Rapid thermal processing chamber for in-situ x-ray diffraction

Rapid thermal processing (RTP) is widely used for the processing of photovoltaics (PV), electronics, and energy storage materials.^1–4 RTP involves fast heating ramp rates, generally in the range of 10-100 °C s⁻¹ and processing times of seconds to minutes.⁵ Compared to conventional thermal processes (e.g., belt-furnace), advantages include low thermal cost, high throughput, and better access to metastable phases. So far, RTP process optimization has evolved empirically based on the microstructure and properties obtained after repeated process iterations. Thus, the reaction pathways and accompanying phase transformations during RTP are poorly understood, primarily due to a lack of in-situ characterization on a time-scale characteristic of the fast heating rates. More awareness of the reaction pathways would better enable process optimization and foster increased adoption of RTP.

RTP is also ideally suited for synthesis of metastable phases. Formation of a specific material phase depends strongly on the processing conditions, specifically the heating and cooling rates.^6,7 Rapid addition or removal of energy during synthesis can result in kinetic constraints, allowing access to metastability. This is potentially valuable since metastable phases often have different and useful properties compared to equilibrium phases, and this can be beneficial for certain applications.⁸ Since the phase evolution depends strongly on the process conditions, understanding the relationship between key process conditions and the reaction pathway is paramount.^9–11 Therefore, an in-situ RTP facility for the characterization of the reaction pathways as a function of processing parameters (e.g., ramp rate, processing temperature, and precursor) is particularly useful here.

Up to now, there has been no capability for in-situ characterization during RTP with heating rates as high as 100 °C s⁻¹, as commonly applied in industry practice, and previous technologies have not allowed simultaneous recording of diffraction data with time resolution of the order of 100 ms. Beamline X20C at the National Synchrotron Light Source (not presently operational) was equipped with a RTP tool, which could heat at <35 °C s⁻¹ up to a maximum temperature of 1100 °C.¹² This instrument has proven particularly useful to gain important insights to better understand silicide formation for electronic applications. However, the chamber volume is large and cannot be used to process materials with volatile components, such as CuInGa(S,Se)₂ (CIGS) and Cu₂ZnSn(S,Se)₄ (CZTS). In addition, the energy dispersive diffraction (EDDI) beamline at BESSY II has a facility for in-situ characterization during thermal processing with ramp rates up to 20 °C s⁻¹ and with time resolution only of the order of 1 s.¹³ This has been used to study the formation of CuInSe₂ and CuGaSe₂ films for PV.

Here, we report the design and development of a RTP instrument for in-situ time resolved X-ray diffraction (XRD) experiments with greater capabilities than have been achieved previously. We demonstrate in-situ processing at rates as high as 100 °C s⁻¹ up to 1200 °C, while simultaneously recording diffraction patterns at time intervals of 125 ms. In-situ XRD measurements on a Bi₂O₃-SiO₂ glass frit reveal a metastable phase progression with clear dependence on select processing parameters.

In industry processing, heating rates in RTP can be of the order of 100 °C s⁻¹. Therefore, the in-situ characterization apparatus must provide this ability to enable relevant studies. The apparatus must also enable a variable sample orientation relative to the incident X-ray beam, since this allows variation of the beam footprint and the extent of X-ray absorption on the sample. Achieving a uniform temperature distribution over the sample area and avoiding hot-spots are vital for collecting representative diffraction data. In addition, some PV materials require specific processing atmospheres, for instance, CIGS or CZTS requires some positive S/Se vapor pressure. Therefore, the RTP is designed with the following considerations:

1. temperature ramp rate up to 100 °C s⁻¹;

2. temperature uniformity over the sample area of 20 × 20 mm²;

3. processing under inert and/or reactive atmosphere;

4. robustness and flexibility to vary the incident X-ray angle from 0 to 30°;

5. maximum temperature up to 1200 °C;

6. convenient and fast sample changing ability.

