Ultrafast Electron Dynamics in Epitaxial Graphene Investigated with Time- and Angle-Resolved Photoemission Spectroscopy

6.1 MB·

by  Søren Ulstrup

Ultrafast Electron Dynamics in Epitaxial Graphene Investigated with Time- and
Angle-Resolved Photoemission Spectroscopy
Søren Ulstrup,1 Jens Christian Johannsen,2 Alberto Crepaldi,3 Federico Cilento,3 Michele
Zacchigna,4 Cephise Cacho,5 Richard T. Chapman,5 Emma Springate,5 Felix Fromm,6 Christian
Raidel,6 Thomas Seyller,7 Fulvio Parmigiani,3,8 Marco Grioni,2 and Philip Hofmann1
1Department of Physics and Astronomy, Interdisciplinary Nanoscience Center (iNANO), Aarhus University, Denmark
2Institute of Condensed Matter Physics, E´cole Polytechnique F´ed´erale de Lausanne (EPFL), Switzerland
3Sincrotrone Trieste, Trieste, Italy
4IOM-CNR Laboratorio TASC, Area Science Park,Trieste, Italy
5Central Laser Facility, STFC Rutherford Appleton Laboratory, Harwell, United Kingdom
6Lehrstuhl fu¨r Technische Physik, Universit¨at Erlangen-Nu¨rnberg,Germany
6 7Institut fu¨r Physik, Technische Universit¨at Chemnitz , Germany
1 8Department of Physics, University of Trieste, Italy
0
2 In order to exploit the intriguing optical properties of graphene it is essential to gain a better
understandingofthelight-matterinteractioninthematerialonultrashorttimescales. Excitingthe
n
Diracfermionswithintenseultrafastlaserpulsestriggersaseriesofprocessesinvolvinginteractions
a
between electrons, phonons and impurities. Here we study these interactions in epitaxial graphene
J
supportedonsiliconcarbide(semiconducting)andiridium(metallic)substratesusingultrafasttime-
5
and angle-resolved photoemission spectroscopy (TR-ARPES) based on high harmonic generation.

For the semiconducting substrate we reveal a complex hot carrier dynamics that manifests itself in
]
l an elevated electronic temperature and an increase in linewidth of the π band. By analyzing these
e effects we are able to disentangle electron relaxation channels in graphene. On the metal substrate
-
r thishotcarrierdynamicsisfoundtobeseverelyperturbedbythepresenceofthemetal,andwefind
t that the electronic system is much harder to heat up than on the semiconductor due to screening
s
. of the laser field by the metal.
t
a
m
The excitation of a crystal with an intense ultrafast (ARPES)istheidealmethodtomeasurethebandstruc-
-
d lightpulsebringstheelectronsofthematerialoutofequi- ture and many-body effects of condensed matter mate-
n librium. Theequilibriumconditionsarerestoredthrough rials. The advent of ultrafast intense laser sources and
o
aseriesofrelaxationprocessesondifferenttimescalesin- the development of high harmonic generation (HHG) of
c
cludingcarrierscattering, hotcarrierthermalizationand extreme ultraviolet (XUV) pulses have carried this tech-
[
finallythermalequilibriumwiththelattice[1]. Obtaining nique into the time domain. By creating a transient

1
a consistent picture of such processes for graphene is not distribution of electrons in the unoccupied bands of a
v
straightforward due to the peculiar low-energy spectrum material with an infrared pump pulse and subsequently
5
4 consisting of a Dirac cone that is formed by the π and performing ARPES with an XUV pulse, it is possible
7 π∗ bands crossing in proximity of the Fermi level, which tosimultaneouslyacquirespectralanddynamicinforma-
0 places the optical properties of the material somewhere tion about the out-of-equilibrium carrier excitation and
0
between those of metals and semiconductors [2, 3]. This relaxation processes at arbitrary momenta in the Bril-
.
