Transcript
Polycrystalline diamond UV-triggered MESFET receivers This article has been downloaded from IOPscience. Please scroll down to see the full text article.2012 Nanotechnology 23 075202(http://iopscience.iop.org/0957-4484/23/7/075202)Download details:IP Address: 192.167.226.218The article was downloaded on 23/02/2012 at 09:58Please note that terms and conditions apply.View the table of contents for this issue, or go to the journal homepage for more HomeSearchCollectionsJournalsAboutContact usMy IOPscience IOP P UBLISHING N ANOTECHNOLOGY Nanotechnology 23 (2012) 075202 (6pp) doi:10.1088/0957-4484/23/7/075202 Polycrystalline diamond UV-triggeredMESFET receivers G Conte 1,2 , E Giovine 2 , M Girolami 1 , S Salvatori 1 , A Bolshakov 3 ,V Ralchenko 3 and V Konov 3 1 University Roma Tre, Via Vasca Navale, 84, 00146 Rome, Italy 2 Institute for Photonics and Nanotechnology, CNR, Via Cineto Romano, 42, 00156 Rome, Italy 3 A M Prokhorov General Physics Institute, RAS, 38 Vavilova Street, 119991 Moscow, RussiaE-mail: gconte@fis.uniroma3.it Received 11 August 2011, in final form 27 December 2011Published 20 January 2012Online at stacks.iop.org/Nano/23/075202 Abstract Optically triggered UV sensitive receivers were fabricated on polycrystalline diamond assurface channel MESFETs. Opaque gates with asymmetric structure were designed in order toimprove charge photogeneration mainly within the gate–drain region. Photogenerated holescontributed to the channel charge by assistance of the local electric field, in such a wayimproving the current signal at the drain contact. The sensitivity to UV light is demonstratedby using 3 ns wide laser pulses at 193 nm, well over the diamond bandgap. The receivertransient response to such laser pulses shows that the photogeneration process is only limitedby the pulse rise time and charge collection at the drain contact completed in a time scale of afew nanoseconds. Such opaque gate three-terminal devices are suitable for application inemerging photonic technologies, for power-management system optical receivers, wherecopper wires and EM shielding can be replaced by lightweight optical fibers.(Some figures may appear in colour only in the online journal) 1. Introduction The electronics industry is ever more focused on thedevelopment and optimization of devices based on widebandgap semiconductors with high breakdown field andlarge saturation velocity values, mandatory for devicesoperating in the high frequency and power output regimes.Moreover, in photonics applications requiring small dark current, sensitivity to UV radiation, solar blind behaviorand highly selective response in the active spectral regionare being proposed [1]. On the other hand, owing to the potentiality of enabling high performance transistorsoperating in the microwave region, chemical vapor deposition(CVD) polycrystalline diamond is currently receiving a greatdeal of interest for the fabrication of field effect activedevices [2–4]. According to expectations, surface channel MESFETs suitable for high frequency and high powerapplications have been demonstrated [5, 6]. Moreover, the increasing interest in imaging ultraviolet sources like excimerlaser and synchrotron radiation UV facilities is also fosteringresearch efforts to develop active matrix array detectors basedon such material, showing solar blind spectral response andsuitable for obtaining optical signal gain. Because of itsunpaired physical characteristics and the availability of largearea specimens, polycrystalline diamond is today expectedto be the ideal candidate for wide-area UV sensitive arrayreceiver fabrication [7–9]. Selectivity, sensitivity, optical gain and high speed operation are mandatory requirements forthis kind of detectors used to transform optical stimuli toelectric signals. These requirements can be met by opaqueSchottky-gate field effect transistors (OpFETs) where carriergeneration is triggered by UV light impinging within thegate–drain area. The first UV activated diamond FET wasdemonstrated by Lansley et al [10]. Since that time, mainly due to the difficulties in modeling the quasi 2D hole Fermi gas(2DHG) formed below the Schottky contact as a consequenceof the hydrogenation process, no mathematical modeling of OpFETs fabricated on diamond is found in the literature.Among the proposed theories, the so-called surface transferdopingmodel[11]proposedbyMaier etal istodaycommonlyaccepted. Under the framework of the electrochemical surfacetransfer doping model [12], redox couples in the adsorbed 10957-4484/12/075202 + 06$33.00 c 2012 IOP Publishing Ltd Printed in the UK & the USA Nanotechnology 23 (2012) 075202 G Conte et al Figure 1. Arrhenius plot of the diamond sample used, before andafter surface hydrogenation. water layer provide the surface acceptor levels necessary forinitiating electron transfer across the diamond/air interface,therebyresultinginanaccumulationofsubsurfaceholesinthediamond. The surface conduction of these holes contributesto the formation of Schottky p-type junctions with somemetals like aluminum and chromium. Modeling of opaquegate optically activated