TY - JOUR
AU1 - Nakaji, D.
AU2 - Grillo, V.
AU3 - Yamamoto, N.
AU4 - Mukai, T.
AB - Abstract Monochromatic cathodoluminescence CL images of the threading dislocations in Si-doped n-GaN were observed by the TEM–CL technique. We studied dependence of the contrast and the FWHM of the dislocation image on sample thickness, accelerating voltage and temperature. The CL spectra were measured at various temperatures and were analyzed to find the property of the band edge (BE) emission used for the CL imaging. The observation showed that the FWHM of the dislocation contrast monotonically increases with sample thickness and decreases with accelerating voltage, which qualitatively agrees with the behavior expected from the theory. On the other hand, the temperature dependence of the FWHM shows an anomalous behavior. This dependence can be explained by the mobility and lifetime of holes as a function of temperature. The relation between the FWHM of the dislocation contrast and the diffusion length is also discussed. cathodoluminescence, transmision electron microscopy, GaN, dislocation Introduction Optical properties of defects in semiconductors have been studied by a transmission electron microscope (TEM) combined with a cathodoluminescence (CL) detection system (TEM–CL) [1–6]. The TEM–CL technique is a unique method, because structural information of the defects can be simultaneously obtained from TEM observation together with their optical measurement. The CL technique combined with a scanning electron microscope (SEM–CL) is widely used because it allows simple for and easy handling of samples; however, the TEM–CL technique has an advantage in high spatial resolution in comparison with the SEM–CL technique, because the carrier generation volume is much smaller than that in the SEM–CL due to the use of thin film samples. The optical properties of dislocations in III–V and II–VI compound semiconductors have been intensively studied by these CL techniques. The dislocations generally show dark contrast in the CL images, since they act as non-radiative recombination centers. To evaluate the semiconductor parameters such as diffusion length [7,8], analysis of the dislocation contrast in the SEM–CL has been performed based on several theoretical approaches. However, very few works have been carried out for evaluating dislocation contrast in the TEM–CL. Recently we developed a simulation program for calculating the CL contrast of a dislocation using the Monte Carlo method [9] and applied it to the threading dislocations in GaN [10]. The dislocation contrast in GaN is narrower in width compared with that in GaAs and other compound semiconductors, because the diffusion length of carriers in GaN is short. Therefore, the dislocations in GaN are proper object for evaluating spatial resolution of the TEM–CL technique. The dislocations in the GaN hexagonal lattice have three different Burgers vectors, a, c, and a+c. The threading dislocations running parallel to the c axis are classified into edge, screw and mixed types. More than 90% of the threading dislocations are reported to be edge type [11]. Hino et al. [11] showed, by using etching technique and photoluminescence measurement, that the three types of threading dislocations affect the luminescence in different ways. They suggested that the screw dislocations act as strong non-radiative centers, while the edge dislocations are less active compared with the other two types. However, our previous CL observation showed that the edge dislocation reveals considerably strong contrast when compared with the mixed one [10]. So all the threading dislocations are visible in the CL image and it is difficult to identify the types only from their CL contrasts. The spatial resolution of the TEM–CL depends on several parameters such as sample thickness, temperature and accelerating voltage of TEM. In the present study, we observed monochromatic CL images of the threading dislocations in Si-doped n-GaN by the TEM–CL after changing those parameters. The CL spectra were taken at various temperatures and were analyzed to find the property of the band edge (BE) emission used for the CL imaging. We found that the blue emission is attributed to the recombination of the free exciton (FX) and bound exciton (D0,X) whose energies are close to each other. Dependences of the CL contrast has tendency that a full width of half maximum (FWHM) of the dislocation contrast monotonically increases with sample thickness and decreases with accelerating voltage, which is similar to the behavior expected from the theory. However, the temperature dependence of the FWHM shows no monotonical change. This dependence can be understood mainly from the behavior of the hole mobility, as will be discussed later. The relation between the FWHM of the dislocation contrast and the diffusion length will also be discussed. General theory Here, we briefly discuss the general theory of the TEM–CL technique. Cathodoluminescence in semiconductors is due to radiative recombination of excess minority carriers or excitons. There are two stages before the recombination: the first is the carrier generation process, and the second is the carrier diffusion process. In the first stage an incident electron beam generates electron hole pairs in a semiconductor material indirectly through the generation of secondary electrons and plasmons. The incident electrons are spread inside a thin film sample mainly by elastic scattering and a beam spread δb(μm) at the bottom of the sample is given by   \[{\delta}_{\mathrm{b}}\ =\ 6.25\left(\frac{Z}{E}\right)\left(\frac{{\rho}}{A}\right)^{{\frac{1}{2}}\ t\frac{3}{2}},\] (1) where Z is the atomic number, E (keV) is an energy of the incident electron, ρ (g cm−3) is the density of the material, A (g mol−1) is the atomic weight and t is sample thickness. The excess carrier generation volume has a cone shape with a base diameter of δb. However, the excess carriers are mainly concentrated in the central region, so the net diameter of the carrier distribution should be smaller than δb. In the second stage, the excess carriers take a random work from the source point by diffusion. The minority carriers are important for the luminescence process, because the majority carriers exist everywhere to recombine with the minority carriers. The excess minority carrier concentration p(r, t) obeys a diffusion equation of the form   \[\frac{{\partial}p}{{\partial}t}\ =\ D{\nabla}^{2}p\ {-}\ \frac{p}{{\tau}}\ +\ g\left(r\right),\] (2) where D is the diffusion coefficient of the minority carrier, and τ is the minority carrier lifetime. The second term on the right side in eq. (2) indicates the decrease due to the recombination processes. In general, the observable lifetime τ is expressed as   \[\frac{1}{{\tau}}\ =\ \frac{1}{{\tau}_{\mathrm{r}}}\ +\ \frac{1}{{\tau}_{\mathrm{nr}}},\] (3) where τr and τnr are the radiative and non-radiative recombination lifetimes, respectively. The carrier generation function g(r) represents the spatial distribution of the excess carriers generated by the electron scattering process in a sample. The steady-state carrier distribution can be given by solving eq. (2) with ∂p/∂t = 0 under proper boundary conditions. In a thin sample case the surface recombination occurs at the top and bottom surfaces (z=0 and z=t, respectively) and then the boundary condition is written as,   \[D\frac{{\partial}p\left(\mathbf{r}\right)}{{\partial}z}{\vert}_{z=0,t}\ =\ \mathrm{v}_{\mathrm{s}}p\left(r\right){\vert}_{z=0,t{^\prime}}\] (4) where νs is the surface recombination rate. Under the steady-state condition the CL intensity is given by the integration of the radiative recombination rate in eq. (2) over a sufficiently large volume.   \[I_{\mathrm{CL}}\ =\ {\int}\frac{p\left(\mathbf{r}\right)}{{\tau}_{\mathrm{r}}}d\mathbf{r}\ =\ G\frac{{\tau}}{{\tau}_{\mathrm{r}}},\] (5) where G is the total generation rate of the excess minority carriers, and the absorption and surface reflection of the emitted light after the radiative recombination are ignored for simplicity. Contrast of a defect in CL images are defined by   \[C\ =\ \frac{I_{\mathrm{CL}}\left(P\right)\ {-}\ I_{\mathrm{CL}}\left(D\right)}{I_{\mathrm{CL}}\left(P\right)}\ {\times}\ 100\left(\%\right),\] (6) using the CL intensities at perfect and defect regions. If the non-radiative recombination channel characterized by the lifetime τD is added to the recombination process in the defect region, the minority carrier life time in the defect region is given by τ′ where 1/τ′=1/τ+1/τD. Then the contrast of the defect can be written as   \[C\ =\ \frac{{\tau}\ {-}\ {\tau}{^\prime}}{{\tau}}\ =\ \frac{1}{1\ +\ \left({\tau}_{D}/{\tau}\right)}.