Supplementary Materials for...The waveguide structure was finally patterned using photolithography...

Post on 28-Sep-2020

3 views 0 download

Transcript of Supplementary Materials for...The waveguide structure was finally patterned using photolithography...

www.sciencemag.org/cgi/content/full/science.aat8687/DC1

Supplementary Materials for

Quantum-critical conductivity of the Dirac fluid in graphene Patrick Gallagher, Chan-Shan Yang, Tairu Lyu, Fanglin Tian, Rai Kou, Hai Zhang,

Kenji Watanabe, Takashi Taniguchi, Feng Wang*

*Corresponding author. Email: fengwang76@berkeley.edu (F.W.)

Published 28 February 2019 on Science First Release DOI: 10.1126/science.aat8687

This PDF file includes: Materials and Methods Supplementary Text Figs. S1 to S17 References

1

Materials and Methods Sample fabrication Individual flakes of the graphene heterostructure were first prepared on oxidized silicon wafers by mechanical exfoliation from bulk crystals of graphite (NGS, graphit.de), hBN (Watanabe and Taniguchi), and WS2 (HQ Graphene). The heterostructure was assembled using a dry transfer technique (18) and subsequently deposited on a fused quartz substrate.

Photoconductive switches were separately prepared starting from a bulk wafer supplied by BATOP GmbH, which consisted of two epitaxial layers grown on a semi-insulating GaAs substrate. The top epitaxial layer was a 2.6 micron thick GaAs film grown at low temperature (300ยฐC) to achieve a nominal carrier recombination time of 0.5 ps. Beneath this layer was a 500 nm thick etch stop made of Al0.9Ga0.1As. The top layer was etched into squares (50 micron by 50 micron) and rectangles (200 micron by 50 micron) by defining an etch mask (S1818 photoresist, MicroChem) and subsequently etching in a citric acid solution (~10 minutes in 6 parts citric acid monohydrate:6 parts deionized water:1 part hydrogen peroxide). The etch stop layer was then dissolved in 10:1 buffered hydrofluoric acid (~6 hours) followed by a thorough rinse in deionized water and gentle drying with a low nitrogen flow. Etched GaAs squares and rectangles which remained loosely attached to the semi-insulating GaAs substrate were mechanically transferred to our quartz substrate using PDMS (Sylgard 184, Dow Corning) as an adhesive.

The waveguide structure was finally patterned using photolithography (LOR 3A and S1818 photoresists, MicroChem). Electrode metal (5 nm titanium, 200 nm gold) was deposited using an electron-beam evaporator. Figure S1 shows a large-area photograph of the completed device. Figure S2 shows the cross-section of the heterostructure beneath the waveguide. Terahertz measurements Laser pulses illuminating the emitter and detector were split off from the output of a mode-locked Ti:sapphire laser (Atseva) with center wavelength 800 nm, pulse width ~150 fs, repetition rate 80 MHz, and total output power 800 mW. Pulsed optical pump excitation was generated from the same Ti:sapphire output using a supercontinuum fiber (Newport SCG-800); the fiber output was filtered to transmit wavelengths between 1050 nm and 1500 nm to avoid exciting the WS2 gate electrode. All optical pump fluences reported are the fluences absorbed by the graphene sheet, but should be considered order-of-magnitude values, as geometrical factors and local field effects were not precisely measured. The probing electric field applied to graphene was kept below 10 mV/micron so that the energy accumulated between collisions would not exceed ๐‘˜๐‘˜B๐‘‡๐‘‡e, preserving approximate electron-hole symmetry (9) for our studies of the Dirac liquid. We verified that reducing the probing field by a factor of ~6 does not change our results near charge neutrality.

The lengths of the emitter and optical pump beam paths were adjusted using two mechanical delay stages. All data shown in the main text were collected by fixing the optical pump stage and adjusting the length of the emitter beam path to measure the time-domain waveform (for Fig. 2, the optical pump excitation was blocked). The length of the detector beam path was modulated at 50 Hz with ~50 micron amplitude using a mirror mounted on a piezoelectric shaker, while the optical pump beam was mechanically chopped at 225 Hz. The 50 Hz and 275 Hz components of the current signal were simultaneously collected to measure, respectively, the time derivative of the transmitted pulse ๐‘‘๐‘‘๐‘‘๐‘‘/๐‘‘๐‘‘๐‘‘๐‘‘ and its change ๐‘‘๐‘‘(๐›ฅ๐›ฅ๐‘‘๐‘‘)/๐‘‘๐‘‘๐‘‘๐‘‘ upon optical pump excitation. The 50 Hz and 275 Hz signals are related to ๐‘‘๐‘‘๐‘‘๐‘‘/๐‘‘๐‘‘๐‘‘๐‘‘ and ๐‘‘๐‘‘(๐›ฅ๐›ฅ๐‘‘๐‘‘)/๐‘‘๐‘‘๐‘‘๐‘‘ by

2

frequency-dependent calibration constants of order 1; these constants were carefully measured to avoid scaling errors in the extracted ๐›ฅ๐›ฅ๐œŽ๐œŽ. Since the photon energy of the optical pump is below the bandgap of WS2 and hBN, the ๐‘‘๐‘‘(๐›ฅ๐›ฅ๐‘‘๐‘‘)/๐‘‘๐‘‘๐‘‘๐‘‘ signal exclusively measures the conductivity change of graphene.

The sample was mounted in the vacuum space of an optical cryostat cooled to 77 K by a liquid nitrogen bath. The 300 K measurements shown in the inset of Fig. 2 of the main text were collected by allowing the nitrogen bath to empty and warm to room temperature. Curve-fitting All fits reported in this work were performed using LMFIT (41), an optimization package for Python. Error bars displayed in figures and written in the text indicate standard deviations determined by LMFIT.

Supplementary Text Determination of charge neutrality Since the Dirac fluid behavior under study should only appear at chemical potentials ๐œ‡๐œ‡ <๐‘˜๐‘˜B๐‘‡๐‘‡, we require the sample to be tuned quite precisely to charge neutrality. Fig. S4 shows the evolution with ๐‘‘๐‘‘gate of the measured current signals proportional to ๐‘‘๐‘‘๐‘‘๐‘‘/๐‘‘๐‘‘๐‘‘๐‘‘ and ๐‘‘๐‘‘(๐›ฅ๐›ฅ๐‘‘๐‘‘)/๐‘‘๐‘‘๐‘‘๐‘‘ at fixed delay between terahertz emitter and detector. The gate voltage corresponding to the extrema of these curves identifies charge neutrality to within ~10 mV in gate voltage, or ~7 meV in Fermi energy. In the worst case, our experiments use ๐œ–๐œ–F = 7 meV as charge neutrality, which corresponds to ๐œ‡๐œ‡ = 2.6 meV at 77 K and ๐œ‡๐œ‡ = 0.7 meV at 300 K. At all experimental temperatures ๐œ‡๐œ‡ < ๐‘˜๐‘˜B๐‘‡๐‘‡, with ๐œ‡๐œ‡ โ‰ช ๐‘˜๐‘˜B๐‘‡๐‘‡ at high temperatures. Transmission line model of the waveguide

We model the coplanar waveguide as a transmission line with impedance per unit length ๐‘๐‘โ€ฒ = ๐‘…๐‘…โ€ฒ โˆ’ ๐‘–๐‘–๐‘–๐‘–๐‘–๐‘–โ€ฒ and admittance per unit length ๐‘Œ๐‘Œโ€ฒ = ๐บ๐บโ€ฒ โˆ’ ๐‘–๐‘–๐‘–๐‘–๐‘–๐‘–โ€ฒ, as shown in Fig. S5. Writing expressions for the voltage ๐‘‘๐‘‘(๐‘ง๐‘ง) and the current ๐ผ๐ผ(๐‘ง๐‘ง) in the circuit and expanding to first order in the infinitesimal element ๐›ฅ๐›ฅ๐‘ง๐‘ง leads us to solutions that are forward and backward moving waves,