Figure 1 shows the in-situ RTP design concept schematically. The RTP reactor consists of a metallic body (front cap assembly, central block, and end cap assembly) that houses a quartz reaction chamber. The sample is secured inside the reaction chamber on a sample holder connected to the front-cap assembly. The sample holder is made up of two quartz rods (3 mm in diameter) having small projections on the top to guide the sample placement, while minimizing conductive heat transfer. Quartz does not block the heat radiation reaching the sample platform/sample. A Si wafer (375 μm, 25 × 25 mm²) platform is placed on the sample holder, and the sample of interest is placed over the sample platform. A thermocouple permanently bonded with the sample platform was used to monitor temperature; a schematic of this configuration is shown in Figure 1(b). The thermocouple was bonded to the wafer using a ceramic bonding agent under applied pressure and heat (Thermoelectric Company, West Chester, PA, supplied the thermocouple bonded Si wafer). Both ends of the quartz chamber are secured and sealed at the front-cap and end-cap flanges using o-rings. A controlled atmosphere inside the chamber is achieved via gas inlet and exit ports in the end-caps. Two pairs of tungsten halogen lamps configured above and below the sample and perpendicular to the reaction chamber axis provide rapid heating. This configuration is used because only the middle one-third of the 120 mm long lamp emits uniform radiation. Orienting the lamp axes parallel to the quartz tube would have required the use of a quartz chamber with comparable length, which would have restricted the XRD collection angle to less than 30°. Installing the lamps perpendicular to the quartz chamber axis allowed for a shorter chamber length, which in turn minimizes the distance between the sample and the exit X-ray window (30 mm). A large solid angle of collection, 42°, is achieved in this fashion.

The compact RTP body is made of an aluminum alloy. Aluminum exhibits good reflectivity for IR radiation (∼0.98). However, at elevated temperatures, aluminum surfaces readily oxidize, which lowers reflectivity, and results in heat absorption. Thus, the aluminum walls are gold-coated to a thickness of ∼7 μm. The chamber block is plumbed to allow water-cooling via a chiller to keep it relatively cool even after repeated high temperature experiments. The RTP chamber was mounted on the beamline diffractometer (HUBER Diffraktionstechnik GmbH & Co. KG, Rimsting, Germany) by means of a coupling attached to the bottom of the chamber. The Huber coupling consists of a thread ring (Type 1001-1) and a base (Type 1002-2) as depicted in Figure 2.

X-rays enter the reaction chamber through the rectangular X-ray window on the front-cap (Figure 1). Scattered X-rays are collected through the half-circular exit window with an area detector (Pilatus DECTRIS 300 K). Thin Kapton^® (50 μm) films are used as windows, which minimize X-ray absorption losses. The Kapton^® windows were aluminum-coated (1000 Å) to enhance their reflectivity and minimize heat loss. A low background signal results from parasitic scattering of X-rays by the aluminum coated Kapton windows. To minimize this background, the X-ray entrance window is positioned with an incline of 30° from vertical in the front-cap assembly, and a curtain close to the window is provided, which blocks parasitic X-ray scattering from the window.

Typically, a RTP experiment lasts for several minutes. Therefore, a simple and convenient sample loading and unloading procedure is particularly useful because it enables facile measurement of a series of samples, with systematically varying process parameters. It follows that the front-cap assembly is mounted on sliding guide rails and bearings from Thomson Linear^® (Radford, VA) to ensure fast and smooth sample change. With experience, sample replacement can take as little as 2 min.

In RTP processing, temperature uniformity over the sample area is achieved by applying uniform and continuous radiative heat from lamps. The radiative heat transfer between two surfaces depends on their relative orientation, radiative properties, and temperature. The view factor is defined as the fraction of radiation emitted or reflected from a surface that reaches another surface and considers these variables. The view factor has been calculated for the RTP apparatus to estimate the heat transfer between the lamps and the sample surface for lamp separation distances of 25 and 34 mm using a commercial MathCAD^® package (Figure 3).

This calculated heat distribution was compared with the experimentally observed light intensity values obtained using fiber optic cable mounted in an array over a 25 × 25 mm² graphite section. The normalized intensity measured at nine locations varied from 0.92 to 1, as shown in Figure 4, which agrees well with the simulation results. This shows that good temperature uniformity can be achieved over a sample area of 20 × 20 mm² with 34 mm lamps separation.

