Chapter 3 Experimental
3.2 Atomic force microscopy (AFM)
In the early 1986s, Gerd Binning and Calvin F. Quate invented atomic force microscopy (AFM) and realized that it is able to measure the interactive force between atoms of the probe tip and sample surface by using special probe which consists of a flexible, elastic cantilever and sharp tip.
Fig. 3-2-1 shows the configuration of AFM. The sharp tip locates at the free end of a flexible cantilever attached to the scanner. The scanner controls two independent movements of the cantilever: scanning along the sample surface (X-Y plane) and movement perpendicular to the surface (along the Z-axis). The scanner is made of piezoelectric material that expands or shrinks depending on the applied voltage. When the cantilever is bent by the repulsive or attractive force from the interaction between the tip and the sample surface, it changes the deflection. During scanning, the detection system measures the cantilever deflection from its initial position and then sends a signal proportional to the deflection to the scanner control system. The feedback signal is used to move the probe up or down by the piezoelectric crystal to bring the parameter back to its original values. Simultaneously, the probe displacement is recorded by computer and interpreted as the specimen topography.
There are three operation modes in AFM, i.e. contact, non-contact, and
tapping mode. Contact mode has the highest resolution of all three modes in which the tip touches the surface and scans over the sample. But in this way, the tip or sample may easily be damaged by the scanning process. Non-contact mode is preferred to avoid the probe deformation since it utilizes the long range Van der Waal’s force between the tip and specimen. However, the sensibility and resolution is limited in the interference from ambient environment. Tapping mode is the compromise of the above, which consists of contact and non-contact mode. Detection of tapping mode is more sensitive than that in non-contact mode and less destructive on probes or specimen than that in contact mode.
Scanning probe microscopy (SPM) system used in our lab is Slover P47H, manufactured by the “Molecular Devices and Tools for Nano Techology (NT-MDT)” in Russia. It can be operated in multi-modes such as AFM for morphology measurements. In our studies, the InGaN dots morphology was measured by tapping mode in order to optimize the resolution and avoid probe destruction. The AFM probe, which has a cantilever about 50 or 80 μm and a sharp tip with a radius of curvature about 10 nm (Fig. 3-2-2), is also from NT-MDT.
Fig. 3-2-1 Schematic diagram of AFM system.
Fig. 3-2-2 The SEM image of scanning probe and the tip.
3.3 Macro-photoluminescence (Macro-PL)
Macro-PL system consists of two parts, one for the infrared range and the other for visible signals as shown schematically in Fig. 3-3-1.
I. The infrared signals:
For infrared signals, a He-Cd laser (Kimmon IK5552R-F) operating at 442 nm lines was used as the excitation light source with maximum output power ~ 92 mW. The beam passed through a 442 nm band-pass filter to ensure that the excitation source is pure from 400 nm to 2000 nm. The laser beam which was focused by a f=15 cm lens onto the sample, with a beam size ~250 μm. The luminescence signals were collected by lens, in which the laser intensity was blocked by a 850 nm long- pass filter (Thorlabs), and then coupled into a monochromator (ARC Pro 500). The dispersed signals were detected by a extended InGaAs detector (EOS), and processed using a lock-in amplifier (Stanford SR530), and then sent into Acton Spectra Hub for PL analysis.
II. The visible signals:
Laser operating at the 325 nm UV line was used as the excitation light source with the maximum output power ~ 22 mW. The configuration of PL in visible region is similar to the infrared, except that the luminescence signals are reflected by a UV mirror and feed into a optical fiber, which was finally coupled
into the monochromator (ARC Pro 500). The signals were dispersed by the monochromator, whose spectral resolution is about 0.25 nm when both the entrance and exit slits where opened to 250 μm. The dispersed signals were detected by a photomultiplier tube (Hamamatsu R943-02) using the normal applied voltage -2000 V, along with a photon counter (Hamamatsu C1230) for detection.
Low-temperature macro-PL was carried out using a closed cycle cryogenic system (APD HC-2D). The temperature was varied from 13 K to 300 K by Lakeshore 330 temperature controller.
