2024 Volume 33 Issue 11
Article Contents

Lei Du, Hao-Ran Tao, Liang-Liang Guo, Hai-Feng Zhang, Yong Chen, Xin Tian, Chi Zhang, Zhi-Long Jia, Peng Duan, Guo-Ping Guo. In situ non-destructive measurement of Josephson junction resistance using fritting contact technique[J]. Chinese Physics B, 2024, 33(11): 110309. doi: 10.1088/1674-1056/ad73b6
Citation: Lei Du, Hao-Ran Tao, Liang-Liang Guo, Hai-Feng Zhang, Yong Chen, Xin Tian, Chi Zhang, Zhi-Long Jia, Peng Duan, Guo-Ping Guo. In situ non-destructive measurement of Josephson junction resistance using fritting contact technique[J]. Chinese Physics B, 2024, 33(11): 110309. doi: 10.1088/1674-1056/ad73b6

In situ non-destructive measurement of Josephson junction resistance using fritting contact technique

  • Received Date: 22/07/2024
    Accepted Date: 26/08/2024
    Available Online: 01/10/2024
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Figures(4)

Article Metrics

Article views(747) PDF downloads(4) Cited by(0)

Access History

In situ non-destructive measurement of Josephson junction resistance using fritting contact technique

Abstract: Conventional four-probe methods for measuring the resistance of Josephson junctions can damage superconducting thin films, making them unsuitable for frequency measurements of superconducting qubits. In this study, we present a custom probe station measurement system that employs the fritting contact technique to achieve in situ, non-destructive measurements of Josephson junction resistance. Our experimental results demonstrate that this method allows for accurate prediction of qubit frequency with an error margin of 17.2 MHz. Moreover, the fritting contact technique does not significantly affect qubit coherence time or the integrity of the superconducting film, confirming its non-destructive nature. This innovative approach provides a dependable foundation for frequency tuning and addressing frequency collision issues, thus supporting the advancement and practical deployment of superconducting quantum computing.

1.   Introduction
  • Superconducting quantum computing has emerged as a promising approach for building scalable and fault-tolerant quantum computers.[1] The key component in superconducting qubits is the Josephson junction, which allows for the creation of quantum superposition and entanglement states essential for quantum computation.[2] However, as the number of qubits in a system increases, several challenges arise, including three-dimensional circuit integration,[3,4] readout fidelity,[5,6] and coherence time.[7,8] Among these challenges, frequency collisions stand out as a primary obstacle, directly impacting the selective control of qubits and the fidelity of two-qubit gates, particularly in fixed-frequency transmon structures.[9–12]

    Frequency collisions occur due to fabrication errors in the chip manufacturing process, causing the frequencies of qubits to deviate from their design values.[13,15] These deviations lead to severe frequency collisions, necessitating measurement and assessment under cryogenic environments, which significantly consumes cryogenic resources and extends the testing cycle. Recently, post-processing techniques such as thermal annealing and laser annealing have shown excellent capabilities in tuning qubit frequencies,[16–18] making the prediction of qubit frequencies at room temperature essential for implementing these techniques.

    In transmon qubits, the qubit frequency (fq) is determined by the Josephson junction energy (EJ) and the charging energy (EC), satisfying the relation fq∝8EJEC.[19] The capacitance (Cq) of the qubit, typically taking the form of a coplanar waveguide capacitor, can be precisely fabricated. However, the critical current (Ic) of the Josephson junction, determined by its small area and barrier thickness, is susceptible to fabrication errors, which significantly contribute to qubit frequency deviations from the design value.[20–22] The relationship between Ic and the Josephson junction resistance (Rn) is given by the Ambegaokar–Baratoff relation Ic=πΔ2eRn, where Δ is the superconducting energy gap. Therefore, by measuring Rn at room temperature, we can estimate fq.

    Conventional four-probe methods for measuring the resistance of Josephson junctions involve the probe scrubbing the film surface, mechanically penetrating the surface oxide layer and the film, and moving forward to establish a stable contact.[24] This method can leave micron-scale damage in the qubit capacitor area, potentially affecting the capacitor parameters and introducing impurities, which can become sources of decoherence.[18] Furthermore, commonly used superconducting films such as tantalum and niobium are relatively hard and not easily penetrated;[25,26] this leads to high contact resistance between the probe and the film, resulting in inaccurate measurements. Consequently, in situRn measurement using traditional methods may lead to measurement inaccuracies and potential damage to the superconducting film.

    To mitigate these issues, researchers often use non-in situ methods by preparing Josephson junction test structures at the edge of the superconducting chip to estimate fq.[27–30] This approach involves fabricating additional test junctions with identical parameters on the periphery or non-critical regions of the chip. However, this method has significant limitations:

    (i) Inability to detect random failures It cannot detect randomly failing qubits within the operational areas of the chip,[13] limiting the effectiveness of the measurements.

