Anwar, M.S.
Citation
Anwar, M. S. (2011, October 19). Spin triplet supercurrents in thin films of ferromagnetic CrO2. Casimir PhD Series. Retrieved from https://hdl.handle.net/1887/17955
Version: Corrected Publisher’s Version
License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden
Downloaded from: https://hdl.handle.net/1887/17955
Note: To cite this publication please use the final published version (if applicable).
Supercurrents through CrO 2
This chapter concerns measurements of supercurrents through the half metal- lic ferromagnet CrO2 grown on TiO2 and hexagonal Al2O3 (sapphire). The current was observed to flow over a distance of the order of a micrometer between two superconducting amorphous Mo70Ge30electrodes deposited on a CrO2film. The critical current Icincreases as a function of decreasing temper- ature. Upon applying an in-plane magnetic field, Icgoes through a maximum at the rather high field of 80 mT. For films deposited on TiO2, supercurrents were not detected through the junctions until a thin layer of Ni(2 nm) was deposited before sputtering superconducting leads, with a 5 nm thick Cu layer between CrO2 and Ni to decouple the magnetizations of both ferromagnets.
The critical current density in this case is significantly higher than that of the sapphire based junctions, which suggests homogeneous spin activity of the interface. The long ranged current is argued to be carried by odd-frequency spin triplet pairing correlations.1
1Parts of this chapter have been published in,
M. S. Anwar, F. Czeschka, M. Hesselberth, M. Porcu, and J. Aarts, Long-range supercurrents through half-metallic ferromagnetic CrO2, Phys. Rev. B 82, 100501(R) (2010).
M. S. Anwar and J. Aarts, Inducing supercurrents in thin films of ferromagnetic CrO2, Supercond. Sci. Technol. 24, 024016 (2011).
61
5.1 Introduction
It has been discussed in Chapter 2 that conventional superconductivity and ferromagnetism are antagonistic in nature. In principle, spin-singlet Cooper pairs from the superconductor (S) dephase quickly over a coherence length ξF = p~DF/hex (dirty limit) in the ferromagnet (F) under the influence of its exchange field hex (and DF the electronic diffusion constant in the F- metal), The order of magnitude for ξF ranges from 1 nm (strong magnet) to 10 nm (weak magnet). Such dephasing would not occur with equal spin triplet Cooper pairs. By allowing the symmetry of the wave function to be odd in frequency (or in time), it is possible to construct spin-triplet pairs with s-wave symmetry. The triplet Cooper pairs will suffer from spin flip scattering, but since the spin diffusion length is usually much larger than ξF of the singlets, this may lead to a Long Range Proximity (LRP) effect in the magnet. As described in Chapter 1 and 2 it was predicted that spin-triplet correlations could be induced at an S/F interface and sustained in the F-layer under the condition of the presence of an inhomogeneous exchange field [10, 25, 54], for instance from domain walls or unaligned magnetic moments. More recently, it was shown that for a long-ranged Josephson current to exist, two sources of triplet components are required [24]. A simple realization for this would be an S/F’/F/F”/S system in which the magnetizations of the F’, F” layers are non-collinear with the central F layer.
In early 2006, the experimental observations of Keizer et al. [11] and Sosnin et al. [12] revealed the existence of LRP effect in ferromagnets, possibly generated via spin-triplet superconductivity. In the first case, a supercurrent was measured in SFS devices where superconducting electrodes of NbTiN with separations up to 1 µm were placed on unstructured 100 nm thick films of CrO2 (a half metallic ferromagnet or HMF) which were grown on TiO2 substrates. In the second case, the LRP effect was observed in ferromagnetic Ho wires of lengths up to 150 nm using an Andreev interferometer geometry.
No other experiments were reported for quite some time, but this is now rapidly changing. Long ranged superconducting correlations were reported in single crystalline Co nanowires, reaching a distance of a micron [17], although the presence of magnetic anisotropy or inhomogeneity are not clear. A very recent theoretical investigations [55] suggested that it is not necessarily long ranged triplet supercurrent because a long ranged spin singlet is also possible for the 1D case of ferromagnetic nanowires. At the same time, reports came out on Josephson junctions where Co layer were used in combination with thin PdNi, CuNi or Ni layers [15, 56] and/or Ho layers [16] to provide magnetic inhomogeneity. No decay of the Josephson current was found up to a thickness of 50 nm of the Co layer, which is significantly larger than what was found
without such engineered inhomogeneities [58]. In the first case, non-collinear magnetization exists between the PdNi layer which is out-of-plane magnetized, and the Co layer which has in-plane magnetization; in the second case the spiral ferromagnetism of the Ho layer provides the required inhomogeneity.
