chapter 3. high-frequency (>100 ghz) electronic devices · high-frequency (>100 ghz)...

Chapter 3. High-Frequency Electronic Devices

Chapter 3. High-Frequency (>100 GHz) Electronic Devices

Academic and Research Staff

Professor Qing Hu

Graduate Students

Edouard A. Garcia, Rajesh K. Gupta, Brian R. Jacobson, Zachary K. Lee, Jurgen H. Smet, Rolf A. Wyss

Undergraduate Students

Elliot E. Hui, Jason S. Kim

Technical and Support Staff

Barbara A. King

3.1 Facility for Millimeter-wave and THzFrequencies

Professor Hu's laboratory is equipped with variousmillimeter-wave and infrared sources which cangenerate coherent and incoherent radiation up to30 THz. These include Gunn oscillators at W-bandfrequencies (75-110 GHz); a frequency doubler,tripler, and quadrupler using Schottky diodes at200, 300, and 400 GHz; an optically pumped far-infrared laser which generates coherent radiation upto 8 THz; and an infrared Fourier transformspectrometer which is capable of performing linearspectroscopy from 45 GHz to 30 THz and beyond.The laboratory is also equipped with variouscryogenic millimeter-wave and infrared detectors.These include Si composite bolometers, InSb hot-electron bolometers, SIS (superconductor-insulator-superconductor) receivers, and high-To Josephsondetectors. There are many infrared cryostats whichcan cool the devices from 0.3 K to 77 K.

3.2 Millimeter-wave and InfraredSuperconducting Focal-plane ReceiverArrays

Sponsors

MIT Lincoln LaboratoryNational Aeronautics and Space Administration

Grant NAG2-693

Project Staff

Brian R. Jacobson, Edouard A. Garcia, ProfessorQing Hu in collaboration with MIT Lincoln Labora-tory

Although the range that includes millimeter-waveand far-infrared frequencies is one of the most

underdeveloped, these frequencies have greatpotential for applications in remote sensing andcommunication. The millimeter-wave and far-infrared frequency range falls between the twoother frequency ranges in which conventional semi-conductor devices are usually operated: (1) micro-wave frequency range and (2) near-infrared andoptical frequency range.

Semiconductor devices which utilize the classicaldiffusive transport of electrons such as diodes andtransistors have a high frequency limit. This limit isset by the time it takes for electrons to travel acertain distance. Currently, electron mobility and thesmallest feature size which can be fabricated bylithography limit the frequency range to below 100GHz. It is not likely that this limit can be pushedmuch higher. Semiconductor devices based onquantum mechanical interband transitions, however,are limited to frequencies higher than those corre-sponding to the semiconductor energy gap, which ishigher than 10 THz for most bulk semiconductors.Therefore, a large gap exists from 100 GHz to 10THz in which very few devices can operate.

The gap energies of conventional superconductorssuch as Nb are in the range of 100 GHz to 2 THz.This coincidence makes superconducting devicesnatural candidates for millimeter- and submillimeter-wave applications. At millimeter-wave frequencies,the superconducting video detectors have quantumefficiency e/hw, that is, a transport of one electronfor one incoming photon; while the superconductingcoherent receivers have their sensitivities limitedonly by the zero-point fluctuation of vacuum. Suchreceivers have been used widely in astrophysicalstudies. Additional applications are feasible inspace-based communication and far-infrared spec-troscopy which requires the ultimate sensitivity.


We are currently developing a novel quasiopticalscheme to couple the millimeter-wave and infraredsignals to superconducting devices through a com-bination of an etched horn antenna and a planarantenna supported by a thin (- 1 Em) membrane,as shown in figure 1. This scheme combines theadvantages of the easy fabrication of lithographicthin-film structures and the high antenna efficienciesof horn antennas. Because of the absence of sub-strate losses in this scheme, it is expected that aTHz receiver could be constructed using all-Nbsuperconductor-insulator-superconductor (SIS) junc-tions. More important, because of the cumbersomemechanical structures used in conventional wave-guide technology, microwave receivers have beensingle-element devices, as opposed to CCDimaging arrays at optical frequencies. Spatial scanhas been achieved mechanically. Using the novelquasioptical scheme mentioned above, focal-planedetector arrays can be fabricated lithographically ona single Si wafer, as shown in figure 2, so that far-infrared imaging becomes feasible.

