THIN FILM BaxSrl.xTiO3 Ku- and K-BAND PHASE SHIFTERS GROWN ON MgO SUBSTRATES F. W. VAN KEULS a, C. H. MUELLEW, F. A. MIRANDA a,and R. R. ROMANOFSKY _, J. S. HORWITZ b, W. CHANG b, and W. J. KIM b _XIASA Glenn Research Center, Cleveland, OH 44135, U.S.A. _Xlaval Research Laboratory, Washington, DC 20375, U.S.A. We report measurements of gold circuits fabricated on four BaxSr_.xTiO3 films doped with 1% Mn grown on MgO substrates by laser ablation. Low frequency measurements of er and tan8 on interdigital capacitors are compared with high frequency measurements of phase shift and insertion loss on coupled microstrip phase shifters done on the same films. The variation in temperature of both high and low frequency device parameters is compared. Annealed and unannealed films are compared. Room temperature figures of merit of phase shift per insertion loss of up to 58.4°/dB at 18 GHz and 400 V dc bias were measured. Kevwords: Ba_Sr_.xTiO3; phase shifters; tunable microwave devices INTRODUCTION Major efforts are underway to develop new tunable microwave devices at frequencies above about 10 GHz where Si becomes too lossy. Inexpensive, compact low-loss devices are desired for a variety of communications and military applications. Recent improvements in the deposition of thin film ferroelectrics have made these materials a candidate for these roles. At ambient temperatures, the greatest attention has been given to the ferroelectric, Ba_Sr_.xTiO3 (BSTO). In BSTO, changing x from 0 to 1 causes the maximum of the dielectric constant to shiftfrom20K to 395 K. Tunable ferroelectric devices are generally used above the Curie temperature (Tc) in the paraelectric regime, where a large dielectric constant is tuned with an applied dc electric field. BSTO films suffer from a high loss tangent, which can be lowered by introducing dopants _'2_. Another avenue for improving films consists of annealing to remove lattice imperfections and increase tunability t31.This paper examines four BSTO films as a function of temperature. The films were patterned with interdigital capacitors for low frequency measurements and coupled microstrip phase shifters (CMPS) for high frequency measurements. DESIGN AND EXPERIMENTAL DETAILS The Naval Research Laboratory deposited these BSTO ferroelectric films using on-axis laser ablation at a temperature of 750 C and a dynamic oxygen pressure of 100 mtorr. The BSTO was deposited to a thickness of 500 nm on substrates of 508 gm thick (100) single crystal MgO. All four films were doped with 1% Manganese. Two films had Ba:Sr ratios of 50:50 and the other two had 60:40. One films of each composition was bomb annealed at 1100 ° C for 6 hours t_l.The 60:40 composition samples have a higher Tc, which usually leads to higher tuning and loss at room temperature. After BSTO deposition, the films were metallized using electron-beam evaporation at NASA Glenn Research Center with a 15 nm chrome (Cr) adhesion layer followed by a 2 I.tm thick Au film. Standard lift-off chemical etching techniques were used to fabricate Au/Cr/BSTO/MgO interdigital capacitors, two different circuits were patterned on these BSTO films. After measurements were complete, the Au was etched off with 1:1 KI:H20 solution and the Cr seed layer was removed with perchloric acid. A second circuit consisting of a Au/Cr/BSTO/MgO four-element CMPS was patterned on the BSTO films in the samemanner.Finally, aCr/Au groundplanewith thesamemetallicthicknesses wase-beamevaporatedonthebackofthesubstrateforthephaseshifters. The interdigital capacitors were measured at 