The role of the surfaces in the photon absorption in Ge nanoclusters embedded in silica
© Salvatore et al; licensee Springer. 2011
Received: 28 September 2010
Accepted: 11 February 2011
Published: 11 February 2011
Skip to main content
© Salvatore et al; licensee Springer. 2011
Received: 28 September 2010
Accepted: 11 February 2011
Published: 11 February 2011
The usage of semiconductor nanostructures is highly promising for boosting the energy conversion efficiency in photovoltaics technology, but still some of the underlying mechanisms are not well understood at the nanoscale length. Ge quantum dots (QDs) should have a larger absorption and a more efficient quantum confinement effect than Si ones, thus they are good candidate for third-generation solar cells. In this work, Ge QDs embedded in silica matrix have been synthesized through magnetron sputtering deposition and annealing up to 800°C. The thermal evolution of the QD size (2 to 10 nm) has been followed by transmission electron microscopy and X-ray diffraction techniques, evidencing an Ostwald ripening mechanism with a concomitant amorphous-crystalline transition. The optical absorption of Ge nanoclusters has been measured by spectrophotometry analyses, evidencing an optical bandgap of 1.6 eV, unexpectedly independent of the QDs size or of the solid phase (amorphous or crystalline). A simple modeling, based on the Tauc law, shows that the photon absorption has a much larger extent in smaller Ge QDs, being related to the surface extent rather than to the volume. These data are presented and discussed also considering the outcomes for application of Ge nanostructures in photovoltaics.
PACS: 81.07.Ta; 78.67.Hc; 68.65.-k
Nanostructured materials represent a promising route of development for photovoltaics (PV) because of the unique optical and electronic properties caused by the quantum confinement of electrons and holes, allowing to increase the efficiency of the sunlight-electricity conversion [1–8]. It has been argued that quantum dots (QDs) permit to gather a great part of solar energy in a variety of modes, among which multiple exciton generation [1, 6], intermediate band formation , or modulation of the solar absorption based on the size tuning due to the quantum confinement effect (QCE) . Actually, confined Si (2- to 5-nm QDs) shows a threshold for light absorption (optical bandgap, E g opt spanning over 2.0 to 2.8 eV [9, 10], well larger than that of bulk Si (1.1 eV) . Since the actual PV module production is largely dominated by Si (mono, poly-crystalline, or amorphous), the enhancement of energy conversion efficiency through Si-based or Si-compatible nanostructures could lead to a breakthrough in the PV market.
Recently, the variation of the Si QD optical bandgap was experimentally shown to rely not only on the size tuning but also on the deposition technique (comparing sputtering and chemical vapor deposition methods) and on the amorphous-crystalline (a-c) phase of the nanoclusters . Moreover, theoretical calculations confirmed that the amorphization of Si nanoclusters reduces the fundamental gap and increases the absorption strength [12, 13]. Some trial PV devices have been fabricated with Si QDs (size of 3 to 8 nm) embedded in SiO2, exhibiting a conversion efficiency up to 10% . In similar devices, a poor carrier transport has been evidenced as a limiting factor for cell performance and a maximum open circuit voltage of 410 mV was measured, well below that of single-junction mono-crystalline Si solar cell . Thus, at present, PV cells based on Si QDs do not show encouraging characteristics. On the other hand, passing from bulk to confined Si, E g opt hops from 1.1 to about 2.0 eV, opening a not-negligible break in the solar energy harvesting by Si. Thus, new nanostructured materials, Si compatible, are required to fill this gap.
Recently, Ge QDs are attracting a larger attention for their potential applications in PV because of the lower fabrication temperature and of the larger excitonic Bohr radius (approximately 20 nm) with respect to Si (approximately 5 nm) [11, 16], this allowing in principle an easier modulation of the electronic properties by the QCE. Moreover, since the electronic bandgap of bulk Ge (0.66 eV) is well lower than that of bulk Si (1.1 eV) , the QCE in Ge QDs could allow the modulation of E g opt within the energy range (1.1 to 2.0 eV) where bulk or confined Si fails. Up to now, Ge QDs embedded in SiO2 have been widely studied for optoelectronic applications [16–20], with a nearly size-independent photoluminescence which was not attributed to simple confinement effect but probably to the QD/matrix interface [16, 19]. Only a few studies have been performed on nanoscaled Ge clusters for PV application, mainly focused on their fabrication within SiO2 matrix [21, 22], or on the combination with titania nanoparticles . In addition, the sunlight absorption in these nanostructures has been poorly characterized, and a univocal consensus on the underlying mechanism has not been reached.