The first step of this project is to demonstrate thefeasibility of fabricating high-quality SIS devices onfree-standing SiN membranes and then to demon-strate that these devices can survive thermalcycling. Using the microfabrication facilities inGroup 86 at MIT Lincoln Laboratory, we have fabri-cated several wafers of SIS junctions with a criticalcurrent density of 4,000 A/cm 2. We then etched theSi wafer underneath anisotropically in a KOH sol-ution. A view of an SIS junction with a log-periodicantenna on a 1-pm thick SiN membrane is shownin figure 3. Extreme care was taken to protect theSIS junctions from the KOH etchant during theanisotropic etching. Figure 4 shows the I-V curvesof the SIS junction measured before and after theSi wafer was etched away. There are no noticeablechanges in the I-V characteristics of the device.Clearly, our work has established the feasibility ofbuilding millimeter-wave SIS receivers using micro-machined horn antenna structures.

3.3 Photon-assisted QuantumTransport in Quantum Point Contacts

Sponsor

National Science FoundationGrant ECS 91-09330

Project Staff

Rolf A. Wyss, Professor Qing Hu, inwith Cristopher C. Eugster, ProfessorAlamo

collaborationJesis A. del

Figure 1. Example of anisotropic etching in a <100>silicon wafer. The opening of the wafer yields a pyramidalhorn with a flare angle of 35.50 .

Figure 2. Perspective view of a two-dimensional hornimaging array.

Quantum transport has been one of the most activefields in solid-state physics in recent years.Advances in material preparation have madequantum phenomena profound in electron transportfor many semiconductor quantum devices such asquantum point contacts, quantum dots, quantumwires, quantum wells, superlattices, etc. In cleansamples and at low temperatures, electrons cantravel through the whole sample without sufferingphase-destructive scattering. Extensive work hasbeen done to study various features of such phase-coherent quantum transport. However, so far theexperiments reported are limited to dc transportmeasurements or far-infrared spectroscopymeasurements.

It is well known in the field of superconducting tun-neling that photons can assist the tunneling

110 RLE Progress Report Number 135


Figure 3. Perspectives on a silicon nitride membranestructure. (top) View of device seen through the mem-brane using backside illumination on microscope.(bottom) View of same device using top-side illumination;the 700 angle of the anisotropic etch can be made out bynoticing the dark rim around the border of the device.

process, provided the tunneling is elastic so thatelectrons do not suffer inelastic scattering. In abroad sense, elastic tunneling is a phase-coherentquantum transport in a classically forbidden region.Therefore, all the results of photon-assisted tun-neling can be applied to the study of photon-assisted quantum transport in semiconductordevices. This will provide a new dimension to studythe exciting quantum transport phenomena. Novellong-wavelength optoelectronic devices may alsoemerge from this research.

In this NSF-sponsored project, weinteraction between far-infraredballistic electrons in quantum point

are studying thephotons and

contact devices,

Figure 4. DC I-V curves for a device on SiN membrane(top) before etching, and (bottom) after etching the Siwafer in KOH.

whose schematic is shown in figure 5(a). We havefabricated several antenna-coupled quantum pointcontact devices using a combination of optical andelectron-beam lithography. Figure 6 shows theSEM photographs of one of the devices. The split-gate electrodes (the vertical leads) also serve asthe terminals of a log-periodic antenna, whose func-tion is to couple the far-infrared radiation of milli-meter wavelength to the point contact of submicrondimensions.

The dc transport measurement of the drain/sourcetransport showed fifteen quantized conductancesteps, indicating the high quality of the devices.Under a coherent far-infrared radiation (285 GHz), aprofound photon-induced drain/source current isproduced throughout a gate voltage range in whichthe device exhibits the behavior of a one-dimensional electron system. The amplitude of thephoton-induced current is comparable to that corre-sponding to a quantized conductance step, and itoscillates with the gate voltage with a period corre-


(a) Drain

"ZJ//

d

Gatate

Source

(b)n= 2

- n=l

photon

Ef

Figure 5. (a) Schematic of a quantum point contact. (b)Schematic of an electron waveguide which can be usedto analyze the transport of electrons through the quantumpoint contact shown in (a).

sponding to the gate voltage between adjacent con-

ductance steps, as shown in figure 7.

Our preliminary analysis suggests that the observedfar-infrared photon-induced drain/source current is

mainly due to heating (or a bolometric effect). Sincethe excitation in the two-dimensional electron gas is

gapless, energy absorption can occur everywherewithin the electron gas. Thus, the effect of photon-

assisted quantum transport will always be accom-

panied by an effect due to heating. This aspect is

quite different from photon-assisted tunneling in

superconducting tunnel junctions in which the

superconducting energy gap prevents energy

absorptions in the leads. Our analysis also sug-

gests several improvements must be made to

enhance the effect of photon-assisted quantumtransport and minimize the heating effect.

