A new coherent soft X-ray branchline at the Advanced Light Source facility has been designed and is near completion. Using the third harmonic from the U8 undulator, this branch will operate from 200 eV to 1000 eV. This increased energy range will allow us to bring interferometric techniques previously used in the extreme ultraviolet region to the soft X-ray region.
Coherent radiation from undulator beamlines has been used to directly measure the index of refraction of several metals [1,2]. We have used this same interferometric technique to measure the indices of refraction of silicon, ruthenium, and TaSiN, materials important for EUV lithography. We now wish to use this technique to directly measure the refractive index of materials with absorption edges in the 500-800 eV range.
Also, we have begun work to use an at-wavelength phase shifting point diffraction interferometer (PS/PDI) to measure the quality of zone plates at soft X-ray wavelengths (below the oxygen K edge of 543.1 eV). The results of this experiment will aid in the evaluation of zone plates currently used in soft X-ray microscopy
The soft X-ray microscope XM-1 [1] at Lawrence Berkeley National Laboratory (LBNL) is a full-field transmission microscope (Figure 1). It has the unique capability to image wet samples as thick as 10 µm in air, while it provides elemental and chemical contrast over a 10-µm-in-diameter object field. All of these features are accompanied by very high resolution (20 nm), which is made possible by the fine features of the micro zone plate (MZP) [2]. XM-1 has been found to be very useful in biological and material studies. Recently, material scientists have taken advantage of XM-1's special capabilities to investigate the magnetic properties of different materials (e.g., Co and Fe).
My research is to measure and improve the resolution of XM-1, which depends in part on the performances of the MZP. Several test objects containing line and space patterns (grating) with different periods and duty cycles were utilized to measure the contrast as a function of spatial frequencies. Recently, we employed multilayers, a structure of two interleaving material layers, for resolution measurement because of their precisely controlled structure quality. Figure 2 shows the X-ray image of a 25 nm Si/25 nm Cr multilayer imaged at 600 eV. These measurements were compared with the simulation results obtained from a program called SPLAT, which was developed by Professor Andrew Neureuther's group in our department. One way to enhance the resolution of the microscope is to reduce the outermost zone width and increase the absorption of light by the opaque zones using thicker zones, which demand a high aspect ratio. A bilayer resist process in which the zone pattern was written into a high-resolution resist layer and then transferred to a thick hardbaked polymer by ICP plasma etching, has successfully yielded 5:1-aspect-ratio MZPs with the smallest zone width of 35 nm [3]. Currently, an effort to decrease the zone width is being undertaken.
Figure 1: The layout of the soft X-ray microscope XM-1 at beamline 6.1.2 of LBNL's Advanced Light Source (ALS) synchrotron radiation facility. Bending magnet radiation from the ALS is used to illuminate the sample.
Figure 2: The soft X-ray image of 25 nm Cr/25 nm Si multilayer cross-section at 600 eV.
High aluminum (Al) content AlxGa1-xAs is known to be controllably oxidizable under elevated temperature in the presence of water vapor. Oxidized AlxGa1-xAs exhibits properties that make it desirable for device applications, such as being highly selective on aluminum composition, being electrically insulating, and having a low refractive index. Such characteristics make AlOx a desirable material choice for making efficient grating structures.
We proposed a novel buried AlOx/semiconductor horizontal cavity grating structure. Defect-free high Al content AlxGa1-xAs grown on nonplanar substrates with 0.9 < x < 1 was achieved using low-pressure MOCVD [1]. After oxidation, we obtained a saw-tooth pattern oxide front with a period half of the original lithography defined pitch. An RCE (resonant cavity enhanced) photodetector structure was designed and fabricated using this AlOx grating as its DBR mirrors. This structure can find applications for making microcavity single-wavelength lasers and optoelectronic devices.
Figure 1: SEM cross-sectional views of four different samples, each with four pairs of GaAs/AlGaAs. Darker regions are AlGaAs layers. Samples (a) and (b) have trenches along [01-1] directions, but were grown at two different temperatures, 600°C for (a) and 640°C for (b). The thickness enhancement ratios between the sidewall and the planar region for samples (a), (b), and (d) are 2.20, 1.32, and 1.17, respectively. Sample (c) shows the convergence of two sidewalls. Sample (d) has the trench aligned along [011] direction. The no-growth plane (111)B is circled in (d), too. The nominal aluminum for all these samples is 0.92.
Figure 2: Oxide front pattern of AlxGa1-xAs on nonplanar substrates with trenches aligned in the [01-1] direction. (a) A schematic shows the nonplanar substrate . (b) Oxide front pattern bright field image. (c) Oxide front pattern top view from the SEM with the backscattered detector.
A controllable variable optical buffer is one of the most critically sought after components in optical communications and signal processing. In such a buffer, optical data would be kept in optical format throughout the storage time without being converted into electronic format. The buffer must be able to turn on to store and off to release optical data at a very rapid rate by an external command. It is generally believed that an optical buffer would enable many new optical systems such as packet-switched networks [1-4], optical signal processing, phase-arrayed antennas, and nonlinear optics.