Neither Ho nor Co are fully spin polarized, and the triplet decay will mainly be set by the spin diffusion length, of the order of 100 nm in both materials.
That makes the CrO2 case with its significantly larger decay length of special interest, but here the issue of reproducibility has hampered progress. The original report mentioned large variations in the magnitude of the critical (su- per)current Ic between different samples, and many of them not showing any effect [11]; no other reports on experiments with CrO2 were published. That has much to do with the fact that, in the simple geometry of S contacts on an F layer, it is not fully clear where the inhomogeneous magnetization re- sides which is needed for the triplet generation, although a possibility seems to be the presence of disordered moments at the interface. Here we report new observations of supercurrents in CrO2, using devices which are different from the earlier ones in various aspects. We have grown CrO2 films on Al2O3 (sapphire) rather than on TiO2, which leads to significant differences in film morphology as discussed in Chapter 3; and the superconducting contacts are made from amorphous (a-)Mo70Ge30, rather than from NbTiN. Again we find significant values for Ic even at a separation of about 1 µm between the elec- trodes, and only small sensitivity to applied magnetic fields up to 0.5 T. We did not succeed, however, to find supercurrents in films grown on TiO2, until we put an extra ferromagnetic layer, with thickness less than corresponding coherence length, between the S and CrO2. We used 2 nm thick Ni as an extra layer with 5 nm thick Cu to magnetically decouple the Ni and CrO2and observed long ranged supercurrents. Our observations strengthen the conclu- sion that odd frequency triplets can generally be induced in ferromagnets via inhomogeneous magnetization and/or disordered magnetic moment at the in- terface, leading to LRP effects. Also, we look more closely at the behavior of the resistance R of our devices around the superconducting transition tem- perature Tc and show that the characteristic behavior of the clean S/HMF interface is a jump of R to a higher value, rather than the simple decrease usually encountered.
5.2 Device fabrication
A special issue in the device preparation lies in the growth of CrO2 films, which was discussed in detail in Chapter 3. The S/F/S devices were fab- ricated by sputtering a-Mo70Ge30 superconducting contacts on unstructured
Figure 5.1: Scanning electron microscopy image of a device fabricated on CrO2 (dark gray) using a-MoGe (light gray) electrodes. Two junctions are visible with gaps along the c-axis and the b-axis of the TiO2substrate. Both junctions are measured independently. The inset shows a close-up of one of the gaps, of the order of 700 nm.
100 nm thick CrO2 thin films deposited on untreated and pretreated sub- strates (films deposited on sapphire were 80 nm thick), in a lateral geometry.
In principle, the best structure to work with would be a simple CrO2bar, con- tacted on both sides by superconducting contacts. Such a bar can be made lithographically by direct etching, but the problem is that the insulating sub- strate does not allow the next e-beam step of defining contacts. In particular, it is difficult to etch a structure into the film and then define electrodes on the bare substrate. Instead, we made the devices by (RF-)sputtering 60 nm of a-Mo70Ge30 superconducting electrodes with a superconducting transition temperature Tc = 6 K through a lift-off mask onto the unstructured film us- ing MMA/PMMA-950K bilayer resist. The choice of the superconductor was guided by the need to avoid the formation of unwanted oxides at the interface, as was found to occur for Nb and Al [59] (although not for NbTiN). Before sputtering the superconductor and after developing the structure, the film sur- face was cleaned briefly with an O2 reactive ion plasma, in order to remove resist or developer residues. This is important for more rough films, like those
deposited on sapphire. Next, Ar-ion etching was applied immediately prior to deposition, in order to remove newly formed Cr2O3 on the film surface, and the Mo70Ge30was sputtered without breaking the vacuum. The width of the electrodes was 30 µm and the gap between the electrodes was varied from 1 µm down to 500 nm. Two gaps were made perpendicular to each other and with the electrode directions along the crystallographic b-axis and the c-axis, respectively, to see the effect of magnetic anisotropy, especially for samples prepared with TiO2. Both junctions were measured independently. A picture of the device geometry is shown in Fig. 5.1.