๐‘‘๐‘‘(๐‘ง๐‘ง) = ๐‘‘๐‘‘+๐‘’๐‘’๐‘–๐‘–๐‘–๐‘–๐‘–๐‘– + ๐‘‘๐‘‘โˆ’๐‘’๐‘’โˆ’๐‘–๐‘–๐‘–๐‘–๐‘–๐‘– ๐ผ๐ผ(๐‘ง๐‘ง) = ๐‘๐‘0โˆ’1(๐‘‘๐‘‘+๐‘’๐‘’๐‘–๐‘–๐‘–๐‘–๐‘–๐‘– โˆ’ ๐‘‘๐‘‘โˆ’๐‘’๐‘’โˆ’๐‘–๐‘–๐‘–๐‘–๐‘–๐‘–)

where the propagation constant is

๐‘˜๐‘˜ = ๐œ”๐œ”๐‘›๐‘›0๐‘๐‘

= ๐‘–๐‘–โˆš๐‘–๐‘–โ€ฒ๐‘–๐‘–โ€ฒ๏ฟฝ๏ฟฝ1 + ๐‘–๐‘–๐‘…๐‘…โ€ฒ

๐œ”๐œ”๐ฟ๐ฟโ€ฒ๏ฟฝ ๏ฟฝ1 + ๐‘–๐‘–๐บ๐บโ€ฒ

๐œ”๐œ”๐ถ๐ถโ€ฒ๏ฟฝ (S1)

and the characteristic impedance is

๐‘๐‘0 = ๏ฟฝ๐‘…๐‘…โ€ฒโˆ’๐‘–๐‘–๐œ”๐œ”๐ฟ๐ฟโ€ฒ

๐บ๐บโ€ฒโˆ’๐‘–๐‘–๐œ”๐œ”๐ถ๐ถโ€ฒ.

The resistance per unit length of the gold traces is ๐‘…๐‘…โ€ฒ โˆผ 4 kฮฉ/m. In contrast, a simple estimate using two parallel wires of circular cross-section finds ๐‘–๐‘–โ€ฒ โˆผ 5 ร— 10โˆ’7 H/m, which at ๐‘–๐‘– = 0.5 ร— 1012 rad/s yields ๐‘–๐‘–๐‘–๐‘–โ€ฒ โˆผ 250 kฮฉ/m โ‰ซ ๐‘…๐‘…โ€ฒ. We therefore take ๐‘…๐‘…โ€ฒ = 0 for all frequencies. In the region with no graphene, we also have ๐บ๐บโ€ฒ = 0, leaving ๐‘˜๐‘˜ = ๐‘–๐‘–๐‘›๐‘›0/๐‘๐‘ =๐‘–๐‘–โˆš๐‘–๐‘–โ€ฒ๐‘–๐‘–โ€ฒ and ๐‘๐‘0 = ๏ฟฝ๐‘–๐‘–โ€ฒ/๐‘–๐‘–โ€ฒ. We find ๐‘›๐‘›0 = 1.56, ๐‘๐‘0 = 133 ฮฉ, and ๐‘–๐‘–โ€ฒ = 3.89 ร— 10โˆ’11 F/m to within 4% over our spectral range using 2D mode simulations in COMSOL, assuming the dielectric constant of fused quartz to be ๐œ–๐œ– = 3.85 at terahertz frequencies (42).

3

Our transmission line thus reduces to a segment of length ๐‘‘๐‘‘ (the graphene-containing region) with unknown conductance per unit length ๐บ๐บโ€ฒ surrounded by two semi-infinite, lossless regions. This problem maps directly onto the wave optics problem of normal transmission through a slab of thickness ๐‘‘๐‘‘ and index ๐‘›๐‘›1 immersed in a homogeneous medium of index ๐‘›๐‘›0 (Fig. S6). Following Eq. S1, the unknown index of refraction ๐‘›๐‘›1 is

๐‘›๐‘›1 = ๐‘›๐‘›0๏ฟฝ1 + ๐‘–๐‘–๐บ๐บโ€ฒ

๐œ”๐œ”๐ถ๐ถโ€ฒ. (S2)

For an incoming right-moving wave with frequency ๐‘–๐‘– and amplitude ๐ด๐ด0 immediately to the left of the slab of thickness ๐‘‘๐‘‘, the transmitted right-moving wave immediately to the right of the slab has amplitude ๐‘‘๐‘‘๐ด๐ด0, where the transmission coefficient is

๐‘‘๐‘‘ = ๐œ๐œ01๐œ๐œ10๐‘’๐‘’๐‘–๐‘–๐‘˜๐‘˜1๐‘‘๐‘‘

1+๐œŒ๐œŒ01๐œŒ๐œŒ10๐‘’๐‘’2๐‘–๐‘–๐‘˜๐‘˜1๐‘‘๐‘‘ (S3)

with ๐œŒ๐œŒ01 = ๐œ๐œ01 โˆ’ 1 = ๐‘›๐‘›0โˆ’๐‘›๐‘›1

๐‘›๐‘›0+๐‘›๐‘›1.

For Fig. 2 of the main text, we experimentally measured the ratio ๐‘‘๐‘‘๏ฟฝ/๐‘‘๐‘‘๏ฟฝ0 of transmission at ๐œ–๐œ–F โ‰  0 to the โ€œreferenceโ€ transmission at charge neutrality. We extract ๐บ๐บโ€ฒ by numerically inverting the expression ๐‘‘๐‘‘/๐‘‘๐‘‘0 = ๐‘‘๐‘‘๏ฟฝ/๐‘‘๐‘‘๏ฟฝ0, where both ๐‘‘๐‘‘ and the reference transmission ๐‘‘๐‘‘0 are given by Eq. S3, but ๐‘‘๐‘‘0 is known from the conductivity at charge neutrality measured in Fig. 3.

The magnitude of ๐‘‘๐‘‘0 is very nearly equal to 1 owing to the minimal conductivity of charge-neutral graphene. To proceed analytically, we ignore the conductivity of graphene in the reference configuration, leading to

๐‘ก๐‘ก๐‘ก๐‘ก0

=2๏ฟฝ1+ ๐‘–๐‘–๐บ๐บโ€ฒ

๐œ”๐œ”๐ถ๐ถโ€ฒ๐‘’๐‘’โˆ’๐‘–๐‘–๐œ”๐œ”๐‘›๐‘›0๐‘‘๐‘‘

๐‘๐‘

2๏ฟฝ1+ ๐‘–๐‘–๐บ๐บโ€ฒ

๐œ”๐œ”๐ถ๐ถโ€ฒ cos๏ฟฝ๐œ”๐œ”๐‘›๐‘›0๐‘‘๐‘‘๐‘๐‘

๏ฟฝ1+ ๐‘–๐‘–๐บ๐บโ€ฒ

๐œ”๐œ”๐ถ๐ถโ€ฒ๏ฟฝโˆ’๐‘–๐‘–๏ฟฝ2+๐‘–๐‘–๐บ๐บโ€ฒ

๐œ”๐œ”๐ถ๐ถโ€ฒ๏ฟฝ sin๏ฟฝ๐œ”๐œ”๐‘›๐‘›0๐‘‘๐‘‘๐‘๐‘

๏ฟฝ1+ ๐‘–๐‘–๐บ๐บโ€ฒ

๐œ”๐œ”๐ถ๐ถโ€ฒ๏ฟฝ . (S4)

In the thin-film limit ๐œ”๐œ”๐‘›๐‘›0๐‘‘๐‘‘๐‘๐‘

๏ฟฝ1 + ๐‘–๐‘–๐บ๐บโ€ฒ

๐œ”๐œ”๐ถ๐ถโ€ฒโ‰ช 1, which is quite valid near charge neutrality and largely

valid in the Fermi liquid regime, we can analytically invert the above equation to find ๐บ๐บโ€ฒ = 2

๐‘๐‘0๐‘‘๐‘‘๏ฟฝ๐‘ก๐‘ก0๐‘ก๐‘กโˆ’ 1๏ฟฝ.

