Superconducting nanowire-based devices are increasingly being used in complex circuits for applications such as photon detection and amplification. To keep up with the growing circuit complexity, nanowire processing is moving from single layer fabrication to heterogeneous multilayer processes. Hydrogen silsesquioxane (HSQ) is the most common choice of negative-tone electron-beam resist for patterning superconducting nanowires. However, HSQ has several limitations, including an inability to be removed without a strong reagent that damages the superconducting film, making it unsuitable for multilayer fabrication. As a result, it is vital to consider alternative resists that can be removed through less harmful solvents. Here, the authors explore the use of ma-N 2400 series deep ultraviolet photoresist as an electron-beam resist for fabricating superconducting nanowire devices. They demonstrate that ma-N can be used to pattern dense lines as narrow as 30 nm and isolated features below 20 nm in width. They also examine the reproducibility of 36 identical superconducting devices by comparing their minimum dimensions and switching currents. Through this analysis, they conclude that ma-N 2400 is a suitable electron-beam resist for fabricating nanoscale devices and has the potential to expand the use of nanowire-based technologies into more advanced applications.

Advancements in the nanoscale patterning of materials using electron-beam lithography (EBL) have facilitated the development of a new family of superconducting devices based on thin-film nanowires. These devices are used in a wide variety of applications, including photon detection,1,2 digital circuitry,3,4 and memory,5 and are becoming increasingly important with the rising need for low-power technologies. Traditionally, superconducting nanowires are patterned monolithically, requiring a single EBL step. Since the functionality of these nanowire devices depends strongly on the exact nanoscale dimensions of their features, proper choice of EBL resist is vital to ensure precise control of their geometries. Hydrogen silsesquioxane (HSQ) was first chosen as a negative-tone EBL resist for superconducting nanowire processes6 since it offers advantages such as low line-edge roughness and ∼5 nm resolution.7 

Despite these benefits, there are significant difficulties in using HSQ with superconducting materials. First, exposed HSQ cannot be easily removed without the use of a strong reagent such as hydrofluoric acid, which damages the quality of the superconducting film and, thus, limits device performance. This is particularly hindering as nanowires are gaining popularity in more complex circuits that involve integration with other technologies, such as magnetic memory elements.8 Such complex circuit environments often demand multilayer processes, which require exposed resist to be removed between layers. Since HSQ cannot be removed without damaging the underlying device, it is clearly unsuitable for these applications. Additionally, HSQ has a high areal dose density (∼3500–4000 μC/cm2 in a 125 kV EBL system) in comparison to many positive-tone resists, which prolongs the write time for complex circuits with many components. Finally, HSQ often has difficulty adhering to superconducting materials such as niobium nitride (NbN), requiring a pretreatment step that can damage the superconducting film.9 As a result, there is a need for investigating alternative negative-tone electron-beam resists to support the advancement of superconducting devices toward more complex applications.

Here, we report on the use of ma-N 2400 series deep ultraviolet photoresist (©microresist technology) as an electron-beam resist for patterning superconducting nanoscale devices. ma-N 2400 series resist is primarily composed of a polymeric bonding agent (phenolic resin) and a bisazide photoactive compound.10 In contrast to HSQ, exposed ma-N can be removed with solvents like N-methyl-2-pyrrolidone (NMP), which are generally harmless to superconducting films like NbN and, thus, allow for multilayer processing. While this work focuses only on single layer processing, the results may be extended to multilayer fabrication based on the ability to remove ma-N without inflicting damage. Additionally, ma-N’s dose is roughly four times lower than that of HSQ in 125 kV EBL systems, which could significantly reduce the write time for large patterns. While previous studies have reported minimum feature sizes of 50 nm or more,10,11 we are able to pattern repeated lines of widths down to 30 nm and individual features with minimum dimensions less than 20 nm. To demonstrate ma-N’s suitability for patterning superconducting devices, we compare the performance and geometry of 36 identical devices and conclude that the reproducibility is satisfactory. These efforts also enrich our understanding of superconducting nanowire uniformity, in general, since there are very few systematic studies on lithographic variations in nanowire patterning.