The devices fabricated with CrO2 films on TiO2 were of two different types. One only had CrO2 as ferromagnetic layer, while in second we used a Ni(2 nm)/Cu(5 nm) bilayer over the CrO2after cleaning and before sputtering the superconductor. The Cu layer is used to decouple the two ferromagnetic layers. For smaller gap (<500nm) devices, bilayer PMMA-495K/950K was utilized to avoid a short through the resist because of rather bigger undercut occurring with MMA.
One other type of device was also prepared, where superconducting elec- trodes were deposited on a Au layer on Si substrate in the same way as the devices with CrO2/TiO2 were prepared, in order to compare the behavior of low resistance normal metal with that of the equally low resistance CrO2.
Critical currents as function of temperature and applied field were mea- sured in a mu-metal shielded Oxford cryostat, where the temperature can be controlled in the range of 1.6 - 10 K. The temperature dependent zero bias resistance of the junctions was measured using a dc current source (Keithley 220) and nanovoltmeter (Keithley 2182).
5.3 Results; S contacts on CrO
25.3.1 TiO2 based devices
The devices prepared with 100 nm thick CrO2 films deposited on pretreated TiO2 substrates always showed a sharp up jump in the resistance R at the critical temperature Tc as shown in Fig. 5.2a, if they were fabricated with proper cleaning to remove Cr2O3. Such an R(T ) behavior is anisotropic with about 6-fold increase in RN from 580 mΩ along b-axis and only about 2-fold increase along the c-axis. We come back to the values for RN in the discussion.
Figure 5.2b shows the R(T ) behavior of devices which were fabricated with- out any cleaning. The resistance jumps down at Tc as expected, with about 40% decrease, followed by an increase of about 15%. For these devices the RN is about 720 mΩ, which somewhat higher than that of cleaned devices.
These observations of an up-jump suggest that either it is connected with the
geometry of the devices or the HMF nature of the CrO2. The device with Au instead of CrO2 as bottom film shows a normal down jump in R(T ) (see inset of Fig. 5.2b), which also indicates that the behavior is peculiar for the ferromagnet, and has probably to do with the high degree of spin polarization of the CrO2. Such behavior has not been reported before and to unravel the physics needs more experimentation. The results of some more experiments to understand this unexpected behavior of the resistance are given in Appendix B.
Figure 5.2: Normalized resistance (with normal state resistance at T > 6 K) as a function of temperature R(T ) for TiO2based devices (a) when fabricated with proper cleaning to remove Cr2O3. It shows an up-jump at Tc along both the b- and the c-axis with a relatively larger jump along the b-axis. (b) For a device fabricated without cleaning, the resistance is showing a normal down jump at Tc with a dip structure before saturating.
These devices did not show supercurrents, which might be related with the magnetic anisotropy of the CrO2 thin films. Most of our films showed uniaxial anisotropy at lower temperatures, so in this sense a true comparison with the earlier work of Ref. [11] was not made. In some of our experiments we used thin films deposited on pretreated TiO2 substrate of the order of 40 nm thickness, where the presence of biaxial anisotropy is expected. We still did not find any supercurrent, just the same up-jump in resistance was found at Tc. It seems that the generation of supercurrents in CrO2 deposited on TiO2 is difficult because of poor control over the magnetic anisotropy (which in turn depends on the substrate cleaning procedure) and over the interface cleanliness.
5.3.2 Sapphire based devices
A number of devices were prepared with 80 - 100 nm thick CrO2films deposited on sapphire substrates in the way as described above, and 3 out of roughly 10 showed a supercurrent. We call them AS, BS, and CS. Another device DS is also discussed which does not show any supercurrent. Device BS was slightly different from the other two in that it consisted of three parallel electrodes rather than one, with a distance between electrodes of 100 µm, and the three gaps measured in parallel. Figure 5.3 illustrates the data of the zero bias resistance as a function of temperature, measured with a current of 10 µA the junctions AS and DS. At Tc, junction DS shows an up-jump similar to the TiO2 based junctions. For junction AS, the up-jump is still there but with decrease in temperature the resistance goes down to zero, and a clear signatures of a proximity effect is observed.
Figure 5.3: Temperature dependent normalized resistance (with normal state resistance at T > 6 K) for sapphire based junctions, for junction DS (circles), a large resistance jump (at Tc of the S electrodes) is observed but no super- current. For junction AS (square) the up-jump is followed by a decrease to zero resistance.
Figure 5.4 shows the current-voltage (I-V ) characteristic of device AS, taken between 6 K (just below Tc) and 2.5 K. We observe a clear zero resis- tance supercurrent branch, with a maximum value for Ic of 170 µA at 2.5 K.
