- Email: [email protected]
Nb Josephson junctions
Physica B 165&166 North-Holland
(1990)
77-78
EXCESS LOW-FREQUENCY FLUX NOISE IN dc SQUIDS INCORPORATING Nb/AI-OXIDE/Nb JOSEPHSON JUNCTIONS* Martin E. HUBER and Michael W. CROMAR National Institute of Standards and Technology, 325 Broadway, Boulder, CO 80303-3328 U. S. A. We have fabricated thin-film dc SQUIDS (Superconducting Quantum Interference Devices) incorporating Nb/Al-Oxide/Nb tunnel junctions. The spectral density of the voltage noise, Sv, of stripline SQUIDS is characterized between 1 Hz and 2000 Hz. In this frequency range, Sv is proportional to the square of the aV/acP)over a significant ran e of bias conditions with an unusual frequency dependence. In responsivi a 7 pH SQ?%I , the spectral density of the Blux noise, SQ, at 1 Hz is less than lo-11 @$/Hz, where @, E h/2e. The observed noise does not appear to be environmental; also, it is independent of the value of the junction shunt resistance and whether the stripline material is a Pbln alloy or Nb. Subject to the constraint of a constant product of the junction area, critical current density, and SQUID self-inductance in the SQUIDS studied, Sm is inversely proportional to the junction area.
INTRODUCTION The intrinsic noise level of dc SQUIDS (Superconducting Quantum Interference Devices) can, in theory, be very low (l), making them ideal sensors for electric currents and magnetic fields (2). In practice, excess noise of one form or another is often observed. We are studyin the intrinsic noise of dc SQUIDS (3) incorporating 9ow-leakage Nb/AIOxide/Nb Josephson tunnel junctions (4). One type of excess noise which has been observed before (56) is flux-like. If biased with a constant current, the voltage V across the dc SQUID varies with applted ma netic flux CDwith a smallsignal responsivity &I/ & , which is a function of a. According to models of dc SQUID dynamics (1): the spectral densit of the voltage noise Sv is not srmply related to (aVIK(o)2. Proporttonality between the two thus indicates the presence of a flux noise with spectral density S@= S&9/0@)*. There are usually no obvious sources of flux noise present, however, which is why the noise is called “flux-like”. We, too, observe a flux-like noise, for which So is independent of the operating point and has an unusual frequency dependence (neither frequency independent nor a power law function). We report on the dependence of‘the value of So on physicai oarameters such as the shunt resistance. SQUID geometry, and SQUID-loop materials. ’ 2. SQUID PREPARATION AND MEASUREMENT The SQUIDS are fabricated on Si in a thin-film process (3) incorporating Nb/Al-OxidelNb Josephson junctions. Resistors are Aulns and the insulation is SiO. The u per wiring layer is either a Pbln alloy or Nb. All SQ e IDS studied are of a low-inductance, self-shielded (stripline) geometry (3). The two junctions are located at opposite ends of a Ion narrow stripline (the SQUID “body”) and connect tRe upper
and lower conductors of the stripline. The current is injected midway between junctions in one layer and is collected from a symmetric point on the other layer. The nominal design parameters are listed in Table 1. The insulation thickness is 300 nm and the critical current density Jc is about 700 A/cm*. We measure Sv in the range 1 Hz to 2 kHz with the SQUID biased at a constant current. The SQUID is connected to a resistor in series with a transformer for current gain,; the secondary winding is connected to a commercral dc SQUID operating in a flux-locked mode. 3. RESULTS AND DISCUSSION To determine whether the noise was in fact fluxlike, SQUIDS were first biased at three different drive currents. At each drive current, Sv and aV/a were measured at various values of the applied flux. We observed proportionality between Sv and @V/a@)*, where the value of the slope is Sa, (Fig. 1). Since such noise could result from external sources, it was necessary to rule out that possibility before considering intrinsic sources. The standard SQUID mount is a circuit board with BeCu contacts, TABLE 1 Nominal parameters for the stripline SQUIDS tested. R is the shunt resistance for a single junction and L is the SQUID self-inductance. Junction A :
