Photoluminescence and Raman Spectroscopy on truncated Nano Pyramids Physics of low Dimensions, FFF042 Josefin Voigt & Stefano Scaramuzza 10.12.2013, Lund University 1
Introduction In this project truncated GaN nano pyramids containing one or more InGaN quantum wells are examined by using photoluminescence (PL) spectroscopy and Raman spectroscopy. InGaN/GaN quantum wells are interesting because depending on the composition of the InGaN or the quantum well width they are able to emit light from the whole visible spectrum. Structures like nanopyramids containing such quantum wells are quite novel and thus it is interesting to study them. Sample An illustration and a SEM image of the examined samples is shown in Figure 1. After applying a GaN layer on a Si substrate a SiN mask is deposited on the GaN featuring holes. The GaN grown is continued on this spots which results in a formation of the truncated pyramids. The quantum wells are made by depositing a thin layer of InGaN on top of the pyramids and by covering it again with GaN. The pyramids have less strain between the InGaN and the GaN compared to pyramids grown on a single flat surface. The examined samples are presented in Table 1. Sample 249 is a reference sample containing no InGaN. Figure 1: SEM image [4] (left) and Illustration (right) of a truncated nanopyramid. Table 1: Samples and their properties. TEG and TMI stand for the flow rates in sccm of the precursors triethylgallium and trimethylindium. 2
Techniques In photoluminescence spectroscopy a sample is illuminated by a light source of a defined wavelength. The incident light gives enough energy to the electrons to lift them from the valence band into the conduction band leaving a hole in the valence band. If the electrons occupy a higher state than the ground state in the conduction band they will relax into the ground state under emission of phonons. The same happens with the holes in the valence band. Once electrons and holes are in their ground state, they recombine under emission of a photon of a characteristic energy for the considered material. This way materials and their properties can be studied. In Raman spectroscopy a sample is illuminated by a certain wavelength laser where due to the Raman effect a photon is absorbed exciting an electron to a virtual energy level. The subsequent relaxation of the electron to a different vibrational or rotational state, causes a photon of different energy to be emitted. This difference in energy between the original photon and the emitted photon causes a frequency shift, calculated as 1/λ 0 1/λ R where λ 0 is the wavelength of the laser and λ R is the wavelength of the Raman radiation. These shifts are material specific, since only certain modes are allowed for the state after recombinations for electrons, according to certain selection rules. The shifts can thus be used to decide which materials can be found in certain samples. Results Photoluminescence Spectroscopy A PL spectrum has been acquired for each sample. Figure 2 shows the PL spectrum of the reference sample 249 and 278. It has been acquired with a laser of 325 nm wavelength. According to [2] a narrow peak from the GaN around 3.48 ev and a broader peak from the InGaN quantum well around 3.1 ev is expected. According to [3] the InGaN well peak is expected to be in the range between 2.9 ev and 3.1 ev. The GaN signal should stay constant for each sample while the InGaN peak depends on different factors as e.g. the size of the quantum well and is thus expected to slightly vary between the samples. There are other peaks in the spectrum which might result from other materials (Si substrate) or impurities in the different materials used in the sample. The important difference between the presented plots in Figure 2 is the peak in sample 278 at 2.95 ev. It is assumed to belong to the InGaN well for different reasons. First it is not visible in the reference sample. Second according to [3] and [4] we should expect it in this energy region. Third it is the only peak varying its position depending on the sample. Since it is assumed that several peaks in the region 3.0 3.5 ev belong to impurities in the GaN a new laser wavelength of 375 nm is used. The impurity peaks are expected to decrease since the lasers energy (3.3 ev) is slightly below the GaN bandgap of 3.47 ev ([1]). 3
The obtained result is presented in Figure 3. Again the reference sample 249 and the sample 278 are presented. Energies over 3.2 ev are neglected since the laser light could not be filtered before it reached the detector and its energy is around 3.3 ev. Thus we would have too much contribution around these energies. The spectrum of the reference sample is similar to the one obtained with the previous wavelength. The spectrum of sample 278 although shows that the peak around 2.95 ev is better defined than with the previous laser. Figure 2: PL spectrum of sample 249 and 278 acquired with a 325 nm laser. Figure 3: PL spectra of sample 249 and 278 acquired with a 375 nm laser. 4
