- Email: [email protected]
Study of the light emission in Ge layers and strained membranes on Si substrates
TSF-34821; No of Pages 4 Thin Solid Films xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Thin Solid Films journal homepage: www.elsevier.com/locate/tsf
Study of the light emission in Ge layers and strained membranes on Si substrates A. Gassenq a,⁎, K. Guilloy a, N. Pauc a, J.-M. Hartmann b, G. Osvaldo Dias b, D. Rouchon b, S. Tardif a, J. Escalante a, I. Duchemin a, Y.-M. Niquet a, A. Chelnokov b, V. Reboud b, Vincent Calvo a a b
Université Grenoble Alpes CEA, INAC-SP2M, F-38000 Grenoble, France CEA-LETI-Minatec Grenoble, 17 rue des Martyrs, 38054 Grenoble, France
a r t i c l e
i n f o
Article history: Received 6 July 2015 Received in revised form 16 November 2015 Accepted 17 November 2015 Available online xxxx
a b s t r a c t The influence of pattern design and tensile strain on light emission was investigated in Ge layers and suspended membranes. The optical properties were examined by micro-photoluminescence and reflectivity. Tensile strain was extracted from micro-Raman spectroscopy. It has been shown that Fabry–Pérot interference fringes can dominate the photoluminescence spectra. It is crucial to remove them in order to analyze the photoluminescence changes coming from tensile strain; especially if Fabry–Pérot oscillations are in the same energy range compared to the stress-induced spectral shift. This study highlights the fact that this interference must be taken into account in order to examine the strain in suspended Ge layers. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Integrating laser sources which are compatible with microelectronics is a major challenge in silicon photonics. To overcome the indirect nature of the Ge bandgap, n-type doping, tensile strain or Sn alloying has been proposed to fabricate low-threshold Ge-on-Si lasers [1–3]. For the strain approach, thick Ge layers grown on Si (001) substrates are naturally tensile-strained due to the difference in thermal expansion coefficients between Ge and Si that comes into play during cooling to room temperature. Membrane patterning has thus been used to focus and locally amplify the tensile strain in Ge [4,5]. Tensile strain in Ge devices may also be induced either externally [6] or by using SiN layers stressors [7,8]. Ge photoluminescence (PL) is enhanced when the crystal is tensilestrained [4,5,9,10]. However, the PL enhancement factor due to applied tensile strain is difficult to accurately quantify [4] for Ge membranes, particularly if the finite thickness of the Ge layer gives rise to Fabry– Pérot interferences (FPI) in the same energy range as the red shift due to the tensile strain. In order to reduce interferences, very thick (28 μm) [6] or ultra-thin Ge layers (24 nm) [11] can be used. They are however not fit for photonics applications which typically require a Ge thickness of around 0.35 μm, corresponding to the condition for monomode light propagation at a 2 μm wavelength. Anti-reflection (AR) coatings have been proposed to solve this problem, but without quantification of the PL enhancement [12].
⁎ Corresponding author. E-mail address: [email protected] (A. Gassenq).
We study here the impact of tensile strain on Ge membrane light emission by taking into account the influence of FPI and we quantify the effect of an AR coating. Firstly, the emission spectrum and the reflectivity of a suspended Ge layer are simulated numerically. Then, the influence of FPI on PL spectra is investigated experimentally, at first in Ge layers grown on bulk Si or Silicon-On-Insulator (SOI) substrates, without strain variations. Results from strained Ge membranes will be then detailed for uniaxial (~1.2%) and biaxial (~0.5%) tensile strains. Finally, the impact of an AR coating is quantified. 2. Modeling In order to evaluate the impact of the FPI on the PL spectrum of a Ge membrane, the emission and the reflection spectra of a suspended Ge layer have been calculated by finite difference time domain calculations using Rsoft software. Fig. 1(a) shows the simulated system: a 1 μm-thick Ge layer (nGe = 4.1 [13]) is embedded in air (nair = 1) and the detection cell is placed above, at a distance hd N 1 μm. A broad band light source is positioned at a distance hs. The normal reflectivity of the suspended layer has been calculated using hs N hd. The emission properties have been calculated by integrating the detected spectra for 0 b hs b 1 μm to obtain what we called the calculated flat PL spectrum. This is an image of PL if the generated carriers are homogeneously distributed in a layer with an isoenergetic emission. The impact of FPI on PL emission can then be evaluated using this theoretical approach. Fig. 1(b) shows the resulting flat PL and reflectivity spectra. Since the refractive index contrast between air and Ge is high, a very high fringe contrast is found. The flat PL maximum is those localized at the reflectivity minimum (e.g. close to 0.76 eV).
http://dx.doi.org/10.1016/j.tsf.2015.11.039 0040-6090/© 2015 Elsevier B.V. All rights reserved.
Please cite this article as: A. Gassenq, et al., Study of the light emission in Ge layers and strained membranes on Si substrates, Thin Solid Films (2015), http://dx.doi.org/10.1016/j.tsf.2015.11.039
2
A. Gassenq et al. / Thin Solid Films xxx (2015) xxx–xxx
Fig. 1. Reflection and emission modeling: (a) model design and (b) corresponding calculated reflectivity (solid line) and calculated flat normalized PL spectrum (dots).
