csjulcover 5/7/06 8:58 am page 1 compound semiconductoriopp.fileburst.com/cs/cs_12_06.pdf ·...

COMPOUNDSEMICONDUCTOR

July 2006 Volume 12 Number 6

C O N N E C T I N G T H E C O M P O U N D S E M I C O N D U C T O R C O M M U N I T Y

Cree kicks off SiCconsolidation withIntrinsic deal p5

IntegrationTelecom-boom trendmakes a reappearance inoptical networks. p14

French fancyISBLLED delegates get a taste of non-polar GaN in Montpellier. p20

GAN ELECTRONICS

Big guns take aim at wide-bandgap sector

HEADLINE NEWS TECHNOLOGY CONFERENCE REPORT

CSJulCover 5/7/06 8:58 am

SiC SUBSTR ATES • SiC EPITA X Y • GaN EPITA X Y

©2006 INTRINSIC Semiconductor Corporation

At INTRINSIC, one of our most important deliverables is service. Sure, we

supply high quality substrates and epitaxial services as well — including one

of our newest offerings, a zero-micropipe SiC substrate. But we believe it’s

our dedication to customer service that has made us the go-to supplier for

the world’s most demanding manufacturers. It’s how we keep customers

not merely supplied, but satisfi ed.

Call us today at 1.703.437.4000, or e-mail us at [email protected].

We want to serve as your materials and solutions partner.

Visit www.intrinsicsemi.com to learn more.

The Measure of Quality.

INTRINSIC offers wafers in 2″, 3″, and 4″ sizes.

For some, customer service is a department.For INTRINSIC, it’s a way of doing business.

INTRINSIC ad_02.indd 1 2/13/06 4:14:08 PM

mailto:[email protected]

http://www.intrinsicsemi.com

J U L Y 2 0 0 6 V O L U M E 1 2 N U M B E R 6

C O N N E C T I N G T H E

C O M P O U N D

S E M I C O N D U C T O R

C O M M U N I T Y

Compound Semiconductor July 2006 compoundsemiconductor.net 1

TECHNOLOGY

14 Optical integration moves back onto the agenda:Photonic integration was one of the hot new ideas that emerged during the telecoms boom, but it was subsequently shelved by cash-strapped component vendors. Now, as Roy Rubenstein reports, the technology could be set to make a big revival.

17 Simulations provide additional insights into GaN HFET reliability: GaN HFET reliability can be improved by adding field plates to the structures. This is believed to reduce carrier trapping at the surface, but simulations by Ric Borges, Nelson Braga, Bo Wu and Vidas Mickevicius of Synopsys show that it also decreases the electron temperature and bulk trapping throughout the device.

20 Researchers seek material solutions to GaN deficiencies at ISBLLED 2006: Researchers are turning to alternative capping layers and GaN and non-polar sapphire substratesto push the output power of LEDs to levels suitable for everyday lighting. Richard Stevenson reports.

23 Fujitsu proves reliability of GaN HEMTs: Fujitsu’s GaN program has led to proven device reliability, a fabrication process using low-cost conducting 3 inch SiC substrates, and record output power from MIS-HEMTs, says Toshihide Kikkawa.

27 Product Showcase

28 Research Review: German collaboration extends room-temperature VCSEL emission to 2.3 µm...Cermet makes ZnO LEDs on n-type substrates...InAs HEMT prototype breaks efficiency record.

Hot topic: materialsRichard Stevenson reports from theISBLLED conference, which was held inMontpellier, France, in May. p20

Photonics revivalOptical integration is back in fashion astransmission speeds increase toward100 Gbit/s. p14

Main cover image: RF Micro Devices is now utilising part of its Greensboro,NC, fab to manufacture GaN-based transistors on 3 inch SiC wafers.

Compound Semiconductor’s circulation figures are audited by BPA International

INDUSTRY

5 Headline news: Novalux bags $22 million finance with supply deal...Cree prepares to buy Intrinsic for $46million...Japanese firms target new GaN FET applications.

6 The Month in RFICs: Nitronex attracts $22 million infunding...Cree and RFMD launch a host of GaNtransistors...Filtronic invests $83 million in fab...Outsourcing strategies suit AWSC and IQE.

8 The Month in HB-LEDs: Cool-white chip delivers higherlevels of efficacy...US Department of Energy showers $10 million on solid-state lighting... Avago aims for high-power sector...Highlink hikes up wafer capacity.

10 The Month in Optoelectronics: Californian optical component maker doubles wafer fab capacity in Texas...iSuppli analysis predicts Blu-ray player deadlock...Sharp urges III-V solar firms to co-operate...Ten-layer laser has broadband promise...Datacoms push upturn in active components market...IPG steps up capacity with Veeco MBE reactor kit.

13 Behind the headlines: Carolina’s big guns board the GaN train With some major developments at Cree, RF Micro Devices and Nitronex, there’s no doubting where the push to commercialize GaN-based RF devices is coming from.

High-power lasersThe Veeco MBE reactor kit will help IPGincrease its diode manufacturing volumesin the future. p11

Brave new worldCree and RF Micro Devices havelaunched new ranges of GaNtransistors, putting them in directcompetition with each other for thefirst time. p6

CSJulContents1 5/7/06 9:22 am Page 1

http://compoundsemiconductor.net

compoundsemiconductor.net July 2006 Compound Semiconductor2

E D I T O R I A L

Ready for prime timeIt appears that GaN’s time has come. The major compoundsemiconductor players in both Japan and the US have signaledtheir intention to attack the as-yet unmaterialized market forGaN-based microelectronic devices.

Some of the biggest names in the business – Cree, RF MicroDevices and Toshiba – have either firmed up plans for ramping

up manufacturing, released products for RF applications or outlineddevelopment roadmaps, and the omens look good for the wide-bandgaptechnology. RFMD is traditionally regarded as a fairly conservativecompany, and would not be launching a raft of new GaN-based products forboth WiMAX and cellular infrastructure applications without someseriously good information on the likely buying habits of its customer base.

If it all goes to plan, 2006 could well go down as the year in which GaNmicroelectronics started to hit the big time. RFMD is confident that its top-tier cellular infrastructure customers are now convinced of not just the

technological potential of the wide-bandgapmaterial, but also the reliability of the productsthat are based on it. Now armed with what itbelieves to be a very convincing set of reliabilitystats, RFMD is turning on the evangelism.

Let’s hope that the charm offensive works,because there are plenty of suppliers of SiCsubstrates who will be more than willing to

produce the wafers needed to support any future volume ramp. Thatsupplier base is already starting to consolidate, with the news that Cree is toswallow up the Virginia firm Intrinsic Semiconductor, a pioneer when itcomes to defect reduction in wide-bandgap material.

Intrinsic’s zero-micropipe (ZMP) substrates represent the cutting edge ofthe SiC substrate industry, and could turn out to be a crucial technology forensuring the reliability of electronic GaN devices that are fabricated on topof them. The only problem would have been the cost of producing them.

This is where the hook-up between Cree and Intrinsic really starts tomake sense. Cree’s volume manufacturing prowess, coupled withIntrinsic’s technology, should lead to larger, lower-defect substrates, whichwill be ideal for the likes of RFMD to produce reliable, high-performancedevices at a price point that the cellular industry will find hard to resist.

Michael Hatcher Editor

“Armed with a veryconvincing set ofreliability stats,RFMD is turning onthe evangelism.”

Air Products & Chemicals Inc 4

Bandwidth Semiconductor 8

Cedova 7

CremIC 25

CS Week 3

INTRINSIC Semiconductor IFC

KLA-Tencor 12

ORS 22

Raboutet 9

Riber 16

Shiva Technologies 26

Synova 25

Taylor Hobson 26

Veeco OBC

Advertisers’ Index

Editor Michael [email protected]: +44 117 930 1013. Fax: +44 117 925 1942

Features editor Richard [email protected]: +44 117 930 1192

Consulting editor Tim [email protected]: +44 117 930 1233

Senior sales executive David [email protected]: +44 117 930 1032. Fax: +44 117 920 0977

Business development manager Rosemarie [email protected]: +1 215 627 0880. Fax: +1 215 627 0879

Circulation manager Jackie [email protected]: +44 117 930 1218. Fax +44 117 930 1178

Publisher Sarah [email protected]: +44 117 930 1020

Senior production editor Ruth LeopoldAd production Joanne Derrick, Mark TrimnellArt director Andrew GiaquintoTechnical illustrator Alison ToveyPublishing director Richard Roe

SubscriptionsAvailable free of charge to qualifying individualsworking at compound semiconductor fabs andfoundries. For further information visitcompoundsemiconductor.net/subscribe. Subscriptionsfor individuals not meeting qualifying criteria:individual £86/$155 US/7125; library £193/$348US/7280. Orders to Compound Semiconductor, WDIS, Units 12 & 13, Cranleigh Gardens IndustrialEstate, Southall, Middlesex UB1 2DB, UK. Tel: +44 208 606 7518; Fax: +44 208 606 7303. General enquiries: [email protected].

9173 average total qualified circulation**December 2005 BPA audit statement

Editorial boardMayank Bulsara Atlas Technology (USA); Andrew Carter Bookham Technology (UK); Jacob TarnEpistar/Gigacomm (Taiwan); Ian Ferguson GeorgiaInstitute of Technology (USA); Toby Strite JDSU(USA); Mark Wilson Motorola (USA); Dwight StreitNorthrop Grumman (USA); Joseph Smart Crystal IS(USA); Colombo Bolognesi Simon Fraser University(Canada); Shuji Nakamura University of California atSanta Barbara (USA)

©2006 IOP Publishing Ltd. All rights reserved.

US mailing information: Compound Semiconductor(ISSN 1096-598X) is published 11 times a year for $148 by Institute of Physics Publishing, Dirac House,Temple Back, Bristol BS1 6BE, UK. Periodicalspostage paid at Middlesex, NJ 08846. POSTMASTER: send address corrections toCompound Semiconductor, c/o PO Box 177,Middlesex, NJ 08846. US agent: Pronto MailersAssociation Inc, 200 Wood Avenue, PO Box 177,Middlesex, NJ 08846.

CSJulLeader2 5/7/06 9:29 am Page 2










COMPOUND SEMICONDUCTORWEEK 2006

November 12–15, 2006San Antonio,Texas, USA

Event organized by

THE KEY CONFERENCENOVEMBER 13–14, 2006PART OF COMPOUND SEMICONDUCTOR WEEK 2006

This two-day conference will be packed with top invitation-onlyspeakers from the key players in the compound semiconductorindustry, respected market analysts and cutting-edge start-ups, and itwill focus on the following key areas:

• GaAs–silicon convergence;

• silicon carbide power devices;

• alternative III-nitride technologies and applications;

• multi-junction solar cells;

• new laser application markets.

Confirmed speakers include senior representatives from:

IBM • Sony • JDSU • Cree • Freescale • Massachusetts Institute

of Technology • Telesoft Ventures • Infineon Technologies •SemiSouth • Yole Developement • Kyma Technologies • Group4

Laboratories • SEMATECH • IMEC • GA Tech • Sensor ET •NRL • Spectrolab • APT • Emcore • OSU

If you need to know about the materials, technologies and applicationsthat will drive the compound semiconductor market of the future, makesure you don’t miss this event.

Sign up to receive regular program updates online

compoundsemiconductor.net/csweek

GOLD SPONSORS

November 12–15, 2006 San Antonio, Texas, USA

COMPOUND SEMICONDUCTORWEEK 2006Conferences and Exhibition

http://compoundsemiconductor.net/csweek

What have we contributed to the electronics industry?

About this much.

©2006 Air Products and Chemicals, Inc.

Air Products team members,

Soo Yon Lee and Young Shin

Kwak, use their experience

and understanding to deliver

smart, efficient solutions in

Asia. They’re great examples

of our company’s commitment

to what matters most to you.

In this business, small is huge. That’s

why, for over 30 years, companies

throughout the world have counted

on Air Products’ materials, equipment

and integration expertise.

We offer the broadest range of

bulk and specialty materials in the

industry, including those from

Schumacher,® ACT® and Solkatronic.®

Our experienced team will thoroughly

integrate our gas and chemical

materials into your manufacturing

processes to optimize results. You

can even tap into our experience online

at www.airproducts.com/focused8.

Register there for our free e-newsletter,

Solution Updates, and get a free

USB drive filled with in-depth

technical papers and industry

information. It’s just one more way

we’re doing big things for you.

tell me more www.airproducts.com/focused9

http://www.airproducts.com/focused8

http://www.airproducts.com/focused9


I NDUSTRY H E A D L I N E N E W S

Laser chip developer Novalux has completeda new round of equity financing worth “up to”$21.7 million.

This includes a substantial investment fromUnaxis Optics, which is based in Liechtenstein.The two companies have also signed a devel-opment and licensing agreement under whichCalifornia-based Novalux will supply UnaxisOptics with red, green and blue laser chips.

The European firm will mass-producemodules based on the lasers and distributethem throughout the display industry forapplications including high-end televisionsand personal projectors.

Earlier this year Novalux, which owns aGaAs wafer fab in Sunnyvale, CA, joinedforces with display industry giant Seiko-Epson, to which it is also supplying laser chips.

Features provided by projector systemspowered with Novalux’s extended-cavitysurface-emitting lasers (NECSELs) are saidto include images with brighter, truer colorsthan those from “conventional” lamp systems,and higher reliability. Novalux laser sourcescan reduce the total number of componentsrequired and this could allow volume produc-tion of top-performance optical devices atmore competitive prices, Unaxis Optics stated.

As well as large displays and micro projec-tors for high-definition TVs, the laser modulescould also find applications in cell phones andgaming consoles, mini projectors, and head-up displays that project driver information ontovehicle windscreens.