\] (7) Then if τD becomes much smaller than τ, the contrast becomes close to the maximum, C=1. Experimental We used a sample of a Si-doped GaN layer with a doping concentration of 1.8×1018 cm−3 grown on a sapphire substrate. Density of the threading dislocations is nearly 108 cm−2. Plan-view TEM samples were made by Ar ion milling to remove the sapphire substrate and the bottom part of the GaN epilayer. The samples were examined under a TEM (JEM2000FX) combined with a CL detection system. An accelerating voltage is 80 kV for the CL measurement except for the experiment measuring the accelerating voltage dependence of the dislocation contrast. The samples were cooled by a liquid He holder in the temperature range from 20 K to room temperature. The spectral resolution for the CL spectrum measurement is 0.4 nm in wavelength and that for the CL imaging is 2 nm. Results Figure 1 shows CL spectra of a Si-doped GaN sample taken at various temperatures from 20 K to room temperature. Four peaks, marked P1–P3, are seen in the CL spectrum at 20 K. A large peak at a wavelength of 357.0 nm (3.473 eV) is identified as a mixture of the emissions associated with the (D0, X) bound exiton transition and the free exiton transition, because the FWHM of the peak is large in a highly doped n-type sample and the energy difference between these two peaks is 8 meV. The P1 peak shifts to longer wavelengths with increasing temperature. The temperature dependence of the peak energy is shown in Fig. 2a. This behavior of the P1 peak is well explained by Varshni's law for the band gap energy, as indicated by a broken line in Fig. 2a, using the parameters in the literature [12]. Peak P2 is attributed to the donor to acceptor pair transition, (D0, A0), because the peak energy does not follow Varshni's law, which is characteristic of the donor acceptor transition. Peak P3 is a LO phonon replica of P2 and the energy separation between them is always the same at any temperature, which corresponds to the LO phonon energy of 75 meV. The line width of peak P1 changes with temperature, as shown in Fig. 2b. The line width can be analyzed with the formula [13]:   \[{\Gamma}\ =\ {\Gamma}_{0}\ +\ {\Gamma}_{1}\frac{1}{\mathrm{exp}\ \left(\frac{{\hbar}{\omega}_{1}}{kT}\right)\ {-}\ 1}\ +\ {\Gamma}_{2}\frac{1}{\mathrm{exp}\ \left(\frac{{\hbar}{\omega}_{2}}{kT}\right){-}1},\] (8) where the first term Γ0 is due to the imperfection and residual strain, the second and third terms are due to the exciton–phonon interaction and k is the Boltzman factor. The solid line in Fig. 2b is a fitting curve with the parameters of ℏω1=17.9 meV (144 cm−1) ℏω2=92 meV (743 cm−1) which correspond to the low frequency branch of the E2 mode and LO phonon mode, respectively [13]. This temperature dependence indicates that the P1 emission originates from the exciton associated transition rather than the donor to valence band transition. Figure 2c shows the Arrhenius plot of the integrated intensity of the P1 peak. The data is fitted with the formula   \[I_{\mathrm{CL}}\ =\ I_{0}\frac{1}{1\ +\ C_{1}\mathrm{exp}\left({-}E_{1}/kT\right)\ +\ C_{2}\mathrm{exp}\left({-}E_{2}/kT\right)}.\] (9) Fig. 1 View largeDownload slide CL spectra from a Si-doped GaN measured at various temperatures. Fig. 1 View largeDownload slide CL spectra from a Si-doped GaN measured at various temperatures. Fig. 2 View largeDownload slide (a) Temperature dependence of the peak energies of the BE emissions, (b) line width of the P1 emission and (c) Arrhenius plot of the integrated intensity of the P1 emission. Fig. 2 View largeDownload slide (a) Temperature dependence of the peak energies of the BE emissions, (b) line width of the P1 emission and (c) Arrhenius plot of the integrated intensity of the P1 emission. From the fitting, the activation energies are obtained as E1=4.2 meV and E2=42 meV. The former is the localization energy of exciton at the neutral donor and the latter is the donor binding energy, though the latter value is slightly larger than the reported one (35 meV) [14] which