BACKGROUND OF THE INVENTION
The monolithic integration of several optoelectronics devices in optoelectronics integrated circuits (OEICs) and photonic integrated circuits (PICs) is of considerable interest for the development of telecommunications systems.
In OEICs, optical devices such as lasers and electronic devices such as transistors are integrated on a single chip for high speed operation since parasitic reactance in the electrical connections can be minimized from the closely packed devices.
PICs are a subset of OEICs with no electrical components, in which only photons are involved in the communication or connection between optoelectronics and/or photonic devices. The driving forces for PICs are to improve the complexity of next-generation optical communication links, networking architectures and switching systems, such as in multiple channel wavelength division multiplexing (WDM) and high speed time division multiplexing (TDM) systems. In PICs, besides gaining from the low cost, size reduction, and increased packaging robustness, the main advantage is that all the interconnections between the individual guided-wave optoelectronics devices are precisely and permanently aligned with respect to one another since the waveguides are lithographically produced.
In the integration process, complex devices are built up from components that are very different in functionality such as light emitters, waveguides, modulators and detectors. Each component needs different material structures to achieve optimized performance. As a result, the ability to modify the bandgap energy and the refractive index of materials is important in order to realize OEICs and PICs. A number of techniques have emerged for this purpose, including growth and regrowth, selective area epitaxy or growth on a patterned substrate and quantum well intermixing (QWI).
Growth and regrowth is a complicated and expensive technique which involves growing, etching and regrowing of quantum well (QW) layers at selected areas on bulk material. These layer structures are overgrown with the same upper cladding but a different active region. This approach suffers from mismatches in the optical propagation coefficient and mismatches in the dimensions of the waveguide at the regrown interface. In addition, this process gives low yield and low throughput, and therefore adds cost to the final product.
Selective area growth utilizes differences in epitaxial layer composition and thickness produced by growth through a mask to achieve spatially selective bandgap variation. Prior to epitaxy growth, the substrate is patterned with a dielectric mask such as SiO2, in which slots with different widths are defined. The growth rate in the open areas depends on the width of the opening and the patterning of the mask. No growth can take place on top of the dielectric cap. However, surface migration of the species can take place for some distance across the mask to the nearest opening. The advantage of this approach is a reduction in the total number of processing steps such that essentially optimum laser and modulator multiple quantum well (MQW) sections can be accomplished in a single epitaxial growth stage. This process works well under a precisely controlled set of parameters but is difficult to manipulate in a generic fashion. In addition, this technique gives poor spatial resolution of around 100 μm, and hence the passive section generally has a relatively high loss.
QWI is based on the fact that a QW is an inherently metastable system due to the large concentration gradient of atomic species across the QWs and barriers interface. Hence, this allows the modification of the bandgap of QW structures in selected regions by intermixing the QWs with the barriers to form alloy semiconductors. This technique offers an effective post-growth method for the lateral integration of different bandgaps, refractive index and optical absorption within the same epitaxial layers.
The QWI technique has been gaining recognition and popularity for which several potential applications in integrated optoelectronics have been identified, for example bandgap-tuned electroabsorption modulators, bandgap-tuned lasers, low-loss waveguides for interconnecting components on an OEIC or PIC, integrated extended cavities for line-narrowed lasers, single-frequency distributed Bragg reflector (DBR) lasers, mode-locked lasers, non-absorbing mirrors, gain or phase gratings for distributed feedback (DFB) lasers, superluminescent diodes, polarization insensitive QW modulators and amplifiers and multiple wavelength lasers.
Current research has been focused on QWI using approaches such as impurity free vacancy induced disordering (IFVD), laser induced disordering (LID) and impurity induced disordering (IID). Each of these QWI techniques has its advantages and shortcomings.
The IFVD method involves the deposition of a dielectric capping material on the QW materials and subsequent high temperature annealing to promote the generation of vacancies from the dielectric cap to the QW materials and hence enhance the intermixing at selected areas. For instance, in GaAs-AlGaAs QW materials, SiO2 is known to induce out-diffusion of Ga atoms during annealing, hence generating group III vacancies in the QW material. The thermal stress at the interface between the GaAs and the SiO2 layer plays an important role. The thermal expansion coefficient of GaAs is ten times larger than that of SiO2. During high temperature annealing, the bonding in the highly porous SiO2 layer deposited using plasma-enhanced chemical vapor deposition (PECVD) may be broken due to the stress gradient between the GaAs and SiO2 film. Thus, the out-diffusion of Ga helps to relieve the tensile stress in the GaAs. These Ga vacancies then propagate down to the QW and enhance the interdiffusion rate of Ga and Al, and hence result in QWI. After the intermixing process, the bandgap in the QW material widens and the refractive index decreases.