4.0 16.0 7.8
Junction Separation (u ) 5: 250 350
D E
32.5 16.0
175 500
ID
- Contribution of the U.S. Government, not subject to copyright in the United States. 1990 - Elsevier
Science
Publishers
B.V. (North-Holland)
SQUID Width (u 1 1: :;
R ( ) 5: 2:4
L (D 1 26: 12:4
30 40
A.6’ 1:2
i.1 6:8
78
M.E. Huber, M. W. Cromar
0.25
Nominal Junction Area (um2)
0.00
0.00
0.25
0.50
(%/&@2
0.75
1.00
(nV2/pQ0’)
FIGURE 1 Sv vs. @V/acP)2for a type C SQUID at various bias points and frequencies (a,, t h/2e). Although the data are clustered near the oriain. a linear refationship also exists on an expand& scale. sandwiched between solid Nb disks. The SQUID mount is enclosed linder of high permeability alloy specially desi at low temperatures, and the cylinder is foil. There are also three layers of room temperature high-permeability shieldin and the Dewar is non-magnetic metal. We reduce CBthe low-fre uency shielding in stages, including re lacing tB e mount with an unshielded circuit boar s incorporating In contacts. No increase in So was observed until the SQUID was operated with neither superconducting nor normal-metal magnetic shielding. SQUIDS with half the normal insulation thickness (and hence less area to couple
FIGURE 3 Sa vs. nominal junction area “a” for SQUID types A through E. The line is a fit to Sa = l/a. to external fields exhibited no reduction in S@(Fig. 2). In addition, $ QUJDs fabricated on sapphire rather than Si exhibited no variation in SQ. As a whole, these results implicate non-environmental sources as the origin of the excess noise. Figure 2 summarizes the results of all fabrication runs, Including those in which the junction shunt resistance and stripline materials were changed. Notice that although S@is process dependent below 100 Hz, it is not correlated to the features studied; variation is observed even in SQUIDS processed under identical conditions on different wafers (the “nominal” points). Above 100 Hz, SO is the same for all fabrication runs. The shunt resistances of the “low-R” SQUIDS were 0.4 times the values listed in Table 1; we conclude that Sa is not a function of the shunt resistance. Also, althou h Wellstood et al. (6) have observed less noise in S5 UfDs with Pb loops than those with Nb loops, there is no such dependence in these SQUIDS. Finall , at 107.5 and 1075 Hz, we average the value of &a for all SQUIDS of a given ty and plot Sa vs. the nominal junction area (Fig. 3 . Subject to the constraint a*Jc*L = constant for the r QUlDs studied, we observe So 0~ t/a (‘a” is the junction area). The specific parameter dependence of the noise is under investigation. REFERENCES
(1) C. D. Tesche and J. Clarke, J. Low Temp. Phys. 29 II 977) 301.
(2) T. Ryhan&H. (3)
1o2 10’ Frequency (Hz) FIGURE 2 .Sa vs. frequency for type C SQUIDS at selected frequencies for all fabrication runs. The bias current is about 0.8 times the maximum zero-voltage current.
(4) (5) (6)
Seppa, R. Ilmoniemi, and J. Knuutila, J. Low Temp. Ph s. 76 (1989) 287. M. W. Cromar, J. A. Beall, & . Go, K. A. Masarie, R. H. Ono. and R. W. Simon. IEEE Trans. Maa. 25 (1989)‘1065. H. A. Huggins and M. Gurvitch, J. Appl. Phys. 57 (1985) 2103.. R.-R Koch,-J. Clarke, W. M. Goubau, J. M. Ma&is, C. M. Pe rum, and D. J. Van Harlin en, J. Low gremp. Phys. 51 (1983) 207. F. C. 5: elistood, C. Urbina, and J. Clarke, IEEE Trans. Mag. 23 (1987) 1662.
