High resolution X-ray diffraction and grazing incidence small an-

High resolution X-ray diffraction

The X-ray techniques are a tool used for the structural characterisation of the crystal structure. Several methods are possible, each offering different information about the crystal. In comparison with TEM, X-ray diffraction (XRD) is based on the interference of photons that are elastically scattered at the electron cores of the crystal atoms. In a XRD setup the sample is illuminated with X-rays ¹ and the resulted scattered light is collected by a detector. Both the sample and the detector can be rotated in this setup. In Fig. 2.22 the typical rotation, i.e., scan directions are indicated. In particular, in a ω/2θ scana coupled rotation of the sample, aroundω, and the detector, around2θ, takes place.

For the ZnSe material system, this kind of scan is typically performed across the (004) reciprocal lattice point. In this configuration the penetration depth of the light is around 7µm and consequently, all epitaxial layers forming a laser structure are detectable. The main information that can be extracted from aω/2θ scan is, the lattice matching of the epilayers with the GaAs substrate.

Since the intensity of the scattered light depends on the scattering volume, a first rough estimation of the layer thicknesses is also possible. A more precise information concerning the thicknesses can be extrapolated, if the sample contains layers with very smooth interfaces. This layers give rise to thickness oscillations, called also thickness fringes, observed as a modulation of the scattering signal. However, in a multilayer sample it is difficult to assign the thickness fringes to the correct layer. For sufficiently thick layers the crystalline quality can be judged by estimating the width of the layer peak [62] inωdirection.

Further insight into the crystalline quality of a sample is obtained by performing a mapping of the reciprocal space. Such a reciprocal space map is obtained by measuring

1λ= 1.54056A (Cu K˚ _α1line)

ω 2θ

G

scattering vector

sample

ω scan

incident wave vector

k k

origin of the reciprocal space

reciprocal lattice point

scan

incident beam

scattered wave vector

θ Ewald sphere

ω/2θ

Figure 2.22: Schematic view of the possible scans in the X-ray diffraction [20].

multipleωscans along theω/2θdirection around a reciprocal lattice point, as indicated in Fig. 2.22. In such a measurement configuration the relaxation of the layers can easily be detected since their in-plane lattice constant differs from the substrate lattice con-stant. In the mapping of an asymmetrical reflex, e.g. of the (224) reflex, the layer peak has a differentq_x position with respect to the substrate. When the layer peak positions are all at the same q_xvalue of the substrate, i.e.,∆q_x = 0, the structure is fully strained.

The signal of a fully relaxed layer is placed on the line that connects the substrate spot and the origin of the reciprocal space. The triangle formed by the substrate peak posi-tion, the layer peak position in the fully strained case, and in the fully relaxed case is the so-calledrelaxation triangle[8].

Grazing incidence small angle X-ray diffraction

In comparison with conventional X-Ray techniques, the diffuse scattering techniques are an important tool for characterisation of self-organised crystal structures. Investiga-ting the diffuse scattering around the reciprocal lattice points, information about strain, shape, size and correlation effects can be obtained. The advantage of these scattering techniques is the higher resolution obtained in reciprocal space which is a direct con-sequence of the high angular resolution of just a few seconds of arc. Due to the high X-ray penetration depth, the volume fraction of the mesoscopic structures is small com-pared with the illuminated volume. Therefore the contribution of the useful scattering signal has to be increased. Grazing incidence techniques decrease significantly the sig-nal coming from the underlaying layers and are suitable for the study of surfaces and interfaces. Among them, the grazing incidence small angle X-ray scattering (GISAXS) technique is suitable to determine the ordering in self-assembled quantum dot struc-tures and can average statistically over a large assemble of dots. The setup correspond-ing to such an experiment in shown in Fig. 2.23, correspondcorrespond-ing to the SAXS beamline of the ELETTRA synchrotron beamline. An X-ray energy of 8 Kev corresponding to a 1.55 ˚A wavelength is employed. Using a torroidal mirror, the beam is focused on the detector behind the sample. A slit system sets a convenient beam size at the sample

2.6 Complementary experimental techniques

Figure 2.23: Schematic setup of the SAXS beamline at ELETTRA synchrotron facility (Italy). Schematic drawing made by T. Schmidt.