FET devices was investigated byShubha et al [13] and Alsunaidi et al [14] on GaAs based structures.Noisestudiesandanalysisofdynamicperformancewere also reported [15, 16]. On other hand, the analysis and interpretation of the observed responses of diamond basedOpFETs to the impinging radiation cannot be completelyexplained by these models based on p–n junctions. We believethat overgap UV radiation, penetrating only a few micronsinside the bulk diamond, with the assistance of the localelectric field at the gate–drain junction, contributes to theimprovement of an already formed 2DHG channel, whereaselectrons are swept out to the ground electrodes. Resultson UV sensitive opaque gate MESFET receivers fabricatedon polished polycrystalline diamond are here reported in amore complete form compared to the previous preliminarywork [17]. The very fast response to ArF excimer laser radiation in the nanosecond regime and the sensitivity togate and drain voltage variations demonstrate the suitabilityof these architectures for application in emerging photonictechnologies able to work as fast switching optical receivers. 2. Experimental details 2.1. Technology issues The CVD diamonds used in this work were 0 . 8 × 0 . 8 cm 2 , 0.027 cm thick, polycrystalline substrates growninto a microwave plasma using a CH 4 /H 2 gas mixture.The specimens with polished surfaces were cleaned inorganic solvent and dipped into a strong oxidant mixture Figure 2. Scheme of the asymmetric gate design aimed atimproving the photogeneration area. Figure 3. Optical microscope view of one of the developed deviceswith maximized sensitive area. of HClO 4 , HCl and HNO 3 . The lower roughness sidewas then hydrogenated at 700 ◦ C inside an Astex S1500apparatus for 35 min. During this treatment, the diamondreleased oxygen termination and surface dangling bonds aresaturated with hydrogen. Local hydrogen termination inducesa band bending that is responsible for the p-type conductionexplained as a 2D hole Fermi gas. In figure 1 we report theconductivity Arrhenius plot of the sample before and after thesurface hydrogenation.The average sheet resistance after hydrogenation wasmonitored over the whole substrate surface by four-pointVan der Pauw measurements and was 14 ± 2 k . Planardevices with different structures and layouts were fabricatedby standard optical UV lithography, together with teststructures: transfer length measurement (TLM) pads, circulardiodes, Hall bar and multi-grid contacts. The designed mask contains a total of 60 devices with different geometriesand gate width / length ( W / L ) ratios, and structures separatedinto four quadrants. The structure of the asymmetric gate(i.e. different source-to-gate and gate-to-drain distances)is aimed at improving carrier photogeneration inside thegate–drain area (see figure 2). Moreover, structures withdifferent source-to-drain distances were designed to evaluatethe minimum detectable light. Results on this point willbe reported in another paper. Figure 3 reports an opticalview of a device realized to maximize photogeneration. On 2 Nanotechnology 23 (2012) 075202 G Conte et al Figure 4. Total resistance R T measured on different transfer length(TLM) structures versus the pad distance d . Z is the pad length. Thetransfer length L T resulted to be 2.5 µ m; ρ C is the specificresistance, R c and R Sh are the contact and sheet resistances,respectively. R is the figure of merit of the linear fit. the other hand, the design is also aimed at the use of ground–signal–ground (GSG) probe heads for high frequencymeasurements.Thermally evaporated gold was used for the drain andsource contact fabrication, whereas more than 200 nm of aluminum was deposited for the realization of a UV-opaquegate (i.e. complete absorption at λ = 193 nm) electrode.The Ohmic contact resistivity, measured for three differenttest structures (C3-B, C5-B, C14-B) realized on the samediamond sample (Ra 270), resulted to be 2 . 5 × 10 − 5 cm 2 ,as reported in figure 4. This value is two orders of magnitude higher than expected for high frequency and high powerdevices, where the series resistance must be reduced to avoidpower loss effects. The contact resistivity is dependent on thehydrogenation recipe, average surface roughness, orientationand dimensions of the grains.An Ar / O 2 mixture and reactive ion etching processingwere used to remove hydrogen from the device’s non-activeareas, accomplishing at the same time leakage currentreduction and structure isolation. Double and single gateasymmetric structures were designed in order to increase andexploit UV photogenerated charges. UV radiation impingingwithin the gate–drain region creates charges in the channel,contributing to the transistor effect. Active area dimensions inthe range 400 µ m 2 to 12 . 000 µ m 2 were investigated. 