Material characteristics analysis - Experiment techniques

Chapter 3: Experiment techniques

3.2 Material characteristics analysis

3.2-1 High resolution X-ray diffraction system

X-ray offers a non-invasive way to analyze sample properties, which interferences with regularly arranged atoms of crystal and reflect constructively if it meets the Bragg’s condition.

By varying the incident and received angles, sample information such as composition, layer thickness, lattice constant, and quality can be deduced. Our system is a commercial BeDe D1 high resolution X-ray diffraction meter, which uses crystals on the work path to get a better resolution. Measurements are usually performed at the (004) surface of the zinc-blende crystal and operated in mode. By comparing the signals of epi-layers to the substrate, compositions of epi-layers can be checked. Quality of layers can be judged from its full width of half maximum (FWHM). Fig 3.9 gives the example, which is three different GaAsSb layers with different composition grown on an InP substrate.

-240 -10.000

170

950

101

102

103

104

105

106

-6000 -4000 -2000 0 2000 4000 6000

Peak Position

Intensity (cps)

OMEGA-2THETA (arcsec)

Fig. 3.9 X-ray data scanned in mode for the example of three GaAsSb layers grown on an InP substrate, where peak position values are noted.

3.2-2 Photoluminescence Spectroscopy

Photoluminescence spectroscopy is a simple and non-destructive tool to study samples. In principle the samples are excited by a light source and electron-hole are carriers generated.

The carriers relax to lower electronic levels of the structure and recombine spontaneously to release photons. Intensity and spectrum of the luminescence can be collected and measured to analyze sample quality, quantized energy levels, emission wavelength, impurity levels, etc.

Sample properties also can be characterized by varying excitation power and temperature.

Our homemade PL system setup is depicted in Fig. 3.10. A continuous Ar+ laser with a tunable wavelength of 488nm or 514.5nm serves as an excitation source, which is conducted via reflections of mirrors and normally incident to a 2-inch diameter CaF2 focusing lens. It is focused on the sample surface with a spot size ~100m diameter. The spontaneous emission from the sample goes through a CaF2 window then collected and transformed into a parallel light by the CaF2 lens. It goes successively through another 2-inch CaF2 lens, a laser filter, and is finally focused on the entrance slit of our monochromator, Horiba iHR 550. Samples are placed on a copper chuck mounted to a helium close-cycled cryogenic system, which is capable of tuning temperature from ~10K to 390K inside the chamber under a vacuum~10^-1 torr. The monochromator has gratings to disperse the light and is integrated with detectors to detect photons. We have two photo detectors for mid-IR regime; one is a thermal electric cooled InGaAsSb photodetector with optical response to ~2.5m, while another is a liquid nitrogen cooled InSb photodetector with optical response to ~5m. The signals are handled with a conventional lock-in technique to suppress noise. A chopper is inserted in the Ar+ laser light path and modulated at ~200Hz frequency. Photo detector signals and chopper triggers are fed to a SR-530 lock-in amplifier. The lock-in amplifier can process and output the data to a computer via IEEE 488 bus.

3.2-3 The setup for optically-pumped lasers measurements

Optically-pumped lasers are different from electrical injected p-n junction lasers, which establish carrier population inversions via optical pumping. The light of the pumping source can penetrate through the thick cladding layer and generate photo carriers inside the active region. Optical pumping provides the way to archive lasing when electrical injection is not favorable due to reasons like large turn on voltage and high resistivity. It has advantages that can be exploited to produce high output power laser and good beam quality. GaSb-based

Fig. 3.10 The Scheme of the PL setup of our lab. The excitation source form an Ar+ laser beam (488nm/ 514nm) is directed via mirror reflections, going through chopper, density filter, and focused on to sample surface via a CaF2 lens. The sample is placed on a copper chuck mounted on a closed cycle cryogenic system capable of temperature tuning from ~10K to 390K. The spontaneous emission is collected by two CaF2 lenses, passing through a laser filter, and arrives at the entrance slit of the IHR550 monochromator. The singles from the photo detector, which is in the exit slit of the monochromator, finally go to the computer after processing with a conventional lock-in technique.

optically pumped “W” lasers have been demonstrated with higher output peak powers and maximum operation temperatures over others [43-45]. It also offers a possibility to perform lasing at dual wavelengths [46].