1 MHz using a HP4192A LF Impedance Analyzer. The interdigital capacitor consisted of 100 identical fingers with a finger width of 25 _tm and a gap between fingers of 15 p.m. The finger length was 0.6909 cm long. The capacitance and tan5 measurements were made in a closed cycle He refrigerator at temperatures from 30 K to 330 K and excitation voltages of 50 mV or 100 mV. The performance of the CMPS circuits at microwave frequencies was evaluated by measuring the transmission ($2_) and reflection (S_) scattering parameters between 10 and 20 GHz using an HP 8510 C network analyzer. All loss measurements quoted here include the losses due to the SMA launchers which are estimated to be 0.25 dB. Two phase shifters were measured at temperatures down to 100 K in a closed cycle He refrigerator. The measurements were done in vacuum to protect from dielectric breakdown of the air in the large dc electric fields between coupled microstrip sections. However, tests of CMPS circuits in air have been done after coating the films in thin film bonding wax. No breakdown occurred and the wax had a negligible effect on device performance. The phase shifter design consists of n-coupled microstrip sections in series. Each section functions as a single pole broadband filter whose passband shifts with de bias applied to the ferroelectric. The phase shift is proportional to n. The circuits used in this study were four- coupled section phase shifters. A schematic of a single coupled microstrip section is given in Fig. 1. A photograph of the entire circuit is shown in Fig. 1. The dimensions of the coupled length, 1= 457 lam, the gap between coupled section, s = 10 pm, and the coupled section width, w = 56 I.tm. The total circuit length is 1 cm. In Fig. 2, the dc bias is applied at the top two radial stubs while the bottom three stubs which include the input and output microstrips are held at ground. These phase shifters are fairly narrowband, about12%bandwidth,andtheoptimalfrequencyofoperation, fopt,depends upon the c, and thickness of the ferroelectric film. A detailed discussion of the device properties has been given elsewhere t4_. K----1 ----_ Figure 1. Schematic of a single coupled microstrip section for this design on 508 _m thick MgO, s = 10 Fig. 2. A four element coupled _tm, 1= 457 _tm, and w= 56 iam. The microstrip phase shifter on a 508 _tm total length is I cm. thick MgO substrate. The device was first successfully demonstrated using YBa2Cu3OT._ atop a 1 I.tm SrTiO 3ferroelectric film on a 254 lam thick LaAIO3 (LAO) substrate [41.At 40 K and 16 GHz, those 1cm long cryogenic devices demonstrated 484 ° of phase shift using 375 V dc bias with a figure of merit (K) of 80°/dB phase shift per maximum insertion loss. Transferring this exact LAO design to a room temperature Au/0.3 pm BSTO/LaAIOi structure quickly achieved 200 ° phase shift and 43°/dB figure of merit (K) at 14 GHz and using 400 V dc bias I51. RESULTS The 1 MHz interdigital capacitor results are shown in Fig. 3 and Fig. 4. The dielectric constant of the film was then derived from the measured capacitance by using the conformaI mapping and partial capacitance formulas of Gevorgian et a/. [61 The relationship between C and e, for this 100 finger electrode configuration is, C (pF) = 28.969 + 0.19780er (1) The error in the capacitance measurement and the uncertainties in the dimensions are about 5%. Judging from test circuits on LaAIO3 and MgO, the errors in the partial capacitance analysis of C(Cr) are probably less than 20%. The tan5 measurements shown in Fig. 4 are more sensitive to stray capacitance and calibration errors, including those induced by changing temperature. These measurements have uncertainties of approximately +_0.003 except for Sample 1 which seems to suffer from an admixture of the capacitance. 