may be due to the difference in doping concentration. Figure 3a is a TEM image of a thin GaN sample and (b) is a monochromatic CL image of the same area taken at room temperature and at a peak wavelength of 366 nm. Comparison between these images indicates that dark dots in the CL image correspond to the threading dislocations. The threading dislocations showed a line contrast with finite length in the TEM image when the sample was tilted from the incident electron beam, as seen in Fig. 3c. The incident direction is nearly parallel to the [011] direction and the tilt angle is 31°. The sample has a wedge shape with an edge on the left, and the sample thickness at the threading dislocations can be measured from their length in the TEM image. Figure 4a shows thickness dependence of the CL intensity of the P1 emission, measured from Fig. 3. The CL intensity is zero below 200 nm and linearly increases with sample thickness in the range thicker than 200 nm. Then the CL intensity is expressed as a function of thickness t as   \[I_{\mathrm{CL}}\left(t\right)\ =\ I_{0}\left(t\ {-}\ t_{0}\right),\] (10) where t0 is a threshold thickness. The threshold thickness should be the same order of magnitude as a carrier diffusion length, because the surface recombination is dominant in the thin region. The width of the dark dot contrast due to the threading dislocation is also a measure of a carrier diffusion length. Figure 4b shows the FWHM of the dark dot contrast as a function of sample thickness, which is measured in Fig. 3. The FWHM of the dot monotonically increases with thickness, though the increase is very gradual. A broken line in the figure indicates the beam spread given by eq. (1). This line exceeds the observed FWHM at ∼400 nm, which means that the value given by eq. (1) overestimates its contribution to the resolution of CL image. It is also noted that the contrast of the dark dot decreases with thickness as shown in Fig. 4b. The contrast is defined by eq. (6), where ICL(D)is an intensity at a dislocation. A high contrast and small FWHM can be realized at the thin region near the threshold thickness, but the intensity is too weak to observe the CL image with a sufficient S/N ratio. It is better to observe the CL image at the thicker region. Fig. 3 View largeDownload slide (a) A TEM image of a thin GaN sample and (b) a monochromatic CL image of the same area taken at room temperature and at the peak wavelength of 366 nm. (c) A TEM image of the tilted sample from the incident electron beam direction. Fig. 3 View largeDownload slide (a) A TEM image of a thin GaN sample and (b) a monochromatic CL image of the same area taken at room temperature and at the peak wavelength of 366 nm. (c) A TEM image of the tilted sample from the incident electron beam direction. Fig. 4 View largeDownload slide (a) Thickness dependence of the CL intensity of the P1 emission and (b) the FWHM and contrast of the dislocation image as a function of sample thickness. Fig. 4 View largeDownload slide (a) Thickness dependence of the CL intensity of the P1 emission and (b) the FWHM and contrast of the dislocation image as a function of sample thickness. The dark dot contrast of the threading dislocation also depends on accelerating voltage. Figures 5a–e are CL images taken at various accelerating voltages from 40 to 100 kV. They were taken at room temperature with a wavelength of 366 nm. Figure 5f is a TEM image of the same area with a thickness of 510 nm. It is seen that with decreasing accelerating voltage the dark dot contrast of threading dislocations becomes blurred and the contrast becomes weak as plotted in Fig. 6. A broken line indicates the beam spread δb given by eq. (1), which is inversely proportional to accelerating voltage. The observed FWHM changes in a similar way as the beam spread δb. In this case the contribution due to the carrier diffusion is fixed, and the change in FWHM is only caused by the beam spread variation. Unfortunately, the CL intensity becomes weak during the electron beam illumination at an accelerating voltage >100 kV. This is because the high energy electron irradiation produces damages such as point defects which act as non-radiative recombination centers. The threshold voltage to