“The laser technology from Novalux is anideal addition to our own technologies andproducts,” said Thomas Limberger, CEO ofUnaxis. “This development is an innovation

that should reshuffle the market for front andrear projection displays and represent thefuture direction of projection systems.

“The partnership with Novalux makes sensefor Unaxis since the technology is comple-mentary to our product portfolio and can beapplied across the range of micro display andscanning devices,” he added.

“Our NECSELtechnology is ideal for manycurrent and future projection display applica-tions,” said Jean-Michel Pelaprat, chair andCEO of Novalux, adding: “The new arrange-ment will provide consumers with a lower-cost, longer-lifetime solution than other lightsources, along with an unmatched viewingexperience. This partnership positions bothcompanies to launch RGB laser lighting intothe projection display marketplace.” Matthew Peach is a contributing editor atOptics.org.

Novalux bags $22 millionfinance with supply deal

Cree is to acquire fellow SiC substrate devel-oper Intrinsic Semiconductor in a deal that isvalued at $46 million. The Durham, NC-basedchipmaker will stump up $43.5 million in cash,while the remaining $2.5 million is to be paidout through the assumption of outstandingIntrinsic stock options.

Although both companies manufacture andsell SiC substrates, Intrinsic has pioneered thedevelopment of so-called “zero-micropipe”material, a high-performance product line thathas enabled the development of advanced SiCmicroelectronics.

Cree says that the acquisition will combinethat expertise with its own volume manufac-turing experience, stimulating the develop-ment of large-diameter SiC wafers beyond the3 inch material that is starting to becomewidely used today, enabling high-quality 4 inchand possibly even 6 inch production.

Cree itself has almost completed the con-version of its LED production lines to 3 inchmaterial, while RF Micro Devices will manu-facture its new range of GaN-based productson wafers of the same size.

“These substrates should not only supportour cost roadmap for LEDs, but – more impor-tantly – they should also enable us to morerapidly commercialize higher-power devicesfor motor-control applications and hybrid[electric] vehicles,” commented Cree chiefChuck Swoboda.

Cengiz Balkas, the man who has ledIntrinsic until now, added that the combinationof Cree and Intrinsic expertise would yield aunique opportunity to make a new generationof cost-effective SiC devices earlier than hadpreviously been expected.

Intrinsic, which is based in Dulles, VA,recently began sampling 4 inch SiC substratesand has device prototyping capability. LikeCree, the company uses a physical vapor trans-port method to produce its substrates.

Cree prepares to buyIntrinsic for $46 million

V E N T U R E C A P I T A LM A T E R I A L S

Unaxis Optics will mass-produce television backlightmodules featuring the RGB laser sources from Novalux.

Among a raft of new GaAs- and GaN-basedproducts aimed at the wireless communica-tions infrastructure business, the InternationalMicrowave Symposium, held in San Franciscoin June, saw Toshiba reveal details about itsplans for the development of wide-bandgapmicroelectronics.

The Japanese company says that havingproved the basic design concept with itsdevelopment of GaN-based power FETs withan output of 174 W at 6 GHz, it is currently

refining the technology to overcome heatdissipation issues. It then plans to develop aseries of GaN products for satellite and point-to-point communications.

But perhaps surprisingly, the first appli-cation area that Toshiba is considering ismedicine. “Initially, our plans are to developan X-band device in the 9 GHz, 50 Wrange formedical applications, targeted for the thirdquarter of 2006,” said Ichizo Kobayashi, seniormanager at Toshiba’s electronic systems

department. Kobayashi believes the devicescould be used inside RF heating systems to treat tumors.

Meanwhile, Panasonic researchers havecome up with a new vertically-arrangedtransistor structure that is said to reduce thedevice area to just one-eighth that of a con-ventional planar device. “This is the world’sfirst demonstration of a GaN vertical transistorapplicable to high-power switching devices,”said the Osaka-based firm.

Japanese firms target new GaN FET applications

NO

VALUX

G A N E L E C T R O N I C S

CSJulHeadlineNews5 5/7/06 9:51 am Page 5



I NDUSTRY T H E M O N T H I N R F I C S

Nitronex attracts $22 million in fundingV E N T U R E C A P I T A L

GaN-on-silicon specialist Nitronex has closeda $21.8 million venture-capital funding round.The Raleigh, NC, company, which agreed adeal to supply GaN-based power amplifierproducts for WiMAX applications to a Koreancustomer late last year, says that it will usesome of the new funds to expand its manufac-turing operations.

Several new products for WiMAX andcellular applications are said to be in the latestages of development, reliability testing andassembly qualification, while Nitronex alsoplans to accelerate recruitment in all areas ofthe business.

The latest funding round was led by AlloyVentures and included new investors ARCHVenture Partners, Diamondhead Ventures andIntersouth Partners.

Nitronex CTO Kevin Linthicum toldCompound Semiconductor that the latest fund-ing round would be the company’s last, as itnow has the money to switch to a larger, state-of-the-art manufacturing facility. Although thenew location isn’t yet confirmed, epitaxy oper-ations are set to be switched first, with the

move expected to be completed in early 2007.Staff numbers at the company are expected todouble to reach 60 people by then.

The investment in Nitronex is the latest signthat the market for GaN-based RF transistorssuch as HEMTs and HFETs is set to become asignificant one in the near future.

While its major rivals will manufacture theirproducts on SiC substrates, Nitronex takes adifferent approach, using a proprietary GaN-on-silicon solution. Linthicum added thatNitronex would be building its own 6 inchwafer-capable deposition tool so that somechip processing stages would be easier to out-source in the longer term.

According to its supporters, the use of sili-con substrates will ultimately mean that GaN-based transistors can be manufactured at acompetitive price. However, others, includingCree and RFMD, contend that the thermalmanagement benefits of SiC substrates out-weigh the advantages of silicon, especiallynow that many suppliers are offering 3 inchSiC and 4 inch is becoming more available.

Linthicum’s response is that while using a

silicon substrate does pose some thermal dif-ficulties, these can be overcome through theuse of innovative packaging solutions thateffectively dissipate excess heat.

Despite the attentions of Cree and RFMD,Linthicum says that the extra competition isgood news for the GaN RF business: “We seeit as a plus. Commercial customers will wantmore than one supplier.” The CTO added thatsuccess in the emerging market would ulti-mately be determined by having the mostmanufacturable process, and by being price-competitive with silicon LDMOS technology.Linthicum sees the Japanese company EudynaDevices as its major competitor currently inthe GaN RF space.

Korea-based consumer electronics giantSamsung is driving a move towards the con-vergence of cellular and WiMAX technolo-gies for “beyond 3G” that could create demandfor GaN-based RF components in the long run,suggested Linthicum. He believes that thedrive is towards orthogonal frequency-divi-sion multiplexing, an approach that almosteliminates interference between channels.

Compound industry heavyweights RF MicroDevices (RFMD) and Cree used the MTT-Sevent in San Francisco in June to launch theirnew ranges of GaN-based transistors designedfor applications in wireless communications.

The two North Carolina companies, whichhave never previously been in such direct com-petition with each other, are targeting both thecellular infrastructure market and the emer-ging market for amplifiers in WiMAX basestations and customer premises equipment.

RFMD already has a business unit dedicatedto cellular infrastructure applications that,although small, operates at a very healthy profitmargin. The Greensboro firm will be seekingto build on existing good relations with top-tier base-station equipment vendors, as wellas the high-volume manufacturing capacityprovided by its large fabrication facility.

Durham-based Cree has rather less experi-ence in the cellular base-station business, buthas the advantage of control over the supplyof SiC substrates that both it and RFMD use tomanufacture GaN-based transistors. However,Cree is not believed to be RFMD’s primarysupplier of substrate material. See “Carolina’s big guns board the GaNtrain” on p13 for more details.

Cree and RFMD launch a host of GaN transistorsW I D E - B A N D G A P T R A N S I S T O R S

Cree and RFMD’s entry could mark the beginning of the commercial GaN RF sector. RFMD is using part of itsexisting GaAs fab to manufacture its wide-bandgap transistors on 3 inch SiC substrates, one of which is shown here.

RFM

D

CSJulRFICsNews6-7 5/7/06 10:01 am Page 6


INDUSTRY T H E M O N T H I N R F I C S


The compound semiconductor business in theUK is set to receive a huge boost from the saleof Filtronic’s wireless infrastructure unit toPowerwave Technologies.

US-based Powerwave is set to purchase theunit in a cash-and-stock deal valued at around$345 million. Should it get the go-ahead fromFiltronic’s shareholders and regulatory author-ities, the UK firm will receive $150 million incash and 20.7 million Powerwave shares.

Filtronic has indicated that much of that cashwill be invested in its GaAs fabrication facil-ity in the north of England. The facility is cur-rently used primarily to manufacture PHEMTswitches for RF Micro Devices in the US.

“[The Filtronic board] has committed to aplan to expand the semiconductor facility togive Filtronic the potential to increase the sizeof this business by up to three times in the nexttwo years, at an estimated cost of £45 million($83 million),” said a Filtronic spokesperson.

The huge increase planned by Filtronic is aresponse to the burgeoning demand for GaAs-based PHEMT switches that are increasinglyused in cell-phone handsets to switch betweenfrequency bands. It said that the large invest-ment is needed to meet demand forecasts fromvarious cell-phone handset module suppliersand also to support its supply agreement withUK-based defense firm Selex, which makes

electronics systems for military applications.Because advanced handsets increasingly

operate in as many as four bands, and futurehandsets are expected to feature yet more aswireless local-area network and WiMAX con-nectivity become standard features, thedemand for PHEMTs is rising fast.

Filtronic added: “The [compound semi-conductor] division is seeing strong long-termmarket growth for switches based on GaAsPHEMTtechnology, with the expectation thatthey will be used in around 80% of handsetsproduced by 2008.”

While Filtronic is keeping its defense elec-tronics business unit, which also utilizes GaAstechnology, it seems clear that the firm’s focuswill be on exploiting the full potential of its6 inch wafer fab in Newton Aycliffe to becomea giant supplier of GaAs-based products.

In a trading update, Filtronic added that itscompound semiconductor business had met atarget of break-even operations in its mostrecent quarter, with revenues up 45% on theprevious year. “A key part of the strategy forthe group is to achieve further substantial valuein compound semiconductors,” said the com-pany. “We have considerable opportunity toincrease our supply of switch components formobile handsets. The aim is to enhance ourposition as the number one supplier.”

Filtronic invests $83 million in fabG A A S M A N U F A C T U R I N G

Taiwan’s Advanced Wireless SemiconductorCompany (AWSC) and UK-based IQE bothappear to be benefiting from a move by GaAschip manufacturers to outsource more ele-ments of the semiconductor fabrication pro-cess to pure-play foundry operations.

With wafer volumes now back close to thelevels seen in 1999, and manufacturingefficiencies thanks to larger wafer sizes andreduced overheads, many foundries are hope-ful of producing profits in the near future.

According to IQE, the upturn in orders andsales reflects a maturation of the industry sincethe boom of the late 1990s. In hindsight, thefragmentation of the compound semiconduc-tor business at that time resulted in very poorfuture sales visibility for foundry companies,with apparent demand for products muchhigher than the real demand.

Thanks to consolidation across the businessand a change in the approach to foundry usageby the larger device manufacturers, companies

such as IQE are no longer seen just as a way to“top up” excess wafer demand when required.

Instead, says IQE, device manufacturers arebuilding consistent levels of foundry out-sourcing into their manufacturing strategies.IQE is seeing the strongest growth in its MBEoperations in the US for wireless applications,and in particular for GaAs PHEMT switchesthat are finding increased application in cell-phones operating in various frequency bands.

Meanwhile, Taiwan-based AWSC is report-edly planning to expand its weekly capacity of4 inch GaAs wafers from 1100. The localDigiTimes news website reports that AWSCwill near-double this production to 2000wafers a month by year-end, and to 3000wafers a month by the end of 2007. The com-pany’s monthly sales have more than doubledsince 2005, according to Taiwan StockExchange figures.

Again, demand is driven by advanced cell-phone architectures, and in particular therequirements of the US company SkyworksSolutions. Skyworks has been using AWSC’scapacity extensively to ensure that it can meetrapidly rising demand without the need for amajor fab upgrade of its own or a switch tomanufacture on larger-size wafers.

Outsourcing strategiessuit AWSC and IQE

F O U N D R I E S

High Tech Campus 12b 5656 AE Eindhoven The Netherlands

Tel: +31 40 2512845 Fax: +31 84 7308840www.cedova.com [email protected]

epitaxial growthwafer processinganalyses and test

For individual processes or the complete flow from substrate to chip.

Foundry service

Are you looking for a pure play foundry?Cedova is the solution for your business.

Our team is ready to work with you in building the future of your company

CSJulRFICsNews6-7 5/7/06 10:01 am Page 7

http://www.cedova.com




I NDUSTRY T H E M O N T H I N H B - L E D S

LED manufacturer Cree has revealed LEDefficacy test results for a cool-white LED of131 lm/W, measured at 20 mA. The correlatedcolor temperature was 6027 K and the US Nat-ional Institute of Standards and Technology inGaithersburg, Maryland, confirmed the results.

Although the result is for a prototype deviceand does not reflect the performance of cur-rent production LEDs, it certainly points theway to commercial LEDs in the near futurewith efficacies exceeding 100 lm/W.

Tests were performed using prototype whiteLEDs using Cree “EZBright” LED chips thatmeasure less than 0.3 × 0.3 mm. These chipsare designed to be incorporated into whiteLEDs, which are used in backlighting appli-cations in LCD screens on mobile phones,PDAs, televisions and monitors, as well as forindoor and outdoor LED display, camera flash,

gaming and indicator applications.Cree sells EZBright chips to external cus-

tomers, but does not use them for internal whiteLED production. Its high-power XLamp LEDproducts incorporate 1 × 1 mm chips and theselarger devices are designed to operate at a cur-rent of 350 mA. As a result, they have a lowerefficacy than the smaller-chip devices.