Lock-in amplifier Spectral Hub
cryogenic stage
InGaAs detector
He-Cd laser (Kimmon IK5552R-F) mirror
mirror
Focus lens f=15 cm
325 nm
Long-pass filter Fiber coupler
Fiber mirror
chopper
442 nm band-pass filter
PMT
Fiber adaptor Coupling lens
monochromator monochromator
Photon counter
PC (infrared signals ) PC (visible signals )
coupling lens
He-Cd laser (Kimmon IK5552R-F) mirror
mirror
Focus lens f=15 cm
325 nm
Long-pass filter Fiber coupler
Fiber mirror
chopper
442 nm band-pass filter
PMT
Fiber adaptor Coupling lens
monochromator monochromator
Photon counter
PC (infrared signals ) PC (visible signals )
coupling lens f=30 cm
Fig. 3-3-1 Schematic diagram of macro-PL system.
3.4 Near-field scanning optical microscopy (NSOM)
The optical coupling system and the scanning probe are the most important parts of NSOM measurements, which are described respectively in this section.
I. Experimental setup
Fig. 3-4-1 showed the schematic diagram of NSOM system. The scanning measurement was performed with Solver SNOM (Olympus based) developed by NT-MDT. A He-Cd laser (Kimmon) operated at the 325nm is used as the excitation light source which is reflected from the edge filter (Semrock, LP02-364RU) then coupled into the optical fiber and illuminated on the sample through the probe. The PL signals were collected by the same probe and passed through the edge filter that blocked the excitation laser beam. The PL signals were dispersed by the monochromator (ARC Pro 500) and detected by the Hamamatsu R955 photomultiplier tube. The PMT signals were processed by the lock-in amplifier (Stanford SR530) and the Solver SNOM controller. Our system collected the PL signals and topography images at the same time to show the NSOM mapping image.
II. Fabrication of NSOM probe
The optical probe is the most critical part of the near-field microscope for achieving high resolution images. NSOM probes have been fabricated from a
variety of materials, including cleaved crystal, semiconductor structures, glass pipettes, and tapered optical fibers [10]. Following fabrication of the tapered tip, the sides of the probe are coated with an opaque metal film (usually aurum, platinum, or aluminum) to prevent light loss in regions of the waveguide other than the aperture. There are many different ways to fabricate and characterize near-field optical probes. There are a variety of techniques that have proven to be useful for fiber tapering, but the task is most commonly accomplished by chemical etching or by heating and pulling, and in some cases both methods are combined to achieve the desired taper characteristics. Our NSOM probe was made by chemical etching of BOE solution (HF: NH4F = 1:6) and coated with platinum metal film by using ion sputter (Hitachi E-1010). Then, the probe was adhered to a tuning fork with the resonance frequency at 32.768 KHz. Fig. 3-4-2 showed the SEM image of coated tip.
Fig. 3-4-1 Schematic diagram of NSOM system.
3.5 Time-Resolved Photoluminescence (TRPL)
For TRPL measurements, the excitation pulse of width about 120 fs and the repetition rate locked to the frequency of 80 MHz was generated from a mode-locked Ti:Sapphire laser (Tsunami). Its output beam at λ=700~720 nm was frequency-doubled to the UV region by a BBO (Beta Barium Borate ) crystal. The laser beam was focused by a f = 15 cm focusing lens onto the sample to a spot about 500 μm in diameter, then we used a 400 nm long-pass filter (10LWF-400-B) to remove the second harmonic light. The TRPL signals were dispersed through a single-grating monochromator (ARC Spectro PRO-500) with a 1200 lines/mm grating and detected by an avalanche photodiode (APD), which are then sent to the picosecond timing discriminator (EG&G-9307) to define the arrival time. The delay line (EG&G-425A) is incorporated to postpone the signals long enough to arrive at the picosecond time analyzer (PTA EG&G-9308) after 50 ns interpolation dead time as the Start pulse. The reference frequency of 80 MHz is directly sent to the constant fraction discriminator (EG&G-584) and then marks the arrival time of the analog pulse by sending a timing logic pulse to the Stop input of the picosecond time analyzer (PTA EG&G-9308). Finally, the signals were received by a GPIB card and recorded by a computer for data processing.