generate point defects locates between 100 and 200 kV for the typical III–V and II–VI compound semiconductors. Thus, a suitable accelerating voltage to observe CL images with sufficient resolution and contrast and without sample damages for GaN is ∼80 kV. Fig. 5 View largeDownload slide (a)–(e)CL images taken at various accelerating voltages from 40 to 100 kV and (f) a TEM image of the corresponding area. Fig. 5 View largeDownload slide (a)–(e)CL images taken at various accelerating voltages from 40 to 100 kV and (f) a TEM image of the corresponding area. Fig. 6 View largeDownload slide The FWHM and contrast of the dislocation image as a function of accelerating voltage. Fig. 6 View largeDownload slide The FWHM and contrast of the dislocation image as a function of accelerating voltage. The contrast of the threading dislocations in GaN changes with temperature as shown in the CL images in Figs. 7a–e. Figure 7f shows a TEM image of the same area at a thickness of 500 nm. The FWHM of the dark dot contrast is small at low temperatures below 100 K, increasing with temperature up to ∼200 K, and then decreasing with temperature to room temperature. The contrast also changes correspondingly with the FWHM, i.e. the contrast is strong when the FWHM is narrow. The FWHM and contrast measured from the CL images in Fig. 7 are shown in Fig. 8. Fig. 7 View largeDownload slide CL images taken at various temperatures; (a) 40 K, (b) 100 K, (c) 120 K, (d) 200 K, (e) 300 K and (f) A TEM image of the tilted sample from the incident electron beam direction. Fig. 7 View largeDownload slide CL images taken at various temperatures; (a) 40 K, (b) 100 K, (c) 120 K, (d) 200 K, (e) 300 K and (f) A TEM image of the tilted sample from the incident electron beam direction. Fig. 8 View largeDownload slide The FWHM and contrast of the dislocation image as a function of temperature. Fig. 8 View largeDownload slide The FWHM and contrast of the dislocation image as a function of temperature. Discussions The CL image of the dislocation is contrasted with a spatial resolution of the TEM–CL technique, which is limited by three factors: (1) an incident electron beam diameter, (2) an electron beam spread in a sample and (3) the diffusion length of a minority carrier. The incident beam diameter does not change, if an experimental condition such as a beam current is fixed. The beam spread depends on accelerating voltage and sample thickness. These two parameters give the region of excitation such as carrier generation, which is an effective probe of the CL measurement for insulators. In case of semiconductors the generated excess carriers move to surrounding region by diffusion. The dislocation acts as a non-radiative recombination center and traps the carriers passing close to the dislocation line to give rise to the dark contrast in the CL image. We then can determine the diffusion length by analyzing the CL contrast of the dislocation. The sample is n-type GaN, and we consider the behavior of holes as a minority carrier. First we consider a simple case assuming that the carrier generation function g(r) is constant in space [g(r)=G] and the carrier concentration is zero inside a cylindrical region of a radius r0 around a dislocation line. Using a cylindrical coordinate (r,θ,z), eq. (2) is reduced to   \[D\left(\frac{d^{2}}{dr^{2}}\ +\ \frac{1}{r}\ \frac{d}{dr}\ {-}\ \frac{1}{L^{2}}\right)p\left(r\right)\ +\ G\ =\ 0,\] (11) where L is the diffusion length of the minority carrier and is related to D and τ as \(L\ =\ \sqrt{D{\tau}}\) . Only the radial component is left because of the cylindrical symmetry of p(r) around a dislocation running along the z axis. If, considering an infinite medium and the conditions that p(r)=0 for r≦r0 and p(∞)=Gτ, a solution for the excess hole concentration at a distance r from the dislocation is given as   \[p\left(r\right)\ =\ G{\tau}\left[1\ {-}\ \frac{K_{0}\left(r/L\right)}{K_{0}\left(r_{0}/L\right)}\right],\] (12) where K0 is the zeroth-order modified Bessel function. This equation is approximated as   \[p\left(r\right)\ =\ G{\tau}\left[1\ {-}\ \mathrm{exp}\left(\frac{r}{L}\right)\right].