“This is the highest level of efficacy that hasbeen publicly reported for a white LED andraises the bar for the LED industry,” explainedScott Schwab, Cree general manager, LEDchips. “This result once again demonstratesCree’s leadership in LED technology and pro-vides a glimpse into the future as to why webelieve LED-based lighting products couldnot only save energy, but also change the waypeople use light.” Tim Whitaker is editor of LEDs Magazine.

The US Department of Energy (DoE) hasannounced details of five projects that it isfunding under its solid-state lighting productdevelopment program. The projects, whichinclude a 30% cost share, total $10 million invalue and feature three efforts that are focusedon III-V LED technology and two based onorganic LED materials.

Awarded via the DoE’s National EnergyTechnology Laboratory, the projects areexpected to contribute to achieving the DoE’starget of developing solid-state lighting witha product system efficiency of 50% by 2025.

Color Kinetics, General Electric GlobalResearch and Osram Sylvania have beenawarded the three projects that incorporate III-V LED technology.

Color Kinetics plans to develop replace-ment lamps for existing 60 W incandescentbulbs with an efficacy of 80 lm/W. It will develop what it describes as a novel hybrid-LED source that combines direct emissionwith phosphor down-conversion.

General Electric’s research wing will focuson using novel nanophosphors to convertviolet LED emission into white light. Also planning to replace incandescent sources, it isaiming for an efficacy of 96 lm/W.

Osram Sylvania’s target is to increase exter-nal quantum efficiency using so-called remotephosphors and a multilayer thin-film coatingtechnique to extract more light from LEDs.

The product development required underthe project guidelines includes production offield-ready prototypes focused on a targetedmarket application with fully defined price,efficacy and other performance features.

All of the projects are covered under theDoE’s “exceptional circumstances determi-nation”, which was issued in June 2005 andcovers ownership of intellectual property.

Cool-white chip delivershigher levels of efficacy

US Department of Energy showers$10 million on solid-state lighting

F U N D I N G

Color Kinetics, one of the recipients of funding underthe DoE’s latest solid-state lighting initiative, hasinstalled a number of LED-based systems in residentialsettings, including this meditation room in the US.

COLOR KINETICS

D E V E L O P M E N T

CSJulHBLEDNew8-9 5/7/06 10:12 am Page 8



INDUSTRY T H E M O N T H I N H B - L E D S

Compound Semiconductor July 2006 compoundsemiconductor.net

Avago Technologies, the chipmaker formerlyknown as Agilent’s semiconductor productsgroup, has entered the high-power LED mar-ket with the first in a series of 1 W emitters.

Designed for use in applications such as out-door lighting and display backlights, the In-GaN-based LEDs are available in blue, greenand, when coupled with a color-convertingphosphor, white. Previously, Avago concen-trated on lower-power LED products and didnot compete in such a direct manner with com-panies such as Lumileds, Nichia and Cree in

the high-power arena.The ASMT-MX00 products are the first in

a series of devices operating at 1 Wpower dis-sipation and 350 mAforward current.

The line-up features green emitters that pro-duce a luminous flux of 40 lm, 10 lm in blueand 35 lm in white. White color temperaturesranging from 4500K to 10000K are available.

Avago was sold by Agilent in 2005 to twoinvestment banks, Kohlberg Kravis Robertsand Silver Lake Partners, while its 50% stakein Lumileds was sold to Philips.

From our Web pages...visit compoundsemiconductor.net for daily news updates

...Veeco hooks up with KOPTIVeeco Instruments is to deliver one of its latestGaNzilla MOCVD machines to the KoreaPhotonics Technology Institute (KOPTI) under anew collaborative relationship to help advancesolid-state-lighting technology. Both KOPTI andother LED manufacturers in the country,including Samsung and LG Innotek, will be ableto use the machine, get process support andliaise with key experts at the site.

...Nakamura wins gongGaN pioneer Shuji Nakamura has won theMillennium Prize Foundation’s second“millennium prize”, following in the footsteps ofInternet trail-blazer Tim Berners-Lee. As well asthe gong, which will be awarded at a ceremonyin Helsinki on 8 September, Nakamura receives71 million ($1.26 million). The biennial prize isawarded for technological innovations thatimprove the quality of human life.

...Strategies Unlimited confirms slowdownThe latest report on the growth of the HB-LEDmarket from Strategies Unlimited has confirmedits earlier assertation that the sector hasentered a period of slower growth than it hasexperienced before. Forecaster Bob Steelerevised his estimate of the value of the marketsector in 2005 down slightly to $3.9 billion.That represents annual growth of 6.2%, aneffect caused by the market for cell-phonehandset applications reaching saturation.

...Citizen plans ultra-bright chip rampJapanese LED manufacturer Citizen Electronicsis set to begin the mass production of 3.5 Wwhite LEDs that produce 245 lm, claims a reportin the Nihon Keizai Shimbun newspaper.Volume manufacturing will begin in Septemberat a unit situated next door to the company’sheadquarters in Fujiyoshida, in the Yamanashiprefecture of Japan.

9

Avago aims for high-power sector

Highlink hikes up wafer capacityM A N U F A C T U R I N G

Highlink Technology Corporation has orderedtwo huge MOCVD reactors in a bid to signifi-cantly step up its wafer fabrication capacity.The Taiwanese LED manufacturer will takedelivery of the two Aixtron machines, whichwill be used to make high-brightness GaNemitters, over the next few months.

The AIX 2800G4HT tools are the largest-capacity offerings for GaN growth from theGerman equipment vendors and support42 × 2 inch wafer manufacturing. As well asproviding a large ramp-up in capacity, the toolsfeature an improved gas inlet that will improveboth wafer yields and device uniformity.

Gary Yu, president of Highlink, commented:

“The build-up in customer demand means that we now require a major increase in pro-duction capacity.”

Sales of MOCVD equipment to Taiwan-based LED manufacturers have picked upstrongly in the first half of 2006, with Aixtronand its US rival Veeco Instruments reportingimproved order books. The upturn follows aperiod of retrenchment and consolidation inthe Taiwanese LED sector, caused by an over-crowded market and overcapacity of supply.

Mergers and acquisitions between some ofthe region’s key companies, such as UnitedEpitaxy Company and Epistar, have sincestabilized the industry.

C H I P S

CSJulHBLEDNew8-9 5/7/06 10:13 am Page 9




http://www.raboutet.fr


I NDUSTRY T H E M O N T H I N O P T O E L E C T R O N I C S

Market analysts at iSuppli say that the next-generation DVD sector will drive demand forGaN-based blue lasers to at least 65 milliondevices in 2010.

This year, they expect the debut modelsfrom companies such as Sony, Toshiba andSamsung to result in the shipment of 1.6 mil-lion players and recorders, each one of whichwill require a blue laser diode.

However, unlike some commentators,iSuppli senior analyst Chris Crotty says thatthere will not be a clear winner in the battlebetween the Blu-ray and high-definition (HD)DVD formats. “This is not a repeat of VHS ver-sus Beta,” said Crotty, referring to the previ-ous battle between consumer electronicsgiants. “The most likely outcome is stalemate,with the savvy manufacturers introducingdual-format players as early as the 2006holiday season.”

In a tear-down analysis of Toshiba’s newHD-DVD player, iSuppli estimated the bill ofmaterials at $674, with the laser-containingoptical drive unit accounting for $200 of thattotal. Toshiba is clearly subsidizing its ownplayer in a bid to steal a march on its rivals,

since it retails for just $499 in the US.Samsung and Philips have both launched

Blu-ray players in the past month, whileaccording to reports in the Japanese press,Toshiba has ditched any remaining efforts tounify the Blu-ray and HD-DVD standards.

Californian optical component makerdoubles wafer fab capacity in Texas

M A N U F A C T U R I N G

Finisar, the optical component maker based inSunnyvale, CA, has posted another profitablequarter, with total sales for fiscal 2006 reach-ing $364.3 million.

In the latest three-month period, ending on30 April, revenue from sales of components andsubsystems rose by 9.2% sequentially to$91.9 million. Finisar reported a net profit of$1.7million for the quarter on total sales (inclu-ding network testing gear) of $102.4 million.

Those sales of optical components and sub-systems represent a 42.5% improvement com-pared with one year ago. “Fiscal 2006 markedan important turning point for our company,”said CEO Jerry Rawls. “We worked our way

back from one of the most difficult marketdownturns in history and succeeded in achiev-ing our plan for returning to profitability.”

To meet what it expects to be a continuedrapid increase in demand for optical compo-nents, Rawls says that the company is in theprocess of doubling its wafer fab capacity inTexas. Currently based in Richardson, Finisaris switching production to a brand new clean-room facility in Allen.

“The first reactor is plumbed and wired anddoing the first calibration runs,” explainedRawls. The existing Richardson fab will firstbe replicated with new equipment in Allen,before the company switches the Richardson

set-up to the same site.The biggest growth for Finisar has been seen

in sales to the local and storage-area networksectors, which have grown by 41%. Sales tometro access applications also rose by 27%.

For the full fiscal 2006 period, Finisar hada deficit of $24.9million, compared with a lossof $88.6 million in the previous year.

Finisar, which has acquired Honeywell’sadvanced optical components division andmuch of the fiber-optic business previouslyowned by Infineon Technologies over the pastfew years, also operates a chip fab in California. A jury in Texas awarded Finisar $79 m indamages from DIRECTV on June 26.

Sharp, the world’s biggest manufacturer ofsolar cells, has said that concentrator systemsbased on III-Vmultijunction cells will becomean essential part of the global effort to combatgreenhouse gas emissions.

Takashi Tomita, corporate executive direc-tor at the Japanese electronics giant, is also urg-

ing companies in the photovoltaic industry toco-operate on the development and promotionof the technology to ensure that its advantagescan be properly exploited.

Although Sharp focuses on silicon-basedsolar energy generation, the company pro-duced multijunction cells with a conversionefficiency of 37% in its laboratories last year,a figure that is comparable with the best effortsof III-V specialists Emcore and Spectrolab.

Tomita says that although 85% of solar cellsmanufactured in 2005 were made from crys-talline silicon, the higher conversion efficiency

and scalability of III-V cells means that theportion of the system cost attributable to thecell itself is relatively small. This means thatoverall system cost could be dramaticallyreduced through volume manufacturing in thenear future, Tomita added.

Sharp has also identified the areas of theglobe that would be best suited to concentra-tor systems based on III-V cells. These areregions where there is abundant solar irradia-tion through clear skies, and include southernAustralia, southern Europe and northernAfrica, Mexico and the Middle East.

iSuppli analysis predicts Blu-ray player deadlockG A N L A S E R S

Sharp urges III-V solarfirms to co-operate

S O L A R C E L L S

Samsung began shipping the industry’s first Blu-ray disc player to hit the US market in late June. iSuppli expectsthat 1.6 million next-generation players featuring a blue diode laser will ship this year, rising to 65 million in 2010.

SAM

SU

NG

CSJulOptonews10-11 5/7/06 10:17 am Page 10


INDUSTRY T H E M O N T H I N O P T O E L E C T R O N I C S


Quantum dot (QD) laser specialist NLNanosemiconductor has come up with a novelstructure that delivers singlemode operationover a 25 nm-wide range of wavelengths.

The broadband laser chip will enable cost-effective integration of III-Vcomponents withsilicon photonics such as arrayed waveguidesand light modulators, says the company fromDortmund, Germany, which has just completedSeries B funding to raise 75 million ($6.3 mil-lion), and opened an office in San Jose.

As well as having potential applications inoptical communications, the broadband sourcecould be used in place of superluminescentLEDs for the medical imaging technique thatis known as optical coherence tomography. Inthis application, the higher power of the lasershould lead to better-quality images that couldbe used for the much earlier diagnosis of malig-nant tumors, for example.

“We managed to develop a way to get a laser

to lase with a very wide emission spectrum,which is of course the opposite of the funda-mental nature of lasing – a very narrow spec-trum,” said company COO Alexey Kovsh.

To create the broadband emission, Kovshand colleagues designed a special epitaxialstructure featuring ten layers of quantum dotson a 3 inch GaAs substrate. “Each layer wasgrown under a different regime to make thegain spectrum broader,” he explained.

Although this sounds like a very complexprocess, Kovsh says that the laser is compati-ble with a manufacturing environment. Themaximum output power achieved in continu-ous-wave mode so far is 400 mW, while anemission of 150 mW over a wavelength rangeof 20 nm has been shown to be reproducible,Kovsh told Compound Semiconductor.

In the longer term Kovsh believes that theQD lasers under development could find wide-spread use in both optical interconnects and

optical clock systems for high-performancecomputing applications.

Acknowledging that such applications areunlikely to take off before 2010, Kovsh admitsthat the company is now searching for shorter-term market opportunities.

Although large volumes are not yetrequired, he says that the quantum-dot processis at a level of maturity where the laser couldbe switched into mass production if necessary.

Some improvements are still needed, how-ever. For instance, the laser currently operatesat up to 80 ºC in a stable fashion. Operation at100 ºC will be required for final implementa-tion in an uncooled system.

Kovsh says that the NLNanosemiconductorteam is working to optimize chip dimensionsand the number of QD layers, and is carefullyadjusting each of those layers to improve theoverall device performance and extend thelaser’s emission band further still.

Ten-layer laser has broadband promise

Datacoms push upturn in active components marketM A R K E T S

The rapid build out of fiber-based enterpriseand access networks is expected to drive asharp upturn in the market for active opticaltelecom components such as lasers, modula-tors and detectors.

According to analysts at CommunicationsIndustry Researchers (CIR), the value of theactive components sector will near-quintuplefrom just under $1 billion today to $4.8 billionin 2011. “The big surge in growth is due to therapid penetration of fiber into enterprise andaccess networks, and the larger bandwidthapplications riding over them,” said CIR.