The inset shows data for device BS (the one with three parallel electrodes) measured at 2 K.
Figure 5.4: Current (I) versus Voltage (V ) measurements for device AS at 2 K, 3.15 K, 4 K and 6 K. The values of the critical current are indicated and the RN is around 10 Ω until the temperature is below the Tc. The inset shows an I-V characteristic for device BS at 2 K.
From these measurements Ic was determined as the first deviation from the linear I-V characteristic around zero bias (equivalent to the peak in the derivative dI/dV ). The temperature dependence Ic(T ) is given in Fig. 5.5 for all three samples. All devices have very similar values for the critical current, even for the case of three parallel electrodes. The behavior close to Tc is concave rather than linear. In Fig. 5.6 we present the effect of applying a magnetic field Ha on Ic in device AS at a temperature of 3 K. The field was applied in the plane of the film, with a direction either parallel to the long axis of the electrodes, or perpendicular to that axis. In the first configuration we do not find effects up to 500 mT. In the second configuration we find large changes, however. Starting from zero field, Icincreases by about 10% and goes through a maximum around 80 mT before dropping down to a level which at 500 mT is about 10% below the zero field value. Sweeping back, the behavior is different, with a relatively sharp jump back to the zero field level, but no peak as in the forward sweep. Continuing in the negative field quadrant, no
structure in Ic(Ha) was found. A point to note is that the maximum lies well outside the hysteresis loop of the magnetic film. The coercive field Hc is of the order of 10 mT only (see Fig. 3.11). Unfortunately, the samples proved fragile and could only be cooled down a few times before the supercurrent disappeared. However, there was no slow degradation during a measurement.
5.3.3 Discussion
The results are best discussed in comparison with the previous report on super- currents in CrO2 [11]. Firstly, we can compare their magnitudes by assuming that the current flows homogeneously across the bridge and through the full thickness dCrO2 of the layer. In our case (dCrO2 ≈ 100 nm, bridge width 30 µm, current 100 µA) we find a critical current density at 2 K of about 3
×107 [A/m2]. Keizer’s data (dCrO2 = 100 nm, bridge width 2 µm, typical current 1 mA) correspond to 5 ×109[A/m2], and from this point of view there appears to be a large difference between the two results. Comparing the field dependence, in Ref. [11] a Fraunhofer pattern was detected with a distance between maxima of about 90 mT. Assuming this to be equivalent to one flux quantum Φ0in the junction area of 310 nm × dCrO2, a value of roughly 80 nm is found for dCrO2, quite close to the nominal thickness and suggesting that the full film thickness was partaking in the supercurrent (c.q. in the shielding from the magnetic field). In the dataset presented here (Fig. 5.6) a Fraunhofer pattern is not clearly visible, but there is a maximum at 80 mT followed by discontinuities around 150 mT and 250 mT, and a small maximum at 300 mT.
Taken together, this suggests a period of 100 mT. For a junction area of 700 nm ×dCrO2, this corresponds to dCrO2 ≈ 30 nm, which indicates that in our case the current is not flowing through the full thickness of the layer.
The picture then emerging is that, although the results are qualitatively the same, the growth on Al2O3 leads to weaker junctions. Since the TEM picture in Fig. 3.2b shows that in our devices grain boundaries will always be in the path of the current, this actually seems a reasonable conclusion. Another point to discuss is that the maximum in the field data is not found at zero field, which in Ref. [11] was ascribed to the finite sample magnetization. That is probably not a sufficient explanation since saturation of the magnetization is reached at a significantly smaller field value. However, it has been argued from the magnetoresistance behavior that also intergrain tunneling plays a role (as shown in Chapter 4 for instance see Figs. 4.7 - 4.9), and the intergrain coupling may well still change at higher field than where the magnetization loop has closed.
There is another way to gauge the strength of the junction. According to diffusive theory, Ic for a long S-N-S junction (N a normal metal) is propor-
Figure 5.5: Critical current Ic versus temperature T for devices AS (°), BS (4), CS (¤). The inset shows lnIc - 3/2 lnT versus T1/2. Dashed lines correspond to a Thouless energy of 72 (91) µV for device AS (BS and CS).
tional to T3/2exp (p(2πkBT )/ET h), with ET hthe Thouless energy given by (~D)/L2, D the diffusion constant of the N metal and L the junction length [60, 61]. Plotting ln(Ic) - 3/2 ln(T) versus√
T (inset of Fig. 5.5) shows that the relation holds well at low temperatures, with values for ET hof 72 (91) µV for device AS (BS,CS). This in turn can be used to estimate the maximum critical current from the relation eIcRN = 10.8 ET h [61]. The normal resis- tance RN of the junction is 10 Ω, which would yield a value for Ic of 75 µA.