\] (13) This formula was sometimes used for evaluating the diffusion length from the CL contrast of dislocations in semiconductors [15–17]. Figure 9a shows a CL intensity profile across the threading dislocation at a sample thickness of 500 nm. Plots show an observed profile from the CL image of Fig. 3 and a solid line is a fitting curve. We use Gaussian formula for fitting instead of eq. (13), since both give nearly the same value of L. From this fitting the value of L is 82 nm, which is a half of the FWHM. However, eq. (13) expresses an emission intensity distribution under the uniform illumination such as in the case of photoluminescence (PL) and electroluminescence (EL). Obviously in the case of CL, the profile expressed by eq. (13) should be convoluted with a spread function of the carrier generation region. In that sense the value is overestimated. However, this treatment ignores the surface recombination which reduces the effective diffusion length in the thin sample. Thus this treatment is only a qualitative one and gives an approximate value of the diffusion length. Fig. 9 View largeDownload slide (a) Profile of the CL intensity across the dislocation taken at room temperature, (b) a calculated intensity profile of the dislocation contrast, (c) The trajectories of the incident electrons and (d) the excess carrier distribution at the recombination points after the diffusion process. Fig. 9 View largeDownload slide (a) Profile of the CL intensity across the dislocation taken at room temperature, (b) a calculated intensity profile of the dislocation contrast, (c) The trajectories of the incident electrons and (d) the excess carrier distribution at the recombination points after the diffusion process. The situation is complicated in the real TEM–CL measurement. The carrier generation function g(r) is not spatially uniform but is spread in a finite region around an incident beam position due to the electron scattering process. It is difficult to derive an analytical solution for the diffusion equation using this g(r) in the case of a threading dislocation, because the cylindrical symmetry is broken. Surface recombination is also important, which restricts the lateral spread of carriers in the diffusion process. In order to simulate the observed CL image, we have to calculate the CL intensity emitted from the whole sample region as a function of the electron beam position. The theoretical approach has been done by numerical calculations using the Monte Carlo simulation [9]. The excess minority carrier distribution is calculated by this simulation program. In order to calculate the dislocation contrast, several parameters such as the diameter of a cylindrical region around a dislocation and the carrier lifetimes inside the region are involved. Figure 9b shows calculated profiles of the dislocation contrast by the Monte Carlo simulation with proper parameter values; r0=20 nm and τd=80 ps for the edge-type threading dislocation. The surface recombination rate is taken to be 5×104 cm s−1. These values were used for the CL image fitting of the dislocations in the previous study [10]. The sample thickness is 500 nm, and the accelerating voltage is 80 kV. Profiles were calculated for various diffusion lengths of the perfect region. The diffusion length of 256 nm gives a good fit. This value is similar to the previously reported value of 250 nm by Rosner et al. [16], who measured with Si-doped GaN at room temperature by SEM-CL. Figure 9c shows the trajectories of the incident electrons where the carriers are generated at each scattering point. Figure 9d shows the excess carrier distribution at the recombination points after the diffusion process in the perfect region. The carrier distribution expands in the sample from (c) to (d). The FWHM of the distribution is 160 nm, which is shorter than the diffusion length. The diffusion length obtained here is longer than the one previously obtained for the threading dislocation in the undoped GaN (L=158 nm) at room temperature. Here we obtain the diffusion length for holes, Lp, in Si-doped GaN with the doping concentration of 2×1018 cm−3 at room temperature. The diffusion length generally decreases with doping concentration. However, we must be careful of the other factors which influence the dislocation contrast such as surface condition and impurity decoration along