The absorption spectrum (α) of Ge QDs has been experimentally measured, and it was shown that the two main peaks visible in α of bulk Ge (i.e., the E 1 and E 2 direct transitions at 2.1 and 4.3 eV, related to the band structure of bulk Ge ) disappear by shrinking the QD size below 3 nm, suggesting that the band structure of bulk can be altered by the confinement . Later on, Tognini and co-workers evidenced a relevant blueshift of E 2 (due to the QCE) and a weakening of E 1 with size reduction of Ge QDs embedded in Al2O3, while Heath et al. concluded that E 1 and E 2 transitions are apparently unaffected by confinement in Ge QDs produced with ultrasonic methods . For PV application, the E g opt of embedded Ge QDs is a crucial parameter, but experimental measurements are still lacking. Several theoretical studies predict that it increases up to 5 eV by reducing the QD size below 1 nm, while it is fairly constant at a value of 1.5 eV for size larger than 6 nm [28, 29].
In order to verify these calculation results and to test the application of Ge QDs for PV, some open questions are whether the size of such nanostructures is the only parameter determining the sunlight absorption and to which extent, and whether there is some effect related to the structural phase (a or c) of Ge QD or to the QD-matrix interfaces. In this paper, we report an experimental investigation on the photon absorption in Ge QDs (2 to 10 nm in size) embedded in silica, providing the thermal evolution of the absorption spectra in connection with the a-c transition and the QD ripening. An optical bandgap of 1.6 eV has been found with clear evidence that light absorption is mediated by electronic states localized at the interface between Ge QDs and the hosting matrix.
Ge QDs embedded in silica have been obtained by magnetronco-sputtering of SiO2 and Ge targets (Ar atmosphere,nominal deposition temperature 400°C), upon fused silicasubstrates. Thermal annealing in the 600°C to 800°C range(1 h, N2 ambient) promoted the phase separation of SiGeOfilm into SiO2, GeO2, and Ge clusters (due toprecipitation of the exceding Ge). The thickness of the SiGeO film (approximately 280 nm) was measured by transmission electron microscopy (TEM), and the elemental composition was determined by Rutherford backscattering spectrometry (RBS, 2.0 MeV He+ beam). The spectra, simulated with SIMNRA software , revealed that in the as-deposited sample, the Si, Ge, and O contents are 24, 16, and 60 at.%, respectively, homogeneous in depth. Because of the annealing, the overall Ge amount contained in the SiGeO film slightly decreases from 3.0 × 1017 cm-2 (in the as-deposited sample) to 2.6 × 1017 cm-2 (800°C-annealed sample) due to the Ge out-diffusion through the surface, as already evidenced in the literature . Normal transmittance (T) and the 20° reflectance (R) spectra in the 200- to 2000-nm wavelength range were measured, by using a Varian Cary 500 double beam scanning UV/Visible/NIR spectrophotometer (Agilent Technologies, Inc., Santa Clara, CA, USA) for extracting the absorption coefficient of the films, as described in Ref. . Cross-section transmission electron microscopy in high resolution (HR-TEM) or scanning mode (STEM) was used to verify the formation of Ge clusters, to measure their size distribution, and to evidence the crystalline phase. The observations were carried out using a JEOL 2010F microscope (JEOL Ltd., Tokyo,Japan) operating at 200 kV equipped with a Schottky field-emission gun, a Gatan imaging filter (GIF) for compositional mappings, and a JEOL STEM unit, with an annular dark-field detector operated in high angle (HAADF) mode for Z contrast imaging. In addition, c-Ge clusters have been characterized also with glancing-incidence X-ray diffraction (GI-XRD) analysis, using the K α radiation of Cu (λ = 0.154 nm), fixing the incidence angle at 0.5° and performing the 2θ scan. Basing on the (111), (110), and (220) Bragg diffraction peaks of the GI-XRD spectra (not shown), the average QD size was estimated by applying the Scherrer formula .
where d, T S, and R S are, respectively, thickness, transmittance and reflectance of the sample, while T Q is the transmittance of the quartz substrate. The overall indetermination on α, also including errors in d, T, and R, has been estimated to be about 5%, while the dynamic range for α in our measurements was approximately 1 × 103 to 2 × 105 cm-1.
To investigate the role of the QD structural phase, we induced the c-a transition of the Ge QDs in the sample annealed at 800°C by means of an ion implantation process followed by 550°C, 1-h annealing. The ion implantation parameters (1.3 × 1014 Ge/cm2, 600 keV, max Ge concentration lower than 0.01 at.%) were chosen to induce the c-a transition in a 500-nm-thick c-Ge film, which is enough to ensure the full amorphization of our Ge QDs . Post-implant thermal treatment is needed to anneal the matrix damage without inducing re-crystallization of Ge QDs. The absorption spectrum (closed triangles) of the amorphized Ge QDs is reported in Figure 4a. The c-a transition of Ge QDs does not modify the onset of light absorption neither the spectrum itself, except that for the disappearance of the direct resonance peaks as expected because of the lost crystalline order within the Ge QDs. It should be remarked that the c-a transition in Si QDs embedded in SiO2 actually modifies the absorption by lowering the optical bandgap of about 0.4 eV . This effect has been predicted to occur in both Si and Ge QDs by theoretical calculations of the electronic bandgap [12, 13]. Thus, the data presented in this work evidence a divergence in the behavior of Ge QDs with respect to Si ones. Moreover, in Ge QDs, the α spectra at 800°C (both c- or a-Ge QDs) are halved with respect to as-deposited sample, while the Ge content reduction due to Ge out-diffusion was measured to be less than 20%. Thus, annealing at high temperatures clearly induces a not-negligible fall in the light absorption efficiency of Ge QDs, while QD structural phase does not affect the onset of light absorption.