3. Photoluminescence of Ge layers PL measurements were performed on Ge layers grown on bulk Si or thin SOI substrates. Since the refractive index contrast is lower for Ge on Si than for Ge on SOI, we expect the impact of FPI on PL to be smaller for Ge grown on Si than for Ge on SOI. Fig. 2(a–c) shows the PL spectra measured on in-situ phosphorous doped (1019 cm−3) Ge layers grown on bulk Si or on thinned-down optical SOI substrates. The Ge layer grown on bulk Si(001) is 0.7 ± 0.2 μm-thick and the Ge layer grown on SOI is 1.4 ± 0.2 μm-thick (starting Si thickness: close to 60 nm; buried SiO2 thickness: 2 μm). Both Ge layers were grown at 400 °C, 13 kPa using GeH4 and PH3 gaseous precursors in an Epi Centura Reduced PressureChemical Vapor Deposition machine. The tensile strain in bulk Ge layers at the end of the fabrication process is around 0.2% resulting from differences in thermal expansion coefficients between Ge and Si. Room temperature (RT) PL measurements were performed using a Horiba iHR 320 spectrometer with a 5 mW green pump laser focused to a 5 μmdiameter spot. The emitted PL signal was collected with an optical fiber, sent through the spectrometer and detected with an extended InGaAs photo-detector. In addition, macroscopic reflectivity measurements (~4 mm2 measured surface) were carried out in a Bruker IFS55 Fourier transform infrared spectrometer with a Tungsten input lamp and an InSb cooled detector. Several measurements were performed at different locations of the 200 mm wafers: the Ge layer thickness decreases slightly when moving
from center to the edges (±0.2 μm). The influence of Ge thickness on PL can easily be seen by changing the pump spot location. For Ge layers grown on bulk Si (Fig. 2(a)), Ge thickness variations have little effect on the spectrum shape. The Γ and L valley emissions are close to 1.65 μm and 1.8 μm, respectively, in good agreement with theoretical data [3]. However, thickness variations of Ge layers grown on SOI have a major impact on PL spectra because of the presence of Fabry–Pérot oscillations (coming from the high index contrast between Ge and SiO2) (Fig. 2(b)). The actual shape of the Ge emission can be extracted by fitting the maximum PL signal over all the measured thicknesses (Fig. 2(c)), which results in a thickness-independent PL spectrum with the Γ and L valley emissions at the correct energies. We have seen in the previous section that FPI shapes the emission and the reflectivity of the Ge layer. In order to study this effect, Fig. 2(d) and (e) show normalized reflectivity spectra and the associated PL spectra measured for different thicknesses of the Ge layer grown on SOI. As expected, a good agreement is found between the energies of PL maxima and reflection minima, which confirm experimentally that the PL spectral shape is strongly dominated by FPI, as can also be seen in the layer stack reflectivity. Note that this effect is highly dependent on the measurement method. Indeed, the detected oscillation contrast will depend on the numerical aperture of the measurement setup. 4. Strain induction by membrane processing Ge membranes were processed in the 0.7 ± 0.2 μm thick Ge layer grown on Si presented above to locally enhance the strain. Membrane patterning was performed by ultra-violet lithography followed by dry etching in an inductively coupled plasma reactor. Ge membrane underetching was conducted in a tetramethyl ammonium hydroxide 25% solution at 60 °C. Fig. 3(a) and (b) present the membrane designs for uniaxial and biaxial stress induction, respectively. The angle “a” is fixed at 26°. Tensile stress was tuned by changing the parameters x, d and the under-etching which is measured by Scanning Electron Microscopy (SEM) imaging using a Zeiss Ultra 55 apparatus at 15 kV operating voltage. Finite Element Method simulations using COMSOL Multiphysics were performed using a 2D linear elastic model assuming no stress along the direction perpendicular to the membrane plane. We clamped the pattern away from the region of structuration and applied the initial stress on the free standing membrane. Fig. 3(c) and (d) shows the results of simulations for both designs. In the case of the micro-cross design, the
Fig. 2. Room temperature photoluminescence spectra measured in (a) Ge grown on Si, and (b,c) Ge on SOI; (d) normalized reflectivity and (e) corresponding photoluminescence measured in Ge on SOI (di: Ge layer thicknesses).
Please cite this article as: A. Gassenq, et al., Study of the light emission in Ge layers and strained membranes on Si substrates, Thin Solid Films (2015), http://dx.doi.org/10.1016/j.tsf.2015.11.039
A. Gassenq et al. / Thin Solid Films xxx (2015) xxx–xxx
3
Fig. 3. Design and tilted SEM imaging for (a) uniaxial stress induced by micro-bridge patterning and (b) biaxial stress induced by micro-cross patterning; Finite Element Method simulations of strain in (c) uniaxially and (d) biaxially stressed Ge membranes.
bi-axial strain εbi is defined by (ε100 + ε010)/2 while the uniaxial strain εuni is along the [100] direction in the micro-bridge design. The corresponding (x,d) parameters are provided in the figures. 5. Strain and photoluminescence measurement in Ge membrane Raman spectroscopy, PL and reflectivity measurements were combined to investigate the effects of pattern design and strain on light emission. Strain was measured using a Renishaw InVia micro-Raman spectrometer with a 785 nm wavelength incident laser and a 1 μmdiameter spot. The laser intensity was low enough to avoid heating effects [14]. The Raman shift was measured by fitting the spectra with Lorentzian functions. A bulk Ge substrate was systematically used as a reference for 0% deformation. The Raman spectra measured on the Ge membranes are shown in Fig. 4. The extracted strains are indicated in the scale assuming linear relationships between strain and Raman shift, with a 424 cm−1 coefficient for bi-axial stress (micro-crosses) and a 154 cm−1 for uniaxial stress (micro-bridges) [4,15,16]. For the uniaxial design (Fig. 4(a)), two micro-bridges (d = 5 μm) are compared.
