Study on ingan gan quantum structures and their applications in semiconductor saturable absorber mirror 5

4.12 without the AR coating, and the LT and HT GaN buffers were shown together as GaN buffer, there were large variations of the refractive indices at the interfaces of sapphire substrat

Chapter 5

Optimization of GaN-based SESAM

In Chapter 4, the non-monolithic GaN-based SESAM with broad stopband and high maximum reflectance was successfully fabricated using dielectric DBR and

AR coatings However, severe interference-induced reflectance fluctuations, which might lead to the instability in passive mode-locking, were found within the stopband

of the SESAM Optimization in the device structure was therefore necessary to suppress the reflectance fluctuations before the SESAM can be reliably used in passive mode-locking In this chapter, Section 5.1 investigates the origin of the interference fringes within the stopband of the GaN-based SESAM It is found that the thick GaN buffer is the major cause of the reflectance fluctuations By performing simulations on the layer structures, a new SESAM structure with thinner GaN buffer

to suppress the reflectance fluctuations was then designed Subsequently, the fabrication of the new SESAM structure is described in Section 5.2, which, according

to the simulation, is able to suppress the reflectance fluctuations The fabrication process includes wafer bonding, laser lift-off of the sapphire substrate, inductively

coupled plasma (ICP) etching of the GaN buffer and the PECVD deposition of the AR coating Finally, in Section 5.3, the surface morphology, crystal quality of the SESAM structure and the emission property of the QWs will be analyzed to examine the effects of the fabrication process on the characteristics of the SESAM This whole chapter will be focused on the GaN-based SESAM sample fabricated by the InGaN/GaN 8-QW saturable absorber, as discussed in Chapter 4

5.1 Simulation

First of all, the origin of the interference fringes within the stopband of the GaN-based SESAM was investigated As shown in the unmodified GaN-based SESAM structure sketched in Fig 5.1 (same as Fig 4.12 without the AR coating, and the LT and HT GaN buffers were shown together as GaN buffer), there were large variations of the refractive indices at the interfaces of sapphire substrate / air (Δn ≈ 0.7), GaN buffer / sapphire substrate (Δn ≈ 0.8), and GaN cap / first SiO2 layer of DBR (Δn ≈ 1) These three interfaces are referred to as interfaces A, B and C, respectively, as indicated in Fig 5.1 The cavities formed by the above interfaces are the sources of the light interferences Therefore, there are two major sources which can cause the interferences: the cavity formed by interfaces A and B and that formed

by interfaces B and C According to the period of the interference fringes (16 nm in average as shown in Fig 4.13) which is related to the thickness and refractive index

of the layers between two interfaces, the observed interference fringes within the SESAM stopband were caused by the cavity formed by the interfaces B and C The

cavity formed by interfaces A and B would give a high-frequency oscillation beyond the wavelength resolution of the spectrophotometer used, so the effect of the sapphire substrate on the reflectance fluctuations was not observed in our experiment

Figure 5.1 Schematic structure of the GaN-based SESAM (This is same as Fig 4.11 without the AR coating.)

The interference-induced reflectance fluctuations caused by the cavity formed by interfaces B and C can be controlled by thinning the GaN buffer and other epilayers (GaN barriers / cap and QWs) The purpose of thinning these layers is to increase the period of the interference fringes so as to reduce the reflectance fluctuations within a small wavelength span For example, in the oscillation fringes in Fig 4.13, there is a valley at 417 nm and an adjacent peak at 425 nm The oscillation period around this wavelength region is about 16 nm, and the valley-to-peak

reflectance difference is about 6.5 %, which is also the oscillation magnitude If the period of the fringes is largely increased, assuming that the oscillation magnitude is unchanged, the reflectance difference between 417 nm and 425 nm will be much less than the oscillation magnitude (6.5 %) Hence, the reflectance fluctuations within this wavelength span (417 – 425 nm) are suppressed For the thinning of the GaN-based layers, obviously the dimension of the QW structure (the barriers, QWs and cap) can not be modified by this process as it provides the desired nonlinear saturable absorption characteristics Besides, only part of the GaN buffer can be removed as one layer of GaN is needed to form the barrier for the InGaN/GaN QW

Based on the above, we can reduce the reflectance fluctuations within the stopband by removing the sapphire substrate and thinning the GaN buffer But then the interface between the thinned GaN buffer and air may also cause interference This can be reduced by an AR coating on the top of the thinned GaN buffer For this