According to the recently developed theory of

photon-assisted quantum transport, the photon-assisted process mainly occurs near the classicalturning points of the subbands under concern.

Thus, to enhance the photon-assisted quantum

process, it is crucial to concentrate the radiation

field in the central region of the point contacts. This

requires use of thick (- 2000A) antenna terminals

made of good conductors. The quantum effect can

also be enhanced by irradiating the devices at fre-

quencies much higher than 300 GHz so that the

photon energies are much greater than the thermalbroadening of the Fermi surface.

Figure 6. (a) SEM (with a magnificationquantum point contact with log-periodicCentral region of the quantum point contact.of the split-gate electrodes is 0.15 Ipm.

of 400) of aantenna. (b)The opening

3.4 High-Tc SuperconductingJosephson Devices

SponsorDefense Advanced Research Projects Agency

Contract MDA972-90-C-0021

Project Staff

Rajesh K. Gupta, Professor Qing Hu, in collab-

oration with Twente University, The Netherlands


0 0.9

6 0.3

4 - o

2 T0K .0.3

Sf = 285 GHz

-2.2 -2 -1.8 -1.6 -1.4 -1.2

VGS (V)

Figure 7. Drain/source conductance and the photon-induced drain/source current as functions of the gatevoltage in a region where the conductance steps(measure without radiation) are well defined.

Many superconducting analog devices have beendemonstrated to have higher sensitivities, higherspeed and frequency limit, and lower power dissi-pation than competing semiconductor devices.Among them, the most successful ones are Super-conducting QUantum Interference Devices(SQUIDS) and millimeter-wave detectors.

The discovery of superconductors with a supercon-ducting transition temperature higher than liquidnitrogen temperature (high-To superconductors) hasopened up new and exciting possibilities in elec-tronic device technology. A high temperatureversion of the superconducting devices mentionedabove will find a much wider range of applicationswherever refrigeration is a problem. The keyelement of all superconducting analog and digitaldevices is the Josephson junction. This junction isformed by two superconductors weakly coupledtogether electrically. The supercurrent flowingthrough the Josephson junction oscillates at theJosephson frequency f = 2eV/h (500 GHz/mV)when a voltage is applied across the junction. Mostof the useful applications of Josephson devicesresult from this high-frequency oscillation. Thesedevices include high-frequency and high-speedsignal processors and high-sensitivity SQUID mag-netometers whose high precision results from aver-aging over many cycles of the Josephsonoscillations.

In collaboration with the Dutch group headed byProfessor Horst Rogalla, which is the leading groupin fabricating high-quality high To Josephson junc-tions, we have studied the response of a


YBCO/PBCO/YBCO ramp-type junction to coherentradiation at 176 GHz and 270 GHz. The I-V charac-teristic of the junction closely resembles the predic-tion of the resistively-shunted-junction (RSJ) model.The critical current and normal resistance productIcR, of the junction is 0.25 mV at 5 K. Themillimeter-wave radiation is coupled to the junctionvia a quasioptical structure that focuses the radi-ation onto the junction through a yttrium-stabilizedZrO2 substrate. At 176 GHz, we have observed asmany as six Shapiro steps at the maximum powerlevel of our Gunn Oscillator-pumped frequencydoubler. This implies a Josephson oscillation up to1 THz. Shapiro steps are still visible up to 65 K, asshown in figure 8.

The amplitudes of the zeroth, first, and secondShapiro steps, as functions of the square root of theradiation power, agree remarkably well with aBessel function fit, as shown in figure 9. ThisBessel-like behavior is an indication that the junc-tion is voltage-biased at the radiation frequency.This is in great contrast to other experimentscarried at lower frequencies, where a current-biasedbehavior is usually observed. For a voltage-biasedRSJ-like junction, the detector response can be pre-dicted analytically. The current responsivity of ourdevice, defined as the induced dc current per unitRF power, is calculated to be 2 x 103 A/W, which iscomparable to the quantum efficiency e/hw- 1.4 x 103 A/W at this frequency.

We have also measured the response of thehigh-To Josephson junction to coherent radiation at270 GHz. At this frequency, due to a combination ofthe heavy RF loss in the ZrO2 substrate and thelack of radiation power, we have observed only thefirst Shapiro step.

Figure 8. I-V curves of the high-Tc Josephson junctionirradiated by coherent radiation at 176 GHz and taken atdifferent temperatures. Shapiro steps are still visible at 65

113

400

200

0 o

-200

-400


Aln

2°c

0.6

00.6

0 1 2 3

20a (- PRF1/ 2)

4 5

Figure 9. Measured amplitudes (normalized by the crit-ical current in the absence radiation) of the zeroth, first,and second Shapiro steps as functions of the square rootof radiation power at 176 GHz near 5 K. The solid linesare Bessel functions.

