YBa₂Cu₃O_7−δ superconducting quantum interference devices with metallic to insulating barriers written with a focused helium ion beam

Recent advances in focused ion beam Josephson junctions¹ made from high temperature superconductors (HTS) have renewed interest in using HTS for a broad range of applications in superconducting electronics. Antennas^2,3 and other circuits for communications^4–6 would benefit greatly from higher operating temperatures. Whereas the unique properties such as small capacitance, and large energy gap may have uses in superconducting computing,^7,8 THz devices,⁹ and nano superconducting quantum interference devices (SQUIDs).¹⁰ In the last decade, a great deal of progress has been made on HTS SQUIDs; today, the sensitivities and noise properties are approaching those of low transition temperature niobium SQUIDs.¹¹ Historically, high quality HTS SQUIDs are very difficult to reproducibly fabricate. They are hand processed one at a time, individually tested with only the top few percent selected. This has kept costs high and prohibitive for most applications, especially those requiring multiple sensors like magnetoenceptahalography (MEG).

One promising approach for manufacturing large numbers of SQUIDS is to use Josephson junctions with barriers created using ion irradiation. The key to this technique is that HTS materials like YBa₂Cu₃O_7−δ (YBCO) are very sensitive to point defects in the crystal lattice caused by ion irradiation. Increasing irradiation levels has the effect of reducing the superconducting transition temperature (T_C) and increasing resistivity.¹² For high doses of irradiation, YBCO becomes insulating and no longer conducts or superconducts. Using ion implantation through a mask fabricated with electron beam lithography, we have demonstrated large numbers of uniform devices on 100 mm wafers.^13,14 Unfortunately, these devices are proximity effect junctions with large excess currents and low critical voltages $VC=ICRN$⁠, where I_C is the maximum Josephson supercurrent and R_N is the normal state resistance. The occurrence of these undesirable properties can be understood in the framework of a model put forth by Blonder et al. (BTK).¹⁵ In this work, they describe a crossover from metallic to insulating behavior at a normal-superconductor interface using a parameter related to the strength of the barrier. For metal-superconductor interfaces (weak barrier), there is a large transport component (excess current) due to Andreev reflections. For insulator-superconductor interfaces (strong barrier), this mechanism is suppressed. In the case of proximity effect junctions, the physical length of the normal region is too long to allow for a strong barrier resulting in non-Josephson excess current.

An additional problem with masked ion damage junctions is that there is a large variance in properties from wafer to wafer because the superconducting transport properties depend exponentially on the length of the barrier. Slight variations of even 1 nm in the critical dimension of the implant mask can dramatically change $ICRN$⁠. Eliminating the mask and directly writing devices with a scanned beam substantially improves reproducibility and quality.

The idea of directly writing Josephson devices into the plane of YBCO films was conceived of many years ago. Zani et al. directly wrote Josephson weak links into a (YBCO) film using a focused silicon ion beam¹⁶ and others used high energy electron beams.^17,18 Unfortunately, like the masked ion damage junctions, these earlier direct-write devices were also proximity effect junctions because the tools at the time were not capable of creating a narrow (∼1 nm) and strong barrier.

In this letter, we report fabrication of HTS SQUIDs with a very finely focused 500-pm diameter helium ion beam from a gas field ion source.¹⁹ Narrow insulating barriers can be written with this technique because the beam size is of the order of the tunneling length, and the electrical transport properties of the Josephson barriers can be continuously tuned from metallic to insulating by simply increasing the ion fluence.¹ 

DC washer SQUIDs were patterned with conventional photolithography and ion milling from single layer 120 nm thick YBCO films grown by reactive coevaporation on cerium oxide buffered sapphire with a sputtered gold contact. The design consisted of a 1 mm × 1 mm square washer with a YBCO multi-turn planar input coil Fig. 1(a). The gold contact was removed in the area intended for junctions [Fig. 1(b)] and thickness of the YBCO film in this region was ion milled down to ∼30 nm. This was to ensure that the 30 kV helium ion beam would completely penetrate the YBCO and create a uniform barrier. The helium beam was scanned across of the arms of the SQUID to form 4 μm long junctions with very narrow barriers ∼1–2 nm. Several devices were fabricated using helium doses ranging between 2 × 10¹⁶ and 9 × 10¹⁶ ions/cm². These doses are much smaller than those typically used to remove YBCO (∼10¹⁸ ions/cm²). SQUIDs written with doses less than 4 × 10¹⁶ ions/cm² had superconductor-normal metal-superconductor (SNS) junctions and at higher doses superconductor-insulator-superconductor (SIS) junctions. The SIS and SNS SQUIDs presented in this work were fabricated with doses 9 × 10¹⁶ and 4 × 10¹⁶ ions/cm², respectively.