This compares well to the measurements, but a problem is that the measured RN is much larger than expected for the CrO2bridge. Using a typical specific resistance, measured in various films, of 10 µΩcm, we rather estimate the nor- mal resistance of the junction to be 4 mΩ. This points to a low transparency T of the S/F barrier, which would correct the prefactor of ET hroughly with T and yield an estimate for the theoretical IcRN lower than the measured value.
This issue requires further study.
Overall, the numbers suggest in several ways that the junction critical cur- rents are smaller than what can in principle be obtained. On the other hand, in our working devices the current densities are large enough to conclude that the effect is intrinsic, rather than carried by filamentary normal metal shorts
Figure 5.6: Dependence of the critical current Ic on a magnetic field applied in the plane of the junction, perpendicular to the long axis of the electrodes (H||Ic). Closed circles denote the forward sweep (FS) from zero field to high position field, filled circles denote the backward sweep (BS).
in the ferromagnetic matrix, for which also otherwise no signs exist. Our premise is that the supercurrent is of triplet nature, and a difficulty lies in the preparation of the ’spin-active’ interface, which should both provide the differ- ence in spin scattering and unaligned magnetic moments [26]. Experimentally, the CrO2 film surface is sensitive to oxidation and has to be cleaned before the superconducting electrodes are deposited. The Ar-etching will not only remove unwanted oxides, but may also damage the surface in such a way that the required scattering or magnetization disorder is not present. Especially the fact that device BS, which consists of three parallel electrodes rather than one, does not show a larger Ic, strongly suggests that the triplet generation takes place at isolated spots under the electrodes rather than homogeneously over their width. Another hindrance is the finite lifetime of the devices, which is probably due to the grainy nature of the films and thermal expansion dif- ferences between film and substrate. These may not be the only bottlenecks, however. One common factor between the earlier experiments using TiO2and the present ones with sapphire is that the films have more than one easy axis of magnetization. In the case of sapphire, this is due to the strong polycrys-
talline nature of the growth, particularly evident in Fig. 3.6. In the case of TiO2, it was due to the peculiar circumstance that strain relaxation in the film can lead to a change in the easy-axis direction, with biaxial behavior occurring around a film thickness of 100 nm [35].
Comparative analysis; TiO2 versus sapphire
The next point to address is the behavior of devices on different substrates.
A simple first question is whether the interfaces are always sufficiently clean to make good superconducting contacts. It might be expected that, because of the difference in crystalline disorder, this is a bigger problem for sapphire- based devices than for TiO2-based devices and therefore not relevant to the observations. Still, what the experiments do show is the behavior which ap- pears to be connected with clean contacts. For both substrates, the resistance of the S/F/S devices jumps up when the surfaces are cleaned, including for the devices where a supercurrent is detected. The behavior of the interface therefore does not appear to be an essential difference between the two types of films. We believe the difference rather has to do with the basic physics gov- erning the Josephson current. As was discussed in Chapter 2, the condition for the Josephson current is that inhomogeneous magnetization exists at both superconducting contacts in order for the supercurrent to be generated and removed again [24, 62, 63].
Going back to the (FL,F,FR) model of Ref. [24], it was shown that no triplet supercurrent can flow when either FL and F, or F and FR have their magnetization parallel. We now assume that in our experimental system the inhomogeneous magnetization is not present as disordered moments at the interface, but actually resides in different grains having different directions of their magnetization. We schematically depict such a situation in Fig. 5.7a, where we draw two superconducting contacts on top of a ferromagnetic film.