Tsunami pulsed Laser
Fig. 3-5-1 Schematic diagram of TRPL system.
Chapter 4 Results and discussion
In this chapter, we will discuss the surface morphologies, alloy compositions and emission properties of these In-rich InxGa1-xN dots (x >0.87) depending on growth temperature by AFM, XRD, PL, NSOM, and TRPL measurements.
Not only the NIR emission but also the visible emission was observed, in which NIR signals agree with XRD results. Furthermore, NSOM mapping of the visible emission provided the spatial correlation with the non-dot region from AFM morphologies. Finally, we will propose an energy scheme to explain both the NIR and visible emissions as shown in PL spectra.
4.1 Surface morphology of InGaN dots
As shown in AFM images of Fig. 4-1-1, the sample grown at 550 °C appears to be a rough film formed by coalescent islands. For growth temperatures in the range of Tg = 600-725 °C, individual InGaN dots were observed. With the increasing Tg, the average height and diameter of the InGaN dots increased from 24 to 144 nm and 75 to 410 nm, respectively; whereas the dot density decreased from 4.0×109 to 2.2×107 cm-2, see Table 4-1 and Fig. 4-1-2. For Tg between 600 and 700 °C, the decreasing dot density with the increasing Tg can be attributed to
the enhanced migration length of adatoms between In and Ga adatoms, that resulted in the formation of much larger and less dense dots at higher growth temperatures [11]. For Tg > 700 °C, the dot density decreased rapidly and dropped to zero at 750 °C, due possibly to the desorption of metallic In from the growing surface.
To further study the nucleation mechanism of InGaN dots, we have also prepared a series of samples with InN dots grown at different Tg under similar growth conditions. The density of InN dots as a function of Tg was also included in Fig. 4-1-2. We found that the densities of InGaN and InN dots show a similar dependence on Tg, implying that the nucleation of InGaN dots is governed by the surface migration of In adatoms, rather than Ga or both. This can be realized from the very different migration capabilities of In and Ga adatoms on the GaN surface. Since the bond strength of InN (7.7 eV/atom) is weaker than GaN (8.9 eV/atom), In adatoms (or its adsorbed precursor molecules) are expected to have a considerably longer migration length on the GaN surface, leading to a much higher nucleation rate and hence governing the formation of InGaN dots.
This increasing trend is also a consequence of the different migration capabilities of In and Ga adatoms, which considerably hinders the incorporation of Ga into InN islands to form InGaN alloys. Because the Ga migration length is
relatively shorter, they are unable to travel long enough to reach the already formed InN islands (or In-rich InGaN islands) during the deposition periods of TMIn and TMGa. This effect is expected to be more pronounced at higher growth temperatures, due to the even larger difference in their migration capability. For the case of 725 °C grown sample, the InxGa1−xN dots become highly In-rich, with x up to 0.99. Therefore, it can be inferred that most of the deposited Ga adatoms are very likely to be distributed among these In-rich islands, forming a thin Ga-rich layer. We would like to point out that the deduced In content from Fig. 4-1-2 does not follow the prediction of equilibrium solubility of GaN in InN [12], where the Ga content is expected to increase with the growth temperature based on thermodynamic considerations. This implies that the incorporation of Ga into InN during the growth of InGaN dots is kinetically inhibited, rather than under thermodynamic equilibrium.
(a) Sample a (Tg = 550 ℃) (b) Sample b (Tg = 600 ℃)
(c) Sample c (Tg = 650 ℃) (d) Sample d (Tg = 700 ℃)
(e) Sample e (Tg = 725 ℃) (f) Sample f (Tg = 750 ℃)
Fig. 4-1-1 AFM morphology images of InGaN dots
750℃ - - Average height
(nm) Average height
(nm) Growth
temperature Tg (℃)
Table 4-1 Average height/diameter and dot density of InGaN dots taken from 5×5 μm2 image.
500 550 600 650 700 750
107
Fig. 4-1-2 The dot density versus growth temperature as a function.
4.2 X-ray diffraction (XRD) spectrum of InGaN dots
From the reference [13], the diffraction peak of InN (0002) plane at about 31.3°, and that of GaN (0002) plane occurs at 34.6°. The XRD (θ-2θ) scans of the InGaN samples were investigated between these two peaks as shown in Fig.