Fig. 4. Raman spectra of Ge membranes under various states of stress: (a) uniaxial stress, (b) biaxial stress, and (c) biaxial stress with AR coating.
The first (circle markers, x = 84 μm, thickness = 0.85 μm) and the second (square markers, x = 45 μm, thickness = 0.75 μm) were obtained using 3 h and 4 h 30 min under-etching times, yielding 1.1% and 1.4% strains along [100], respectively. The measured Raman spectral shifts and the corresponding strain values are provided in the figure legend. In the case of biaxial strain (Fig. 4(b) and (c)), the thickness of the Ge layer was 0.75 μm, d = 5 μm and the under-etching time was 3 h. The strain was tuned between 0.3% and 0.5% using the parameter x, as indicated in the legend. No AR layer was added to the membranes in Fig. 4(b) while the membranes in Fig. 4(c) were coated with 125 nm of SiO2 and 33 nm of amorphous Si deposited by e-gun evaporation on top of the sample at the end of the processing. Raman spectroscopy measurements were performed after the AR layer deposition. No effect of AR coating on the strain was observed. Fig. 5 shows the RT photoluminescence of the Ge membranes previously measured by Raman spectroscopy (Fig. 4). Macroscopic reflectivity measurement results performed over the chip containing the membranes of interest have been also added to the figure. Note that the reflectivity contrast is underestimated for macroscopic reflectivity measurements (with mainly a 2D Ge layer on Si) compared to the actual membrane reflectivity (suspended Ge). The fringe positions are however not strongly affected, as the Ge thickness is similar over the measured surface. Therefore, it is possible to make the same type of PL/reflectivity correlation as discussed in Sections 2 and 3. Fig. 5(a) and (b) show the data measured in the Ge micro-bridges: the PL spectra were measured from outside the bridge (location D in Fig. 3(a)) to the central part of the bridge where the strain is maximum (location A in Fig. 3(a)). The theoretical Ge emission energy corresponding to the measured strain is displayed as a vertical dashed line [17,18]. A red shift of the PL maximum is observed when the strain increases from the slightly strained part (curve D) to the highly strained part (curve A) of the membranes. However, the PL enhancement is completely different between both micro-bridges. This results from interferences which occur differently as a function of Ge layer thicknesses. Indeed, the emission maximum is not localized at the Γ bandgap but at the minimum of the reflectivity (as in the modeling section). We show here that the emission intensity in the center of the membrane is more affected by layer thickness than by induced strain. Fig. 5(c) and (d) present the micro PL measurements in the Ge micro-cross previously measured by Raman spectroscopy (Fig. 4(b) and (c)), with and without the AR coating, respectively. In the
Please cite this article as: A. Gassenq, et al., Study of the light emission in Ge layers and strained membranes on Si substrates, Thin Solid Films (2015), http://dx.doi.org/10.1016/j.tsf.2015.11.039
4
A. Gassenq et al. / Thin Solid Films xxx (2015) xxx–xxx
Fig. 5. Room temperature reflectivity and photoluminescence in Ge membranes for (a) 1.1% uniaxial strain (membrane thickness = 0.85 μm), (b) 1.4% uniaxial strain (membrane thickness = 0.75 μm), and 0.3 to 0.5% biaxial strain (c) without and (d) with anti-reflection coating. The vertical lines show the Gamma bandgap energies calculated for the measured strains at the center of the membranes.
absence of the AR coating and for low strain (0.3%) the emission is limited by FPI (high reflectivity). For higher strains, the PL intensity increases due to a combination of the red shift induced by the biaxial tensile strain and constructive interference due to lower reflectivity at longer wavelengths. Once again, the maximum of the PL emission is not localized at the bandgap. When the membranes are coated with AR layers (Fig. 5(d)), the PL spectra are then centered on the Γ bandgap. For low strained (0.3%), the PL emission is markedly larger due to interference minimization by AR layers. For higher strain (0.5%), PL enhancement is lower than in the sample without AR layers. Therefore, PL enhancement in strained Ge membranes varies dramatically if no antireflective coating is used which can lead to major interpretation errors if FPI occur in the same energy range as the red shift induced by the strain. 6. Conclusion In this article, the influence of tensile strain on Ge light emission was investigated. Strain was measured using Raman spectroscopy while the light emission properties were probed using photoluminescence. We have shown that Fabry–Pérot interferences can strongly shape the light emission. Without anti-reflection layers or adapted dimensions, straininduced PL enhancement cannot be easily measured in suspended Ge membranes. Acknowledgments The authors would like to thank the Platforme de Technologie Amont in Grenoble for the clean room facilities. This work was supported by the CEA DSM-DRT Phare projects “Photonics” and “Operando”, as well as the CEA Enhanced Eurotalent project “Straintronics”. References [1] B. Dutt, D.S. Sukhdeo, D. Nam, B.M. Vulovic, Z. Yuan, K.C. Saraswat, Roadmap to an efficient germanium-on-silicon laser: strain vs. n-type doping, IEEE Photonics J. 4 (2012) 2002–2009, http://dx.doi.org/10.1109/JPHOT.2012.2221692. [2] S. Wirths, R. Geiger, N. von den Driesch, G. Mussler, T. Stoica, S. Mantl, Z. Ikonic, M. Luysberg, S. Chiussi, J.M. Hartmann, H. Sigg, J. Faist, D. Buca, D. Grützmacher, Lasing in direct-bandgap GeSn alloy grown on Si, Nat. Photonics 9 (2015) 88–92, http://dx. doi.org/10.1038/nphoton.2014.321. [3] M. El Kurdi, G. Fishman, S. Sauvage, P. Boucaud, Band structure and optical gain of tensile-strained germanium based on a 30 band kp formalism, J. Appl. Phys. 107 (2010)http://dx.doi.org/10.1063/1.3279307.