Accurate temperature monitoring is crucial to enable efficient temperature control, especially when heating at rates as high as 100 °C s⁻¹. Fine gauge thermocouples are well suited here due to their fast response time. The in-situ RTP chamber is equipped with fine gauge Microtemp K-type thermocouples (wire size 0.125 mm) from Omega Engineering Inc. (Stamford, CT). The thermocouple response time in still water at 38 °C is 0.04 s, which is fast enough for the current setup. Two thermocouples are used to monitor the temperature during experiments—one pressed onto the top of the sample and another permanently bonded to the Si sample platform. The thermocouple bonded with the Si sample platform is designated for feedback control, while the sample thermocouple measurements are indicative of the temperature at the top surface of sample.

Temperature control during rapid heating requires a fast and efficient feedback control system. The high frequency control system installed for regulating temperature during rapid heating enables lamp power output regulation on a 10 s of milliseconds timescale. The temperature control system consists of a 16-channel thermocouple reader, eight analog input channels for the photodiodes (to monitor lamp intensities), four analog outputs for controlling four individual lamps, and two power supplies (1500 W each). A LabVIEW^® program reads the temperature through the thermocouple and regulates the power output of the lamps via a proportional integral differential (PID) control program to achieve a desired temperature profile. A temperature profile can have several segments of varying ramp rates and hold times. In the PID algorithm, the temperature and power supply control signals update every 75 ms. The program compares the measured temperature and the pre-set temperature, and then the PID control module adjusts the power output accordingly. The PID controller interface allows manual PID value assignment, which can be customized for different temperature ranges, profiles, and thermal loads.

For in-situ experiments, the PID temperature control program was integrated with the beamline control software (SPEC) to synchronize data acquisition with RTP. This was accomplished by coupling the diffraction data collection command within SPEC to the RTP temperature controller. At the start of diffraction data acquisition, transistor-transistor logic (TTL) signal of 5 V is sent from SPEC, which triggers the temperature controller to energize the RTP. Diffraction data are collected continuously at a defined time resolution for specified length of time, which can be defined at the start of every run depending on the ramp rate and maximum temperature.

During heating at rates as high as 100 °C s⁻¹, the sample temperature may differ significantly from the value measured by the thermocouples, which are indicative of the sample surface temperature and the Si sample platform temperature. This is due to the combination of thermal mass differences, relative optical absorption and transient heat effects, heat loss through thermocouple wires, and contact resistance between the thermocouple hot junction and sample.¹⁴ As a result, the thermocouples may not accurately represent the sample temperature at a given instant. Therefore, an internal calibration standard is required to quantify and correct this error.

A silver film was chosen as an internal standard, and the temperature dependence of the lattice parameter, which can be measured in-situ during heating, was used to determine the sample temperature. A silver film (2000 Å) on a silicon substrate was used for the calibration procedure. At faster heating rates, the difference in temperature recorded by the thermocouple and actual sample temperature is larger. Therefore, the maximum ramp rate, 100 °C s⁻¹, was chosen for monitoring the difference in temperature of the sample and the thermocouple. Diffraction data were collected every 125 ms with a detector exposure time of 100 ms. The time-resolved XRD data obtained from the Ag film are shown in Figure 5. Silver XRD peaks undergo changes in both width and position in response to heating and cooling in the RTP. Initially, XRD peaks are broad and spread over more than 10 pixels, indicating fine grain microstructure. On heating, the XRD peaks become sharper due to grain growth, and shift to lower Q due to lattice expansion. The peak position reveals the temperature.

Figure 6 shows a comparison between the recorded temperature and the temperature determined from the silver lattice parameter.¹⁵ The accuracy in estimating the temperature in this fashion depends on the certainty in determining the XRD peak position and is smaller when the peak is spread over multiple pixels on the detector. In the initial stages, when the grain sizes of Ag film were finer; XRD peaks spread over more than 10 pixels and the error bars were smaller. However, on heating, the XRD peaks become sharper, due to grain growth, which results in larger error. Nevertheless, the calibration demonstrates close agreement between thermocouple reading and the temperature determined from the lattice parameter at a ramp rate of 100 °C s⁻¹, with a maximum error of ±30 °C. The error in calibration can be further improved by increasing the sample-detector distance.

