Research

New spin-off at FBH: Thin film coating at atmospheric pressure

FBH research: 12.12.2012

: Fig. 1: Plasma source at atmospheric pressure

: Fig. 2: Experimental set-up for sputter coating at atmospheric pressure

: Fig. 3: Deposited gold layer

Many of the commodities we use every day, such as computers, flat-panel displays, mobile phones, eye glasses, and architectural glasses contain thin films deposited by cathode evaporation (sputtering). Up to now, this requires bulky and costly vacuum coaters. Researchers at FBH successfully transferred this process to atmospheric pressure. This promises substantial cost reductions by eliminating expensive vacuum technology. Also, it opens up new coating applications not feasible today due to the needs of vacuum environment. The technology is to be commercialized in near future by founding a spin-off company.

The technology is based on an atmospheric plasma source developed for medical treatments during a BMBF-funded project (fig. 1). An experimental set-up (fig. 2) combines the plasma source with a sputter target and an additional sputtering power supply. This allows sputter deposition of thin films at atmospheric pressure (fig. 3). Combining the microwave plasma with a dc current for ion extraction leads to a stable operation and thus enables the desired thin-film deposition. International patents are pending.

The technology was presented to the community for the first time in October 2012 on the 13th International Conference on Plasma Surface Engineering in Garmisch-Partenkirchen, Germany, and raised a lively discussion. Up to now, this process was considered to be impossible, even by experts. The great market oportunities of this invention led employees of the FBH to go for the foundation of a spin-off in order to commercialize the technology. An EXIST research transfer project proposal was submitted and successfully passed the review.

Publications

R. Bussiahn, R. Gesche, S. Kühn, K.-D. Weltmann, "Integrated Microwave Atmospheric Plasma Source (IMAPlaS): thermal and spectroscopic properties and antimicrobial effect on B. atrophaeus spores", Plasma Sources Sci. Technol., vol. 21, no. 065011 (2012).

J. Liebmann, J. Scherer, N. Bibinov, P. Rajasekaran, R. Kovacs, R. Gesche, P. Awakowicz, V. Kolb-Bachofen, "Biological effects of nitric oxide generated by an atmospheric pressure gas-plasma on human skin cells", Nitric Oxide, vol. 24, no. 1, pp. 8-16 (2011).

Fast and efficient switching with GaN-based power-transistors

FBH research: 20.11.2012

: Fig. 1: 100 mΩ / 300 V GaN switching transistor with 4.5 mm × 2.4 mm die size on a 3" wafer

: Fig. 2: Pulsed output characteristic of a GaN power transistor in p-GaN gate technoloy with 150 mm gate width showing an on-state resistance of RON R_ON = 80 mΩ for 5 V gate bias

: Fig. 3: Switching transients from off- to on-state (l) and from on- to off-state (r) for a GaN transistor with 150 mm gate width for 150 V blocking voltage and 6 A on-state current. The rise and fall times are 20 ns, respectively. Measurements: TU Berlin

Light weighted and small volume electronic power converters are desirable for automotive and other mobile applications. Technically, such features are achieveable by increasing the switching frequency, which enables smaller and more light weighted passive components such as, for example, inductors. However, a key requirement is the availability of efficient power switching transistors.

The transition time between transistor on-state (low conduction losses due to low voltage drop at high current) and transistor off-state (low losses due to low leakage current at high bias) should be as short as possible to minimize switching losses. Current flow and drain bias are simultaneously present in this period and thus generate thermal loss. The gate charge required to open or to close the transistor channel determines the switching speed. The product of on-state resistance R_ON and gate charge Q_G is thus a benchmark for switching losses inherent to a certain transistor technology.

GaN-based high-voltage switching transistors simultaneously show a low area-specific on-state resistance and a low gate capacitance. This enables particularly low switching losses and therefore efficient high-frequency converters.

The FBH develops normally-off GaN transistors based on a p-GaN gate technology. Switching transistors with 150 mm gate width have been realized showing an on-state resistance R_ON = 100 mΩ and a maximum pulse current of 75 A for 300 V blocking strength.

A double-pulse switching set-up with inductive load and a time resolution < 5 ns has been realized at the TU Berlin, Power Electronics Department to analyze the switching properties of the FBH transistors. The rise and fall time of the switching events were both determined to 20 ns only when the transistors were switched between an off-state voltage of 150 V and a current load of 6 A. For 1 MHz switching transistor losses would thus be present only for < 10% of the total cycle. The gate charge required for the transistor switching event was determined as Q_G = 5.5 nC. The resulting product R_ON x Q_G should be as small as possible. For the tested GaN transistors optimized for 300 V blocking voltage R_ON x Q_G is only 0.55 ΩnC and therefore 4 times better than compared to competitive Si devices featuring the same blocking voltage.

Publications

O. Hilt, F. Brunner, E. Cho, A. Knauer, E. Bahat-Treidel, J. Würfl, "Normally-off High-Voltage p-GaN Gate GaN HFET with Carbon-Doped Buffer", Proc. Int. Symp. on Power Semiconductor Devices & IC's (ISPSD), San Diego, CA, May 23-26, pp. 239-242 (2011).

O. Hilt, E. Bahat-Treidel, E. Cho, S. Singwald, J. Würfl, "Impact of Buffer Composition on the Dynamic On-State Resistance of High-Voltage AlGaN/GaN HFETs", Proc. Int. Symp. on Power Semiconductor Devices & IC's (ISPSD), Bruges / Belgium, 3-7 June, pp. 345-348 (2012).

N. Badawi, P. Knieling, S. Dieckerhoff: "High Speed Gate Driver Design for Testing and Characterizing WBG Power Transistors", Proc. EPE-PEMC (2012).

3-dimensional integration of SiGe-BiCMOS and InP-HBT technology

FBH research: 06.11.2012

: Fig. 1: The separately processed InP and SiGe-BiCMOS Wafers are adjusted by capable alignment marks and bonded using BCB with an accuracy better than < 10 µm.

: Fig. 2: After the wafer bond process, the InP substrate is removed, and the structures processed before are exposed. The interconnects between InP and Si technology are formed by dry etching and plated with gold.

: Fig. 3: Microstrip fabricated in transfer-substrate-technology with via holes in the subjacent metallization layer of the BiCMOS wafer.

Indium Phosphide (InP) Hetero-Bipolar Transistor (HBT) high frequency circuits fabricated at FBH by far exceed the performance of silicon devices in terms of RF power and frequency range. Due to its high electron mobility and breakdown voltage, InP technology opens up application fields that are difficult to access with RF-CMOS and SiGe-BiCMOS technology. The capability to generate high RF power at high frequencies enables millimeter wave imaging applications, e.g. high-resolution radar systems and terahertz scanning. Also, for future wireless communication systems with data rates exceeding 100 Gbit/s, amplifiers and mixers in InP Technology operating at 300 GHz and above are key enabling components.

Silicon is the dominating material of modern semiconductor technology with a market share of 98,5%. Established process modules, high integration density and yield enable a cost-efficient production of integrated circuits with high complexity. However, as application frequencies rise above 100 GHz, the limited RF power performance of silicon-based circuits at these frequencies becomes a bottleneck.

In order to profit from the advantages of both technologies, InP and silicon, the joint "HiTeK" project funded within the SAW context. FBH and the Leibniz Institute for Innovative Microelectronics (IHP) aim to establish a technology platform for heterogeneous integrated circuits at terahertz frequencies (0,1-1 THz) [1]. Within this project, the established InP HBT transfer-substrate-process at FBH [2], is combined with IHP's Silicon-Germanium BiCMOS process. First, InP and silicon wafers are processed separately at FBH and IHP, respectively. Since IHP's production line uses a wafer size of 200 mm in diameter, the wafers need to be cut into 3" diameter wafers for compatibility with FBH's process line. Utilizing benzocyclobutene (BCB) based wafer bonding, the 3-dimensional integration of both technologies is performed at FBH (see fig. 1). In order to meet the requirements for lateral wafer-to-wafer alignment accuracy of less than 10 µm, the wafer bond process has been improved for the new materials system InP/BCB/Si. After wafer bonding, the InP substrate is completely removed, and the previously created structures in InP and BiCMOS technology are exposed (see fig. 2). By dry etching, short and thus low-impedance vertical interconnects (vias) are defined between InP and Si-BiCMOS. Within the BiCMOS technology, aluminum is used as an interconnect metal. Since aluminum tends to form an oxide passivation layer which could increase the contact resistance of the vias, special attention needs to be focused on the via process. By optimizing the fabrication procedures, via resistance decreased from multiple ohms down to 300 mOhm. In a first technology run, passive microstrips have been tested which were designed in accordance to the given requirements. The test structures consist of gold transmission lines on the InP side, two via transitions in and out of the silicon circuit layers, and an aluminum transmission line on the silicon side. The propagation loss in the gold transmission line on the InP side amounts to 0,4 dB/mm. We can therefore extract a loss of 0,36 dB per via. The low loss and the attained frequency behavior are matching well with the simulation. This result is a first proof of the monolithic high-frequency integration of our InP-HBT and SiGe-BiCMOS technologies. Wafers with active circuits are currently being processed and are expected to be completed in the coming months.