[4] M.J. Süess, R. Geiger, R.a. Minamisawa, G. Schiefler, J. Frigerio, D. Chrastina, G. Isella, R. Spolenak, J. Faist, H. Sigg, Analysis of enhanced light emission from highly strained germanium microbridges, Nat. Photonics 7 (2013) 466–472, http://dx.doi.org/10. 1038/NPHOTON.2013.67. [5] D.S. Sukhdeo, D. Nam, J.-H. Kang, M.L. Brongersma, K.C. Saraswat, Direct bandgap germanium-on-silicon inferred from 5.7% b100N uniaxial tensile strain, Photonics Res 2 (A8) (2014)http://dx.doi.org/10.1364/PRJ.2.0000A8. [6] M. El Kurdi, H. Bertin, E. Martincic, M. de Kersauson, G. Fishman, S. Sauvage, A. Bosseboeuf, P. Boucaud, Control of direct band gap emission of bulk germanium by mechanical tensile strain, Appl. Phys. Lett. 96 (2010) 041909, http://dx.doi.org/ 10.1063/1.3297883. [7] K. Guilloy, N. Pauc, A. Gassenq, P. Gentile, S. Tardif, F. Rieutord, V. Calvo, Tensile strained germanium nanowires measured by photocurrent spectroscopy and Xray microdiffraction, Nano Lett. 15 (2015) 2429–2433, http://dx.doi.org/10.1021/ nl5048219. [8] G. Capellini, G. Kozlowski, Y. Yamamoto, M. Lisker, C. Wenger, G. Niu, P. Zaumseil, B. Tillack, A. Ghrib, M. de Kersauson, M. El Kurdi, P. Boucaud, T. Schroeder, Strain analysis in SiN/Ge microstructures obtained via Si-complementary metal oxide semiconductor compatible approach, J. Appl. Phys. 113 (2013)http://dx.doi.org/10. 1063/1.4772781. [9] Y. Huo, H. Lin, R. Chen, M. Makarova, Y. Rong, M. Li, T. Kamins, J. Vuckovic, J. Harris, Strong enhancement of direct transition photoluminescence with highly tensilestrained Ge grown by molecular beam epitaxy, Appl. Phys. Lett. 98 (2011)http:// dx.doi.org/10.1063/1.3534785. [10] D.S. Sukhdeo, D. Nam, J.-H. Kang, M.L. Brongersma, K.C. Saraswat, Bandgapcustomizable germanium using lithographically determined biaxial tensile strain for silicon-compatible optoelectronics, Opt. Express 23 (2015) 16740, http://dx. doi.org/10.1364/OE.23.016740. [11] J.R. Sanchez-Perez, C. Boztug, F. Chen, F.F. Sudradjat, D.M. Paskiewicz, R.B. Jacobson, M. Lagallya, R. Paiella, Direct-bandgap light-emitting germanium in tensilely strained nanomembranes, Proc. Natl. Acad. Sci. 108 (2011) 18893–18898, http:// dx.doi.org/10.1073/pnas.1107968108. [12] D. Nam, D. Sukhdeo, A. Roy, K. Balram, S.-L. Cheng, K.C.-Y. Huang, Z. Yuan, M. Brongersma, Y. Nishi, D. Miller, K. Saraswat, Strained germanium thin film membrane on silicon substrate for optoelectronics, Opt. Express 19 (2011) 25866, http://dx.doi.org/10.1364/OE.19.025866. [13] H.H. Li, Refractive index of silicon and germanium and its wavelength and temperature derivatives, J. Phys. Chem. Ref. Data 9 (1980) 561–658, http://dx.doi.org/10. 1063/1.555624. [14] M. Süess, R. Minamisawa, R. Geiger, K. Bourdelle, H. Sigg, R. Spolenak, Powerdependent Raman analysis of highly strained Si nanobridges, Nano Lett. 14 (2014) 1249–1254, http://dx.doi.org/10.1021/nl404152r. [15] I. De Wolf, Micro-Raman spectroscopy to study local mechanical stress in silicon integrated circuits, Semicond. Sci. Technol. 11 (1996) 139, http://dx.doi.org/10.1088/ 0268-1242/11/2/001. [16] F. Cerdeira, C.J. Buchenauer, F.H. Pollak, M. Cardona, Stress-induced Raman shifts of diamond and zincblende semicond, Phys. Rev. B 5 (1972) 580, http://dx.doi.org/10. 1103/PhysRevB.5.580. [17] C.G. de Walle, Band lineups and deformation potentials in the model-solid theory, Phys. Rev. B 39 (1989) 1871–1883, http://dx.doi.org/10.1103/PhysRevB.39.1871. [18] C. Boztug, J.R. Sánchez-Pérez, F.F. Sudradjat, R. Jacobson, D.M. Paskiewicz, M.G. Lagally, R. Paiella, Tensilely strained germanium nanomembranes as infrared optical gain media, Small 9 (2013) 622–630, http://dx.doi.org/10.1002/smll.201201090.
