ixers wit Structures H. G. LeDuc Center for Space Microelectronics Technology, Jet Propulsion Laboratory , Pasadena, CA 91109, USA.

Transcription

1 3584 IEEE TRANSACTIONS ON APPLIED SUPERCONDUCMVITY, VOL. 7, NO. 2, JUNE lyy xers wt Structures Me Bn, M. C. Gads*, J. Zmudznas and T. G. Phllps Downs Laboratory of Physcs 32047, Calforna Insttute of Technology, Pasadena, CA 91125, USA. H. G. LeDuc Center for Space Mcroelectroncs Technology, Jet Propulson Laboratory , Pasadena, CA 91109, USA. I Abstract We recently reported [SI a quasoptcal SIS mxer whch used Nb/Aloxde/Nb tunnel junctons and a normalmetal (Al) tunng crcut to acheve an uncorrected recever nose temperature of 840 K (DSB) at 1042 GHz. Were we present results on several dfferent devce desgns, whch together cover the GHz frequency range. The mxers utlze an antreflectoncoated slcon hyperhemsphercal lens, a twnslot antenna, and a twojuncton tunng crcut. The broadband frequency response was Radal st measured usng Fourer transform spectrometry (FTS), and s n good agreement wth model calculatons. Heterodyne tests were carred out from 400 GHz up to 1040 GHa, and these measurements agree well wth the FTS results and Transformer wth calculatons based on Tucker's theory. Keywords SIS Mxers, Submllmeter Recevers Mcrostnp nductor \ f 'Mcrostrp transformer I. INTRODUCTION HE theoretcal upper frequency lmt for SIS mxers s twce the gap frequency, or about 1.4 THz for mxers usng Nbtrlayer tunnel junctons [1][3]. To date, good results have been obtaned up to 800 GHz usng allnobum crcuts [4][6]. Modern SIS mxers generally use ntegrated tunng crcuts to resonate the juncton capactance. However, snce the RF loss of superconductors ncreases dramatcally above the gap frequency, the nose temperatures of allnobum mxers ncrease substantally above 700 GHz. To reduce ths RF loss above the gap frequency, tunng crcuts made of a hgh conductvty normal metal such as alumnum have been fabrcated, and encouragng recever performance has been obtaned [7], [8], [9]. In prevous papers, we descrbed results for a 1 THz quasoptcal I recever wth an alumnum tunng crcut [8], [lo]. An uncorrected doublesdeband (DSB) recever nose temperature of 840 K was obtaned at 1042 GHz, wth the devce cooled to 2.5 K. Corrected for the losses and thermal nose from the LO/sgnal njecton beamspltter, the recever nose temperatures are N 770 K from 822 GHz Ths work was supported n part by grants from NASA (NAGW 107 and NAG2744), NASA/JPL, and a NSF PYI grant to J.Z. The juncton fabrcaton was performed at the Center for Space Mcroelectroncs Technology, Jet Propulson Laboratory, Calforna Insttute of Technology, and was sponsored by NASA, Offce of Space Access Technology. *Present address: SubmllmeterWave Superconductve Sensor Development Group, Jet Propulson Laboratory , Calforna1 Insttute of Technology, Pasadena, CA 91109, USA. Emal: me@tacos.caltech.edu. Fg. 1. Schematc layout of the mxer and ts equvalent crcut. to 1042 GHz (at 2.5 K). In ths paper, we report on the ex' tenson of ths work to other frequency bands, n an effort to verfy our understandng of the devce performance. We dscuss the desgn and gve test results for Alwrng mxers optmzed for four frequency bands centered at 450,750, 850, and 1050 GHz. The frequency response n drect detecton was obtaned usng Fourer transform spectroscopy (FTS), and agrees wth the predcted response. An ana ss of the mxer performance based on Tucker's theory s also presented. II. MIXER DESIGN The desgn detals for the planar twnslot antenna, t