We then compute the change in transmission coefficient ๐›ฅ๐›ฅ๐‘‘๐‘‘ when ๐บ๐บโ€ฒ changes by a small amount ๐›ฅ๐›ฅ๐บ๐บโ€ฒ โ‰ช ๐บ๐บโ€ฒ to find

๐›ฅ๐›ฅ๐บ๐บโ€ฒ = โˆ’ 2๐‘๐‘0๐‘‘๐‘‘

๐›ฅ๐›ฅ๐‘ก๐‘ก๐‘ก๐‘ก

, which is used in Fig. 3 and 4 of the main text. Extracting conductivity from measured conductance Extracting the conductivity ๐œŽ๐œŽ of the graphene sheet between the waveguide traces is complicated by the fact that ๐บ๐บโ€ฒ obtained from the experiment also includes serial capacitances and resistances beneath the traces. The situation can be modeled as in Fig. S7A. In the sliver of length ๐›ฅ๐›ฅ๐‘ง๐‘ง, the graphene beneath the waveguide trace contributes a serial resistance ๐œŒ๐œŒt๐‘ ๐‘ /๐›ฅ๐›ฅ๐‘ง๐‘ง looking along ๐‘ฅ๐‘ฅ๏ฟฝ, where ๐œŒ๐œŒt = ๐œŽ๐œŽtโˆ’1 and ๐‘ ๐‘  are the resistivity and length of the graphene beneath the traces, respectively. The capacitance between sliver and trace is ๐‘๐‘t๐‘ ๐‘ ๐›ฅ๐›ฅ๐‘ง๐‘ง, where ๐‘๐‘t is the capacitance per unit area of the hBN. By further dividing the sliver into chunks of length ๐›ฅ๐›ฅ๐‘ฅ๐‘ฅ, forming an RC network of infinitesimal resistors ๐‘…๐‘…t = ๐œŒ๐œŒt๐›ฅ๐›ฅ๐‘ฅ๐‘ฅ/๐›ฅ๐›ฅ๐‘ง๐‘ง and capacitors ๐‘–๐‘–t = ๐‘๐‘t๐›ฅ๐›ฅ๐‘ฅ๐‘ฅ๐›ฅ๐›ฅ๐‘ง๐‘ง (Fig. S7B), we can model the sliver of length ๐›ฅ๐›ฅ๐‘ง๐‘ง as its own transmission line with characteristic

4

impedance 1๐›ฅ๐›ฅ๐‘–๐‘– ๏ฟฝ

๐‘–๐‘–๐œŒ๐œŒt๐œ”๐œ”๐‘๐‘t

. Defining ๐‘๐‘ser = ๏ฟฝ๐‘–๐‘–๐œŒ๐œŒt๐œ”๐œ”๐‘๐‘t

and accounting for the regions beneath both traces as

well as the graphene of conductivity ๐œŽ๐œŽ between the traces, we find the total conductance ๐บ๐บ = ๐บ๐บโ€ฒ๐‘‘๐‘‘ = ๐‘‘๐‘‘(2๐‘๐‘ser + ๐‘ค๐‘ค๐œŽ๐œŽโˆ’1)โˆ’1, (S5)

where we have ignored the effects of the finite length ๐‘ ๐‘ . To mitigate the effect of ๐‘๐‘ser on the total conductance, in all measurements we heavily doped the graphene region beneath the waveguide traces. The doping in this region was controlled by adjusting the voltage difference between graphene sheet (at voltage ๐‘‘๐‘‘graphene) and the grounded waveguide traces; i.e., the waveguide traces served as a local top gate. All figures in the main text were collected using ๐‘‘๐‘‘graphene = โˆ’2 V. For our studies of charge neutrality (Fig. 3), this condition doped the region beneath the traces to ๐œ–๐œ–F = 106 meV, which practically eliminated the effect of ๐‘๐‘ser on ๐บ๐บ. Solid curves in Fig. S8A,B show ๐บ๐บ = ๐‘‘๐‘‘

๐‘ค๐‘ค๐œŽ๐œŽ

(i.e., ๐บ๐บ calculated from Eq. S5 when ๐‘๐‘ser = 0) for charge-neutral graphene at electron temperatures 77 K and 300 K, with ๐œ๐œโˆ’1 = 4 THz and ๐œ๐œโˆ’1 = 8.5 THz, respectively. Dotted and dashed curves show ๐บ๐บ calculated from Eq. S5 under the same conditions, but assuming finite- and infinite-length RC networks beneath the traces, respectively, with nonzero ๐‘๐‘ser. All curves fall nearly on top of each other, demonstrating the negligible effect of ๐‘๐‘ser. On the other hand, at the highest doping levels studied in Fig. 2 of the main text, the additional doping beneath the traces provided by the condition ๐‘‘๐‘‘graphene = โˆ’2 V does not entirely suppress the effect of ๐‘๐‘ser (Fig. S8C,D). The conductance ๐บ๐บ computed using a finite-length RC network with ๐‘ ๐‘  = 16 micron (dotted lines, Fig. S8C,D) furthermore produces oscillations about the infinite-length result (dashed lines) which become increasingly visible for larger ๐œ–๐œ–F. To extract the conductivity ๐œŽ๐œŽ displayed in Fig. 2 of the main text, we used an infinite RC network for ๐‘๐‘ser, with ๐œŽ๐œŽt under the traces determined by the known doping and an assumed scattering rate of 0.6 THz. We chose an infinite RC network since the irregular shapes of our graphene and WS2 flakes (Fig. 1C of the main text) should largely average out any oscillations resulting from finite length. Validity of the transmission line model

The transmission line model should be quantitatively accurate if the conductivity of graphene is large enough that the electric field lines distribute uniformly across the graphene sheet; in this case, graphene can be modeled as a simple resistive element with conductance per unit length ๐บ๐บโ€ฒ. This condition will be satisfied if |๐บ๐บโ€ฒ| โ‰ซ ๐‘–๐‘–๐‘–๐‘–โ€ฒ, so that the admittance due to the line capacitance is negligible. In the opposite limit, |๐บ๐บโ€ฒ| โ‰ช ๐‘–๐‘–๐‘–๐‘–โ€ฒ, the graphene is a small perturbation on the waveguide mode. The mode will accordingly look like that of the empty waveguide, in which ๐ธ๐ธ๐‘ฅ๐‘ฅ between the traces is not constant (Fig. S9). Additional circuit elements would be required to account for the non-uniform field distribution across the sheet. The line capacitance of our waveguide is ๐‘–๐‘–โ€ฒ = 1 ๐‘’๐‘’2

โ„Ž ยตmโˆ’1 THzโˆ’1. At low frequencies in

the Fermi liquid regime (Fig. 2 of the main text), |๐บ๐บโ€ฒ| โ‰ซ ๐‘–๐‘–๐‘–๐‘–โ€ฒ, but for ๐‘–๐‘– โˆผ 6 ร— 1012 rad/s, ๐‘–๐‘–๐‘–๐‘–โ€ฒ approaches |๐บ๐บโ€ฒ|. Figure S10 tracks the evolution of the mode profile (simulated with the COMSOL 2D mode solver) with increasing frequency for ๐œ–๐œ–F = 119 meV between the traces. The field ๐ธ๐ธ๐‘ฅ๐‘ฅ within the graphene between the traces is extremely homogeneous at ๐‘–๐‘– =1 ร— 1012 rad/s, but spatially varies by ~20% at ๐‘–๐‘– = 5 ร— 1012 rad/s. We note that ๐ธ๐ธ๐‘ฅ๐‘ฅ between the traces (Fig. S10D,F) shows the opposite curvature to that of the empty waveguide

5

(Fig. S9C), implying that even at high frequencies, the graphene sheet (including the heavily doped region at ๐œ–๐œ–F = 159 meV beneath the traces) still strongly affects the field distribution.