The selectivity of this technique can be obtained using an SrF2 layer to inhibit the outdiffusion of Ga, hence suppress the QWI process. Using this technique, devices such as multiple wavelength bandgap tuned lasers and multiple channel waveguide photodetectors have been successfully demonstrated.
Although IFVD is a successful technique when employed in GaAs/AlGaAs system, this technique gives poor reproducibility in InGaAs/InGaAsP systems. Furthermore, due to the poor thermal stability of InGaAs/InGaAsP materials, the IFVD process, which requires high temperature annealing, is found to give low bandgap selectivity in InGaAs/InGaAsP based QW structures.
Laser induced disordering (LID) is a promising QWI process to achieve disordering in InGaAs/InGaAsP QW materials due to the poor thermal stability of the materials. In the photoabsorption-induced disordering (PAID) method, a continuous wave (CW) laser irradiation is absorbed in the QW regions, thereby generating heat and causing thermal induced intermixing. Although the resulting material is of high optical and electrical quality, the spatial selectivity of this technique is limited by lateral flow to around 100 μm. A modification of the PAID method, known as pulsed-PAID (P-PAID), uses high-energy Q-switched Nd:YAG laser pulses to irradiate the InP-based material. Absorption of the pulses results in disruption to the lattice and an increase in the density of point defects. These point defects subsequently interdiffuse into the QW during high temperature annealing and hence enhance the QW intermixing rate. Though P-PAID can provide spatial resolution higher than 1.25 μm and direct writing capability, the intermixed materials give low quality due to the formation of extended defects.
Of all the QWI methods, impurity induced disordering (IID) is the only process which requires the introduction of impurities into the QW materials in order to realize the intermixing process. These impurities can be introduced through focused ion beam, furnace-based impurity diffusion and also ion implantation.
IID is a relatively simple and highly reproducible intermixing process. It has the ability to provide high spatial resolution for the integration of small dimension devices and bandgap shifts can be controlled through the implantation parameters. This technique is commonly used to achieve lateral electrical and optical confinement in semiconductors such that low threshold current and single lateral-mode operation can be obtained. Furthermore, the IID process is of considerable interest for the integration of WDM systems, such as multiple wavelength laser sources, low-loss waveguides, modulators and even detectors.
The IID effect is widely accepted to consist of two stages. The first stage is to implant impurities into the QW material. The subsequent stage is to anneal the material to induce diffusion of both impurity and point defects into the QWs and barriers, and hence interdiffusion of matrix elements between QWs and barriers. In an InGaAs/InGaAsP QW system, the interdiffusion of Group V elements from barrier to well, which results in blueshifting of the bandgap energy, are believed to be caused by the diffusion of point defects generated during the implantation process, the self-interdiffusion at elevated temperature (thermal shift), and the diffusion of the implanted species.
During implantation, impurities as well as point defects, such as Group III vacancies and interstitials, are introduced into the material in selected areas. The diffusion of these point defects and impurities at elevated temperature enhances the interdiffusion rate between the QWs and barriers and hence promotes intermixing after annealing. Under the influence of injected impurities, the compositional profile of the QW is altered from a square to a parabolic-like profile. As a result, after the interdiffusion process, the local bandgap increases and the corresponding refractive index decreases.
Using the IID technique, selective area intermixing across a wafer can be obtained by using an SiO2 implant mask with various thicknesses. However, this technique involves multiple lithography and etching steps which complicate the fabrication process.
A paper entitled “Integration process for photonic integrated circuits using plasma damage induced layer intermixing”, Electronics Letters, Volume 31, 449, 1995, by B S Ooi, A C Bryce, and J H Marsh, describes a quantum well intermixing process based on reactive ion bombardment damage. In this technique, a high RF power, and hence high damage, H2 plasma process was used to introduce point defects on the surface of samples which were then annealed to diffuse the point defects into the QW region. The plasma exposure was performed using a parallel plate reactive-ion etching (RIE) machine. Similarly, L M Lam et al, in a paper entitled “Plasma Immersion Ar+Ion Implantation Induced Disorder in Strained InGaAsP Multiple Quantum Wells”, Electronic Letters, Volume 34, No. 8, 16 April 1998, disclose a plasma immersion ion implantation process that uses an RIE machine. In each of these techniques, the QWI process is based on ion bombardment damage and requires multiple cycles to affect fairly modest degrees of bandgap shift.