the dislocation. These factors can change the surface recombination rate and the lifetime in the defect region. Further systematic study would be necessary to get a more complete understanding of the dislocation contrast. The diffusion length depends on temperature, which reflects the change in CL contrast of the threading dislocation with temperature as seen in Fig. 7. The FWHM of the dislocation contrast is small at 40 K and increases with temperature up to ∼200 K, and then decreases to room temperature. This behavior should be explained by the property of the diffusion length. The diffusion length is written by \(L\ =\ \sqrt{D{\tau}}\) , and the diffusion constant D is expressed using the Einstein's relation as D = (kT/e)μ, where μ is the carrier mobility. Then the diffusion length is written as   \[L\ =\ \sqrt{\frac{kT}{e}{\mu}\left(T\right){\tau}\left(T\right)}.\] (14) The mobility is expressed as   \[{\mu}\ =\ \frac{e{\tau}_{\mathrm{d}}}{m^{{\ast}}},\] (15) where m* is the effective mass and τd is the mean free time for collision of the carriers. There are many types of collisions in semiconductors; collision with ionized impurities is dominant at low temperatures and that with phonons becomes dominant at sufficiently high temperatures. The data plots in Fig. 10 show the hole mobility in Mg-doped p-GaN with an expected dopant concentration of 1.7×1019 [18]. Such temperature dependence of mobility can be calculated by solving the Boltzmann transport equation iteratively using several factors for acoustic deformation potential scattering, piezoelectric scattering, polar optical phonon scattering and ionized impurity scattering [19]. At low temperatures the mobility increases with temperature mainly due to the ionized impurity scattering, which has temperature dependence of T3/2. The collision with phonons has T−3/2 dependence, and so the mobility decreases with temperature at higher temperature range. A bold solid line in Fig. 10 shows the diffusion coefficient D calculated from this data and using Einstein's relation. Temperature dependence of the carrier lifetime τ can be qualitatively estimated from the dislocation contrast through eq. (7). In Fig. 8 the dislocation contrast monotonically decreases with temperature. This indicates that the carrier lifetime monotonically decreases with temperature, since the lifetime in the defect region τD is independent of temperature. Consequently the diffusion length expressed by \(L\ =\ \sqrt{D{\tau}}\) shows the anomalous behavior seen in Fig. 8. Fig. 10 View largeDownload slide The hole mobility in p-GaAs taken after the literature [18] and the diffusion coefficient calculated from the mobility. Fig. 10 View largeDownload slide The hole mobility in p-GaAs taken after the literature [18] and the diffusion coefficient calculated from the mobility. Conclusion In the present study, we examined the monochromatic CL contrast of the threading dislocations in Si-doped n-GaN by the TEM–CL. The contrast and the FWHM of the dislocation image depend on sample thickness, accelerating voltage and temperature. The CL spectra were measured at various temperatures and were analyzed to find the property of the BE emission used for the CL imaging. We found that the blue emission is attributed to the recombination of the FX and bound exciton (D0, X) whose energies are close to each other. The observation showed that the FWHM of the dislocation contrast monotonically increases with sample thickness and decreases with accelerating voltage. Such behavior qualitatively agrees with that expected from eq. (1), though quantitatively the calculated value overestimates the spread of carriers. On the other hand, the temperature dependence of the FWHM shows an anomalous behavior. This dependence can be explained from the hole mobility and lifetime as a function of temperature. The relation between the FWHM of the dislocation contrast and the diffusion length is also discussed. References 1 Petroff P M, Logan R A, and Savage A ( 1980) Non-radiative recombination at Dislocations in III–V compound semiconductors. Phys. Rev. Lett.  44: 287–291. Google Scholar 2 Yamamoto N, Spence J C H, and Fathy D ( 1984) Cathodoluminescence and polarization studies from individual dislocations