The development of waveguides and photonic crystal(PhC) waveguides has been a topic of intense research effortduring the past few years.1–6 This is due to the broad application range from nano and micro-opto-electro-mechanicalsystems to classical and quantum telecommunications,thanks to the growing demand for more efficient and fastercommunication tools offering a reduced footprint. Most ofconventional waveguides are fabricated on Si1 using the socalled silicon-on-insulator (SOI) technology, which is wellestablished and presents a remarkable mechanical stability.However, it is intrinsically limited to passive devices owingto the indirect bandgap of Si. Waveguides based on directbandgap III-V semiconductors such as GaAs2 or InP3,4exhibit attractive features but at the expense of significantgrowth and fabrication effort.7 In addition, InP presents acouple of issues like fragility, high temperature sensitivity,and relatively high cost. Thanks to their wide-bandgap, IIInitrides are potentially suitable for operation around 1.5 lm,although most of the past research focused on the visiblespectral range. They offer low free carrier absorption8 andnegligible two-photon absorption.9 They also exhibit a considerable mechanical hardness and a high thermal stability ofthe refractive index,10 which is one order of magnitude largerthan that of InP.5,6 Furthermore, III-nitride epilayers have already been grown on silicon substrates with device quality,as shown by the fabrication of high-efficiency blue lightemitting diodes.11 Actually, GaN on silicon may offer aunique template for integrated photonics.
In addition, membranes are highly desirable in order toguarantee a refractive index contrast as high as possiblebetween GaN and its cladding-like due to its low refractiveindex (!2.3 at 1.5 lm).14 It is not required for SOI-basedwaveguides since the lower refractive index of SiO2 compared to Si provides enough contrast to confine the lightwithin the slab. The fabrication of freestanding III-nitridephotonic structures by means of sacrificial layers has beenreported by several groups using doping-selective andbandgap-selective photoelectrochemical etching and withdifferent III-nitride sacrificial layers such as InGaN,15 AlN,16and InAlN.17 The main disadvantage of those approaches isthe limited airgap thickness that can be achieved (only a fewhundreds of nanometers). This might be particularly criticalfor structures operating in the near infrared.
Another approach consists in substrate underetching. Todo so, heteroepitaxy is carried out on foreign substrates,which usually exhibit a large lattice-mismatch with GaN.This commonly leads to a high dislocation density (e.g.,>109 cm"2 for GaN grown on Si).18 The most common substrates used in III-nitride technology are SiC, Si, and sapphire. When considering substrate under-etching techniques,SiC and Si are excellent candidates since sapphire etchingremains very challenging.19,20 SiC offers a lower lattice mismatch with GaN (e.g., 3.5% for 6H-SiC (0001)) compared toSi (17% for Si (111)), but its cost is considerably larger. Inaddition, using Si is crucial for the forthcoming integrationof photonics and electronics by taking advantage of both itslow cost and well-established technology and the uniqueoptoelectronic properties of III-nitrides. Indeed, in recentyears, many efforts have been devoted to the integration ofIII-V semiconductors on Si platforms opening perspectivesin silicon photonics (see Refs. 12, 13, and referencestherein). For instance, Pernice et al.21 recently reportedultra-high quality factors (Q) in passive nanobeam cavitiesusing AlN-on-SOI.
In this work, we report on the achievement of IIInitride-based freestanding waveguides and PhC cavitiesoperating in the near infrared with Q factors of several thousands. These photonic structures are fabricated from GaNepilayers by underetching the Si (111) substrate.
l microscope images of the actual structures are displayed.In Fig. 2(a), a L3 cavity together with the W1 waveguide isshown. One can observe the high regularity of the PhC lattice. The smooth and vertical profiles of the holes sidewallsare shown in the inset of Fig. 2(a). Note that our processallowed us achieving a high verticality of the sidewalls(< 5%) with an aspect ratio of 3.24 The pink-green contrast inthe optical image (Fig. 2(b)) delimits the undercut areas(pink-like) and the 3 lm thick airgap is clearly evidenced inFig. 2(c). Such large airgap confirms the high selectivitybetween III-nitrides and Si (111), which makes this fabrication process promising for the integration of III-nitrides onSi platforms.
Fig1
The W1 waveguide was used to couple light into the L3(lateral coupling) and DH (in-line coupling with a barrier often periods (10a)) cavities. The emission peak of the DH(L3) cavity is centered at !1611 nm (!1598 nm) and has alinewidth of !300 pm (!730 pm).28 Note that further adjustment of the wavelength to perfectly match the telecommunication window is straightforward by adapting the PhClattice. The experimental Q factors are about equal to 2200and 5400 for the L3 and DH cavities, respectively. The corresponding theoretical values, estimated by finite elementsimulations, are !2900 and !50 000, respectively. Giventhat all the structures were processed simultaneously, thelarger difference between the theoretical and experimentalvalues in the case of the DH cavities can be ascribed to twomain factors. First, an infinite barrier was considered in thetheoretical calculations instead of a finite one in the actualstructure (10a). Second, a larger sensitivity of such cavitiesto hole position-accuracy is expected compared with its L3counterpart. In the case of L3 cavities, the fact that both experimental and theoretical Q factor values are comparable isattributed to the low-defect fabrication process and to the relative long coupling distance between the W1 waveguide andthe cavity (8 rows in this case). Indeed, a similar cavityplaced only 6 rows away from the W1 waveguide exhibits aQ factor of !1900.