AR coating, the substrate is now GaN instead of the sapphire substrate for the AR coating used for the unmodified SESAM in Fig 4.12 Hence, the single quarter-wavelength SiO2 AR coating, as illustrated in Table 4.1, should be used Reflectance simulations using Essential Macleod Thin Film Design Program were then carried out on the various structures Figure 5.2 shows the simulated reflectance spectra of the SESAM: (a) for the unmodified structure shown in Fig 5.1; (b) after sapphire substrate removal; (c) after further thinning of the GaN buffer until ~350-nm GaN is left; and (d) eventually with the SiO2 AR coating on the thinned GaN buffer

in Fig 5.2 (a), disappeared after the removal of the sapphire substrate as shown in Fig 5.2 (b) Furthermore, after thinning the GaN buffer, the period of the interference-induced oscillation fringes was increased and there was only one interference minimum falling into the DBR stopband as can be seen from Fig 5.2 (c) Due to the large refractive index difference between the air and the thinned GaN buffer, the surface reflectance could be as high as about 30 % The SESAM at this stage acted as resonant type We can reduce the reflectance from the thinned GaN buffer with an AR coating to make this SESAM anti-resonant After adding a SiO2

layer with the optical thickness of 1/4 λ (λ = 425 nm) as an AR coating, the surface reflectance could be reduced to less than 0.7 %, resulting in the reduction of the oscillation amplitude as shown in Fig 5.2 (d) These simulation results show that, by sapphire substrate removal, GaN layer thinning and the additional AR coating, the interference-induced reflectance fluctuations within the stopband of the SESAM can

be effectively suppressed

Figure 5.2 Simulated reflectance spectra of the SESAM (a) for the unmodified structure shown in Fig 5.1 (taken from the sapphire substrate side), (b) after sapphire substrate removal (taken from the GaN buffer side), (c) after further thinning part of the GaN buffer (taken from the thinned GaN buffer side), and (d) eventually with the SiO2 AR coating on the thinned GaN buffer (taken from the AR coating side)

5.2 Experiments

Figure 5.3 Experimental reflectance spectra of the SESAM (a) for the unmodified structure shown in Fig 5.1 (measured from the sapphire substrate side), (b) after laser lift-off of the sapphire substrate (measured from the GaN buffer side), (c) after the ICP etching of the GaN buffer (measured from the etched GaN side), and (d) eventually with the SiO2 AR coating on the etched GaN layer (measured from the AR coating side)

According to the above reflectance simulations of various structures, a series

of experiments was subsequently conducted A Shimadzu UV-VIS-NIR scanning spectrophotometer was used to perform the reflectance measurements The recorded reflectance spectra of the SESAM after each step are shown in Fig 5.3

Figure 5.3 (a) is the measured reflectance spectrum of the unmodified SESAM structure sketched in Fig 5.1, and it shows severe interference-induced reflectance fluctuations over the DBR stopband Comparing Fig 5.3 (a) with the corresponding simulation result, Fig 5.2 (a), one can see that the high frequency fringes did not appear in Fig 5.3 (a) The high frequency oscillations have a period of less than 1 nm, which is unable to be picked up by the spectrophotometer with a wavelength accuracy of ± 0.5 nm The low frequency oscillation predicted in simulation (Fig 5.2 (a)) has the same period as that observed in the experiment (Fig 5.3 (a))

5.2.1 Wafer bonding

Before the sapphire substrate removal, the SESAM sample as sketched in Fig 5.1 was first bonded to another supporting substrate for easier handling Here a second sapphire substrate was used as the supporting substrate because it is chemically stable at elevated temperatures (ex 280oC) A benzocyclobutene (BCB)

3025 polymer [Niklaus2001] was chosen as the bonding medium because it could sustain the high temperature during the final PECVD deposition of the AR coating

To perform the wafer bonding, the SESAM was cut into a 3 mm × 3 mm square piece

which was then bonded to an 8 mm × 8 mm supporting sapphire The bonding procedure was as follows: 1) Both the SESAM and the supporting substrate were cleaned with acetone, methanol and rinsed with deionized (DI) water 2) After drying

in an oven at 200oC, the supporting substrate was spin-coated with an adhesion promoter, followed by ~ 3-μm BCB 3025 polymer 3) The supporting substrate was placed on a hotplate at 100oC for 10 minutes to evaporate the solvent 4) The SESAM was then attached to the supporting substrate with the DBR side facing the bonding interface, so that the sapphire substrate to be removed was exposed Pressure is applied to remove any trapped air 5) The bonded structure was finally cured at 280oC

on a hotplate for 1hr The structure after wafer bonding is illustrated in Fig 5.4, and it was ready for laser lift-off of the sapphire substrate

Figure 5.4 Schematic structure of the SESAM sketched in Fig 5.1 after being bonded to the supporting sapphire and ready for laser lift-off