The part of the film underneath the left (right) contact we call FL(R), the part in the middle is called F. For films on sapphire, crystallites with different magnetization can meet regularly, and the same kind of inhomogeneity is present everywhere, in particular under the superconducting contacts. This invokes the right conditions for a supercurrent. For films on TiO2the situation is very different. If the films are uniaxial, no inhomogeneity actually exists and no supercurrent can flow. If the films are biaxial, or if the easy axes are randomly distributed over the b- and c-axis, inhomogeneity can exist (see Fig. 5.7b), but if the long axis of a grain is of the order of the gap size, similar conditions on both ends may still not be present. The biaxial magnetic nature of the film is then very important, but as we noticed, also very sensitive to the treatment of the substrate surface before growth. The right condition
Figure 5.7: Schematic of the structure of grains and magnetization directions which can exist under the superconducting contacts (boxes marked S) in films grown (a) on sapphire and (b) on TiO2. FL, F and FR denote the three ferro- magnets (or parts of the ferromagnetic film) which should have non-collinear magnetizations in the spirit of the S/FL/F/FR/S model proposed by Houzet and Buzdin [24]. In (b) the drawn arrows picture grains with uniaxial magne- tization which are therefore collinear. The dashed arrows picture grains with biaxial magnetization, which can be non-collinear.
may not always come about at the same film thickness. We have to note in this respect that we have not yet performed experiments on films with the exact characteristics as described in Ref. [11]. A question which still has to be answered is the consequence of this model for the magnetic field dependence of Ic. When saturation of the magnetic moment is reached, around 50 mT, most of the magnetic moments have become aligned, and the supercurrent should vanish. Instead Icbasically retains its value up to at least 0.5 T [13], indicating that the magnetic moments involved in triplet generation are strongly pinned.
Another point to mention is the finite lifetime of the superconducting de- vices, which stop to function after cooling down a few times. This is probably due to the grainy nature of the films and thermal expansion differences between film and substrate. All in all, the interest in CrO2for studying very long range proximity effects is undiminished, but the materials science problems posed by the material are serious. Different routes involving the engineering of inhomo- geneous magnetization, rather than relying on its fortuitous presence might be a better way to control the generation of spin triplet Cooper pairs. Next, we therefore explore the idea of generating artificial magnetic inhomogeneity
with multilayer ferromagnet as was used to induce a long range proximity in Co based junctions [15, 16, 56].
5.3.4 Results; N/F/S contacts
Again devices are fabricated as described above. This time additional lay- ers of Ni(2 nm)/Cu(5 nm) were deposited on CrO2 thin films, prior to the deposition of a-MoGe. The Cu layer is being used to decouple both the fer- romagnetic layers. A supercurrent was measured successfully in three devices out of five. Here we describe the characteristics of three samples with two junctions prepared on each, named AT, BT and CT. Both junctions were showing a supercurrent on sample AT, fabricated with 30 µm width of the superconducting leads and a gap of 600 nm and 800 nm, termed AT-a and AT-b respectively. Samples BT and CT were prepared with 5 µm width and a gap of 700 nm and here only one junction was showing a measurable criti- cal current on each sample. The smaller width here was chosen to lower the absolute value of the critical current.
For these multilayer samples, R(T ) behaves differently from what we en- countered before. They show down jumps or only small up-jumps at Tc. On sample AT, both junctions were showing a supercurrent. They only have a down jump at Tc, as shown in Fig. 5.8a. Figure 5.8b illustrates the up jump in the resistance at Tc for BT. For junction CT the behavior is similar but with a huge jump to 12 Ω like sapphire based junctions. First, there is a small dip at 6 K, followed by a jump up with further decrease in the temperature.
The resistance remains flat between 6 to 5.8 K and then starts to decrease slowly. The normal resistance of these junctions is much lower than sapphire based junction because of better quality films on TiO2substrates with a lower density of grain boundaries.
Figure 5.9-a shows the I-V characteristics for sample AT-b. There is a zero resistance branch up to 3.4 mA followed by a bend to some non-zero resistance until the main transition occurs at 15 mA. There is a clear transition to the normal state with RN=100 mΩ, which is value of R(T ) just above Tc. It suggests this transition is due to the superconducting leads. The bending of the IV curve for currents higher than 3.4 mA might be connected to the higher applied currents, junction geometry and/or heating effects. Then: We take a voltage criterion of 1 µV to determine Ic On the other hand the I- V’s in Fig. 5.9-b for CT illustrate the presence of a supercurrent branch with critical current of the order of 1 mA. There is no bending prior to the transition to normal state, which indicates good behavior of junctions with lower critical currents. Hysteretic behavior is also very clearly present, which can be considered as an intrinsic property of a Josephson junction.
Figure 5.8: Temperature dependent resistance (a) for junctions AT-a and AT-b with 600 nm and 800 nm gaps. (b) For junction BT.