twojuncton SI tunng crcut, the mxer block, and t optcal layout have been presented elsewhere 151, [Ill,[l Fg. 1 dsplays the schematc structure representatve of th mxers used n ths work. The mcrostrp lne between the two junctons serves as the tunng nductor to resonate o the juncton capactances. The bent mcrostrp lnes nectng the junctons to the radal stubs serve as mped transformers. A vrtual ground s formed at the center the tunng nductor, due to the antsymmetrc exctator from the two slot antennas. Therefore, only half of thc crcut needs to be analvzed. we have wrtten a cmputer progra,m to smulate optmze the mxer crcut. The crcut model nclude! the frequencydependent mpedance of the antennas, m /97$ IEEE

2 TABLE I DESIGN PARAMETERS FOR THE TUNING CITCUITS. f Cj 85 ff/pm2 Cj = 65 ff/pm2 (GHz) nd. transf. nd. transf x x x x x x x x x x x x x x x x mm thck quartz plate, whch for some measurements had a Teflon antreflecton (AR) coatng. The 77 K IR flter was ether an ARcoated quartz plate or a sheet of Ztex flm [13]. The mxer optcs nclude a slcon hyperhemsphercal lens, wth ARcoatngs optmzed for each of the four frequency bands, and a polyethelene lens mounted on the mxer block to further shape the beam. Our prevous paper [5] gves more detals on the recever optcs and mxer block constructon. crostrp lnes, and tunnel junctons. The propertes of the normalmetal A1 mcrostrp lnes are calculated usng a method whch ncludes the surface mpedance of AI as gven by the anomalous skn effect for fnte thckness flms [lo]. The smulaton program was used to generate devce desgns for center frequences at 450, 750, 850, and 1050 GHz, by maxmzng the RF couplng averaged over a 100 GHz bandwdth. Several devces were desgned for each band to allow for parameter varatons, partcularly n the juncton area and specfc capactance. For each band, the tunng crcut was optmzed for a sngle value of the juncton area, 1,7 pm2, but two dfferent values of the specfc capactance were used, 65 ff/pm2 and 85 ff/pm2. For each tunng crcut desgn, we ncluded three nomnal juncton areas (1.2 pm2, 1.7 pm2, and 2.3 pm2) on the mask layout. Thus there were four bands, eght tunng crcuts, and 24 dfferent devce desgns. For the desgn, we assumed R,A = 20 Rpm2, whch s approprate for Nb/Aloxde/Nb junctons wth J, M 10 ka cm2, p(300 K) = 2.45 prcm for alumnum, and p(300 K)/p(4 K) M 10 for the AI flms. The measured resstvty rato was about 5 [SI. The A1 flms are , thck, whle two thcknesses of S0 (er = 5.6) were used for the mcrostrp sectons: , for the tunng nductor, and , for the mpedance transformer. The parameters for the eght tunng crcuts are gven n Table I. The dmensons of the mcrostrps are gven as w x 1 (n pm). The length 1 of the nductor s defned as the dstance between the centers of the two junctons. All devces use a short (2.5 pm) secton of mcrostrp, wth the same wdth as the nductor (w = 5 pm), to connect the transformer secton wth the nductor secton. The devces were fabrcated usng a modfed Nb/Aloxde/Nb juncton xocess [8]. No sgnfcant lthographc undercut was ob ;erved for the batch of devces tested. The results presented are labeled usng three or four dgt lumbers that dentfy the devce desgns. The frst one or,wo dgts refers to the frequency band: 4, 7, 8, 10 correspond to 450, 750, 850, and 1050 GHz, respectvely. The ext dgt gves the juncton sze, wth 1, 2, 3 referrng to L.2 pm2, 1.7 pm2, and 2.3 pm2. The last dgt represents pecfc capactance assumed for the desgn, wth 6 and 8 ndcatng 65 ff/pm2 and 85 ff,fpm RECEIVER PERFORMANCE The devces