Tunable lasers are now one of the key

growth engines, with tens of thousands of suchdevices being shipped annually. By 2011, thisniche alone is expected to become worth morethan $460 million.

Meanwhile, the market-research companyHeavy Reading is predicting a huge upturn inthe deployment of fiber-to-the-home networks.“Over the next 15 to 20 years, copper accessnetworks will largely be replaced by a fiberaccess network, creating massive opportuni-ties,” it claimed in a new report.

The emergence of complex, data-hungryservices such as video delivery over IP net-works has sparked the beginning of a mass

migration to fiber access in several countries,notably Japan, Sweden and the US already,with China, France, Korea and the Netherlandsset to follow in the next couple of years.

That technology shift will boost demand foractive optical devices at the access level of thenetwork, as well as the need for more band-width in the optical backbone. The HeavyReading report predicts that by 2011 more than62% of Japanese residences will enjoy high-speed fiber access to services, equivalent to31 million connected households. In the US itexpects 10.3 million households to be fiber-connected, or 9% of available connections.

IPG steps up capacity with Veeco MBE reactor kitH I G H - P O W E R L A S E R S

Fiber laser system integrator and laser chipmanufacturer IPG Photonics is expanding itsmanufacturing base with the purchase of anMBE reactor from Veeco.

IPG, which owns a semiconductor fab inOxford, MA, has become one of the world’sleading manufacturers of laser diodes in thepast couple of years thanks to booming salesof high-power fiber laser systems for applica-tions in industrial machining.

The Gen200 MBE tool that it has orderedfrom Veeco enables 4 × 4 inch epiwafer pro-duction and will help IPG in its bid to increase

its diode manufacturing volumes in the future.IPG is in the process of constructing extra

fab cleanroom capacity at its headquarters andindicated to Compound Semiconductor that itplans to significantly increase its productionof GaAs-based lasers, which emit at 980 nm.

Having manufactured an estimated aggre-gate of 1 MW in total laser chip output powerin 2005, the company is hoping to double thatfigure this year and then to reach 6 MW in2008. Each IPG chip typically emits 8 W,although the firm is also working on increas-ing that figure.

D E V I C E D E V E L O P M E N T

IPG makes 8W laser diodes at its fab in Oxford, MA.

IPGPH

OTO

NIC

S

CSJulOptonews10-11 5/7/06 10:17 am Page 11


Think that epi-ready sapphire waferis defect free? Think again.

Detect and classify those critical transparent wafer defects with Candela.™ Our Candela Optical Surface

Analyzers are unique surface inspection systems that automatically detect and classify defects on transparent

materials like sapphire, SiC, LiNbO3, and glass. Faster than operator-dependent microscope inspection,

these systems can detect surface contaminants and defects on wafer substrates that hurt device performance.

Measures scattered light, reflected light, phase shift, and surface topography

Detects and classifies epi and coating defects

Accommodates 50 to 300 mm wafers

Automated cassette-to-cassette operation

For more product information, go to:

©20

05 K

LA-T

enco

r Cor

pora

tion.

www.kla-tencor.com/candela

Accelerating Yield®

http://www.kla-tencor.com/candela


I NDUSTRY B E H I N D T H E H E A D L I N E S

Much hope, and indeed hype, has long accompanied theprospects for wide-bandgap RF technology based onGaN and SiC transistors. This now seems set to becomea commercial reality. Not only have industry big gunsRF Micro Devices (RFMD) and Cree both launched awide range of high-power GaN-based transistors, buttheir North Carolina neighbor Nitronex has corneredanother $22 million in equity finance.

With all three now set to plow greater effort and cashinto pushing the technology at customers in the broad-band wireless access and infrastructure sector, 2006 lookslikely to go down in history as the turning point for GaNmicroelectronics.

That’s largely because Cree and RFMD are such keyplayers in the compounds business, and have been centralto the success of GaAs and SiC-based electronics. Thesuccess or otherwise of their GaN offerings may welldetermine the fate of the overall sector.

RFMD’s involvement is arguably the most significantdevelopment. The Greensboro-based device maker is theworld’s major GaAs wafer producer and not a companythat launches a range of new products based on a newmaterial platform without a great deal of forethought.“RFMD is fairly conservative,” admits Jeff Shealy, headof RFMD’s wireless infrastructure unit, lending weightto the theory that GaN’s time has now come. He says thatwhile the market opportunity for GaN has been there allalong, only now has the reliability of products reached apoint where customers are serious about deploying it.Coupled with this is the widespread availability of 3 inchSiC substrates on which to manufacture GaN devices.

Proving reliability is perhaps the most critical prob-lem that GaN has faced. Armed with a substantial dataset from a long run of GaN-on-SiC wafers, Shealy is inthe process of educating potential customers in the base-station business who want, unsurprisingly, to see water-tight stats before they commit to a new line of products.Shealy believes that they are now finally waking up tothe advantages that the wide-bandgap technology canbring, and that the cellular industry in particular is com-mitted to backing GaN technology for applications inwideband-CDMAbase stations.

Shealy says that RFMD is looking to compete in thehigh-end cellular sector for final-stage, linearized ampli-fiers. Already sampling to the top-tier vendors of cellu-lar base-station equipment, its new family of GaN HEMTtransistors features nine different products. Transistorsfor cellular applications include four HEMTs, the mostpowerful being a 120 W device. On the WiMAX front,RFMD launched new 50 Wproducts for both the 2.5 and3.5 GHz frequency bands.

Of the three North Carolina companies, RFMD looksto have the best chance of success within the cellular busi-ness. Crucially, it is in a position to exploit its experiencein high-volume GaAs manufacturing to give it the edgeover a relative upstart such as Nitronex. And althoughCree also has plenty of volume-manufacturing expertise

thanks to its great success with GaN-based LEDs, it isnot nearly so well established in cellular infrastructure.“RFMD has great customer relationships and a qualifiedhigh-volume facility,” said Shealy. Although its GaNprocess will need individual qualification, which is setto begin in September, cellular customers are alreadyconvinced that RFMD can deliver as a company.

Having now converted much of its 4 inch GaAs capa-city to 6 inch, RFMD has been able to apply its smallermanufacturing line for back-end processes such asmetalization and lithography to 3 inch GaN-on-SiCwithout the need for costly redevelopment. “We plan oncompeting [for this market] in the context of our GaAsfab,” said Shealy. “It means that GaN looks no differentto any other wafer.”

Akey difference is at the epitaxy stage, where RFMDuses MOCVD for GaN deposition rather than MBE. Thelatter is its standard method for GaAs wafer epitaxy. Thecompany uses both internal and external sources of GaNepitaxial material and has two MOCVD machines of itsown. Durham-based Cree is also targeting WiMAX andcellular applications with its own family of 2.4–3.9 GHzHEMTs. At the MTT-S exhibition in mid-June, Creereleased higher-power versions, including a 120Wdevice.

Head-to-headMeeting as competitors for the first time, one side issuethat Cree and RFMD might need to resolve is the supplyof SiC substrates. Despite the locality of the two com-panies and its dominance of the SiC market, Cree is notcurrently RFMD’s primary supplier of wide-bandgapmaterial. Even though the two will compete directly forthe first time with their GaN products, the market forthese new devices is at such an early stage that for Creeto supply RFMD with material should be viewed as aminor conflict of interest.

Nitronex, based just down the road from Cree inRaleigh, is the young pretender that stands against thetwo established chip makers. The key difference is itstechnology, GaN-on-silicon. The thinking here is thatsilicon wafers can already be scaled much larger than SiC,and ultimately provide a more economic materials solu-tion with tougher thermal problems dealt with throughinnovative packaging solutions. Having recently sealedan additional $22 million in venture capital, Nitronex ispoised to move into a new state-of-the-art facility in early2007, in preparation for device production.

While acknowledging the potential threat posed by itsnear neighbors, Nitronex CTO Kevin Linthicum is happyto see their involvement: “Commercial customers likeNokia and Siemens will want more than one supplier,”he said. “The GaN train is leaving the station and Creeand RFMD don’t want to be left behind.”

Of course, the relative merits of the North Carolina triowill be rendered irrelevant if silicon LDMOS continuesto dominate the infrastructure scene. As always for III-Vdevelopers, silicon will pose the biggest threat of all.

Charlotte

Greensboro

Raleigh

Durham

North Carolina

South Carolina

Virginia

Carolina’s big guns board the GaN trainWith some majordevelopments at Cree,RF Micro Devicesand Nitronex, there’sno doubting where thepush to commercializeGaN-based RF devicesis coming from.

R F I C S

Cree, RFMD and Nitronex areall located in North Carolina.While Cree and Nitronex wereboth spawned from the researchgroup headed by Bob Davis atNorth Carolina State University,RFMD gained its wide-bandgapexpertise through theacquisition of Cornell Universityspin-out RF NitroCommunications in 2001.

CSJulBehindHeadlines13 5/7/06 10:37 am Page 13



T ECHNOLOGY I N P O P T O E L E C T R O N I C S

Optical integration movePhotonic integration was one of the hot new ideas that emergedduring the telecoms boom, but it was subsequently shelved bycash-strapped component vendors. Now, as Roy Rubensteinreports, the technology could be set to make a big revival.

Today, the development of monolithically integratedInPcircuits featuring several different optical compo-nents is being driven by high-speed transmission, inparticular the move towards 40 Gbit/s and the new100Gbit/s standards. Indeed, the most widely deployedInP photonic integrated circuit (PIC) to date is anelectro-absorption modulated laser (EML) that com-bines a laser and modulator, while there is also a smallermarket for tunable lasers that integrate several func-tions in a single InP chip. “Tunable lasers are sellingin reasonable volumes – several tens of thousands ayear – while EMLs are in the hundreds of thousands,”says Vladimir Kozlov, founder of transceiver marketanalyst company Lightcounting.

US firm Apogee Photonics is targeting this marketwith its 10 Gbit/s lasers. The company has its origins intwo InP start-ups: ASIP and ThreeFive Photonics, thelatter having brought to market such monolithic devicesas a multi-wavelength receiver and an optical perfor-mance monitor. Now, Apogee’s focus is to exploit itsselective-area growth and asymmetric twin-waveguidetechnologies to optimize the optical functions in itsproducts. “We connect functions using tapers, and withprecise control of the composition and structure we candetermine the performance of the laser and modulator,”says Milind Gokhale, Apogee’s CTO.

One example is Apogee’s uncooled EML, whichoperates over a wide temperature range. “It is fair tosay that we are using one of the best integration plat-forms to make more down-to-earth products that themarket needs in large quantities,” says Erik Pennings,Apogee’s director for product marketing.

Pennsylvania-headquartered InP-chip specialistCyoptics is also making EMLs, as well as tunable lasersunder contract for other firms. Like Apogee, its staffhave made some exotic PICs in the past. “We made adistributed Bragg grating with phase control, a powerdetector, and a modulator for a tunable laser over the C-band [1530–1565 nm],” says Robert Hartman,Cyoptics’vice-president for device design and devel-opment. “We even had a version with a semiconductoroptical amplifier that was finished in development.”But the market wasn’t ready for such devices, he says,and didn’t want to pay any more than a 10% premiumfor single-wavelength (untuned) devices.

Now, however, Cyoptics is seeing renewed interest

in PICs. “Regarding photonic integration, some of themeetings I had [at OFC/NFOEC 2006] were the best infive years,” says Stefan Rochus, director for marketingand business development, who added that particularinterest is being shown in the emerging 100 Gbit/sEthernet standard.

In fact, optical chips operating at these transmissionrates are already being built and deployed by Infinera,a US-based vertically-integrated company that man-ufactures InPchips at its Californian fab and uses themto build systems. The company has even detailed a labdemonstration of a 1.6 Tbit/s optical chip comprising40 channels operating at 40 Gbit/s. “The reasoning [forsuch a device] is purely economic,” explains JagdeepSingh, CEO of the US systems vendor. “It is the lowestcost way to deliver bandwidth.”

Even though this 1.6 Tbit/s device will not bedeployed for several years, it highlights the perfor-mance hike that can be achieved. More significantly,Infinera’s PIC strategy could signal a forked path inthe optical networking industry. Indeed, Infinera’s PICsare the optical industry’s sole example to date ofMoore’s Law-style progress.

For now, though, carriers – and by implication sys-tem vendors – want to maintain flexibility by usingonly a fraction of a system’s capacity. More channelsare lit, or a channel is upgraded from 2.5 to 10 Gbit/s,

Do it all: InP is already the material of choice for active optical devices, and its ability to s

“A highlyintegrateddevice is notnecessarily the bestsolution.” Emmanuel DesurvireAlcatel

CSJulPHOTONIC14-16 5/7/06 10:42 am Page 14



TECHNOLOGY I N P O P T O E L E C T R O N I C S

only when the need arises. “For any WDM system, ourphilosophy is pay-as-you-grow,” says EmmanuelDesurvire, senior director for photonic technologiesin Alcatel’s photonic-network product group.

When designing a new system, vendors such asNortel Networks look at the platform specification interms of the interfaces – and the density of interfaceswanted by the carrier – and the reach. “We look at whatnew technologies can be brought to bear,” says MauriceO’Sullivan, manager of physical layer and opticaldesign at Nortel. Depending on the maturity of theintegration process, Nortel will consider the technol-ogy if it delivers power, size and, in some cases,improvements in system performance.

The cost of integrationAlcatel stresses that adequate trade-offs between differ-ent performance criteria must be made when consider-ing integrated products. Adopting an integrated devicewill probably require the line card to be redesigned, andthat adds cost. “The technology also has to be matureand proven – we don’t take chances in the field,” saysDesurvire. “A highly integrated device is not neces-sarily the best solution. Throwing away what you havecan have an impact across the [network] architecture.”

