doi: 10.1038/nphys4110ย ยท we find ฮพ~590 nm, i.e. ๐ฟ๐ฟ/ฮพ~0.65. note that in contrast to the...
TRANSCRIPT
In the format provided by the authors and unedited.
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Supplementary Information for โTunnelling Spectroscopy of Andreev States in Grapheneโ
Landry Bretheau1โ *, Joel I-Jan Wang1โ , Riccardo Pisoni1,2, Kenji Watanabe3, Takashi Taniguchi3, Pablo Jarillo-Herrero1* 1 Department of Physics, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, United States 2 Politecnico di Milano, Piazza Leonardo da Vinci, 32, 20133 Milano, Italy 3 National Institute for Materials Science, Namiki 1-1, Tsukuba, Ibaraki 305-0044, Japan โ These authors contributed equally to this work. *e-mail: [email protected]; [email protected]; [email protected] Content
1. Device Fabrication 2. Measurement Setup 3. Scattering Characterization via Transport Measurements 4. Grapheneโs Density of States in the Normal Regime 5. Potentials effects of the Graphite Tunnel Probe 6. Grapheneโs Proximitized DOS in the Superconducting Regime 7. Supercurrent Spectral Density 8. Data on additional devices
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS4110
NATURE PHYSICS | www.nature.com/naturephysics 1
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
1. Device Fabrication The experiment was performed using a van der Waals heterostructure schematized in
Fig.S1a. From top to bottom, it consists of a thin and narrow graphite tunnelling probe (width = 150 nm), an hBN monolayer, a graphene monolayer, an hBN bottom layer (thickness =15 nm), and a graphite bottom gate. To assemble this stack with various thin films, we apply a polymer-based dry pick-up and transfer approach1,2 using a polycarbonate (PC) film3 (Fluka Analytical, part # 181641) to achieve higher efficiency in picking up monolayer graphene or hBN from the SiO2 substrate.
To start with, pre-patterned graphite probes on SiO2 substrate are picked-up and transferred onto a hBN monolayer previously identified with Raman spectroscopy4. The stack is then shaped into isolated rectangles, each of which defines the length L of a graphene weak link. To pattern the thin films, we perform electron-beam lithography using Poly(methyl methacrylate) (PMMA 950A5 from Microchem) as both the resist and etching mask (thickness ~ 250 nm). After lithography, the graphite is etched by reactive ion etching in oxygen (15 sccm), whereas the hBN is etched in an oxygen (5 sccm), argon (10 sccm), and CHF3 (10 sccm) environment.
We then employ another PC film to successively pick up the probe/tunnelling barrier stack, a monolayer graphene flake, and an hBN bottom layer. All of this is eventually transferred onto a graphite flake deposited on the SiO2 substrate (285 nm thermal oxide on p-doped Si substrate from NOVA electronics). After dissolving the PC film in chloroform, the heterostructure is annealed in forming gas (Ar/H2) at 350 ยฐC for at least 3 hours in order to remove organic residue and to reduce the area with bubbles. Fig.S1b shows an AFM micrograph of the assembled stack.
At last, we perform electron-beam lithography and we deposit 7 nm of Ti and 70 nm of Al by thermal evaporation. After lift-off in acetone, the device is ready for measurement (see Fig.S1c).
2. Measurement Setup Figure S2 shows the schematics of our measurement setup. The sample is anchored
to the mixing chamber of a dilution refrigerator at 20 mK, well below the transition temperature of aluminium (TC ~ 1.1K). All DC lines are heavily filtered using both discrete RC filters and distributed low-pass copper tape filters5, which are placed at the cryogenic level before reaching the device. ๐๐๐ผ๐ผ/๐๐๐๐ measurements are performed at low frequency (10-100 Hz) using room-temperature amplification and standard lock-in techniques with an excitation voltage of 5-10 ยตV.