To test the RTP, we have used a Bi₂O₃-SiO₂ glass frit, which is an alternative to conventional lead-oxide based additives to screen-printed silver-pastes for making contacts to Si solar cells.^16,17 This serves as an interesting test system to demonstrate the in-situ RTP capabilities and to assess the dependence of reaction pathways on processing conditions. The amorphous frit contains about 60 mol. % Bi₂O₃, 30 mol. % SiO₂ and 10 mol. % B₂O₃. Ag nanoparticles were mixed with the frit, as is done in industry, and serve as an effective internal standard to determine the sample temperature. In-situ processing of Bi₂O₃-SiO₂ frit was carried out in air at two different heating ramp rates: 5 and 100 °C s⁻¹.

XRD patterns at varying temperatures are shown in Figure 7. The amorphous Bi₂O₃-SiO₂ frit exhibits several metastable crystalline phases during firing before melting at around 825 °C (close to the maximum firing temperature of 850 °C). During heating at a ramp rate of 100 °C s⁻¹, the Bi₂O₃ first crystallizes at around 400 °C to a metastable δ phase, which is a solid solution of SiO₂ in Bi₂O₃ (Note that Bi₂O₃ can dissolve close to 25 mol. % SiO₂ at around 600 °C in metastable state.¹⁸). On heating to 430 °C, another metastable compound Bi₂SiO₅ forms. At around 550 °C, all the δ-phase converts to Bi₂SiO₅, which eventually melts at around 825 °C. On the other hand, on heating at the rate of 5 °C s⁻¹, an equilibrium phase Bi₁₂SiO₂₀ appears at around 750 °C along with a predominant Bi₂SiO₅, while the system melts at slightly higher temperature close to 840 °C.

A schematic of the phase progression observed at different heating ramp rates is depicted in Figure 8. Although, the melting point of pure Bi₂SiO₅ phase is about 845 °C, in the presence of δ phase, this can melt at lower temperatures. It is important to note that during heating at rates of 100 °C s⁻¹, only metastable phases appear before melting. On the other hand, during slower heating ramp rate of 5 °C s⁻¹, a thermodynamically stable Bi₁₂SiO₂₀ phase is also observed, which generally melts congruently at about 900 °C.^18,19

These results indicate that the formation of stable phases is kinetically hindered during fast heating ramps, 100 °C s⁻¹ in this case. Conversely, when heating slowly (5 °C s⁻¹), the equilibrium Bi₁₂SiO₂₀ phase formed, and a consequent increase in the melting temperature is observed. This shows that the phase formation sequence during firing of bismuth-oxide based frit materials is sensitive to the heating rate. Furthermore, this simple demonstration illustrates that an in-situ characterization instrument, capable of probing the microstructural evolution of complex material systems under the actual processing conditions used in applications, is indispensable for understanding key reaction pathways.

A compact RTP chamber for in-situ diffraction was designed and fabricated. The RTP instrument enabled in-situ XRD data acquisition during fast thermal processing up to ramp rates of 100 °C s⁻¹, with time resolution of 125 ms. Samples sizes 20 × 20 mm² can be handled effectively, with good temperature uniformity during heating, using an experimental configuration designed to minimize parasitic heat-loss and provide optimal heat transfer. Close agreement between the temperatures measured by thermocouple and that derived from a silver calibration standard was demonstrated. Finally, the measurements on a Bi₂O₃-SiO₂ glass test system revealed the ability to resolve metastable phase evolution during RTP using the new apparatus.

This project is funded through the Bridging Research Interactions through collaborative the Development Grants in Energy (BRIDGE) program under the SunShot initiative of the Department of Energy (DE-EE0005951). In-situ measurements were carried out at the Stanford Synchrotron Radiation Laboratory, a national user facility operated by Stanford University on behalf of the U.S. Department of Energy, Office of Basic Energy Sciences, under Contract No. DE-AC02-76SF00515. We thank Ron Marks for assistance with SSRL beam line 7-2.