were measured n a test dewar, whose presure wndow was ether a 25 pm thck Mylar sheet or a A. Fourer Transform Spectroscopy The recever response as a functon of frequency was measured wth an FTS system usng the SIS juncton as a drect detector. A descrpton of the FTS system as well as our measurements for allnobum crcuts can be found n the paper by Gads et al. [5]. As n ou,r prevous work, we fnd that the heterodyne nose temperatures correlate well wth the drectdetecton frequency response measured on the FTS (see Fg. 4) nn t Fg. 2. The response measured on the FTS (sold lnes) vs. the calculated RF couplng (dashed lnes) for fve devce desgns (418, 428, 716, 816, and 1028). These devces are all from the same fabrcaton batch. All the smulatons are calculated usng a specfc capactance value of 85 ff/pm2 and the nomnal desgn values for the rest of the parameters. The alumnum rrcrostrp lnes are calculated usng the nonlocal theory of the anomalous skn effect, usng the measured resstvty rato of p(300 K)/p(4 K) = 5. Fg. 2 shows a comparson of the theoretcal and expermental frequency response measured for devces n each of the four frequency bands. The smulaton curves gve the RF couplng effcency, whch s the fracton of the power receved by the antenna that s dsspated n the tunnel junctons. Snce we cannot measure the absolute response, the vertcal scalng of the FTS data s adjusted to best match the smulaton. In addton, the juncton specfc capactance used n the smulatons was adjusted to ft the

3 3586 shapes of the FTS spectra. There s good agreement between theory and experment for a specfc capactance of 85 ff /pm2, whch s the same value we nfer from measurements on allnb crcuts [5]. The agreement between smulaton and experment s reasonably good, gven that the measurements are affected by water absorpton lnes at 557, 752, 986, 1094, 1114, 1163, 1211, and 1229 GHz, and also by FabryPerot resonances (wth a 30 GHz spacng) from the quartz IR flter. B. Heterodyne Measurements We have measured the nose temperatures of several devces, ncludng those n Fg. 2, usng the Yfactor method wth a room temperature hot load and a lqud ntrogen temperature cold load. Fg. 3 dsplays recever nose temperatures as a functon of frequency for several devces at 4.2 K. The ncrease n nose temperature wth frequency s partly due to the decreasng output power of the LO S. At the hghest frequences, thck Mylar beamspltters (25 pm and 51 pm) were needed to couple suffcent LO power nto the relatvely largearea junctons. For the 982 GHz measurement at 4.2 K, whch was obtaned usng a soldstate multpled (x9) Gunn oscllator source [14], the recever nose temperature drops from 1450 K to 950 K after correctng for the loss and thermal nose assocated wth the beamspltter. Ths calculaton s supported by the sgnfcantly lower uncorrected nose temperature that was measured for same devce at 1042 GHz. At ths frequency, a farnfrared laser was used as the LO source, whch produces a good deal more power than the multpled Gunn LO. Thus, a much thnner beamspltter (10 pm) could be employed L W a E, 1000 c. I 0 BOO & L m BOO Fg. 3. The DSB nose temperatures of four AIwrng SIS recevers measured at 4.2 K. The open markers are uncorrected DSB recever nose temperatures, whle the correspondng sold markers nclude a correcton for the LO njecton beamspltter. The hghfrequency devces were also tested at a pumpedlhe temperature of 2.5 K. Upon coolng to 2.5 K, the nose decreases by 100 K 300 K over the band, n part due to the reduced dark current and ts assocated