With Nortel and Alcatel selling their optical-component divisions, it has also become more com-

plicated to explore how integration can benefit systemdesign. “It is difficult to predict what optical compo-nent players are doing and for them to second-guesssystems performance,” says Michel Belanger, seniortechnical advisor in Nortel’s next-generation opticalnetwork group.

Indeed, Infinera’s decision to make systems allowsthe company to put its PIC technology at the heart ofits design. “If you don’t have control of the optical tech-nology, you can’t have a differentiated product,” saysDave Welch, Infinera’s chief strategy officer.

Infinera’s digital transport network (DTN) platformaddresses the cost issue of optical–electrical–opticalconversions by integrating discrete transponder func-tionality into its transmit and receive PICs. Infinera willnot detail the resulting cost savings, but one carriersuggests that its DTN platform is 30% cheaper thanequivalent dense wavelength-division multiplexing(DWDM) systems. Since every wavelength is con-verted into the electrical domain, whether it is droppedor passed, several other key benefits result. The signal’scondition can be monitored, switching and groomingcan be performed, and the signal can be properly regen-erated – unlike in the optical domain where both signaland noise are only amplified.

Welch claims that Infinera’s InP manufacturingprocess is robust enough to achieve high yield. “We

es back onto the agenda

ility to support passive waveguides makes it the ideal platform for monolithic integration.

Look inside: Infinera’s DTN (digital transport network) is the first transport platform to be based on large-scale InP photonic integrated circuits. Its features include protection switching, grooming,signal regeneration, bandwidth management and digital performance monitoring.

INFIN

ERA

INFIN

ERA




TECHNOLOGY I N P O P T O E L E C T R O N I C S

make distributed-feedback lasers in a similar fashionto everyone else, as we do our modulators,” he says.What is different is that Infinera gets its engineers todesign around the process. “We design what the processcan manufacture.”

Infinera’s design also trades off optics and electron-ics. Forward-error correction and electronic dispersion-compensation techniques are used to relax the opticalspecifications, placing more of the link-budget burdenon the electronics. Nortel makes a similar point aboutthe importance of integrated components – not justoptical, but also analog and digital silicon chips.

“Infinera can basically do what others thoughtimpossible because they expanded the domain overwhich they did their design trade-offs to span systemsto processing,” says Karen Liu, research director forcomponents at Ovum-RHK.

And while the industry is focused on a pay-as-you-grow approach, Infinera’s strategy is to deliver wave-length blocks – in chunks of 10 × 10 Gbit/s – cheaplyenough, whether or not they are all needed. Welchargues that a line card costs roughly the same, regard-less of what is on it, and the company’s PIC shifts theindustry from 10 to 100 Gbit/s on a line card. Infinera’sprototype 40 × 40 Gbit/s PIC will also deliver a further10-fold hike in three to four years’ time.

But Ovum-RHK’s vice-president for communica-tions components Daryl Inniss is less optimistic. “A

tier-one carrier will not wholly replace its network withInfinera equipment in the next 5 or 10 years,” he says.“It’s a good idea, a radical idea, but the path they offeris way ahead of incumbent systems.”

Baby stepsAs to whether the pace of monolithic integration willhasten in the next five years, most believe not. But thatdoesn’t mean there won’t be exciting developments.“Infinera’s 10×10Gbit/s is all-in-one-go, but there willbe some baby steps by the industry towards its solu-tion,” says Inniss.

Perhaps the most exciting is the placing of a tunablelaser within an XFP package that supports line-sidetransmission distances of 80 km and greater. Agility(recently bought by JDSU) has an InP monolithicallyintegrated tunable laser that is sufficiently small to fitwithin an XFP package.

This development indicates that a variety of trans-ceiver types will converge to one form factor, with onelaser and one receiver, says Inniss. Unit volumes willgo up, while the price of such a tunable, pluggableDWDM interface will dip below $1000.

Meanwhile, Infinera believes it is only a matter oftime before someone breaks away from the pack to makea PIC triplexer. At the same time, advances in hybridtechnology continue to reduce packaging costs, and onlytime will tell which integration method will win.

About the authorRoy Rubenstein is a freelancetechnology journalist who isbased in Omer, Israel. Contact: www.iviht.com.


http://www.iviht.com


http://www.riber.com


T ECHNOLOGY T R A N S I S T O R M O D E L I N G

Simulations provide additionalinsights into GaN HFET reliabilityGaN HFET reliability can be improved by adding field plates to the structures. This is believed to reduce carriertrapping at the surface, but simulations by Ric Borges, Nelson Braga, Bo Wu and Vidas Mickevicius ofSynopsys show that it also decreases the electron temperature and bulk trapping throughout the device.

GaN HFETs are attracting considerable attention ashigh-power and high-frequency devices for radar,avionics and wireless base-station transmitters, thanksto the unique material properties of III-N material.However, the commercialization of these devices hasbeen hampered by reliability issues that have beenattributed to carrier trapping in either the bulk or sur-face of the device. These problems are being addressedby experimental studies that can produce epitaxy- andprocessing-related improvements, and simulations thatcan optimize device design.

Simulations provide key insights into deviceoperation and the degradation mechanisms that affectreliability. They can contain structural details such aslayer thicknesses, doping profiles and trap concen-trations. The software-based approach can also aid thetailoring of a structure to specific market applicationsrequiring particular performance characteristics, andcan be an effective tool for selecting and optimizingthe design.

In the silicon industry, process and device simulation– referred to as technology computer-aided design – iswidely used to develop and optimize various tech-nologies ranging from highly scaled CMOS to powermanagement, non-volatile memory and image sensors.Over the past decade this approach has also beenincreasingly used for III-V technologies, and recentadvances to III-N devices has promoted the use ofsimulation for improving GaN HFETs.

Investigating current collapseAt Synopsys we have built the simulator softwareSentaurus Device (see “Simulation software” box,opposite, for details) to explore a physical mechanismthat is referred to as either current collapse or disper-sion, and is defined as the drain current (ID) degrada-tion under operational or stress conditions. Thismechanism limits GaN HFET reliability and is conse-quently inhibiting device commercialization. However,its origin is still under debate despite the various mod-els that have been proposed, such as electron trappingin the GaN buffer and AlGaN barrier layers, and at thesurface (Binari et al. 2002, Vetury et al. 2001).

Our simulations can help to understand the natureof current collapse, provide a detailed insight into

device operation, and point towards mitigation strat-egies to avoid current collapse. However, they cannotpredict the formation of non-ideal device propertiessuch as traps.

The software has simulated a GaN HFET(see figure1, p18) with a room-temperature Hall mobility of

The simulation of GaN HFETs presents several challenges and needs to cater for thepolarized wurtzite crystal structures of AlGaN, InGaN and GaN, which have dipoles acrossthe crystal in the [0001] direction. These dipoles can cause spontaneous (pyroelectric)polarization, while pseudomorphic heterostructures fabricated from these III-Ns alsohave strain-induced (piezoelectric) polarization.

At Synopsys, our software, which can be run on either PCs or workstations, provides arigorous calculation of the polarization fields from the stress in the structure – includinglocal stress sources such as the gate metal and dielectric passivation – and leads to in-plane polarization vector components. The resulting polarization-induced charges areincluded in our calculation.

Quantization effects at the AlGaN/(In)GaN heterointerface create a two-dimensionalelectron gas channel. A quantum-mechanical solution is needed to properly account forthe electron spatial distribution in this channel. Our model uses a quantum potentialcorrection to the continuity equation, known as density gradient. This closely matches thesolution to the full Schrödinger equation (see figure, above right), while improving thesimulation’s robustness and providing a five-to-ten fold cut in computational time. Oursimulation also accommodates hot-carrier effects and dynamic trapping/detrapping,which is needed to simulate the physical mechanisms behind device degradation.

Simulation software

distance (nm)

ener

gy (e

V)

–0.4–5 0 5 10

–0.2

0.0

0.2

0.4

AlGa

NGa

N

depth (nm)

elec

tron

dens

ity (c

m–3

)

–2 0 4 6

4×1019 AlGa

N

GaN

2

VG= –5 V

VG= 0 V2×1019

0

The interface between the GaN and AlGaN layers can confine electrons in the conduction band infour different energy states (see left figure). Although calculations using Schrödinger’s equationcan be used to simulate the electron density in the channel at VG = 0 V and VG = –5 V (see rightfigure, solid line), time savings can be made using the density-gradient approach, which involves aquantum-mechanical correction to the continuity equation (see right figure, square markers).

CSJulMODELING17-19 5/7/06 11:03 am Page 17



TECHNOLOGY T R A N S I S T O R M O D E L I N G

1100–1200 cm2/V-s and sheet carrier concentration of8–9 × 1012 cm–2 (Braga et al. 2004). The calculationsused values for the band structure, carrier mobility andsaturation velocity of GaN, InN and AlN that are shownin table 1. Interpolation of the corresponding binaryparameters provided values for AlGaN and InGaN.

Trapping behaviorTo investigate trapping behavior and gain insight intothe current-collapse mechanisms, we inserted a back-ground of acceptor/electron bulk traps with a densityof NT = 5 × 1017 cm–3 into our model of this particulardevice. Since the motivation behind this study is toobtain a qualitative insight, the traps were defined witha single energy level 1 eV above mid-band. However,our simulation software can accommodate multipletrap levels that are commonly ascribed to GaN epitax-ial material (see Binari et al. 2002) in either the bulk orat the surface.

Using these conditions we could match our simu-lations of direct-current (DC) I-V characteristics withmeasured data if we used an interface charge of1.15 × 1013 cm–2 and an electron saturation velocitythat is 40% of the theoretical bulk value. The valueused for the interface charge, which equates to a par-tially relaxed AlGaN layer, is lower than that associ-ated with theoretically-ideal strain conditions, whichwould have a value of 1.5–1.7 × 1013 cm–2. However,

this value is consistent with observations from severalexperimental groups who have seen a reduction of thepiezoelectric component. A reduction in saturationvelocity has already been proposed, due to modifiedscattering of the two-dimensional electron gas (2DEG)at the heterointerface.

We also compared the experimentally-measured DCID–VD output characteristics with simulations that bothignore and include hot-electron effects (see figure 2).Our drift-diffusion (DD) model assumes that the car-riers are in thermal equilibrium with the lattice, whilethe hydrodynamic hot-electron compatible (HD) modelaccounts for carrier heating and non-local electric fieldeffects. The measured and simulated curves createdwith the hydrodynamic model have a decline in theirID value after the peak – known as a negative differen-tial conductance – that is absent in the curve producedwith the DD model. We believe that this feature is notattributable to self-heating (Deng et al. 1999), but is infact evidence for hot-electron capture at bulk trapsunder sufficiently high drain bias.

The HD model also shows that at large drain biaseselectrons in the channel are significantly heated andcan exit the AlGaN barrier and then spread toward theGaN bulk – an effect that is not predicted with theDD model (see figure 3). At these high drain biaseselectrons occupy more traps. This increases the energyof the conduction band under the gate edge located

Dielectric constant 9.5 15.3 8.5Energy gap (eV) 3.47 0.8 6.2Electron affinity (eV) 3.4 5.8 1.9Electron mobility (cm2/V-s) 1100 2400 300Electron saturation velocity (cm/s) 1.2 × 107 2.6 × 107 1.5 × 107

Effective conduction band density of states (cm–3) 2.65 × 1018 1.3 × 1018 4.1 × 1018

Energy relaxation time (ps) 0.1 0.1 0.1

GaN InN AlN

The properties of III-Ns

Table 1. The Synopsys simulations use the room-temperature values for III-Ns that are shown in the table. The values are taken from the work by M E Levinshtein and colleagues (M E Levinshtein et al. 2001 Properties of Advanced Semiconductor Materials: GaN, AlN, InN, BN and SiGe, Wiley, New York).

distance (µm)

dist

ance

(µm

)

–0.1

–0.04

–0.02

0.00

0.02

0.04

0.06

0.08

0.10

gate

0.0 0.1 0.2 0.3 0.4 0.5 0.6

10191018

1016

1014

1012

1010

108 106

(a) distance (µm)–0.1

–0.04

–0.02

0.00

0.02

0.04

0.06

0.08

0.100.1 0.2 0.3 0.4 0.5 0.6

(b)

10191016

1012

108

106

0.0

gate

Fig. 3. The simulated electron density map of the cross-section of the AlGaN/InGaN HEMT generated using thehydrodynamic model shows that the electrons spread over the AlGaN barrier into the GaN bulk (a). The simple drift-diffusion model simulation, which does not consider hot-electron effects, confines the electrons to the channel (b).

source draingate

i – Al0.3Ga0.7 N – 25 nm

i – GaN – 2 µm

i – AIN

i – In0.015Ga0.985N – 4 nm

Fig. 1. Our simulations model intrinsically (i)-dopedAlGaN/InGaN HEMTs with a 5 × 1017 cm–3 trappingdensity and σTn = 1 × 10–15 cm–2 capture cross-section.

VD (V)

I D (mA/

mm

)

0 2 4 6

800

600

400

200

08 10

VG = 0 V

VG = –1 V

VG = –2 V

VG = –8 V

Fig. 2. The hydrodynamic transport model (solid lines)that includes hot-electron effects is a better fit to theexperimental data (squares) than the drift-diffusionmodel (dotted lines). Simulations have only beenperformed for VG = 0 and –2 V.




TECHNOLOGY T R A N S I S T O R M O D E L I N G

toward the drain, and creates a potential barrier toelectron flow.

The loss of electrons from the drain channel, whichleads to current collapse, can be prevented by improv-ing the electron confinement in this region. In fact thisimprovement has already been observed in a AlGaN/InGaN/GaN double heterojunction FET structure thatfeatures a potential barrier between the InGaN chan-nel and the underlying GaN layer (Simin et al. 2001).