The pumping source is a commercial 1064nm pulsed fiber laser. The photon energy of 1064nm is 1.17eV, which is small than the band gap of the cladding layer material, In0.52Al0.48As, for our lasers. This obviates pumping light absorption by the cladding layer, which is important for optical pumping consideration. The fiber laser can be controlled by a computer which can set the output power percentage, pulse duration, and repetition rate. The output power with maximum 100% percentage can reach a peak power of ~10KW. There are eight pulse durations of 10ns, 30ns, 20ns, 50ns, 75ns, 100ns, 150ns, and 200ns can be selected.

The repletion rate can be synchronized by triggers from outside TTL signals and operated arbitrarily from single shot to 500 KHz. Fig 3.11 shows the measured output power as function of output percentage for 20ns pulse duration operated under 100Hz, 1KHz, and 10KHz.

A general optical pumping experiment setup is depicted in Fig. 3.12. The fiber laser is shaped into a size of 3mm x 200m via a cylindrical lens. It then focuses on the top on sample surface with the along axis in parallel with cavity direction. The mid-IR laser emission from the sample cleaved edge is collected as the same way as the PL setup when measuring the laser spectrum. As for the total output laser intensity (L-L measurements), the mid-IR light from cleaved mirror is collected directly by a aspherical lens (with numerical valure > 0.5) and detected by a thermal-electric cooled InAs photodiode which has integrated with a front filter to block the 1064nm scattering light of the pumping source. The photo detector signal is fed to the lock-in amplify which processes and transfers the data to computer. A function generator is used to generate TTL signals which are sent to the fiber laser and the lock-in amplify for the reference input.

Fig. 3.11 The 1064 nm fiber laser output peak power as function of output percentage for 20ns pulse duration operated under 100Hz, 1 KHz, and 10 KHz.

Collected by PL system

as measuring 1064nm

pulsed laser

Cylindrical lens

Sample with cleaved laser cavity

Filter

Aspherical lens

TE cooled InAs detector

TTL signals

Function generator Fiber laser

Computer

Lock-in amplify

Fig. 3.12 The setup for optical pumping measurements (L-L curve). If the spectrum needs to be analyzed, the output light is collected using the PL system, which is equipped with the monochromator.

Chapter 4 2~3  m mid infrared light sources using InGaAs/GaAsSb “W” type quantum wells on InP substrates

After introducing theoretical considerations and analysis techniques in chapter 2 and 3, this chapter moves to real growth of our designed type-II “W” quantum wells. The samples were studied via temperature and excitation power dependent photoluminescence experiments. The characteristics of the type-II “W” QWs, such as blue shift of PL spectrum along with increasing excitation power density and the trade-off situation between long emission wavelength and optical momentum matrix element, were discussed and compared with simulation results.

4.1 The growth of ternary alloys lattice-matched to InP substrate

Prior to the growth of real structure, it is important to assure the ternaries, In_0.53Ga_0.47As, In0.52Al0.48As and GaAs0.49Sb0.51, have correct compositions to match with the lattice constant of InP substrate. For the ternaries with two group-III species, it was done by adjusting the growth rates of the two group-III species. In our lab, the indium growth rate was calibrated on the InAs substrate, which has a different lattice constant (6.06Å ) with the GaAs substrate (5.65 Å ) used for gallium and aluminum growth rate calibrations. It should be considered as to derive the correct alloy mole fraction ratio. Taking In0.53Ga0.47As as an example, the indium growth rate (g_In) and gallium growth rate (g_Ga) roughly follow the relation (g_In/6.06³) /(( g_Ga/5.65³)+(g_In/6.06³)) =0.53. Notice that the growth rates are all normalized by its own cubic lattice constant. Conventionally we grew three layers with slightly various growth rate ratio of two III-species, each 200nm thick, on InP substrate. The lattice match condition is

finally checked by XRD measurement. Fig. 4.1 shows the XRD results of InGaAs and InAlAs for the lattice match check.

As seen by the Fig. 4.1, the lattice match condition were found and confirmed by one of the three peaks coincided with the InP substrate. The growth rate of indium is 0.5 m/hr; the growth temperature is around 500^oC (pyrometer value); the group V to total group III BEP ratio is around 15-20. The samples were also performed with low temperature PL measurements as shown in Fig 4.2 to check band gap energies.

-420-220

-6000 -4000 -2000 0 2000 4000 6000

Peak Position

-6000 -5000 -4000 -3000 -2000 -1000 0 1000 2000 3000 4000 5000 6000 Peak Position

Intensity (cps)

OMEGA-2THETA (arcsec)

(a) InGaAs (b) InAlAs

Fig. 4.1(a) InGaAs and (b) InAlAs XRD results for lattice match check on InP substrate. Each sample grown with three layers of slightly various compositions, one of which is nearly coincided with the main peak of InP substrate.

1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60

0.0

Fig. 4.2 InGaAs and InAlAs low temperature PL results for lattice match check on InP substrate.

The bang gap for InGaAs is around 0.8eV and for InAlAs is around 1.5eV.

The GaAsSb ternary composition was mainly controlled by Sb2/As2 BEP ratio. The lattice match condition is shown by the XRD data in Fig. 3.7. The low temperature PL measurement is shown in Fig. 4.3, which reveals band gap energy around 0.81 eV.

4.2 The growth condition dependence of Sb fraction in GaAsSb alloy

The GaAsSb includes two group V species that makes it is more difficult to control the composition. The Sb fraction of the alloy is not linear with the As2/Sb2 BEP ratio, which is shown in Fig. 4.4 according to the work [47].

Fig. 4.3 GaAsSb low temperature PL result for lattice match check on InP substrate. The bang gap energy is around 0.81eV.

0.5 0.6 0.7 0.8 0.9 1.0

0.0 0.2 0.4

Normalized intensity

Energy (eV) LT

0.81eV GaAsSb for InP lattice match

Fig. 4.4 Antimony composition of GaAsSb as a fuction of Sb and As beam equivalent pressure ratio FSb/(FSb+FAs) [ref. 42].

The Sb incorporation is also greatly influenced by the growth temperature [48]. It has a higher sticking coefficient and incorporation efficiency for the Sb2 molecular beam under lower growth temperature. The growth temperature effect for the Sb incorporation of the GaAsSb alloy in our MBE system was evaluated by the following experiment. Three GaAsSb layers each with 150nm thickness were grown on InP substrate at pyrometer temperatures, in turn, of 485^oC, 470^oC, and 460^oC, while the As2 and Sb2 BEP were kept the same ( Sb2 : 9.3x10^-7 torr, As2 : 8x10^-7 torr). The compositions then were decided by XRD data assuming strain is relaxed in each layer, which is shown in Fig. 4.5. The composition and lattice constant values are listed in the Table.

Tem. (^oC) 460 470 485

Com. GaAs0.25Sb0.75 GaAs0.32Sb0.68 GaAs0.4Sb0.6

Lattice con.( Å ) 5.99 5.95 5.92

Fig. 4.5 XRD examinations of antimony incorporation of three GaAsSb layers grown on InP substrate at different temperatures, the growth temperatures, compositions ,and lattice constants of each layer listed in the Table.

As seen by the Fig. 4.5, the growth temperature indeed influences the Sb incorporation of GaAsSb ternary dramatically. The temperature difference from 485^oC to 460^oC leads to a change of the Sb mole fraction from 0.6 to 0.75 in GaAsSb alloy. For precisely determination of the Sb fraction of GaAsSb pseudomorphic layer in “W” structure, it had better to separately grow the GaAsSb layer with same condition during the same run for the XRD check of the composition.

4.3 “W” type quantum wells sample growth

In the following, the mid-IR light sources between 2 m to 3 m were studied and designed using the type-II “W” type QWs [27, 28, 49]. We demonstrate that the emission wavelength can be extended longer than 2.56m at room temperature.

Our samples were grown on S-doped (001) InP substrates. The As2 and Sb2 species were used through the equipped needle-valve controlled cracking cells. The wafer surface temperature was monitored by an infrared pyrometer. For the multilayer growth consideration, the structure of the “W” QWs designed here are the symmetric In_0.53Ga_0.47As/

GaAsSb/In0.53Ga0.47As layers sandwiched between two 2nm In0.36Al0.32Ga0.32As tensile strain layers to compensate the compressive strain in the GaAsSb layer. Nine such “W” type QWs spaced by 25 nm In_0.52Al_0.48As barrier layers were grown and placed between two 200 nm In0.52Al0.48As layers, as shown in Fig. 4.6. In order to optimize the “W” structure, this study was performed systematically by varying the thickness of the InGaAs/GaAsSb layers, the composition of the GaAsSb layer and the growth temperature. The growth parameters are listed in Table 4.1, where three series of samples are grouped and labeled as A, B, and C.

Groups A and B are samples with variable thicknesses of InGaAs and GaAsSb, respectively.

Group C contains samples with different Sb mole fraction in the GaAsSb layer by adjusting the Sb₂/As₂ beam equivalent pressure (BEP) ratio. The growth temperature is also changed

from group A to group C.

Fig. 4.6 The structure of the designed “W” type QWs.

Table 4.1 The InGaAs/GaAsSb layer thickness, the Sb2/As2 BEPratio, the summarized PL peak wavelength, and FWHM of the “W” type QW samples in group A, B, and C.

4.4 Photoluminescence results and discussion

After the samples were grown, the photoluminescence (PL) measurement was carried out using a 488nm Ar⁺ laser as the excitation source and a thermal electric cooled InGaAs(Sb) detector. All the PL spectra were calibrated by the response spectrum of a 1000^oC black radiation source. The PL spectra measured at 20K are presented in Fig 4.7(a). The peak emission wavelength (peak) covers the range from ~2 m to ~2.5 m. As summarized in Table 4.1, the peak can be extended from 2.05 m to 2.47 m by increasing the InGaAs layer thickness from 4 nm to 10 nm as shown in group A, and can be extended from 2.17 m to 2.42 m with the increase of GaAsSb layer thickness from 2 nm to 4 nm (in group B). The sensitivity of peak to the InGaAs/GaAsSb layer thicknesses agrees well with our calculations.

Notice that the full width of half to maximum (FWHM) of the PL spectra increases from 13.9meV to 22.7meV when the GaAsSb layer thickness is reduced, as indicated in group B.

This is caused by the energy level broadening due to the thickness fluctuation when the GaAsSb layer is thin. The peak can be also extended by the increase of the Sb-content, as shown by group C samples. The Sb fraction is determined by fitting the peak with the calculation results. The peak goes from 2.37 m to 2.48 m, around a 23 meV difference, as the Sb mole fraction increases from 0.74 to 0.78. We can also see the effect of growth temperature by comparing these three groups of samples. It is found that the samples grown at a lower temperature have a longer emission wavelength and a narrower PL spectrum. This indeed indicate that the Sb incorporation is more efficient and the fluctuation of alloy composition is less at lower growth temperatures [50, 51]. Figure 4.7(b) shows the integrated PL intensity as a function of the emission wavelength. It clearly shows the trade-off situation as predicted by the simulation. The calculated curve, which is based on the assumption that the integrated PL intensity is proportional to the square of the wave function overlap, fits very nicely with the experimental results.

Fig. 4.7 (a) PL spectra of samples in group A, B, C, and (b) integrated PL intensity (normalized at sample A1) plotted against the peak wavelength. The calculated result is plotted as the solid curve.

The optical transition of the type-II heterostructure has been extensively studied previously [52, 53].The peak usually blue shifts with the excitation power (P_ex), and the amount of shift is linearly dependent on the one-third power of P_ex because of the band bending effect caused by the accumulation of spatially separated electrons and holes in the adjacent triangular interface potential wells. However, the power dependent emission of a “W” type QW has never been studied in detail. We found that, although peak shifts to a shorter wavelength as Pex

is increased, the energy shift does not obey the P_ex^1/3 law, especially under low P_ex. The power dependent spectra of sample A1 and C1 are presented in Fig. 4.8(a), and the amount of energy shift is plotted as a function of the P_ex in Fig. 4.8(b) along with the ideal P_ex^1/3 curve for comparison (both axes in log scale). Since the curves are not linear in this log-log plot, the energy shift vs. Pex does not follow any power law. We also notice that sample A1 has a more pronounced energy shift than sample C1. This power dependent behavior is not due to the heating effect since the integrated PL intensity is linearly proportional to Pex, as shown in the inset of Fig. 4.8(b). The possible reason for the amount of blue shift to deviate from the P_ex^1/3 law is the state filling effect of the localized states. The localized states are caused by interface roughness and alloy composition fluctuation, which are common in the ternary alloy [54, 55]. Under low excitation powers, the generated carriers occupy the localized states with lower energies. As the amount of carriersis increased with increased Pex, higher energy states are occupied and therefore the peak of the emitted light shifts to a shorter wavelength. The extended tail in the low energy side of the PL spectra shown in Fig. 4.7(a) and Fig. 4.8(a) is an indication of the radiation from these localized states.

Fig. 4.8 (a) Power dependence PL spectra of sample A1 and sample C1, and (b) the energy shifts versus Pex along with the simulation results and the ideal Pex1/3

curve for comparison. The inset shows the power dependence of the integrated PL intensity.

We have performed a simulation for such effect by assuming a joint density of states associated with the localized states as erfc((<E₀>-E)/E), where erfc is the complementary error function, <E0> and E refer tothe average and the standard deviation of the transition energy to take into account of the inhomogeneous broadening effect [56]. E values of 30meV and 18meV were used in the calculation for samples A1 and C1. When the Fermi level is raised because of a higher pumping power, the emission peak blue shifts to a higher energy.

This behavior is represented schematically in Fig 4.9. The calculated results are shown together with the experiment data in Fig. 4.8(b). Excellent agreement between the calculated and the experimental results was achieved. Since sample A1 has a broader PL spectrum, it has a larger E and a more pronounced states filling effect. Therefore it has a larger energy shift compared to that of sample C1. We have also estimated the blue shift contributed by the band bending effect. The bending potential profile is solved by the Poisson equation, which depends on the charge distributions. However, the charge distributions, assumed proportional to the wavefunction probability, are solved by the potential profile depended Schrödinger equation. We solve the two coupled equations iteratively until the results are convergent. It turns out the blue shifts along with the carrier densities are not much, for example the mount of blue shift is around 5meV for the structure A1 as the 2D carrier density is increased from 2x10¹¹ cm^-1 to 1x10¹²cm^-2, which is over the density range Ar+ laser excited in the experiments. The reason is that both electron and hole have been under the strong quantum confinement in the “W” structure. The band bending effect caused by the charge separation merely perturbs the system without altering the quantum levels much, which might be

在文檔中第二型砷化銦鎵/砷銻化鎵“W”量子井之光學特性及光激發中紅外線雷射之研究 (頁 65-0)