    (ii) Dependence on fabrication uniformity The accuracy of frequency prediction is highly dependent on the uniformity of the fabrication process.[14] Any variations can lead to significant discrepancies between the test junctions and the actual operational junctions.

    (iii) Localized nature of adjustments Techniques like laser annealing require precise knowledge of the actual resistance of the operational junctions to fine-tune the qubit frequencies effectively. The non-in situ nature of the test junction measurements does not provide this level of precision.

    In this study, we address these limitations by utilizing the fritting contact technique to achieve in situ, non-destructive measurements of Josephson junction resistance. Fritting contact involves electrically breaking down the metal oxide layer by applying a controlled voltage, resulting in minimal physical damage to both the electrode and the probe while maintaining stable electrical contact. Our experimental results demonstrate that this method allows for accurate prediction of qubit frequency with an error margin of 17.2 MHz. Furthermore, we show that fritting contact does not significantly affect qubit coherence time or the integrity of superconducting film, confirming the non-destructive nature of this method. This advancement provides a significant technological foundation for frequency tuning and addressing frequency collision issues in multi-qubit systems, thereby accelerating the scalability and application of superconducting quantum computing.

2.   Method
  • Fritting contact is a method used to achieve electrical contact by electrically breaking down the metal oxide layer. This method involves applying a low force to bring the probe into contact with the electrode and then applying a high voltage to the probe, inducing electron transition or ionization within the metal oxide layer to form a stable electrical contact. The advantage of this method lies in its minimal physical damage to both the electrode and the probe while maintaining the stability of the electrical contact.[31–33] To employ the fritting contact method, we built a custom probe station measurement system, as shown in Fig. 1(a). The chip is vacuum-adhered to a four-axis motorized stage. The probe, driven by a piezoelectric platform (P-611Z, PI), approaches the chip at a descent speed of 5 nm/s. The contact force between the probe and the chip is monitored in real-time using a micro-force sensor (μ-Force 20g, Chief SI) positioned between the stage and the probe, with a resolution of 5 μN. When the probe makes contact with the sample, the contact force starts to appear and increases to a preset value, at which point the stage stops moving, completing the contact process. This entire process is fully automated. The probe is made of tungsten, with a tip diameter of 200 nm. Imaging is performed using a CCD camera and a light source, with the microscope lens angled to ensure observation of the probe tip during the fritting contact process. The entire setup is placed on an active vibration isolation table to mitigate environmental noise, ensuring accurate contact force measurements.

    Figure 1(b) illustrates the circuit used during the fritting contact and Josephson junction resistance measurement. First, after the probe contacts the qubit capacitor, a 20 ms pulse voltage signal is applied between contact points A and B, fritting the oxide layer. The pulse signal reduces circuit heating and minimizes damage to the film. Once electrical contact is established, the current limiter of the SMU is activated to prevent chip damage. The contact resistance is then measured. Electrical contact at points C and D is established using the same method. Finally, the SMU switch is turned off, and the circuit is switched to Josephson junction measurement circuit. Two lock-in amplifiers are used as an AC signal source and voltmeter for four-terminal measurements. The AC signal in the circuit eliminates the effects of thermoelectric and contact potentials that give rise to low-frequency noise. To protect Josephson junction, a 1 MΩ resistor is connected in series in the circuit to limit the current. The chip unit structure studied is shown in Fig. 1(b). The substrate is high resistivity (> 10 kΩ⋅cm) intrinsic (100) plane silicon, and qubit is transmon structure formed by an aluminum coplanar waveguide capacitor and an Al/AlOx/Al junction in parallel. The aluminum film is prepared by electron beam evaporation, and the Josephson junction is fabricated using the Manhattan-style technique, with an area of ∼ 0.04 μm2. Qubits are driven and read out through the readout bus.

3.   Results and analysis
  • Fritting voltage and contact force are two critical parameters that determine the success rate of fritting contact and the extent of electrode damage.[34,35] To minimize damage to the superconducting film, we selected a contact force of 10 μN for testing, which is significantly lower than the mechanical breakdown contact force of 30 mN–100 mN. Figure 1(c) shows the variations in voltage and current during a typical fritting contact. As the applied voltage gradually increases, naturally oxidized aluminum is electrically broken down, which results in a transient current in the circuit. Subsequently, the current limiter activates, rapidly reducing the current to 10 μA. This method allows us to identify the threshold fritting voltage and control the heat at the contact point, preventing damage to the aluminum film. Figure 1(d) illustrates that the contact resistance decreases as the limiting current increases. This is due to thermal expansion at the contact point caused by the heat from the higher current, which increases the contact area and consequently decreases the contact resistance. At a limiting current of 100 mA, the contact resistance is minimized to within 1 Ω. To protect the Josephson junction barrier, we set the limiting current to 10 μA, and the higher contact resistance is eliminated by using the four-terminal measurement method.