1 intermediacy leads to a series of intriguing traits such louin zone (BZ) [15]. Such time-resolved ARPES (TR-
0 as constant absorption [4] and the possibility of achiev- ARPES) measurements have been undertaken with near
6
ing carrier multiplication [5–8]. Combined with a high UV laser sources. The probing photon energy is how-
1
carrier mobility and gate tunable carrier doping, these ever limited to ∼ 6 eV [16], which is not enough to col-
:
v form the basis of very attractive perspectives for photo- lect TR-ARPES data from electronic states with large
i
X electric devices such as efficient solar cells or photode- in-plane momentum k due to the constraints imposed
(cid:107)
tectors [9–11]. Since electronic or optoelectronic devices byenergyandmomentumconservation. Forexample, to
r
a that exploit these properties inevitably involve hot car- photoemitelectronsfromtheDiracconeatk =1.7˚A−1
(cid:107)
rier transport, it is crucial to determine how hot carriers ingraphene,photonenergiesofatleast16eVareneeded.
are generated and how they decay when graphene is ex- With XUV-based TR-ARPES this has come within ex-
cited by light. perimental reach as exemplified by recent TR-ARPES
studies of monolayer graphene [17–21], bilayer graphene
The light-matter interaction in graphene has been in- [22, 23] and graphite [24].
tensely studied using all-optical approaches [7, 12–14],
however in order to directly access the electron dynam- Here we discuss the hot carrier dynamics in epitax-
ics of the Dirac particles a technique that simultane- ial graphene samples grown on both semiconducting and
ously offers energy-, momentum- and time-resolution is metallic substrates on the basis of TR-ARPES measure-
called for. Angle-resolved photoemission spectroscopy ments. The response to laser excitations for both types
2
GR/H/SiC
-1.5
1.8 (a) (c) (d) max
-1k (Å)y1.7 K -1.0 eV)--21..00
1.6 -0.1 Γ0.0 0.1 -0.5 π* 0.95 E (bin0.0 min
k (Å-1) V) e 1.7
0.0 (b) x E (ebin -0.24 eV V 1.01.3 ky (Å-1) pos
0.0
V) π -2.0
E (ebin 01..50 max 0.5 (eV)Ebin-10..00 n0eg
1.7
1.5 min 1.0 1.0 k (Å-1)
1.3 1.7 -0.3 0.0 0.3 1.3 y
ky (Å-1) kx (Å-1) -500 0 100 200 500 1000 2000 4000
t (fs)
FIG. 1: Equilibrium and out-of-equilibrium photoemission measurements for GR/H/SiC: (a)-(c) ARPES data obtained with
synchrotron radiation revealing (a) the Fermi contour of graphene around the BZ corner, (b) a single intense linear branch in
the Γ¯−K¯ direction and (c) both branches along a line orthogonal to Γ¯−K¯. In (c) the bands have been linearly extrapolated
(dashed lines) to show the position of the Dirac point (horizontal dashed line) at -0.24 eV. (d) TR-ARPES spectra (top row)
and difference spectra (bottom row) at the time delays marked on the timeline. The excitation is caused by a pump beam of
0.95 eV initially leading to vertical transitions as sketched by the red arrow in (c). The equilibrium Fermi level is marked by
dashed lines in (d).
of substrates are important for the operation of actual and the photoemitted electrons [31]. The samples were
photoelectric devices as semiconducting substrates are kept at a temperature of 300 K, and the total time, en-
employed for gating and metallic contact pads will al- ergy and angular resolution were set to 60 fs, 350 meV
ways be in close proximity of the graphene sheet [9]. To and 0.3◦, respectively. In order to map the equilib-
study the implications of these substrates for the elec- rium electronic structure of the samples with high res-
tron dynamics around the Dirac cone, so-called quasi- olution, synchrotron-based ARPES measurements were
free standing graphene samples are employed as they performed with a photon energy of 47 eV at the SGM-3
exhibit an unperturbed Dirac dispersion. Specifically, beamline on the ASTRID light source in Aarhus, Den-
we use hydrogen intercalated graphene on 6H-SiC(0001) mark [32]. The energy- and angular-resolution in these
(GR/H/SiC) [25, 26] and oxygen intercalated graphene measurements were 18 meV and 0.1◦, respectively, and
on Ir(111) (GR/O/Ir) [27, 28]. Both samples are hole- the samples were kept at a temperature of 70 K using a
doped but with different carrier concentrations. Parts of He cryostat.
theresultsforGR/H/SiChavebeenpublishedpreviously
GR/H/SiC is produced ex situ by thermal decomposi-
[17] but are reviewed here to make the new data and the
tion of the Si-terminated face of SiC(0001) in an argon
comparison to GR/O/Ir more accessible.