3. Scattering Characterization via Transport Measurements To estimate the scattering properties of graphene in the normal regime, itโs useful to
measure in transport a graphene-based junction. However, since the superconducting electrode shunts the graphene flake within a SQUID loop, one cannot directly measure in transport the graphene probed by tunneling spectroscopy. Instead, we measure a graphene junction between two neighboring loops on a device that is fabricated using the same procedure as the device presented in the main text.
Figure S3a shows the normal state resistance ๐ ๐ ! of this graphene junction (๐ฟ๐ฟ ร ๐๐ = 350 ๐๐๐๐ ร 1 ๐๐๐๐) as a function of backgate voltage. A lower bound of the
NATURE PHYSICS | www.nature.com/naturephysics 2
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
mean free path, ๐๐!, can be estimated using the formula6 ๐๐! = ๐๐โ/(2๐๐!๐๐!), where โ is Planckโs constant, ๐๐ the electron charge and ๐๐ = ๐ฟ๐ฟ/(๐๐๐ ๐ !) the conductivity extracted
from the measurement. The Fermi wave vector ๐๐! = ๐๐๐๐! ๐๐! /๐๐ depends on the
backgate voltage ๐๐!, with ๐๐! โ 2.3 ๐๐๐๐/๐๐๐๐! the gate capacitance per unit square. The extracted mean free path is plotted in Fig.S3b. In the n-doped region (๐๐! > 0), one gets ๐๐! ~ 140 ๐๐๐๐~ ๐ฟ๐ฟ/2.7, which suggests that the junction is neither ballistic nor diffusive but in an intermediate regime. Note that the extracted ๐๐! is dramatically suppressed in the p-doped region. This is due to p-n junctions that form at the metal/graphene interface and does not reflect the intrinsic quality of the graphene. To infer the superconducting length ฮพ, we use the formula for the (worst case) diffusive relation ฮพ = โ๐ท๐ท ฮ, with ๐ท๐ท =๐ฃ๐ฃ!๐๐! 2 the Einstein diffusion coefficient in 2D. We find ฮพ~590 nm, i.e. ๐ฟ๐ฟ/ฮพ~0.65.
Note that in contrast to the graphene flake within the loop, which is encapsulated by a monolayer hBN and the graphite probe, the graphene between two loops is not covered with any other material. This means that this junction is more likely to be contaminated by residue from nanofabrication, and the extracted transport properties such as mean free path or coherence length should be regarded as a lower bound for the actual weak link where tunneling spectroscopy is performed. Based on this estimate, the graphene-based Josephson junction should rather be in the short junction limit (but not infinitely short), and each conduction channel should contain 1 or 2 pairs of ABS7,8.
4. Grapheneโs Density of States in the Normal Regime
Grapheneโs normal state DOS is shown as a function of the backgate voltage in Fig. S4a. The overall V-shape curve reflects the DOS corresponding to the linear Dirac cone dispersion, with a minimum at the CNP ๐๐!"# ~ 0.05 ๐๐, thus demonstrating the weak doping induced by the graphite probe. On top of it, one can see sharp resonances, which disperse as a function of both energy and backgate voltage in the manner of Coulomb diamonds9 (see Fig.S4b). They are probably associated with spurious quantum dots at the interfaces with hBN. A detailed analysis suggests that they correspond to 10-20 nm size features with typical addition energy of ~5-15 meV. This energy scale is huge compared to the superconducting gap that appears as a narrow dip at zero energy, and these quantum dots features have little interplay with the superconducting proximity effect and do not affect our measurements.
