Sample information

The experiments were performed on both GaAs and InP samples with embedded quantum dots. The details as well as the fabrication procedure of these samples are described briefly in this chapter.

The growth and fabrication of these samples was no part of my project and will therefore not be discussed in great detail.

3.3.1 Sample growth

The GaAs samples were grown using Molecular Beam Epitaxy (MBE), while the InP samples were grown using MetalOrganic Vapour Phase Epitaxy (MOVPE). The quantum dots were grown using the Stranski-Krastanov growth method. Both samples had low density quantum dots.

MBE growth is performed at Ultra-High Vacuum (UHV) pressures of 10^-8 Pascal or lower.^[38] Under these conditions the elements of the material that you wish to grow are heated so they evaporate and travel towards the surface. Due to the UHV the distance between the sources and the substrate is shorter than the mean-free path of the particles. Because of this, the particles do not interact with each other and thus the particles will not reach equilibrium until they interact with the surface. On the surface the particles will redistribute over the surface, to find the lowest energy configuration to reach equilibrium.^[39]

MOVPE growth is performed at higher pressures. Gas compounds with the elements which need to be grown are released inside the growth chamber. At the substrate pyrolysis of these gas compounds takes place, causing the elements to be deposited on the substrate while the remaining gas is released from the chamber.^[39]

The Stranski-Kranstanov growth method is based on the different lattice constant for the quantum dot material and the surrounding material. Due to this mismatch strain will build up at the surface between both the materials. Therefore it is energetically favourable to keep this interface area as small as possible. However, the surface of the material is also the cause for an energy increase due to the dangling bonds. This effect favours a large interface area between the two materials since this causes a smaller surface area. The trade-off between these two effects causes the arising of 3D structures at a certain critical thickness.^[40] Apart from the growth of quantum dots, a thin layer on the surface will also be formed. This is called the wetting layer and gives rise to a large emission at higher energies than the dot emission.

3.3.2 Device fabrication

To embed the photonic crystal structure in the grown slabs, several fabrication steps were needed.

These steps will be described here briefly for completeness. The fabrication process for GaAs and InGaAsP were slightly different so they will be described simultaneously here and only when there is a difference this will be highlighted. The fabrication of the structures consisted of four steps which will all be described separately here. The steps are also drawn schematically in Figure 22.

First the SiNx mask is deposited using PECVD. Afterwards a layer of ZEP positive resist was spin-coated on (Figure 22 (a)). This material is resistant to most etching materials, but can be etched using Electronic Beam Litography (EBL). In this way a precise pattern where the air holes should come could be patterned in the ZEP.

31

Next the sample is echted down using Reactive Ion Etching (RIE). This process uses reactive ions to transfer the pattern into the mask and to etch away the remaining photoresist (Figure 22 (c)). The ions used for both samples were CHF3.

After this the holes are transferred to the active layer by using Inductively Coupled Plasma (ICP). This process uses plasma to etch away the material. This is used because it can create the deep straight sidewalls needed for a photonic crystal (Figure 22 (d)). For GaAs a Cl2/N2 plasma was used and for InP a Cl2/Ar/H2 plasma was used. For both cases, the plasma temperature was kept at 200°C. These parameters were optimized to prevent etching of the SiNx mask.

The final step consists of wet etching the sample. With GaAs HF was used while with InP a 4:1 HCL:H20 mixture was used. This process has a high selectivity for the buffer layer while keeping the active layer almost undamaged. Therefore the wet etchant will flow through the holes and create the undercut needed to ensure confinement in the third direction (Figure 22 (e)).

Figure 22: Schematic drawing of the various fabrication steps used to construct the photonic crystal structure on the samples. When there is a difference between the GaAs and InP layers the GaAs layer is shown in grey while the InP layer is shown in green.

3.3.3 Sample characterisation

Both the GaAs sample and the InP had embedded quantum dots in the active layer. To characterise the sample the emission spectrum of these quantum dots have been measured by measuring the PL of an unprocessed part of the sample. The InAs dots in the GaAs sample were constructed in such way that they would emit around 1300nm while the InAs dots in the InGaAsP layer were designed to emit around 1500nm. Both wavelengths are interesting to study for telecommunication purposes.

