Energy gap measurement of nanostructured thin aluminium films for use in single Cooper-pair devices
aa r X i v : . [ c ond - m a t . s up r- c on ] J un Energy gap measurement of nanostructured thinaluminium films for use in single Cooper-pairdevices
N A Court, A J Ferguson † , and R G Clark Australian Research Council Centre of Excellence for Quantum ComputerTechnology, University of New South Wales, Sydney, NSW 2052, AustraliaE-mail: [email protected]
Abstract.
Within the context of superconducting gap engineering, Al-Al O -Altunnel junctions have been used to study the variation in superconducting gap, ∆,with film thickness. Films of thickness 5, 7, 10 and 30 nm were used to form the smallarea superconductor-insulator-superconductor (SIS) tunnel junctions. In agreementwith previous measurements we have observed an increase in the superconductingenergy gap of aluminium with a decrease in film thickness. In addition, we find grainsize in small area films with thickness ≥
10 nm has no appreciable effect on energy gap.Finally, we utilize 7 and 30 nm films in a single Cooper-pair transistor, and observe themodification of the finite bias transport processes due to the engineered gap profile.PACS numbers: 74.78w, 73.23.Hk, 85.25.Cp
Submitted to:
Supercond. Sci. Technol. † Present address: Microelectronics Research Centre, Department of Physics, Cavendish Laboratory,Cambridge, CB3 0HE, UK. nergy gap measurement of nanostructured thin aluminium films
1. Introduction
By contacting superconducting materials with different gap energies it is possible tomodify the energetics of quasiparticle states in different regions of a superconductingnanostructure. This allows quasiparticles to be either confined in or excluded fromcertain parts of a device. This principle of quasiparticle gap engineering is used in photondetection where it is desirable to confine excess quasiparticles resulting from a photonabsorption event [1–3]. In addition, the interplay between Coulomb blockade and anengineered gap profile can be used to suppress unwanted quasiparticle tunnelling (QPpoisoning) in single Cooper-pair transistors (SCPTs) and Cooper-pair boxes (CPBs)[4–6].Methods of gap engineering include the use of different superconducting materialssuch as aluminium and tantalum [1, 3], superconductor-normal metal bilayers [2],oxygen-doped aluminium [4, 7] and more recently different thickness aluminium films[5, 6, 8]. This work utilizes the latter method in which changes in the superconductinggap of more than 50% may be achieved [9]. Early experiments on thin aluminiumfilms observed the enhancement of the superconducting gap and transition temperatureswell above bulk values. This enhancement has been attributed to different types ofdisorder found in thin film structures such as grain size, average lattice constant andthe presence of an oxide layer surrounding each grain [10–13], as well as the film thicknessitself [9, 10, 14].Controlling the superconducting gap profile using different thickness aluminiumfilms is a useful concept. In particular, there is no need for introduction of differentmaterials, which can complicate the evaporation process. However, the use of films asthin as 5 nm in Coulomb blockade devices such as the SCPT [8] involves evaporationonto a liquid nitrogen cooled stage and requires reliable fabrication techniques. It is alsoimportant to identify any unusual characteristics due to added disorder in such filmssuch as the presence of multiple superconducting gaps.Aumentado et al [4] have recently used oxygen doping to investigate the parityof the supercurrent in a gap engineered SCPT. The parity (whether the supercurrentoccurs at odd or even integer gate charge) was found to be strongly influenced by thedifference in gap energies of the island and leads δ ∆ = ∆ i − ∆ l . Their explanationof this behaviour involves the tunnelling of thermal, or non-equilibrium, quasiparticlesfrom the leads onto the device island. Essentially δ ∆ is the energy cost of transferring aquasiparticle from the leads to the island, and consequently, it alters the correspondingquasiparticle tunnel rates and occupation probability. For the case of positive δ ∆ thequasiparticle tunnel rates can be suppressed, and the subsequent use of radio-frequencytechniques have enabled several experiments in which quasiparticle tunnelling rates werestudied on microsecond timescales [6, 15]. The ability to create well known δ ∆ in singleCooper-pair devices is the main motivation behind this study.Here we present measurements of small area ( ≤ ×
100 nm ) aluminiumsuperconductor-insulator-superconductor (SIS) junctions. From the maxima in nergy gap measurement of nanostructured thin aluminium films (a)(b) (c) Figure 1.
Thin film Al-Al O -Al junctions were fabricated using a Dolan bridgeresist structure and oxidized in-situ. (a) shows a scanning electron micrograph (SEM)of a typical device structure, with a junction size of ∼ ×
75 nm . SEM images of thegrain structure of an aluminium thin film (thickness 10 nm) in (b) a room temperatureevaporated device T ∼
293 K and (c) device evaporated onto a liquid nitrogen cooledstage at T ∼
173 K . differential conductance measurements we infer the peak density of states, and thusdetermine superconducting gap energies, ∆, of d = 5, 7, 10 and 30 nm aluminiumfilms [16]. We focus on single thickness film (S1-I-S1) junctions for more reliability inthe determination of gap energies. We also investigate films with different grain sizes,by changing evaporation temperatures, to ascertain if there is any appreciable effect onenergy gap. Finally, we briefly present measurements on a gap engineered SCPT (30nm - 7 nm - 30 nm). We discuss both the 2 e -supercurrent and the finite bias resonances.
2. Thin-film Junction Fabrication
Figure 1 shows a typical test sample with a junction area of ∼ ×
75 nm . A number ofdevices were measured with resistances ranging between ∼
20 kΩ − · cm) silicon substrate with 200 nm of thermallygrown silicon dioxide on the surface. To enable sufficient contact to the thinner films( ≤
10 nm) a single layer of PMMA was patterned to define thin overlap areas ( ∼ − nergy gap measurement of nanostructured thin aluminium films Figure 2.
Normalized differential conductance as a function of applied bias voltagefor 4 different thickness films. 2∆ was determined from the peak in differentialconductance. The 4.2 K resistance of each device is also shown. (a) d = 30 nm,2∆ = 430 µ eV (b) d = 10 nm, 2∆ = 476 µ eV (c) d = 7 nm, 2∆ = 584 µ eV and (d) d= 5 nm, 2∆ = 608 µ eV. (d) The dash-dot line indicates dI/dV = 0 and the points atwhich dI/dV crosses from > < Liquid nitrogen could be introduced into the chamber via a modified feed through intothe top of the oxide chamber, making contact with the stage onto which the substratewas mounted.The temperature of the stage was monitored via a thermocouple in contactwith the back of the stage, giving an indication of the substrate temperature duringevaporation. Cooled evaporations were performed at T ≃
173 K whilst roomtemperature evaporations were at T ≃
293 K. The variation in temperature duringevaporation was ≤ × − mbar O for 4 min. We found that a steady evaporation rateis particularly crucial in achieving continuous films of 10 nm, for a room temperatureevaporation, and 5 nm, for a cold evaporation. The aluminium was evaporated at a rateof 0.1 nm · s − .The grain structure of the films depends strongly on evaporation temperature. Ascanning electron micrograph of the grain structure of a 10 nm film evaporated at 293K is shown in figure 1(b). Large structure is observed with grains of up to 40 nmin diameter seen. In contrast grain structure in the 10 nm film evaporated at 173 K(figure 1(c)) is significantly smaller with the largest grain size <
10 nm in diameter.
3. Results and Discussion
Measurements were performed in a dilution refrigerator with a base temperature ofapproximately 100 mK. The differential conductance ( dI/dV ) was measured with astandard low-frequency ac-lock-in technique with a modulation amplitude of 10 µ V. nergy gap measurement of nanostructured thin aluminium films Figure 3.
Observed superconducting energy gap ∆ of aluminium as a function offilm thickness. Each point represents a single junction whose gap energy was extractedvia differential conductance measurements. Films were evaporated onto a substrate incontact with a liquid nitrogen cooled stage at a temperature of ∼
173 K with grainstructure similar to that seen in figure 1(c).
Figure 2 shows plots of typical differential conductance vs applied bias voltage for fourdifferent thickness films.Maxima corresponding to the peak in density of states at the gap edges allowsaccurate determination of 2∆ [18]. There is a clear dependence on the observed 2∆ as afunction of film thickness ranging from 2∆ = 430 µ eV for the 30 nm film (see figure 2(a)),to an increase of almost 50% for 5 nm films with 2∆ = 608 µ eV (see figure 2(d)). Thedifferential conductance has been rescaled to a normalized resistance due to differencesin junction area as indicated by the normal state resistances given in figure 2. It is ofinterest to note that only singular peaks occur in the quasiparticle density of states.This is opposed to multiple gap structures that are sometimes observed in disorderedfilms [19].Subgap structure in the 5 nm devices (see figure 2(d)) is more complicated, perhapsdue to the lower junction resistance (R = 26 kΩ) of this sample. We see negativedifferential resistance regions ( dI/dV <
0) which lead to peaks in the integrated dI/dV (not-shown). Similar behaviour was also seen in a second device (R = 21 kΩ). Atpresent we cannot attribute this behaviour to specific transport processes. However, wenote that multiple Andreev reflection is unlikely to be the origin since the positions donot coincide with integer multiples of 2∆ /n [20].Figure 3 shows the distribution of the measured energy gap ∆ as a function offilm thickness, d. We observe a increase in ∆ as the thickness of the film is decreased.The magnitude of this increase is consistent with other studies which infer gap energiesfrom the critical temperature of the films [10, 11, 14]. From the standard distributionof gap energies taken at each thickness we see a spread of up to 15 µ eV away fromaverage values [Table 1]. The magnitude of this variation is likely to be due to changesin conditions (e.g. evaporation rate and substrate temperature) between successiveevaporations. We support this by noting that junctions made in the same evaporation(for example the 5 nm and the 10 nm room temperature evaporated films) tend to have nergy gap measurement of nanostructured thin aluminium films Table 1.
SIS parameters for cold and room temperature evaporations. Number ofsamples measured, evaporation temperature, evaluated gap energies and standarddeviation for each thickness film. d Number of Approx. Junct. Tevap ∆ ± σ ∆(nm) Samples area (nm) (K) ( µ eV)5 2 100 ×
100 173 307 ± ×
30 173 298 ± ×
70 173 250 ± ×
70 293 236 ± ×
70 173 209 ± ×
50 293 208similar characteristics. The gap energies agree with those obtained in [5] and Table 1indicates the variation in ∆ that can be easily achieved both by room temperature andcooled substrate evaporations.Films with larger grain structure [evaporated at 293 K, see Table 1] have peaks at eV = 2∆ which fall within the standard deviation of energies for both the 10 and 30nm films evaporated at 173 K. Consequently, we see no indication that the grain sizehas a strong influence on energy gap. Our results agree with the conclusions of previousinvestigations [10, 12] that grain size alone does not account for the enhancement inenergy gaps.To review, we have presented differential conductance measurements of small areaSIS junction which show an enhancement of superconducting gap with decreasingthickness films. A variation of up to 15 µ eV from average values for each thicknessis accounted for due to variability in evaporation conditions. We see no significantdependence of gap energies on grain size for thicker films.
4. Thin-Films in the Single Cooper-Pair Transistor
The aim of this work is to be able to design superconducting single Cooper-pair deviceswith well-controlled superconducting gap profile, hence in this section we briefly describeelectrical transport measurements on such a sample. We made SCPTs using the samefabrication procedure and junction areas as the SIS tunnel junctions (see figure 4(a)).The measurements were performed at milliKelvin temperature in a dilution refrigeratorby a radio-frequency reflectometry technique [21]. The SCPT was embedded in aresonant LC tank circuit and the reflection coefficient of an incident radio-frequencysignal at the circuit resonance frequency is related to the differential conductance of theSCPT [6, 15]. We use the same rf-setup as described in [6]. nergy gap measurement of nanostructured thin aluminium films δ ∆ = ∆ i − ∆ l ∼ µ eV for a 30 nm - 7 nm - 30 nm SCPT (figure 4(b)). The 4.2 K resistance of thisdevice was 54 kΩ and the charging energy E C = e / C Σ = 190 µ eV, as determinedfrom normal-state Coulomb diamonds measured at B = 3 T. Estimating the Josephsonenergy per junction from the 4.2 K resistance and the Ambegoakar-Baratoff relation( E J ∼ h ∆ i ∆ l i +∆ l ) e R ) we find E J = 30 µ eV.In the Coulomb diamonds (see figure 4(c)) we see peaks in the reflected powercorresponding to 2 e -periodic supercurrent in n g at zero bias. Supercurrent peaks arenot observed in the case of a device without gap engineering and their presence is dueto a reduced quasiparticle occupation probability on the island. The appearance of thesupercurrent due to the quasiparticle barrier on the island is consistent with previousstudies [4, 5].Coulomb blockade of quasiparticle tunnelling occurs for eV ds > i +2∆ l , and fromthe threshold of this process we can estimate the superconducting gaps of the leadsand island. In this device 2∆ i +2∆ l = 1.08meV, and taking the ∆ nm = 209 µ eV, then∆ nm = 331 µ eV. We note that this is larger than expected for a 7 nm film and is againlikely due to be caused by a variation in evaporation conditions. Qualitatively we noticethe presence of a large quasiparticle co-tunnelling current in the gap engineered devices.At finite bias, but for eV ds < i +2∆ l , there are e -periodic features correspondingto the sequential tunnelling of both Cooper-pairs and quasiparticles (figure 4(c)). Weobserve e -periodic peaks related to resonant Josephson quasiparticle (JQP) and doubleJosephson quasiparticle (DJQP) cycles [22–24]. The JQP cycle consists of the coherenttunnelling of a Cooper-pair through one junction followed by two quasiparticles throughthe other. The condition for Cooper-pair tunnelling must be satisfied and, additionally,the energetics must permit the subsequent tunnelling of two quasiparticles. This resultsin e -periodic repetition of a pair of crossed-lines in the range E C +2∆ < eV ds < E C +2∆.In principle, the primary change to the JQP cycle due to the modified gap is thatthe thresholds change so that the cycle occurs in the range E C + ∆ i + ∆ l < eV ds < E C + ∆ i + ∆ l . The lower threshold is plotted in figure 4(c) with the estimated valuesfor the superconducting gaps and appears to correspond to the start of the JQP cycle.In the DJQP cycle, coherent tunnelling of Cooper-pairs between the island andjunctions is allowed with quasiparticle tunnelling events determining which junctionis on resonance [25]. For the cycle to be possible, the quasiparticle events must bepermitted to occur but, due to the need to satisfy Cooper-pair tunnelling across bothjunctions, its position must remain fixed at eV ds = ± E C . The DJQP cycle can beseen in figure 4(c).Additional transport resonances appear just above the DJQP, and parallel to theJQP. Unlike the previously described processes, these differ between devices and maybe due to transport though higher order bands in the SCPT or more complex cyclesinvolving both Cooper-pair and quasiparticle tunnelling.To summarize, the fabrication of a gap engineered SCPT with a positive δ ∆ hasa number of different effects on the transport processes. The primary difference is the nergy gap measurement of nanostructured thin aluminium films
30 nm 30 nm7 nm (c)
Figure 4.
We utilize the our knowledge of energy gap and film thickness to design aSCPT with δ ∆ ∼ µ eV using a 7 nm film (∆ ∼ µ eV) for the island and 30 nm film(∆ ∼ µ eV) for the leads. (a) Scanning electron micrograph of a device similar tothat measured showing the contrast difference between the 7 nm island and the 30 nmleads. (b) Schematic profile of the change in thickness across the SCPT. (c) Observedsuperconducting Coulomb diamonds of a SCPT device with a 7 nm island and 30 nmleads. 2 e -periodic supercurrent peaks can be clearly seen at zero bias, while at finitebias resonances corresponding to e -periodic transport are seen (indicated in figure). A corresponds to eV ds = 2∆ i +2∆ l , B corresponds to the lower threshold for the JQPcycle occurring at eV ds = E C + ∆ i + ∆ l , C shows the condition for the JQP cycle, and D shows the DJQP occurring at eV ds = ± E C . appearance of a supercurrent which is discussed in greater detail in [4]. Furthermore,the thresholds for Coulomb blockade of quasiparticle tunnelling and the Josephsonquasiparticle resonance are shifted. Explanation of the additional resonant features willrequire a more detailed study of the energetics perhaps involving knowledge of higherorder bands.
5. Conclusion
We have observed an increase in the superconducting energy gap of aluminium, as afunction of decreasing film thickness in small area SIS junctions. We find that our resultsare consistent with previous studies and that cold evaporation of thin films produceshigh quality small area junctions which can be used in superconducting gap engineered nergy gap measurement of nanostructured thin aluminium films e -periodic supercurrent and a modification of the finite bias transportprocesses. In addition, for this technique to be used in quantum nanostructures, suchas the Cooper pair box, a measurement of the charge noise in the cold-evaporated filmand fluctuator density of the tunnel junctions remains to be undertaken. Acknowledgments
The authors would like to thank T. Duty for helpful discussions and D. Barber and R.P. Starrett for technical support. This work is supported by the Australian ResearchCouncil, the Australian Government, and by the US National Security Agency (NSA)and US Army Research Office (ARO) under Contract No. W911NF-04-1-0290.
References [1] M C Gaidis, S Friedrich, D Prober, S Moseley, and A Szymkowiak 1993
IEEE Trans. Appl.Supercond. Phys. Rev. Lett. Phys. Rev. Lett. Phys. Rev. Lett. Appl. Phys. Lett. Phys. Rev. Lett. Phys. Rev. B. Phys. Rev. Lett. Sov. Phys. JETP Can. J. Phys. Phys. Rev. B. Phys. Rev.
Phys. Rev. B. J. Appl. Phys. Phys. Rev. B. Introduction to Superconductivity . Dover Publications Inc.[17] G. J. Dolan 1977
Appl. Phys. Lett. Phys. Rev. Lett. Phys. Rev. B. Physica (Amsterdam)
Science , Phys. Rev. Lett. Phys. Rev. Lett. Phys. Rev. B. Phys. Rev. Lett.89