Publications:

[1] M. Lisker, A. Trusch, M. Fraschke, P. Kulse, Y. Borokhovych, B. Tillack, I. Ostermay, T. Krämer, F.J. Schmückle, O. Krüger, V. Krozer, W. Heinrich, "InP-Si BiCMOS Hetero Integration for Broadband Radio Links", submitted to Smart Systems Integration, International Conference & Exhibition, 2013.

[2] T. Kraemer, M. Rudolph, F.J. Schmueckle, J. Wuerfl, G. Traenkle, "InP DHBT Process in Transferred-Substrate Technology With f_t and f_max Over 400 GHz", IEEE Transactions on Electron Devices, Vol. 56 (9), p. 1897 – 1903, 2009.

Narrow bandwidth red emitting diode lasers

FBH research: 26.10.2012

: Fig. 1: Voltage, output power, and efficiency as a function of driving current of a DBR-RWL emitting at 635.3 nm.

: Fig. 2: Spectrum of a DBR-RWL with a grating period of 972 nm at 15°C.

Narrow bandwidth diode lasers emitting in the red spectral range will be an alternative for commonly used He-Ne-Lasers in metrology and Raman spectroscopy in the future. Furthermore, such kinds of lasers enable high brightness red-emitting light sources for entertainment and display technology by wavelength multiplexing. Scientists from the FBH just recently succeeded to integrate distributed-Bragg-reflector (DBR) surface gratings in waveguide structures for diode laser emitting in wavelength range 630 nm…640 nm. Monolithic integration makes such laser sources very compact and stable.

For that purpose, FBH used a waveguide structure containing a p-side AlGaAs cladding instead of the typically used AlInGaP cladding layer. Such a structure enables an easier production of surface gratings using dry etching technology. FBH developed red-emitting ridge-waveguide (RW) diode lasers with 10^th order (about 1000 nm period) surface gratings as wavelength selective reflector by using standard i-line stepper lithography. Such technology is very reproducible and paves the way to mass production.

First results turned out very promising: The new DBR-RWLs emit up to almost 100 mW optical output power in the wavelength range between 630 nm and 640 nm (Fig. 1). A grating period of 972 nm, for example, resulted in an emission of 635.35 nm (Fig. 2). The emission is spectrally narrow with less than 9 pm bandwidth and a side mode suppression ratio in excess of 18 dB. With customized coating a small linewidth necessary for high precision metrology and a higher side mode suppression ratio was achieved. A very simple and, at the same time, fast wavelength tuning can be performed by current modulation. Research on tapered laser structures with such monolithic integrated Bragg reflectors will lead to an output power above 100 mW with narrow spectral bandwidth in future.

Publication:

D. Feise, W. John, F. Bugge, G. Blume, T. Hassoun, J. Fricke, K. Paschke, G. Erbert, "96 mW longitudinal single mode red-emitting distributed Bragg reflector ridge waveguide laser with tenth order surface gratings", Opt. Lett., vol. 37, no 9, pp. 1532-1534 (2012).

Optimizing Multifinger Transistors for MM-Wave Power Amplification

FBH research: 09.10.2012

: Transistor variations in TS-InP technology (upper pad is the base, lower pad is the collector, emitter is grounded)

: Short-circuit current-gain cut-off frequency f_T vs. maximum frequency of oscillation f_max for single and multifinger TS-InP transistors. Emitter size type I: 0.8 x 5 μm², type II: 0.9 x 10 μm²

To achieve higher output power levels of mm-wave power amplifiers, there are only a limited number of practical methods available. Heterojunction Bipolar Transistors (HBT) with sufficient gain at mm-wave or higher frequencies are restricted by the emitter size for a given technology. This limits the achievable output power per emitter finger. To increase the circuits overall power level it is then necessary to resort to power combining, which can be done on the circuit level by combining the outputs of different transistors, or on transistor level by increasing the number of fingers per transistors.

The last approach has the advantage that losses in the power combining network are avoided. The disadvantage is that parasitics, due to the necessary changes in transistor geometry, reduce the cut-off frequency as compared to the 1-finger transistor.

Work on multifinger transistors in FBH's Transferred-Substrate InP HBT technology has led to the development of 4-, 6-, and 8- finger transistors with transit frequency (f_max) values of 323, 283, and 256 GHz, respectively. Even though 1- and 2-finger transistors achieve approximately 375 GHz (the emitter finger size is in all cases 0.8 x 5 μm²). The realized multi-finger HBTs are suited for amplifier applications in the mm-wave range around 100 GHz. The top figure shows the short-circuit current gain cut-off frequency and the maximum frequency of oscillation for transistors with varying number of fingers. The figure shows two types of transistors (type I and type II) which exhibit an emitter area of 0.8 x 5 μm² and 0.9 x 10 μm², respectively. One clearly observes the general degradation in transit frequencies with increasing finger number, but within each group there are still differences that can be exploited to find the optimal transistor geometry.

Power measurements on matched amplifiers have indicated that the maximum output power level is 14-15 dBm@77 GHz (3.5 mW/μm² emitter size) for a 2-finger transistor. This should allow almost 20 dBm output power for 8-finger transistors in the W band. At higher frequencies the number of fingers needs to be reduced or the emitter width must be downscaled.

The work is supported by DLR (Raumfahrt-Agentur, Deutsches Zentrum für Luft und Raumfahrt) under reference 50 RA 1103.

Publication:

T. Jensen, T. Kraemer, T. Al-Sawaf, V. Krozer, W. Heinrich and G. Tränkle , "Multifinger InP HBT’s in Transferred-Substrate Technology for 100 GHz Power Amplifiers", to be published at European Microwave Conference, Amsterdam, October 2012.

Growth process modification to improve the quality of Al0.45Ga0.55N

FBH research: 27.09.2012

: Fig. 1 Cross section images of AlGaN layers (a) arrow: no material is grown on top of the ridges, (b) AlN buffer layer leads to growth on top of the ridges.

: Fig. 2 Cross section image of a flat coalesced AlGaN layer grown at 400 hPa.

Hydride vapor phase epitaxy (HVPE) offers the possibility to efficiently grow thick layers of AlN, GaN and Al_xGa_1-xN at high growth rates. Binary AlN and GaN layers with a thickness above 50 µm are already commercially available. They serve as the basis for growth of devices like UV-LEDs. In particular, Al_xGa_1-xN in a medium composition range (x≈0.5) would be an ideal base layer for the fabrication of UV-B LEDs emitting at wavelengths between 280 – 320 nm. However, such HVPE grown ternary layers are not yet available mainly due to the bad layer morphology that suffers from polycrystalline growth and composition inhomogeneity.