At charge neutrality, the magnitude of the line admittance either approaches or exceeds |๐บ๐บโ€ฒ| throughout our spectral range. Figure S11 tracks the evolution of the mode profile with increasing frequency for ๐œ–๐œ–F = 0 between the traces and the experimentally relevant doping ๐œ–๐œ–F =106 meV beneath the traces. The field ๐ธ๐ธ๐‘ฅ๐‘ฅ within the graphene between the traces is clearly less homogeneous than it is at higher doping. Here the curvature of ๐ธ๐ธ๐‘ฅ๐‘ฅ (Fig. S11B,D,F) generally trends toward that of the empty waveguide, although the curvature inverts near the waveguide traces, likely owing to the region of heavily doped graphene beneath the traces. These simulated mode profiles qualitatively suggest that the transmission line model should work well to describe the Fermi liquid regime, but that it may be somewhat less accurate at charge neutrality. To quantitatively evaluate our models, we simulated the effective refractive index of the waveguide mode for various Drude conductivities ๐œŽ๐œŽ between the waveguide traces and ๐œŽ๐œŽt beneath the traces. We then used the transmission line model (Eq. S2) and series impedance model (parameterized by ๐œŽ๐œŽt; Eq. S5) to extract the apparent conductivity ๐œŽ๐œŽsim between the traces, and compared ๐œŽ๐œŽsim to the input ๐œŽ๐œŽ. Figure S12 shows an example of our results for higher doping. Here ๐œŽ๐œŽ corresponds to ๐œ–๐œ–F = 119 meV, with ๐œ๐œโˆ’1 = 0.6 THz and ๐‘‡๐‘‡ = 77 K. We find that ๐œŽ๐œŽsim matches well with ๐œŽ๐œŽ. More generally, we observe this level of agreement between ๐œŽ๐œŽsim and ๐œŽ๐œŽ throughout the Fermi liquid regime discussed in Fig. 2 of the main text, confirming that the transmission line model adequately describes the physics of the waveguide for high doping. Figure S13 compares ๐œŽ๐œŽ and ๐œŽ๐œŽsim at charge neutrality between the waveguide traces for electron temperatures 77 K and 293 K. The difference between ๐œŽ๐œŽ and ๐œŽ๐œŽsim is less than 2 ๐‘’๐‘’2/โ„Ž, but clearly does depend on frequency. The extent to which this disagreement impacts our fits is examined in the next section. All of the above simulations were performed using a 2D mode solver, even though the real sample is three-dimensional. We verified for a handful of input conductivities that the apparent conductivity ๐œŽ๐œŽsim does not appreciably change if we instead perform a 3D simulation of the transmission and compute ๐œŽ๐œŽsim using Eq. S4. Curve-fitting near charge neutrality The frequency-dependent disagreement between ๐œŽ๐œŽ and ๐œŽ๐œŽsim at charge neutrality (Fig. S13) indicates that the conductivity change ๐›ฅ๐›ฅ๐œŽ๐œŽ shown in Fig. 3 of the main text, extracted using the transmission line model, does not perfectly measure the true change in graphene conductivity upon optical heating. As such, ๐›ฅ๐›ฅ๐œŽ๐œŽ may not exactly obey a difference in Drude functions. We nonetheless find that fitting ๐›ฅ๐›ฅ๐œŽ๐œŽ to a difference in Drude functions remains a reasonable approximation to extract ๐œ๐œโˆ’1 and ๐‘‡๐‘‡e. To demonstrate this, we simulated the effective refractive index of the waveguide mode for charge-neutral graphene of conductivity ๐œŽ๐œŽ(๐‘–๐‘–; ๐œ๐œโˆ’1,๐‘‡๐‘‡e) =๐ท๐ทgr๐œ‡๐œ‡=0(๐‘‡๐‘‡e)๐œ‹๐œ‹โˆ’1(๐œ๐œโˆ’1 โˆ’ ๐‘–๐‘–๐‘–๐‘–)โˆ’1 parameterized by a wide range of ๐œ๐œโˆ’1 and ๐‘‡๐‘‡e. We then used wave

mechanics (Eq. S3) to compute the simulated transmission ๐‘‘๐‘‘sim(๐‘–๐‘–; ๐œ๐œโˆ’1,๐‘‡๐‘‡e) for all ๐œ๐œโˆ’1 and ๐‘‡๐‘‡e, interpolating between discrete parameter values. We finally fit the measured relative transmission change ๐›ฅ๐›ฅ๐‘‘๐‘‘/๐‘‘๐‘‘ to the function ๐›ฅ๐›ฅ๐‘‘๐‘‘sim/๐‘‘๐‘‘sim = ๐‘‘๐‘‘sim(๐‘–๐‘–; ๐œ๐œโˆ’1,๐‘‡๐‘‡e)/๐‘‘๐‘‘sim(๐‘–๐‘–; ๐œ๐œ0โˆ’1,๐‘‡๐‘‡0) โˆ’ 1. The free fit parameters were ๐œ๐œโˆ’1, ๐œ๐œ0โˆ’1, and ๐‘‡๐‘‡e, with ๐‘‡๐‘‡0 = 77 K, as in Fig. 3 of the main text. The ๐›ฅ๐›ฅ๐‘‘๐‘‘sim/๐‘‘๐‘‘sim curves fit the data well (Fig. S14A,B). The extracted temperatures (Fig. S14C) are very similar to those extracted from the fits shown in Fig. 3 (reproduced in Fig. S14D,E,F), while the extracted scattering rates are also similar but slightly lower than those in

6

Fig. 3. Most critically, the scattering rates in both cases are well described by the expression ๐œ๐œโˆ’1 = ๐œ๐œeeโˆ’1 + ๐œ๐œdโˆ’1, with ๐œ๐œeeโˆ’1 = ๐‘–๐‘–๐‘˜๐‘˜B๐‘‡๐‘‡e/โ„ the quantum-critical scattering rate for charge-carrier interactions and ๐œ๐œdโˆ’1 โˆ ๐‘›๐‘›imp๐‘‡๐‘‡eโˆ’1 the scattering rate due to charged impurities. The fits to ๐›ฅ๐›ฅ๐‘‘๐‘‘sim/๐‘‘๐‘‘sim yield ๐‘–๐‘– = 0.16 and ๐‘›๐‘›imp = 1.3 ร— 109 cmโˆ’2, as opposed to ๐‘–๐‘– = 0.20 and ๐‘›๐‘›imp =2.1 ร— 109 cmโˆ’2 from the fits to a difference in Drude functions. The data shown in Fig. 4 of the main text mostly concern graphene at higher conductivities, for which the transmission line model should work better than it does at charge neutrality. We expect some fitting errors owing to our application of the transmission line model to extract ๐›ฅ๐›ฅ๐œŽ๐œŽ, but these errors will not be substantial, and will not change our qualitative demonstration of two-mode conductivity. Validity of the transient heating approach The measured conductivities in Fig. 3 and 4 were collected as the electron system cooled down following optical heating, using an optical pump/terahertz probe technique. Since the terahertz pulse has a nonzero width, the electron system is effectively probed over a (narrow) range of temperatures. In this section, we demonstrate that this technique introduces negligible error in the extracted scattering rates and electron temperatures. We model the sample response using a classical equation of motion, analogous to standard derivations of the Drude response, but with a time-varying scattering rate (denoted ๐œ๐œsโˆ’1(๐‘‘๐‘‘)) to model the effect of temperature decay:

๐‘š๐‘š๏ฟฝฬˆ๏ฟฝ๐‘ฅ + ๐‘š๐‘š๏ฟฝฬ‡๏ฟฝ๐‘ฅ๐œ๐œs(๐‘ก๐‘ก)

= ๐‘’๐‘’๐ธ๐ธ๏ฟฝ๐‘‘๐‘‘ โˆ’ ๐‘‘๐‘‘probe๏ฟฝ, (S6) where ๐‘‘๐‘‘probe indicates the arrival time of the incident pulse. To make direct comparisons with our experiment, we choose the functional form ๐œ๐œsโˆ’1(๐‘‘๐‘‘) = ๐‘–๐‘–๐‘˜๐‘˜๐‘‡๐‘‡s(๐‘‘๐‘‘)/โ„ with the experimental value ๐‘–๐‘– = 0.20. The temperature decay is given by ๐‘‡๐‘‡s(๐‘‘๐‘‘) = 77 K + ๐‘‡๐‘‡s(0)exp(โˆ’๐‘‘๐‘‘/๐‘‘๐‘‘d) with decay constant ๐‘‘๐‘‘d = 8 ps (Fig. S15C), similar to the decay behavior extracted from our data. To calculate the apparent optical conductivity of the model system described by Eq. S6, we numerically solve for ๏ฟฝฬ‡๏ฟฝ๐‘ฅ(๐‘‘๐‘‘), using a Gaussian probing pulse ๐ธ๐ธ๏ฟฝ๐‘‘๐‘‘ โˆ’ ๐‘‘๐‘‘decay๏ฟฝ as shown in Fig. S15A. This pulse is quite realistic for our experimental conditions, as can be judged from Fig. S15B, which compares the time derivative ๐‘‘๐‘‘๐ธ๐ธ/๐‘‘๐‘‘๐‘‘๐‘‘ of this pulse to the experimentally measured ๐‘‘๐‘‘๐ธ๐ธ/๐‘‘๐‘‘๐‘‘๐‘‘. The computed velocity ๏ฟฝฬ‡๏ฟฝ๐‘ฅ(๐‘‘๐‘‘) is then converted to a current density using the expression