where hν, B, and E opt g are the incoming photon energy, the Tauc constant, and the optical bandgap, respectively. The photon absorption leads to transitions between the extended electronic states from the valence band toward the conduction band, being E opt g the energy difference and B proportional to the convolution of the density of electronic states (DOS) in the two energy bands. The Tauc plots, (α hν)1/12 versus hν, of selected samples are reported with symbols in Figure 4b, while lines are the linear fit used to determine B and E opt g. For reference, a thin (120 nm) amorphous Ge film was deposited on quartz, and its Tauc plot (stars) is also reported with its fit. Tauc plots have a linear slope over a wide range of energy, and the very good agreement between fits and experimental data justifies the Tauc approach.
The optical bandgap of a-Ge results 0.8 eV, in good agreement with the literature , while the samples containing Ge QDs always exhibit an E opt g of approximately 1.6 eV (well larger than not-confined Ge), independently of the annealing temperature and of the structural phase (a or c). A similar E opt g has been reported in the literature only for one sample containing Ge QDs in a TiO2 matrix , without variation of annealing temperature or structural phase. In order to account for the E opt g of QDs, quantum confinement effect can be invoked since the size is well below the excitonic Bohr radius. In Figure 2, the QD size enlargement was reported, but it is not accomplished by a reduction of the E opt g, as expected if only the confinement rule applies. Such a contrast indicates that the confinement rule alone cannot account for the mechanism of photon absorption in Ge QDs, or it is masked by a stronger phenomenon.
These surface electronic states can be related to the presence of Ge dangling bonds or Ge-O or Ge-Si bonds located near the QD surface, or to the surface itself which induces an atomic rearrangement with different bond angle and bond length than in the bulk. To test the presence of dangling bonds, we annealed some samples (as deposited, or annealed at 700°C or 800°C) in forming gas ambient (Ar/H = 95:5 mixture, 1 h at 450°C) which is known to saturate dangling bonds in disordered structures. The optical T and R of these samples were unaffected by the forming gas treatment, so we can state that the observed behavior in the light absorption is not influenced by dangling bonds. On the other hand, a strong Fermi-level pinning near the top of valence band in bulk Ge has been recently evidenced, preventing the formation of a reliable n-channel MOSFETs device [38–40]. Such an effect was shown to be caused by native defects at the Ge surface, which modify the density of acceptor-like and donor-like states nearby the surface with respect to those in the bulk, and thus largely vary the electronic properties through a significant upwards band bending close to the surface. Actually, surface states in semiconductors typically induce a shift of the charge neutrality level (CNL) towards one of the energy bands. In Si, or in GaP or in GaAs, the CNL at the surface is located above the valence band by about one third of the respective energy bandgap , while in Ge it was recently shown to be above the valence band by only one eighth of the bandgap [38–40]. In addition, Schottky barrier heights in metal/Ge contacts are shown to be weakly dependent on the metal work functions [38–40], denouncing a very large density of interface states . Thus, Ge surface largely dominates the electronic properties nearby the surface, much more than in other semiconductors, through a strong pinning of the Fermi level and a significant band bending. Since such a band bending is expected to extend largely for undoped Ge, quantum dots as large as 10 nm can show an overwhelming surface effect on the energy band structure. In this scenario, the expected quantum confinement effect could be masked by the influence of surface states and then the theoretical calculation should reconsider these states for the optical bandgap determination.
In conclusion, we have produced and characterized Ge QDs (2 to 10 nm in size) embedded in silica by thermal annealing of a SiGeO film produced by magnetron sputtering. The light absorption spectra of the investigated Ge QDs have been measured, demonstrating that the optical bandgap of these nanostructures, both in the amorphous or crystalline phase, is pinned at about 1.6 eV, regardless of the QD size and then of the confinement extent. Moreover, we showed that for a given amount of clustered Ge, the probability of photon absorption is larger the smaller is the QD size. By modeling the photon absorption mechanism, we evidenced that it is related to the surfaces of Ge QDs rather than to their volume, through the mediation of the electronic states localized at the interface between Ge QDs and the hosting matrix. This behavior has been discussed in comparison with the Fermi-level pinning observed in metal/Ge contacts. The reported surface effect on the light absorption in Ge QDs should be kept into account for both the electronic gap calculations and for any application in photovoltaic devices. As far as the optical bandgap is concerned, Ge QDs, in conjunction with confined and bulk Si, give the chance to efficiently modulate the onset of light absorption from 1.1 eV (bulk Si) up to more than 2 eV (Si QDs).
The authors wish to thank I. Crupi and S. Gibilisco (MATIS-IMM-CNR) for the fruitful discussions, and C. Percolla and S. Tatì (MATIS-IMM-CNR) for the technical assistance.
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.