Please cite this article as: A. Gassenq, et al., Study of the light emission in Ge layers and strained membranes on Si substrates, Thin Solid Films (2015), http://dx.doi.org/10.1016/j.tsf.2015.11.039
Contents lists available at ScienceDirect
Thin Solid Films journal homepage: www.elsevier.com/locate/tsf
Study of the light emission in Ge layers and strained membranes on Si substrates A. Gassenq a,⁎, K. Guilloy a, N. Pauc a, J.-M. Hartmann b, G. Osvaldo Dias b, D. Rouchon b, S. Tardif a, J. Escalante a, I. Duchemin a, Y.-M. Niquet a, A. Chelnokov b, V. Reboud b, Vincent Calvo a a b
Université Grenoble Alpes CEA, INAC-SP2M, F-38000 Grenoble, France CEA-LETI-Minatec Grenoble, 17 rue des Martyrs, 38054 Grenoble, France
a r t i c l e
i n f o
Article history: Received 6 July 2015 Received in revised form 16 November 2015 Accepted 17 November 2015 Available online xxxx
a b s t r a c t The influence of pattern design and tensile strain on light emission was investigated in Ge layers and suspended membranes. The optical properties were examined by micro-photoluminescence and reflectivity. Tensile strain was extracted from micro-Raman spectroscopy. It has been shown that Fabry–Pérot interference fringes can dominate the photoluminescence spectra. It is crucial to remove them in order to analyze the photoluminescence changes coming from tensile strain; especially if Fabry–Pérot oscillations are in the same energy range compared to the stress-induced spectral shift. This study highlights the fact that this interference must be taken into account in order to examine the strain in suspended Ge layers. © 2015 Elsevier B.V. All rights reserved.
1. Introduction Integrating laser sources which are compatible with microelectronics is a major challenge in silicon photonics. To overcome the indirect nature of the Ge bandgap, n-type doping, tensile strain or Sn alloying has been proposed to fabricate low-threshold Ge-on-Si lasers [1–3]. For the strain approach, thick Ge layers grown on Si (001) substrates are naturally tensile-strained due to the difference in thermal expansion coefficients between Ge and Si that comes into play during cooling to room temperature. Membrane patterning has thus been used to focus and locally amplify the tensile strain in Ge [4,5]. Tensile strain in Ge devices may also be induced either externally [6] or by using SiN layers stressors [7,8]. Ge photoluminescence (PL) is enhanced when the crystal is tensilestrained [4,5,9,10]. However, the PL enhancement factor due to applied tensile strain is difficult to accurately quantify [4] for Ge membranes, particularly if the finite thickness of the Ge layer gives rise to Fabry– Pérot interferences (FPI) in the same energy range as the red shift due to the tensile strain. In order to reduce interferences, very thick (28 μm) [6] or ultra-thin Ge layers (24 nm) [11] can be used. They are however not fit for photonics applications which typically require a Ge thickness of around 0.35 μm, corresponding to the condition for monomode light propagation at a 2 μm wavelength. Anti-reflection (AR) coatings have been proposed to solve this problem, but without quantification of the PL enhancement [12].
⁎ Corresponding author. E-mail address: [email protected] (A. Gassenq).
We study here the impact of tensile strain on Ge membrane light emission by taking into account the influence of FPI and we quantify the effect of an AR coating. Firstly, the emission spectrum and the reflectivity of a suspended Ge layer are simulated numerically. Then, the influence of FPI on PL spectra is investigated experimentally, at first in Ge layers grown on bulk Si or Silicon-On-Insulator (SOI) substrates, without strain variations. Results from strained Ge membranes will be then detailed for uniaxial (~1.2%) and biaxial (~0.5%) tensile strains. Finally, the impact of an AR coating is quantified. 2. Modeling In order to evaluate the impact of the FPI on the PL spectrum of a Ge membrane, the emission and the reflection spectra of a suspended Ge layer have been calculated by finite difference time domain calculations using Rsoft software. Fig. 1(a) shows the simulated system: a 1 μm-thick Ge layer (nGe = 4.1 [13]) is embedded in air (nair = 1) and the detection cell is placed above, at a distance hd N 1 μm. A broad band light source is positioned at a distance hs. The normal reflectivity of the suspended layer has been calculated using hs N hd. The emission properties have been calculated by integrating the detected spectra for 0 b hs b 1 μm to obtain what we called the calculated flat PL spectrum. This is an image of PL if the generated carriers are homogeneously distributed in a layer with an isoenergetic emission. The impact of FPI on PL emission can then be evaluated using this theoretical approach. Fig. 1(b) shows the resulting flat PL and reflectivity spectra. Since the refractive index contrast between air and Ge is high, a very high fringe contrast is found. The flat PL maximum is those localized at the reflectivity minimum (e.g. close to 0.76 eV).
http://dx.doi.org/10.1016/j.tsf.2015.11.039 0040-6090/© 2015 Elsevier B.V. All rights reserved.
Please cite this article as: A. Gassenq, et al., Study of the light emission in Ge layers and strained membranes on Si substrates, Thin Solid Films (2015), http://dx.doi.org/10.1016/j.tsf.2015.11.039
2
A. Gassenq et al. / Thin Solid Films xxx (2015) xxx–xxx
Fig. 1. Reflection and emission modeling: (a) model design and (b) corresponding calculated reflectivity (solid line) and calculated flat normalized PL spectrum (dots).
