7.78 nm. Comparing this value with those given in Table 2, it is concluded that, in the first steps of the sputtering process, the roughness average slightly increases. However, further sputtering of the surface leads to smoother surfaces with smaller roughness averages. The nanodots appear when the surfaces have been smoothed up to surface average values in the range of 3–8 nm. Finally, a compositional analysis of the different surfaces has been developed by using energy dispersive X-ray analysis (EDX). The results have shown Ga enrichment at the surface of the irradiated samples, especially at the Au-diffused surface. This behaviour has also been observed by Facsko et al. [2] by using Auger spectroscopy in undoped GaSb sputtered samples and it is attributed to Sb losses from the surface during the sputtering. However, an exact quantification of these losses is not possible by using EDX analysis. This is due to the fact that, while these losses are expected only in the first 2–3 nm [2] the volume analysed by the EDX is close to 1 μm3. 4. Concluding remarks In conclusion we have analysed the effect of Au-diffused impurities on the formation of nanodots on GaSb substrates created by low energy argon ion sputtering. Three separate sputtering processes have been performed on both pure an Audiffused GaSb samples in order to find out which is the threshold of required eroded surface before the nanodots are formed. In both cases it has been proved that the nanodots start to be formed, somewhere in the range between 10 and 40 nm, for the experimental conditions used in this work. Although the thickness (12 nm) of the Au-diffused layer resulted to be smaller than the formation threshold, the Au impurities proved to have remarkable effects in increasing the size and height of the dots and in the formation of cracks. The formation of dot arrays at the nanometre scale was confirmed both by atomic force and ultrahigh resolution secondary electron microscopies both in pure GaSb and Au-diffused GaSb samples. The hexagonal
This work has been supported by the Ministry of Education and Science of Spain under a Ramon y Cajal Contract and by the project MAT2003-09873. References [1] A.A. Tseng, K. Chen, C.D. Chen, K.J. Ma, IEEE Trans. Electron. Packag. Manuf. 26 (2003) 141. [2] S. Facsko, T. Bobek, H. Kurz, T. Dekorsy, S. Kyrsta, R. Cremer, Appl. Phys. Lett. 80 (1) (2002) 130. [3] V.A. Shchukin, D. Bimberg, Rev. Mod. Phys. 71 (1999) 1125. [4] O.G. Schmidt, N.Y. Jin-Phillipp, C. Lange, U. Denker, K. Eberl, R. Schreiner, H. Gräbeldinger, H. Schweizer, Appl. Phys. Lett. 77 (2000) 4139. [5] H. Heidemeyer, et al., PRL 91 (2003) 196103. [6] H. Heidemeyer, et al., J. Cryst. Growth 261 (2004) 444. [7] S. Kiravittaya, O.G. Schmidt, Physica, E, Low-Dimens. Syst. Nanostruct. 23 (2004) 253. [8] S. Kiravittaya, O.G. Schmidt, Appl. Phys. Lett. 86 (2005) 206101. [9] S. Kiravittaya, A. Rastelli, O.G. Schmidt, Appl. Phys. Lett. 88 (2006) 043112. [10] S. Kiravittaya, M. Benyoucef, R. Zapf-Gottwick, A. Rastelli, O.G. Schmidt, Appl. Phys. Lett. 89 (2006) 233102. [11] S. Facsko, H. Kurz, T. Dekorsy, Phys. Rev., B 63 (2001) (165329-1-5). [12] S. Facsko, T. Dekorsy, C. Koerdt, C. Trappe, H. Kurz, A. Vogt, H.L. Hartnagel, Science 285 (1999) 1551. [13] S. Facsko, T. Bobek, T. Dekorsy, H. Kurz, Phys. Status Solidi, B Basic Res. 224 (2001) 537. [14] V. Valbusa, C. Boragno, F. Buatier de Mongeot, J. Phys., Condens. Matter 14 (2002) 8153. [15] G. Carter, J. Phys., D, Appl. Phys. 34 (2001) R1. [16] P.S. Dutta, H.L. Bhat, V. Kumar, J. Appl. Phys. 81 (1997) 5821. [17] G.N. Panin, P.S. Dutta, J. Piqueras, E. Dieguez, Appl. Phys. Lett. 67 (1995) 3584. [18] A.F.W. Willoughby, Rep. Prog. Phys. 4 (1978) 1665. [19] C.A. Wang, J. Cryst. Growth 272 (2004) 664. [20] Z.Y. Liu, D.A. Saulys, T.F. Kuecha, Appl. Phys. Lett. 85 (2004) 4391. [21] E. Papis, A. Kudla, T.T. Piotrowski, K. Golaszewska, E. Kaminska, A. Piotrowska, Mater. Sci. Semicond. Process. 4 (2001) 293. [22] G.P. Schwartz, G.J. Gualtieri, J.E. Griffiths, C.D. Thurmond, B. Schwartz, J. Electrochem. Soc. 127 (1980) 2488. [23] F. Frost, A. Schindler, F. Bigl, Phys. Rev. Lett. 85 (2000) 4116. [24] R.M. Bradley, J.M.E. Harper, J. Vac. Sci. Technol., A 6 (1988) 2390. [25] M.A. Makeev, A.L. Barabasi, Appl. Phys. Lett. 71 (1997) 2800. [26] A.S. Nowick, W.R. Heller, Adv. Phys. 12 (1963) 251.