atmosphereuntilaso-calledbufferlayerisformed. Inter-
The TR-ARPES measurements on both samples were calationofhydrogenthendecouplesthebufferlayerfrom
performed at the ARPES end-station at the Artemis fa- the substrate, which yields the pristine quasi-free stand-
cility, Rutherford Appleton Laboratory [29, 30]. Pump inggraphenesheetwiththeDiracconeclearlydefinedin
and probe pulses were generated by a 1 kHz Ti:sapphire the synchrotron ARPES measurements in Fig. 1(a)-(c)
amplified laser system, which provided infrared pulses [25, 26]. This quasi-free-standing monolayer graphene is
centered at a wavelength of 785 nm with a duration of p-dopedwithacarrierconcentrationof≈4×1012 cm−2,
30 fs and an energy of 10 mJ per pulse. A p-polarized that places the Dirac point 240 meV above the Fermi
probebeamwithaphotonenergyof33.2eV,correspond- energy as seen by the linear extrapolation of the bands
ing to the 21st harmonic of the HHG spectrum gener- in Fig. 1(c). The TR-ARPES data are acquired along
ated in a pulsed Ar gas jet synchronized with the drive the Γ¯ −K¯ direction marked by the arrow on the Fermi
laser, was selected using a time-preserving monochro- contour map in Fig. 1(a). Along Γ¯ −K¯ the intensity
mator. Tunable pump pulses were obtained by seed- peaksinthelinearbranchinthefirstBZwhileitiscom-
ing an optical parametric amplifier (HE-Topas), and 1.6 pletely extinguished in the second BZ due to well-known
eV energy photons were generated by higher harmon- matrix element effects [33, 34], which leads to the single
ics of the HE-Topas light. The pump beam was s- intense branch in Fig. 1(b). Both branches are visible
polarizedtoavoidanyinterferencebetweenthelaserfield when scanning along a direction orthogonal to Γ¯−K¯ as
3
seen in Fig. 1(c). (a) E (eV) (b)
bin
The TR-ARPES measurements are performed along -0.8 1.0 < 0
Γ¯ − K¯ for a pump excitation of ¯hω = 0.95 eV with a u.) --00..46 300 K
pump fluence of 340 µJ/cm2 that induces vertical tran- b. -0.2 110900 0fs K
s1i(tci)o.nsSbnaetpwsheeontstohfetπheanddispπe∗rsbioanndasreasacsqkuetircehdedfoirntFimige. nsity (ar 000...204 0.5 1320000 f sK
steps of 100 fs both before (negative time delays) and e 0.6
after (positive time delays) the arrival of the optical ex- Int 0.8
1.0
0.0
citation, as shown in Fig. 1(d). As the position of the
1.4 1.7 2.0 0.0 0.8
chemicalpotentialvariesafteropticalexcitation[35],the
k (Å-1) E (eV)
binding energy scale refers to the position of the Fermi (c) y bin
levelatnegativetimedelays,wherethesystemisinequi- 1900
librium. The top panels of Fig. 1(d) display the photoe- τ = 200 fs
mission intensity as a function of the delay time, while 1500 1
thebottompanelsdisplaythedifferencespectraobtained K)
bysubtractingthephotoemissionintensitysignalatneg- T (e1100
τ =3000 fs
ative time delays from the corresponding spectra at pos- 2
700
itive time delays. Contrary to the synchrotron ARPES
data, the second branch of the π band is slightly visible,
300
which may be caused by the poorer angular resolution
leading to photoemission intensity from a broader range 0 1000 2000 3000 4000
t (fs)
of angular distributions. A duplicate of the most intense
branch is also visible in the binding energy range from
FIG. 2: Extraction of the electronic temperature from an
E = −2 eV to E = −1 eV, which is caused by a
bin bin
analysis of momentum distribution curves (MDCs): (a) Sub-
smallamountofphotonsassociatedtothe23rdharmonic
set of MDCs (markers) fitted by double Lorentzian functions
in the XUV pulse.
(lines)fortheTR-ARPESspectrumatatimedelayof100fs.