1000 .... ,.... ,.... _.... _.... _.... ,.. 1500 .... L .... I .... , .... *.... i .... I"' 0 0 V/=m cooflng • 0,AV'/_._rn,,_wormcblo°,.% a00 **21..3_33V_//_m, wcoromiong,.q 2.33 V//Jrn warrn_ng 1000 _ 6OO 400 500 0 0 V//tm eool;ng 0 V/_n_. warmln 9 1.11 V//._m wormir_l 200 4'- 1.94 V/_V coormg 50:50 a= depos;ted 60:40 A=-deposlted 1.94 V//am warming 0 .... 510' ==jI1O0.... 150I, ,,,200l .... 250I , , ..1.,300 00 *'''! 50.... 10I,,,0.lvl..! t50 200.... 25I0.... 30I0t • Temperature (K) Temperature (K) a) c 4000 .... ,o, ...,..,.o.=I....,. .°._;...._%I .... I.... I J''' 4000 .... 0 0IV../.,./J*m coIolr;rr'ig ' j ' ' ' ' I.... I .... I ' ' • 0 Y/._ warming • 1,11 V//.cm worming o_m_,,_e4w__ 3000 3000 + 1.94 V//_.m cooling _- _ = 1.9_' V.//am worming ,J._p "_ _2ooo 2000 ,'Y % 10oo 1000 i -" •,,e.,,..e''' _'_ 60:4o on.,o_I , i i . I0.... _0I0, L ,.._J1._J5_L0J--_ 20I0.... 25I0,, , •30I0, , °o&El o__1 . , .1Io.o... 1I5.o... 20t0, , , ,25I0.... 30I0, . Temperature (K) Temperature (K) b) d) Figure 3. The dielectric constant as a function of temperature for all four samples as derived from the capacitance of an interdigital electrode at l MHz. a) Sample l: BSTO 50:50 as deposited, b) Sample 2: BSTO 50:50 annealed, c) Sample 3: BSTO 60:40 as deposited, d) Sample 4: BSTO 60:40 annealed. 0.025 .... i .... i .... i .... i .... i .... i'' 0.025 .... !.... i _, ,_ l, ,* ,i .... t .... t ' ' 0 0V/,_rr,Vroofing 0 0 V/p'n cool_g 0V/urn wormin9 • 0 V_ warmin 0.020 ; 1,..3333¥_///_c.o.o.h.ng,.9 0020 ._ _ _ 1.11 V/H,m warming _2.33 V/,=m coofin9 lib.:I. _ 1,94 V cod;ng = 2.33 V//,v,n warming 0.015 0,015 *o • %,'. _ 0.010 _-_it $0:50 o= deposlted 9 0.010 "_,_ ""_x_x 60:40 ks-depollted r _,,__ :_.,. 0.005 0,005 ,',{% ..,,-_Y 0.000 0,000 I i ,T.... Ii ,a ,"_r]"_ar_, I i •' ' I..... , ,, i .... I ,, i i I ti i •I .... i .... t , , 50 100 150 200 250 300 0 50 1(20 150 200 250 300 Temperature (K) Temperature (K) c) a) 0.025 .... i .... I' ' ,"r'¢, ,,, i .... i .... i . , 0.025 _,,,i .... i .... i .... 0t0....V/_'n _c.o..o.ling i'' tl_ _, O 0 V cooling • 0 V _ wormln 0.020 "_"'-"t X_'x_,,,•,_2.113VvZ/j_/'_n_m5cwoa.or,ml°i,nl;ngooi'_,_, 0,020 /2/J_' _,__ *+ 11..9161VV//jT.4li"nil"t wcomol'imnging t . i . _ * I,g4 V/;.cm worming 0.0t5 _k[\__ * 2.33 V/_m worming 0.015 I ,,, ,{% .g 0.010 '/,' 0.005 0.005 50:50 onneded 60:40 onneoled _. 0.000 0.000 • ,t5,,0,,I,,i,I, 1O0 150 200 250 300 ii , ,5/ O.... lOi0.... 15I0• ,._20l0, _a 250i .... 300I., Temperature (K) Temperature (K) b) d) Figure 4. The loss tangent as a function of temperature for all four samples, measured using an interdigital electrode at 1 MHz. a) Sample l" BSTO 50:50 as deposited, b) Sample 2: BSTO 50:50 annealed, c) Sample 3: BSTO 60:40 as deposited, d) Sample 4: BSTO 60:40 annealed. There are several observations that can be made from these low frequency data. While the films certainly have varying dielectric properties before annealing, it appears that annealing increases the maximum dielectric constant of these films by more than a factor of 2. The tan3 is also increased, although only by 20% in the case of the 60:40 samples. The dielectric constant is quite high with a maximum in Sample 2 of 3850. Even higher than one would expect from an undoped BSTO sample. This result echoes that of Wu and Barnes p]who found that bulk BSTO with 1% Mn doping