An ideal optical buffer needs to be compact, have low-power consumption, have variable buffering time, and most importantly, provide enough buffering capacity and bandwidth. We propose the first semiconductor all-optical buffer [5] based on the electromagnetically-induced transparency (EIT) effect [6] in QDs (quantum dots). We establish the conditions and formulation necessary to achieve a large slow-down factor. The light pulses can slow down significantly with negligible dispersion, making it desirable for making optical buffers with an adjustable storage. Narrow linewidth QD fabrication is found to be critical to the overall device performance. Figure 1 shows the schematics of the device and Figure 2 shows the simulation results of signal packet buffering.
Figure 1: Schematics of a semiconductor all-optical buffer based on quantum dot waveguide structures. Signal and pump copropagate (or counterpropagate) in the same waveguide. Signal slow down is induced by the pump via electromagnetically-induced transparency (EIT).
Figure 2: (a) Signal propagation through a quantum dot waveguide with the length of 1 cm. (b) Control of the slow-down factor with the pump power density.
We have developed a model [1] for thermal oxidation of AlGaAs using the continuity equation based on the principle of oxidant mass conservation (Figure 1). Oxidant transport is found to be related to the structure-dependent oxidation process (Figure 2). The model takes into account several processing control parameters all in one single equation and therefore can be easily applied to the process control of device fabrication. The relevant control parameters in the device processing include layer thickness, oxidation temperature, oxidation time, spacing between two devices, Al composition, and mesa geometry. Theoretical calculations agree well with reported experimental values. We also apply this model to study the batch process control of VCSEL fabrication. It is found that, in the order of importance, the following parameters will contribute to the fluctuation in the final aperture size: (1) Al composition (can be completely eliminated if AlAs is used); (2) oxidation temperature; (3) initial mesa size; (4) oxidation time; (5) AlGaAs layer thickness; and (6) spacing between two devices.
Figure 1: Master equation used to model the thermal oxidation of AlGaAs. Rho is the oxidant concentration, D is the diffusion coefficient, and v results from the oxidant transport blockade via several possible mechanisms including the following: the pressure difference built up by oxidation reaction by-product out-diffusion or external forces such as surface tension at the oxide/AlGaAs interface, the internal stress in the oxide, and the oxidant diffusion path termination due to formation of porous AlOxHy and the stuffing of pores in the oxide with As containing reaction by-products.
Figure 2: Schematics of thermal oxidation of AlGaAs. (a) Cross-sectional view. (b) Top view for lateral oxidation in a straight mesa configuration. (c) Top view lateral oxidation in a circular mesa configuration.
A polymer wafer bonding and laser lift-off process has been developed and implemented to integrate GaN heterostructures and functional devices with dissimilar substrates. The integration process takes advantage of the high quality of GaN growth on sapphire and the complementary functionality of the receptor substrates. The process developed in this study uses an SU-8 photo-polymer both as a bonding material and as a delamination material for layer transfer. SU-8 was spun onto an optically transparent handle wafer (e.g., quartz or glass) and then bonded to the GaN-on-sapphire donor wafer. The pre-exposure bake (PEB) temperature has an effect on the Tg or the maximum working temperature of the resist. In this case, 200-300 C PEB was chosen to achieve good SU-8 bonding above 160 C. UV curing of the SU-8 through the backside of the handle wafer could be performed after PEB to yield higher bonding strength at low temperatures. However, the bonding strength of UV-cured samples degraded very quickly at temperatures above 170-200 C. This was expected to be a result of crosslinking, which prevented the relaxation of the resist from residue internal thermal stresses. Laser lift-off of the GaN from the sapphire substrate was then performed with a 38 ns-KrF laser pulse (248 nm wavelength) transmitted through the backside of the sapphire with a fluence of 500-600 mJ/cm2, yielding a GaN/SU-8/silica glass final structure. For the double-transfer experiments, the GaN/SU-8/silica glass structure obtained was then bonded to a Si substrate using the Pd/In transient-liquid-phase method. Laser ablation or heat treatment above Tg of the UV-uncured resist was used to separate the GaN film from the silica-glass substrate.
Recent developments in layer transfer processes for multi-material integration relies greatly on the “paste-and-cut” approach. Exemplified in both thermal ion-cut [1], mechanical ion-cut [2], and ELTRAN®[3] processes for SOI wafers, as well as laser liftoff transfer of GaN onto Si [4], the “paste-and-cut” approach consists of (1) the bonding of two materials' units and (2) separating the bonded system along a prescribed layer other than the bonded interface. Our work investigates a low temperature layer transfer process using a mechanical ion-cut process consisting of direct wafer bonding (Si/Si and Si/SiO2) and crack-initiated separation at the ion implantation region. Experimentally measured strengths of bonded Si/Si and Si/SiO2 interfaces, and of the hydrogen-implanted cut layer indicate that layer transfer occurs when the strength of bonding interface is greater than that of the implanted layer. In summary, the mechanical ion-cut renders three modes of separation depending on relative interface strengths. The schematics illustration of the three modes and corresponding images for the Si/SiO2 system are shown in Figure 1. The process temperature for partial mechanical transfer starts as low as 105°C for Si/Si and 170°C for Si/SiO2, compared to 400-600°C for the conventional thermal ion cut.