To a first approximation, the excess intensity (red) (b) Integrated Lorentzians as a function of binding energy
aboveE =0inthedifferencespectracanbeviewedas (markers)atthegiventimedelays. ThefitsconsistofFermi-
bin
the electrons injected by the pump, while the depletion Dirac functions broadened with Gaussians that take the en-
of intensity below E = 0 is due to the corresponding ergyresolutionintoaccount(lines),whichprovidethestated
bin
electronictemperatures. (c)Fittedelectronictemperatureas
holes. Within the optical excitation at t = 0, the excess
afunctionoftimedelay(markers). Thelinecorrespondstoa
intensity increases. This signifies a pumping of electrons
fitting function that describes the rising edge and two expo-
(holes) into the π∗ (π) band. Already during the opti-
nential decays with the given time constants τ and τ , con-
1 2
cal excitation the distribution of holes and electrons ap- voluted with a Gaussian that takes the time resolution into
pears uniform along the branch i.e. we do not observe account.
the filling (emptying) of the discrete states in the π∗ (π)
band at h¯ω/2 above (below) the Dirac point. In fact,
excited electrons are observed at energies substantially function of binding energy yields the energy- and time-
largerthantheexpectedtransition,whichpointstowards dependent occupation function as shown in Fig. 2(b)
Auger-like carrier scattering processes. A thermal distri- for three time delays. Fitting these to Fermi-Dirac (FD)
butionofhotelectronsisthusestablishedalreadywithin distributions, convoluted with a Gaussian function ac-
our time resolution. After the peak signal at t = 100 fs counting for the finite energy resolution, yields the time
the excess intensity decays. Within the first 200 fs the dependenceoftheelectronictemperature,whichisshown
removal of electrons appears efficient from regions of the in Fig. 2(c). Note that at t = 100 fs additional inten-
branch at large negative binding energies while excited sity is seen in the experimental data above the FD fit in
carriers in close vicinity of the Dirac point relax slowly Fig. 2(b), which may be a hint of a non-thermal effect.
on a timescale larger than 1 ps. However, this is difficult to conclude because an asym-
The time dependent temperature of electrons can metry in the matrix elements of the π and π∗ bands can
be extracted from the photoemission intensity using affect the shape of this curve [22, 23, 34], and so can a
a method involving the momentum distribution curves slight misalignment of the cut through the Dirac point.
(MDCs) of the spectra at each time delay [36]. This is We can rule out that the effect is due to separate dis-
illustrated for the spectrum at t = 100 fs in Fig. 2(a), tributions of electrons and holes as observed by Gierz et
where selected MDCs are shown along with the result al. [18] since the threshold fluence for achieving this is
of fits composed by two Lorentzian peaks (one for each 2 mJ/cm2 [37], and we are well below this regime at 340
branch) and a fixed parabolic background. Integrating µJ/cm2. We therefore assume that a thermalized hot
the main Lorentzian peak and plotting this integral as a electron distribution is present for all time delays after
4
(a) (b) holes. To study this further an MDC through the pho-
u.) Ebin = 1 eV Ebin = 1 eV toemission intensity at a binding energy of 1 eV binned
b. u.) over ±0.1 eV is fitted to a Lorentzian peak convoluted
ar0 b. with a Gaussian that takes the momentum resolution of
Diff. ( y (ar the experiment into account. This is done for all spectra
y sit in our time series. Examples are given for a spectrum
Intensit t = 100 fs Inten tt <= 0100 fs bexecfoitreedthsiegnaarlriavtalto=f t1h0e0pfusminp Fpiugl.se3a(nbd). foTrhtehcehpaneagke
of linewidth is tracked via the full-width at half max-
1.4 1.5 1.6 1.7 1.4 1.5 1.6 1.7
k (Å-1) k (Å-1) imum (FWHM) of the Lorentzian fits as a function of
(c) y y time delay as shown in Fig. 3(c). The dynamics of the
0.10
linewidth is seen to duplicate the dynamics of the hot
τ = 200 fs
1 electron temperature with a double exponential fit con-
-1)Å sisting of 200 fs and 3000 fs decay constants describing
M ( 0.09 the data. We can rule out that the broadening of the
H τ =3000 fs
W 2 spectraisaspacechargeeffectcausedbyacloudofpho-
F
toemitted secondary electrons since we do not observe
any rigid shift in kinetic energy of the spectra when the
0.08 pump arrives at the applied fluence of 340 µJ/cm2. If
we increase the pump fluence above 1 mJ/cm2 we ob-
0 1000 2000 3000 4000
t (fs) servesuchshifts, whichareknowntobecausedbyspace
charge [38]. Furthermore, the broadening we observe is
FIG. 3: Analysis of the time dependence of the π-band’s on a timescale of picoseconds while space charge effects
linewidth: (a) MDC of the difference spectrum at a binding persist over nanoseconds [39].