Ic(T ) was measured for both junctions BT and CT in the temperature range of 2.5 K to 6 K. It shows an almost linear dependence on temperature as illustrated in Fig. 5.10 for sample BT. A Thouless energy analysis can still be attempted at lower temperature which is of the order of Eth = 54 µeV (see the inset of Fig. 5.10). We did not measure the Ic(T ) for the larger junctions because the supercurrents disappeared after the third cooling and an intermediate inspection with an electron microscope.
Figure 5.11a illustrates the effect of a magnetic field on the Ic for both larger junctions AT-a,b, with the field applied in-plane but perpendicular to the current. It shows that Ic is quite sensitive to the applied field, with max- imum suppression happening below 50 mT with a sharp decrease at 30 mT.
It is interesting to note that a peak is appeared for both AT a,b junctions.
The field dependence of Ic for junctions BT is presented in Fig. 5.11b for all three configurations of the applied field, parallel to the junction, perpendicular to the junction and out of plane (perpendicular). Neither for the bigger nor for the smaller junctions there is the evidence for the presence of Fraunhofer patterns. For parallel configuration, a Fraunhofer pattern cannot be expected but with this configuration the effect of the alignment of the magnetizations of both magnetic layers (Ni and CrO2) can be seen. A 100 nm thick CrO2
film and a 2 nm thick Ni layer can be saturated with 200 mT, which should yield maximum suppression of Ic. It is obvious that Ic is more sensitive to the field for the bigger junctions AT-a and AT-b compared to the small junction BT. There is about 60% decrease in Ic for AT-a and about 85% decrease for AT-b at 200 mT. On the other hand, for junction BT there is a only about
Figure 5.9: Current (I) versus Voltage (V) measured at 4.2 K for SFS devices fabricated on CrO2/TiO2 with a 2 nm thick Ni layer. (a) For junction AT-b (width 30 µm, gap 800 nm). Zero voltage current branch starts to bend at current about 3.4 mA. Inset shows the low voltage curve. (b) For junction CT a zero voltage supercurrent branch is very clear for a junction with width is only 5 µm and gap of 600 nm (CT). This measurements was performed at Twente University.
20% suppression of Ic for the in-plane perpendicular configuration at 200 mT.
5.3.5 Discussion
The claim from the measurements on the multilayer samples is that very large critical currents are now flowing through the CrO2bridge. In discussing these results we address a number of important issues. We compare the residual re- sistance of the CrO2bridge in the supercurrent measurements with the normal state resistance of the bridge; we discuss the possibility of depairing currents in the superconducting leads in comparison with the measured critical current densities of the junctions; a Thouless analysis is also performed to gauge the size of the allowed current densities in accordance with the dimensions of the junctions; and we discuss the effects of applying a magnetic field.
The Ic measured in the multilayer junctions can be compared with the Ic of sapphire based junctions. The current density at 4.2 K, (dCrO2 ≈ 100 nm, junction width 30 µm and 5 µm, current ≈ 3.4 mA and 0.5 mA respectively) is of the order of 1.1 × 109 A/m2 for AT and 1 × 109 A/m2 for both BT and CT. In all these cases, it is 100 times larger than that of sapphire based
Figure 5.10: Critical current Ic as function of temperature for junction BT. Inset; fit with the Thouless energy expression yielding Eth = 54 µeV. Open symbol : Ic for junction CT at 4.2 K
junctions, while it is in good comparison with earlier observations of Keizer et al. [11]. This suggests that a uniform spin active interface is present at the interface, due to the additional layer of Ni (2 nm) and it makes these multilayer junctions more promising to generate a long ranged supercurrent.
An important question is whether the I-V characteristics such as shown in Fig. 5.9 are truly from the CrO2 films. For this we take another look at the normal resistance of the bridge. Taking ρ = 10 µΩcm, a film thickness of 100 nm, bridge width 5 µ, junction length 700 nm, RN comes out to be 140 mΩ (25 mΩ for the 30 µm wide contacts). This is significantly higher than what is measured: the zero-bias residual resistance in the superconducting state is not more than a few mΩ, which is the sensitivity of the voltmeter. Note that the measured resistance in the normal state is higher than the above estimate. This is because, when the superconducting leads become normal, the geometry of the sample is a very different one, with both high resistance MoGe and low-resistance CrO2contributing.