in diamond. Phil. Mag. A  49: 609–629. Google Scholar 3 Mitsui T and Yamamoto N ( 1997) Distribution of polarized-cathodoluminescence around the structural defects in ZnSe/GaAs(001) studied by transmission electron microscopy. J. Appl. Phys.  81: 7492–7496. Google Scholar 4 Yamamoto N ( 2002) Development of CL for semiconductor research I; TEM–CL study of microstructures and defects in semiconductor epilayers. Lecture Note in Physics 588 . pp. 37–51 (Springer Verlag, Berlin). Google Scholar 5 Mita T, Yamamoto N, Mitsui T, Heun S, Francioci A, and Bonard J M ( 2001) Cathodoluminescence study of the Y0 emission from ZnSe films. Solid State Phenomena 78/79: 89–94. Google Scholar 6 Akiba K, Yamamoto N, Grillo V, Genseki A, and Watanabe Y ( 2004) Anomalous temperature and excitation power dependence of cathodoliminescence from InAs quantum dots. Phys. Rev. B  70: 165322-1-9. Google Scholar 7 Davidson S M ( 1977) Semiconductor material assessment by scanning electron microscopy. J. Microsc. 110: 177–204. Google Scholar 8 Yacobi B G and Holt D B ( 1986) Cathodoluminescence scanning microscopy of semiconductors. J. Appl. Phys.  59: R1–R24. Google Scholar 9 Grillo V, Armani N, Rossi F, Salviati G, and Yamamoto N ( 2003) A random walk simulation approach to cathodoluminescence processes in semiconductors. Inst. Phys. Conf. Ser.  180: 565–568. Google Scholar 10 Yamamoto N, Itoh H, Grillo V, Chichibu S, Keller S, Speck J S, DenBaars S P, Mishra U K, Nakamura S, and Salviati G ( 2003) Cathodoluminescence characterization of dislocations in GaN using a transmission electron microscope. J. Appl. Phys.  94: 4315. Google Scholar 11 Hino T, Tomiya S, Miyajima T, Yanashima K, Hashimoto S, and Ikeda M. ( 2000) Characterization of threading dislocations in GaN epitaxial layers. Appl. Phys. Lett.  76: 3421–3423. Google Scholar 12 Jain S C, Willander M, Narayan J, and Van Overstraeten R ( 2000) III-nitrides: growth, characterization, and properties. J. Appl. Phys.  87: 965–1006. Google Scholar 13 Chichibu S, Okumura H, Nakamura S, Feuillet G, Azuhata T, Sota T, and Yoshida S ( 1997) Exciton spectra of cubic and hexagonal GaN epitaxial film. Jpn. J. Appl. Phys.  36: 1976–1983. Google Scholar 14 Gil B (ed) ( 1998) Group III Nitrides Semiconductor Compounds. (Clarendon Press, Oxford). Google Scholar 15 Brantley W A, Lorimor O G, Dapkus P D, Haszko S E, and Saul R H ( 1975) Effect of dislocations on green electroluminescence efficiency in GaP grown by liquid phase epitaxy. J. Appl. Phys.  46: 2629–2637. Google Scholar 16 Rosner S J, Carr E C, Ludowise M J, Girolami G, and Erikson H I ( 1997) Correlation of cathodoluminescence inhomogeneity with microstructural defects in epitaxial GaN grown by metalorganic chemical-vapor deposition. Appl. Phys. Lett.  70: 420–422. Google Scholar 17 Sugawara T, Sato H, Hao M, Naoi Y, Kurai S, Tottori S, Yamashita K, Nishino K, Romano L, and Sakai S ( 1998) Direct evidence that dislocations are non-radiative recombination centers in GaN. Jpn. J. Appl. Phys.  37: L398–L400. Google Scholar 18 Nakayama H, Peter H, Khan M R H, Detchprohm T, Hiramatsu K, and Sawaki N ( 1996) Electrical transport properties of p-GaN. Jpn. J. Appl. Phys.  35: L282–L284. Google Scholar 19 Huang D, Yun F, Reshchikov M A, Wang D, Morkoc H, Rode D L, Farina L A, Kurdak C, Tsen K T, Park S S, and Lee K Y ( 2001) Hall mobility and carrier concentration in free-standing high quality GaN templates grown by hydride vapor phase epitaxy. Solid State Electronics  45: 711–715. Google Scholar Author notes 1Department of Physics, Tokyo Institute of Technology, Tokyo 152-8551 and 2Department of Research and Development, Nichia Chemical Industries, Ltd., Kaminaka, Anan, Tokushima 774, Japan © The Author 2005. Published by Oxford University Press on behalf of Japanese Society of Microscopy. All rights reserved. For permissions, please email: journals.permissions@oxfordjournals.org
TI - Contrast analysis of dislocation images in TEM–cathodoluminescence technique
JF - Journal of Electron Microscopy
DO - 10.1093/jmicro/dfi026
DA - 2005-06-01
UR - https://www.deepdyve.com/lp/oxford-university-press/contrast-analysis-of-dislocation-images-in-tem-cathodoluminescence-BSl0YvPxe0
SP - 223
EP - 230
VL - 54
IS - 3
DP - DeepDyve
ER -