shot nose. For nstance, the 1042 GHz recever nose drops from 1170 K to 840 K upon coolng. Ths 40% mprovement n the recever nose temperature can be largely explaned by the reducton n the subgap leakage current and the ncrease n the gap voltage. Tucker theory calculatons based on these facts ndcate a 20% drop n the mxer nose temperature and a 35% reducton n the converson loss. Therefore, a 25%30% reducton n the recever nose temperature can be expected from these effects. The mcrostrp lnes may have a lower loss at 2.5 K due to a decrease n the A1 resstvty. However, we measured only a 3% decrease n, the AI resstvty. The resultng ncrease n the RF couplng eecency s calculated to be neglgble, only 0.4%, and so ths effect does not sgnfcantly contrbute to the decrease of the recever nose temperature upon coolng x 2 03 s. a, E e 1250 P W Q LL n: L m v) 750 Cl Fg. 4. The FTS spectrum and heterodyne recever performance for devce 716. The measured and smulated RF couplng effcences1 are plotted vs. frequency. The measured recever nose temperatures at a bath temperature of 4.2 K are also shown. Note that the recprocal of the recever nose temperature correlates well wth the FTS drect detecton curve. In Fg. 4, the recever nose temperatures measured for devce 716 at 4.2 K are plotted as a functon of frequency, along wth the FTS spectrum and the calculated frequen response. The broad bandwdth (wth a 3 db band GHz) s typcal of the Alwrng devces, and result of the relatvely low Qfactor of the alumnum crostrp tunng nductance. The recprocal of the recev nose temperature s also plotted, multpled by a scal factor to allow a comparson wth the FTS drectdetect measurement and the crcut calculaton. There s clearly a good overall correlaton between the heterodyne and dd rect detecton measurements, and both expermenta sults agree reasonably well wth the calculaton. Note the calculaton models only the crcut behavor, and not nclude the optcal transmsson or the antenna bea effcency. The frequency dependence of these effects could account for some of the dscrepances between the exper! mental and theoretcal curves, especally consderng thd very broad band response of the mxer. C. Mxer Performance Analyss As an example, we analyze the recever performance r terms of the front end optcs, the mxer, and the IF amplp 500

4 ~ [Dashed: 3587 TABLE I1 ESTIMATED TRANSMISSION AND NOISE OF THE OPTICS: DEVICE 1028 AT 982 GHZ AND 2.5 K For 25 pm beamspltter: RF gan GRF = 0.12 = 9.2 db, RF nose TRF = K 3 For 51 pm beamspltter: RF gan GRF = 96 = 10.2 db, RF nose TRF = EC fer chan, for the case of devce 1028 at 982 GHz and 2.5 K. At 982 GHz, a multpled Gunn LO source was used whch had a nomnal output power of 60 pw [14], and we were unable to obtan optmum pumpng of the juncton when usng a 25 ym thck Mylar beamspltter. Although usng a 51 pm thck beamspltter mproved the pumpng condton, the (uncorrected) recever nose temperature ncreased due to the larger sgnal loss and thermal nose njecton of the thcker beamsplt ter.? v U. 0.1 (a) L I b measured 1.20 calculated m L ; N calculated IV I s 150 wth 51 pm b.s. c (a = 0.58) P, \ Bas voltage (mv) ' 1.00 *' Bas voltage (mv) Fg. 6. Measured and calculated IF output power (a) and Yfactor (b) for devce 1028 at 982 GHz and 2.5 K bath temperature. The calculaton assumes a LO pumpng strength a = 0.47 (see Fg. 5). ' Bas voltage (mv) Fg. 5. Unpumped and pumped dc IV curves for devce 1028 at 982 GHz and 2.5 K. The thcker beamspltter (51 pm) couples more LO power onto the juncton. The calculated IV curves for the two pumpng cases are also plotted. To calculate the mxer performance usng Tucker's SIS theory n the 3port approxmaton [l], the LO pumpng strength a = evlo/hv and the RF embeddng admttance Ye seen by the juncton must be known. The LO voltage a s easly calculated from the unpumped and pumped IV curves. Wth a = 0.47 for 25 pm beamspltter and a = 0.58 for 51 ym beamspltter, the calculated IV curves follow the measured ones very well (see Fg. 5). Snce we I cannot measure the RF embeddng admttance, the value predcted by our crcut program was used: at 982 GHz, Ye = (4.2 +j0.6) G,, where G, = (14 s the normalstate conductance of the juncton. Usng Tucker's theory, the mxer gan Gmx and the mxer nose temperature Tmx are calculated. The theory predcts that the optmum pumpng for the mxer would occur around a = 0.7. To calculate the recever performance, we also need to sccount for the optcal losses and thermal nose, the losses of the alumnum tunng crcut, and the nose of the IF amplfer. The transmsson and nose of' the optcs and tunng crcut are estmated and lsted n Table 11. The estmates are calculated values and were not adjusted to match the expermental results. In the table, Tamb s the ambent temperature of an optcal componlent, T;Fb s the Planck effectve temperature, and T, s the added nose referred to the nput of the component [lo]. Note the use of a 51 ym beamspltter ncreases the RF loss by 1 db and the RF nose by approxmately 100 K, when compared to the use of a 25 pm beamspltter. The shot nose technque [15] [17] was used to obtan the IF amplfer nose TIF and gan GIF. The IF output power &F can be calcullated accordng to 8~ = ~ B A V [(Te$ + Tmx)Gmx + ~F]GIF, (1) where Te' s the Planck effectve temperature of the hot or cold load as seen by the juncton. In other words, T$! ncludes the loss and thermal nose of the front end optcs and the RF tunng crcut. For the case of a 25 pm beamspltter, the calculated and measured IF output power vs. bas voltage wth hot and cold loads at the recever nput are shown n Fg. 6. The calculaton s only N 5% lower than the measurement on the frst photon step below the gap. The agreement s qute good, gven the dffcultes

5 3588 TABLE I11 RECEIVER PERFORMANCE OF DEVICE 1028 AT 982 GHZ AND 2.5 K. TABLE IV DSB MIXER CONVERSION LOSS AND NOISE TEMPERATURE. Devce f Bath T Loss Nose hw/k~ name (GHz) (K) (db) (K) (K) and uncertantes nherent n estmatng the optcal and crcut losses a pror. Fg. 6(b) shows the smlarly good agreement between the measured and calculated Yfactor. The calculated results are lsted n Table 111, where we compare the use of the two beamspltters. Note that when a 51 pm beamspltter s used, the Tucker theory predcts slghtly better mxer performance (lower mxer nose TZ: and smaller mxer converson loss GZ:), because the juncton s better pumped. However, the recever performance s worse snce the thcker beamspltter has greater RF sgnal loss and thermal nose. If enough LO power were avalable and a 10 pm beamspltter could be used, the uncorrected recever nose s calculated to be N 850 K, whch s comparable to the result at 1042 GHz wth a FIR laser LO. The same analyss has been carred out for dfferent devces at several frequences. The calculated DSB mxer converson losses and nose temperatures are gven n Table IV. Note that the calculated mxer nose s on the quantum nose level (hulkg) below the gap frequency, and s only several tmes the quantum level above the gap frequency and nto THz regme. The recever performance at 1 THz s controlled largely by the loss n the alumnum tunng crcut. IV. CONCLUSIONS We have demonstrated SIS mxers wth alumnum tunng crcuts at frequences from 400 GHz to 1 THz. Our crcut desgn program was verfed by comparng ts predctons wth Fourer transform spectrometer measurements. Sgnfcant advances n THz SIS mxers wll requre tunng crcuts wth lower losses, whch may be possble usng hghqualty NbN flms. ACKNOWLEDGMENT We are very grateful to P. Zmmerman for the loan of hs 1 THz multpler chan, and thank G. Blake and P. Stockman for ther help wth the FIR laser measurements. REFERENCES [l] J. R. Tucker and M. 3. Feldman, Quantum detecton at mllmeter wavelengths, Rev. Mod. Phys., vol. 57, no. 4, pp , [Z] D. Wnkler and T. Claeson, Hgh Frequency Lmts of Superconductng Tunnel Juncton Mxers, J. Appl. Phys., vol 62, pp , [3] M. J. Wengler and D. P. Woody, Quantum Nose n Heterodyne Detecton, IEEE J. Quantum Electroncs, vol 23, pp , [4] G. de Lange, C. E. Honngh, J. J. Kupers, H. H. A. Schaeffer, R. A. Panhuyzen, T. M. Klapwjk, H. Van de Stadt, and M. M. W. M. de Graauw, Heterodyne Mxng wth Nb Tunnel Junctons Above the Gap Requency, Appl. Phys. Lett., vol 64, pp , [5] M. C. Gads, H. G. LeDuc, M. Bn, D. Mller, J. A. Stern, and J. Zmudznas, Characterzaton of LowNose QuasOptcal SIS Mxers for the Submllmeter Band, IEEE Bans. Mcrowave Theory. Tech., vol 44, pp , [6] C. E. Honngh, S. Haas, D. Hottgenroth, K. Jacobs, J. Stutzk, Fxed Tuned Wavegude Mxers Around 450 GHz, 670GHz, and 810 GHz for a Dual Channel Recever, Proc. Seventh Id. Symp. Space Terahertz Tech., March 1214, 1996, pp. 6370, Unversty of Vrgna, Charlottesvlle, Vrgna. [7] H. Van de Stadt, A. Baryshev, P. Deleman, T. Klapwjk, S. Kovotonyuk, G. de Lange, I. Laptskaya, J. Mees, R. Panhuzen G. Prokopenko, and H. Schaeffer, A 1 THz Nb SIS Heterodyn Mxer wth Normal Metal Tunng Structure, Proc. Sxth Intl Symp. Space Terahertz Tech., March 2123, 1995, pp. 6671, Caltech, Pasadena, Calforna. [8] M. Bn, M. C. Gads, J. Zmudznas, T. G. Phllps, and H. G LeDuc, Lownose 1 THz nobum superconductng tunnel junc ton mxers wth a normal metal tunng crcut, Appl. Phys. Lett., vol 68, pp , [9] H. Van de Stadt, A. Baryshev, J. R. Gao, H. Golsten, Th. de Graauw, W. Hulshoff, S. Kovotonyuk, H. Schaeffer, and N. Whyborn, Proc. Seventh Intl. Symp. Space Terahertz Tech., March 1214, 1996, pp , Unversty of Vrgna, Charlottesvlle, Vrgna. [lo] M. Bn, M. C. Gads, D. Mller, J. Zmudznas, T. G. Phllps, and H. G. LeDuc, Desgn and characterzaton of a quasoptcal SIS recever for the 1 THz band, Proc. Seventh Intl. Symp. Space Terahertz Tech., March 1214, 1996, pp , Unversty of Vrgna, Charlottesvlle, Vrgna. [ll] J. Zmudznas, H.G. LeDuc, J.A. Stern, and S.R. Cypher, Juncton Tunng Crcuts for Submllmeter SIS Mxers, Bans. Mcrowave Theory Tech., vol 42, pp , [12] J. Zmudznas, N.G. Ugras, D. Mller, M. Gads, H.G. L and J.A. Stern, LowNose Slot Antenna SIS Mxers, %ns. Appl. Superconductvty, vol 5, pp , [13] Norton Performance Plastcs, 150 Dey Rd., Wayne, NJ 07 [14] Radometer Physcs, Bergerwessen Str. 15, 5309 Meckenhem,, Germany. [15] S. Rudner, M. J. Feldman, E. Kollberg, and T. Claeson, Superconductornsulatorsuperconductor mxng wth arrays at mllmeterwave frequences, J. Appl. Phys., vol. 52, pp. 6366, [16] D. P. Woody, R. E. Mller, and M. J. Wengler, GHz recevers for rado astronomy, IEEE 13ans. Macrowave Theorg Tech., vol. 33, no. 2, pp. 9095, [17] N. B. Dubash, M. J. Wengler, and J. Zmudznas, Shot and photonnduced correlatons n 500 GHz SIS detectors, Bans. Appl. Supercond., vol. 5, no. 2, pp , 1995.