Figure 1: Schematic illustration of (a) three separation modes for an implanted Si-SiO2 pair and (b) images of corresponding separation surfaces.
We study plasma instabilities in inductive discharges using SF6 and Ar/SF6 gases. Instabilities occur in transition from capacitive to inductive mode of the discharge [1]. The capacitive coupling plays a crucial role in the instability process. A variable electrostatic (Faraday) shield has been used to control the capacitive coupling from the excitation coil to the plasma. An increase of the shielded area reduces the capacitive coupling and leads to the reduction of stable capacitive and unstable regions of the discharge. The plasma instability disappears when the shielded area exceeds 65 percent of the total area of the coil. A global model of instability gives a slightly higher value of 85 percent for instability suppression with the same discharge conditions (Ar/SF6 1:1, 5 mTorr). The variations of plasma parameters have been explored for Ar/SF6 gas pressures between 5 and 40 mTorr. Space and time variations of plasma density have been observed at higher pressures during the instability. The instability decays toward the center of the chamber more rapidly as the gas pressure increases. Weaker plasma instabilities have been also observed in pure inductive discharges for an Ar/SF6 gas pressure above 20 mTorr. This weaker instability, not related to the E-H transition, is probably an attachment instability. Measurements of the chemical composition of the plasma have allowed improvements in the instability model.
As packing density increases in the fabrication of semiconductors, the fine control of dry etching is strongly required for transferring the fine patterns from weak and thin photo resist. The dry etching is well known to be controlled by ion energy, ion flux, radical flux, and radical species. These control parameters are strongly related to the electron temperature, electron density, and sheath potential of the bulk plasma. In our research, independent control of electron temperature, electron density, and sheath potential of the bulk plasma will be demonstrated through simulations of a circuit model and a global model for the CW and pulse modulated dual frequency capacitively coupled plasma (CCP). This research is focused on the roles of frequency, electrode gap, shape of pulse, chemistry, etc.
1Postdoctoral Researcher
In this research, we are developing a model to investigate the properties of electronegative plasma driven by time modulated power. We have started with a study on the O2 plasma to develop a general model for electronegative discharges. The calculations are done utilizing a global model of a cylindrical plasma discharge [1,2]. This model assumes uniform spatial distribution of plasma parameters over the volume of bulk plasma, with the plasma density in the bulk dropping sharply to edge values at the walls. The global model consists of particle balance and power balance equations. In the particle balance equations, the reaction rates of O2+, O- ions, and electrons have been included for simplicity. The O+ ions and metastable O atoms will also be included later. We are developing a simulation code that can be used in the simulation with variable power modulation periods, duty ratios, and modulation pulse shapes.
As a preliminary examination of the simulation codes, calculations for a sinusoidal modulation of power with a duty ratio of 50% have been performed. The pulse period was varied from 1 microsecond to 1 millisecond. In the shorter modulation period, we observed that the electron temperature falls rapidly during the off cycle whereas the plasma density does not. And the time-average plasma density and the time-average electron temperature showed dependency on the modulation period.
We are working on the optimization of the simulation codes to include more reaction chemistry and are planning to do more simulation with various duty ratios or modulation shapes. Later we will apply the results from this research to the modeling of dual frequency discharge.
Figure 1: Plasma densities averaged over one period of the modulated power with various modulation periods.
The dotted line represents plasma density of the CW driven plasma.
Figure 2: Electron temperatures averaged over one period of the modulated power with various modulation periods.
The dotted line represents electron temperature of the CW driven plasma.
Dual frequency discharges give independent control of ion bombarding energy and plasma density. A high frequency RF source, between 20-80 MHz, controls the plasma density, while a low frequency RF source, between 1-20 MHz, sets the sheat bias and ultimately the ion impact energy. Particle-in-cell (PIC) simulations will be used to characterize the performance of such devices, including kinetic and nonlinear effects. Collisions with background gas are modeled with a Monte Carlo collision (MCC) model, including electron-neutral scattering, excitiation, ionization, and ion-neutral scattering and charge exchange.
One dimensional cylindrical simulations run in a matter of hours, and allow asymmetric electrodes. Preliminary PIC-MCC results compare favorably with global models. Two dimensional simulations are planned to correctly model the grounded third electrode (wall) of realistic ion etch reactors.
1Graduate Student (non-EECS)
Magnetized DC discharges are used in plasma processing for sputtering deposition. Enhanced confinement of electrons is achieved using permanent magnets for magnetron configurations. While such processes are commonly used, the physics are complex and substantial deviations from analytic models are often seen, when such models are even available.
Using a kinetic description of plasmas including collisions with neutrals, particle-in-cell (PIC) methods will be used to accurately model DC discharges, both magnetized and unmagentized. Insight will be gained into the ion energy and angular distributions at the cathode, important to the sputtering deposition process.
Figure 1: Depiction of a DC discharge