Ferrite gelcasting for wide bandgap semiconductors
A University of Nottingham PhD candidate explains how gelcasing can be used for processing magnetic cores in complex shapes and dimensions.
Wide bandgap (WBG) semiconductor technology developments and market trends have a future of power conversion that will lead to very compact high-switching frequency power electronics converters. In the paper IEEE ITRW working group position paper — packaging and integration, author EPSRC Centre for Power Electronics Director, Mark Johnson, classified heterogeneous integration and packaging as enabling technologies to unlock the full potential of WBG devices.
A radical change of design and manufacturing philosophy is needed, from assemblies of discrete components, each designed and packaged separately, to fully integrated assemblies consists of power devices, gate drives, filters, sensing, and control functions. At present, one of the major limitations to this trend is posed by magnetic components, for example transformers and inductors, materials and related manufacturing techniques.
Soft ferrites, manganese-zinc (MnZn) and nickel-zinc (NiZn), are the only class of material that has the potential to operate efficiently at the maximum frequency capability of WBG devices, from 500kHz to a few MHz. Currently, the standard manufacturing process requires high temperature and high pressure, thus increasing the overall process complexity, and inhibiting the design of novel shapes to be used to confine and guide the magnetic flux.
Since gelcasting, contrarily, does not need pressure, it can be used for processing magnetic cores in a large variety of complex shapes and dimensions, keeping manufacturing costs low and the processing time has been reduced.
In 1996, at the Oak Ridge National Laboratory, USA, gelcasting was used for the first time to produce ferrite cores. In more recent years, Virginia Tech, USA, PhD candidate, Lanbing Liu, used this method for power electronics application in the paper titled, NiCuZn ferrite cores by gelcasting: processing and properties, published in IEEE Transactions on Industry Applications. At University of Nottingham, UK, an analogous process is under investigation. The scope is to take works previously proposed and optimise them for heterogeneous integration purposes. To this end, a reduced solid load of 65 weight percentage (wt%) and gelation without catalyst have been applied to improve casting, and de-airing steps are needed.
Heterogeneous integration consists of a functional and physical integration of heterogeneous power electronics components, structures and materials with the aim of minimising parasitic elements, increasing operating frequency and power density. This concept, together with a modular approach based on locally optimised switching cells, aims to be a successful route to enable the new wide bandgap devices to switch at their maximum potential.
A feature of these combined approaches is that each commutation cell comes with gate drivers, filters and all the elements needed for its own independent operation, including magnetic components. The challenge is to provide magnetic cores that ensure optimum exploitation of the available space. For this reason, the core manufacturing technique must allow modifications in geometry and dimensions without extensive effort or mass production requirements.
Another aspect is related to the energy density of the inductive components. In this case, the limit is fixed by the maximum level of magnetic flux density that magnetic materials can tolerate without entering in saturation. There are soft magnetic materials, such as amorphous and nanocrystalline alloys, which show saturation levels higher than 1 Tesla (T). Saturation level higher than 1T is unusually high for soft magnetic materials. It is unfortunate that they can only operate at relatively low frequencies of 10-20kHz because their relative permeability begins to drop off, which causes the saturation level to do the same.
The only viable solution for applications above 1MHz are NiZn ferrites, however, they do not exhibit high saturation magnetic flux density, typically 300 millitesla (mT). Therefore, to avoid saturation and excessive losses, the equivalent permeability of the magnetic path must be kept small. In recent years, quasi-distributed air gap cores have been used in many applications, indeed, they offer the advantage of obtaining low permeability, hence avoiding saturation, by introducing multiple small air gaps.
The way to manufacture these cores consists of cutting bars of high permeability materials and shaping the magnetic path by combining them. Although this may appear as a practical solution, core losses increase drastically when the material grains are damaged for cutting. The ideal solution would use a closed core with a permeability that matches the application requirements. In this activity, the researchers proved that gelcasting – a forming technique which is based on the casting of a liquid slurry – can offer both shape flexibility and permeability modulation.