Figure 2.24: Scattering geometry for the case of a position sensitive detector (top) and a CCD camera (bottom). Schematic drawing made by T. Schmidt.

position. In grazing incidence geometry the incident angleα_i on the sample surface is chosen well above the critical angle of total external reflectionαc which is in this case 0.32 allowing an adequate information depth. The information is collected on a position sensitive detector (PSD), providing one dimensional information, or on the CCD camera providing 2D information (see Fig. 2.24). Whereas the PSD, which provides a larger dy-namic range, records the intensity at vanishing parallel momentum transferq_k = 0, the additional real space coordinate x on the CCD camera resolvesq_k in the direction trans-verse to the incoming beam. From the measurement of the scattered intensity alongq_⊥ measured with PSD in dependence on grazing incidence angleα_i, one can distinguish two intense features. Specularly reflected beam corresponding to an exit angleα_f equal with the α_i is observed followed by the Yenoda peak where theα_f was equal with the critical angle for external reflection α_c. The correlation signal is detected in the diffuse scattering intensity atq_k 6= 0.

Chapter 3

Degradation of Cd-rich quaternary quantum well laser diodes

3.1 Research status and motivation

The world of telecommunication relies on the use of optical data transmission employ-ing laser and light emittemploy-ing diodes as signal emitter. The use of plastic optical fibre (POF) with polymethyl methacrylate (PMMA) core for transmitting the signal reduces the costs drastically in comparison with the doped silica fibres. The main obstacle are the wavelength dependent transmission losses in the POF [63], therefore an appropriate light source is needed. The transmission losses in these plastic fibres could be signifi-cantly reduced by the use of light sources emitting around 560 nm where the PMMA has its absolute loss minimum.

ZnSe based laser structures grown on GaAs substrate are an appropriate candidate due to the good structural quality with which they can be produced. The epitaxy group of Prof. Hommel (Institut f ¨ur Festk ¨orperphysik Bremen University) has a rich experi-ence in growing ZnSe structures on GaAs [26, 8, 25, 20]. Using ternary CdZnSe QWs the wavelength of the emitted light can be tuned from 471 nm to 550 nm [64]. The longer wavelength region can be achieved by using quaternary CdZnSSe QWs with high Cd content [20]. In this situation a highly strained QW is grown on GaAs.

However, the lifetime of ZnSe based laser diodes still remains a problem, even af-ter the strong reduction of the stacking fault density experimentally obtained. Under current injection, the quantum efficiency decreases (the devices become darker) and the best laser diode obtained in Bremen could be operated in continuous wave (cw) mode for only 4 min [20]. The degradation has been related to problems connected with the nitrogen doping used to obtain p-type conductivity (nitrogen related deep complex centers act as efficient nonradiative centers during operation [65, 66]) and with the in-stability of the QW during operation, which is regarded to be the main obstacle for the application of II-VI materials for light emitters.

To investigate the mechanism behind the gradual darkening of the light emitting de-vices during operation several experimental methods can be employed, such as: electro-luminescence (EL), photoelectro-luminescence (PL) or cathodoelectro-luminescence (CL) (Sec.2.6). In terms of device relevance, the EL topography is most important [67, 68, 69]. The EL ex-periments revealed the generation of dark defects during operation of the device. They are well known features commonly found in degraded opto-electronic devices [70, 71]

and identified also as main cause of the degradation in conventional III-V devices. They appear as dark spot defects (DSD) and dark line defects (DLD).

The combination of EL measurements and TEM studies of defect characterization is one of the most successful tool in order to understand the degradation mechanism. For example, it was observed that in II-VI structures the DSD are seen directly after turn-on of the current, and nucleated at the site of a pre-existing defect, usually a stacking fault [72]. Furthermore, the defects responsible for the reduced light emission are confined to the active region [42]. During operation, these defects grow in number and form a network of triangular shape [73] consisting of dislocation loops which have at the final stage of degradation a density of 10¹⁰cm⁻²[43]. However, there are some discrepancies regarding the Burgers vectors of these defects. Guha et al. reported vectors of the type (a/2)h110i lying in h100i junction plane, Hua et al. reported that the dislocations are oriented along theh430idirection [42, 43] and Tomiya et al. from Sony found networks involving dislocation dipoles with Burgers vectors of (a/2)h011i inclined at 45 to the h001ijunction plane [72]. The reason for the discrepancies between the different reports on ZnSe-based devices is unclear, but may be attributed to different growth conditions and sample preparation techniques.

In the case of DLD, it can be distinguished between two different types. The DLD aligned in h110i direction are correlated with misfit dislocations in the quantum well [67]. They do not grow significantly under current injection and in consequence have limited impact on the process of degradation [72]. They indicate a partially relaxed QW as was also verified using samples grown in Bremen [74]. The other type of DLD is oriented alongh100idirections and appear during the degradation experiments, under current injection. Therefore their influence on the process of degradation is enhanced.