5. Potential effects of the Graphite Tunnel Probe
The question of the invasiveness of the probe is crucial in any tunnelling spectroscopy experiment. First of all, letโs mention that prior to this experiment, we performed a similar experiment using a gold tunnelling electrode in place of the graphite one. The charge doping of the graphene by the gold electrode was so big that we could not explore regions in chemical potential close to the charge neutrality point. We therefore end up using graphite as a tunnel probe, as it limits the doping in the graphene (directly underneath the monolayer hBN barrier), owing to the small work function difference between graphite and graphene. We then performed a calibration experiment in which we used a graphite electrode and an hBN tunnel barrier to measure the DOS of bulk aluminium. The measured spectra were consistent with a BCS DOS for Al, which
NATURE PHYSICS | www.nature.com/naturephysics 3
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
allowed us to validate this experimental scheme and demonstrate that the graphite DOS is constant at the scale of aluminiumโs superconducting gap.
Second, itโs essential that the coupling rate between the graphite probe and the graphene flake is small enough so that it is in the tunnelling regime and that the relevant energy features are not dramatically broadened. Although itโs hard to quantitatively extract the tunnel rate from the measurements, one can estimate it via the magnitude of the measured differential conductance in the normal regime. Indeed, extending the Breit-Wigner formula to the multi-channel case, the conductance reads:
๐บ๐บ(๐๐) =2๐๐!
โ4ฮ!"ฮ!"
(๐๐๐๐ โ ๐๐!)! + (ฮ!" + ฮ!")!
!
!!!
where ๐๐! is the energy level of the conduction channel m, ฮ!" and ฮ!" the coupling rates to the tunnel and superconducting probes, and ๐๐ the number of conduction channels. Therefore:
๐บ๐บ!"# =2๐๐!
โ4ฮ!"/ฮ!"
(1+ ฮ!"/ฮ!")!
!
!!!
~2๐๐!
โ 4๐๐ฮ!/ฮ!
(1+ ฮ!/ฮ!)!
Performing a good tunnelling spectroscopy requires both ฮ! โช ฮ! and ฮ! โช ๐ฟ๐ฟ๐ฟ๐ฟ, where ๐ฟ๐ฟ๐ฟ๐ฟ~ฮ~170 ยตeV is the relevant energy scale of the experiment. In our case, at ๐๐! = 1๐๐, which corresponds to ๐๐~100 , we typically measure ๐บ๐บ~4 ยต๐๐. Therefore ฮ!/ฮ!~3. 10!! โช 1 as required in the tunnelling regime. ฮ! can be estimated as ๐๐ โ !!
!~ ๐๐ โ 1.73 ๐๐๐๐๐๐, with ๐๐ < 1 the average contact transparency. Then itโs clear that
ฮ! < 3.10!! โ 1.73 ๐๐๐๐๐๐~ 0.5 ยต๐๐๐๐ โช ฮ and we can safely assume that we are performing spectroscopy with a weakly invasive probe.
6. Grapheneโs Proximitized DOS in the Superconducting Regime
We now switch to additional measurements in the superconducting regime. - Figure S6 shows the DOS in the superconducting state, as a function of both energy
and phase, for various values of the gate voltage. The energy spectra depend strongly on the graphene carrier density, which is tuned from hole-type through the CNP to electron-type. As already mentioned in the main text, the larger the carrier density, the stronger the DOS modulates with phase. On top of that, some spectra (a, d, w, x) are dramatically different. There, the differential conductance is modified due to neighbouring quantum dot resonances that create voltage-dependent backgrounds.
- To get rid of this effect, we subtract a bias voltage dependent background defined as the average DOS over one period in magnetic field. The corresponding background-subtracted spectra are shown in Fig. S7. After subtraction, all spectra share similar features as they display a checkerboard pattern that weakly depends on the carrier density.
- At very large carrier density ๐๐ > 3ร10!"/๐๐๐๐! , one can measure a complete closing of the induced gap and a flat DOS at ๐๐ = ๐๐ (Fig. S4 b,u,v). This is emphasized in Fig. S8, which shows the DOS at a gate voltage ๐๐! = 2.4 ๐๐. There, disorder is negligible and more ballistic ABS reach zero energy at ๐๐ = ๐๐.