energy of 1 eV, showing tails (marked by arrows) of excess The dynamics of the hot electrons and the holes can
intensityawayfromthemainpeak. Thelineisaguidetothe
be understood from the diagram in Fig. 4. We merely
eye. (b)MDCs(markers)atabindingenergyof1eVbinned
discuss the electrons here but equivalent arguments can
over ±0.1 eV before and immediately after the arrival of the
be made for the holes. Initially, the pump pulse creates
pump. Fits are shown by lines. (c) Full-width at half maxi-
mum (FWHM) of the MDC fits as a function of time delay. a non-thermal population of excited electrons in the π∗
The line is a fitting function, similar to the one used for the band by direct transitions as shown in Fig. 4(a). These
electronic temperature in Fig. 2(c), where the decaying part excited carriers thermalize via the electron-electron in-
is described by two exponential decays with the stated time teraction on a time-scale of ∼30 fs [13, 40–42], which is
constants τ and τ .
1 2 faster than the time resolution of our experiment. This
leaves the electronic system at a well-defined tempera-
ture, which in our case reaches a maximum of 1900 K.
theopticalexcitation. Fittingtheelectronictemperature Auger-like processes such as the intraband scattering of
as a function of time delay with a function that consists electrons and holes depicted in Fig. 4(b) as well as in-
of an exponential rising edge and two exponential de- terbandandintervalleyprocessesbetweenthetwoK¯ and
cays with time constants τ1 and τ2, convoluted with a K¯(cid:48) points(notshown)allcontributetothisfastthermal-
Gaussian that takes the time resolution into account, re- ization [3]. The electron thermalization can be accom-
veals that the hot electron relaxation consists of a fast panied by carrier multiplication, i.e. the generation of
τ1 =200fsdecayandaslowτ2 =3000fsdecayasshown more than one electron-hole pair per absorbed photon if
in Fig. 2(c). the rate of impact ionization processes exceeds that of
Theopticalexcitationofthecarriersdoesnotonlylead the inverse processes of Auger recombination. These in-
to their re-distribution, the spectral function is also af- terband processes are sketched in Fig. 4(c) [5, 43]. The
fected, as seen in the changes of the π band’s linewidth. carrier multiplication effect is a function of fluence, time
Such a change can be directly discerned in the difference and doping [44, 45]. These non-thermal processes are so
spectra in the bottom row of Fig. 1(d). Apart from the fast that we are not able to directly observe them with
obvious loss of intensity in the centre of the band (blue), the time resolution of our experiment.
theintensitydistributionbroadensinmomentumandthe Upon thermalization the hot electronic system can ef-
extent of this effect appears to follow the timescales of ficiently lose energy by emission of optical phonons. In-
the hot electron signal at negative binding energies. An traband transitions by optical phonons cause the elec-
MDC through the difference spectrum at t=100 fs and trons to cascade down to the region close to the Dirac
abindingenergyof1eVisshowninFig. 3(a),wherethe point while interband transitions mediate electron-hole
tails of excess intensity are clearly seen away from the recombinationacrosstheDiracpoint. Bothprocessesare
main depression of intensity due to the pump-induced sketched in Fig. 4(d). In addition to these cooling chan-
5
FIG. 4: Timeline of electron dynamics on the Dirac cone in graphene: (a) At time zero electrons (black spheres) and holes
(yellow spheres) are generated by an infrared pump via vertical transitions (straight arrows). (b)-(c) Within 30 fs the excited
electrons and holes undergo scattering processes (b) within and (c) between the bands by Auger recombination (black curled
arrows) and impact excitations (magenta arrows). (d)-(f) This leads to a thermalized hot electron distribution that decays
efficientlywithin200fsby(d)emissionofopticalphonons(bluewiggledarrows),whichisfollowedby(e)slowersupercollisions
involving acoustic phonons (green wiggled arrows) and impurities (green dashed arrows). (f) Direct acoustic phonon emission
also occurs, but is very inefficient.
nels, intervalley scattering by zone boundary phonons the π band part in Fig. 4(b) [52]. After the peak sig-
also contribute to the cooling (not sketched) [46]. These nal at t=100 fs these additional channels are closed by
cooling channels are blocked for electrons with energies the optical and acoustic phonon processes that mediate
below the typical optical phonon energies in the range of electron-holerecombinationbetweentheπandπ∗ bands.
150−200 meV or when the temperature of the optical This interpretation is supported by the strong similarity
phononsreachesthatoftheelectrons[2,7,47]. Ourdata with the time dependence of the hot electron tempera-
suggest that this occurs within the initial 200 fs through ture, which is essentially caused by the same processes.
the initial fast decay of the electronic temperature.