The improvement of HVPE grown Al_0.45Ga_0.55N crystal quality is a current research topic at FBH. For that purpose, the sapphire substrate is patterned with parallel trenches which then are laterally overgrown (epitaxial lateral overgrowth, ELOG) by material growing from the top of the ridges. This method is expected to improve the material quality and especially to avoid crack formation. As shown in Fig. 1a, distinct polycrystalline growth occurs from the sidewalls of the trenches. Facets are offered on the stepped surface on which the deposition of Ga-rich material is favored resulting in inhomogeneity in the Al-distribution in the Al_0.45Ga_0.55N layer. To achieve the aim of a coalesced flat layer by growth from the top of the ridges ways to suppress growth in the trenches were investigated. The deposition of a 500 nm thick AlN layer prior to growth was found to support the Al_0.45Ga_0.55N deposition on the top of the ridges, but the material grown from the sidewalls of the trenches still dominates the surface morphology (Fig. 1b). The surface mobility of the precursors on the sapphire substrate can be influenced by the total pressure during growth. A decrease in total pressure from 800 hPa (Fig. 1b) to 400 hPa (Fig. 2) suppresses precursor diffusion into the trenches and therefore leads to a flat coalesced surface. Growth on such a coalesced Al_0.45Ga_0.55N layer allows for an increase in layer thickness and hence offers the possibility to fabricate substrates for UV-B LEDs.

Publication

S. Hagedorn, E. Richter, U. Zeimer, D. Prasai, W. John, M. Weyers "HVPE of Al_xGa_1-xN layers on planar and trench patterned sapphire" J. Cryst. Growth, vol. 353, no. 1, pp. 129-133 (2012).

Flip-chip mounted UV-B LEDs with optical powers in the milliwatt range

FBH research: 14.09.2012

: Fig. 1: Optical power-current characteristics of a mounted UV-B LED with an output power of ≥5 mW and emission spectrum at a current of 20 mA.

: Fig. 2: Degradation measurements on a flip-chip mounted UV-B LED at a current of 20 mA; lifetime >300 hours.

The application fields of ultraviolet (UV) emitters in the 300 nm to 320 nm wavelength range are manifold: ranging from material processing such as hardening or surface treatment of plastics, paints and finishes to medical applications like phototherapy of skin diseases. One of the benefits of UV light generated by light emitting diodes (LEDs) is that the wavelength can be optimally tailored to fit the application. Furthermore, LED light sources can be realized with compact form factors and, unlike mercury lamps, do not contain toxic substances. Within the regional growth core Berlin WideBaSe, FBH develops UV-B LEDs with high optical power and efficiency in close collaboration with TU Berlin (TUB) and local industry partners. After successful optimization of the fabrication technology we managed to increase the optical power of the devices into the milliwatt range, which is comparable or even better than international record values for this wavelength range.

The design and the epitaxial growth conditions of the semiconductor layer structure were found to be the key to improve the LED efficiency. Using AlN buffer layers deposited on sapphire substrates at temperatures ≥1400 °C together with thick Si-doped AlGaN layers, templates for light emitting layers were fabricated whose dislocation density is in the range <10¹⁰ cm^-2. Moreover, the composition and the doping of the electron blocking layer, which is intended to prevent electrons from overflowing to the p-side of the diode, were carefully optimized. Selected wafers of this first generation were used to process bottom-emitting LEDs which were flip-chip mounted on AlN submounts using hard-solder and measured on copper heat sinks in a calibrated integrating sphere. Fig. 1 shows that the LEDs feature maximum output powers of up to 5 mW at an emission wavelength of 300 nm. Moreover, degradation studies as presented in Fig. 2 show lifetimes of the LEDs of several hundred hours with very little degradation after the burn-in. Further optimization steps allowed to improve the performance of the LEDs even more. UV-B LEDs of the second generation, which were fabricated after these optimizations, reproducibly feature external quantum efficiencies in the range 0.5 to 1 % in on-wafer measurements. In the near future those LEDs will be available as mounted chips and allow us to build first demonstrators for the applications mentioned above.

Powerful brilliant light sources: Truncated-tapered semiconductor optical amplifiers improve MOPA system performance

FBH research: 24.08.2012

: Fig. 1: Schematic of the optical system used for in- and out-coupling of the semiconductor power amplifier in top view (upper) and side view (under).

: Fig. 2: Measured dependence of output power, voltage and efficiency as a function of QCW bias current for a 400 mW external seed power. Inset: optical pulse profiles at 40 and 60 A.

One of the current challenges in the development of semiconductor-based light sources is to combine optical powers in the Watt class with diffraction-limited beam quality and spectrally stable and narrow-band emission. Every improvement enables new applications in material processing, optical free-space communication, as well as in display and LIDAR technology. Significant progress has recently been achieved in this field using "Master-Oscillator-Power-Amplifier (MOPA)" systems. The output stage is based on a semiconductor optical amplifier with a large-area gain region. The MO produces spectrally narrow emission, with diffraction limited beam quality. The optical emission from the MO of around 0.5 W is coupled into the amplifier. The amplifier (power amplifier – PA) has an electrically pumped area with a so-called "truncated" taper-profile. This design enables a high output power and at the same time good beam quality (see Fig. 1). Low optical loss and a tailored modal gain are both critical factors in the design of semiconductor structures for use in such amplifiers, with the gain matched to the output of the MO. The gain must produce sufficient amplification; however if gain is too large the beam quality is degraded.

Based on variations to the vertical wave guide and the number of quantum wells, optimized semiconductor layer structures have been successfully developed that allow single amplifiers to achieve more than 17 W diffraction limited optical output under quasi-continuous-wave (QCW) conditions. The spectral properties are defined by the MO, with a spectral width of under 40 pm. A total power of more than 50 W was emitted within a narrow spectral band, as seen in Fig. 2. Further developments are now planned, that will produce compact MOPA-based systems that are optimized for CW-operation. This means in particular that the power conversion efficiency must be increased, and the optical components miniaturized. This optimization process will lead to MOPA systems being very attractive for use in commercial applications.

Publication

X. Wang, G. Erbert, H. Wenzel, B. Eppich, P. Crump, A. Ginolas, J. Fricke, F. Bugge, M. Spreemann, G. Tränkle, "High-power, spectrally stabilized, near-diffraction-limited 970 nm laser light source based on truncated-tapered semiconductor optical amplifiers with low confinement factors", Semicond. Sci. Technol., vol. 27, no. 015010 (2012).

Flexible and Efficient Microwave Power Amplifier for Future Mobile Base Stations

FBH research: 14.08.2012

: Circuit of the H-Bridge Class-D PA for the 900 MHz Band

: Realized H-Bridge Class-D PA module

The continuously increasing demands on today’s wireless communication infrastructure (more subscribers, more applications, higher data rates) require broadband channels (e.g. LTE) and spectrum-efficient modulation schemes, with an optimized exploitation of the available (limited) bandwidth. One consequence is that the RF power amplifier (PA) has to operate at high peak-to-average power ratios (PAPR), which significantly reduces energy efficiency when using a common linear PA (e.g. class-AB). In order to improve efficiency at usual UMTS PAPR (6 - 8 dB), one commonly applies the Doherty concept or envelope tracking.

Beyond this , the H-bridge class-D PA topology has been considered as a proper candidate to achieve high efficiency. It uses two complementary voltage-switching class-D PAs with two final stage transistors each acting as a switch. The realizations so far show either fully integrated solutions with output powers well below 1 W or use technologies with f_T and power density limitations lower than GaN.

At FBH a GaN-based H-bridge class-D PA for the 900 MHz band has been developed and realized. The figure shows the concept and the realized PA module.

The demonstrator achieved a maximum efficiency of 50% and a saturated output power of 8 W. Increasing the PAPR to 10 dB efficiency is degraded to 22% with a special pulse-width modulated signal. Also, other modulation approaches, e.g. band-pass delta-sigma modulation, have been tested, but did not improve the performance. The H-bridge PA represents a compact, efficient and flexible alternative for future mobile base stations, which allows to expand the digital part within the base-station architecture. The realized PA advances the state-of-the-art in mainly two ways:

This is the first microwave H-bridge realization in the 10 W of peak output power range.
The H-bridge topology offers new degrees of freedom with regard to ternary coding schemes, which is a key issue for efficiency and thus PAPR performance of the PA.

The results confirm FBH's leading role in this field of research. They were presented at the IEEE International Microwave Symposium in Montreal..

Publication

A. Wentzel, C. Meliani, G. Fischer, W. Heinrich, ”An 8 W GaN-based H-Bridge Class-D PA for the 900 MHz Band Enabling Ternary Coding”, IEEE MTT-S International Microwave Symposium Digest 2012, WEPG-14, Montreal, Canada.