Here, we describe the ma-N process used to pattern the devices and structures, which aligns closely with ma-N recipes in previous literature.11,12 All steps particular to fabricating superconducting nanowires, including reactive ion etching and measurement, follow our standard methods.

All chips written using ma-N 2401 (the most dilute of the ma-N 2400 series) followed the same EBL procedure. Prior to exposure, ma-N 2401 was spun at 3 krpm for 60 s, yielding a thickness of about 92 nm, and then baked at 90 °C for 60 s. Afterwards, patterns were written using a 125 kV Elionix system (ELS-F125) with a 500 pA current, corresponding to a beam spot size of 2 nm. Doses ranged from 800 to 1000 μC/cm2, depending on the pattern. Following exposure, chips were developed in Microposit MF CD-26 at room temperature for 10 s with gentle agitation by hand and immediately rinsed in de-ionized water and blown dry using a nitrogen gun. The fidelity of the write was assessed by examining the patterns using a scanning electron microscope (Zeiss Sigma HD). Some features written only in ma-N 2401 without an underlying superconducting film were sputtered with gold-palladium prior to imaging to enhance the contrast. For patterns that were written on superconducting films, pattern transfer to the underlying film was accomplished using reactive ion etching in CF4 (Plasmatherm RIE, 50 W rf power, 10 mTorr chamber pressure). The resist could be stripped by submerging the chip in heated NMP at 60 °C for 1 h, offering a significant advantage over HSQ, which can only be removed with a strong reagent like hydrofluoric acid. We did not observe any adverse effects of prolonged immersion in heated NMP on the quality of the superconducting film.

An AJA sputtering system was used to deposit an NbN film13 on a 1 × 1 cm2 die cut from a 4 in. Si wafer with thermal oxide. The sheet resistance of the film was 150 Ω/sq. Prior to patterning the superconducting devices, 50-nm-thick gold pads for making electrical contact were written using direct write photolithography (Heidelberg μPG 101) with a bilayer photoresist process followed by lift-off. After patterning the contact pads, nanowire structures were written with ma-N 2401 using the EBL methods described above. The devices were then imaged using a scanning electron microscope (Zeiss Sigma HD). The resulting micrographs were processed using prosem software (GenISys) to determine the minimum dimension for each device.

After imaging, the current-voltage characteristics of the devices were measured in a cryogenic probe station. The samples were cooled down to 4.89 K. A bias current was applied using a DC battery source (Stanford Research Systems SIM928) in series with a 100 kΩ resistor. The bias current was ramped from 0 to 30 μA, and the voltage was read out through a Keithley multimeter. The current-voltage characteristics for each device were obtained, and the point at which the voltage jumped from roughly zero to an appreciable value on the order of millivolts was recorded as the device’s switching current.

We first present the resolution we obtained using ma-N 2401 and characterize its etch resistance in a fluorine chemistry relative to that of HSQ. Afterwards, we discuss the reproducibility of superconducting devices that were patterned using ma-N resist by comparing their minimum dimensions and electronic characteristics.

Using the methods described above, we patterned a variety of geometries on a 1 × 1 cm2 Si die with 300-nm-thick thermal oxide. Figure 1 displays several of these features. As shown in Fig. 1(a), it was possible to resolve 30-nm lines with a 120-nm pitch, a smaller resolution than what has been previously claimed.10,11 Besides repeated lines, we also patterned isolated features. Figure 1(c) shows an example of one such feature, a T-shaped pattern with the same geometry as a common three-terminal superconducting nanowire device, the nanocryotron (nTron).3 As indicated on the figure, the minimum feature size of this pattern was ∼18 nm, measured using image analysis software (prosem). This result demonstrates that the resolution of ma-N resist is superior to what has been previously thought and that it is suitable for patterning isolated features narrower than 20 nm.

Fig. 1.

