Mixing it in academia and industry
Professor Wang Nang Wang is not your typical academic. He has not devoted his entire career to university research, but has instead worked for several years in executive roles at optoelectronic chip manufacturers. Although he didn t really enjoy running companies, his time spent leading Quantum Optotech and founding Arima Optoelectronics has given him experiences that have clearly shaped his outlook and today all of his research is focused on improving the performance of real-world devices.
The importance he places on the potential commercial impact of his work is reflected in his interest in patent literature. If many generic patents are held in a particular area he steers clear of this topic, as any breakthrough will not lead to successful commercialization. In fact it just helps the existing manufacturers, says Wang, who are simply shown how to improve their own product s performance.
This desire to avoid technologies covered by a string of patents has led Wang to pursue a new technique for making GaN substrates, which have subsequently formed a high-quality basis for optoelectronic device growth. With a focus on improving commercial GaN LED and laser performance in particular, Wang s first move was to design a novel reactor. This first vertical HVPE reactor was built in 2001 and has led to the filing of seven patents.
A vertical showerhead configuration has been employed for the first two generations of this tool and will also be used for a third reactor that is under construction. According to Wang, the advantage of this design is its high degree of symmetry, which allows easy scaling up from the 2-inch GaN that has been grown so far to 3, 4 and ultimately 6-inch material.
Wang s first reactor suffered from a very narrow process window for making high-quality material and a tendency for producing polycrystalline films at high growth rates. "We also didn t have a compatible process for growing materials that are flat and free from cracks."
The second reactor addressed some of these faults, along with the introduction of a new fabrication process that Wang refers to as nano-pendeo epitaxy. This technique involves taking a silicon or sapphire substrate, patterning it with sub-micron-sized pores and then depositing GaN onto this structure. GaN nanocolumns are formed as a result, which can disrupt many of the dislocations created during conventional hetero-epitaxy. "There are no threading dislocations [in our material]," said Wang, "and the dislocation density falls to 8 × 106 /cm2 when we go over 300 µm."
When it is finished this fall, the third-generation reactor will be easier to use than the previous tools and will produce faster growth rates, thanks to the combination of modifications to the showerhead, the source design and control software. "This allows us to run efficiently for a few days of operation," explained Wang. The growth efficiency, the proportion of process gases converted into GaN material, is expected to be 55% for the latest reactor, compared to 30–35 and 45% for the earlier two versions, respectively.
Wang s team is also equipped with an MOCVD reactor, a modified AIX 200 horizontal tool. This single-wafer reactor is used to grow optoelectronic structures for various collaborative programs and six patents are in the pipeline relating to device growth and non-polar materials.
The team s LED development, which will aid the UK project NoveLELS – aiming to develop very bright devices for airplane wings and cockpits – has produced polar emitters that are almost free from so-called "droop". This effect, a reduction in lumen per watt as drive current increases, plagues all of today s LED products and is a barrier to manufacturing very bright, efficient devices. Commercial LEDs that Wang has looked at produce 130 lm/W at 20 mA, but this falls to 80 lm/W at 350 mA. In comparison, his devices – which have not been packaged and are measured at the wafer level – are slightly worse at low current drive, but typically produce 100 lm/W at 350 mA.
To solve the droop problem, Wang incorporated a resonant tunnel junction into the LEDs. This cools electrons before they enter the quantum wells and leads to higher recombination efficiencies. This gain is seen in LEDs grown on both sapphire and free-standing GaN, but the latter platform produces higher efficiencies and reliabilities.
Earlier this year, Philips Lumileds claimed that it had also solved the efficiency droop problem, but it did not disclose the details behind this breakthrough. "Lumileds have a lot of tricks up their sleeves to improve the epitaxial structure and they might have a different way," said Wang. Although he doesn t know exactly how Lumileds improved the high-current performance, he believes that excellent thermal management, which is a hallmark of its LEDs, will have contributed to the success.
Wang s LEDs also benefit from very good thermal management. This advantage comes from forming free-standing devices with thin GaN layers, which are made by using wet etching or mechanical rotation to remove the foreign substrate. The thinner chip reduces heat confined within the device, which can also be attached to another carrier that can act as a heat sink.
This free-standing chip approach has also been employed in Blurayds, a UK government-funded project directed at developing high-power 405 nm laser diodes that is led by Wang and includes the University of Oxford, Advanced Optical Coatings of Plymouth and Arima Optoelectronics (UK) Ltd. With this approach, the laser can benefit from an air interface that improves optical confinement thanks to an increase in the refractive index contrast at this boundary. "You can also put the laser on another carrier, such as silicon," said Wang, "and align the cleaving facet of GaN with the natural facet of silicon, which means that you have a better cleaving facet." This can aid laser manufacture, as it can improve the yield of the cleaving process that defines the cavity s dimensions.
Lasers made with Wang s epiwafers, which also contain a resonant tunnel junction to cool electrons, have threshold currents below 50 mA and maximum output powers of 50–60 mW. These diodes also feature an undisclosed facet coating to improve performance that is not titanium dioxide or silicon dioxide. "With titanium dioxide, the performance of 405 nm lasers can trail off, because titanium dioxide has a cut-off wavelength around 380 nm. If this material is not of sufficient crystalline quality it can absorb at the laser s wavelength," explained Wang. This will reduce the laser s output power, an issue that he is trying to avoid with his novel coatings that also improve the threshold current density.
Wang s partnerships with device developers – he co-founded Arima in 2000 – have been very useful because they have provided feedback about material quality and helped to drive improvements in substrate and epitaxial growth.
Currently, his team is only producing a few wafers each month, but Wang is once again sensing a real commercial opportunity and shipments should increase through the launch of a spin-off company, NanoGaN, which will produce both substrates and epiwafers. This venture already has £250,000 ($499,000) from Bath University s "Sulis" commercialization fund, showing that even when Wang works in academia, his commercial nous leads him to playing a role within the manufacturing side of our community.