5.2.2 Laser lift-off

As the double-side polished sapphire was used as the substrate for the unmodified SESAM, it is transparent to the excimer laser beam Wong et al have demonstrated the damage-free separation of GaN thin films from sapphire substrates

by pulsed KrF excimer laser [Wong1998] The transparency of sapphire and the thermal decomposition of GaN at the interface were combined to realize the separation of the sapphire substrate from the rest of the structure

In our work, a Novaline 100 KrF excimer laser with a wavelength of 248 nm and a pulse width of 25 ns was used The laser beam spot-size was 2 mm × 2 mm The laser pulse with an energy of 285 mJ was incident through the sapphire substrate and irradiated at the sapphire / GaN interface The pulsed UV irradiation and short GaN optical absorption length resulted in the localized heating of the GaN at the interface [Wong1999] A distinct change in the interfacial region from transparent to a metallic silver color was observed, which indicated that the interfacial GaN had decomposed into metallic gallium and N2 gas Four laser pulses were used to cover each corner of the SESAM sample, respectively The sapphire substrate was then detached from the GaN film leaving a shiny gallium surface The excess gallium on the surface was subsequently removed in diluted HCl (HCl : H2O = 1:1)

Figure 5.3 (b) shows the reflectance spectrum of the SESAM after removing the surface gallium Figure 5.3 (b) is essentially similar to Fig 5.3 (a), because the interference caused by the sapphire substrate was not resolvable in Fig 5.3 (a) as

simulated reflectance spectrum), it was found that the oscillation magnitude obtained from the sample was much smaller than that from the simulation Note that the nature

of the surface morphology was not considered in the simulation; however, the sample showed a poor surface morphology after laser lift-off of the sapphire substrate, which will be shown later in Section 5.3 The rough surface of the sample would compromise the cavity effect, therefore resulting in the much smaller oscillation magnitude in the reflectance spectrum

5.2.3 ICP etching

Thinning of the GaN buffer was then performed in inductively coupled plasma (ICP) chamber with the substrate holder maintained at the temperature of 6oC and a chamber working pressure of 5 mTorr BCl3 and Cl2 were used as the etchants

A similar GaN buffer thinning process after laser lift-off has also been demonstrated

by Wong et al using an ion milling technique [Wong19991]

The ICP etching process was controlled until there was ~ 350-nm GaN buffer left The remaining GaN buffer was not only used to maintain the integrity of the QW structure, but also to prevent the QW regions from being damaged during the ICP etching The reflectance spectrum after GaN etching is shown in Fig 5.3 (c) It agrees well with the simulated result shown in Fig 5.2 (c) The period of the oscillation fringes was greatly increased

5.2.4 AR coating deposition

SiO2 ofquarter-wavelength optical thickness was then deposited onto the

thinned GaN buffer as an AR coating This layer was expected to reduce the surface reflectance to less than 0.7% The AR coating deposition was also conducted in the Nextral (NE) D200 Unaxis PECVD system described in Section 2.2 at the chamber temperature of 280oC The backside coating process on the exposed GaN buffer after laser lift-off was also reported by Song et al in the fabrication of a vertical injection blue light emitting diode [Song1999]

The final reflectance spectrum after AR coating is shown in Fig 5.3 (d) It is clear that the interference-induced reflectance fluctuation is effectively suppressed The experimental results agree very well with the simulation

We also noted that the oscillation fringes outside the high-reflectance stopband showed some deviation from the simulated spectra This is because the layer thicknesses in the real sample are not exactly the required thicknesses (λ/4n) These deviations were un-avoidable due to the limitation in the very precise thickness control during the crystal growth and etching A tiny variation in the thickness of one

of the layers would cause the interference fringes to change outside the DBR stopband Fortunately, this deviation does not play any role in the operation of the SESAM because it is outside the working wavelength range of the SESAM In fact, the great agreement between the experimental and the simulation results within the DBR stopband wavelength region indicates that our thickness controls during the growth and etching processes are rather good

5.3 Characterizations

property of the SESAM sample during and after the optimization of the structure were analyzed by the characterization techniques of AFM, XRD, SEM and PL to examine the effects of the fabrication processes on the characteristics of the SESAM

5.3.1 Surface morphology

The surface morphology of the SESAM sample after each optimization step was characterized using a Digital Instruments Nano-scope III AFM under tapping mode

Figure 5.5 AFM images (5 μm × 5 μm) of the surface profiles of (a) the eight-period quantum well (8-QW) saturable absorber (taken from the GaN cap surface), (b) the SESAM after laser lift-off of the sapphire substrate (taken from the exposed GaN buffer surface), (c) the SESAM after the ICP etching of the GaN buffer (taken from the etched GaN surface), and (d) the SESAM eventually with the SiO2

AR coating on the etched GaN layer (taken from the AR coating surface) The lighter colors represent higher features, while the darker colors represent lower features