Figure 2 compares electrical transport measurements for typical SIS and SNS SQUIDs at 4 and 50 K, respectively. The current-voltage characteristics (I–V) of both the SIS [Fig. 2(a)] and SNS [Fig. 2(b)] SQUIDs exhibit resistively shunted junction characteristics well-described by the Stewart-McCumber model^20,21 for small voltages. The $I0R$ products, where I₀ is the critical current of the SQUID, for the SIS and SNS SQUIDs are 270 and 60 Μv, respectively. The Stewart-McCumber parameter $βC≡2πI0R2C/Φ0=0.032≪1$ for the SIS SQUID, which explains why there is not hysteresis in the I–V characteristics. This comes about from the very small capacitance (⁠$C≃3$ fF) due to the small electrodes. The SIS SQUID exhibits a rise in I–V ∼ 250 μV that is independent of temperature, which rules out the possibility that it is a secondary critical current. It may possibly be due to a standing wave resonance in the input coil or the SQUID bias leads.

The SQUIDs were DC biased above the critical currents and the voltages were measured as functions of magnetic field [Figs. 2(c) and 2(d)]. Both the SIS and SNS devices exhibited a well-behaved voltage modulation of approximately 3/4 and 1/2 $I0R$⁠, respectively. Measurements were taken for a larger magnetic field range to observe the Fraunhofer envelopes of the junctions. The first minima occur near 25 μT for both devices [Figs. 2(e) and 2(f)]. The shape is in a good agreement with that predicted for planar geometry junctions.^22,23 However, the second minima occur at 50 and 70 μT for the SIS and SNS SQUIDs, respectively. This discrepancy is not understood and merits additional study.

I–V for the SIS SQUID (Fig. 3) measured for a much higher voltage range shows non-linear insulator-like characteristics. Using a lock-in amplifier, we differentiate the I–V curve and observe a gap-like feature at ∼20 mV, similar to our previous results¹ on SIS YBCO junctions (33 mV) and others reported in the literature.^24,25 We attribute the different value to the different material used in this work.

To characterize the noise, we connected the SNS SQUID to a Tristan Technologies iMAG^® 300 series flux-locked-loop and measured the output on a signal analyzer. The results are shown in Fig. 4 with and without bias reversal to reduce the critical current noise.²⁶ The $1/f$ knee occurs around 1 kHz and the white noise level is $∼ 2 μΦ0/Hz−1/2$ [Fig. 4(a)]. For low frequencies, the noise of the SQUID is $20 μΦ0/Hz−1/2$ at 10 Hz, which corresponds to a field noise of 20 pTHz^−1∕2. This value can be substantially improved by using a larger sized washer or incorporating the SQUID into a multi-turn flux transformer.

In conclusion, we report a fabrication method for high quality HTS SQUIDs that is scalable and can be used to create large numbers of SQUIDs on a single wafer. This method may allow for writing Josephson junctions directly into a flux transformer eliminating the need for using flip-chips. Furthermore, the large resistance of these SQUIDs could eliminate impedance matching transformers and simplify SQUID read-out electronics.

This method of directly patterning the SQUID should not only increase HTS sensor yields but also reduce inter-sensor variability in terms of noise, phase delay, and critical current. With these improvements, we believe that the financial barrier to large channel count HTS arrays is removed, mainly due to the large reduction in sensor fabrication labor. Decreasing the cost of the sensors makes high channel count systems more economical than conventional niobium SQUID systems because the dewars do not require an intricate thermal shield arising from the 50-fold increase in latent heat of vaporization for liquid nitrogen vs liquid helium. Also because of the simplified cryogenics, the sensors can be mounted much closer to room temperature, improving signal-to-noise.

This work was supported by AFOSR FA9550-07-1-0493 and M. K. Ma was supported by UC Scholars. The authors thank D. Rosenstock and T. J. Wong for helping with experimental setup.