3.5 Far-infrared (THz) Lasers UsingMultiple Quantum Wells

Sponsor

U.S. Army - Research OfficeGrant DAAL03-92-G-0251

Project Staff

Jurgen H. Smet, Zachary K. Lee, Professor QingHu, in collaboration with Professor Clifton G.Fonstad, Jr.

Semiconductor quantum wells are man-madequantum mechanical systems in which the energylevels can be chosen by changing the sizes of thequantum wells. Typically, the frequency corre-

sponding to the intersubband transitions is in thefar-infrared or THz range. Naturally, long-wavelength photoelectric devices, such as far-infrared lasers, which utilize the intersubbandtransitions have been proposed and subsequentlystudied.

Significant progress has been made recently towardthis goal. Large oscillator strengths of intersubbandtransitions have been observed in far-infraredabsorption spectroscopy experiments. An intersub-band spontaneous emission with a power level of- 10- 7W has been observed. Although these prelim-inary results are encouraging, two major challengesto building a quantum-well far-infrared laser stillremain. One is to achieve a high degree of popu-lation inversion between two subband levels, andthe other is the confinement of photons within theactive region. Optical pumping suffers from lowefficiency and low modulation speed. Thus, it willbe difficult to achieve a sufficient gain to maintain alasing oscillation, and, therefore, unsuitable forcommunication applications. As far as optical con-finement is concerned, the usual dielectric wave-guide confinement method, commonly used inoptical and near-infrared laser systems, is not appli-cable at far-infrared frequencies because the con-finement (on the order of a wavelength) is too largecompared to the dimensions of the active region.

In this project, we will use a novel multiplequantum-well (MQW) device to circumvent thesetwo problems. In this device, whose energy banddiagram is shown in figure 10 with and without a dcbias voltage, electrons are injected selectively intoan upper subband level of a wide lasing quantumwell through a narrow filter quantum well. Afterrelaxing into a lower subband in the lasing well, theelectrons are then removed selectively to a collectorthrough another narrow filter quantum well. Thisfilter well prevents the electrons in the uppersubband level from tunneling out to the collector.Using this method of selective injection andremoval, a high degree of population inversion canbe achieved, provided the tunneling rate of theelectrons from the ground state to the collector isgreater than the relaxation rate of the electronsfrom the upper subband to the ground state. Ourcalculation indicates that this should be achievedeasily.


1InGaAs wells

d = 360 Ad, = 230 A

QuantumElectron well lasinginjector medium

Figure 10. Band profiles (not drawn in proportion) of theproposed far-infrared laser device under (a) zero-bias,and (b) lasing condition in which Vbias- (E2 - E,)/e E -20 mV. The device consists of onenarrow quantum-well energy filter on both sides of thelasing quantum well. The width of the narrow filter wells(made of Ino53Ga 0 47As) is approximately 230 A so theenergy of its first subband lies in between the first twosubbands in the wide lasing well whose width is approxi-mately 360 A, and will have a radiation frequency of 5THz due to the intersubband transitions.

In the proposed MQW laser device, a parallel-platewaveguide that confines the photons can be formedbetween a heavily doped semiconductor injectorand collector. At a doping level of 2 x 1018 /cm3 , wecan raise the plasma frequency in the injector andcollector to about 17 THz. It has been reported thatthe electron mobility, t, for such a high doping levelis 5 x 103 cm 2/Vs. This plasma is mostly reflectivewith a penetration depth as short as 0.8 tm for alasing frequency in the range of 1-8 THz. This con-finement is sufficiently tight so that the lasingthreshold current density is on the order of severalhundred A/cm 2, which should be achievable.

The first step of this project is to fabricate multiple-quantum-well structures with different well widthsand barrier thicknesses and to perform tunnelingspectroscopy on them to determine the relativepositions of the subbands in different quantumwells. Because of the small energy differenceswhich are associated with the relatively widequantum wells, the tunneling spectroscopy must beperformed at liquid helium temperatures. Our initial

Electroncollector


measurements on the tunneling I-V characteristicsrevealed many features which could be attributed toelastic tunneling and LO phonon-assisted inelastictunneling.