We now turn to the GR/O/Ir case, which permits
A slow decay via acoustic phonons then sets in, which an investigation of the effect of larger hole doping and,
isresponsibleforthehotcarrierrelaxationtimeconstant more importantly, the implications of a metal substrate
of3000fs. Duetotheverysmallenergytransferinvolved for a laser TR-ARPES experiment. These samples are
in intraband acoustic phonon processes as illustrated in prepared on clean single crystal Ir(111) substrates by
Fig. 4(f), such processes have to occur a multitude of temperature-programmed growth cycles, where ethylene
timestobringthehotelectronintotheπ band. Roughly, gasatabackgroundpressureof5×10−7 mbarisdosedin
the energy transfer in these processes scales as the ratio an ultra high vacuum chamber while ramping the tem-
v /v , where v = 2·104 m/s is the sound velocity and perature of the Ir crystal between 520 K and 1520 K.
s F s
v = 106 m/s is the Fermi velocity in graphene. The This procedure routinely yields single crystal monolayer
F
acousticphononmediateddecayisthereforesoinefficient graphene of high quality [53, 54]. Oxygen is then inter-
that the relaxation would take several hundreds of pi- calated by applying a high oxygen pressure and elevated
coseconds[48]. Insteadthree-bodycollisionswithdefects temperatures. These samples are generally strongly hole
and acoustic phonons can become the dominant contri- doped with a doping level that can be slightly tuned by
bution,bringingtherelaxationtimedowntotheorderof the amount of intercalated oxygen [27]. In this study
afewpicoseconds[49–51]. Theseso-calledsupercollisions we employ a sample with a carrier concentration of 4
are especially relevant for carriers near the Dirac point ×1013 cm−2, placing the Dirac point 750 meV above the
region as shown in Fig. 4(e), and mediate a faster relax- Fermi energy as shown in the synchrotron ARPES data
ationacrosstheDiracpointthatwouldotherwiseimpose in Figs. 5(a)-(c). The Fermi contour is therefore signif-
abottleneckforacousticphononcooling. Thispictureof icantly larger than in the GR/H/SiC case in Fig. 1(a).
thephononcoolingofhotcarriersingrapheneissubstan- A single sharp and quasi-free standing branch is seen in
tiated by a phenomenological three-temperature model the Γ¯ −K¯ direction in Fig. 5(b) and both branches are
as described in our previous study [17]. discerned orthogonal to this direction in Fig. 5(c).
The time-dependent broadening of the π band seen in Since GR/O/Ir exhibits a larger hole doping than
Fig. 3(c) reflects the initial pump-induced opening of GR/H/SiC it is necessary to tune the pump energy to
additional decay channels for the holes as sketched in ¯hω = 1.60 eV in order to enable direct transitions be-
6
GR/O/Ir
-1.5 max
(a) (c) (d)
1.8
-1k (Å)y1.7 K -1.0 π* 1.60 eV eV)--21..00
1.6 Γ -0.75 eV (n
bi min
E 0.0
-0.1k 0(Å.0-1) 0.1 eV)-0.5 π 1.0 1.7
(b) x E (bin 1.3 ky (Å-1) pos
0.0
0.0
V) -2.0 0
e
(bin 0.5 max eV)-1.0
E 0.5 (n neg
1.0 Ebi0.0
1.7
1.5 min 1.0 1.0 k (Å-1)
1.3 1.7 -0.3 0.0 0.3 1.3 y
ky (Å-1) kx (Å-1) -500 40 100 200 500
t (fs)
FIG. 5: Equilibrium and out-of-equilibrium photoemission measurements for GR/O/Ir: (a)-(c) Synchrotron ARPES measure-
mentsof(a)Fermicontour,andπbandsin(b)theΓ¯−K¯ directionand(c)orthogonaltothisdirection. Thelinearextrapolation
(dashed lines) of the bands in (c) reveal the Dirac point position (horizontal dashed line) at -0.75 eV. (d) TR-ARPES spectra
(top row) and difference spectra (bottom row) at the time delays marked on the timeline. The pump energy is tuned to
¯hω = 1.60 eV enabling the vertical transitions sketched by the red arrow in (c). The equilibrium Fermi level is marked by
dashed lines in (d).