(a) (b) (c)

(d) (e)

32

1100 1200 1300 1400 1500 1600 1700

0,0 0,2 0,4 0,6 0,8 1,0

Intensity (a.u.)

Wavelength (nm) InP

GaAs

Figure 23: PL spectrum of an unprocessed sample of InAs dots in InP (black curve) and GaAs (red curve) measured using a normal PL setup.

In the emission spectrum of Figure 23 it can be seen that the InAs dots in InP and GaAs emit around a slightly higher wavelength than desired at room temperature. However the inhomogeneous broadening of the peak is large enough to still get sufficient emission at the desired wavelength. This inhomogeneous broadening is caused by the fact that the dots are created randomly and thus do not always have the same size and corresponding emission energy. It can also be seen that the inhomogeneous broadening is much larger for the dots on InP than for the dots on GaAs.

3.3.4 Cavity characterization

To get an idea what kind of parameters gave the best results for the cavity processing a large number of single Noda cavities, without any FP cavity nearby, were fabricated. All the cavities had slightly different parameters. The crystal parameters changed were: Lattice mismatch for the cavity, lattice constant and filling factor. The measurements were performed on 2 different samples and several identical cavities to check reproducibility.

Figure 24 shows a typical spectrum of a cavity mode obtained from one of the single Nodas. It can be seen that the broad emission from the quantum dot is now suppressed except for one peak where the emission is strongly enhanced. This peak corresponds to the resonant wavelength of the cavity.

From the linewidth of the peak the Q-factor of the mode can be determined via where is the Full Width at Half Maximum (FWHM) of the peak.

33

1550 1560 1570 1580

0 1000 2000 3000 4000 5000

Intensity (a.u.)

Wavelength (nm)

Figure 24: Typical spectrum of a single Noda cavity obtained with the micro-PL on InP, showing the Noda peak in the red square, the barrier modes in yellow and the modes from the lower edge of the photonic bandgap in blue.

These spectra have been measured for around 320 structures and the Q-factor of all these spectras have been compared. The results of these measurements are shown in Figure 25. From this figure it can be seen that a decreasing filling factor (Figure 25 (a)) or an increasing lattice constant (Figure 25 (b)) in general increases the Q-factor of a cavity. However, this also increases the wavelength. When we plot the Q-factor versus wavelength for these cavities (Figure 25 (c)) we see a much clearer dependence where an increasing resonant wavelength causes an increase in Q-factor. This can be explained by the fact that the bandgap for this material was around 1.4μm. Therefore close to this bandgap absorption would be strongly enhanced and this would lead to a decrease in Q-factor.

The dependence on the lattice mismatch was also determined but this showed no clear trends for the Q-factor. This only changed the energy difference between the Noda modes and the barrier modes but did not mean higher confinement. This can be explained by the fact that two competing effects are taking place in this case. An increase in lattice mismatch would increase the barrier and therefore decrease the in-plane losses. However, the higher barrier also causes more scattering, thus increasing the out-of-plane losses. Since in the final structure the in-plane losses enable coupling with the FP cavity for this purpose a small lattice mismatch was desirable. However this introduced more stringent demands on the fabrication process.

34

Figure 25: Q-factor dependence on Filling factor (a), lattice constant (b) and resonant wavelength (c) of measured single Noda cavities.

0,26 0,27 0,28 0,29 0,30 0,31 0,32 1000

1500 2000 2500 3000 3500

Q-factor(-)

Filling Factor(-)

420 430 440 450 460

500 1000 1500 2000 2500 3000 3500

Q-factor (-)

Lattice constant (nm)

1440 1460 1480 1500 1520 1540 1560 500

1000 1500 2000 2500 3000 3500

Q-factor(-)

Wavelength(nm)

(a) (b)

(c)

35

4 Experimental results