3. Photoluminescence of Ge layers PL measurements were performed on Ge layers grown on bulk Si or thin SOI substrates. Since the refractive index contrast is lower for Ge on Si than for Ge on SOI, we expect the impact of FPI on PL to be smaller for Ge grown on Si than for Ge on SOI. Fig. 2(a–c) shows the PL spectra measured on in-situ phosphorous doped (1019 cm−3) Ge layers grown on bulk Si or on thinned-down optical SOI substrates. The Ge layer grown on bulk Si(001) is 0.7 ± 0.2 μm-thick and the Ge layer grown on SOI is 1.4 ± 0.2 μm-thick (starting Si thickness: close to 60 nm; buried SiO2 thickness: 2 μm). Both Ge layers were grown at 400 °C, 13 kPa using GeH4 and PH3 gaseous precursors in an Epi Centura Reduced PressureChemical Vapor Deposition machine. The tensile strain in bulk Ge layers at the end of the fabrication process is around 0.2% resulting from differences in thermal expansion coefficients between Ge and Si. Room temperature (RT) PL measurements were performed using a Horiba iHR 320 spectrometer with a 5 mW green pump laser focused to a 5 μmdiameter spot. The emitted PL signal was collected with an optical fiber, sent through the spectrometer and detected with an extended InGaAs photo-detector. In addition, macroscopic reflectivity measurements (~4 mm2 measured surface) were carried out in a Bruker IFS55 Fourier transform infrared spectrometer with a Tungsten input lamp and an InSb cooled detector. Several measurements were performed at different locations of the 200 mm wafers: the Ge layer thickness decreases slightly when moving
from center to the edges (±0.2 μm). The influence of Ge thickness on PL can easily be seen by changing the pump spot location. For Ge layers grown on bulk Si (Fig. 2(a)), Ge thickness variations have little effect on the spectrum shape. The Γ and L valley emissions are close to 1.65 μm and 1.8 μm, respectively, in good agreement with theoretical data [3]. However, thickness variations of Ge layers grown on SOI have a major impact on PL spectra because of the presence of Fabry–Pérot oscillations (coming from the high index contrast between Ge and SiO2) (Fig. 2(b)). The actual shape of the Ge emission can be extracted by fitting the maximum PL signal over all the measured thicknesses (Fig. 2(c)), which results in a thickness-independent PL spectrum with the Γ and L valley emissions at the correct energies. We have seen in the previous section that FPI shapes the emission and the reflectivity of the Ge layer. In order to study this effect, Fig. 2(d) and (e) show normalized reflectivity spectra and the associated PL spectra measured for different thicknesses of the Ge layer grown on SOI. As expected, a good agreement is found between the energies of PL maxima and reflection minima, which confirm experimentally that the PL spectral shape is strongly dominated by FPI, as can also be seen in the layer stack reflectivity. Note that this effect is highly dependent on the measurement method. Indeed, the detected oscillation contrast will depend on the numerical aperture of the measurement setup. 4. Strain induction by membrane processing Ge membranes were processed in the 0.7 ± 0.2 μm thick Ge layer grown on Si presented above to locally enhance the strain. Membrane patterning was performed by ultra-violet lithography followed by dry etching in an inductively coupled plasma reactor. Ge membrane underetching was conducted in a tetramethyl ammonium hydroxide 25% solution at 60 °C. Fig. 3(a) and (b) present the membrane designs for uniaxial and biaxial stress induction, respectively. The angle “a” is fixed at 26°. Tensile stress was tuned by changing the parameters x, d and the under-etching which is measured by Scanning Electron Microscopy (SEM) imaging using a Zeiss Ultra 55 apparatus at 15 kV operating voltage. Finite Element Method simulations using COMSOL Multiphysics were performed using a 2D linear elastic model assuming no stress along the direction perpendicular to the membrane plane. We clamped the pattern away from the region of structuration and applied the initial stress on the free standing membrane. Fig. 3(c) and (d) shows the results of simulations for both designs. In the case of the micro-cross design, the
Fig. 2. Room temperature photoluminescence spectra measured in (a) Ge grown on Si, and (b,c) Ge on SOI; (d) normalized reflectivity and (e) corresponding photoluminescence measured in Ge on SOI (di: Ge layer thicknesses).
Please cite this article as: A. Gassenq, et al., Study of the light emission in Ge layers and strained membranes on Si substrates, Thin Solid Films (2015), http://dx.doi.org/10.1016/j.tsf.2015.11.039
A. Gassenq et al. / Thin Solid Films xxx (2015) xxx–xxx
3
Fig. 3. Design and tilted SEM imaging for (a) uniaxial stress induced by micro-bridge patterning and (b) biaxial stress induced by micro-cross patterning; Finite Element Method simulations of strain in (c) uniaxially and (d) biaxially stressed Ge membranes.
bi-axial strain εbi is defined by (ε100 + ε010)/2 while the uniaxial strain εuni is along the [100] direction in the micro-bridge design. The corresponding (x,d) parameters are provided in the figures. 5. Strain and photoluminescence measurement in Ge membrane Raman spectroscopy, PL and reflectivity measurements were combined to investigate the effects of pattern design and strain on light emission. Strain was measured using a Renishaw InVia micro-Raman spectrometer with a 785 nm wavelength incident laser and a 1 μmdiameter spot. The laser intensity was low enough to avoid heating effects [14]. The Raman shift was measured by fitting the spectra with Lorentzian functions. A bulk Ge substrate was systematically used as a reference for 0% deformation. The Raman spectra measured on the Ge membranes are shown in Fig. 4. The extracted strains are indicated in the scale assuming linear relationships between strain and Raman shift, with a 424 cm−1 coefficient for bi-axial stress (micro-crosses) and a 154 cm−1 for uniaxial stress (micro-bridges) [4,15,16]. For the uniaxial design (Fig. 4(a)), two micro-bridges (d = 5 μm) are compared.