To interpret the fine features of the I-V character-istics with a greater certainty, we have employed apowerful method, magnetotunneling spectroscopy,to study the relative positions of the subbands. Inaddition to spectroscopy, this study has a practicalpurpose for laser applications. The nonradiativeintersubband relaxation rate due to phonon emis-sions is much greater than that of the radiativerelaxation rate. This is mainly because the trans-verse motion of electrons in the lasing quantum wellis unrestricted, so there is a large volume in thephase space into which the electrons can be scat-tered. Similar to the concept employed in atokamak to confine plasmas, a strong magnetic fieldwill restrict the electron motions in the transversedirections. This magnetic field-induced transverseconfinement, along with the longitudinal confine-ment provided by the quantum well, will effectivelyreduce the dimensionality of the electron system tothat of a zero-dimensional system, exactly like thatof quantum dot systems. Therefore, thenonradiative relaxation rate can be substantiallyreduced. Because the magnetic field does notaffect the longitudinal motion of electrons, electricalpumping is still possible. This is difficult to imple-ment for quantum dot systems which are formedgeometrically.

We have performed magnetotunneling spectroscopyon a double quantum-well structure which consistsof 115-A and 200-A In0.53 Gao.47As quantum wells,both are confined by 30 A of Ino.52AIo 48As respec-tively on the top and bottom and separated by a60-A Ino.52AI0 .48 As middle barrier. When a longi-tudinal magnetic field is applied, the tunnelingcurrent is modulated by the field, as shown in figure11(a). The magnetooscillation is periodic with theinverse of the magnetic field 1/B. Thus, a Fouriertransform of this magnetooscillation in 1/B gives apeak which centers at a characteristic field Bo.

The physical origin of this magnetooscillation isscattering-mediated inter-Landau level tunneling.Suppose the energy in the first well is given byE1 + (n, + 1/2)heB/m*, and the energy in thesecond well is E2 + (n2 + 1/2)hB/m*, whereE1 and E2 are the energies of the subbands in thefirst and second well under concern. In the absenceof LO phonon emissions, energy conservationrequires that the two energies are equal. Thus,(n2 - nl)/heB/m* = E, - E2 or 1/B = (n2 - ni)he/m*(E1 - E2). This implies that whenever 1/B is aninteger times the inverse of a characteristic field1/Bo = he/m*(E - E2), the energy conservation issatisfied and the tunneling current will be enhanced,


which gives the oscillatory behavior of the tunnelingcurrent as a function of 1/B, as shown in figure11(a).

This magnetooscillation provides us with veryimportant information about the relative positions ofthe subbands under concern. The energy differencebetween the two subbands is related to Bo throughE, - E2 = heBJm*. Thus, the energy difference canbe measured directly using the magnetotunnelingspectroscopy. Since the energy difference E, - E2changes linearly with the bias voltage, so is thecharacteristic field Bo. By measuring Bo as a func-tion of the bias voltage, we can determine the frac-tion of the total applied bias voltage dropped acrossthe two wells.

100

80

60

(a)

40

0 5 10 15Magnetic Field B [T]

S0.8 I (b)

0.6o0

L 0.4LL

0.2

05 15 25 35

Fundamental Frequency Bf [T]

Figure 11. (a) Tunneling current (at a fixed bias voltage)as a function of the strength of a magnetic field applied inthe same direction as that of the tunneling current. (b)Fourier transform of the tunneling current as a function of1/B.

3.6 Publications

Feng, S., and Q. Hu, "Far-infrared Photon-assistedTransport Through Quantum Point ContactDevices." Submitted to Phys. Rev. B, 1993.

Gupta, R., Q. Hu, D. Terpstra, G.J. Gerristsma, andH. Rogalla. "Near-millimeter-wave Response ofHigh-To Ramp-type Josephson Junctions," Sub-mitted to Appl. Phys. Lett., 1993.

Hu, Q. "Photon-assisted Quantum Transport inQuantum Point Contacts." Appl. Phys. Lett. 62:837 (1993).

Hu, Q., R.A. Wyss, C.C. Eugster, J.A. del Alamo, S.Feng, M.J. Rooks, and M.R. Melloch. "A NovelSubmillimeter-wave Detector Using QuantumPoint Contacts." Proceedings of the FourthInternational THz Conference, University ofCalifornia at Los Angeles, April 1993.

Wyss, R.A., C.C. Eugster, J.A. del Alamo, and Q.Hu. "Far-infrared Photon-induced Drain/SourceCurrent in a Quantum Point Contact." Sub-mitted to Appl. Phys. Lett., 1993.

Thesis

Garcia, E.A. Fabrication of AII-Nb SIS Tunneltions for Submillimeter-wave Detectors.thesis. Dept. of Electr. Eng. and Comput.MIT, 1992.

Junc-S.M.Sci.,


chapter 3. high-frequency (>100 ghz) electronic devices · high-frequency (>100 ghz)...

Documents