The nucleation site for these DLD are the DSD. Qualitatively, one observes that the DLD are actually DSD growing along the above mentioned directions [75].

Once that the presence of dark defects was experimentally identified as main cause of the degradation in ZnSe based laser structures, the next question which arises is how this defects were generated. It was found that their formation is directly connected with the recombination of the carriers in the QW, with an excitation energy at least as high as the band gap energy of the QW [76]. Furthermore Chang et al. developed a model for their growth under current injection [77]. Some of the carriers injected into the system recombine non-radiatively at a defect site and the absorbed energy is released into crystal lattice via generation of phonons. This leads then to local heating and strong vibration of the defect atoms, which reduces the barrier for the motion of the defect, enabling the defect migration or even new defect generation (recombination enhanced defect reaction – REDR) .

In the case of generation and growth of dark defects in ZnSe based devices two types of mechanisms are important which are combined together [78]. A recombination en-hanced dislocation glide (REDG) process is responsible for the expansion of the defects inh110idirection no point defects being necessary. The formation of theh100iDLD oc-curs via a dislocation climb (REDC) process requiring the presence of point defects for growth [79, 80]. For very low defect densities (<10⁻⁴ cm⁻²) [81] and optimized growth conditions, no dark features are observed and a gradual darkening of the active region occurs. Still this darkening is connected with REDR [78, 79] caused by the diffusion of point defects in the material.

3.1 Research status and motivation In the following, the parameters that can influence the degradation mechanism are pointed out. Since REDR are related to the diffusion of the point defects (generated mainly in the growth process), the situation is complicated by the presence of p-type doping of ZnSe material, which can act as an additional source of point defects. For p-type nitrogen doped structures a decrease of the QW emission is observed, which is not the case for non-p-type doped structures [82]. Under current injection or optical pumping the N-acceptors can transform into three-fold positively charged interstitial N complexes [83], which can migrate in the presence of the electric field. Due to the migration towards the n-type region, point defects are generated in the QW. Another group reported the generation of DLD and the darkening of the active region also for undoped structures [84]. The experimental results obtained in this respect in the epitaxy group of Prof. Hommel (Uni Bremen) revealed also the influence of the p-type doping, but is different in comparison with Ref. [82]. This discrepancy is most probably con-nected with the type of N-source used. However, by low n-type doping of the QW, the stability of the devices could be improved [85], which fits with the fact that Cd exhibits an increased diffusion tendency in the presence of p-type doping [86, 87].

However, for a device under operation there is another aspect that drives the degra-dation: the heat generated during current injection. The influence of heat onto devices was studied in pulse-mode under varying pulsing conditions [88, 89]. Lifetime mea-surements made for different laser structures at constant output power and current densities around the threshold revealed the same time evolution. After an initial drop of the operating current, a low increase is found which is attributed to the defect an-nealing processes [90, 91, 86]. A stronger increase was observed only in the last 10%

of the operating time. This emphasizes that the heat generation does not change the process of degradation, but influences its velocity. Accordingly, the REDR model is also valid at high current injection levels and thus, represents the fundamental degradation mechanism of these devices.

Usually, the active region of the ZnSe laser diodes consists of CdZn(S)Se QW with high Cd content up to 40%. Therefore, this QW is under high compressive strain. Since it was mentioned before that a relaxed QW can produceh110idark lines in EL topogra-phy experiments, it plays also an important role in the degradation process. Moreover, the REDR mechanism is based on the diffusion of point defects and it is well known that in the presence of p-type doping Cd has a strong tendency to diffuse [87, 86]. Tem-perature dependent annealing measurements revealed also that the point defect dif-fusion starts from GaAs/ZnSe heterointerface, which acts as reservoir for point defects [92, 93, 94]. The temperature stability of the PL emission of CdZnSe QW depends on the distance from the interface. In consequence, the point defect diffusion plays an impor-tant role for the Cd diffusion process [95]. Combined with consimpor-tant-current degradation experiments made for different light emitting devices, there is a strong suggestion that a strained quantum well and/or Cd diffusion are major driving forces for the point defect diffusion based degradation.

Since microstructural investigations of the defect formation mechanisms are rare and not really conclusive [42, 96], this work is focused on the microstructural changes in the active region of the device due to the QW degradation. In combination with the EL measurements, used to analyze the output intensity and peak wavelength of the device, the microstructural sensitive methods comprising TEM, as well as HRXRD, can provide a clearer picture of the device degradation process.