After acquiring the PL spectra with the new wavelength the peak corresponding to the InGaN well for each plot is determined. This information can be further used to estimate the size of the quantum well. Under assumption of different parameters one can plot the energy of the quantum well peak as a function of the well width L. Since the composition x of the In x Ga 1 x N is not known the calculations are made for x=0.2, 0.4, 0.6 and 0.8. The plot is shown in Figure 4. Figure 4: PL peak energy of the quantum well in dependence of its thickness for different compositions x. By comparing the found energies of the PL spectrum with the plot in Figure 4 we can estimate the size of the quantum well in the pyramids. The results are summarized in Table 2. Table 2: PL peak energies of the quantum wells and the resulting estimated quantum well widths. Assuming an InGaN monolayer thickness of 5 Å we get a thickness of 1 10 monolayers per well. Either high well times or flow rates seem to be required to gain a bigger quantum well, assuming an increase of material leads to more growth. It is possible to be the opposite way as well but it seems unlikely to increase the layer thickness when growing for a longer time. Another way to determine the composition of the samples is to measure the well sizes using SEM. When analysing the SEM images the wells appear to be around 2nm 4nm even though the resolution of the SEM images makes it hard to find an exact value. Comparing this well size with figure 4 we can estimate a composition around In 20 Ga 80 N. For a more exact value of the composition we would need a more exact value of the sizes of the quantum wells, which could be achieved by using TEM. 5
Results Raman Spectroscopy To correctly calculate the width of the wells from the position of the peaks in the PL spectrum a correct value of the composition of the InGaN is desirable, as mentioned before. As has been shown in [6] and [7] this can be obtained by studying the Raman spectrum. Therefore Raman spectroscopy was performed on the samples. When using a laser of 632.5 nm the Raman spectra for the samples were as shown in Figure 5. Figure 5: Raman spectra for the different samples, obtained using a 632.5 nm laser. These peaks of approximately 525 cm 1 and 570 cm 1 correspond well to the expected values found in [8] of peaks from the GaN substrate for cross sectional insidence. The first peak correspond to the A 1 (LO) vibrational mode and the second peak correspond to the E 1 (TO) and E 2 vibrational modes. It is also possible that Si from the substrate contribute to the peak at around 525 cm 1, since Si has a TO shift of 521 cm 1 for zincblende structure, which makes up the substrate. As can be seen all the spectra are approximately equivalent, even that of sample 249, this indicates that we do not get a signal from the quantum well since this spectra does not differ from the others. From [6] we expect a frequency shift in the range of 600 cm 1 to 750 cm 1 for a thin film of InGaN, which can not be seen in figure 5. The reasons for this could be several. One major cause could be that the layer of InGaN is very thin and when exposing the sample the laser is focused on a very tiny spot, thus exposing very little of the thin InGaN layer and more of the surroundings. To see if the laser was too low in energy or penetrated too deep into the sample to actually detect the quantum wells close to the surface, another setup, with a laser of 532 nm, was tried. We did not however see any peaks from quantum wells in this case either. 6
Conclusion Using PL spectroscopy and the assumption that the composition of the quantum well was between In 20 Ga 80 N and In 80 Ga 20 N the wells of the samples were estimated to be 0.20 nm 5.60 nm, with some differences for the different wells (see table 2). This corresponds to a thickness of 3 18 monolayers per well. The size of the well seems to be clearly influenced by the well time and the flow rates of the precursors during growth. Had we been able to measure the sizes of the quantum wells with a good certainty it would have been possible to determine the composition of the wells. However SEM showed to have to low resolution to be able to do this and we would have to use TEM. The sizes of the wells would also have been possible to determine if we had succeeded in determining the composition using raman spectroscopy. Unfortunately this did not work, probably due to the small amount of material constituting the quantum well not giving enough signal. Sources [1] John H. Davies, The physics of low dimensional semiconductors, Cambridge University Press, 1997. [2] Chul Woo Lee, Sung Taek Kim, Ki Soo Lim, Photoluminescence Studies of GaN and InGaN/GaN Quantum Wells, Journal of the Korean Physical Society, Vol. 35, No. 3, pp. 280 285, 1999. [3] M. S. Minsky, S. B. Fleischer, A. C. Abare, J. E. Bowers, E. L. Hu, Characterization of high quality InGaN/GaN multiquantum wells with time resolved photoluminescence, Applied Physics Letters, Vol. 72, No. 9, pp. 1066 1068, 1998. [4] Ioffe Physico Technical Institute, Electronic archive: New Semiconductor Materials. Characteristics and Properties, Accessed: 9.10.2013, http://www.ioffe.ru/sva/nsm/. [5] Zhaoxia Bi, Presentation: MQW on truncated GaN pyramids, Lund University, 2013. [6] Robert Oliva Vidal, Master thesis: Optical emission and Raman scattering in InGaN thin films grown by molecular beam epitaxy, Raman Spectroscopy Group, ICTJA, CSIC Lluis Solé i Sabarís, Barcelona [7] S. Hernández, R. Cuscó, D. Pastor, L. Artúsa, K. P. O Donnell, R. W. Martin, I. M. Watson, Y. Nanishi, E. Calleja, Raman scattering study of the InGaN alloy over the whole composition range, Journal of applied physics 98, 013511, 2005 [8] Z. C. Feng, W. Wang, S. J. Chua, P. X. Zhang, K. P. J. Williams, G. D. Pitt, Raman scattering properties of GaN thin films grown on sapphire under visible and ultraviolet excitation, J. Raman Spectrosc., 32, 840 846, 2001 7