tween the π and π∗ bands as sketched in Fig. 5(c). In for metals h¯ω is in the range of 6−15 eV and the ap-
p
the present experiment a substantially larger pump flu- plied pump energy is ¯hω ∼ 1.60 eV [55]. The timescale
ence has been used, 31 mJ/cm2, owing to the weaker of this screening is determined by the plasma frequency
response of the material. A time step of 40 fs is used to and therefore below 1 fs, which is much faster than our
acquire these data. This yields the spectra in Fig. 5(d). experiment. Population of the π∗ bands at h¯ω/2 above
ComparedtoFig. 1(d)astrongerbackgroundisobserved theDiracpointisnotatanypointintimeobservedinour
in addition to the main π band branch, which is caused data. Furthermore, contrary to the GR/H/SiC case we
by the underlying metal substrate [28]. The difference donotevenobservepumpingoftheπ∗band. Theexcited
spectra in the bottom row of Fig. 5(d) show that the carriers in GR/O/Ir decay back to the equilibrium state
excessintensitydecaysfast,andthatitappearslocalized onatimescaleoflessthan400fsasindicatedbythedif-
in close vicinity of the equilibrium Fermi level. ferencespectruminFig. 6(b). Thissuggeststhatthelife-
The fast decay is studied further in Fig. 6, where dif- timeofthehotcarriersisreducedingraphenesupported
ferencespectraareshownimmediatelyafterthepumping by a metal surface. The observed significant surplus of
stage at t=40 fs and at a later stage at t=400 fs. The intensity above the Fermi level in Fig. 6(a), even for k||
observed response at 40 fs in Fig. 6(a) to the large ap- away from the graphene dispersion could help to explain
pliedfluenceappearstobeextremelyweak,andasurplus thisultrafastdecayoftheexcitedcarriersingraphene. It
of intensity is observed for only a small range of binding seems reasonable to assume that this background inten-
energiesabovetheequilibriumFermilevel. Thisobserva- sity corresponds to carriers excited in the metal. In Fig.
tion clearly points towards a substantial screening of the 6(c), we compare the dynamics of the background inten-
electric field of the pump light by the underlying metal, sitytotheintensitycorrespondingtotheexcitedcarriers
and consequently a highly reduced population of excited ingraphenebyintegratingtheintensityinthetwoboxed
carriers in the π band branches. Since the pump beam regions shown in Figs. 6(a)-(b). For both regions, the
is s-polarized with respect to the scattering plane it is integrated intensity is shown to follow a single exponen-
orientedparalleltothesurface. Withinaclassicaldipole tialdecaywiththesamedecayconstantof≈250fs. The
approximationthisfieldiscancelledoutatthesurfaceof observed identical dynamics in the two regions may be
the metal. This description is valid only if the plasma due to the fact that the relaxation dynamics of excited
frequencyω ofthemetalissubstantiallylargerthanthe carriersingrapheneonthismetalsubstrateinvolveade-
p
pumpfieldfrequencyω,whichisfulfilledinourcasesince caythroughthestatesofthemetal. However, giventhat
7
by the European Union through the project Concept-
(a) t = 40 fs (b) t = 400 fs (c)1.0 box 1
-1 Graphene, and by the German Research Foundation in
2 1 2 1 τ = 250 fs
u.) the framework of the SPP 1459 Graphene.
b. box 2
V) 0 ar
E (ebin pos nsity (0.5 τ = 250 fs
e
1 nt
0 I
[1] S. Sundaram and E. Mazur, Nature Materials 1, 217
0.0 (2002).
2 neg
[2] W.-K. Tse and S. Das Sarma, Phys. Rev. B 79, 235406
1.3 1.7 1.3 1.7 0.5 1.0 1.5 2.0
k (Å-1) k (Å-1) t (ps) (2009).
|| ||
[3] E. Malic, T. Winzer, E. Bobkin, and A. Knorr, Phys.
FIG. 6: Dynamics in GR/O/Ir: (a)-(b) difference spectra at Rev. B 84, 205406 (2011).
(a)40fsand(b)400fs. (c)Intensityintegratedintheboxed [4] R.R.Nair,P.Blake,A.N.Grigorenko,K.S.Novoselov,
regionsshownin(a)-(b)andnormalizedtothemaximumin- T.J.Booth,T.Stauber,N.M.R.Peres,andA.K.Geim,
tensityinbox1. Linesaresingleexponentialfitswithadecay Science 320, 1308 (2008).
time of τ =250 fs in both cases. [5] T. Winzer, A. Knorr, and E. Malic, Nano Letters 10,
4839 (2010).