NATURE PHYSICS | www.nature.com/naturephysics 4
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
- Fig. S9 compares energy spectra of opposite carrier density. When the graphene is hole-doped, the phase modulation of the DOS is smaller with a V-shaped induced gap, whereas it is U-shaped with a stronger phase modulation in the case of electronic-type carrier density. We attribute this to p-n junctions that form at the graphene/aluminium interface and reduce the contactsโ transparency, thus repelling the ABS from the gap edges toward lower energy.
7. Supercurrent Spectral Density
In the case of a pair of discrete ABS at energies ยฑ๐ธ๐ธ!, the DOS reads: ๐ท๐ท๐ท๐ท๐ท๐ท ๐ธ๐ธ,๐๐ =๐ฟ๐ฟ ๐ธ๐ธ + ๐ธ๐ธ!(๐๐) + ๐ฟ๐ฟ ๐ธ๐ธ โ ๐ธ๐ธ!(๐๐) , where ๐ฟ๐ฟ is the Dirac distribution and ๐ธ๐ธ! > 0 is the Andreev excitation energy that depends on the phase ๐๐. The energy in the ground state reads ๐ธ๐ธ!"(๐๐) = โ๐ธ๐ธ!(๐๐) =
!!
๐ ๐ ๐ ๐ ๐ ๐ (โ๐ธ๐ธ)๐ท๐ท๐ท๐ท๐ท๐ท ๐ธ๐ธ,๐๐ ๐๐๐๐!!! . Following Samuelsson and
co-authors8, the supercurrent spectral density is defined as ๐ฝ๐ฝ! ๐ธ๐ธ,๐๐ = !!!
!!!!"
(โ๐ฟ๐ฟ ๐ธ๐ธ +๐ธ๐ธ! + ๐ฟ๐ฟ ๐ธ๐ธ โ ๐ธ๐ธ! ), where ๐๐! = โ/2๐๐ is the reduced flux quantum. It quantifies the amount of supercurrent carried by Andreev states at energy ๐ธ๐ธ and phase ๐๐. Therefore, one can see that ๐ฝ๐ฝ! ๐ธ๐ธ,๐๐ = โ 1
!!
๐๐๐๐๐๐๐๐๐๐๐๐
!! ๐๐,๐๐ ๐๐๐๐, the integration boundaries satisfying
๐ฝ๐ฝ! 0,๐๐ = 0. Using this definition, the supercurrent in the ground state reads: ๐ผ๐ผ! ๐๐ =1!
sgn(โ๐ธ๐ธ)๐ฝ๐ฝ! ๐ธ๐ธ,๐๐!!! ๐๐๐๐ = โ 1
!!
!!!!"
. We extend these relations to the case of continuous DOS, which enables one to
extract the supercurrent spectral density and the ground state supercurrent from a measured DOS. ๐ฝ๐ฝ! is a basic quantity naturally expressed in units of ๐๐/โ and that makes evident the link between Andreev physics and Josephson effect. To get proper units, the tunnelling conductance is converted into a DOS using the proportionality factor ๐๐ = 34.72 ๐๐๐๐V!!๐๐S!!. We obtained this factor by fitting the V-shaped gate-dependent tunnelling conductance from Fig.S4a to the ideal DOS of graphene in the normal regime
๐ท๐ท๐ท๐ท๐ท๐ท(๐ธ๐ธ!) =! ! !
! (โ!!)!๐ธ๐ธ! , where the Fermi energy reads ๐ธ๐ธ! = ๐ ๐ ๐ ๐ ๐ ๐ ๐๐! โ๐ฃ๐ฃ! ๐๐ !! !!
!. In
the main text, Fig. 4 shows both the Andreev spectrum in graphene and its corresponding supercurrent spectral density and supercurrent-phase relation for a gate voltage ๐๐! = 2.4 ๐๐.