We simulated the effect of the potential barrier oncurrent collapse by carrying out transient simulationsat differing indium concentrations that are akin to gate-lag measurements. These HD simulations involvesetting the drain voltage to 0.1 V and giving the gate a

1 µs gate pulse from 0 to –5 V, which is then restoredback to 0 V. These simulations confirm the improve-ment in current collapse resulting from the use ofInGaN layers to provide superior channel confinementof the 2DEG.

Field-plate enhancementsOne of the most compelling attributes of GaN HFETsis their high-voltage operation, which produces a higheroutput impedance and wideband impedance match-ing. Recent work suggests that field-plate structuresare effective at reducing current collapse at highvoltages (see, for example, Compound SemiconductorJanuary/February 2006, p25). This is because fieldplates reduce the electric field at the drain-edge of thegate, which leads to a lower electron temperature.

We investigated the electron temperature in GaNHFETs with no field plate, a source field plate and agate field plate. When the device is built without a fieldplate, hot electrons diffuse into the bulk and are trapped,but when a gate field plate is used, the electrontemperature is reduced (see figure 4).

The impact of field plates on current collapse canbe assessed with transient simulations. Using condi-tions identical to those for the transient simulationalready described, but with the exception of anincrease in the drain voltage to 6 V, revealed that thegate field-plate structure yields an almost completerecovery of the drain current, indicating very smallgate lag (see figure 5). This analysis is consistent withreduced hot-electron diffusion and trapping, andshows that the addition of field plates will improvedevice reliability.

Our hot-electron-based model of current collapsesuggests a different mechanism from the “virtual gate”model, which argues that high electric fields at thedrain-edge of the gate cause electrons to be injectedfrom the gate into surface traps (Vetury et al. 2001).However, since surface trapping of electrons is wellknown in compound semiconductors, it is plausiblethat both models play a role in current collapse.

We believe our simulations, which show that fieldplates and double heterostructures can suppress thehot-electron effects in the bulk region of a transistor,can accurately model device behavior, and are aninvaluable tool for optimizing this promising technol-ogy toward its eventual commercialization.

AcknowledgmentsWe would like to thank Michael Shur of the RensselaerPolytechnic Institute, Remis Gaska of SensorElectronic Technology, and Grigory Simin and AsifKhan of the University of South Carolina for guidanceand experimental support.

ReferencesS C Binari et al. 2002 Proc. IEEE 90 1048.N Braga et al. 2004 J. Appl. Phys. 95 6409.J Deng et al. 1999 MRS Fall Meeting Proceedings.G Simin et al. 2001 Jpn. J. Appl. Phys. 40(2) L1142.R Vetury et al. 2001 IEEE Trans. Elec. Dev. 48 560.

About the authorsRic Borges (top left) isSynopsys’ technical marketingmanager, and was previouslydirector of device engineering atNitronex Corporation. Nelson Braga (top right) isSynopsys’ manager ofcorporate applicationsengineering, and was previouslyan associate professor at theUniversity of Sao Paulo, Brazil. Bo Wu (bottom left) iscorporate applications engineer at Synopsys, focusingon heterostructure devicesimulation. Vidas Mickevicius (bottomright) is senior manager ofcorporate applicationsengineering at Synopsys, andhas more than 20 years ofsemiconductor experience.

Fig. 4. GaN HEMTs with gate field plates (c) and source field plates(b) have lower electron temperatures near the drain-edge of the gate than comparable devices without a field plate (a). Note thatthe field plates are not shown in the plots.

time (ns)

norm

alize

d dr

ain

curre

nt

1.0

0.8

0 1000 2000 3000

no field platesource field plategate field plate

0

0.6

0.4

0.2

Fig. 5. Transient gate-lag simulations show that the source and gate field-plate structures both aid the recovery of the drain current,and will consequently improve device reliability.




T ECHNOLOGY C O N F E R E N C E R E P O R T

Researchers seek material solutionsto GaN deficiencies at ISBLLED 2006Researchers are turning to alternative capping layers and GaN and non-polar sapphire substrates topush the output power of LEDs to levels suitable for everyday lighting. Richard Stevenson reports.

At many of today’s conferences you will often see apresentation by a major LED manufacturer outlininga roadmap that begins with the indicator lights of the1970s and points towards the future growth of solid-state lighting. This map gives the impression that per-formance improvements are inevitable and it is easyto avoid thinking about the technological break-throughs that are required to maintain progress.

However, the message was a little different at therecent International Conference on Blue Lasers andLEDs (ISBLLED). At the Montpellier meeting WernerGoetz, director of epitaxial technology at PhilipsLumileds, didn’t speak in vague terms about the needto improve the output of phosphor-converted blue andgreen LEDs to 150 lm/W to win business in generallighting, LCD backlighting, automotive and projec-tion applications – he actually detailed the improve-ments in various LED characteristics that are neededto hit this target.

In his opinion, to produce 150 lm/Wgreen LEDs forred-green-blue light sources will require an increasein extraction efficiency from 65 to 85% and a hike ininternal quantum efficiency (IQE) from 20 to 40%.These gains will have to be delivered in high-power

devices operating at typically 350 mA. For phosphor-converted blue LEDs similar extraction efficiencyimprovements are required, but the IQE must beincreased from 50% to almost 100% – a value that isalready produced in GaAs-based red and infra-redLEDs. Goetz says that there are already laboratoryresults at Philips Lumileds and other claims within theindustry of 85% extraction efficiency in GaN-basedLEDs. “We now have to focus on increasing the IQEthrough improvements in material quality.”

Material issues…Goetz believes that all of the material quality improve-ments must be compatible with uniform, reproducibleand high-throughput manufacturing, and he says thattoday’s issues can be sub-divided into problems asso-ciated with either the n-type, active or p-type regionsof GaN LEDs. The low p-type doping efficiency, lowsolubility of the magnesium dopant and high levels ofimpurities such as oxygen and carbon are all associ-ated with the GaN/AlGaN p-type region that provideshole injection and electron confinement. Meanwhile,the performance of the active region is hindered by lowefficiencies at higher currents, difficulties associated

ISBLLED 2006 was held onMay 15–19 in the city ofMontpellier, France, andattracted 228 delegates.

MO

NTPELLIER

TOU

RIS

T BO

ARD

CSJulISBLLED20-22 5/7/06 11:12 am Page 20



TECHNOLOGY C O N F E R E N C E R E P O R T

with growing high-quality InGaN material at longerwavelengths and polarization effects. The latter couldbe improved through the use of non-polar substrates.Goetz says that the typically silicon-doped n-typeregion suffers from tensile strain and growth-relatedissues that are linked to the difficulties associated withdeposition on a sapphire substrate.

According to Goetz, there are also several reasonswhy it is particularly difficult to grow highly efficientgreen LEDs. Firstly, the additional indium contentincreases the piezoelectric strain between the InGaNand GaN layers, which leads to increased carrier leak-age and a reduced IQE due to the quantum-confinedStark effect (see figure 1). Higher-indium-contentlayers also require growth at lower temperatures, butunder these conditions adatoms are less mobile, theimpurity incorporation and the density of “V”-shapeddefects increases and it is more difficult to crackammonia – the standard nitrogen source.

Goetz notes that another problem with today’s dev-ices is reduced external quantum efficiency at high cur-rents (see figure 2). According to him, all commercialInGaN-based LEDs, regardless of color, have peakefficiency at a current density of 1–10 A/cm2. Goetz

said that improvements can be made through devicedesign to reduce current density, but also added thatthe origin of the IQE reduction is not fully understood.

… and solutionsMany of the challenges outlined by Goetz are alreadybeing addressed by several research groups who pre-sented at the conference. Russell Dupuis from GeorgiaInstitute of Technology described an increase in greenLED output produced by switching from a p-dopedGaN layer to a p-doped InGaN layer.

Dupuis explained that the conventional fabricationof green LEDs structures involves high-temperaturegrowth of a p-doped GaN layer, which degrades theactive region by increasing indium compositional fluc-tuations in the InGaN quantum well. His research grouphas tried to reduce this effect with lower temperaturegrowth of a magnesium-doped In0.04Ga0.96N layer thatincreases hole concentration and conductivity.

The team compared MOCVD-grown LEDs madewith a magnesium-doped GaN layer deposited at 930ºCand a magnesium-doped In0.04Ga0.96N layer grown at840 ºC. In both cases the quantum wells were grown at740 ºC, and an inductively coupled plasma etch with asilicon-dioxide mask defined 230 × 230 µm mesadevices that were given Ti/Al/Ti/Au and Ni/Au n-typeand p-type ohmic contacts, respectively.

At a drive current of 100 mA the device with theInGaN p-doped layer had an electroluminescence peakalmost three times higher than the standard structure,and at lower currents this difference was even more pro-nounced. The new structure’s only downside is a smallincrease in the forward voltage from 3.1 to 3.2V, whichDupuis attributes to the abrupt heterojunction betweenthe p-type InGaN layer and the upper GaN quantum-well barrier. However, he thinks that the magnesium-doped InGaN layer could improve current spreadingand reduce diode series resistance, and that the forward-voltage value is still less than most of the latest devices.

One of Goetz’s other suggestions to improve GaNLED output – the use of “non polar” substrates thathave a different crystalline orientation – was a hot topicamong the delegates, including GaN LED and laserpioneer Shuji Nakamura. Unfortunately, his first resultsusing these substrates showed that the device outputpowers were significantly lower than those of standardLEDs. The best blue devices made by the Universityof California, Santa Barbara professor’s team had anoutput of 1 mW at 20 mAdrive current, while the bestgreen LEDs produced 0.26 mW at 250 mA.

Nakamura believes that these low output powers aredue to either stacking faults or point defects. Althoughthe output powers may be lower than expected, thenovel devices are showing some of the other benefitsthat Nakamura had predicted. For example, the emi-ssion wavelength of the non-polar LEDs does notchange with different drive current, which means thattheir color is extremely stable.

These devices also have a lower series resistanceand the carrier concentration in the p-doped layer is anorder of magnitude higher than for standard devices.

15

polar QW nonpolar QW

1050–5–10depth (nm)

VB

–3.4

–3.6

–3.2

0.4

0.2

0.0

3.35 eV

ener

gy (e

V) CB

e–

h+

AlGaN AlGaNGaN

151050–5–10depth (nm)

VB

–3.4

–3.5

–3.3

0.2

0.1

0.0

3.49 eVen

ergy

(eV)

CBe–

h+

AlGaN AlGaNGaN

(a) (b)

Fig. 1. The quantum-confined Stark effect, which occurs in quantumwells (QWs) grown on conventional polar substrates, involves theseparation of electrons and holes with a built-in electric field (a). Thiseffect reduces the internal quantum efficiency. Researchers arelooking to avoid this condition by using non-polar substrates thatallow the growth of devices that do not suffer from the Stark effect (b).

24

20

16

12

8

4

100 101 102 103 104

forward current (mA)

exte

rnal

qua

ntum

effi

cien

cy (%

)

Luxeon ILuxeon III

Luxeon K2

temperature = 25°Cchip: 1×1 mm2

Fig. 2. Commercial high-brightness LEDs, such as the Philips Luxeonrange of 1 ×1mm chips targeting the solid state-lighting market,operate well-outside the external quantum efficiency “sweet spot”.

PATR

ICK

WAL

TERE

ITPH

ILIP

S LU

MIL

EDS

Held in the heart of the historiccity of Montpellier (above),ISBLLED featured a concert fordelegates on the first night ofthe event (below).

MO

NTPELLIER

TOU

RIS

T BO

ARD

ISB

LLED

“The onlyproblem [withnon-polar LEDs]is output power.”Shuji Nakamura




TECHNOLOGY C O N F E R E N C E R E P O R T

Their polarized emission makes them suitable for directbacklighting of LCD displays, which normally requirea polarizing light filter to operate with conventionalLED backlights. “The only problem is output power,”concluded Nakamura.

Improving the device platform The materials-related problems in the n-type region ofthe LED that Goetz described can be addressed byswitching from sapphire to GaN substrates. It is anapproach that has been pursued by Sumitomo ElectricIndustries’Katsushi Akita, who explained: “Our moti-vation is to produce GaN with lower threading dislo-cation densities that could lead to high extractionefficiencies at high current densities.”

Sumitomo’s engineers compared the performanceof MOCVD-grown LEDs with identical active layersthat are built on either c-plane sapphire or HVPE-grown GaN that had a threading dislocation density of less than 1 × 106 cm–2. They grew two types of LED

on each substrate: one featured a 3 nm thick quantumwell and the other a 5 nm thick quantum well. Theresults of external quantum efficiency (EQE) mea-surements performed with a pulsed injection current(10 kHz duty cycle, duty = 5%) are shown in the table.

Akita believes that the decrease in EQE at highercurrents in both of the 3 nm quantum-well devices isdue to carrier localization and that the devices fabri-cated on GaN contain non-radiative centers (NRCs)that are unrelated to threading dislocations. The den-sity of these NRCs is lower in the LEDs with 5nm thickquantum wells because these devices have a lowerindium concentration (10% rather than 14%), said Akita,before backing up this claim with cathodoluminescenceimages of the structures that showed fewer defects.

All the chips were then mounted p-side down toimprove thermal management and increase light extrac-tion, but this only benefited the LED with 5 nm thickquantum wells that was grown on GaN. For this devicea 20 mA drive current produced 17 mW at 30% EQE,while a 200mAcurrent produced 148mWat 26% EQE.

Other presentations focused on the use of low-polarityGaN epitaxial layer overgrowth to eliminate stackingfaults and the benefits of selective-area growth byHVPE. The need to generate substantial improvementsin external quantum efficiency is critical to the futuresuccess of solid-state lighting, and the discussions atISBLLED should go some way to ensuring that LEDmanufacturers’device roadmaps can be trusted.

“

take control

For accurate monitoring in any type of reactor environment, with real-time results you can

trust, ORS is your answer.