4-2-1. The diffraction curve of InN film grown on GaN was also included as the reference to determine the In compositions of InGaN dots.
For the InGaN samples, a broad diffraction feature corresponding to the ternary InGaN dot shift gradually toward the InN (0002) as Tg was increased until 725 °C. However, the ternary InGaN feature disappeared for the 750 °C grown sample. It is due to the significant desorption of In from the growing surface at indium nucleation critical point ~750 °C. This result is consistent with its surface morphology. Besides, the narrowing trend of FWHM reflects better quality at higher growth temperature in the range of Tg = 550-700 °C as shown in Fig. 4-2-2. Furthermore, using the Vegard’s law formula for InxGa1-xN alloy as the following :
Eg(x) = x · Eg,InN + (1-x) ·Eg,GaN - b · x · (1-x) (1) where the bowing parameter b ~1.43 eV was adopted [14], we have estimated that the In content (x) of the InxGa1−xN dots increases from 0.85 to 0.99 as Tg was increased from 550 to 725 °C as shown in Fig. 4-2-3.
Nevertheless, there is a continuous tail distributed between the InN (0002) and the GaN (0002) diffraction peaks that rised gradually toward the GaN (0002) peak, especially for samples grown at higher Tg. It can be inferred that the Ga adatoms would be deposited and very likely formed a thin Ga-rich layer [15], so that it is revealed as a continuous tail covering wide diffraction angles.
Our XRD results agree fairly well with the PL peak energy and are consistent with the PL emission efficiency that will be discussed in the next section.
31 32 33 34 35
x=0.96
550
oC 600
oC 650
oC 700
oC 725
oC
x=0.85 x=0.99 x=0.97
x=0.92
Int ensi ty ( a.u. )
2 θ (degree)
InN bulk InN (0002) GaN (0002)
750
oC
Fig. 4-2-1 X-ray diffraction of InGaN dots grown at different temperatures.
500 550 600 650 700 750
Fig. 4-2-2 The XRD and FWHM pattern of the InGaN dots grown at different temperatures.
500 550 600 650 700 750
0.82
Fig. 4-2-3 The estimate In content (x) from XRD of the InGaN dots samples
4.3 Optical properties of InGaN dots
4.3.1 The Macro -PL spectrum results
From the macro-PL results in Fig. 4-3-1(a), no PL signals were detected for the 550 °C grown sample. It is believed that low-temperature growth is detrimental to their optical quality. This result agrees with the surface morphology feature in which we saw a rough film formed by coalescent islands in the AFM images. For the growth temperatures Tg ranging from 600 to 725 °C, individual InGaN dots were observed, but for Tg = 750 °C the dot density dropped to zero. Concurrently, the PL spectra exhibit a NIR emission band from 0.77 eV to 0.92 eV as the growth temperature Tg was increased from 600 to 725
°C, but absent again at 750 °C. Moreover, our PL spectra feature agrees not only to the surface morphology but also to the XRD results.
From the redshift of NIR peak energy with the increasing Tg, we obtained the corresponding In content (x), which is close to that determined by XRD. Besides, the x-ray FWHM indicate better quality data of samples grown in the range of Tg
=600 to 700 °C, which can be compared with the PL emission efficiency as shown in Fig. 4-3-2. While the x-ray FWHM decreases, the integrated NIR PL intensity increased rapidly.
However, for the Tg = 750 °C InGaN sample the AFM results show no dot
formation, implying that the NIR band emission is originated from the In-rich InGaN dots. Another evidence of the NIR PL emission from the dot region is described as follows: it is expected that the PL emission peak should reflect the In content estimated by Vegard’s law as the dashed line in Fig. 4-3-3. The NIR emission peak follows the dashed line closely, implying that the NIR emission is from the dots. Hence, we attributed the NIR emission to a band to band transition from the InGaN dots regions.