Fig. 4. Raman spectra of Ge membranes under various states of stress: (a) uniaxial stress, (b) biaxial stress, and (c) biaxial stress with AR coating.
The first (circle markers, x = 84 μm, thickness = 0.85 μm) and the second (square markers, x = 45 μm, thickness = 0.75 μm) were obtained using 3 h and 4 h 30 min under-etching times, yielding 1.1% and 1.4% strains along [100], respectively. The measured Raman spectral shifts and the corresponding strain values are provided in the figure legend. In the case of biaxial strain (Fig. 4(b) and (c)), the thickness of the Ge layer was 0.75 μm, d = 5 μm and the under-etching time was 3 h. The strain was tuned between 0.3% and 0.5% using the parameter x, as indicated in the legend. No AR layer was added to the membranes in Fig. 4(b) while the membranes in Fig. 4(c) were coated with 125 nm of SiO2 and 33 nm of amorphous Si deposited by e-gun evaporation on top of the sample at the end of the processing. Raman spectroscopy measurements were performed after the AR layer deposition. No effect of AR coating on the strain was observed. Fig. 5 shows the RT photoluminescence of the Ge membranes previously measured by Raman spectroscopy (Fig. 4). Macroscopic reflectivity measurement results performed over the chip containing the membranes of interest have been also added to the figure. Note that the reflectivity contrast is underestimated for macroscopic reflectivity measurements (with mainly a 2D Ge layer on Si) compared to the actual membrane reflectivity (suspended Ge). The fringe positions are however not strongly affected, as the Ge thickness is similar over the measured surface. Therefore, it is possible to make the same type of PL/reflectivity correlation as discussed in Sections 2 and 3. Fig. 5(a) and (b) show the data measured in the Ge micro-bridges: the PL spectra were measured from outside the bridge (location D in Fig. 3(a)) to the central part of the bridge where the strain is maximum (location A in Fig. 3(a)). The theoretical Ge emission energy corresponding to the measured strain is displayed as a vertical dashed line [17,18]. A red shift of the PL maximum is observed when the strain increases from the slightly strained part (curve D) to the highly strained part (curve A) of the membranes. However, the PL enhancement is completely different between both micro-bridges. This results from interferences which occur differently as a function of Ge layer thicknesses. Indeed, the emission maximum is not localized at the Γ bandgap but at the minimum of the reflectivity (as in the modeling section). We show here that the emission intensity in the center of the membrane is more affected by layer thickness than by induced strain. Fig. 5(c) and (d) present the micro PL measurements in the Ge micro-cross previously measured by Raman spectroscopy (Fig. 4(b) and (c)), with and without the AR coating, respectively. In the
Please cite this article as: A. Gassenq, et al., Study of the light emission in Ge layers and strained membranes on Si substrates, Thin Solid Films (2015), http://dx.doi.org/10.1016/j.tsf.2015.11.039
4
A. Gassenq et al. / Thin Solid Films xxx (2015) xxx–xxx
Fig. 5. Room temperature reflectivity and photoluminescence in Ge membranes for (a) 1.1% uniaxial strain (membrane thickness = 0.85 μm), (b) 1.4% uniaxial strain (membrane thickness = 0.75 μm), and 0.3 to 0.5% biaxial strain (c) without and (d) with anti-reflection coating. The vertical lines show the Gamma bandgap energies calculated for the measured strains at the center of the membranes.
absence of the AR coating and for low strain (0.3%) the emission is limited by FPI (high reflectivity). For higher strains, the PL intensity increases due to a combination of the red shift induced by the biaxial tensile strain and constructive interference due to lower reflectivity at longer wavelengths. Once again, the maximum of the PL emission is not localized at the bandgap. When the membranes are coated with AR layers (Fig. 5(d)), the PL spectra are then centered on the Γ bandgap. For low strained (0.3%), the PL emission is markedly larger due to interference minimization by AR layers. For higher strain (0.5%), PL enhancement is lower than in the sample without AR layers. Therefore, PL enhancement in strained Ge membranes varies dramatically if no antireflective coating is used which can lead to major interpretation errors if FPI occur in the same energy range as the red shift induced by the strain. 6. Conclusion In this article, the influence of tensile strain on Ge light emission was investigated. Strain was measured using Raman spectroscopy while the light emission properties were probed using photoluminescence. We have shown that Fabry–Pérot interferences can strongly shape the light emission. Without anti-reflection layers or adapted dimensions, straininduced PL enhancement cannot be easily measured in suspended Ge membranes. Acknowledgments The authors would like to thank the Platforme de Technologie Amont in Grenoble for the clean room facilities. This work was supported by the CEA DSM-DRT Phare projects “Photonics” and “Operando”, as well as the CEA Enhanced Eurotalent project “Straintronics”. References [1] B. Dutt, D.S. Sukhdeo, D. Nam, B.M. Vulovic, Z. Yuan, K.C. Saraswat, Roadmap to an efficient germanium-on-silicon laser: strain vs. n-type doping, IEEE Photonics J. 4 (2012) 2002–2009, http://dx.doi.org/10.1109/JPHOT.2012.2221692. [2] S. Wirths, R. Geiger, N. von den Driesch, G. Mussler, T. Stoica, S. Mantl, Z. Ikonic, M. Luysberg, S. Chiussi, J.M. Hartmann, H. Sigg, J. Faist, D. Buca, D. Grützmacher, Lasing in direct-bandgap GeSn alloy grown on Si, Nat. Photonics 9 (2015) 88–92, http://dx. doi.org/10.1038/nphoton.2014.321. [3] M. El Kurdi, G. Fishman, S. Sauvage, P. Boucaud, Band structure and optical gain of tensile-strained germanium based on a 30 band kp formalism, J. Appl. Phys. 107 (2010)http://dx.doi.org/10.1063/1.3279307.