๐‘—๐‘—(๐‘‘๐‘‘) =๐ท๐ทgr๐œ‡๐œ‡=0๏ฟฝ๐‘‡๐‘‡s(๐‘ก๐‘ก)๏ฟฝ

๐œ‹๐œ‹๏ฟฝ๐‘š๐‘š๐‘’๐‘’๏ฟฝ ๏ฟฝฬ‡๏ฟฝ๐‘ฅ(๐‘‘๐‘‘), (S7)

where ๐ท๐ทgr๐œ‡๐œ‡=0(๐‘‡๐‘‡s(๐‘‘๐‘‘)) is the temperature-dependent Drude weight of charge-neutral graphene. (We

note that in the standard Drude picture, ๐ท๐ท = ๐‘›๐‘›๐‘’๐‘’2๐œ‹๐œ‹๐‘š๐‘š

; substituting this in for ๐ท๐ทgr๐œ‡๐œ‡=0(๐‘‡๐‘‡s(๐‘‘๐‘‘)), Eq. S7

reduces to the usual formula ๐‘—๐‘—(๐‘‘๐‘‘) = ๐‘›๐‘›๐‘’๐‘’๏ฟฝฬ‡๏ฟฝ๐‘ฅ.) The conductivity for a given time ๐‘‘๐‘‘probe is finally computed using the expression

๐œŽ๐œŽ(๐‘–๐‘–) = ๐‘—๐‘—(๐œ”๐œ”)๐ธ๐ธ(๐œ”๐œ”)

. We computed ๐œŽ๐œŽ(๐‘–๐‘–) for four different delays ๐‘‘๐‘‘probe (Fig. S15C). The real and imaginary parts of these computed conductivities (Fig. S15D and S15E) are seen to closely follow the Drude form. Using the Drude weight ๐ท๐ทgr

๐œ‡๐œ‡=0(๐‘‡๐‘‡) for charge-neutral graphene at electronic temperature ๐‘‡๐‘‡, we performed Drude fits to the computed conductivities with the temperature ๐‘‡๐‘‡ and scattering rate ๐œ๐œโˆ’1 as fit parameters. The values (๐‘‡๐‘‡, ๐œ๐œโˆ’1) retrieved from the fits (Fig. S15F, open circles) are seen to fall very close to the parameters (๐‘‡๐‘‡s, ๐œ๐œsโˆ’1) of the equation of motion at

7

time ๐‘‘๐‘‘ = ๐‘‘๐‘‘probe (Fig. S15F, crosses). These results confirm that the transient cooling effect introduces negligible error in our extraction of the scattering rate and electron temperature. Disorder scattering at charge neutrality In charge-neutral graphene, unscreened and singly charged impurities of density ๐‘›๐‘›imp produce a scattering rate ๐œ๐œdโˆ’1 given by (12)

๐œ๐œdโˆ’1 = ๐œ‹๐œ‹4

27๐œ๐œ(3)๐‘›๐‘›imp

โ„๏ฟฝ ๐‘’๐‘’2

4๐œ‹๐œ‹๐œ–๐œ–0๐œ–๐œ–๏ฟฝ2 1๐‘–๐‘–B๐‘‡๐‘‡e

, where ๐œ๐œ(3) โ‰ˆ 1.202 and ๐œ–๐œ– is the effective dielectric constant of the medium. We assume ๐œ–๐œ– = 4 for graphene encapsulated in hBN. Drude weights of the zero- and finite-momentum modes The Drude weight ๐ท๐ทF of the finite-momentum mode is (8, 12)

๐ท๐ทF = ๐œ‹๐œ‹(๐‘›๐‘›๐‘’๐‘’๐‘ฃ๐‘ฃF)2/๐‘Š๐‘Š, where ๐‘›๐‘›๐‘’๐‘’ is the charge density, ๐‘ฃ๐‘ฃF is the Fermi velocity, and ๐‘Š๐‘Š is the enthalpy density:

๐‘Š๐‘Š = 3(๐‘–๐‘–B๐‘‡๐‘‡e)3

๐œ‹๐œ‹(โ„๐‘ฃ๐‘ฃF)2 โˆซ ๐‘˜๐‘˜๏ฟฝ2 ๏ฟฝ 1๐‘’๐‘’๐‘˜๐‘˜๏ฟฝโˆ’๐œ‡๐œ‡/(๐‘˜๐‘˜B๐‘‡๐‘‡e)+1

+ 1๐‘’๐‘’๐‘˜๐‘˜๏ฟฝ+๐œ‡๐œ‡/(๐‘˜๐‘˜B๐‘‡๐‘‡e)+1

๏ฟฝ ๐‘‘๐‘‘๐‘˜๐‘˜๏ฟฝโˆž0 .

The Drude weight of the zero-momentum mode is ๐ท๐ทZ = ๐ท๐ทgr โˆ’ ๐ท๐ทF, where

๐ท๐ทgr = 2 ๐‘’๐‘’2

โ„2๐‘˜๐‘˜B๐‘‡๐‘‡e log ๏ฟฝ2 cosh ๏ฟฝ ๐œ‡๐œ‡

2๐‘–๐‘–B๐‘‡๐‘‡e๏ฟฝ๏ฟฝ

is the total Drude weight of graphene. Data from an additional sample

We studied the scattering rate as a function of electron temperature at charge neutrality on another sample of similar construction, which we refer to as โ€œSample 2โ€ (see photograph in Fig. S16). Owing to a noisy terahertz emitter, the data from Sample 2 are substantially noisier and more limited than the data from the sample described in the main text (โ€œSample 1โ€). Nonetheless, we were able to measure the conductivity change as a function of optical pump fluence to a reasonable degree of precision (Fig. S17A,B). These measurements are similar to those in Fig. 3 of the main text, except that here we have varied the fluence rather than the pump-probe delay time.

Following the procedure used in Fig. 3, we performed fits to a difference in Drude conductivities and extracted the scattering rate ๐œ๐œโˆ’1 as a function of electron temperature ๐‘‡๐‘‡e (Fig. S17C). The scattering rates of Sample 2 are again found to scale approximately linearly with electron temperature, with a very similar slope to that observed in Sample 1: fitting directly to the expression ๐œ๐œeeโˆ’1 = ๐‘–๐‘–๐‘˜๐‘˜B๐‘‡๐‘‡e/โ„, we find ๐‘–๐‘– = 0.16 (๐›ผ๐›ผ = 0.21) for Sample 2, similar to the value ๐‘–๐‘– = 0.20 (๐›ผ๐›ผ = 0.23) found for Sample 1. We note that compared to Sample 1, Sample 2 shows a much less pronounced upturn in the scattering rate at low temperatures. We hesitate to interpret this strongly given the large noise in this dataset.

The conductivity data for Sample 2 (extracted as for Fig. 3) are seen to collapse onto the universal curve ๐œŽ๐œŽU (Fig. S17D) with ๐‘–๐‘– = 0.16. We suggest that the quantum-critical scattering observed at charge neutrality in both Samples 1 and 2 is a robust feature of clean graphene.

8

Fig. S1. Large-area photograph of the waveguide device.

Fig. S2. Cross-sectional view of the heterostructure beneath the waveguide electrodes.

Fig. S3. Fast Fourier transforms of the pulses in Fig. 2A, inset.

9

Fig. S4. Determination of charge neutrality. (A) Current ๐ผ๐ผ measured at 50 Hz, proportional to the transmitted waveform ๐‘‘๐‘‘๐‘‘๐‘‘/๐‘‘๐‘‘๐‘‘๐‘‘, as a function of the voltage difference ๐‘‘๐‘‘gate โˆ’ ๐‘‘๐‘‘graphene. The emitter-detector delay is held fixed to the peak of the transmitted waveform, and ๐‘‘๐‘‘graphene =โˆ’2 V. A larger current corresponds to higher transmission. Charge neutrality is reached at ๐‘‘๐‘‘gate โˆ’ ๐‘‘๐‘‘graphene = โˆ’0.53 V, the gate voltage of maximum transmission. (B) Current ๐›ฅ๐›ฅ๐ผ๐ผ simultaneously measured at 275 Hz, proportional to the transmission change ๐‘‘๐‘‘(๐›ฅ๐›ฅ๐‘‘๐‘‘)/๐‘‘๐‘‘๐‘‘๐‘‘ upon optically heating the graphene. The minimum of this signal corresponds to charge neutrality. Inset: finer scan of ๐›ฅ๐›ฅ๐ผ๐ผ over a smaller gate voltage range, demonstrating that charge neutrality can be determined to within ~10 mV in ๐‘‘๐‘‘gate, which translates to ~7 meV in ๐œ–๐œ–F.

10

Fig. S5. Repeating element of the transmission line model of our waveguide, characterized by impedance ๐‘๐‘โ€ฒ๐›ฅ๐›ฅ๐‘ง๐‘ง along the waveguide traces and admittance ๐‘Œ๐‘Œโ€ฒ๐›ฅ๐›ฅ๐‘ง๐‘ง between the traces.