(1990)
77-78
EXCESS LOW-FREQUENCY FLUX NOISE IN dc SQUIDS INCORPORATING Nb/AI-OXIDE/Nb JOSEPHSON JUNCTIONS* Martin E. HUBER and Michael W. CROMAR National Institute of Standards and Technology, 325 Broadway, Boulder, CO 80303-3328 U. S. A. We have fabricated thin-film dc SQUIDS (Superconducting Quantum Interference Devices) incorporating Nb/Al-Oxide/Nb tunnel junctions. The spectral density of the voltage noise, Sv, of stripline SQUIDS is characterized between 1 Hz and 2000 Hz. In this frequency range, Sv is proportional to the square of the aV/acP)over a significant ran e of bias conditions with an unusual frequency dependence. In responsivi a 7 pH SQ?%I , the spectral density of the Blux noise, SQ, at 1 Hz is less than lo-11 @$/Hz, where @, E h/2e. The observed noise does not appear to be environmental; also, it is independent of the value of the junction shunt resistance and whether the stripline material is a Pbln alloy or Nb. Subject to the constraint of a constant product of the junction area, critical current density, and SQUID self-inductance in the SQUIDS studied, Sm is inversely proportional to the junction area.
INTRODUCTION The intrinsic noise level of dc SQUIDS (Superconducting Quantum Interference Devices) can, in theory, be very low (l), making them ideal sensors for electric currents and magnetic fields (2). In practice, excess noise of one form or another is often observed. We are studyin the intrinsic noise of dc SQUIDS (3) incorporating 9ow-leakage Nb/AIOxide/Nb Josephson tunnel junctions (4). One type of excess noise which has been observed before (56) is flux-like. If biased with a constant current, the voltage V across the dc SQUID varies with applted ma netic flux CDwith a smallsignal responsivity &I/ & , which is a function of a. According to models of dc SQUID dynamics (1): the spectral densit of the voltage noise Sv is not srmply related to (aVIK(o)2. Proporttonality between the two thus indicates the presence of a flux noise with spectral density S@= S&9/0@)*. There are usually no obvious sources of flux noise present, however, which is why the noise is called “flux-like”. We, too, observe a flux-like noise, for which So is independent of the operating point and has an unusual frequency dependence (neither frequency independent nor a power law function). We report on the dependence of‘the value of So on physicai oarameters such as the shunt resistance. SQUID geometry, and SQUID-loop materials. ’ 2. SQUID PREPARATION AND MEASUREMENT The SQUIDS are fabricated on Si in a thin-film process (3) incorporating Nb/Al-OxidelNb Josephson junctions. Resistors are Aulns and the insulation is SiO. The u per wiring layer is either a Pbln alloy or Nb. All SQ e IDS studied are of a low-inductance, self-shielded (stripline) geometry (3). The two junctions are located at opposite ends of a Ion narrow stripline (the SQUID “body”) and connect tRe upper
and lower conductors of the stripline. The current is injected midway between junctions in one layer and is collected from a symmetric point on the other layer. The nominal design parameters are listed in Table 1. The insulation thickness is 300 nm and the critical current density Jc is about 700 A/cm*. We measure Sv in the range 1 Hz to 2 kHz with the SQUID biased at a constant current. The SQUID is connected to a resistor in series with a transformer for current gain,; the secondary winding is connected to a commercral dc SQUID operating in a flux-locked mode. 3. RESULTS AND DISCUSSION To determine whether the noise was in fact fluxlike, SQUIDS were first biased at three different drive currents. At each drive current, Sv and aV/a were measured at various values of the applied flux. We observed proportionality between Sv and @V/a@)*, where the value of the slope is Sa, (Fig. 1). Since such noise could result from external sources, it was necessary to rule out that possibility before considering intrinsic sources. The standard SQUID mount is a circuit board with BeCu contacts, TABLE 1 Nominal parameters for the stripline SQUIDS tested. R is the shunt resistance for a single junction and L is the SQUID self-inductance. Junction A :