[4] M.J. Süess, R. Geiger, R.a. Minamisawa, G. Schiefler, J. Frigerio, D. Chrastina, G. Isella, R. Spolenak, J. Faist, H. Sigg, Analysis of enhanced light emission from highly strained germanium microbridges, Nat. Photonics 7 (2013) 466–472, http://dx.doi.org/10. 1038/NPHOTON.2013.67. [5] D.S. Sukhdeo, D. Nam, J.-H. Kang, M.L. Brongersma, K.C. Saraswat, Direct bandgap germanium-on-silicon inferred from 5.7% b100N uniaxial tensile strain, Photonics Res 2 (A8) (2014)http://dx.doi.org/10.1364/PRJ.2.0000A8. [6] M. El Kurdi, H. Bertin, E. Martincic, M. de Kersauson, G. Fishman, S. Sauvage, A. Bosseboeuf, P. Boucaud, Control of direct band gap emission of bulk germanium by mechanical tensile strain, Appl. Phys. Lett. 96 (2010) 041909, http://dx.doi.org/ 10.1063/1.3297883. [7] K. Guilloy, N. Pauc, A. Gassenq, P. Gentile, S. Tardif, F. Rieutord, V. Calvo, Tensile strained germanium nanowires measured by photocurrent spectroscopy and Xray microdiffraction, Nano Lett. 15 (2015) 2429–2433, http://dx.doi.org/10.1021/ nl5048219. [8] G. Capellini, G. Kozlowski, Y. Yamamoto, M. Lisker, C. Wenger, G. Niu, P. Zaumseil, B. Tillack, A. Ghrib, M. de Kersauson, M. El Kurdi, P. Boucaud, T. Schroeder, Strain analysis in SiN/Ge microstructures obtained via Si-complementary metal oxide semiconductor compatible approach, J. Appl. Phys. 113 (2013)http://dx.doi.org/10. 1063/1.4772781. [9] Y. Huo, H. Lin, R. Chen, M. Makarova, Y. Rong, M. Li, T. Kamins, J. Vuckovic, J. Harris, Strong enhancement of direct transition photoluminescence with highly tensilestrained Ge grown by molecular beam epitaxy, Appl. Phys. Lett. 98 (2011)http:// dx.doi.org/10.1063/1.3534785. [10] D.S. Sukhdeo, D. Nam, J.-H. Kang, M.L. Brongersma, K.C. Saraswat, Bandgapcustomizable germanium using lithographically determined biaxial tensile strain for silicon-compatible optoelectronics, Opt. Express 23 (2015) 16740, http://dx. doi.org/10.1364/OE.23.016740. [11] J.R. Sanchez-Perez, C. Boztug, F. Chen, F.F. Sudradjat, D.M. Paskiewicz, R.B. Jacobson, M. Lagallya, R. Paiella, Direct-bandgap light-emitting germanium in tensilely strained nanomembranes, Proc. Natl. Acad. Sci. 108 (2011) 18893–18898, http:// dx.doi.org/10.1073/pnas.1107968108. [12] D. Nam, D. Sukhdeo, A. Roy, K. Balram, S.-L. Cheng, K.C.-Y. Huang, Z. Yuan, M. Brongersma, Y. Nishi, D. Miller, K. Saraswat, Strained germanium thin film membrane on silicon substrate for optoelectronics, Opt. Express 19 (2011) 25866, http://dx.doi.org/10.1364/OE.19.025866. [13] H.H. Li, Refractive index of silicon and germanium and its wavelength and temperature derivatives, J. Phys. Chem. Ref. Data 9 (1980) 561–658, http://dx.doi.org/10. 1063/1.555624. [14] M. Süess, R. Minamisawa, R. Geiger, K. Bourdelle, H. Sigg, R. Spolenak, Powerdependent Raman analysis of highly strained Si nanobridges, Nano Lett. 14 (2014) 1249–1254, http://dx.doi.org/10.1021/nl404152r. [15] I. De Wolf, Micro-Raman spectroscopy to study local mechanical stress in silicon integrated circuits, Semicond. Sci. Technol. 11 (1996) 139, http://dx.doi.org/10.1088/ 0268-1242/11/2/001. [16] F. Cerdeira, C.J. Buchenauer, F.H. Pollak, M. Cardona, Stress-induced Raman shifts of diamond and zincblende semicond, Phys. Rev. B 5 (1972) 580, http://dx.doi.org/10. 1103/PhysRevB.5.580. [17] C.G. de Walle, Band lineups and deformation potentials in the model-solid theory, Phys. Rev. B 39 (1989) 1871–1883, http://dx.doi.org/10.1103/PhysRevB.39.1871. [18] C. Boztug, J.R. Sánchez-Pérez, F.F. Sudradjat, R. Jacobson, D.M. Paskiewicz, M.G. Lagally, R. Paiella, Tensilely strained germanium nanomembranes as infrared optical gain media, Small 9 (2013) 622–630, http://dx.doi.org/10.1002/smll.201201090.
Please cite this article as: A. Gassenq, et al., Study of the light emission in Ge layers and strained membranes on Si substrates, Thin Solid Films (2015), http://dx.doi.org/10.1016/j.tsf.2015.11.039