One another issue should be addressed here, namely whether the large Ic now measured in multilayer junctions actually is the depairing current Idp of the supercurrenting leads. For the sapphire based junctions with their low Ic values this was not relevant. The value for Jdp of a-MoGe superconductor at 4.2 K is about 3 × 1010A/m2[57]. Taking into account that the thickness of
Figure 5.11: Critical current as function of applied magnetic field (a) for bigger junctions AT-a (¤) and AT-b (°), in-plane perpendicular configuration, at 4.2 K. (b) for small junction BT in all three possible configurations of applied magnetic field, parallel configuration (¤), in-plane perpendicular (°), and out of plane perpendicular (4), at 3 K. The vertical dotted lines are indicating the suppression of Ic with in the window of 200 mT
the lead (40 nm) is smaller that the thickness of the bridge, the current density flowing through the lead is still an order of magnitude smaller than Jdp, which is more than an order of magnitude larger than the critical current density for multilayer junctions, which makes it clear that the measured current can be the supercurrent rather than the depairing current.
From Ic(T ), the measured Thouless energy is around Eth = 50 µeV, not much different than that of sapphire based junctions. Estimating the maxi- mum Ic at zero temperature from the relation eIcRN = 10.8ET h, with RN ≈ 90 mΩ (calculated from the resistivity data because of a up jump in R(T ) at Tc) we find 6 mA, substantially higher than the measured value of 0.5 mA at 4.2 K. However, at 4.2 K, where kBT/ETh is about 7, the ration eIcRN/Eth is only 1 [61], which points to a quite reasonable agreement. We also cal- culated the diffusion coefficient D using the relation D = ET h~L2. It yields D ≈ 200 cm2/s. Calculating the same with the density of states relation e2ρoDN = 1, results in D = 90 cm2/s. Both numbers are not much different from the expected values in the range of 50-200 cm2/s.
For the multilayer junctions the magnetic field effects are complicated. In particular, for the in-plane perpendicular configuration, the junctions AT-a
and AT-b are more sensitive to the field than the BT junction. For junctions AT-a and AT-b, the first sharp decrease at 30 mT with a peak like behavior, it might be corresponding to the first flux quanta, which is a quite reasonable value according to the dimensions of the junctions. Unfortunately, there are no clear signatures of the Fraunhofer pattern for both cases. In these junctions, Ic immediately starts to decrease instead of increasing to a peak as in the sapphire based junctions or in Keizer’s measurements [11]. The additional ferromagnetic layer and the geometry of the junctions as full ferromagnetic films might be the reason of unclear field effects.
Another notable point is the non-zero critical current even at 500 mT, which should be high enough to saturate both the CrO2 and Ni [64]. The magnetization direction of Ni or CrO2 should be aligned. That makes the situation of parallel configuration with alignment of magnetizations of left (Ni left) and right (Ni right) with central ferromagnetic layer (CrO2). It suggests that there is a residual magnetic inhomogeneity residing in the CrO2/Cu/Ni sandwich, which is not removed by the magnetic field; or the spin depen- dent magnetic scattering from the two magnetic layers is in itself sufficient to generate triplets.
To understand the behavior of the junctions with and without a possible short we also did some controll experiments, in particular to measure the R(T ).
The results of those experiments are discussed in Appendix C.
5.4 Conclusions
We have provided new evidence that a supercurrent can flow through a half metallic ferromagnet CrO2 film, deposited on a sapphire substrate, over a length of the order of 1 µm. The odd-frequency pairing scenario appears a plausible one, both from the critical current values and from the magnetic field dependence. The analysis shows that the films deposited on sapphire exhibit six fold rotational anisotropy that might provide the required magnetic inho- mogeneity to generate the spin triplet supercurrent. The reproducibility is still low, which suggests a weak control over the preparation of the spin active interface. Such control over the triplet supercurrent generation appears better with an artificially created spin active interface between superconductor and CrO2 film deposited on a TiO2 substrate. It can be done by adding an ad- ditional ferromagnetic layer like 2 nm thick Ni in our case. The supercurrent generate in such a way is in good agreement with previous results. The criti- cal current density of these devices with Ni suggests the presence of a uniform spin active interface. The magnetic field effect on critical current reveals that the presence of non-collinear magnetization might not be the only reason of
triplet supercurrent generation because there is still some non-zero current in the presence of higher fields of the order of 500 mT, which is enough to saturate both the magnetic layers. It also suggests that there are other scat- tering phenomena (different spin dependent magnetic scattering from different magnetic layers) or magnetic moment disorder at the interface (roughness) in- volved in this multilayer magnetic arrangement to generate the odd-frequency spin triplet supercurrent.