2.2. Characterization setup The characterization in the dark was accomplished by usinga Karl-Suss probe station with picosecond heads. TLMand Van der Pauw measurements were carried out with acombination of a Keithley 6221 current generator and a2182A nanovoltmeter. The circular diodes were measured Figure 5. I – V characteristics of different circular Schottky diodes(C8-G, C17-G, etc) with lateral depletion: G, diameter = 250 µ m; M, = 200 µ m; P, = 150 µ m. The hydrogenatedring diameter is 10 µ m. Inset: optical photograph showing theoff-center alignment discussed in the text. by using an HP4140B picoammeter, whereas the MESFETdevices were analyzed by evaluating the drain current, I ds , asa function of the gate–source voltage ( V gs ) and drain–sourcevoltage ( V ds ) . The measurements were conducted at roomtemperature. In figure 5 the I – V characteristics of somecircular diodes are reported: the depletion region of suchdiodes is lateral. It is worth noting the trend at high voltagevalues, addressing the presence of a series resistance. Inreality, as demonstrated by the photo reported in the inset, dueto a wrong mask alignment the distance between the centralaluminum Schottky contact and the external gold one (i.e. thethickness of the circular ring) was not constant at 10 µ m.This resulted in a series of diodes with different depletionregions that activated one after the other with increasingslopes, like the effect of increasing the diode’s quality factor.This interpretation is sustained by the evidence of the initial n = 1 . 2 diode’s quality factor and the series resistance regionfound well over 2 V.To carry out the characterization under pulsed light,the device was wire bonded in a chip carrier and placedin front of the exit of a 193 nm Neweks PSX100 ArFlaser (2 × 3 mm 2 , full width at half maximum ( FWHM ) = 3 . 5 ns, 10–100 Hz, 4 . 5 mJ / pulse). The temporal response of avacuum phototube (Hamamatsu H8496-11) was monitored asa reference for the intrinsic line shape of the laser pulse [17]. Both bias voltages, V ds and V gs , were obtained from a deck of Ruben–Mallory batteries. The drain bias voltage ( V dd ) wasapplied either through a picosecond 5575A bias tee or through100 k resistance to reduce the current flow inside the device.The transient photocurrent signal was measured on the highimpedance input of a LeCroy Wavepro 960 digital samplingoscilloscope (2 GHz bandwidth and 16 GS s − 1 on a singlechannel). 3 Nanotechnology 23 (2012) 075202 G Conte et al Figure 6. Normalized transcharacteristics of a representativeOpFET device. Figure 7. Normalized output characteristics of the same device asreported in figure 6 with W / L = 25. 3. Results and discussion The normalized transcharacteristics in the dark of arepresentative fabricated OpFET device are reported infigure 6. It is worth noting the curve linearity in the controlled regime addressing short channel behavior and the presenceof sub-threshold currents for V ds voltages lower than thethreshold V th = 0 . 0 ± 0 . 2 V. Such a current is mainly dueto the polycrystalline nature of the substrate and to thepresence of surface current paths along grain boundaries.Figure 7 reports the output characteristics of the same Figure 8. Maximum value of normalized I ds for differentbutterfly-shaped devices ( L = 2 µ m; W = 50, 100, 200, 300 µ m).The standard deviation (1 σ ) is 0 . 53 mA mm − 1 whereas the varianceis 0 . 3 mA mm − 1 . asymmetric device with width / length ( W / L ) ratio of 25.As is apparent, after the channel pinch-off all curves showan initial decreasing trend with increasing the V ds voltage,and a stabilization to a lower current value. This overshootis not common to all fabricated devices and from time totime appears for MESFETs fabricated on different qualitydiamond substrates. Among possible causes of this behaviorcan be suggested difference in the field dependent mobilitydue to a non-homogeneous electric field distribution withinthe gate–drain area or nonlocal transport of charge carriers ondifferent conduction paths [18]. The performance of devices with different geometries is quite constant, as shown infigure 8, where maximum drain current I ds values for V gs =− 1 . 5Vand V ds =− 15Varereported.Thisresultisindicativeof good hydrogenation homogeneity, tight distribution of grain dimensions and low average roughness.The voltage dependent transconductance, g m , of therepresentative device is reported in figure 9. The maximumvalue reaches 8 mS mm − 1 and is observed around − 0 . 9 V.The observed decrease for V gs values greater than V gs =− 1 V can be ascribed to series resistance effects or mobilitydegradation due to grain boundary scattering.The response of the tested device to pulsed UV light as afunction of the control gate voltage V gs is reported in figure 10for a device with 2 µ m gate length and active area equal to3 . 2 × 10 3 µ m 2 . The pull-up drain resistance equal to 100 k was biased at V dd =− 8 V. The voltage signal monitored onthe oscilloscope increases with the gate voltage: at zero biasthe device is insensitive to the laser light and a thresholdvoltage is observed around 0.2 V. The signal below this valuecan be considered the device’s noise level. It is worth notingthe asymmetry of the signal’s shape: a fast pulse rising atlower V gs values, whereas the decay appears slower. Withthe increase of the control voltage the pulse shape becomes 4