had the largest value of e, amongst samples with 0, 0.5, 1.0, 2.5 and 5% doping. Another feature to note is that the temperature of maximum dielectric constant for the annealed samples of 50:50 (60:40) BSTO compositions, 183 K (230 K), is well below that of bulk Tc = 250 K (284 K). pl Figure 5 shows the results of insertion loss and phase shift measurements on the four-element coupled microstrip phase shifter circuits patterned on these same films. The results shown here are given at fop, the frequency of maximum tuning to insertion loss ratio, K, for each of these films. The tabulated maximum 0.0 120 & Mognilude 0,0_ 120,-, • Phase 100 -0.5 I / t "0 50:50asdeposited v lg.BCHz 80 rn _, t I=u.z .j -tBO ; -1.0 o t / 4_o •i::l-1.5 I _ .............. _,°_ [n ___o/L ----"'_t__ ,: li t,1 -2.0 . . 7 -2.5 -2.5_0 100 200 300 400 0 100 200 300 400 Bias VolLage (V) Bias Voltage (V) a) c) O.D ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' 120 0.0 .... i .... _ .... i .... 120 A klognitude • Phase 100' 60:40annealed • Pho=$ 100"_ ,_1-0.5 '13 '-0.5 1B,5GHz '0 v 50:50 annealed v v 16 GHz 80 [o [o-1.o {D-10 ,.l,lalA.i_, ali.,iil .._lll'- o -_Y'_,-H.EA_t,.. 60 ¢, "o-1.5 .__ i /_ -4:o = t@_._ 40 li m _2.5'_-_',..i .... I .... t .... 0 -2.5 1O0 200 300 400 0 1O0 200 300 400 Bias Voltage (V) Bias Voltage (V) b) d) Figure 5. The phase shift and insertion loss through coupled microstrip phase shifters on each of the four samples at 298 K. Data is shown at fopt for each sample, a) Sample l b) Sample 2 c) Sample 3 d) Sample 4. phase shift and insertion loss for a 400 V dc biasing range are given in Table I. This frequency, fop,,varies between these films because e, (E) differs for each film. The value of e, (0) in Ku-band frequency range determines the passband of the unbiased circuit. As the bias is increased and gr (E) drops, the device passband shifts to a higher frequency. Thus, er (0) and er (Em_,) determine fop,. The values of fopt show some correlation with the measured e_(0) at 1 MHz, with samples 1-3 following this trend of lower er leading to higher fop,. The fifth column lists fopt, and the last column gives the maximum value of K. The tuning and loss at 16 GHz are also given so that one may compare films at the same frequency. Table I. Four element CMPS measurements on 508 _m MgO substrates at room temperature. All values are given for tuning over a 400 V dc bias range. Ba:Sr Anneal er(0) at fopt Tuning Max. Loss Tuning at Max. loss Max K ratio atll00 IMHz GHz atfopt (dB) atfopt 16GHz w/ (dB) at °/dB Cfor6 and w/400 over 400 V 400 V dc 16GHz w/400 Vdc hrs. 300 K Vdc range ! 50:50 none 506 19.6 57° -! .4 93° -4.9 40.7 2 50:50 yes 946 16 75° -!.95 75° -1.95 38.5 T 60:40 none 1116 15_i i14 ° -2. I 97° -2.75 54.3 4 60:40 yes 132(J 18 80° -I.37 97° -1.8 58.4 These two circuits are not ideal for comparisons between high and low frequency for several reasons. First, the electric fields in the capacitor measurements are limited to 2.33 V/_tm by the 35 V dc maximum of the HP4291A. The CMPS devices act as dc blocks and are biased to 40 Vl_tm. Further tuning with higher voltages is possible and usually leads to higher values of K. None of these films broke down at field strengths of 40 V/_tm. Second, the CMPS circuits are complicated and difficult to model causing considerable uncertainty in backed out values of e, and tan& However, the maximum phase shift of 114 ° seen in sample 3 roughly agrees with a change in e, of 700 (e.g. from 1000 to 300) while the phase shift in Sample 1 at 16 GHz can be modeled by shifting CFfrom 500 to 