Fig. S10 shows the experimental Fourier transform of the supercurrent spectral density, as a function of both energy ๐ธ๐ธ = ๐๐๐๐ and inverse phase 2๐๐๐๐!!. On top of the large 1st harmonic, one can see the 2nd and 3rd harmonics. This demonstrates the anharmonic phase-dependence of ๐ฝ๐ฝ!.
Fig. S11a shows a zoom-in of the Josephson supercurrent-phase relation around zero phase. This is well modelled by the theoretical supercurrent carried by perfectly coupled short ABS with an average transmission coefficient ๐๐ = 0.725 (solid red curve):
๐ผ๐ผ!! = ๐ผ๐ผ!๐๐ sin(๐๐)
4 1โ ๐๐ sin!(๐๐/2)
with ๐ผ๐ผ! = 258 nA. Fig. S11b shows the corresponding Fourier series sin coefficients as a function of the inverse phase 2๐๐๐๐!!. Higher harmonic components can be seen, thus demonstrating that the current-phase relation is strongly anharmonic and that the supercurrent is carried by ABS from well-coupled and well-transmitted short channels.
NATURE PHYSICS | www.nature.com/naturephysics 5
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
8. Data on additional devices For sake of clarity, all measurements presented in the main text were made on the
same particular device (labelled #A), whose SGS junction dimensions are ๐ฟ๐ฟร๐๐ =380 ๐๐๐๐ ร2 ๐๐๐๐. However, we have performed similar measurements for other devices with different dimensions and SQUID loop size, which can be seen in Fig.S1c. In addition to device #A, two other devices were well connected with tunnel resistance in a proper range for spectroscopy measurements: device #B, with dimensions ๐ฟ๐ฟร๐๐=380๐๐๐๐ร600๐๐๐๐ and device #C, with dimensions ๐ฟ๐ฟร๐๐=400๐๐๐๐ร1100๐๐๐๐.
The measured energy spectra are shown in Fig. S12 and S13, for few values of the gate voltage. The magnitude of the measured differential conductance scales with the junctionsโ dimensions, the smallest device (#B) exhibiting conductance as low as 0.1 ยตS close to the charge neutrality point. The observed flux modulations are in good agreement with the designed SQUID loop sizes. These spectra are qualitatively similar to the ones reported in the main text. Although there is no clear influence of the graphene junction width, this shows the reproducibility of our measurements.
In future studies, it would be interesting to vary more broadly the graphene junction length to investigate the crossover between the short and the long junction regime.
NATURE PHYSICS | www.nature.com/naturephysics 6
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Supplementary references 1. Wang, J. I. et al. Electronic Transport of Encapsulated Graphene and WSe 2
Devices Fabricated by Pick-up of Prepatterned hBN. Nano Lett. 15, 1898-1903 (2015).
2. Wang, L. et al. One-dimensional electrical contact to a two-dimensional material. Science 342, 614โ617 (2013).
3. Zomer, P. J., Guimarรฃes, M. H. D., Brant, J. C., Tombros, N. & van Wees, B. J. Fast pick up technique for high quality heterostructures of bilayer graphene and hexagonal boron nitride. Appl. Phys. Lett. 105, 013101 (2014).
4. Gorbachev, R. V et al. Hunting for monolayer boron nitride: optical and Raman signatures. Small 7, 465โ468 (2011).
5. Spietz, L., Teufel, J. & Schoelkopf, R. J. A Twisted Pair Cryogenic Filter. arXiv:cond-mat/0601316, 1-12 (2006).
6. Du, X., Skachko, I. & Andrei, E. Y. Josephson current and multiple Andreev reflections in graphene SNS junctions. Phys. Rev. B 77, 184507 (2008).
7. Wendin, G. & Shumeiko, V. S. Josephson transport in complex mesoscopic structures. Superlatt. and Microstruct. 20, 569โ573 (1996).