Fig. S6. Optical analogue of our experiment. The graphene region of length ๐‘‘๐‘‘ and unknown refractive index ๐‘›๐‘›1 is immersed in a medium of known refractive index ๐‘›๐‘›0.

11

Fig. S7. Modeling serial impedance beneath waveguide traces. (A) Top-down view of the waveguide traces (gold color) and the graphene flake (blue color). A sliver of length ๐›ฅ๐›ฅ๐‘ง๐‘ง, as shown, is analyzed in the text. (B) Cross-sectional view of the sliver of length ๐›ฅ๐›ฅ๐‘ง๐‘ง, modeling the region beneath the traces as RC networks with repeating unit of length ๐›ฅ๐›ฅ๐‘ฅ๐‘ฅ, resistance ๐‘…๐‘…t = ๐œŒ๐œŒt๐›ฅ๐›ฅ๐‘ฅ๐‘ฅ/๐›ฅ๐›ฅ๐‘ง๐‘ง, and capacitance ๐‘–๐‘–t = ๐‘๐‘t๐›ฅ๐›ฅ๐‘ฅ๐‘ฅ๐›ฅ๐›ฅ๐‘ง๐‘ง. The resistance between the traces is ๐œŒ๐œŒ๐‘ค๐‘ค/๐›ฅ๐›ฅ๐‘ง๐‘ง, where ๐œŒ๐œŒ = ๐œŽ๐œŽโˆ’1 is the resistivity of graphene in this region.

12

Fig. S8. Calculated effect of ๐‘๐‘ser on total conductance ๐บ๐บ. (A) Real and (B) imaginary parts of ๐บ๐บ calculated from Eq. S5 for our experimental geometry (๐‘‘๐‘‘ = 9 micron, ๐‘ค๐‘ค = 14 micron, ๐‘ ๐‘  = 16 micron). The graphene region between the traces is held at charge neutrality and an electron temperature of 300 and 77 K, with corresponding scattering rates 8.5 and 4 THz, respectively. Different line-styles show different assumptions about ๐‘๐‘ser: solid curves assume ๐‘๐‘ser = 0, dashed curves assume ๐‘๐‘ser modeled by an infinite RC network, and dotted curves assume ๐‘๐‘ser modeled by a 16 micron RC network. The graphene beneath the waveguide traces is characterized by ๐œ–๐œ–F = 106 meV, ๐‘‡๐‘‡ = 77 K, and ๐œ๐œโˆ’1 = 0.6 THz. (C) Real and (D) imaginary parts of ๐บ๐บ calculated as for (A) and (B), but at ๐œ–๐œ–F = 119, 46, and 33 meV between the traces. The doping beneath the traces is 159, 116, and 111 meV, respectively, as expected for ๐‘‘๐‘‘graphene = โˆ’2 V. The scattering rate is 0.6 THz and the temperature is 77 K throughout the sheet.

13

Fig. S9. Mode of the empty waveguide. (A) Vector plot of the propagating odd mode relevant to our experiment, simulated at ๐‘–๐‘– = 3 ร— 1012 rad/s using the COMSOL 2D mode solver. The half-plane ๐‘ฆ๐‘ฆ < 0 contains the fused quartz substrate, while the half-plane ๐‘ฆ๐‘ฆ > 0 contains vacuum. The waveguide traces, modeled as perfect electrical conductors, are shown as dark rectangles of width 16 micron and thickness 0.2 micron in the region 0 < ๐‘ฆ๐‘ฆ < 0.2 micron. The longitudinal electric field ๐ธ๐ธ๐‘–๐‘– is three orders of magnitude smaller than ๐ธ๐ธ๐‘ฅ๐‘ฅ and ๐ธ๐ธ๐‘ฆ๐‘ฆ. (B) Colormap of the horizontal electric field ๐ธ๐ธ๐‘ฅ๐‘ฅ in units of ๐‘‘๐‘‘/๐‘ค๐‘ค, where ๐‘‘๐‘‘ is the voltage between the traces and ๐‘ค๐‘ค =14 micron is the gap between the traces. (C) Horizontal cut of ๐ธ๐ธ๐‘ฅ๐‘ฅ between the traces at vertical position ๐‘ฆ๐‘ฆ = 0.

14

Fig. S10. Waveguide mode for heavily doped graphene. (A) Colormap of the horizontal electric field ๐ธ๐ธ๐‘ฅ๐‘ฅ at ๐‘–๐‘– = 1 ร— 1012 rad/s in units of ๐‘‘๐‘‘/๐‘ค๐‘ค, where ๐‘‘๐‘‘ is the voltage between the traces and ๐‘ค๐‘ค =14 micron is the gap between the traces. The graphene (dark line) is located at ๐‘ฆ๐‘ฆ = 0 and โˆ’23 micron < ๐‘ฅ๐‘ฅ < 23 micron. The traces of width 16 micron (dark rectangles) are 48 nm above the graphene sheet, following our experimental geometry. The conductivity ๐œŽ๐œŽ of graphene between the traces corresponds to ๐œ–๐œ–F = 119 meV, while the conductivity ๐œŽ๐œŽt beneath the traces corresponds to ๐œ–๐œ–F = 159 meV; the scattering rate is 0.6 THz and the temperature is 77 K throughout the sheet. (B) Horizontal cut of ๐ธ๐ธ๐‘ฅ๐‘ฅ within the graphene and between the traces. (C-F) Analogous to (A,B), but at ๐‘–๐‘– = 3 ร— 1012 rad/s (C,D) and ๐‘–๐‘– = 5 ร— 1012 rad/s (E,F). The period of oscillation beneath the traces agrees with the period calculated from our RC network model of the serial impedance.

15

Fig. S11. Waveguide mode for charge-neutral graphene. Analogous to Fig. S10, except that conductivity ๐œŽ๐œŽ of graphene between the traces corresponds to ๐œ–๐œ–F = 0, ๐œ๐œโˆ’1 = 8.5 THz, and ๐‘‡๐‘‡ = 293 K, while the conductivity ๐œŽ๐œŽt beneath the traces corresponds to ๐œ–๐œ–F = 106 meV, ๐œ๐œโˆ’1 = 0.6 THz, and ๐‘‡๐‘‡ = 77 K.

16

Fig. S12. Verifying the transmission line and series impedance models at high doping. (A) Real and (B) imaginary parts of the conductivity ๐œŽ๐œŽ (solid curves) corresponding to ๐œ–๐œ–F = 119 meV used as input to the simulation, plotted alongside ๐œŽ๐œŽsim extracted from the simulation (filled circles) using the transmission line model and a 16 micron RC network for ๐‘๐‘ser (Eq. S5). The conductivity ๐œŽ๐œŽt beneath the traces, used both as input to the simulation and to calculate ๐‘๐‘ser, corresponds to ๐œ–๐œ–F =159 meV, with ๐œ๐œโˆ’1 = 0.6 THz and ๐‘‡๐‘‡ = 77 K throughout the graphene sheet.

Fig. S13. Error in the transmission line model in studies of charge neutrality. (A) Real and (B) imaginary parts of the conductivity ๐œŽ๐œŽ (solid curves) used as input to the simulation, plotted alongside ๐œŽ๐œŽsim extracted from the simulation (filled circles). The conductivity ๐œŽ๐œŽ between the traces corresponds to ๐œ–๐œ–F = 0, ๐œ๐œโˆ’1 = 8.5 THz, and ๐‘‡๐‘‡ = 293 K (red curves and markers) or ๐œ–๐œ–F = 0, ๐œ๐œโˆ’1 = 3.9 THz, and ๐‘‡๐‘‡ = 77 K (blue curves and markers). The input conductivity ๐œŽ๐œŽt beneath the traces corresponds to ๐œ–๐œ–F = 106 meV, with ๐œ๐œโˆ’1 = 0.6 THz, and ๐‘‡๐‘‡ = 77 K. The series impedance ๐‘๐‘ser is negligible and is ignored in extracting ๐œŽ๐œŽsim.