A near net shape process
The gelcasting process can be divided into two phases – a low-temperature phase where a green body of the desired shape is obtained by casting and polymerisation of a liquid slurry, and a high-temperature phase where the green body is sintered and the grain structure is formed. The first step of the low temperature phase consists of dissolving the monomers in water for hydrogel preparation. Methacrylamide (MAM) and N, N’-Methylenebisacrylamide (MBAM) are used respectively as chain monomer and cross-linker. The monomer ratio of MAM to MBAM, can be modified to improve the strength of the green body. The water solution is then mixed overnight using a magnetic stirrer. Once the monomers are dissolved in water, ferrite powder is added to the solution. LSF120, a NiCuZn ferrite powder is used in this project. The solid load is kept low, 65wt%, to reduce the viscosity and improve the casting even in non-conventionally shaped moulds. Another advantage is that this solution enables the use of high-speed electric mixers instead of ball millers, resulting in complexity reduction for the overall process. During this step a dispersing agent is added to produce homogeneous slurry.
Reducing macro-porosity is essential to getting a stronger green body and better magnetic properties. In this case, a double step de-airing can be performed because no catalyst is used to carry out the polymerisation. The first de-airing in the process removes air introduced by electric mixing. After adding ammonium peroxydisulphate (APS) as polymerisation initiator, the liquid slurry is cast in silicone rubber moulds. Since at this stage the mixture is still liquid, due to the absence of a catalyst, the second de-airing can take place to remove air pockets that may be generated while casting. Vibrating the moulds while de-airing has proven to be extremely helpful to achieve complete removal of air.
In the absence of a catalyst, polymerisation is triggered through increasing the temperature to 80oC for 30 minutes. Once the reaction is complete, the moulds can be removed and the green bodies are ready to be fired. Following this procedure, any shape can be obtained in a cost-effective way by using different silicone moulds. As the purpose of this project was to perform magnetic characterisation of gelcast cores, toroids were chosen as target shape, in order to comply with constant flux density hypothesis during the measurements.
In this way, permeability and loss density measurements are not affected by core geometry, but they describe the material behaviour. Relative permeability as a function of frequency is obtained by small signal analysis using an impedance analyser. A sine wave generator (100kHz–1MHz) and a power amplifier have been used to perform fluxmetric measurements and calculating magnetic losses at excitation level compatible with standard power electronics applications (10mT–100mT).
During the high temperature phase, the green bodies are taken to the furnace for sintering. This is the most delicate step of the whole process – as significant shrinkage occurs because of water and organics removal, the temperature profile must be finely tuned to avoid failure by cracking.
Thermo-gravimetric analysis (TGA) is performed on the green bodies to identify when temperatures organic evaporation occurs.
Using TGA results, the temperature profile has been set with an initial ramp rate of 2°C–200°C, where the temperature is held for one hour to let all the organics burnout, after which the set point increases by 10°C per minute to reach the sintering temperature. During sintering, the ferrite particles diffuse forming the grain structure. By varying the sintering temperature or time, it is possible to obtain different microstructures, e.g. grain size, grain size distribution and micro-porosity percentage, as a result achieve magnetic properties.
Our experimental activity was then focused on correlating sintering temperature, microstructure and magnetic properties. Different sintering temperatures have been investigated, between 1,000°C–1,100°C. Firstly, in order to perform scanning electron microscope (SEM) imaging and characterise the microstructure, the toroidal samples were mounted and polished down to a 0.025μm colloidal silica finish.
Analysis by open source Java image processing programme, ImageJ, was performed to quantify porosity by using the particle counting feature and to quantify grain size by using the linear intercept method.
The initial porosity level at sintering temperature of 1000°C is 27%. By increasing the temperature, it is possible to reduce porosity to 4%. This result proves that nearly fully dense samples can be produced even by using pressure-less sintering process.
Maximum average grain size was obtained in the 1,100°C sample, which was estimated to be 5.62μm, +/-1.23. By analysing the magnetic measurements, it appears that the higher the sintering temperature the lower the magnetic loss density. Indeed the samples sintered at 1,100°C showed better performance even when compared to commercially available solutions, especially at higher frequencies.
This behaviour can be explained by looking at the microstructure – as the grain size increases and porosity at the grain boundaries decreases there is a significant reduction of interference with domain wall movement, hence hysteresis losses are reduced. For similar reasons, magnetic permeability is influenced by grain size as well as porosity. It is worth noting that, in experimental activity, by varying the sintering temperature in a range of 1,000°C-1,100°C, the relative permeability varied from 40–200, enabling a significant tuning of this property.
Gelcasting offers power electronics engineers and researchers additional freedom in magnetic component design, in terms of magnetic properties and shape. Further, gelcasting provides control over the whole manufacturing process, enabling our research group to investigate new magnetic materials for future power conversion applications. Of particular benefit tothe community is the extension of the process to MnZn ferrites and the introduction of dopant to improve the magnetic properties of gelcast cores.