Solid State Ionics 178 (2007) 1576 – 1584 www.elsevier.com/locate/ssi
Influence of Au diffusion on the formation of nanodot hexagonal ordered arrays on GaSb substrates by low energy ion sputtering J.L. Plaza ⁎, E. Diéguez Laboratorio de Crecimiento de Cristales, Departamento de Física de Materiales, Facultad de Ciencias, Universidad Autónoma de Madrid, 28049, Cantoblanco, Madrid, Spain Received 17 November 2006; received in revised form 17 September 2007; accepted 24 September 2007
Abstract We analyse the effect of Au impurities, diffused onto GaSb substrates, on the formation of nanodot hexagonal arrays created by low energy argon ion sputtering. It is concluded that oblique incidence in rotating configuration most probably delays the formation of the nanodots compared to previously reported normal incidence experiments. The presence of cracks induced by the sputtering process has been observed only in the Audiffused sample. For the same sputtering conditions, the dimensions of the dots in the Au-diffused sample are larger than in the case of the pure sample. However, they keep the same shape in both cases. Energy dispersive X-ray analysis revealed that the sputtering induces a more pronounced Ga enrichment in the case of Au-diffused sample. © 2007 Elsevier B.V. All rights reserved. Keywords: Gallium antimonide; Low energy ion sputtering; Nanodots
1. Introduction In the last few years, much effort has been directed to the development of periodic nanostructures and this has become one of the main goals in the field of nanotechnology. In particular, the formation of controlled nanostructures in semiconducting materials appears to be an important issue in the future development of the next generation of optoelectronic devices. Amongst the available nanostructuring tools, electron beam lithography (EBL) is becoming an essential technique in the fabrication of controlled and periodical nanostructures, however, due to its low throughput the role of this technique is being relegated, at the present, to the fabrication of prototypes in the laboratory environment or as support technique in the industry [1]. These difficulties are stimulating a great amount of effort for developing parallel processing techniques to be employed in the fabrication of nanostructures. Amongst them we can mention, for example, the self-assembly of semiconductor nanocrystals [2] or the Stranski–Krastanow (SK) method for the growth of semiconductor heterostructures and quantum dots [3]. ⁎ Corresponding author. Tel.: +34 91 4974784; fax: +34 91 4978579. E-mail address: [email protected] (J.L. Plaza). 0167-2738/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.ssi.2007.09.011
This technique has been recently used in order to create longrange ordered dot arrays [4–6], and fully ordered 3D quantumdot arrays (so-called 3D quantum-dot crystals) [7]. Such ordered quantum dots show extremely small size dispersions [8], narrow photoluminescence peak line widths [9], fine structure splitting and are even suitable for single photon sources [10]. It has been reported [11–13] that low energy ion sputtering (LEIS) at normal incidence, appears to be a very promising technique for the fabrication of nanostructures and quantum dots. This technique represents an attractive alternative to the SK method due to a much more simple (one small high vacuum chamber and an ion gun) and inexpensive experimental equipment need compared to a sophisticated and expensive MBEbased technique like SK. It has been demonstrated, for example, that hexagonal arrays of uniform semiconductor quantum dots, can be developed by LEIS using Ar+ ions [11,12]. In particular, it has been shown by Facsko et al. [12,13] that dot patterns can be spontaneously formed on GaSb and InSb when surfaces of these materials are sputtered with low energy argon ions at normal incidence. They also demonstrated [12] quantum confinement in quantum dots created by LEIS on GaSb layers on AlSb substrates. An additional advantage of this technique is that, by tuning the sputtering time and the energy of the incident ions, the size
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
and the density of the dots can be independently controlled. The formation of these ordered nanostructures has been extensively studied and reviewed both for metals [14] and semiconductors [15]. The self-organization of quantum-dot periodic arrays during sputtering has been attributed to the interplay between surface roughening induced by the ion sputtering and smoothing processes of the surface [11]. Gallium antimonide (GaSb) is a relevant semiconductor in the field of optoelectronics and communications (see for example the review from Dutta et al. [16] and references therein). Very recent papers have been published on the formation of nanostructures on this material by using LEIS, such the series of papers above-mentioned by Facsko et al. [11–13], who studied the formation of hexagonal nanostructured arrays and quantum dots on GaSb. We can also mention the analysis