P. Fischer et al.: Scanning Electroluminescence Microscopy for Light Emitting Diodes 119 phys. stat. sol. (a) 176, 119 (1999) Subject classification: 78.60.Fi; 78.66.Fd; S7.14; S7.15 Scanning Electroluminescence Microscopy: A Powerful Novel Characterization Tool for Light Emitting Diodes P. Fischer1) (a), J. Christen (a), M. Zacharias (a), V. Schwegler (b), C. Kirchner (b), and M. Kamp (b) (a) Institute of Experimental Physics, Otto-von-Guericke University, D-39016 Magdeburg, Germany (b) Department of Optoelectronics, University of Ulm, D-89069 Ulm, Germany (Received July 4, 1999) A novel approach of spectrally resolved scanning electroluminescence microscopy is introduced as a powerful characterization tool directly imaging the spectral emission characteristics of luminescence devices. This fast non-destructive technique allows the direct correlation of morphological and optical properties of final devices at the microscopically identical sample position. It directly visualizes the inhomogeneities in epitaxial growth as well as in processing. Furthermore, it shows where and how much light and with which spectral characteristic is emitted by the device which is most important for device design. The characterization of an InGaN/GaN Light Emitting Diode is given as an example of the power of scanning electroluminescence microscopy. The luminescence intensity maps and the emission peak wavelength images are taken under operation of the device. They directly image the optical quality of the device and yield direct images of the In fluctuations with a spatial resolution of 1 mm. The indium concentration is found to fluctuate from 5% to 8%. The InGaN layer shows a strong spatially localized emission characteristic. 1. Introduction Nowadays the growth of blue and UV Light Emitting Diodes (LED) has gained great attraction because of their large application potential. Intense research on group III-nitride semiconductors successfully led to commercial devices [1 to 7]. To improve the spatial and spectral emission characteristic of LEDs it is helpful to get a direct correlation between morphological and optical properties of the device at the same microscopical position. To demonstrate the great potential of this newly established characterization tool, an InGaN/GaN near UV LED is characterized by highly spatially and spectrally resolved electroluminescence microscopy (m-EL). 2. Experimental The scanning m-EL set-up consists of a modified optical microscope using long distance lenses transparent to UV [8]. The samples are scanned either by an piezo stage or a dc-motor driven scanning stage with a stepping resolution of 1 or 250 nm, respectively. The overall spatial resolution in detection mode is 1 mm. The luminescence is dispersed 1 ) Corresponding author; Phone: ++49-391-6712361; Fax: ++49-391-6711130; e-mail: peter.fischer@physik.uni-magdeburg.de 120 P. Fischer et al. in a 0.5 m spectrometer and detected by a liquid nitrogen cooled CCD camera. The spectral resolution for the m-EL measurements reported here is 0.5 nm. All measurements are performed at room temperature. The whole set-up is computer controlled. At each pixel of the scanned area a complete-EL spectrum is recorded and stored during scanning over typically 128  100 pixels. After a scanning image is completed a three-dimensional data set IEL(x, y, l) is obtained and all types of data cross sections through this tensor can be subsequently generated. Typical examples of such extracted information are local EL spot spectra IEL(l, xi, yi), EL wavelength images, i.e. mappings of the local emission peak wavelength lPeak(x, y), EL spectrum linescans IEL(l, s(x, y)) as well as sets of EL intensity images for different spectral regions IEL(li, x, y). In the EL intensity images the EL intensity is integrated over the chosen wavelength window at each pixel, color-coded plotted and scaled to min/max. The InGaN/GaN LED reported here is grown by low pressure MOVPE in a horizontal reactor on a SiC substrate. The structure consists of 500 nm GaN and 900 nm Si-doped GaN followed by an active region of 50 nm InGaN, 100 nm Mg-doped AlGaN with a nominally Al content of about 4%, and finally 450 nm p-doped GaN (see Fig. 1c). The hole concentration of the Mg-doped GaN is p  2  1017 cm ±± 3. The InGaN layer has an In mole fraction of approximately 6% and is nominally undoped. The doping concentration of the n-type GaN is about 1  1018 cm ±±3. Photolithography is used to define the mesa structure. Chemically-assisted ion-beam etching (CAIBE) transferred the pattern using a conventional photoresist mask. A second lithographic step defines the n- and p-contact area, using lift-off technique and subsequent metallization with Ni/Au contacts. The LED is mounted with the SiC substrate on a chip carrier to investigate the device from the p-contact side. 3. Results and Discussion The area under investigation is depicted in the optical microscope image Fig. 1a. The non-transparent p-contact on top of the mesa is located in the center of the device. The mesa is marked by a white circle. Fig. 1b shows the integral EL spectra for three different injection current densities spatially integrated over all single m-EL spectra. For low current density the spectrum is