150. These values are illustrated in Table III which lists IE3D [81em simulator modeled phase shifts assuming an unbiased e_(0) of 1200. Note that phase shift of S_, Atp(e_) is non-linear, this structure is a more efficient phase shifter at lower e_. TableII. IE3Dmodeledphaseshiftthrough this four element CMPS circuit at 16 GHz assuming er(0) =1200 for the BSTO film at 16 GHz. Modeling assumes that the entire BSTO layer changes er as a function of bias. cr 1200 1000 800 600 500 400 300 200 150 100 Phase 0° 17.2° 40.1° 70.7° 88.8° 110.5° 134.5° 163.2° 179.8° 196.8° ShiR Furthermore, note that the CMPS losses listed in Table I are not solely due to the BSTO film. Mismatch losses vary from 0.15 dB to 0.45 dB for these devices. Conductor loss is estimated to be about 0.5 dB. The dielectric loss in the MgO should account for 0.1 to 0.2 dB of insertion loss. Radiation loss is presumed to be small with the BSTO film present. The remaining loss is due to the BSTO film. For Sample 4 with the lowest loss listed at fort, -1.37 dB, the BSTO film would then account for about 0.62 dB or 45%. For Sample 3, the most tunable film at fort, with losses of-2.1 dB, the ferroelectric could account for about 64% of the loss. Figure 6 shows the temperature dependence of the CMPS phase shift and insertion Ioss as a function of temperature at 15 GHz and while applying 350 V dc for the two 60:40 samples. The as-deposited sample shows tuning increases with tuning that are roughly analogous to the changes in er(0) at 1 MHz. Both graphs indicate about 10% increases while cooling from 300 K to 250 K. The 15 GHz phase shift graph does peak at a lower temperature, however. The tuning in the annealed sample does not show the large increases of er(0) at 1 MHz. The change in er(0) from 300 K to 220 K was 168%, while the phase shift at 15 GHz increases only about 34%. Unfortunately non-linearity in Aq_(er) makes a .... t .... I.... I''''l''''t .... t'' 10 , 10 ++r+), ,,,i,,,,i,,,,i,,, ,i+, ,, I, •" Phose Shift o! S_ at 15_z x Moximum nsert_ Lossot t5 GHz X "_150 150 x 8 I_1 60:4,0 onneoted 8 m v v 100 6_,oJ L.O100 \% 8 wmO X o 40 m x _ 50 x...+.. 50 @ 60:$0 oS deposi|ed 1( 2_ pe]_ 2_ ,_ _,P_ose Shift of 52 ot 15 GHz xxxx x x Uoxmurn Insedion Loss ot 15GHz .... I .... ),,,,I,,,,It=,,tll,_ll 0 50 100 150 200 250 300 0 50 100 150 200 250 300 Bias VoRage (V) Bias Voltage (V) a) b) Figure 6. The phase shift of S21 (A) and maximum insertion loss (x) vs. temperature of a) Sample 3 b) Sample 4. This data was taken at 15GHz and using a 360 V dc bias. conclusion about the value of at e_(0) at 15-18 GHz difficult to draw. These phase shifter results are consistent with modeling where er tunes from 1200 to 550 at room temperature, and from 3200 to 800 at 250 K CONCLUSIONS Four BSTO films of composition 50:50 and 60:40 Ba:Sr ratios doped with 1% Mn were measured at 1 MHz with interdigital capacitor circuits and at 15 to 20 GHz using coupled microstrip phase shifters. Two films were post-annealed at 1100 C. The capacitor measurements as a function of temperature showed that the BSTO 50:50 films had a Tc about 183 Ks while the 60:40 films had a Tc of 230 K. The annealing was found to greatly increase erof these films. The largest value of e, = 3850 was found in the annealed BSTO 50:50 film. Tan8 at 1 MHz were found to be below 0.004 for most of these films. The high frequency phase shifter measurements found room temperature figures of merit, K, up to 58.4°/dB at 18 GHz and 400 V dc bias. The corresponding phase shift per length for a single phase shifter was 419°/cm.