carried out by Panning et al. [17], on the p to n-type conversion on GaSb surfaces by using the same technique. However, more research is needed on the LEIS nanostructuring technique on semiconductor materials. For example, to the best of the authors' knowledge, nothing has been published on the effect of diffused dopants on the formation of nanostructures created by LEIS. It is the scope of this paper to analyse the effect of Au impurities, diffused onto GaSb substrates, on the formation of nanodots created by LEIS using Ar+ ions. Dopant diffusion in semiconductors has been traditionally used [see for example the review by Willoughby [18] and references therein] in order to fabricate semiconductor diodes and other devices based on a single piece of semiconductor material with two different kinds of conductivities (p–n). These technological applications drove us to study the effect of LEIS on impurity-diffused GaSb surfaces instead of simply using intentional doping during the GaSb crystal growth process, in which case the whole material would have only one type of conductivity (p or n). It is also the aim of this work to obtain the threshold conditions from which the nanodots are formed both in pure and Au-diffused GaSb. This is the reason why this work is presented as the result of three successive sputtering experiments, analysing the progressive changes in the surface until the nanodots are formed. Due to the surface nature of the LEIS technique, these fundamental studies, which have been developed in crystalline GaSb substrates, can also be applied to crystalline MBE or LPE-grown GaSb layers deposited on latticed-matched substrates for the fabrication of quantum-dot devices. 2. Experimental In order to carry out a systematic study of the effect of LEIS on Au-diffused GaSb substrates, it was considered necessary the analysis of three different samples submitted to LEIS. In order to fulfil this requirement without introducing possible structural differences between the substrates, a single wafer from a Bridgman-grown GaSb ingot grown in our laboratory (Ga, Sb 6N) was initially taken and polished with 1 μm alumina powder. After polishing, the samples were chemically etched in order to remove the native oxides. The formation of these oxides on GaSb surfaces, when exposed to ambient conditions is a wellknown problem [19–22]. The relevant reactions which take
1577
place in the formation of these native oxides are: Sb2O3 + 2GaSb → Ga2O3 + 4Sb [22]. It has been shown by Raman and X-ray diffraction that these native oxides are mostly formed as β-Ga2O3 and Sb2O3 phases with a possible Sb interfacial layer in between the GaSb surface and the native oxide. It has been reported in the literature that HCl-based solutions are the most effective etchant in order to remove native oxides on GaSb. In particular, HCl:CH3COOH (3:1) solutions combined with methanol rinsing reduced the thickness of the residual native oxide layer up to 1.4–1.7 nm [21]. We have used this procedure in order to prepare clean and native oxide-free GaSb surfaces. The wafer was divided into three pieces of the same size in order to prepare the three different samples. The first as-grown sample, S0, was used as a reference and it was not submitted to any further process. The second one, Sctr, served as a control sample and was exposed to the LEIS process without any other previous treatment. On the third sample, SAu, gold was thermally evaporated by using an electron gun. An Edwards auto 306 Turbo was employed for this purpose. The evaporation parameters were: evaporation time = 12 min, electron beam current = 40 mA, base pressure = 3.4 × 10− 5 mbar. The thickness of the evaporated film was 130 nm. Preliminary Au evaporation– diffusion experiments proved this thickness to be enough in order to get a homogeneous Au film and a continuous Au diffusion. In order to diffuse the Au after the evaporation process into the surface of the substrate, an annealing treatment was carried out. The annealing time was 9 h and the annealing temperature was as low as 250 °C in order to avoid high losses of Sb during the annealing. This process was performed under 300 cm3 (cubic centimeter per minute) of nitrogen flux. The metal layers were removed after the annealing process by using wet chemical etching with HCl:CH3COOH (3:1). This acid does not damage the GaSb surfaces. The etching time was 2 h immersing the sample in the solution at 60 °C in a hot plate inside a laminar flow fume-cupboard. Subsequently, the sample was rinsed in acetone and methanol. This process ensures a native oxide layer thickness below 2 nm as mentioned above. Finally the sputtering process was carried out on the pure and on the Au-diffused GaSb sample by using Ar+ ions. In order to study the progressive effect of the LEIS on the surface of the samples, three different processes were developed with different Ar ion energies and different exposure times. For the sake of clarity, the following notation will be used throughout the text: Sctr1, Sctr2, Sctr3, and SAu1, SAu2 and SAu3 will denote the pure GaSb and Au-diffused samples after the first, Table 1 Sample notation and sputtering parameters Name of Description Sputtering the sample process
Accelerating Sputtering Ion flux voltage (V) time (s) (cm− 2 s− 1)
S0 Sctr1 SAu1 Sctr2 SAu2 Sctr3 SAu3
– 500 500 600 600 600 600
Pure GaSb Pure GaSb Au-diffused Pure GaSb Au-diffused Pure GaSb Au-diffused
No sputtered First First Second Second Third Third
– 200 200 300 300 1800 1800
– 6 × 1015 6 × 1015 1.5 × 1016 1.5 × 1016 1.5 × 1016 1.5 × 1016
1578
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
second and third sputtering processes respectively. The details of the sputtering parameters of all the samples are given in Table 1. As in between the separate sputtering processes, some measurements have been carried out in ambient conditions, the samples were chemically etched in order to keep the native oxide layer to a minimum thickness (1.4–1.7 nm) as previously explained. Therefore at this level we do not expect any important effect induced by the native oxide on each subsequent sputtering process. This later statement is supported by the necessary condition of surface amorphization as reported by Facsko et al. [11–13]. This amorphization is created after the native oxide layer removal. For the case of 1.4–1.7 nm thick oxide layer, this process would be fast. In order to promote a faster nanodot formation after the first sputtering, it was decided to simultaneously increase in the next (second) sputtering process, the three sputtering parameters (accelerating voltage, sputtering time and ion flux). However for third sputtering process only the sputtering time was increased i) in order keep the regime at low energy (as the accelerating voltage, set at the second sputtering, was considered high enough) and ii) an ion flux of 1016 cm− 2 s− 1 is considered by many authors [1–3,23– 25] as the upper limit in which the conditions of the Bradley– Harper model are fulfilled [24]. The Bradley–Harper model explains the ripples and nanostructures formed during LEIS as an instability of fluctuation in the height function, which results from the competition between the roughening (induced by the sputtering which depends on the local curvature) and the smoothing (induced by the surface diffusion) of the surface. In all the cases, the incidence angle was 45° and the target (samples) rotated at 15 rpm. Several researchers have successfully obtained hexagonal arrangement of regular dots by using this set-up (oblique incidence and rotation) [12,22–24]. We decided to perform the successive sputtering processes in the same samples instead of using three different ones sputtered with three different times in order to get a closer value to the actual unknown minimum sputtering time required for the dot
formation. This has been achieved by considering the preliminary results obtained after each sputtering. The system used in our sputtering experiments was a Millatron Ion Miller from Commonwealth Scientific Corp. 3. Results and discussion 3.1. Analysis of the samples after the first sputtering process In order to compare the results, both Sctr1 and SAu1 samples were simultaneously sputtered. The parameters of the first sputtering process are shown in Table 1. They were chosen to be initially the same as those used by Facsko et al. [2] for the sputtering of amorphous and crystalline GaSb surfaces. However, there were two main differences in our experimental set-up compared to Facsko's. In the first place, in our experiment the ion flux was set to 6 × 1015 cm− 2 s− 1. This value is about 40% lower than that used by Facsko which, was 1 × 1016 cm− 2 s− 1 [2]. The second difference concerns to the geometrical arrangement of our sputtering processes. While Facsko used normal incidence, our sputtering experiments were all under rotation (15 rpm) and oblique incidence (45°). Both configurations lead to stable hexagonally ordered dot patterns [2,11–13]. The mass spectrum obtained from the secondary ions corresponding to this first sputtering process is shown in the left-hand side of Fig. 1. The mass spectra corresponding to the second and third successive LEIS processes are also shown and will be referred below. Only the relevant element Au is shown in Fig. 1 as Ga and Sb signals remained constant in the bulk material throughout the process and gave no further information. The presence of Au in sample SAu1 several nm deep into the GaSb surface shown by the Secondary Ion Mass Spectrometry ( SIMS) spectrum is a confirmation of the actual formation of Au diffusion layer. However the Au concentration decays monotonically with depth. This is a common behaviour of impurity diffusion in semiconductors [18]. We can also extract
Fig. 1. Secondary ion mass spectrum of the Au impurity obtained during the consecutive three sputtering processes from the Au-diffused sample (SAu1, SAu2, SAu3). The arrows indicate the extent of the first, second and third sputtering process respectively.
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
1579
information about the thickness of the impurity layer created during the diffusion process, according to the point where the signal reaches a flat shape. This point indicates that the diffusion layer has been almost completely removed from the surface by the erosion effect of the sputtering process. The thickness thus obtained is about 12 nm for Au. The formation of nanodot hexagonal patterns on the surface of the sample was not observed after this first LEIS process. This is not in agreement with the observed formation of dot hexagonal patterns reported by Facsko et al. [2] for the same values of the ion energy and sputtering time. However, the differences in our set-up (lower ion flux and oblique incidence with rotation), compared to Facsko's most probably delay the formation of this kind of nanostructures. 3.2. Analysis of the samples after the second sputtering process With the aim of studying further effects induced by LEIS, a further second sputtering process on samples Sctr1 and SAu1 (thus becoming samples Sctr2 and SAu2) was then carried out. As before, samples Sctr2 and SAu2 were simultaneously sputtered. In order to activate the formation of hexagonal dot patterns, the accelerating voltage, ion flux and the exposure time were further increased as shown in Table 1. The secondary ion mass spectrum for Au corresponding to this second sputtering process is also presented in Fig. 1 (part of the spectrum limited by the arrow named as “2nd”) . In this case, it can be observed that the Au concentration decreases monotonically versus depth but also indicates that the presence of this element is still detectable even at nearly 10.5 nm in depth. Several optical images were taken, in order to study in a preliminary and qualitative way the effect of this second LEIS process on pure and Au-diffused GaSb surfaces. Fig. 2. shows the optical images taken from the reference sample S0 without sputtering and from the two samples Sctr2 and SAu2 after the second sputtering process (Fig. 2B–C). It must be pointed out that, as all the samples come from the same wafer, Fig. 2A is a general reference of the surface morphology of Sctr2 and SAu2 before the sputtering. Prior to any sputtering process, polishing scratches were intentionally left on the surface of the samples (see Fig. 2A), as an optical reference of the erosion effects. The sputtered surface of Sctr2 presented in Fig. 2B shows a lower density of polishing scratches compared to the reference sample shown in Fig. 2B. This is due to the erosion of the surface induced by the impinging Ar ions. However, the presence of cracks is not observed in this sample. This is not applicable to the SAu2 sample where the presence of a high density of defects, in the form of cracks, is observed after the second sputtering as shown in Fig. 2C. This points to the fact that the formation of cracks is greatly enhanced by the presence of impurities on the GaSb surface. This is probably related to the lattice distortion generated by the impurities, evolving into cracks during the sputtering. The Au impurities diffuse most probably via interstitial sites into the GaSb [18]. An externally applied stress field is also supplied by the subsequent ion bombardment. Following the theory of Nowick and Heller [26] the response of a crystal containing
Fig. 2. Optical images from (A) S0, (B) Sctr2 and (C) SAu2 after the second sputtering.
point defects to an externally applied stress field takes the form of “stress induced ordering” or a preferential alignment of the defects. This alignment is observed in Fig. 4 where the cracks, understood as an accumulation of point defects, are more or less aligned along a certain preferential direction. It seems that the formation of this kind of defects depends on a threshold sputtering time, as they have not been observed during the shorter first sputtering process. A surface analysis has also been carried out by using atomic force microscopy, (AFM). This was done in order to determine, the morphology of the sputtered surfaces in more detail, and in order to establish whether or not there were some other unnoticed features on the surface induced by the sputtering that could not be identified by the optical microscope. The images
1580
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
resulting from this AFM analysis are presented in Fig. 3 for S0, Sctr2 and SAu2 (Fig. 3A,B,C respectively). In the case of S0 (Fig. 3A) the polishing marks can be clearly observed. However, in the case of Sctr2 and SAu2 (Fig. 3B–C) the polishing marks are smeared out as a result of the LEIS process and appear to be present in a lower density compared to S0 (Fig. 3A). This is due to the surface erosion caused by the Ar ion beam, which induces the partial removal of the polishing marks. Therefore the LEIS process effectively reduces the roughness of the samples. This effect can be quantified by a proper analysis of the AFM pictures as shown below. The three-dimensional morphology of one of the cracks appearing on the surface of SAu2 is also shown in detail in Fig. 3C. In order to get some information about the dimensions of this crack, a line profile shown in Fig. 3D, was obtained along the line drawn on the AFM image (Fig. 3C). From this profile, it can be easily observed that this crack is around 1000 nm wide by 170 nm deep. These AFM results and a further ultrahigh resolution SEM analysis showed that, as in the first sputtering process, the formation of hexagonal arrays of nanodot structures has not been attained yet. Not even after rising the sputtering time and the ion flux to the level used by Facsko et al. for the formation of hexagonally ordered nanodots. As we previously mentioned, it is likely that the differences in the sputtering configuration (45° oblique incidence and sample rotation versus normal incidence
used by Facsko et al.) might be the cause of this delay in the formation of the nanodots, even for similar ion flux and longer sputtering times. 3.3. Analysis of the samples after the third sputtering process Because of the absence of “dot-like” structures after the second sputtering, a third sputtering process was simultaneously carried out on the samples Sctr2 and SAu2 (thus becoming Sctr3 and SAu3). In order to establish some comparisons with the data above (first and second sputtering processes) it was decided to keep the same values of accelerating voltage and ion flux. Only the sputtering time was increased in this third sputtering process up to 30 min, six times longer than in the previous case (see Table 1). With this longer sputtering time, it was expected to induce the formation of nanodot hexagonal arrays. The corresponding mass spectrum from the third sputtering process is shown in Fig. 1 running along the range indicated by the arrow labelled as “3rd”. We proceed to the analysis of the effects induced by this third sputtering process as before. First, we performed an optical microscopy analysis. Fig. 4 shows the corresponding optical images of the sputtered surfaces. Fig. 4A presents the sputtered surface of Sctr3 where the polishing scratches have almost disappeared due to the erosive effect of the sputtering process.
Fig. 3. AFM images from (A) S0, (B) Sctr2 and (C) SAu2. (D) shows the line profile along the line shown in (C).
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
1581
Two typical AFM images of the surface of Sctr3 and SAu3 are shown in Fig. 6A and B respectively. In disagreement with the SEM results, some differences between the nanostructures formed onto Sctr3 and SAu3 are observed. A first look to these pictures shows that the shape of the dots in Sctr3 appears to be slightly different compared with that of the dots formed on SAu3, which reveals itself as a smoother surface. In addition, both the lateral dimensions and height of the dots in the SAu3 are slightly larger than in Sctr3. The AFM analysis enabled us to determine the 3d structure of the cracks observed in SAu3. One of these cracks is shown in Fig. 6C. In this case and due to the big difference in the scales, the nanodots are not observed. However, from the SEM analysis (Fig. 5C) it has been shown that such structure exists even inside the defects. Obviously, the surface close to these defects shows an irregular topography. This is illustrated by the line (shown in Fig. 6C) profile in Fig. 6D. From this profile it is observed that the crack depth is about 0.3 μm, even deeper than
Fig. 4. Optical images from (A) Sctr3 and (B) SAu3 after the third sputtering.
However, in the case of SAu3, apart from the removal of polishing marks, there is a noticeable increase of the crack density compared to the previous sputtering processes (Fig. 2C). This fact is clearly shown in Fig. 4B. In order to identify possible sputtering-formed nanostructures, we have also performed an UltraHigh Resolution Secondary Electron Microscopy (UHRSEM) analysis in combination with AFM. Some typical UHRSEM images are presented in Fig. 5. As we expected after increasing the sputtering time, the formation of dot arrays at the nanometre scale (about 50–80 nm in diameter) is observed similar to those reported by Facsko et al. [2–13]. However the hexagonal symmetry is slightly distorted. This distortion could be due to the different sputtering conditions (oblique incidence plus rotation) compared to the normal incidence used by Facsko. Fig. 5A shows an SEM image obtained from the Sctr3 surface after the third LEIS process. A similar image, obtained from the SAu3 is shown in Fig. 5B. It is worth noting that, apparently, the structure of the surface regarding to the geometry, arrangement and size of the nanodots, is nearly the same in both samples. However some differences between both samples have been found. We are referring to the above-mentioned formation of cracks in the diffused samples. The structure of these cracks, has also been studied in detail by SEM. An example of two of these cracks, found on the surface of SAu3 after the third sputtering process is shown in Fig. 5C being approximately 4 μm long and 1 μm wide. In order to complete the information given by the SEM, an AFM analysis in Tapping Mode has also been developed on the surface of the two samples after the third sputtering process.
Fig. 5. Ultra-high resolution SEM images from (A) Sctr3 and (B) SAu3 after the third sputtering. (C) Detail of a crack found in SAu3.
1582
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
Fig. 6. AFM images from (A) Sctr3 and (B) SAu3. (C) Detail of a crack found in SAu3. (D) shows the line profile along the line shown in (C).
the cracks produced during the second sputtering. This would indicate that the size and depth of the cracks increases with the sputtering time. A quantitative statistical analysis from AFM and SEM images (Figs. 5 and 6) of the nanodots in Scrt3 and SAu3 has been developed. This analysis has been carried out by using the Grain Analysis Module of the SPIP® (Scanning Probe Image Processor) software. Three main parameters characterizing the lateral dimensions and shape of the nanostructures have been analysed, namely, the size (mean lateral dimension, from AFM, SEM), area (from AFM, SEM) and aspect ratio (mean lateral dimension/height, from AFM). It must be pointed out that due to the irregular form of the dots (see Figs. 5 and 6) the assumption that area = π · (size/2)2 cannot be made for calculating the area. The SPIP® program takes into account the proper area of each dot and performs an statistical analysis based on these results. For the sake of brevity and considering that SEM analysis showed no differences between the samples Sctr3 and SAu3, the SEM-related statistical graphs are not presented here. However, in Fig. 7A–D we do present the column graphs from the AFM analysis. The nominal values of the different parameters are given in Fig. 8 as a compact summary of both AFM and SEM results. From this graph it is concluded that the statistical size, and area distributions obtained from SEM are nearly the same in both Sctr3 and SAu3. Therefore, from the SEM analysis, this
would mean that the dopant diffusion does not seem to have any effect on the lateral dimensions of the nanostructures created by the sputtering process. However, from the AFM results we can conclude that, the effect of the Au impurities diffused onto the GaSb surface tends to increase the size and height of the dots as the mean dot heights recorded for Sctr3 and SAu3 were 10 nm and 16 nm respectively. Although the size of the dots could be slightly over-estimated by the AFM, this over-estimation is always a constant parameter related to the tip dimensions (tip convolution) and therefore, the difference between pure GaSb and Au-diffused dot sizes, which is the important information we were looking for, is not affected. The same aspect ratio for pure and Au-diffused samples means that, although the height and the size of the dots increased in the presence of Au, their general shape remained the same. The reason why AFM and SEM results disagree is most probably related to an SEM dot size under-estimation due to the lack of secondary electrons reaching the SEM detector. This fact is attributed to the valleys in between the dots, which act as obstacles between the point of the surface where the secondary electrons are emitted from, and the detector. This results in darker regions between the dots with minimum or no difference in the contrast between the images of the two samples, thus giving no apparent difference in the dot size. Therefore the actual mean size of the dots must be somewhere in between (close to 50 nm) the size measured by the AFM and that obtained by SEM.
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
1583
Fig. 7. Statistical column graphs obtained from the AFM images shown in Fig. 6B–C. (A) Size, (B) area (C) aspect ratio (diameter/height) and (D) height versus counts which is the number of counted features with a certain value of the parameter (size, aspect ratio… etc.).
However, at this point it is necessary to make some remarks: It has been previously mentioned that, according to the SIMS analysis, by the time the nanodot formation starts (range 10– 40 nm approx. of surface material has been removed), the diffusion layer (only 12 nm deep) has almost been completely removed by the sputtering process. Considering this fact we should not, in principle, expect any major effect induced by the diffused Au impurities on the structure of the nanodots. However, as it has been previously shown by SEM and AFM, the Au impurities induced defects on the lattice appearing as cracks on the surface, which are still present even when the diffusion layer has been removed. Therefore, it can be assumed that the same kind of effects, which induced these cracks could also be the responsible for altering the structure of the dots, increasing their dimensions. However, it is expected that the “penetration depth” of the impurities diffused on the surface of GaSb samples is a key factor in order to induce further changes in the structure of the nanodots. This suggests the possibility to diffuse smaller elements or the use of non-equilibrium techniques, like ionic implantation, in order to increase the thickness of the diffusion layer and to find out whether or not there is any additional effect on the formation of nanodots induced by the presence of impurities. Such experiments are currently under way. In order to quantitatively compare the effects of the consecutive sputtering processes, a roughness analysis from the AFM measurements after the second (Fig. 3) and third (Fig. 6) sputtering process, was carried out by using the roughness
analysis module of the SPIP® software. This analysis gives the average (roughness average) obtained from the roughness values in different regions of the analysed area. The mean roughness basically gives the mean height (in nm) encountered in a surface which is not completely flat (rough surface) and the precise mathematical expression, according to international standards, can be found in the SPIP® software. The results are shown in Table 2. Due to the short sputtering time during the first process, no major difference in the roughness average was detected between the no-sputtered sample and after the first sputtering process. Nearly 50% reduction of the roughness
Fig. 8. Summarized parameters obtained from the statistical analysis from SEM (Fig. 5) and AFM (Figs. 6, 7) images using the SPIP® software.
1584
J.L. Plaza, E. Diéguez / Solid State Ionics 178 (2007) 1576–1584
Table 2 Roughness average analysis from AFM analysis from pure and Au-diffused GaSb samples after the second and third sputtering Sputtering process
Roughness average (nm)
Roughness average (nm)
Pure GaSb
Au-diffused GaSb
2nd Sputt 3rd Sputt
10.4 (Sctr2) 5.37 (Sctr3)
15.2 (SAu2) 7.46 (SAu3)
arrangement of the dots is distorted, especially in the case of Au-diffused samples. The sputtering leads to smoother surfaces with smaller roughness averages. The dot structures appear when the surfaces have been smoothed up to values of the surface average in the range of 3–8 nm. From the compositional point Au-diffused samples showed more pronounced gallium enrichment. Acknowledgement
average can be noticed with the sputtering time (from the second to the third sputtering) for the two samples. A roughness analysis of the pure GaSb reference sample without any previous sputtering treatment gave a roughness average of 7.78 nm. Comparing this value with those given in Table 2, it is concluded that, in the first steps of the sputtering process, the roughness average slightly increases. However, further sputtering of the surface leads to smoother surfaces with smaller roughness averages. The nanodots appear when the surfaces have been smoothed up to surface average values in the range of 3–8 nm. Finally, a compositional analysis of the different surfaces has been developed by using energy dispersive X-ray analysis (EDX). The results have shown Ga enrichment at the surface of the irradiated samples, especially at the Au-diffused surface. This behaviour has also been observed by Facsko et al. [2] by using Auger spectroscopy in undoped GaSb sputtered samples and it is attributed to Sb losses from the surface during the sputtering. However, an exact quantification of these losses is not possible by using EDX analysis. This is due to the fact that, while these losses are expected only in the first 2–3 nm [2] the volume analysed by the EDX is close to 1 μm3. 4. Concluding remarks In conclusion we have analysed the effect of Au-diffused impurities on the formation of nanodots on GaSb substrates created by low energy argon ion sputtering. Three separate sputtering processes have been performed on both pure an Audiffused GaSb samples in order to find out which is the threshold of required eroded surface before the nanodots are formed. In both cases it has been proved that the nanodots start to be formed, somewhere in the range between 10 and 40 nm, for the experimental conditions used in this work. Although the thickness (12 nm) of the Au-diffused layer resulted to be smaller than the formation threshold, the Au impurities proved to have remarkable effects in increasing the size and height of the dots and in the formation of cracks. The formation of dot arrays at the nanometre scale was confirmed both by atomic force and ultrahigh resolution secondary electron microscopies both in pure GaSb and Au-diffused GaSb samples. The hexagonal
This work has been supported by the Ministry of Education and Science of Spain under a Ramon y Cajal Contract and by the project MAT2003-09873. References [1] A.A. Tseng, K. Chen, C.D. Chen, K.J. Ma, IEEE Trans. Electron. Packag. Manuf. 26 (2003) 141. [2] S. Facsko, T. Bobek, H. Kurz, T. Dekorsy, S. Kyrsta, R. Cremer, Appl. Phys. Lett. 80 (1) (2002) 130. [3] V.A. Shchukin, D. Bimberg, Rev. Mod. Phys. 71 (1999) 1125. [4] O.G. Schmidt, N.Y. Jin-Phillipp, C. Lange, U. Denker, K. Eberl, R. Schreiner, H. Gräbeldinger, H. Schweizer, Appl. Phys. Lett. 77 (2000) 4139. [5] H. Heidemeyer, et al., PRL 91 (2003) 196103. [6] H. Heidemeyer, et al., J. Cryst. Growth 261 (2004) 444. [7] S. Kiravittaya, O.G. Schmidt, Physica, E, Low-Dimens. Syst. Nanostruct. 23 (2004) 253. [8] S. Kiravittaya, O.G. Schmidt, Appl. Phys. Lett. 86 (2005) 206101. [9] S. Kiravittaya, A. Rastelli, O.G. Schmidt, Appl. Phys. Lett. 88 (2006) 043112. [10] S. Kiravittaya, M. Benyoucef, R. Zapf-Gottwick, A. Rastelli, O.G. Schmidt, Appl. Phys. Lett. 89 (2006) 233102. [11] S. Facsko, H. Kurz, T. Dekorsy, Phys. Rev., B 63 (2001) (165329-1-5). [12] S. Facsko, T. Dekorsy, C. Koerdt, C. Trappe, H. Kurz, A. Vogt, H.L. Hartnagel, Science 285 (1999) 1551. [13] S. Facsko, T. Bobek, T. Dekorsy, H. Kurz, Phys. Status Solidi, B Basic Res. 224 (2001) 537. [14] V. Valbusa, C. Boragno, F. Buatier de Mongeot, J. Phys., Condens. Matter 14 (2002) 8153. [15] G. Carter, J. Phys., D, Appl. Phys. 34 (2001) R1. [16] P.S. Dutta, H.L. Bhat, V. Kumar, J. Appl. Phys. 81 (1997) 5821. [17] G.N. Panin, P.S. Dutta, J. Piqueras, E. Dieguez, Appl. Phys. Lett. 67 (1995) 3584. [18] A.F.W. Willoughby, Rep. Prog. Phys. 4 (1978) 1665. [19] C.A. Wang, J. Cryst. Growth 272 (2004) 664. [20] Z.Y. Liu, D.A. Saulys, T.F. Kuecha, Appl. Phys. Lett. 85 (2004) 4391. [21] E. Papis, A. Kudla, T.T. Piotrowski, K. Golaszewska, E. Kaminska, A. Piotrowska, Mater. Sci. Semicond. Process. 4 (2001) 293. [22] G.P. Schwartz, G.J. Gualtieri, J.E. Griffiths, C.D. Thurmond, B. Schwartz, J. Electrochem. Soc. 127 (1980) 2488. [23] F. Frost, A. Schindler, F. Bigl, Phys. Rev. Lett. 85 (2000) 4116. [24] R.M. Bradley, J.M.E. Harper, J. Vac. Sci. Technol., A 6 (1988) 2390. [25] M.A. Makeev, A.L. Barabasi, Appl. Phys. Lett. 71 (1997) 2800. [26] A.S. Nowick, W.R. Heller, Adv. Phys. 12 (1963) 251.