17

Fig. S14. Fitting temperature-dependent change in transmission at charge neutrality using simulated transmission. (A) Real and (B) imaginary parts of the measured relative change in transmission ๐›ฅ๐›ฅ๐‘‘๐‘‘/๐‘‘๐‘‘ upon heating the electron system to unknown temperature ๐‘‡๐‘‡e with the optical pump; these data are the same as those shown in Fig. 3 of the main text, where the data were interpreted as the conductivity change ๐›ฅ๐›ฅ๐œŽ๐œŽ = โˆ’ 2

๐‘๐‘0๐‘ค๐‘ค๐›ฅ๐›ฅ๐‘ก๐‘ก๐‘ก๐‘ก

using the transmission line model. Solid curves show fits to a difference in simulated relative transmissions ๐›ฅ๐›ฅ๐‘‘๐‘‘sim/๐‘‘๐‘‘sim = ๐‘‘๐‘‘sim(๐‘–๐‘–; ๐œ๐œโˆ’1,๐‘‡๐‘‡e)/๐‘‘๐‘‘sim(๐‘–๐‘–; ๐œ๐œ0โˆ’1,๐‘‡๐‘‡0) โˆ’ 1. The parameters ๐œ๐œโˆ’1, ๐‘‡๐‘‡e, and ๐œ๐œ0โˆ’1 are free fit parameters, with ๐‘‡๐‘‡0 = 77 K. (C) Scattering rate versus temperature extracted from the fits. The scattering rate is well described by a sum of scattering rates ๐œ๐œโˆ’1 = ๐œ๐œeeโˆ’1 + ๐œ๐œdโˆ’1, with ๐œ๐œeeโˆ’1 = 0.16 ๐‘˜๐‘˜B๐‘‡๐‘‡e/โ„ due to charge-carrier interactions and ๐œ๐œdโˆ’1 โˆ ๐‘›๐‘›imp๐‘‡๐‘‡eโˆ’1 due to charged impurities (๐‘›๐‘›imp = 1.3 ร— 109 cmโˆ’2). (D-F) Reproduction of Fig. 3 in the main text, in which the same dataset (interpreted as ๐›ฅ๐›ฅ๐œŽ๐œŽ) is fit using a difference in Drude functions ๐œŽ๐œŽ(๐‘–๐‘–; ๐œ๐œโˆ’1,๐‘‡๐‘‡e) โˆ’ ๐œŽ๐œŽ(๐‘–๐‘–; ๐œ๐œ0โˆ’1,๐‘‡๐‘‡0).

18

Fig. S15. Estimating error due to transient heating approach. (A) Incident Gaussian pulse ๐ธ๐ธ(๐‘‘๐‘‘ โˆ’ ๐‘‘๐‘‘delay) used in the numerical model. (B) Time derivative ๐‘‘๐‘‘๐ธ๐ธ/๐‘‘๐‘‘๐‘‘๐‘‘ of the pulse used in the model (blue curve), displayed alongside the measured time derivative of the experimental pulse (red curve; Fig. 2A, inset). (C) Time-dependent temperature ๐‘‡๐‘‡s(๐‘‘๐‘‘) used in the model (Eq. S6). The right vertical axis shows the corresponding scattering rates ๐œ๐œsโˆ’1(๐‘‘๐‘‘). Probe pulses are chosen to arrive at four different delay times ๐‘‘๐‘‘delay, as indicated by the colored crosses. (D) Real and (E) imaginary parts of the optical conductivity extracted using Eq. S7. Different colors correspond to the different probe delays in (C). Solid curves are best fits using a Drude model with free parameters ๐œ๐œโˆ’1,๐‘‡๐‘‡. (F) Values of ๐œ๐œโˆ’1,๐‘‡๐‘‡ extracted from the fits (open circles) compared to exact values of ๐œ๐œsโˆ’1,๐‘‡๐‘‡s (crosses) at each probe delay time ๐‘‘๐‘‘delay. The fit results are seen to very nearly reproduce the exact values. For reference, the dashed line shows ๐œ๐œsโˆ’1 = 0.20 ๐‘˜๐‘˜B๐‘‡๐‘‡s/โ„.

19

Fig. S16. Photograph of Sample 2. Scale bar is 15 micron.

20

Fig. S17. Quantum-critical scattering at charge neutrality in Sample 2 (analogous to Fig. 3). (A) Real and (B) imaginary parts of the change in optical conductivity for four different fluences absorbed by graphene. For all fluences, the terahertz probe pulse was timed to arrive at a fixed delay ~3 ps after the optical pump. Solid curves show fits to the change in optical conductivity, with free parameters ๐œ๐œโˆ’1,๐‘‡๐‘‡e. (C) Values of ๐œ๐œโˆ’1,๐‘‡๐‘‡e extracted from the fits. The scattering rate is seen to be linear (๐œ๐œeeโˆ’1 = ๐‘–๐‘– ๐‘˜๐‘˜B๐‘‡๐‘‡e/โ„), with a similar slope ๐‘–๐‘– = 0.16 to that of the sample described in the main text. (D) Real and imaginary parts (open and filled circles) of ๐œŽ๐œŽ at different ๐‘‡๐‘‡e (i.e., different pump fluences), replotted as a function of โ„๐‘–๐‘–/๐‘˜๐‘˜B๐‘‡๐‘‡e. Solid and dashed curves are fits to the real and imaginary parts of the universal function ๐œŽ๐œŽU given in the main text; the fits yield ๐‘–๐‘– =0.16.

References

1. D. Pines, P. Noziรจres, The Theory of Quantum Liquids (Addison-Wesley, 1989).

2. A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, A. K. Geim, The electronic properties of graphene. Rev. Mod. Phys. 81, 109โ€“162 (2009). doi:10.1103/RevModPhys.81.109

3. D. E. Sheehy, J. Schmalian, Quantum critical scaling in graphene. Phys. Rev. Lett. 99, 226803 (2007). doi:10.1103/PhysRevLett.99.226803 Medline

4. J. Crossno, J. K. Shi, K. Wang, X. Liu, A. Harzheim, A. Lucas, S. Sachdev, P. Kim, T. Taniguchi, K. Watanabe, T. A. Ohki, K. C. Fong, Observation of the Dirac fluid and the breakdown of the Wiedemann-Franz law in graphene. Science 351, 1058โ€“1061 (2016). doi:10.1126/science.aad0343 Medline

5. A. Lucas, J. Crossno, K. C. Fong, P. Kim, S. Sachdev, Transport in inhomogeneous quantum critical fluids and in the Dirac fluid in graphene. Phys. Rev. B 93, 075426 (2016). doi:10.1103/PhysRevB.93.075426

6. B. N. Narozhny, I. V. Gornyi, A. D. Mirlin, J. Schmalian, Hydrodynamic Approach to Electronic Transport in Graphene. Ann. Phys. 529, 1700043 (2017). doi:10.1002/andp.201700043

7. D. Y. H. Ho, I. Yudhistira, N. Chakraborty, S. Adam, Theoretical determination of hydrodynamic window in monolayer and bilayer graphene from scattering rates. Phys. Rev. B 97, 121404 (2018). doi:10.1103/PhysRevB.97.121404

8. Z. Sun, D. N. Basov, M. M. Fogler, Universal linear and nonlinear electrodynamics of a Dirac fluid. Proc. Natl. Acad. Sci. U.S.A. 115, 3285โ€“3289 (2018). doi:10.1073/pnas.1717010115 Medline

9. L. Fritz, J. Schmalian, M. Mรผller, S. Sachdev, Quantum critical transport in clean graphene. Phys. Rev. B 78, 085416 (2008). doi:10.1103/PhysRevB.78.085416

10. D. T. Son, Quantum critical point in graphene approached in the limit of infinitely strong Coulomb interaction. Phys. Rev. B 75, 235423 (2007). doi:10.1103/PhysRevB.75.235423

11. S. A. Hartnoll, P. K. Kovtun, M. Mรผller, S. Sachdev, Theory of the Nernst effect near quantum phase transitions in condensed matter and in dyonic black holes. Phys. Rev. B 76, 144502 (2007). doi:10.1103/PhysRevB.76.144502

12. M. Mรผller, L. Fritz, S. Sachdev, Quantum-critical relativistic magnetotransport in graphene. Phys. Rev. B 78, 115406 (2008). doi:10.1103/PhysRevB.78.115406

13. M. Mรผller, S. Sachdev, Collective cyclotron motion of the relativistic plasma in graphene. Phys. Rev. B 78, 115419 (2008). doi:10.1103/PhysRevB.78.115419

14. D. A. Bandurin, I. Torre, R. Krishna Kumar, M. Ben Shalom, A. Tomadin, A. Principi, G. H. Auton, E. Khestanova, K. S. Novoselov, I. V. Grigorieva, L. A. Ponomarenko, A. K. Geim, M. Polini, Negative local resistance caused by viscous electron backflow in graphene. Science 351, 1055โ€“1058 (2016). doi:10.1126/science.aad0201 Medline

15. R. Krishna Kumar, D. A. Bandurin, F. M. D. Pellegrino, Y. Cao, A. Principi, H. Guo, G. H. Auton, M. Ben Shalom, L. A. Ponomarenko, G. Falkovich, K. Watanabe, T. Taniguchi, I. V. Grigorieva, L. S. Levitov, M. Polini, A. K. Geim, Superballistic flow of viscous electron fluid through graphene constrictions. Nat. Phys. 13, 1182โ€“1185 (2017). doi:10.1038/nphys4240

16. Y. Nam, D.-K. Ki, D. Soler-Delgado, A. F. Morpurgo, Electronโ€“hole collision limited transport in charge-neutral bilayer graphene. Nat. Phys. 13, 1207โ€“1214 (2017). doi:10.1038/nphys4218

17. M. C. Nuss, J. Orenstein, in Millimeter and Submillimeter Wave Spectroscopy of Solids (Springer, 1998), vol. 12, pp. 7โ€“50.