Resolution of ma-N 2401 patterned using 125 kV Elionix. (a) 30-nm lines with a 120-nm pitch (dose = 1000 μC/cm2). (b) 60-nm lines with a 120-nm pitch (dose = 1000 μC/cm2). (c) Three-terminal pattern with a minimum isolated feature size of ∼18 nm (dose = 900 μC/cm2). The granularity is due to the gold-palladium sputtering that was used to enhance imaging contrast on the dielectric substrate.

Fig. 1.

Resolution of ma-N 2401 patterned using 125 kV Elionix. (a) 30-nm lines with a 120-nm pitch (dose = 1000 μC/cm2). (b) 60-nm lines with a 120-nm pitch (dose = 1000 μC/cm2). (c) Three-terminal pattern with a minimum isolated feature size of ∼18 nm (dose = 900 μC/cm2). The granularity is due to the gold-palladium sputtering that was used to enhance imaging contrast on the dielectric substrate.

Close modal

To ensure that the use of ma-N 2401 would not limit our ability to transfer patterns onto a superconducting film, we compared its etch resistance to that of HSQ, our traditional negative-tone resist. We first spun 4% HSQ and ma-N onto two identical Si chips, both at 3 krpm for 60 s. After spinning, the ma-N chip was baked according to the procedure described in Sec. II, and the initial thickness of each resist was approximated using multiwavelength ellipsometer (FilmSense FS-1) with a Cauchy model. The HSQ had an initial thickness of about 63 nm, and the ma-N had an initial thickness of about 92 nm.

To compare the etch resistances, we reactively ion etched both chips together in CF4 in 1 min intervals and measured their thicknesses using the ellipsometer after each step. The process was repeated until the resist with the highest etch resistance was reduced to 50% of its initial thickness. The mean squared error, or fit difference, between the measured ma-N ellipsometry data and the model ranged from 0.0033 to 0.0110 over six thickness measurements across the etching process. The mean squared error for the chip coated in HSQ ranged from 0.0018 to 0.0035. These error values are within the range of those reported in previous literature that validated the ellipsometer as an appropriate tool for measuring the thicknesses of thin films,14 suggesting that our thickness measurements are reasonably accurate.

Figure 2 compares the performances of the two resists. Overall, the 4% HSQ had an etch rate of roughly 14.5 nm/min, while the ma-N had an etch rate of 8.9 nm/min. It is important to note that both of these resists were unexposed, so it is possible that their etch resistances would be slightly higher following exposure.15 However, the results shown here indicate that ma-N’s etch resistance is comparable to, if not better than, that of HSQ for a fluorine-based etch; consequently, we can conclude that choosing ma-N over HSQ to pattern superconducting devices is not sacrificing the fidelity of the pattern transfer.

Fig. 2.

Etch rate in CF4 of ma-N 2401 and 4% HSQ (both unexposed). The etch rate of ma-N 2401 was 8.9 nm/min, and the etch rate of 4% HSQ was 14.5 nm/min. Each resist was spun at 3 krpm for 60 s. The substrate with ma-N was baked at 90 °C for 1 min prior to etching. Comparison of the etch rates indicates that the etch performance of ma-N is comparable to that of HSQ.

Fig. 2.

Etch rate in CF4 of ma-N 2401 and 4% HSQ (both unexposed). The etch rate of ma-N 2401 was 8.9 nm/min, and the etch rate of 4% HSQ was 14.5 nm/min. Each resist was spun at 3 krpm for 60 s. The substrate with ma-N was baked at 90 °C for 1 min prior to etching. Comparison of the etch rates indicates that the etch performance of ma-N is comparable to that of HSQ.

Close modal

To evaluate ma-N’s potential as an electron-beam resist for superconducting devices, we used ma-N to pattern 36 identical nTrons on the same chip and compared their performance and minimum dimensions. Figure 3(a) shows a micrograph of one of these devices. As mentioned above, nTrons are three-terminal devices that act similarly to a comparator. An nTron’s functionality depends heavily on its minimum dimension, the gate [see the inset of Fig. 3(a)]. To trigger an nTron, an input signal greater than the gate’s switching current is applied to the gate, creating a resistive “hotspot” that then causes the channel to switch from superconducting to normal.3 Since the ideal switching current of a superconducting structure is the product of the film’s critical current density and the structure’s width (ignoring confounding effects like nonuniformities), the width of the gate is critical in determining when the nTron fires. Thus, by comparing the width and switching current of each nTron’s gate, we can get a sense of how reliably ma-N resolves identical sub-50-nm features.

Fig. 3.

Three-terminal superconducting devices patterned using ma-N 2401. (a) Scanning electron micrograph showing one of the 36 patterned devices. The inset on the right shows an enlarged view of the narrowest (∼30 nm) feature, the gate. (b) Current-voltage characteristics of the gate of one of the devices, with the switching Isw and retrapping Ir currents indicated. (c) Profiles of the gates of the 36 patterned devices. Edge profiles were obtained by analyzing the SEM of each gate using prosem software. The x axis represents the horizontal distance across the gate in nanometers, while the y axis shows the vertical distance, which also defines the gate’s width.

Fig. 3.

Three-terminal superconducting devices patterned using ma-N 2401. (a) Scanning electron micrograph showing one of the 36 patterned devices. The inset on the right shows an enlarged view of the narrowest (∼30 nm) feature, the gate. (b) Current-voltage characteristics of the gate of one of the devices, with the switching Isw and retrapping Ir currents indicated. (c) Profiles of the gates of the 36 patterned devices. Edge profiles were obtained by analyzing the SEM of each gate using prosem software. The x axis represents the horizontal distance across the gate in nanometers, while the y axis shows the vertical distance, which also defines the gate’s width.

Close modal

As described in Sec. II, we measured the switching current of each gate in a probe station using a DC bias current ramp. Figure 3(b) displays the current-voltage characteristics of one sample device, with labels indicating the switching current Isw and the retrapping current Ir, or the point at which the nanowire switches back to the superconducting state. To compare the gate widths, micrographs of each gate (SEM settings: WD = 4.2 mm, mag = 31 550×) were analyzed using prosem software, which measured each feature’s minimum dimension. prosem was also used to generate the gates’ edge profiles, which were postprocessed for statistical data analysis. Figure 3(c) displays the profiles of all 36 gates, offset from each other in the z axis for clarity. The x axis represents the horizontal distance across the gate in nanometers, while the y axis shows the vertical distance, which also defines the gate’s width. By visually comparing the profile of each gate, we can see edge deviations from the desired pattern and observe differences between the structures in a way that is not obvious from simply looking at side-by-side micrographs.

While the edge profiles of Fig. 3(c) provide qualitative evidence of variability between devices, the minimum dimensions and switching currents offer a more quantitative and sensitive comparison. Figure 4(a) shows a histogram of the switching currents for all 36 devices, with a mean of 15.88 μA and a standard deviation of 3.13 μA. These statistics are typical of our nanowire devices, which usually have a standard deviation of 10%–20% of the mean.16 Figure 4(b) displays a histogram of the minimum dimensions measured using prosem. The average minimum gate dimension is 33.7 nm, and the standard deviation is 2.4 nm, suggesting that the patterned geometry was less variable than the switching current.

Fig. 4.

Reproducibility of the gate terminal across all 36 patterned devices. (a) Histogram of the switching currents of the gates. (b) Histogram of the minimum width across the gate, analyzed using prosem software. (c) Plot of the switching current of each device with respect to its minimum measured gate width. The line indicates that deviations from the mean switching current have a roughly linear relationship with deviations in minimum width, with a few outliers that could be due to noise or material defects.

Fig. 4.

Reproducibility of the gate terminal across all 36 patterned devices. (a) Histogram of the switching currents of the gates. (b) Histogram of the minimum width across the gate, analyzed using prosem software. (c) Plot of the switching current of each device with respect to its minimum measured gate width. The line indicates that deviations from the mean switching current have a roughly linear relationship with deviations in minimum width, with a few outliers that could be due to noise or material defects.

Close modal

To get a sense of how the spread in switching currents might be explained, we can plot the switching current of each device as a function of its minimum width, as shown in Fig. 4(c). The trend looks approximately linear, which agrees with our expectations, since the switching current is the product of the current density and the minimum width. However, there are some outliers from the trend, most notably the points that have a lower switching current than predicted. Such premature switching could be caused by a variety of factors, such as material inhomogeneities or grain boundaries in the polycrystalline NbN that cannot be observed by SEM. Nanowires are also susceptible to thermal fluctuations and noise, which may explain why there is a wider spread in switching current than in minimum dimension.17 Despite these variations, the switching current statistics align with our typical device performance,16 indicating that ma-N can produce superconducting devices such as nTrons with similar reproducibility to what we normally obtain using other resists.

We have investigated the use of ma-N 2401 as a suitable negative-tone resist for patterning superconducting devices. In comparison to HSQ, ma-N offers advantages including a lower required dose and the ability to be removed by a solvent that does not damage superconducting films, which is critical for multilayer processes. It may also be cheaper and have a longer shelf life than HSQ and could allow for a combination of both optical and electron-beam lithographies. To demonstrate ma-N’s potential, we showed that it can reliably pattern dense 30-nm-wide lines and individual features with minimum dimensions below 20 nm. This resolution is higher than what has been previously reported. However, it should be noted that HSQ remains the gold standard for resolution and might have superior etch resistance to ma-N in other chemistries like chlorine. The advantages of HSQ in comparison to ma-N are similar to those that it holds over other polymer-based resists and stem from HSQ’s underlying structure as a small, cagelike oligomer.18 Nevertheless, our results indicate that ma-N is a sufficient EBL resist for attaining the dimensions required by most superconducting nanowire devices.

We also tested the process reproducibility and device yield by patterning 36 identical superconducting nTrons, whose minimum feature sizes were ∼30 nm. By comparing each nTron gate’s switching current and minimum width, we were able to evaluate the deviation between identical devices. Our results indicate that the device deviation falls within the typical range for superconducting nanowires, demonstrating that ma-N is an acceptable negative-tone resist for these applications.

Widespread adoption of ma-N as an electron-beam resist could spur the implementation of nanowire devices in complex circuits that use multilayer processes. Since many of these circuits involve integration with other devices, such as standard electronic or magnetic elements, we envision this work encouraging collaboration across different platforms. Additionally, the multilayer processes supported by ma-N could inspire the development of new nanowire devices that incorporate other materials to modify superconducting dynamics. Examples of recent nanowire devices that use normal metal to modulate superconductivity have all been performed with positive-tone resist, partially due to the limitations of HSQ.19,20 As a result, ma-N could facilitate the development of new devices that use multilayer processes. Finally, the use of ma-N is not limited to superconducting films. We have demonstrated that it has ∼30 nm resolution and is processed with solvents that are compatible with many materials, suggesting that it could be used for a wide variety of applications beyond those presented here. Overall, ma-N 2401 appears to be a suitable electron-beam resist for patterning nanoscale features and has the potential to advance the complexity of superconducting nanowire-based circuits.

The authors would like to acknowledge Mark Mondol and James Daley of the NanoStructures Lab for their technical support, Di Zhu and Mina Bionta for comments on the manuscript, and all members of the Quantum Nanostructures and Nanofabrication Group. They would also like to thank Doc Daugherty for support with the prosem software. This research was supported by the Air Force Office of Scientific Research grant under Contract No. FA9550-14-1-0052. Emily Toomey was supported by the National Science Foundation Graduate Research Fellowship Program (NSF GRFP) under Grant No. 1122374.

1.
G. N.
Gol’tsman
 et al,
Appl. Phys. Lett.
79
,
705
(
2001
).
2.
F.
Marsili
,
F.
Najafi
,
E.
Dauler
,
F.
Bellei
,
X.
Hu
,
M.
Csete
,
R. J.
Molnar
, and
K. K.
Berggren
,
Nano Lett.
11
,
2048
(
2011
).
3.
A. N.
McCaughan
and
K. K.
Berggren
,
Nano Lett.
14
,
5748
(
2014
).
4.
M.
Tanaka
,
M.
Suzuki
,
G.
Konno
,
Y.
Ito
,
A.
Fujimaki
, and
N.
Yoshikawa
,
IEEE Trans. Appl. Supercond.
27
,
1800904
(
2017
).
5.
Q.-Y.
Zhao
,
E. A.
Toomey
,
B. A.
Butters
,
A. N.
McCaughan
,
A. E.
Dane
,
S.-W.
Nam
, and
K. K.
Berggren
,
Supercond. Sci. Technol.
31
,
035009
(
2018
).
6.
J. K. W.
Yang
 et al,
IEEE Trans. Appl. Supercond.
15
,
626
(
2005
).
7.
J. K. W.
Yang
,
B.
Cord
,
H.
Duan
,
K. K.
Berggren
,
J.
Klingfus
,
S.-W.
Nam
,
K.-B.
Kim
, and
M. J.
Rooks
,
J. Vac. Sci. Technol. B
27
,
2622
(
2009
).
8.
M.-H.
Nguyen
 et al, preprint arXiv:cond-mat/040426.
9.
E.
Toomey
,
M.
Colangelo
,
N.
Abedzadeh
, and
K. K.
Berggren
,
J. Vac. Sci. Technol. B
36
,
06JC01
(
2018
).
10.
H.
Elsner
,
H.-G.
Meyer
,
A.
Voigt
, and
G.
Grützner
,
Microelectron. Eng.
46
,
389
(
1999
).
11.
B.
Bilenberg
 et al,
J. Vac. Sci. Technol. B
24
,
1776
(
2006
).
12.
M. N.
Mn
,
U.
Hashim
,
M. K.
Md Arshad
,
A. R.
Ruslinda
,
S. F. A.
Rahman
,
M. F. M.
Fathil
, and
Mohd. H.
Ismail
,
PLoS ONE
11
,
1
(
2016
).
13.
A. E.
Dane
,
A. N.
McCaughan
,
D.
Zhu
,
Q.
Zhao
,
C.-S.
Kim
,
N.
Calandri
,
A.
Agarwal
,
F.
Bellei
, and
K. K.
Berggren
,
Appl. Phys. Lett.
111
,
122601
(
2017
).
14.
O.
Medeiros
,
M.
Colangelo
,
I.
Charaev
, and
K. K.
Berggren
,
J. Vac. Sci. Technol. A
37
,
041501
(
2019
).
15.
J. K. W.
Yang
,
V.
Anant
, and
K. K.
Berggren
,
J. Vac. Sci. Technol. B
24
,
3157
(
2006
).
16.
F.
Najafi
,
A.
Dane
,
F.
Bellei
,
Q.
Zhao
,
K. A.
Sunter
,
A. N.
McCaughan
, and
K. K.
Berggren
,
IEEE J. Sel. Top. Quantum Electron.
21
,
2372054
(
2015
).
17.
A. N.
McCaughan
,
D. M.
Oh
, and
S. W.
Nam
,
IEEE Trans. Appl. Supercond.
29
,
1
(
2019
).
18.
A. E.
Grigorescu
and
C. W.
Hagen
,
Nanotechnology
20
,
292001
(
2009
).
19.
E.
Toomey
,
M.
Onen
,
M.
Colangelo
,
B. A.
Butters
,
A. N.
McCaughan
, and
K. K.
Berggren
,
Phys. Rev. Appl.
11
,
034006
(
2019
).
20.
A. N.
McCaughan
,
V. B.
Verma
,
S. M.
Buckley
,
A. N.
Tait
,
S. W.
Nam
, and
J. M.
Shainline
, e-print arXiv: 190310461 (2019).