Reduced dynamic on-state resistance for GaN power transistors

FBH research: 27.07.2012

: Packaged 80 mOhm / 300 V GaN switching transistor with 4.5 mm × 2.4 mm chip size

: Increased dynamic R_ON as a function of voltage blocking strength for different GaN buffer compositions. Carbon concentration in cm^-3. R_ON(stressed) was determined 0.2 µs after opening the transistor from off-state at 65 V drain bias

: Increased dynamic on-state resistance as a function of off-state drain voltage for a transistor with iron-doped GaN buffer

GaN-based high-voltage switching transistors enable particularly efficient power converters due to their low area-specific on-state resistance and low gate capacitance. Due to lower losses per switching cycle, GaN-based converters can operate at higher frequencies than converters with Si-based switches, enabling more compact and light-weighted converters. However, the on-state resistance is increased for short times after some 100 V off-state drain stress. This so called increased dynamic on-state resistance may counteract the GaN-HFET-related fast switching opportunity for high off-state voltages. Charging of the GaN semiconductor and of the semiconductor passivation under high electric fields typically present for the transistor off-state condition are the root causes. The injected charges remain inside the device structure after the transistor is switched to on-state and reduce the electron concentration in the transistor channel – the on-state resistance is increased.

The magnitude of increased dynamic on-state resistance is determined by the GaN buffer composition. Carbon doping is used to elevate the voltage blocking strength of the semiconductor, but it is also a major source for increased dynamic on-state resistance. Target is high electric blocking strength combined with low dynamic on-state resistance.

Recently developed GaN buffer compositions using AlGaN or iron-doped GaN feature electrical blocking strengths comparable to carbon-doped GaN buffer with low doping concentrations. But they show a significantly reduced dynamic on-state resistance. In addition to improved semiconductor material quality, a homogeneous electric field distribution inside the device may reduce the dynamic on-state resistance. Its increase gets reduced to 1/3 when applying gate-connected field plates to the devices.

Publication

Highly reliable silicon carbide photodiodes for visible-blind ultraviolet detector applications

FBH research: 12.07.2012

: Fig. 1: Optical micrograph of the front of a processed photodiode (size 1x1 mm²).

: Fig. 2: Normalized photocurrent of SiC UV photodiodes stressed with a low pressure UV-C lamp. For comparison the dashed lines show the behavior of previous commercial 6H-SiC p-n photodetectors.

: Fig. 3: Normalized photocurrent of 4H-SiC UV photodiodes stressed with a medium pressure UV lamp.

A growing variety of applications for detectors of ultraviolet (UV) radiation depend on highly reliable and long-term stable detection systems. Such applications comprise food, air and drinking water purification, flame and combustion control as well as chemical and biological analysis or sterilization of biological and medical equipment. In a joint project of FBH, Leibniz-Institute for Crystal Growth, and the Berlin-based company sglux GmbH highly efficient and reliable polytype 4H-SiC p-n photodiodes have been fabricated on n-type substrates with active areas of up to 6x6 mm². The n-type and p-type layers were epitaxially grown at Leibniz-Institute for Crystal Growth; the devices were characterized and long-term exposed to UV by sglux GmbH.

At FBH, SiC p-n junction photodiodes were fabricated on 2 and 3 inch n-type substrates. Contact printing lithography using a mask aligner has been used for all patterning and alignment steps. Modern and automated equipment has been used for dry- and wet-chemical processing as well as for deposition of metals and dielectrics to ensure best possible process control. The active area of the p-n junction diodes was defined by mesa etching using an inductively coupled plasma (ICP) with Cl₂-BCl₃ chemistry. The SiC etching technology assured high homogeneity of ± 15 nm over the 3 inch wafer surface and exact control of the total etch depth of 7 µm. Electrical contacts were formed by aluminum-titanium and nickel-chromium alloys on p-type and n-type SiC, respectively. As shown in Fig. 1 the electrical contact at the front has a bonding island in the center (bright). The edge length of the active area of the square chip depicted is 1 mm.

Photodiode chips packaged into a TO housing with an UV transparent window were characterized at room temperature and exposed to high intensity mercury lamp irradiation. The behavior of the photocurrent response under UV light irradiation using a low pressure mercury UV-C lamp (4 mW/cm²) and a medium pressure mercury discharge lamp (17 mW/cm²) has been studied for up to 22 months. The normalized photocurrents of SiC UV photodiodes stressed with a low pressure UV-C lamp and a medium pressure UV lamp are depicted in Figs. 2 and 3, respectively. The devices under test showed an initial burn-in effect, i.e., the photocurrent response dropped by less than 5 % within the first 40 hours of artificial UV aging. Such burn-in effect under UV stress was also observed for previously available 6H-SiC p-n photodetectors (CREE CD‑260‑0.30‑D, on p-type SiC substrates). After initial burn-in no measureable degradation was detected, i.e., no decrease of photocurrent could be measured until now (16,000 hours of irradiation). The aging tests demonstrate the robustness of SiC photodiodes against UV radiation and confirm that the 4H SiC p-on-n photodiodes are suitable for high irradiance UV applications.

This work was partly supported by the Federal Ministry for Economy and Technology based on decision of the German Parliament under grant number KF2194501DB9. Part of the work was funded by the Transfer BONUS program of Berlin's Senate Department for Economy, Technology and Women. Technological support by Frank Kudella and George Brandes is very much appreciated.

Publication

D. Prasai, W. John, L. Weixelbaum, O. Krüger, G. Wagner, P. Sperfeld, S. Nowy, D. Friedrich, S. Winter, T. Weiss, "Highly reliable silicon carbide photodiodes for visible-blind ultraviolet detector applications", J. Mater. Res., 2012, accepted for publication.

Beam characterization for single-mode fiber coupling of tapered lasers

FBH research: 02.07.2012

: Coupling efficiency of the beam from a tapered laser as a function of the axial displacement of the fiber position for two different RW currents; measured (squares) and calculated from measured WDFs (curves).

Tapered lasers show a high market potential for many applications. They deliver a much higher output power than ridge-waveguide (RW) lasers and a significantly better beam quality than broad area (BA) lasers at the same power level. However, for many applications it is necessary to transport the beam using single-mode fibers. Tapered lasers feature a more complex beam profile than RW and BA lasers. This complexity is of such a high level that the commonly used beam propagation factor (M²) is insufficient to predict the coupling efficiency into a fiber.

At the FBH, it could be shown that a more in-depth description of the beam profile using the Wigner distribution function (WDF) allows to precisely predict the coupling efficiency. As an example, a tapered laser with two contacts, one for the ridge-waveguide section and one for the tapered section, was operated using two different RW currents.

While the power is constant and the M² increases only slightly from 4.2 to 5.0, the coupling efficiency deteriorates from 68% to 37% for RW currents of 350 mA and 500 mA, respectively. This difference in coupling efficiency of almost a factor of 2 is not adequately reflected in the M² value. However, when measuring the Wigner distribution function of the beam at the two operation points, coupling efficiencies of 65% and 34% were calculated. The error in the prediction of our WDF based coupling model was 3% only.

This demonstrates the power of the new WDF method, making it a suitable tool to design coupling optics and paving the way for efficient fiber-coupled laser sources powered by tapered lasers.

Publication:

M. Uebernickel, B. Eppich, K. Paschke, G. Erbert, G. Tränkle, "Prediction of Single-Mode Fiber Coupling Efficiencies of a Tapered Diode Laser From Measured Wigner Distribution Functions", Photonics Technology Letters, IEEE Volume 24 , Issue 14 (2012).

Assessment of real wafer temperature of GaN/Saphir with new in-situ sensor

FBH research: 19.06.2012

: Fig. 1: PL wavelength of InGaN/GaN multi quantum wells on GaN/sapphire templates with spacers of different thickness and thus different temperature of the surface. The color code spans the wavelength range from 390 nm (violet) to 460 nm (dark red).

: Fig. 2: Dependence of PL wavelength on the wafer temperature measured by Pyro 400 (red). The pocket temperature shows no correlation with the PL wavelength.

Growing GaN layers on sapphire substrate by metal organic vapor phase epitaxy leads to bowing of the wafers due to lattice mismatch and thermal expansion coefficent differences between substrate and layer. Larger wafers (for example 150 mm) that are increasingly used in production can bow during growth by several 100 µm depending on the layer design. Such bow also affects the temperature of the wafer that cannot be assessed by conventional pyrometry in the infrared as the material is transparent. Inhomogeneous wafer temperature leads to reduced yield since incorporation in InGaN-based LED and laser heterostructures and thus the emission wavelength is very sensitive to the temperature at the growth front.