8. Samuelsson, P., Lantz, J., Shumeiko, V. S. & Wendin, G. Nonequilibrium Josephson current in ballistic multiterminal SNS junctions. Phys. Rev. B 62, 1319โ1337 (2000).
9. Amet, F. et al. Tunneling spectroscopy of graphene-boron-nitride heterostructures. Phys. Rev. B 85, 73405 (2012).
NATURE PHYSICS | www.nature.com/naturephysics 7
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Figure S1. Van der Waals Heterostructure. a, Device schematics. b, AFM image of the stack. c, Optical picture of the device after electron-beam lithography. An encapsulated graphene flake is connected to two superconducting electrodes. On top of it, an insulating barrier of hBN (monolayer) and a graphite electrode enable us to perform tunnelling spectroscopy. The graphite backgate controls electrostatically carrier density in graphene. Magnetic flux ๐๐ threading the superconducting loop imposes a phase ๐๐ = ๐๐/๐๐! across graphene.
NATURE PHYSICS | www.nature.com/naturephysics 8
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Figure S2. Measurement Setup. The sample is anchored to the mixing chamber of a dilution refrigerator at 20 mK. All DC lines are heavily filtered using both discrete RC filters and copper tape filters, which are thermally anchored at 20 mK. ๐๐๐ผ๐ผ/๐๐๐๐ measurements are performed at low frequency (10-100 Hz) using room-temperature amplification and standard lock-in techniques with an excitation voltage of 5-10 ยตV.
Figure S3. Scattering Characterization via Transport Measurements. a, Differential resistance as a function of backgate voltage in the normal regime. This measurement was performed on a similarly fabricated graphene junction between two neighboring SQUID loops. b, Lower bound of the mean free path as a function of backgate voltage. It was extracted using the procedure described in the text.
Vbg
RC filter
RC filter Cu tape filter
Cu tape filter
Current to voltage
converter
D/A converter
Lock-in amplifierSR-830
Stanford Research
MultimeterK2700
Keithley
SourcemeterK2400
Keithley
1/1K
1/10K
Vb+dVb dVb
VbFridge
I+dI
GPIB to computer
T~20 mK
-5 -2.5 0 2.5 5
50
100
150
200
Vg(V )
-5 -2.5 0 2.5 50
500
1500
2500
Vg(V )
a
RN(kโฆ)
1.5
0.5
2.5
0 2.5-2.5 5-50
0 2.5-2.5 5-5
200
150
100
50
b
l e(nm)
NATURE PHYSICS | www.nature.com/naturephysics 9
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Figure S4. Grapheneโs DOS in the Normal State. a, b Differential conductance, dI/dV as a function of gate voltage ๐๐! and energy ๐ธ๐ธ=eV. On the top x-axis, ๐๐! is converted into grapheneโs carrier density ๐๐. On the right y-axis of (a), dI/dV is converted into grapheneโs DOS. a, Horizontal line-cut of (b) (blue dashed-line) at energy ๐ธ๐ธ = 4 ๐๐๐๐๐๐. Inset: Blow-up around the charge neutrality point. The orange dotted lines show the gate voltage values at which the DOS in the superconducting regime is measured in Figs. S6, S7. (b), The differential conductance is gray-coded using a logarithmic scale.
NATURE PHYSICS | www.nature.com/naturephysics 10
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Figure S5. Gate dependence of the proximitized graphene DOS. a-i, Differential conductance ๐๐๐๐/dV as a function of energy ๐ธ๐ธ=eV for different magnetic fields, ๐ต๐ต, and at different gate voltages (indicated in each panel). They correspond to cross-sections at phases ๐๐/๐๐ = 0,0.2,0.4,0.6,0.8,1 of the color-coded data in Fig. 3.
NATURE PHYSICS | www.nature.com/naturephysics 11
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Figure S6. Gate dependence of the proximitized graphene DOS. a-y, Colour-coded DOS as a function of both energy ๐ธ๐ธ=eV and superconducting phase ๐๐, for different gate voltages (indicated in each panel). In each panel, the colour-coded DOS is linearly scaled to maximize the contrast (see Fig.S4 for quantitative values).