dominated by the EL of the p-GaN layer at 440 nm, which corresponds to a DAP recombination for this level of magnesium doping [9]. An insufficient carrier confinement in the active region caused by a low indium content in the active region and insufficient aluminium content in the electron barrier is responsible for the observed emission. With increasing injection current the emission around 400 nm gains intensity and takes over for high injection currents. This luminescence corresponds to an InGaN layer with an indium mole fraction of x = 0.07, which is in perfect agreement with X-ray diffraction measurements. The blue shift observed for the transition from low to medium injection currents is caused by band filling. The red shift of the spectrum for high current density can be explained by thermal effects [10]. The m-EL intensity mappings are depicted in the second row of Fig. 1 in dependence on the current density (64 to 369 A/cm2). It is clearly visual that the intensity map becomes more uniform with increasing current density (see Fig. 1d to f). We conclude that this is due to the saturation of spatially distributed non-radiative recombination centers. The peak±wavelength scanning images are plotted in the third row of Fig. 1. The emission around 440 nm is homogeneously distributed. With higher injection currents the emis- Scanning Electroluminescence Microscopy for Light Emitting Diodes 121 Fig. 1. a) Microscope image, b) integrated EL spectra, c) sample structure and d) to f) EL intensity and g) to i) wavelength images for injected current densities of 64, 168 and 369 A/cm2, respectively sion of the InGaN layer at around 400 nm emerges and shows its localized statistical distribution (see Fig. 1g to i). To investigate the quality of the p-GaN and InGaN layers a separate inspection of both recombination channels involved is required (see Fig. 2b). The intensity and wavelength images within the corresponding wavelength intervals are depicted in Fig. 2c to f. The homogeneity of the EL intensity of the InGaN (Fig. 2c) as well as of the p-GaN (Fig. 2d) is visible in the intensity maps. Inhomogeneities of the EL intensity are attributed to either inhomogeneous injection conditions or of local non-radiative centers in the active region of the device. The small indium mole fraction fluctuation can be seen from the wavelength image Fig. 2e. The emission wavelength fluctuation from 386 to 406 nm corresponds to a fairly low indium fluctuation between x = 0.05 and 0.08, assuming band-band recombination and a bowing parameter of 3.2 eV. The homogeneously distributed emission wavelength of the p-GaN layer can be seen in Fig. 2f. 4. Conclusion In conclusion, a novel characterization tool for light emitting devices is reported. The nondestructive method scanning EL microscopy probes the relevant properties of the 122 P. Fischer et al. Fig. 2. a) Microscope image and b) integral EL spectrum of the LED. EL intensity and emission wavelength mapping for the InGaN and p-GaN layer are separately shown in c), e) and d), f), respectively optically active region of the device under operation. It reveals all deficiencies from epitaxial growth, processing, and spatial design all together. The information obtained from scanning m-EL has high relevance since it has its origin in the optically active region of the device under operation. This is demonstrated for an In0.06Ga0.94N/GaN LED. Due to an imperfect electrical confinement EL of the p-GaN layer is detected for low injection conditions. However, for high injection conditions the device operates as near UV LED at around 400 nm. It clearly turns out that the inhomogeneity of the InGaN layer quality rather than the fairly good low indium fluctuations of 5% to 8% is limiting the quality of the device. Scanning Electroluminescence Microscopy for Light Emitting Diodes 123 Acknowledgements Financial support by the Kultusministerium of Sachsen-Anhalt under contract 1432A/8386B is gratefully acknowledged. The group at Ulm Univ. acknowledges the financial support by the German Ministry of Science and Education (BMBF) under contract 01 BS 802. They are obliged to K.J. Ebeling for fruitful discussions and continuous encouragement. References [1] J. Han, M.H. Crawford, R.J. Shul, S.J. Hearne, E. Chason, J.J. Figiel, and M. Banas, MRS Internet J. Nitride Semicond. Res. 4S1, G7.7 (1999). [2] T. Mukai, D. Morita, and S. Nakamura, J. Cryst. Growth 189/190, 778 (1998). [3] S. Nakamura, M. Senoh, N. Iwasa, S. Nagahama, T. Yamada, T. Matsushita, Y. Sugimoto, and H. Kiyoku, Jpn. J. Appl. Phys. 34, L797 (1995). [4] S. Nakamura, J. Vac. Sci. Technol. A 13, 705 (1995). [5] S. Nakamura, T. Mukai, and M. Senoh, Appl. Phys. Lett. 64, 13 (1994). [6] S. Nakamura, T. Mukai, and M. Senoh, J. Appl. Phys. 76, 12 (1994). [7] M. Kamp, C. Kirchner, M. Mayer, A. Pelzmann, M. Schauler, F. Eberhard, P. Unger, and K.J. Ebeling, Electrochem. Soc. Proc. 97/34, 272 (1997). [8] P. Fischer, J. Christen, M. Zacharias, H. Nakashima, and K. Hiramatsu, Solid State Phenom. 63, 151 (1998). [9] S. Nakamura, T. Mukai, and M. Senoh, Jpn. J. Appl. Phys. 30, L1998 (1991). [10] A. Link, K. Bitzer, W. Limmer, R. Sauer, C. Kirchner, V. Schwegler, M. Kamp, D. Ebling, and K.W. Benz, J. Appl. Phys., accepted.