We make the world’s most advanced intelligent thin-film monitoring systems. Our unique,bespoke hardware and software packages offer innovative solutions for production and R&D.

Our fully customisable, bolt-on monitoring systems can be installed on any film-depositionreactor. Our software programs use fully automated, quantitative, real-time analysis to generate the information you need, second by second, leaving you free to recalibrate yourprocesses at the earliest opportunity.

ORS – unparalleled knowledge and expertise at the forefront of today’s cutting-edge technology.

For more information: Tel: +44 (0)1745 535188 www.ors-ltd.com

How can I takecontrol of my epi production?”

LED on sapphire, 3 nm quantum wells 8.8 5.6

LED on sapphire, 5 nm quantum wells 3.1 4.4

LED on GaN, 3 nm quantum wells 8.3 4.9

LED on GaN, 5 nm quantum wells 7.9 6.6

20 mA (%) 200 mA (%)

EQE values for LEDs on sapphire and GaN with different thickness quantum wells


http://www.ors-ltd.com



T ECHNOLOGY G A N T R A N S I S T O R S

Fujitsu proves reliability of GaN HEMTs Fujitsu’s GaN program has led to proven device reliability, a fabrication process using low-costconducting 3 inch SiC substrates, and record output power from MIS-HEMTs, says Toshihide Kikkawa.

Wireless mobile networks will continue to evolve, andfrom 2010 we can expect to see the roll-out of 4Gservices offering transmission speeds of more than100 Mbit/s. These next-generation networks willrequire faster, more powerful transmission amplifiersthat will consume more energy and take up more space.This could lead to larger base stations operating withhigher running costs if the incumbent and relativelyinefficient silicon LDMOS is deployed, and so effortsare being directed at developing alternative technolo-gies, such as GaN HEMTs, for future networks.

These GaN HEMTs are efficient, can produce highoutput powers from relatively small chips sizes, andhave higher breakdown voltages and cut-off frequen-cies than transistors based on other semiconductormaterials. At Fujitsu we have already demonstratedhighly efficient push–pull amplifiers based on thismaterial delivering 250 W wideband code divisionmultiple access (W-CDMA) signals. However, if theseGaN transistors are to fulfill their promise and providea viable alternative to silicon LDMOS, then improve-ments have to be made in terms of reliability, distor-tion, cost and gate leakage at high output powers.

At Fujitsu Laboratories in Kanagawa, Japan, weaddressed all these issues by investigating the perfor-mance and refining the design of our GaN HEMTs. Ourstandard structure was used to study device reliability(see figure 1a), while modifications to the substratesand top structures provided reductions in cost andimprovements in output power (see figure 1, b and c).

Addressing failure mechanismsOur reliability tests were carried out with the GaNHEMT mounted on a conventional metal/ceramicpackage. We addressed the mechanisms limitingreliability – transconductance dispersion and large-signal drain-current degradation, which is commonlyknown as current collapse – with a fabrication processthat controls the polarization-induced surface chargeswith a moderately-doped n-type cap layer. Thisapproach differs from that of many researchers, whouse an AlGaN cap, or dope the GaN cap either moreheavily or not at all. Our moderately-doped GaN capis not fully depleted, and can suppress the strength of the electric field under the gate and screen thetrapping effect.

We found that device degradation, which is thebiggest obstacle to GaN HEMT commercialization,can occur on three different time scales: either just afew seconds, over one to two hours, or after severalhours. The most rapid degradation produces an expo-

nential increase in the gate leakage current and can beeliminated by improving the interface stability betweenthe gate electrode and the GaN surface.

Sudden device failure can also occur after one to twohours of a high-temperature direct-current stress testand appears as a drain current spike (see figure 2). Thelikelihood of failure depends on the quality of the epi-taxial layers and the processing steps. The transistorcan still operate after it has degraded, and room-temperature leakage current measurements at low gate-source voltages cannot identify the damaged devices(see figure 3, p24). However, if the test is repeated at150 °C the defective device is exposed by a leakagecurrent increase of two to three orders of magnitude.

We also witnessed a gradual degradation in theHEMToutput power of less than 0.5dB during a room-

source gate drain

SiNn-GaN

n-AlGaNu-AlGaN

GaN

SiC

2DEG

sourceTi/Al

gateNi/Au

drainTi/Al

SiNn-GaN

n-AlGaN 2DEG

i-GaN

AIN (~10 µm)

n-SiC (0.04–0.09 Ωcm)

source gate drain

n-GaN

n-AlGaN

GaN

SiC

2DEG

SiN

(a) (b) (c)

Fig. 1. At Fujitsu, we investigated reliability and yield issues with our standard structure, which is madeon semi-insulating SiC substrates (a). The devices are reliable, but costly, and savings can be made byswitching to cheaper conducting SiC substrates and inserting an AlN buffer layer into the structure (b). Athigher output powers gate leakage current from the GaN HEMTs can impede performance, but this effectcan be eliminated by using GaN MIS-HEMTs that have the gate contacted to the SiN layer instead of then-doped GaN cap (c). n-doped, undoped and intrinsically doped layers are designated by n, u and i.

time (s)

drai

n cu

rrent

(au)

104

5

103102101

4

3

2

1

0

Fig. 2. The sudden increase indrain current after several hoursof operation appears to belinked to device processing andepitaxial layer quality. This testwas conducted at 150 °C andat a gate-source voltage of –3 V.

This report is based onToshihide Kikkawa’s paper“Recent progress of highly-reliable GaN HEMTs for massproduction” that won the bestpaper award at this year’s CSMantech show. The otherauthors on the paper were KenjiImanishi, Masahito Kanamuraand Kazukiyo Joshin.

CSJulFUJITSU23,24,26 5/7/06 11:21 am Page 23



TECHNOLOGY G A N T R A N S I S T O R S

temperature P3dB-RF stress test (this compression testcompares RF input and output signals, and determinesthe point at which gain falls by 3dB from the linear gainthat occurs at low input powers). This degradationdepends on the process conditions of the surface layerand the quality of the buffer layer. We addressed theseconcerns by optimizing the transistor’s pulsed I-V characteristics with fabrication process improve-ments, and produced devices with suppressed efficiencydegradation (see figure 4). By preventing any devicedegradation on this, and the two far shorter time scales,we proved the reliability of the GaN HEMTs .

Cutting manufacturing costsGaN HEMT prices will strongly depend on yield andconsequently substrate quality. Today, many SiCsubstrates have pits that can reduce device efficiency– due to an increase in the source-drain current – ifelectrodes are deposited on top of these defects (seefigure 5, p26). This is not a significant issue for rela-tively low-power transistors that have gate electrodesof typically 1 mm, since the yield for these devices isonly weakly dependent on pit density. However, hexag-onal pit densities of less than 1 cm–2 are needed for thelarge-periphery gate die used for transistors deliver-ing higher outputs, such as 100 W.

GaN HEMTs also need to have a low memory effect– a small value for the hysteresis between the inputand output powers. This effect can be examined bylooking at a third-order intermodulated signal that is

produced from two signals and results from the non-linearity of the amplifier. This third-order signal isrelatively weak over a wide frequency range, whichis due to SiC’s good thermal conductivity and highload impedance.

We also built GaN HEMTs on n-doped SiC, whichcould cut production costs by between two-thirds andone-half. These lower cost substrates offer a good mixof price and performance, because although they arenot as cheap as silicon or sapphire, they produce thesame thermal memory effect as semi-insulating SiC.

These substrates do have the problems that are asso-ciated with any conductive substrate, such as parasiticcapacitance and isolation leakage, but we addressedthese issues by inserting a 10 µm thick HVPE-grownAlN buffer between the transistor and a 3 inch SiCsubstrate (see figure 1b). The quality of the oxidizedAlN layer strongly influences the GaN layers’rough-ness and device performance (see table 1). The root-mean-square (RMS) value for surface roughnessincreased with scan area, and we found it was easier todiscriminate between the different surfaces with thesewider scans.

Scaling it upTwo years ago, we reported that we could produce100 W GaN HEMTs on 2 inch conductive SiC sub-strates, and now we are developing the transistors on3 inch material because this will deliver lower manu-facturing costs. However, large substrates bring theirown problems, such as bowing, and MOVPE growthof the GaN HEMTstructure produced an epiwafer bowof 8.7 µm. This level of deformation did not preventthe deposition of high-quality 0.8 µm long gates, andwe were able to make 1 mm-gate-periphery GaNHEMTchips from the entire wafer. The transistors wereoperated at 60 V and delivered a continuous-waveoutput power density of 7.0W/mm, and a power-addedefficiency (PAE) of 70% at the 3G base-stationtransmission frequency of 2.14 GHz. Standard devia-tions for the power density, linear gain and PAE at 50 V across a whole wafer were 0.42 W/mm, 0.2 dBand 3%, respectively.

Next-generation networks will need to operate at

Pout (W/mm) 5.2 5.2 3.1PAE (%) 63 56 51Current collapse small small largeAIN FWHM (ω scan) (arcsec) — 280 970AIN FWHM (ω–2θ) (arcsec) — 25 76GaN FWHM (ω–2θ) (arcsec) 31 39 42RMS @1 µm2 (nm) 0.16 0.27 0.41RMS @25 µm2 (nm) 0.22 0.48 1.95Mobility (cm2/V/s) 1750 1670 1610

SI Device A Device B

The effect of AIN quality on GaN HEMTsTable 1. The AlN buffer-layerquality strongly influences deviceperformance. Device A, whichhas a lower surface roughnessthan device B, has superior RF and DC characteristics thatare comparable with devicesgrown on semi-insulating (SI) substrates.

VGS (V)

gate

leak

age

curre

nt (A

/mm

)

1

10–4

0–2–3

10–8

10–10

10–12

10–14

10–6

–1

(d) 150°C after degradation

(c) room temperatureafter degradation

(b) 150°C before test

(a) room temperaturebefore test

time (h)

∆P ou

t (dB)

0.5

15005000 1000–0.5

–0.4

–0.3

–0.2

–0.1

0.0

0.1

0.2

0.3

0.4

10

8

6

4

2

0

–2

–4

–6

–8

–10

∆l ds

(%)∆Pout

∆lds

Fig. 3. (left) Room-temperaturetests of the gate leakage currentat gate-source voltages (VGS)close to 0 V cannot distinguishbetween degraded andundamaged devices. At 150 °C,however, the degraded deviceshave a much higher leakagecurrent at all VGS values.

Fig. 4. (right) The small changesin output (∆Pout) and drain-source current (IDS) when thedevice is operated at a drain-source voltage of 60 V prove thereliability of these GaN HEMTs.



Unlike abrasive sawing, Laser MicroJet dicing does not suffer from tool wear and this combined with small kerf width results in significant cost savings.

[email protected]

Laser MicroJet dicing is superior to conventional laser cutting thanks to the water jet guiding the laser beam, which prevents heat damage and contamination to the work piece.

Synova has re-invented the laser

The new automatic Laser Dicing System (LDS 300A)

cremIChi-tech semiconductor & IT specialist recruitment consultancy

Investigate • Research • Design • Lead • Collaborate • Out PerformBELGIUM

Our client is at the forefront of forward thinking Research & Development. Employing the world’s greatest engineering brains we are on a mission to enhance ourprinciple teams of Scientists and Researchers all of whom hold exemplary academic records but also specialise in some of the most technologically critical areassuch as Nanotechnology, Solar, Bio, Multimedia and RF. Research into these areas will offer clients the ability to identify future needs and demands up to 10 years

ahead of time, so it’s your creative and imaginative side coupled with your demonstrable technical & academic strengths we want.

Snail Mail to cremIC Search & Selection, Communications House, 26 York Street, London, W1U 6PZ

To apply or enquire contact our retained consultants quoting the unique reference number and call 0044 (0) 870 787 5543 or email [email protected]

The challenge of integrated systems, like autonomous wireless sensing systems, is to join devices andcomponents with different functionalities using different technologies. Various heterogeneous integrationschemes, such as Multi-Chip-Modules, Systems-in-Package, (3D) Wafer-Level-Packaging, 3D-System-In-Chip, require lithographic performance beyond ‘traditional’ IC lithography for planar device processing.To this end we are looking to recruit a Senior Lithography specialist to take responsibility for R&D onlithographic process technology. You will compile and manage a ‘library’ of lithographic processes,which will be the basis to select and customize optimized lithographic process for given requirements ina variety of advanced packaging and interconnect integration schemes. You will define and executeinnovative research in the area of novel resists, dedicated photosensitive polymers, and/or new - highaspect ratio driven - lithographic equipment. You are responsible for deliverables in joined developmentprograms with equipment vendors, publishing research results in journals and conferences, and buildingup intellectual property. You will therefore closely collaborate with other process and integrationscientists and process assistants in the integrated systems department. Minimum Requirements: PhD in Sciences or Electronics, demonstrable Lithography experience, drive and ambitionto be part of an internationally renowned research team.

Quote: 151/5/CSSS Quote: 151/4/CSSJ

3D integration is one of the key routes for continuedminiaturisation of performance electronic systems that will mergeheterogeneous devices which has opened a position for adesigner / researcher to take on the architectural challenges ofbuilding 3D integrated systems containing at least 1 layer of ultrathin Si dies embedded by thin film processing. After addressingand overcoming the technological process challenges, you willidentify system demonstrators in the area of wireless systems forcommunication and/or sensing applications. This complextechnology will be mastered by bringing together our in houseexpertise in the fields of thin film processing, microbumping, 3Dintegration, reliability and RF applications. Minimum Requirements: PhD in Sciences, Physics or Electronics, min5 years in process integration, R&D experience with integrated systemsand/or packaging techniques.