In cooperation with LayTec as manufacturer of in-situ sensors, FBH has tested a new temperature sensor (Pyro 400) that measures the surface temperature of InGaN by pyrometry at around 400 nm. InGaN/GaN quantum wells were grown on GaN templates on 100 mm diameter sapphire substrates monitoring the temperature of the wafer surface as well as that of the pocket holding the wafer. Thin sapphire spacers under the wafer were used to increase the temperature offset between wafer and pocket. The PL wavelength of the quantum wells depend on the spacer thickness and show a redshift by 40 nm for 430 nm spacer (Fig. 1). While the temperature of the pocket is only marginally affected by the spacers, the wafer temperature differs by about 40 K between a wafer directly lying in the pocket and the one with 430 µm distance. This shift of PL wavelength is correlated to the wafer temperature with 1.1 nm/K. Since the temperature difference between pocket and wafer surface is affected, for example, by the thickness grown on the susceptor, a precise knowledge of the wafer temperature is indispensable for an exact adjustment of the emission wavelength of the devices.

Publication

V. Hoffmann, A. Knauer, F. Brunner, S. Einfeldt, M. Weyers, G. Tränkle, K. Haberland, J.-T. Zettler, M. Kneissl "Uniformity of the wafer surface temperature during MOVPE growth of GaN-based laser diode structures on GaN and sapphire substrate", J. Cryst. Growth, vol. 315, no. 1, pp. 5-9 (2011).

GaN-HEMTs in Envelope Tracking systems for telecommunications

FBH research: 07.06.2012

: Fig. 1: Two-tone signal and illustration of the losses (grey) for a fixed supply voltage system (green) compared to a modulated supply voltage system (red).

: Fig. 2: Circuit topology of envelope tracking system based on a HSA as voltage modulator, switch part of the HSA see red rectangle.

In the past years, much effort has been devoted to increase the efficiency in base-station systems for telecommunications. The high peak-to-average-power-ratio (PAPR) signals used in modern wideband systems like WCDMA or long-term-evolution (LTE) force the power amplifiers (PAs) to work in a very low efficient back-off region to meet the system linearity requirements. Therefore, much attention is now devoted to increase the back-off efficiency. One way to achieve back-off efficiency is to modulate the supply voltage with the instantaneous power of the signal, i.e. the envelope. The concept is known as envelope tracking (ET) and the principle is shown in Fig. 1.

When the instantaneous signal power in a modulated signal is low, the voltage difference between supply voltage (Vdd) and signal is large, leading to large losses. If the supply voltage for the amplifier is varied with the instantaneous power of the signal instead, the losses can be minimized and a very efficient transmission of the signal is achieved. One of the most promising transmitter topologies to achieve modulation with the required bandwidth and power is the so-called hybrid switching amplifier (HSA) as shown in Fig. 2.

Fig. 2. Circuit topology of an envelope tracking system based on a HSA as voltage modulator. The switch part of the HSA can be seen in the red rectangle. It consists of a wideband linear amplifier connected to a highly efficient switching stage shown in the red rectangle. When the RF power amplifier (PA) requires more current, the voltage drop across the sense-resistor and the switch (SW) goes active and provides the required current.

Traditionally the switching stage is implemented in Si-MOSFET technology. Therefore, a project was initiated to investigate the FBH GaN-HEMT technology as replacement for Si-MOSFET in hybrid switching amplifiers for wide bandwidth signals. A modular design where the alternative switch technologies could be tested under equal conditions was developed. The project was successful and although un-optimized for the modulator design, the GaN-technology showed already similar performance as Si-MOSFETs. The work was presented at the German Microwave Conference (GeMiC) in Ilmenau in March 2013 and received the best paper award.

The work is now being continued and an improved GaN-HEMT switching stage is being developed. A more powerful HSA that can handle 30 W at a 28 V supply voltage is also being designed. Together with a GaN-HEMT based PA optimized for supply modulation, which was also presented at GeMiC, it will be tested as a full envelope tracking (ET) supply modulated system in the coming months.

In parallel to the design work on the supply modulator, there are ongoing activities to improve the performance of the FBH GAN devices for supply modulated applications. Measurement and post-processing methods have been developed that emulate supply-modulated conditions directly during on-wafer characterization. These were presented last year in the Frequenz magazine.

Publications

R. Perea-Tamayo, O. Bengtsson, P. Landin (University of Gävle, Sweden), W. Heinrich, "A modular hybrid switching amplifier for wide-bandwidth supply-modulated RF power amplifiers", 7^th German Microwave Conference (GeMiC), 2012, 12-14 March 2012, IEEE Conference Publication (2012).

A. Raemer, O. Bengtsson, W. Heinrich, "Software optimization of a supply modulated GaN-amplifier for baseband access ET systems", 7^th German Microwave Conference (GeMiC), 2012, 12-14 March 2012, IEEE Conference Publication (2012).

O. Bengtsson, S. Chevtchenko, R. Doerner, P. Kurpas, W. Heinrich, "Load-Pull Investigation of a High-Voltage RF-Power GaN-HEMT Technology in Supply Modulated Applications", Frequenz. Volume 65, Issue 7-8, Pages 217–224, ISSN (Online) 2191-6349, ISSN (Print) 0016-1136, DOI: 10.1515/freq.2011.030, August 2011.

Increased emission power of 380 nm UV-LEDs

FBH research: 23.05.2012

: Fig. 1: PUI characteristics of a top-emitting 380 nm UV-LED structure with broad (solid line) or conventional n-AlGaN blocking layer (segmented line) underneath the active region.

: Fig. 2: PUI characteristics of a top-emitting 380 nm UV-LED structure grown on a low threading dislocation density GaN/sapphire template (solid line) instead of a conventional template without defect reduction (segmented line).

The FBH is developing top-emitting UV-LEDs for sensor applications in cooperation with TU Berlin and the company Jenoptik. The work is performed within the regional growth core Berlin WideBaSe and focuses on the development of efficient LEDs with emission wavelengths between 360 and 380 nm. It was possible to increase the efficiency of the epitaxial structure by applying a couple of growth and heterostructure adjustments.

Since high emission power at low diode current densities are required for the desired application, a single quantum well structure was used as active region instead of a 5x InGaN/AlInGaN MQW. In addition, the AlGaN blocking layers underneath and on top of the active region that prevent the overflow of minority charge carriers were adjusted in order to improve the majority carrier injection. By increasing the thickness of the AlGaN layer and decreasing the Al mole fraction, the electron injection into the active region is improved as a consequence of the modified band structure. Fig. 1 shows the characteristics of processed top-emitting LEDs with a conventional and a wide n-AlGaN blocking layer underneath the active region.

Furthermore, GaN-based devices on sapphire substrate suffer from the high number of dislocations threading from the GaN/substrate interface to the active region. By applying a defect reducing GaN buffer growth technique, the threading dislocation density in the active region could be reduced from 10⁹ cm^-2 to 3x10⁸ cm^-2. Thus, the recombination efficiency of the quantum wells could be increased. Fig. 2 shows the characteristics of identical LED heterostructures on a conventional GaN/sapphire template compared to a template with reduced defect density. By combining the optimizations it was possible to increase the external quantum efficiency of an LED structure emitting at 380 nm by more than a factor of two. In a next step, it is planned to shift the emission wavelength of the optimized heterostructure toward 360 nm, which is required for sensor applications.

Publications

A. Knauer, H. Wenzel, T. Kolbe, S. Einfeldt, M. Weyers, M. Kneissl, G. Tränkle, "Effect of the barrier composition on the polarization fields in near UV InGaN light emitting diodes", Appl. Phys. Lett., vol. 92, no. 191912 (2008).

A. Knauer, T. Kolbe, S. Einfeldt, M. Weyers, M. Kneissl, and T. Zettler, "Optimization of InGaN/(In,Al,Ga)N based near UV-LEDs by MQW strain balancing with in-situ wafer bow sensor", phys. stat. sol. (a), vol. 206, no. 2, pp. 211-214 (2009).

T. Kolbe, A. Knauer, H. Wenzel, S. Einfeldt, V. Kueller, P. Vogt, M. Weyers, and M. Kneissl, "Emission characteristics of InGaN multi quantum well light emitting diodes with differently strained InAlGaN barriers", phys. stat. sol. (c), vol. 6, no. S2, pp. S889-S892 (2009).

T. Kolbe, T. Sembdner, A. Knauer, V. Kueller, H. Rodriguez, S. Einfeldt, P. Vogt, M. Weyers, and M. Kneissl, "Carrier injection in InAlGaN single and multi-quantum-well ultraviolet light emitting diodes", phys. stat. sol. (c), vol. 7, no. 7-8, pp. 2196-2198 (2010).

T. Kolbe, T. Sembdner, A. Knauer, V. Kueller, H. Rodriguez, S. Einfeldt, P. Vogt, M. Weyers, and M. Kneissl, "(In)AlGaN deep ultraviolet light emitting diodes with optimized quantum well width", phys. stat. sol. (a), vol. 207, no. 9, pp. 2198-2200 (2010).

DFB diode lasers for precision spectroscopy of potassium (767 nm)

FBH research: 07.05.2012

: Fig. 1: Single-quantum well, RW DFB diode laser with 1.5 mm chip length emitting at 767 nm: power, voltage, and conversion efficiency vs. injection current at a mount temperature of 20°C.

: Fig. 2: Emission spectrum vs. injection current of the same diode laser. Individual spectra are normalized to their peaks so that the color coding displays the side mode suppression ratio (range: 0 … 40 dB).

: Fig. 3: FWHM linewidth and intrinsic linewidth (FWHM) of the same diode laser determined by means of a self-delayed heterodyne linewidth measurement. The intrinsic linewidth is inferred from the white noise floor of the frequency noise power spectrum.

FBH has developed narrow linewidth, high-power distributed feedback (DFB) diode lasers for quantum optics experiments on ensembles of ultra-cold potassium atoms. Based on its latest results on the realization of DFB lasers [1] for Rubidium Bose-Einstein condensation (BEC) and atom interferometry applications, FBH has now developed DFB lasers that are suited for the corresponding experiments on potassium. This development was carried out within the framework of a project supported by the German Space Agency DLR. Ultimately, these activities aim at testing the equivalence principle by comparing the free fall of rubidium and potassium ultra-cold atomic ensembles by means of an atom interferometric measurement in space [2].

The development of narrow linewidth, high-power DFB diode lasers becomes a more and more challenging task in terms of laser design and fabrication technology as the emission wavelength is shifted from 780 nm towards shorter wavelength. This is due to increased absorption in the overgrown gratings with decreasing wavelengths. An improved design of the grating layers has been implemented and resulted in better grating and device performance.

The recent results obtained at FBH for single-quantum well, ridge waveguide DFB diode lasers emitting at 767 nm clearly show an electro-optical performance that is comparable to the performance of the DFB diode lasers for 780 nm. With a 1.5 mm long chip an output power of more than 150 mW can be reached at an injection current beyond 250 mA, see Fig. 1. The slope efficiency corresponds to 0.68 W/A and is hence comparable to 0.6 A/W … 0.8 A/W recently reached with DFB lasers for 780 nm. Depending on the actual chip temperature the wavelength can be current-tuned continuously by more than one nanometer with single mode emission being maintained, see Fig. 2 . The side-mode suppression ratio reaches 40 dB at injection current settings beyond 150 mA. A self-delayed heterodyne linewidth measurement reveals a short term linewidth (10 µs) of about 1 MHz full-width-at-half-maximum (FWHM) and an intrinsic linewidth of a few 10 kHz FWHM at large injection current settings, see Fig. 3. The instrinsic linewidth is deduced from the white noise floor of the frequency noise power spectrum and hence excludes technical noise, mostly flicker noise of the current source.

This work is supported by the German Space Agency DLR with funds provided by the Federal Ministry of Economics and Technology (BMWi) under grant number 50WM0940.

Publications

[1] T.-P. Nguyen, M. Schiemangk, S. Spießberger, H. Wenzel, A. Wicht, A. Peters, G. Erbert, G. Tränkle, "Optimization of 780 nm DFB diode lasers for high-power narrow linewidth emission", accepted for publication in Appl. Phys. B

[2] T. van Zoest, et al., "Bose-Einstein Condensation in Microgravity", Science, 328, 1540-1543 (2010).

Gallium-nitride-based high-frequency ICs with improved efficiency

FBH research: 23.04.2012

: Fig. 1: Chip photo of a two-stage X-Band GaN power amplifier MMIC

: Fig. 2: Measurement data of the two-stage X-Band GaN power amplifier MMIC from Fig. 1: Output power, gain, and PAE of the final stage as a function of input power at 10 GHz

Power amplifiers (PA) are key components of any communication, radar and satellite system. As the last element in the transmitter chain before the antenna they dominate the overall properties. The most important figures of merit are output power P_max and power-added efficiency PAE (PAE: Power Added Efficiency). The PAE indicates how much of the consumed power is actually available for the application and how much is dissipated into (thermal) losses. High efficiency is a key issue in view of environment (CO₂ emission) as well as system performance.

The combination of high power, high efficiency and high operating frequency is a problem for most common semiconductor technologies, which is due to physical constraints. In this regard, gallium nitride (GaN) outperforms most of its competitors as it offers both high breakdown electric fields and high electron mobility. This makes it the ideal choice for microwave power amplifiers. It allows realizing PAs with previously unattainable values for output power and PAE.

Various radar and satellite systems operate in the X-band, the frequency range from 8 to 12 GHz. In this frequency range, amplifiers are commonly built as monolithic circuits (MMICs, Monolithic Microwave Integrated Circuits), because through monolithic realization critical tolerances and parasitic properties can be reduced. A GaN MMIC process is available at FBH which also serves this purpose. Recently, the performance of this process has been further improved by reducing the gate length of the GaN transistors to 0.25 μm. Devices achieve efficiencies of 50% and more at 10 GHz in deep-AB operation.

Fig. 1 presents a recently designed power amplifier MMIC realized by using this process. Because the GaN semiconductor layers are grown on silicon carbide (SiC), the substrate is transparent and the metal structure of the circuit seems to float. In order to achieve high gain, a two-stage design is employed. As can be seen from the measurement data in Fig. 2, this circuit reaches a maximum output power P_max of 11 W at 10 GHz. The maximum linear gain is approximately 25 dB and the efficiency of the final stage (PAE) almost 40%. These values can be enhanced further by appropriate circuit optimization. Work on this is ongoing.

Publication

Erhan Ersoy, Chafik Meliani, Serguei Chevtchenko, Paul Kurpas, Mathias Matalla, and Wolfgang Heinrich, "A High-Gain X-Band GaN-MMIC Power Amplifier", presented at 7th German Microwave Conference (GeMiC), Ilmenau, Germany, on 12-14 March 2012.

Efficient Microwave Power Transistors for L- and S-Band Applications

FBH research: 11.04.2012

: Fig. 1: Mounted GaN microwave power transistor of 100-W-class for L- and S-band applications

: Fig. 2: Focused-Ion-Beam (FIB) cross section through Source connected field plate structure

: Fig. 3: Properties of microwave power transistors of 50-W-class in dependence on operation voltage (28 V, 40 V and 50 V). a) Scaling of output power with increasing operation voltage. b) Power added efficiency (PAE) in dependence on operation voltage

Highly efficient GaN- based microwave power transistors are implemented into modern communication and radar systems to an ever increasing degree. Often complex microwave systems are only possible if compact and highly efficient devices can be integrated within a small volume. In cooperation with its spin-off company Berlin Microwave Technologies (BeMiTec) FBH develops highly efficient power transistors for output power levels of more than 100 W in the frequency range between 1 and 3 GHz. Well adjusted optimizing efforts in materials technology (epitaxy), device processing, general device design including layout as well as chip mounting techniques into a suitable microwave package are key for the development of such devices. Consequently our activities led to promising devices which are combining high power density and high absolute power levels with high efficiency. The devices are mounted into a microwave package according to fig. 1 and are available for implementation into microwave systems.

Decisive for the development of transistors showing high output power and efficiency at the same time has been the careful optimization of field plate technology along with the introduction of ballasting resistors between the individual power cells to avoid parasitic oscillations of the packaged power bars. The intention of field plates (see fig. 2) is to influence distribution and maximum intensity of the electric field in internal devices regions such that for all targeted device operation conditions the field maximum always stays below critical field levels. This is a prerequisite for an increase of device operation voltage (for example to 50 V) and for a linear scaling of output power with operation voltage as shown in fig. 3a). The high power added efficiency of close to 70% obtained along with these optimizations qualifies these devices for implementation in microwave amplifier systems operating at 40 V and above. The high operation voltage in turn leads to a higher level of optimum input and output impedance for power matching conditions and therefore enables either very broadband amplifier systems or, by combining multiple power transistors, very powerful microwave amplifiers with power levels above 200 W.

InP high-speed transistors and integrated circuits in transferred substrate technology

FBH research: 30.03.2012

: Fig. 1: FIB cross-section of a TS-HBT with emitter (e), base (b) and collector (c)

: Fig. 2: Extrapolated f_T and f_max of a TS-HBT

: Fig. 3: S-parameter measurements of a 90 GHz power amplifier in TS technology

Research on high-speed transistors is driven by applications for imaging and wide band communications. Recent technical advances of InP-based transistors with several hundred gigahertz (GHz) operating frequencies together with their outstanding material properties qualify them as key components in such systems.

At FBH, a transferred substrate (TS) technology has been established to optimize high frequency and power performance of InP heterojunction bipolar transistors (HBT). The 3" wafer-level process enables lithographic access to both the front- and backside of the HBT aligned to each other. The resulting linear device set-up in Fig. 1 eliminates dominant transistor parasitics and relaxes design trade-offs. The essential step for gaining access to both sides of the epitaxial structure is to completely remove the supporting substrate. Therefore, a robust adhesive wafer bonding procedure via benzocyclobutene (BCB) has been developed. It yields a homogenous, crack- and void-free composite matrix of transistors transferred on a wafer-level scale.

The optimized device topology manifests in excellent HBT performance. Transistors with 2× 0.8×5 μm² emitter area, as depicted in Fig. 2, feature f_T= 376 GHz and f_max = 385 GHz at breakdown voltages BV_CEO > 4.5 V. They combine high frequency performance with saturated output power P_out > 14.2 dBm @77 GHz and an inherently good matching to 50 Ω. The highly scalable device architecture is capable to even further increase high frequency as well as power performance in the future. Power amplifiers have been designed and realized in TS technology for 90 GHz operation. Their S-parameter measurements shown in Fig. 3 confirm a good agreement with modeling.

Currently, the innovative transistor set-up is utilized in an ongoing project to integrate InP-based circuits ontop of BiCMOS wafers heterogeneously.

Publication

T. Al-Sawaf, C. Meliani, W. Heinrich, and T. Krämer, "W-Band Amplifier with 8 dB Gain Based on InP- HBT Transferred-Substrate Technology", Proc. German Microwave Conference, Ilmenau, Germany, 12–14 March 2012.

Diode lasers for precision quantum optics experiments in space

FBH research: 13.03.2012

: Extended cavity diode laser for precision quantum optics experiments on board a sounding rocket

FBH has developed micro-integrated master-oscillator power-amplifier (MOPA) laser and extended cavity diode laser (ECDL) modules for experiments on Rubidium Bose-Einstein condensates on board a sound rocket to be launched in 2013. The MOPA concept is based on optimizing a low power distributed feedback (DFB) or distributed Bragg reflector (DBR) for stable narrow linewidth emission and amplification of its radiation by means of a separate power amplifier chip. The ECDL concept uses optical feedback from an external optical grating. This way, large cavity lengths can be realized which provide a reduction of the frequency noise linear spectral density by 1 to 2 orders of magnitude with regard to monolithic lasers.

Hybrid micro-integration technology is used to integrate laser chips, optics, and electronics on an aluminum nitride ceramic body that takes a volume of only 8 x 2.5 x 1.5 cm³. This amounts to a reduction of the form factor by 3 orders of magnitude with respect to commercial systems. Both module types further omit any moveable parts so that the requirements on mechanical stability for space applications can be met. These MOPA modules have already successfully passed vibration test at 8 g_RMS that simulate the mechanical stress of a sounding rocket launch. Further tests are pending.

The MOPA modules provide an optical power in excess of 1 W at 780.24 nm. Depending on the laser chip used as a master oscillator, a short term (10 µs) emission linewidth below 1 MHz and an intrinsic linewidth as small as a few 10 kHz can be realized. The ECDL provides an output power of typically 50 mW with a short term linewidth of well below 100 kHz and an intrinsic linewidth of a few kHz only. It is hence suited for applications with the most stringent requirements on spectral stability. Coarse frequency tuning is realized by thermally tuning the optical grating which provides a tuning range of approximately 80 GHz.

MOPA and ECDL technology can be transferred to other wavelengths in the 650 nm to 1100 nm wavelength range. It is also considered for the realization of lasers for portable optical clocks.

Publications:

E. Luvsandamdin, G. Mura, A. Wicht, A. Sahm, S. Spießberger, H. Wenzel, G. Erbert and G. Tränkle, "Micro-integrated ECDLs for precision spectroscopy in space", Int. Conf. on Space Optical Systems and Applications (ICSOS 2011), Santa Monica, USA, May 11-13, pp. 383-385 (2011).

Ch. Kürbis, A. Kohfeldt, E. Luvsandamdin, M. Schiemangk, S. Spießberger, A. Wicht, A. Peters, G. Erbert, G. Tränkle, "Mikrointegrierte Lasersysteme für die höchstauflösende Atomspektroskopie und die kohärente Nachrichtenübertragung im Weltraum", 60. Deutscher Luft- und Raumfahrtkongress, Bremen, Germany, Sep 27-29 (2011).

A 24-GHz-CMOS-VCO for Localization purposes

FBH Research: 23.02.2012

: Fig 1. Two versions of a voltage controlled oscillator (VCO) for 24 GHz in 130 nm CMOS technology.

: Fig. 2. Single sideband phase noise of both VCO versions vs. carrier offset frequency.

Just imagine two robot welders, together assembling a car body with high speed and high force. If they crash, the damage will be expensive: Not only the robots will be damaged, but also the assembly line will have to stop. In near future, such pitfalls can be avoided using two tiny localization blocks placed on each robot’s arm. These localization blocks are able to measure the distance to each other and to provide early warning of collisions.

The localization block is based on radar principle at 24 GHz and its core is a voltage controlled oscillator (VCO) generating the radar signal. For high localization accuracy in the range of a fraction of an inch it is necessary to have a radar signal with low phase noise. This is due to the fact that phase noise directly determines the localization accuracy. In the framework of the BMBF project LoWiLo (Low-power Wireless Sensor Network with Localization), such a 24 GHz low phase-noise VCO was developed at FBH [1]. Based on an advanced 130 nm CMOS technology, some versions of cross-coupled VCO were designed and characterized.

The two versions differ in the choice of the frequency-determining spiral inductor with respect to its quality factor and in the way the varactor is coupled, see Fig. 1. Fig. 2 presents the phase noise spectrum of both versions of the VCO. Clearly, version B exhibits a phase noise lower than that of version A by 10 dB at 100 kHz offset, which translates into enhanced localization accuracy.

Publication:

[1] Hossain, M.; Kravets, A.; Pursche, U.; Meliani, C.; Heinrich, W.: "A Low Voltage 24 GHz VCO in 130nm CMOS For Localisation Purposes in Sensor Networks," paper to be presented at German Microwave Conference 2012 in Ilmenau (Session 12, 13 March 2012).

New tool for materials characterization at FBH

FBH research: 13.02.2012

: Fig. 1: Analytics tool for the characterization of degradation processes.

: Fig. 2: High-resolution image of the vertical and lateral structure of a laser with integrated Bragg grating.

: Fig. 3: 20 µm thick HVPE-GaN layer on structured sapphire substrate; (a) CL image (defect density:1,3 x 10⁸ cm^-2), (b) surface structure with regular growth terraces and pits where dislocations terminate.