Moreover, we also observed a visible emission band in the range of 2.0 - 2.4 eV for samples grown at Tg ≥ 650 °C which is shown in Fig. 4-3-1(b). It is significantly strong for the sample grown at Tg = 750 °C, implying that the visible emission is not originated from the dots, because no dot formation in this sample. Therefore, the visible emission is very likely to arise from non-dot region, which may be a thin Ga-rich layer formed during the growth of InGaN dots. To further examine it, we carried out NSOM measurements and identified that the visible emission is in deed from non-dot region as will be discussed in the next section.
0.6 0.8 1.0 1.2
Fig. 4-3-1 PL spectra of In-rich InGaN dot grown at different temperatures.
500 550 600 650 700 750
0.3
Integrated Intensity 0.9
PL In te g r at ed I n ten sity ( a .u .)
Growth Temperature T
g
(
oC )
FWH M (degree)
FWHM
Fig. 4-3-2 PL intensity versus the FWHM measured by X-ray at different temperatures
In molar fraction x
Calculated of bandgap (b=1.43) LT-PL peak energy
Fig. 4-3-3 PL emission peak vs. In fraction by Vegard’s law theoretical prediction
4.3.2 The NSOM spectrum results
NSOM is a powerful tool with nanoscale spatial resolution. Employing this technique, we were able to obtain both the surface morphology and the mapping of visible emissions at room temperature simultaneously. Fig. 4-3-4 shows such a NSOM image of the 725 °C sample probed at its visible emission peak of 2.3 eV along with the corresponding surface morphology. It can be clearly seen that these two images are nearly complementary, with dark spots in the NSOM image coincident very well with the In-rich dots revealed in the AFM image.
This confirms that the visible emission indeed occurred in the relatively flat region outside the In-rich InGaN dots or from the interface between bottom of dots and the surface of HT-GaN buffer layer. As can be seen from Fig. 4-3-4, the visible emission is mainly contributed from regions with less dot density, though there are some bright spots.
μ μ
Fig. 4-3-4 NSOM image of the 725 °C sample mapping is in the upper part, and the corresponding surface morphology is shown below for comparison.
4.4 The visible emission band
Besides the NIR emission band originated from the InGaN dots region, there is also the visible emission band. Whether this emission arises from the thin Ga-rich layer or not needs to be clarified, since most of the deposited Ga adatoms at higher growth temperature are likely to form a thin Ga-rich layer.
Thus, we employed various excitation wavelengths, and powers to examine how the visible signals are generated. In addition, a time correlated detection of TRPL measurement is also presented. The results are interpreted by a proposed scheme of transitions between DAP energy levels.
4.4.1 Excitation wavelength-dependence
For the 725 °C InGaN sample, the excitation wavelength was tuned from 355 nm to 362 nm that covers the range from above to below the GaN band gap. The average incident power on the sample was 20 mW with a beam spot about 500 μm in diameter. The 13 K PL spectra are shown in Fig. 4-4-1. Obviously, the emission exhibits a broad distribution from 1.8 to 2.1 eV depending on the excitation wavelength. The PL intensity decreases as the excitation wavelength is changed from above to below GaN band gap energy, as shown in Fig. 4-4-2.
For excitation wavelength below the GaN band gap (below~358 nm), only a
weak luminescence at low energy side (P1) reflects a continuous band. As the excitation wavelength was tuned to above the band gap, a strong luminescence appeared at 1.95 eV (P2). The PL peak energy does not strongly depend on the excitation photon energy. Furthermore, the (P2) intensity increases monotonically as the excitation wavelength is tuned to above the band gap. As a result, we suggest that the broad band emission must involve GaN related-defect levels for recombination of photo-generated carriers by excitation above the band gap. Moreover, the varying PL spectral feature by tuning across the band gap energy reveals that the thin Ga-rich InGaN layer in the flat regions may not
weak luminescence at low energy side (P1) reflects a continuous band. As the excitation wavelength was tuned to above the band gap, a strong luminescence appeared at 1.95 eV (P2). The PL peak energy does not strongly depend on the excitation photon energy. Furthermore, the (P2) intensity increases monotonically as the excitation wavelength is tuned to above the band gap. As a result, we suggest that the broad band emission must involve GaN related-defect levels for recombination of photo-generated carriers by excitation above the band gap. Moreover, the varying PL spectral feature by tuning across the band gap energy reveals that the thin Ga-rich InGaN layer in the flat regions may not