4.0 16.0 7.8
Junction Separation (u ) 5: 250 350
D E
32.5 16.0
175 500
ID
- Contribution of the U.S. Government, not subject to copyright in the United States. 1990 - Elsevier
Science
Publishers
B.V. (North-Holland)
SQUID Width (u 1 1: :;
R ( ) 5: 2:4
L (D 1 26: 12:4
30 40
A.6’ 1:2
i.1 6:8
78
M.E. Huber, M. W. Cromar
0.25
Nominal Junction Area (um2)
0.00
0.00
0.25
0.50
(%/&@2
0.75
1.00
(nV2/pQ0’)
FIGURE 1 Sv vs. @V/acP)2for a type C SQUID at various bias points and frequencies (a,, t h/2e). Although the data are clustered near the oriain. a linear refationship also exists on an expand& scale. sandwiched between solid Nb disks. The SQUID mount is enclosed linder of high permeability alloy specially desi at low temperatures, and the cylinder is foil. There are also three layers of room temperature high-permeability shieldin and the Dewar is non-magnetic metal. We reduce CBthe low-fre uency shielding in stages, including re lacing tB e mount with an unshielded circuit boar s incorporating In contacts. No increase in So was observed until the SQUID was operated with neither superconducting nor normal-metal magnetic shielding. SQUIDS with half the normal insulation thickness (and hence less area to couple
FIGURE 3 Sa vs. nominal junction area “a” for SQUID types A through E. The line is a fit to Sa = l/a. to external fields exhibited no reduction in S@(Fig. 2). In addition, $ QUJDs fabricated on sapphire rather than Si exhibited no variation in SQ. As a whole, these results implicate non-environmental sources as the origin of the excess noise. Figure 2 summarizes the results of all fabrication runs, Including those in which the junction shunt resistance and stripline materials were changed. Notice that although S@is process dependent below 100 Hz, it is not correlated to the features studied; variation is observed even in SQUIDS processed under identical conditions on different wafers (the “nominal” points). Above 100 Hz, SO is the same for all fabrication runs. The shunt resistances of the “low-R” SQUIDS were 0.4 times the values listed in Table 1; we conclude that Sa is not a function of the shunt resistance. Also, althou h Wellstood et al. (6) have observed less noise in S5 UfDs with Pb loops than those with Nb loops, there is no such dependence in these SQUIDS. Finall , at 107.5 and 1075 Hz, we average the value of &a for all SQUIDS of a given ty and plot Sa vs. the nominal junction area (Fig. 3 . Subject to the constraint a*Jc*L = constant for the r QUlDs studied, we observe So 0~ t/a (‘a” is the junction area). The specific parameter dependence of the noise is under investigation. REFERENCES
(1) C. D. Tesche and J. Clarke, J. Low Temp. Phys. 29 II 977) 301.
(2) T. Ryhan&H. (3)
1o2 10’ Frequency (Hz) FIGURE 2 .Sa vs. frequency for type C SQUIDS at selected frequencies for all fabrication runs. The bias current is about 0.8 times the maximum zero-voltage current.
(4) (5) (6)
Seppa, R. Ilmoniemi, and J. Knuutila, J. Low Temp. Ph s. 76 (1989) 287. M. W. Cromar, J. A. Beall, & . Go, K. A. Masarie, R. H. Ono. and R. W. Simon. IEEE Trans. Maa. 25 (1989)‘1065. H. A. Huggins and M. Gurvitch, J. Appl. Phys. 57 (1985) 2103.. R.-R Koch,-J. Clarke, W. M. Goubau, J. M. Ma&is, C. M. Pe rum, and D. J. Van Harlin en, J. Low gremp. Phys. 51 (1983) 207. F. C. 5: elistood, C. Urbina, and J. Clarke, IEEE Trans. Mag. 23 (1987) 1662.