The ability to control the bandgap across a III-V semiconductor wafer is a key requirement for the fabrication of monolithic PICs. The absorption band edge of QW structures needs to be controlled spatially across a wafer to allow the fabrication of integrated lasers, modulators, and low-loss waveguides. Although QWI techniques offer great advantages over growth and regrowth and selective epitaxial growth techniques for the bandgap engineering process, the spatial control of conventional QWI techniques is indirect and complicated.
The explosive growth of Internet traffic, multimedia services and high-speed data services has exerted pressure on telecommunications carriers to expand the capacity of their networks quickly and cost effectively. Carriers normally have three options to expand capacity, ie install new fibers, increase the bit rate of the transmission system, or employ wavelength division multiplexing (WDM). While the first option has problems of high cost and right-of-way and the second option has limited growth potential because of inherent system limitations, the third option is therefore very attractive because it is capable of manifold increase of the network capacity at a modest cost.
SUMMARY OF THE INVENTION
According to the present invention, a method of manufacturing a photonic integrated circuit comprising a compound semiconductor structure having a quantum well region, in which the method comprises the steps of irradiating the structure using a source of photons to generate defects, the photons having an energy (E) at least that of the displacement energy (ED) of at least one element of the compound semiconductor, and subsequently annealing the structure to promote quantum well intermixing.
The preferred radiation source is a plasma, although there are a number of sources of high energy photons that can be used. Suitable plasma sources include those generated using an electron cyclotron resonance (ECR) system, an inductively coupled plasma (ICP) system, a plasma disk excited by a soft vacuum electron beam, and plasma soft x-ray (SFR) devices. Other suitable sources of high energy radiation include electrical gas discharge devices, excimer lasers, synchrotron devices, flash x-ray devices and gamma ray sources.
The method may include the step of masking a portion of the structure to control the degree of radiation damage. In this manner, the mask may be adapted to prevent intermixing entirely. However, preferably the structure is masked in a differential manner to selectively intermix the structure in a spatially controlled manner by controlling the exposure of portions of the structure in a predetermined manner.
There are a number of suitable forms of exposure masks, including binary masks, phase masks, gray masks, dielectric or metal masks, and photoresist masks. The spatial control of intermixing is advantageously controlled using a variable profile mask pattern. Our co-pending International patent application number (Agent's reference PJF01075WO) describes a method for patterning a structure by exposing a layer of photoresist through a gray scale mask. The degree of quantum well intermixing is controlled in a spatially selective manner in dependence on the optical transmittance characteristics of the gray scale mask. This technique is especially suitable for use in the present invention since it allows a mask to be constructed that can control the exposure of the structure to high energy radiation. The photoresist mask pattern may be used alone to control the degree of exposure or it may instead be used to transfer a mask pattern to an underlying material, such as a layer of dielectric material, through an etching process.
The key feature of the present invention is the use of a radiation source to cause radiation damage to a crystalline structure. To achieve this, a well defined minimum energy transfer is needed. This is called the displacement energy, ED
. Energy transfers exceeding ED
will cause atom displacement, either primary displacement, when a host ion is struck by one of the instant particles, or secondary displacement, when energy transfer is from the host atom previously struck. Energy values of ED
in eV for a range of group III-V materials are given in the table below.
| || |
| || |
| ||GaAs || 9/9.4 |
| ||InP ||6.7/8.7 |
| ||InAs ||6.7/8.3 |
| ||InSb ||5.7/6.6 |
| || |
Whilst it is known that vacuum ultraviolet (VUV) radiation can cause damage to semiconductor structures, this has always previously been investigated on the basis that this damage should be avoided, or at least repaired by annealing to ensure that these defects do not affect the operation of the device.
This novel, low cost, and simple technique can be applied for the fabrication of PICs in general, and WDM sources in particular. By applying a QWI technique in accordance with the present invention, the bandgap energy of a QW material can be tuned to different degrees across a wafer. This enables not only the integration of monolithic multiple-wavelength lasers but further extends to integrate with modulators and couplers on a single chip. This technique can also be applied to ease the fabrication and design process of superluminescent diodes (SLDs) by expanding the gain spectrum to a maximum after epitaxial growth.
The photonic integration research community currently views QWI technology as a promising approach only for two-section photonic devices as conventional QWI processes would otherwise become tedious and complicated. Although it is complex and not cost effective, researchers have instead preferred to use selective area epitaxy for multiple-section integration. The present invention demonstrates that the application of QWI is not limited to two sectional devices. In addition, the technique is more cost effective, and offers a higher throughput and higher yield compared to selective area epitaxy.