18. L. Wang, I. Meric, P. Y. Huang, Q. Gao, Y. Gao, H. Tran, T. Taniguchi, K. Watanabe, L. M. Campos, D. A. Muller, J. Guo, P. Kim, J. Hone, K. L. Shepard, C. R. Dean, One-dimensional electrical contact to a two-dimensional material. Science 342, 614โ€“617 (2013). doi:10.1126/science.1244358 Medline

19. W. Liu, R. Valdรฉs Aguilar, Y. Hao, R. S. Ruoff, N. P. Armitage, Broadband microwave and time-domain terahertz spectroscopy of chemical vapor deposition grown graphene. J. Appl. Phys. 110, 083510 (2011). doi:10.1063/1.3651168

20. L. Ren, Q. Zhang, J. Yao, Z. Sun, R. Kaneko, Z. Yan, S. Nanot, Z. Jin, I. Kawayama, M. Tonouchi, J. M. Tour, J. Kono, Terahertz and infrared spectroscopy of gated large-area graphene. Nano Lett. 12, 3711โ€“3715 (2012). doi:10.1021/nl301496r Medline

21. G. Jnawali, Y. Rao, H. Yan, T. F. Heinz, Observation of a transient decrease in terahertz conductivity of single-layer graphene induced by ultrafast optical excitation. Nano Lett. 13, 524โ€“530 (2013). doi:10.1021/nl303988q Medline

22. K. J. Tielrooij, J. C. W. Song, S. A. Jensen, A. Centeno, A. Pesquera, A. Zurutuza Elorza, M. Bonn, L. S. Levitov, F. H. L. Koppens, Photoexcitation cascade and multiple hot-carrier generation in graphene. Nat. Phys. 9, 248โ€“252 (2013). doi:10.1038/nphys2564

23. S.-F. Shi, T.-T. Tang, B. Zeng, L. Ju, Q. Zhou, A. Zettl, F. Wang, Controlling graphene ultrafast hot carrier response from metal-like to semiconductor-like by electrostatic gating. Nano Lett. 14, 1578โ€“1582 (2014). doi:10.1021/nl404826r Medline

24. A. J. Frenzel, C. H. Lui, Y. C. Shin, J. Kong, N. Gedik, Semiconducting-to-metallic photoconductivity crossover and temperature-dependent Drude weight in graphene. Phys. Rev. Lett. 113, 056602 (2014). doi:10.1103/PhysRevLett.113.056602 Medline

25. S. Kar, D. R. Mohapatra, E. Freysz, A. K. Sood, Tuning photoinduced terahertz conductivity in monolayer graphene: Optical-pump terahertz-probe spectroscopy. Phys. Rev. B 90, 165420 (2014). doi:10.1103/PhysRevB.90.165420

26. D. R. Grischkowsky, Optoelectronic characterization of transmission lines and waveguides by terahertz time-domain spectroscopy. IEEE J. Sel. Top. Quantum Electron. 6, 1122โ€“1135 (2000). doi:10.1109/2944.902161

27. See supplementary materials.

28. C. H. Lui, K. F. Mak, J. Shan, T. F. Heinz, Ultrafast photoluminescence from graphene. Phys. Rev. Lett. 105, 127404 (2010). doi:10.1103/PhysRevLett.105.127404 Medline

29. I. Gierz, J. C. Petersen, M. Mitrano, C. Cacho, I. C. E. Turcu, E. Springate, A. Stรถhr, A. Kรถhler, U. Starke, A. Cavalleri, Snapshots of non-equilibrium Dirac carrier distributions in graphene. Nat. Mater. 12, 1119โ€“1124 (2013). doi:10.1038/nmat3757 Medline

30. H. Yan, D. Song, K. F. Mak, I. Chatzakis, J. Maultzsch, T. F. Heinz, Time-resolved Raman spectroscopy of optical phonons in graphite: Phonon anharmonic coupling and anomalous stiffening. Phys. Rev. B 80, 121403 (2009). doi:10.1103/PhysRevB.80.121403

31. K. Kang, D. Abdula, D. G. Cahill, M. Shim, Lifetimes of optical phonons in graphene and graphite by time-resolved incoherent anti-Stokes Raman scattering. Phys. Rev. B 81, 165405 (2010). doi:10.1103/PhysRevB.81.165405

32. J. H. Strait, H. Wang, S. Shivaraman, V. Shields, M. Spencer, F. Rana, Very slow cooling dynamics of photoexcited carriers in graphene observed by optical-pump terahertz-probe spectroscopy. Nano Lett. 11, 4902โ€“4906 (2011). doi:10.1021/nl202800h Medline

33. M. W. Graham, S.-F. Shi, D. C. Ralph, J. Park, P. L. McEuen, Photocurrent measurements of supercollision cooling in graphene. Nat. Phys. 9, 103โ€“108 (2012). doi:10.1038/nphys2493

34. J. C. W. Song, L. S. Levitov, Energy flows in graphene: Hot carrier dynamics and cooling. J. Phys. Condens. Matter 27, 164201 (2015). doi:10.1088/0953-8984/27/16/164201 Medline

35. A. F. Young, C. R. Dean, I. Meric, S. Sorgenfrei, H. Ren, K. Watanabe, T. Taniguchi, J. Hone, K. L. Shepard, P. Kim, Electronic compressibility of layer-polarized bilayer graphene. Phys. Rev. B 85, 235458 (2012). doi:10.1103/PhysRevB.85.235458

36. T. Sohier, M. Calandra, C.-H. Park, N. Bonini, N. Marzari, F. Mauri, Phonon-limited resistivity of graphene by first-principles calculations: Electron-phonon interactions, strain-induced gauge field, and Boltzmann equation. Phys. Rev. B 90, 125414 (2014). doi:10.1103/PhysRevB.90.125414

37. R. R. Biswas, S. Sachdev, D. T. Son, Coulomb impurity in graphene. Phys. Rev. B 76, 205122 (2007). doi:10.1103/PhysRevB.76.205122

38. G. X. Ni, L. Wang, M. D. Goldflam, M. Wagner, Z. Fei, A. S. McLeod, M. K. Liu, F. Keilmann, B. ร–zyilmaz, A. H. Castro Neto, J. Hone, M. M. Fogler, D. N. Basov, Ultrafast optical switching of infrared plasmon polaritons in high-mobility graphene. Nat. Photonics 10, 244โ€“247 (2016). doi:10.1038/nphoton.2016.45

39. T. V. Phan, J. C. W. Song, L. S. Levitov, Ballistic heat transfer and energy waves in an electron system. arXiv 1306.4972 [cond-mat.mes-hall]. 20 June 2013.

40. P. Gallagher, Data for: Quantum-critical conductivity of the Dirac fluid in graphene. Zenodo (2019); http://doi.org/10.5281/zenodo.2552519

41. M. Newville, T. Stensitzki, D. B. Allen, A. Ingargiola, LMFIT: Non-Linear Least-Square Minimization and Curve-Fitting for Python. Zenodo (2014); http://doi.org/10.5281/zenodo.11813

42. M. Naftaly, R. E. Miles, Terahertz Time-Domain Spectroscopy for Material Characterization. Proc. IEEE 95, 1658โ€“1665 (2007). doi:10.1109/JPROC.2007.898835