Senior Researcher3D Integration – Ultra Thin Chips

Senior LithographyResearch Scientist

CSJulyAdverts25 3/7/06 14:54 Page 1


http://www.synova.ch



TECHNOLOGY G A N T R A N S I S T O R S

higher efficiencies in order to reduce power consump-tion. This means that the GaN HEMTs will have tooperate much closer to their saturation point, but thisleads to very high gate-leakage currents that degradethe device’s reliability and amplification characteris-tics. To address this issue we developed GaN metal-insulator-semiconductor (MIS) HEMTs (see figure 1c)that deliver 110 W output with no forward gate leak-age (see figure 6). These devices avoid having a semi-conductor surface in contact with an AlGaN insulatinglayer that can easily oxidize and degrade performance,

and are the first insulated gate transistors that producegreater than 100 W. Initial reliability tests show thatthey have a stable leakage current over several hours.

Our GaN development program has addressedreliability issues in GaN HEMTs, demonstrated thatthese devices can successfully operate when grownon lower-cost conducting substrates, and led to thefirst GaN MIS-HEMTs capable of output powersgreater than 100 W. We believe this progress showsthat there is a very promising future for GaN transis-tors in tomorrow’s wireless infrastructure market.

About the authorToshihide Kikkawa is a seniorresearcher at Fujitsu, and isresponsible for the fabricationprocesses to produce GaNchips that can be used in high-power amplifiers for RFinfrastructure applications.

VDS (V)

I DS (A

/mm

)

2550 10

10–3

10–4

10–5

10–6

10–7

15 20

(a) no pit

(b) onepit

(c) two pits

Fig. 5. SiC substrates with a high density of hexagonal pits canreduce device yield, particularly for large-sized devices. If the gate electrode touches a pit then the device’s source-drain current(IDS) rises and its efficiency falls.

Pout (dBm)

I g (µA/

mm

)

553530 40

100

45 50

80

60

40

20

0

–20

Schottky gate HEMT

MIS-HEMT

Fig. 6. The insulated-gate MIS-HEMTs have no significant gatecurrent leakage (Ig) even at high output powers, making them strongcandidates for amplifiers in next-generation wireless networks.

3D surface profilingmicro-roughness of semiconductorwafers with sub-angstrom RMS

• clean room ready

• 0.1Å resolution

• fully automated

• 0.02nm noise floor

• non-destructive

• 0.3Å repeatability

visit our website today for details and specificationswww.taylor-hobson.com [email protected]

Talysurf CCI 9150an innovative new instrument for high precision 3D profileanalysis using a patented correlation algorithm to find the coherencepeak and phase position of an interference pattern produced bya selectable bandwidth light source; sub-angstrom resolution isachievable on virtually any surface. Stainless steel and anodizedconstruction for clean room compatability.


http://www.taylor-hobson.com



http://www.shivatec.com




PRODUCT SHOWCASE / CLASS I F I ED

COMPANYBULLETINBANNER

Promote your company message, products and jobs with this banner for only £500 per issue. Contact David Iddon on +44 117 930 1032,or e-mail: [email protected]

Probe StationsLake Shore Cryotronics, IncLake Shore offers cryogenic, magnet-based, load-lock, and high vacuumprobe stations. They provide a platform for measurement of magneto-transport, electro-optical, parametric, high Z, DC, RF and microwave (up to67 GHz) properties of materials and test devices. Typical materialsmeasured include nanoscale electronics, quantum wires and dots,

semiconductors, and spintronic devices.Features include temperatures from 1.5 K to475 K, vertical or horizontal magnet fields, upto six ultra-stable micro-manipulated probearms, and up to 4-inch wafer probe capabilities.Awide selection of options and accessoriesmake it possible to configure a probe station tomeet your specific measurement applications.

Contact: Lake Shore CryotronicsTel: +1 614 891 2244E-mail: [email protected]: www.lakeshore.com

Low-cost semi-insulating AIN/SiCsubstrates for GaN-based HEMTsTechnologies and Devices International, IncLow defect substrates for high power GaN-based transistors and UVemitters. 10-20 microns thick insulating AIN layer provides close latticematch for GaN device epitaxy. High thermal conductivity of both AIN layerand electrically conducting SiC substrate result in excellent heat removalfor high power devices. As grown and polished 2- and 3-inch AIN/SiC

substrates are available.Other types of GaN, AlN,AlGaN and InN epitaxalmaterials are also available.

Contact: TDI, Inc12214 Plum Orchard Drive,Silver Spring, MD 20904, USATel: +1 301 572-7834Fax: +1 301 572 6435E-mail: [email protected]: www.tdii.com

X’Pert PRO MRDPioneering advanced materials researchPANalytical’s X’Pert PRO Materials Research Diffractometer (MRD)delivers fast, flexible and future-proof X-ray analytical performance acrossa comprehensive range of applications for advanced semiconductor, thin

film and nanomaterials research. Key isPANalytical’s proprietary PreFIX system forall major optics and stages. Modularcomponent exchange without realignmentgives virtually limitless capacity to adapt tothe changing needs of any researchlaboratory. PANalytical’s X’Pert PRO MRDXL handles wafers up to 300mm. Automaticwafer loading is available as an option.

Contact: PANalytical B.V., P.O. Box 13,7600 AA Almelo, The NetherlandsTel: +31 (0)546 534444E-mail: [email protected]: www.panalytical.com

High-quality SiC substrates for advanced opto-, high-power and high-frequency electronics.Visit www.sicrystal.de or [email protected]

Damage-free laser dicing of 300 mmwafers: the new LDS 300 ASynova S.A.The LDS 300 is Synova’s newest laser dicing system based on itsinnovative Laser MicroJet technology. Designed to provide chipmakers

with a high-speed tool using alarge working area (up to 300 mmwafers), the LDS 300 allows dam-age- and contamination-freedicing of silicon and compoundsemiconductor material.

Contact: Delphine Perrottet,Synova S.A., Chemin de la Dent-d’Oche, CH-1024 Ecublens,SwitzerlandTel: +41 21 694 35 00Fax: +41 21 694 35 01E-mail: [email protected]

Emerging Markets for GaN ElectronicsElectronic devices based on gallium nitride promise todeliver unrivalled power and thermal performance for RF applications, but widespread adoption of thetechnology depends on whether manufacturers canproduce high-performance devices at low cost.

This new Technology Tracking report, from the publishersof Compound Semiconductor, analyses the challengesfacing GaN device makers as they seek to developcommercial products that can compete effectively withexisting RF technologies.

The report includes critical information in each of thefollowing areas:• Addressable markets • Material choices • Key metrics • Reliability • Market evolution • Organizations

For more information, see www.technology-tracking.com

OUTNOW

CSJulyProdShowcase27 5/7/06 13:24 Page 27

http://www.sicrystal.de



http://www.lakeshore.com


http://www.panalytical.com




http://www.tdii.com

http://www.technology-tracking.com



T ECHNOLOGY R E S E A R C H R E V I E W

Ateam of engineers from the Naval ResearchLaboratory (NRL) in Washington, DC, claimto have broken the record for microwaveamplifier efficiency with a new InAs HEMTdesign. The team’s two-stage narrowbandamplifier dissipates only 350 µW, while deliv-ering 20 dB of gain at 2.5 GHz.

According to the NRLgroup, the amplifier’slow power dissipation could increase the life-time of battery-powered modules that provide

narrowband gain in the S-band (2–4 GHz) andvery low noise.

The InAs HEMTs were grown by MBE onsemi-insulating GaAs (001) substrates, andfeatured an AlGaSb buffer layer to accommo-date the 7% mismatch between GaAs andInAs. The design used a 3 µm source-drainspacing and gate lengths and widths of 0.3 µmand 50µm, respectively. The transistor’s chan-nel mobility was 25 000 cm2/Vs.

Using the InAs transistors to form an AC-coupled two-stage amplifier on a 1 × 0.75 inchcircuit board produced record low-power dissi-pation at a bandwidth of 4%.

The prototype amplifier was built in hybridform for convenience, but the design could eas-ily be adapted for MMIC fabrication.

A German team from laser-diode manufac-turer Vertilas and the Walter Schottky Institutein Garching has extended the continuous-wave (cw) room-temperature emission wave-length of electrically pumped VCSELs from2.0 to 2.3 µm.

This means that the multimode monolithicInGaAlAs/InP VCSELs have the potential tooperate as sources for tunable-diode laser-absorption spectroscopy at wavelengthsbeyond 2 µm, where many important gaseshave absorption peaks. If singlemode versionsof these lasers can be built, they could be used to detect CO2 and CO, which absorb at2004 and 2332 nm, respectively.

That would enable these lasers to replacethe edge-emitting distributed-feedback InPand GaSb lasers that are used today for gasdetection. Edge-emitting lasers are inferiorbecause they do not exhibit a wide tuning

range, or offer high-speed measurement.The latest laser is based on Vertilas’buried

tunnel junction design (see Compound Semi-conductor October 2005 p32). Its emissionwavelength was extended from 2.0 to 2.3 µmby increasing the indium composition in thecenter of the InGaAs quantum well, which sig-nificantly lowers the transition energy. Thecenter of this quantum well contains a 1 nmthick InAs layer, and the InGaAs compositionis varied from this region to the interface withthe barrier, where InGaAs is lattice-matched.

The multimode BTJ-VCSEL produces1.47 mW cw output at 0 ºC and 0.74 mW at20 ºC, but ceases to emit in cw mode at 45 ºC.At 15 mA drive current the VCSEL emits at2300.6 nm, while 30 mA produces lasing at2304.8 nm, giving a maximum room-temp-erature tuning range of 4.2 nm. Side modesuppression is more than 20 dB at 15 mA, and

more than 10 dB at 30 mA.The researchers say that the effective thresh-

old current is 5 kA/cm2, assuming that thecarrier diffusion length is 1.5 µm. This isdouble the threshold current of shorter-wave-length BTJ VCSELs. However, the higherfigure is not surprising because at longer wave-lengths the losses from Auger recombinationare stronger, as are free-carrier and interva-lence band absorption.

Team member Markus Ortsiefer toldCompound Semiconductor that the collabora-tion is now developing singlemode versionsof the VCSELs for sensing applications. Herevealed that these lasers will be implementedinto systems for the detection of CO after theyhave been qualified.

German collaboration extends room-temperature VCSEL emission to 2.3 µm

Journal referenceM Ortsiefer et al. 2006 Electron. Lett. 42 640.

by Richard Stevenson in Montpellier, France.Cermet’s Varatharajan Rengarajan told dele-gates at the recent International Conference onBlue Lasers and LEDs (ISBLLED) that theAtlanta, GA, company has produced electro-luminescence from ZnO LEDs fabricated onn-doped ZnO substrates.

ZnO LEDs offer potential advantages overthose made from other compounds, such aslower threshold voltages, lower growthtemperatures, wet-etch processing techniquesand a larger exciton energy. However, progress

has been hampered by problems with thegrowth of p-doped layers that are needed toform these devices.

Although Cermet can produce its own50 mm ZnO substrates with an etch pit densityof 104 cm–2, p-doping of these substrates isdifficult. However, the company has devel-oped an MOCVD process to produce p-typeZnO layers with a carrier concentration of3 × 1017 cm–3 and a mobility of 4.6 cm2/Vs.

LEDs with a turn-on voltage of 3 V werefabricated using indium dots and multiplemetal layers for the n-type and p-type contacts,respectively. When driven at 140 mA thesedevices showed strong emission at 384 nm,which Rengarajan believes is due to donor–acceptor pair recombination between shallowdonors and nitrogen luminescent centers.

Cermet makes ZnO LEDson n-type substrates

Journal referenceW Kruppa et al. 2006 Electron. Lett. 42 688.

L A S E R S

L E D S

InAs HEMT prototype breaks efficiency recordM I C R O W A V E A M P L I F I E R S

Cermet has fabricated ZnO LEDs with a turn-on voltageof 3 V and an emission peak at 384 nm.

CER

MET

CSJulResRev28 5/7/06 11:25 am Page 28


LED Quarterly InsightsFrom Institute of Physics Publishing,the publishers of LEDs Magazine.Order the complete set of four reports todayfor £595/$995/¤795.www.technology-tracking.com/order

LED Quarterly Insights is a series of four in-depth reports, providing you with incisive analysis of the crucial technologyinnovations and commercial opportunities in the fast-moving LED industry. LED Quarterly Insights is essential reading forequipment and materials suppliers, LED device and module manufacturers, and lighting-systems developers.For more information, please visit our website www.technology-tracking.com or e-mail [email protected].

www.technology-tracking.com

Report 1:High-power LEDsEfforts are still continuing to increase the total lumen outputand improve the efficiency of high-power LEDs. This reportanalyses the technical innovations being made at both thechip and module levels, as well as the measures beingtaken to make high-power LEDs more price-competitive withtraditional light sources.

Report 2:Performance andstandardsSustained growth in the LED industry is being hampered bythe confusion that surrounds the performance metrics usedto characterize LEDs, as well as the many different packagesavailable from LED manufacturers. This issue will analysethe measures that are being taken, and must be taken in thefuture, in order for the LED community to achieve greaterstandardization and continued industry growth.

Report 3: White LEDsThe colour performance of white LEDs continues to be amajor concern for lighting-systems developers and LEDmanufacturers. This report will evaluate current strategies toaddress such issues as colour variation between LED die;techniques for measuring colour output; colour shift duringoperation; and methods to produce white light moreefficiently and with better spectral properties.

Report 4:Packaging and opticsThis edition of LED Quarterly Insights will assess whichpackaging techniques are most likely to yield practical andaffordable LED solutions, and will review new and emergingmethods for optical design that will help to deliver the mostefficient lighting systems.

Published in collaboration with

Competitive intelligence on emerging technologies

Technologytracking

ALL FOURREPORTS

AVAILABLENOW ONCD-ROM

http://www.technology-tracking.com/order




http://www.veeco.com/leds

csjulcover 5/7/06 8:58 am page 1 compound semiconductoriopp.fileburst.com/cs/cs_12_06.pdf ·...

Documents