In the framework of the EFRE project "Application center for high frequency technologies" the Ferdinand-Braun-Institut recently installed a materials analytics tool for characterization of degradation processes in semiconductor devices (Fig. 1). The centerpiece of the tool is an Ultra+ high resolution scanning electron microscope (SEM) with a thermal field emission electron gun from Carl Zeiss NTS which is capable of high-resolution imaging of surfaces. During installation, a lateral resolution of 1.0 nm was demonstrated.

The SEM is equipped with several systems for the analysis of structural and optical properties of semiconductor layers and devices. Among these is an energy-dispersive X-ray spectrometer (EDXS) from Bruker Nano with an energy resolution of 125 eV for quantitative determination of composition of compound materials as well as a system for detection of electron-induced currents (EBIC) allowing for failure analysis of transistors and semiconductor laser diodes.

Moreover, a cathodoluminescence system from Gatan is attached to the microscope for characterization of the optical properties of laser chips in the temperature range from 80 K to 300 K before and after life testing. The Mono-CL4-system provides fast spectral mappings in the wavelength range of 200 nm to 1100 nm. This allows also for high spatial resolution to detect areas where material properties have changed due to the high electrical and optical load during operation.

Some examples illustrate the capabilities of the new tool that considerably improves the equipment potential for characterization of semiconductor layer structures and devices at FBH. Fig. 2 shows the vertical and lateral structure of a laser with an integrated Bragg grating for wavelength stabilization with a high resolution. The contrasts are caused by the different material composition of the etched grating and the waveguide layers showing self organization during growth in the second epitaxial step. The new SEM enables an improved material contrast together with a very good lateral resolution. In Fig. 3(a) 20 µm thick HVPE-GaN layer is depicted which was grown on a structured sapphire substrate. Fig. 3(a) is a cathodoluminescence (CL) image with dark spots at dislocation positions (defect density 1,3 x 10⁸cm^-2) and Fig. 3(b) shows the surface structure with regular growth terraces and small pits where the dislocations terminate at the surface.

Low-threshold GaN-based laser diodes

FBH research: 31.01.2012

: Fig. 1: Current-light power characteristic of a 44x nm RW laser diode with a ridge width of 1.5 µm under CW operation.

: Fig. 2: Threshold current density versus ridge width for laser diodes with two different etching depth of the ridge.

Laser diodes based on GaN are currently available on the market only for a limited number of specific wavelengths. FBH, together with TU Berlin und the company eagleyard Photonics, has started to develop laser diodes with customized wavelengths for use in atom spectroscopy. The current focus is on the mercury lines at 404.7 nm and 435.9 nm. The laser diodes will be operated in an external cavity, which involves a diffraction grating to precisely adjust the lasing wavelength. As required for the desired spectroscopic applications, the lasers have to show a low threshold current. Therefore, ridge waveguide (RW) laser diodes with a small ridge width of 1.5 µm and a resonator length of 600 µm have been fabricated. The narrow ridge is also essential to assure an optimum beam quality. Threshold currents as low as 40 mA have been obtained for devices emitting around 41x nm. The threshold voltage and slope efficiency in pulsed operation were 7.5 V and 0.5 W/A, respectively. Under continuous wave (CW) operation, an output power of 40 mW has been reached as shown in Fig. 1 for a device emitting at 440 nm.

A systematic study of numerous laser diodes has shown that the lasing threshold very much depends on the geometry of the ridge waveguide. The almost square-sectioned ridge waveguide is formed by etching several hundreds of nanometers deep into the semiconductor surface. Its purpose is to laterally confine both the optical mode and the vertical current path. Fig. 2 shows the threshold current density as a function of the ridge width for two batches of laser diodes whose etching depth of the ridge differs only by 175 nm. Although the impact of the ridge depth on the threshold vanishes when the ridge width increases, narrow ridge lasers exhibit more than a factor of two higher threshold current densities in case of shallow etched ridges. Systematic near field and far field measurements along with two-dimensional electro-optical simulations of the devices have been started in collaboration with the NUSOD institute. The anti-guiding effect originating from the high carrier density during lasing, the optical absorption in the region of the lateral mode tails, and the lateral current spreading are considered. Although a comprehensive explanation of the effect still needs to be found, a huge current spreading in the layer structure seems to be unlikely. Rather, the weakening of the mode confinement by the anti-guiding effect is currently favored.

Publications:

L. Redaelli, J. Piprek, M. Martens, H. Wenzel, C. Netzel, A. Linke, Y. V. Flores, S. Einfeldt, M. Kneissl and G. Tränkle, "Effect of ridge waveguide etch depth on laser threshold of InGaN MQW laser diodes", Proc. SPIE, to be published in 2012.

C. Netzel, S. Hatami, V. Hoffmann, T. Wernicke, A. Knauer, M. Kneissl and M. Weyers
"GaInN quantum well design and measurement conditions affecting the emission energy S-shape", phys. stat. sol. (c), vol. 8, no. 7-8, pp. 2151-2153 (2011).

Tapered laser module at 636 nm with > 500 mW for display applications

FBH research: 18.01.2012

: Red-emitting laser module for display applications

Red-emitting diode lasers are needed as compact light sources not only for displays, but also for medical treatment and sensing applications. An important property required for these applications is high radiance (i.e. brightness). In the framework of the InnoProfile initiative „Hybride Diodenlasersysteme“, the FBH succeeded in developing a compact laser module with more than 500 mW output power at 636 nm wavelength. The module emits a nearly diffraction-limited, collimated beam with a radiance of more than 19 MW/cm²/sr. Perceived by a human eye, this corresponds to a luminance of more than 27 TCd/m² [1], which is more than 10,000 times brighter than the luminance of the sun (1.6 GCd/m²) – and a new record value for red-emitting diode lasers.

This result was made possible by the development of red-emitting tapered lasers and their subsequent mounting on CVD-diamond heat spreaders with two contacts. This allows optimal heat extraction from the chip and individual currents through the ridge-waveguide and taper sections of the laser. Additionally, the radiation of the chip was formed by micro-lenses mounted inside the module. The shaped beam could be coupled into an optical fiber featuring a small aperture of only 16 µm with an efficiency of more than 75%. Using this fiber, the laser light can then be guided to its point of use, e.g. a projector head of a laser display system.

Publication:

[1] G. Blume, C. Kaspari, D. Feise, A. Sahm, B. Sumpf, B. Eppich, and K. Paschke, “Tapered diode laser modules at 638 nm with efficient fiber coupling “, IEEE Phot. Technol. Lett. (submitted 05.01.2012).

3D EM simulation is a must in the design of high-frequency CMOS circuits

FBH research: 13.01.2012

: Fig. 1: Multiplier circuit (left) with different passive elements and ports replacing the active elements, ports and metal stack are emphasized on the right side.

: Fig. 2: S parameter S₂₁ as a function of frequency; results for the multiplier circuit from measurements compared to EM simulations and circuit simulator (SpectreRF) using the manufacturer's data

Due to the rapid advances in the miniaturization of semiconductor processes (according to Moore’s law), high-end CMOS circuits can be used today in a frequency range far beyond 10 GHz. In contrary to low-frequency and digital circuits, their functionality is no longer defined primarily by the active elements, i.e., the transistors, but the passive elements such as inductors, capacitors and transmission lines take significant influence. Hence the electrical behavior of these elements must be included already in the basic design steps.

The CMOS processes realize the passive elements through a stack of metal layers separated by dielectrics on top of the semiconductor substrate with the active elements. During circuit design, it is important to use accurate models for the passive structures in the circuit simulator (eg. SpectreRF). It is also indispensable to check the circuit functions in a final step in this way. In this case, a simulation of the entire circuit must be carried out using 3D EM simulation. Active elements are not considered in the EM simulation and are replaced by internal ports. In a later step, the designer implements both the EM simulated data and the properties of the active elements in the circuit simulator in order to obtain an accurate description of the circuit. An example of such a circuit is the multiplier shown in fig. 1.The large ratios of cell sizes and the high number of cells is a challenging task for the simulation, resulting in a mesh with typical 25 million cells. Fig. 2 shows measurement data together with simulation results of the standard models, as provided by the foundry, as well as the results of the in-house EM simulation. As one can see, the results of the standard models deviate considerably from the measurements and using 3D EM simulation leads to a significantly better fit with the measured data.