    To determine the precision of predicting fq using the fritting contact method, we measured Rn of 24 qubits at room temperature and compared the results with the corresponding fq measured at low temperature. The qubits underwent Rn measurement, packaging, and installation into a dilution refrigerator within five hours to minimize the impact of aging and drift on Rn. Here fq is referred to as the bare frequency of qubit, which was first measured using the Ramsey fringe method and was then downshifted by an amount of dispersive shift to rule out the dressing effect arising from the coupling to readout resonator.[36] Figure 2 shows the relationship between fq and Rn, which fits well with a power-law relationship, as expected from transmon theory. The inset in Fig. 2 shows that the residuals between fq and the fitted curve follow a Gaussian distribution with a standard deviation of 17.2 MHz, indicating that we can predict fq with an error of 17.2 MHz using the power-law relationship and Rn measured by the fritting contact method. This level of precision is comparable to the state-of-the-art frequency prediction error of 14.5 MHz demonstrated by IBM on fixed-frequency qubit.[16] In our case, this error mainly comes from the uncertainty of qubit capacitance due to variations in the fabrication process, while fluctuation in the superconducting gap also contributes. We observed a slight deviation between the power-law fitting exponent (−0.48) and the theoretical value (−0.5), which can be attributed to the fact that Rn is measured at room temperature rather than the cryogenic value used in the theory.

    In addition to precise frequency prediction, maintaining high qubit performance is also crucial. We investigated the impact of fritting contact on qubit coherence and the aluminum film. Qubit coherence is a critical parameter for high-fidelity gate operations and is highly sensitive to dielectric loss in materials.[37,38] We analyzed the distribution of energy relaxation time (T1) and phase relaxation time (T2) for 24 qubits before and after Rn measurement, as shown in Fig. 3(a). In the quantile–quantile plot (left panel), T1 and T2 exhibit a good correspondence with the unit slope, indicating that qubit coherence has a similar distribution before and after Rn measurement. The box plots (right panel) provide a detailed comparison of the means, quantiles, and other distributions of T1 and T2. Before (after) resistance measurement, T1 is 42 ± 14 μs (43 ± 14 μs), and T2 is 24 ± 13 μs (22 ± 11 μs), demonstrating that fritting contact did not affect qubit coherence within the statistical error range. Figure 3(b) shows the morphology of the aluminum film when mechanical breakdown of the oxide layer is used to achieve electrical contact between the probe and the electrode. Such indentations may affect capacitance Cq and introduce impurities, which reduce coherence time. We compared the morphology at the contact point before and after fritting contact and found no indentations, as shown in Figs. 3(c) and 3(d). This is achieved by precisely controlling the breakdown voltage and contact force, while the active vibration isolation table effectively isolates environmental noise. Overall, these results indicate that fritting contact is non-destructive for superconducting qubits.

    Laser annealing for frequency tuning requires multiple Rn measurements to ensure tuning accuracy.[16] This demands that Rn measurement is robust. To study the impact of fritting contact on frequency tuning, we collected the parameters of 12 qubits before and after laser annealing. Figure 4(a) shows the variation of Rn with laser annealing, where each qubit undergoes multiple Rn measurements to reach its target value. Figure 4(b) presents the residual distribution between fq and the fitted curve (using the same statistical method as the inset of Fig. 2). The residual values before and after annealing align well with a unit slope, indicating that fritting contact is well suited for frequency tuning technologies such as thermal annealing and laser annealing.

4.   Conclusion
  • In conclusion, we have developed a novel measurement technique utilizing the fritting contact method to achieve in situ, non-destructive measurements of Josephson junction resistance. This technique overcomes the limitations of traditional four-probe methods, which can damage superconducting thin films and lead to inaccurate measurements. Our results show that the fritting contact method allows for precise prediction of qubit frequency with a margin of error of 17.2 MHz. Additionally, the technique does not significantly impact qubit coherence time or the integrity of the superconducting film, confirming its non-destructive nature.

    The successful implementation of this method provides a robust technological foundation for precise frequency tuning, essential for mitigating frequency collision issues in multi-qubit systems. This advancement is crucial for the continued development and scalability of superconducting quantum computing. Future work will focus on further refining this technique and exploring its application in various superconducting quantum computing architectures to enhance performance and reliability.

Figure (4)  Reference (38)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return