NATURE PHYSICS | www.nature.com/naturephysics 12
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Figure S7. Gate dependence of background-subtracted spectra. a-y, Colour-coded background-subtracted DOS as a function of both energy ๐ธ๐ธ=eV and superconducting phase ๐๐, for different gate voltages (indicated in each panel). In each panel, the colour-coded subtracted DOS is linearly scaled to maximize the contrast.
NATURE PHYSICS | www.nature.com/naturephysics 13
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Figure S8. Phase dependence of grapheneโs proximitized DOS at ๐๐! = 2.4 ๐๐. a, b, DOS as a function of energy ๐ธ๐ธ=eV and magnetic field (the top axis shows B converted into the superconducting phase difference ๐๐). The induced superconducting gap fully disappears at ๐๐ = ๐๐ as more ballistic ABS reach zero energy.
NATURE PHYSICS | www.nature.com/naturephysics 14
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Figure S9. Comparison of Andreev spectra in the electron and hole-doped regimes. DOS as a function of both energy ๐ธ๐ธ=eV and superconducting phase ๐๐, for different gate voltages (indicated in each panel) in the hole-doped regime (a-d, i) and electron-doped regime (e-h, j).
NATURE PHYSICS | www.nature.com/naturephysics 15
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Figure S10. Fourier analysis of the supercurrent spectral density. a, Colour-coded supercurrent spectral density ๐ฝ๐ฝ! as a function of both energy ๐ธ๐ธ=eV and superconducting phase difference ๐๐/2๐๐ , at a gate voltage ๐๐! = 2.4 ๐๐ . b, c Corresponding Fourier transform ๐ผ๐ผ๐ผ๐ผ โฑ(๐ฝ๐ฝ!) (absolute value of imaginary part of single-sided discrete Fourier transform) as a function of both energy ๐ธ๐ธ=eV and inverse phase 2๐๐๐๐!!. Cuts at few energy values are shown in (c). Higher harmonic components can be seen, thus demonstrating the anharmonic phase-dependence of ๐ฝ๐ฝ!.
NATURE PHYSICS | www.nature.com/naturephysics 16
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Figure S11. Anharmonic current-phase relation. a, Blue dots: Josephson supercurrent ๐ผ๐ผ! as a function of superconducting phase difference ๐๐/๐๐, at a gate voltage ๐๐! = 2.4 ๐๐ (zoom-in close to zero phase of Fig. 4c). The solid red line corresponds to the theoretical supercurrent carried by perfectly coupled short ABS with an average transmission coefficient ๐๐ = 0.725. b, Corresponding Fourier series sin coefficients ๐ผ๐ผ๐ผ๐ผ โฑ(๐ผ๐ผ!) as a function of inverse phase 2๐๐๐๐!! . Higher harmonic components can be seen, thus demonstrating that the current-phase relation is strongly anharmonic and that the supercurrent is carried by ABS from well-coupled and well-transmitted short channels.
NATURE PHYSICS | www.nature.com/naturephysics 17
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Figure S12. Phase dependence of grapheneโs proximitized DOS for device #B. a-f, Differential conductance as a function of energy ๐ธ๐ธ=eV and magnetic field (the top axis shows B converted into the superconducting phase difference ๐๐), for three different gate voltages (indicated in each row).
NATURE PHYSICS | www.nature.com/naturephysics 18
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS
ยฉ 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Figure S13. Phase dependence of grapheneโs proximitized DOS for device #C. a-b, Differential conductance as a function of energy ๐ธ๐ธ=eV and magnetic field (the top axis shows B converted into the superconducting phase difference ๐๐), for device #C at gate voltage ๐๐! = 3๐๐.
NATURE PHYSICS | www.nature.com/naturephysics 19
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS