On the formation of blisters in annealed hydrogenated a-Si layers
© Serényi et al.; licensee Springer. 2013
Received: 12 September 2012
Accepted: 10 January 2013
Published: 15 February 2013
Skip to main content
© Serényi et al.; licensee Springer. 2013
Received: 12 September 2012
Accepted: 10 January 2013
Published: 15 February 2013
Differently hydrogenated radio frequency-sputtered a-Si layers have been studied by infrared (IR) spectroscopy as a function of the annealing time at 350°C with the aim to get a deeper understanding of the origin of blisters previously observed by us in a-Si/a-Ge multilayers prepared under the same conditions as the ones applied to the present a-Si layers. The H content varied between 10.8 and 17.6 at.% as measured by elastic recoil detection analysis. IR spectroscopy showed that the concentration of the clustered (Si-H) n groups and of the (Si-H2) n (n ≥ 1) polymers increased at the expense of the Si-H mono-hydrides with increasing annealing time, suggesting that there is a corresponding increase of the volume of micro-voids whose walls are assumed from literature to be decorated by the clustered mono-hydride groups and polymers. At the same time, an increase in the size of surface blisters was observed. Also, with increasing annealing time, the total concentration of bonded H of any type decreases, indicating that H is partially released from its bonds to Si. It is argued that the H released from the (Si-H) n complexes and polymers at the microvoid surfaces form molecular H2 inside the voids, whose size increases upon annealing because of the thermal expansion of the H2 gas, eventually producing plastic surface deformation in the shape of blisters.
The electrical and structural properties of hydrogenated amorphous Si, Ge and SiGe are particularly affected by the hydrogen incorporated and its bonding configuration. On one hand, H has proven to be very efficient in reducing the density of open dangling bonds responsible for deep levels in the bandgap. By hydrogenation, their density can be reduced to 1015 to 1016 cm−3 in a-Si, which is quite acceptable for device applications, e.g. in photovoltaic solar cells. On the other hand, the H bonding configuration may negatively affect the microstructure of the amorphous lattice. In a-Si, hydrogen is bonded in two modes: as randomly distributed H bonded at isolated network sites (passivating the dangling bonds) and as H bonded in the form of clusters[1, 3–6]. Smets found that H is silicon-bonded in hydrogenated di-vacancies[1, 7] for low H content. Alternatively, the H clusters are accommodated on the surfaces of voids larger than di-vacancies[4–6]. Nano- and micro-voids have been detected in a-Si[5, 7–10] as well as in a-Ge. Such voids are normally present in as-prepared amorphous materials.
As also recently pointed out by Beyer, voids are still one of the major defects in hydrogenated a-Si. Being empty spaces, they cause density reduction that can change the refractive index, electronic defect states and anomalous stress distribution especially if filled with H or if they form Si-H platelets. Furthermore, the mentioned H clusters that are situated on the inner surfaces of voids can give rise to potential fluctuations in the bulk that deteriorate the electro-optical properties[14, 15]. In a-Si, an increased concentration of Si poly-hydrides, e.g. Si-H2 di-hydrides, was seen to increase the optical bandgap and decrease the refractive index. Voids, and related H bonding configurations, are also believed to be involved in the Staebler-Wronsky effect[17, 18], i.e. degradation of the hydrogenated a-Si properties upon illumination[1, 9].
According to Beyer, cavities in the material are most crucial if they are large enough to accommodate H molecules. In such a case, in fact, hydrogen may desorb as H2 with the consequent reconstruction of dangling bonds and Si-Si weak bonds, which causes deterioration of the electronic properties. This work is a contribution in the field of the relationship between H content, H bonding configuration and voids in hydrogenated a-Si single layers deposited by radio frequency (RF) sputtering and subsequently annealed. It was prompted by the need to improve understanding of our previous results about the presence of blisters in hydrogenated a-Si/a-Ge multilayers sputtered in the same way and submitted to annealing with the aim to produce the a-SiGe alloy by Si and Ge diffusion and intermixing[19, 20]. It is reported here that annealing of the single a-Si layers causes the voids to grow to such a size to form surface blisters detectable by AFM (atomic force microscopy). By using infrared (IR) spectroscopy, it is shown that the annealing causes the formation of (Si-H) n clusters and (Si-H2) n (n ≥ 1) polymers covering the surface of voids. It is then argued that the blisters grow from such voids by accumulation of molecular H2 that had formed by reaction between H atoms released from the (Si-H) n clusters and (Si-H2) n (n ≥ 1) polymers. The results reported here support and confirm our previous hypothesis that ascribed the blisters in a-Si/a-Ge multilayers to the formation of bubbles containing molecular H2[19, 20].
The a-Si layers have been sputtered at a rate of 6.3 nm/min from a high-purity crystalline silicon target in a high-vacuum sputtering apparatus (Leybold Z400, Fergutec, Valkenswaard, The Netherlands) reaching a base pressure better than 5 × 10−5 Pa by a turbo molecular pump. The target was coupled to a RF generator (13.56 MHz) via a network for impedance matching between the generator and its load. The substrate was polished (100) silicon wafer and at a distance of 50 mm away from the target. The layer thickness was approximately 400 nm. Sputtering has been done with a mixture of high-purity argon and hydrogen gases. Both gases have been introduced continuously into the chamber by means of electronically adjustable flow controls. A 1,500-V dc wall potential has been applied to sputter the targets under a plasma pressure of 2 Pa. The samples were annealed in high-purity (99.999%) argon at 350°C for 1 and 4 h.
Controlled layer hydrogenation has been obtained by allowing H to flow continuously into the deposition chamber at different flow rates, namely 0.4, 0.8 and 1.5 ml/min, corresponding to an effective H incorporation in the as-deposited layers of 10.8, 14.7 and 17.6 at.%, respectively, as determined by elastic recoil detection analysis (ERDA). The ERDA measurements were performed with the 1.6 MeV 4He+ beam at the 5 MeV Van de Graaff accelerator of Budapest on a-Si layers 40-nm thick. The recoiled H signal was collected by an Si detector placed at 10° detecting angle to the beam direction, with the sample tilted 85° to the normal. Almost background-free ERDA spectra for H were obtained by placing a 6-μm-thick Mylar foil in front of the detector to stop the forward-scattered He ions. Further details can be found in.
where T0 is the transmission coefficient of the crystalline silicon substrate. Brodsky et al. verified that the equation is correct within ±10% only for αd > 0.1. T0 of the single-side-polished substrate was determined experimentally in relation of the transmission through a double specimen to a single one. We found that in the wavenumber region going from 3,000 to 500 cm−1, T0 monotonically decreases from 23% to 16%. This behaviour can be ascribed to the wavelength-dependent light scattering of the rough back side of the wafer.
Very often, just the integrated intensity I is used since it is proportional to the concentration of H bonds to Si apart from a constant value. This procedure is mostly used in this paper. The sample structure was analysed by AFM with a Veeco Dimension 3100 instrument (Veeco Instruments Inc., Plainview, NY, USA) in the tapping mode.
Being well established that ERDA provides very reliable absolute values of concentration, the ERDA results about the H concentration have been used to check whether IR can reliably follow the qualitative evolution of the Si-hydrogen bonding configurations as a function of annealing time. To this aim, the relative H concentration, CH = NH/NSi with NSi the atomic density of Si (5 × 1022 cm−3), was calculated from deconvoluted IR spectra in the stretching mode range as described in the ‘Methods’ section. Several values for the A of the stretching mode to be included in Equations 2 and 3 have appeared in the literature[1, 22–25]. It should be mentioned that a local oscillator strength (modified by local field effects and screening by the a-Si matrix) involved in the stretching vibrations leads to different A values which are not directly proportional to the hydrogen concentration. The strength of the 2,000 cm−1 stretching band saturates with increasing H concentration up to 6 at.%. The 2,100 cm−1 vibration continues to increase up to a level of approximately 30 at.%; therefore, at least two different values should be used. Well-accepted values are those of Amato et al. and Langford et al.. They also suggested that instead of two different values, A2000 and A2100, an average of them can be used, Aav = 1.4 × 1020 cm−2[23, 24]. Similar results can be obtained by using the proportionality A constant of Brodsky et at. scaled down by a factor of 2 as it was implicitly suggested by them as they wrote that their results are overestimated by a factor of 2[22, 25]. Among the others, Smets et al. suggested instead to use A2000 = A2100 = 9.1 × 1019 cm−2.
According to literature, the vibration mode at approximately 2,000 cm−1 is due to the presence of isolated Si-H mono-hydride bonds[3–6, 13, 16, 22–24]. Such mono-hydrides are generally isolated network sites and are associated with H bonded at isolated dangling bonds and vacancies. With increasing H concentration, the hydrogen chemical potential increases, and more complex bonding configurations can form like clustered Si-H groups in the form of Si-H platelets[3, 13], (Si-H) n groups and poly-hydrides, like Si-H2 and chains of them, (Si-H2) n [3–6, 16, 22–25]. The Si-H platelets should give an IR signature at the frequency of approximately 2,033 cm−1. An IR absorption peak that could be ascribed to Si-H platelets was only observed in the as-deposited sample hydrogenated at the lowest rate of 0.4 ml/min that exhibited a peak at 2,054 cm−1. The poly-hydride bonds instead IR vibrate at approximately 2,100 cm−1[4–6, 22–24]. The clustered (Si-H) n groups also vibrate at approximately 2,100 cm−1[4–6, 13, 16, 22–24]. The Si-H mono-hydrides do not yield any bending mode vibration, whereas Si-H2 and chains of it, (Si-H2) n , do[4–6, 13, 16, 22–24]. This was used to check the contribution of the latter poly-hydrides to the stretching mode absorption at approximately 2,100 cm−1.
Total integrated intensity (cm −1 ) of the IR stretching mode
Annealing time (h)
H = 0.4 ml/min
H = 0.8 ml/min
H = 1.5 ml/min
The origin of surface blisters that form in hydrogenated RF-sputtered a-Si layers submitted to annealing has been investigated by studying the evolution of the Si-hydrogen bonds by means of IR spectroscopy. By increasing the annealing time and/or H content, the blister size increased. Correspondingly, IR spectroscopy showed that the density of the isolated Si-H mono-hydrides decreased, while the concentration of the clustered (Si-H) n groups and (Si-H2) n , n ≥ 1, polymers increased. As both these complexes reside on the inner surfaces of voids, it is concluded that their accumulation at such surfaces favours the void size increase. It was also seen that the total amount of bonded H decreased upon annealing, suggesting that some H is released from its bonds to Si. The H liberated from the (Si-H) n groups and (Si-H2) n polymers decorating the void surfaces is expected to form molecular H2 within the voids. The expansion of the H2 gas would cause further growth of the voids up to a size able to produce surface blistering.
MS is a scientific adviser at the Institute of Technical Physics and Materials Science, Research Centre for Natural Sciences, Hungarian Academy of Sciences, Budapest, Hungary. CF is a senior scientist at the IMEM Institute of the Consiglio Nazionale delle Ricerche, Parma, Italy. ZS is a PhD student and young researcher at the Institute for Solid State Physics and Optics, Wigner Research Centre for Physics, Hungarian Academy of Sciences, Budapest, Hungary. KK is a research professor at the Institute for Solid State Physics and Optics, Wigner Research Centre for Physics, Hungarian Academy of Sciences, Budapest, Hungary. LN is a researcher at the IMEM Institute of the Consiglio Nazionale delle Ricerche, Parma, Italy. AC is a senior research associate at the Institute of Nuclear Research of the Hungarian Academy of Sciences, Hungary. NQK is senior scientist at the Institute of Technical Physics and Materials Science, Research Centre for Natural Sciences, Hungarian Academy of Sciences, Budapest, Hungary.
This work was supported by the Scientific Cooperation Agreement between CNR (Italy) and MTA (Hungary) under the contract MTA 1102, as well as by OTKA under grant nos. K-67969, NF 101329, and CK80126.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.