[6] J. C. W. Song, K. J. Tielrooij, F. H. L. Koppens, and
L. S. Levitov, Phys. Rev. B 87, 155429 (2013).
the graphene electronic structure is well decoupled from
[7] K.J.Tielrooij,J.C.W.Song,S.A.Jensen,A.Centeno,
the substrate in this system [27, 28] it seems more likely A. Pesquera, A. Z. Elorza, M. Bonn, L. S. Levitov, and
that the photoinduced hot electron signal in the π band F. H. L. K. and, Nature Physics 9, 248 (2013).
is weakened compared to the GR/H/SiC case due to a [8] T. Ploetzing, T. Winzer, E. Malic, D. Neumaier,
significantscreeningofthelaserfield. Notethatthefield A. Knorr, and H. Kurz, Nano Letters 14, 5371 (2014).
[9] F.Bonaccorso,Z.Sun,T.Hasan,andA.C.Ferrari,Nat
is not entirely screened since a weak optical pumping is
Photon 4, 611 (2010).
observed in the π band in Fig. 6(a).
[10] T. Mueller, F. Xia, and P. Avouris, Nat Photon 4, 297
Inconclusion,usingTR-ARPESonquasi-freestanding (2010).
graphene on a semiconducting SiC substrate it is possi- [11] N.M.Gabor,J.C.W.Song,Q.Ma,N.L.Nair,T.Tay-
ble to induce excitations from the π to the π∗ band and chatanapat, K. Watanabe, T. Taniguchi, L. S. Levitov,
and P. Jarillo-Herrero, Science 334, 648 (2011).
observedynamicsofhotcarriers. Thesethermalizeonan
[12] M. Breusing, S. Kuehn, T. Winzer, E. Mali´c, F. Milde,
ultrashort timescale that is shorter than the experimen-
N. Severin, J. P. Rabe, C. Ropers, A. Knorr, and T. El-
tal time resolution of 60 fs. The hot carrier dynamics is
saesser, Phys. Rev. B 83, 153410 (2011).
dominated by a fast 200 fs optical phonon mediated de- [13] D. Brida, A. Tomadin, C. Manzoni, Y. J. Kim, A. Lom-
cayandaslow3000fsdecayduetoacousticphononsand bardo, S. Milana, R. R. Nair, K. S. Novoselov, A. C.
impurities (supercollisions). For a more detailed analy- Ferrari, G. Cerullo, et al., Nat Commun 4 (2013).
sis of this, see Ref. [17]. This dynamics is seen in the [14] J.Shang,T.Yu,J.Lin,andG.G.Gurzadyan,ACSNano
5, 3278 (2011).
hot electron temperature, which is extracted by analyz-
[15] T. Rohwer, S. Hellmann, M. Wiesenmayer, C. Sohrt,
ing the optically induced transient population of the π∗
A. Stange, B. Slomski, A. Carr, Y. Liu, L. M. Avila,
band, and it is also reflected in a time-dependent broad-
M. Kallane, et al., Nature 471, 490 (2011).
ening of the π band due to the time-dependence of the [16] F. Schmitt, P. S. Kirchmann, U. Bovensiepen, R. G.
availabledecaychannelsforphotoinducedholes. Wefind Moore, L. Rettig, M. Krenz, J. Chu, N. Ru, L. Perfetti,
thatforgrapheneonametalsubstrate,hereoxygeninter- D. H. Lu, et al., Science 321, 1649 (2008).
calated graphene on Ir(111), it is not possible to observe [17] J. C. Johannsen, S. Ulstrup, F. Cilento, A. Crepaldi,
M. Zacchigna, C. Cacho, I. C. E. Turcu, E. Springate,
suchhotcarrierdynamicsduetothemetalscreeningthe
F.Fromm,C.Raidel,etal.,Phys.Rev.Lett.111,027403
laser field. This is an important result for laser-based
(2013).
experiments as metals are expected to severely influence
[18] I. Gierz, J. C. Petersen, M. Mitrano, C. Cacho, I. C. E.
the measurement, and it may play a role for graphene Turcu,E.Springate,A.Sto¨hr,A.Ko¨hler,U.Starke,and
based photoelectric devices where metallic contact pads A. Cavalleri, Nat Mater 12, 1119 (2013).
can affect the graphene-light interaction. [19] T. Someya, H. Fukidome, Y. Ishida, R. Yoshida,
T.Iimori,R.Yukawa,K.Akikubo,S.Yamamoto,S.Ya-
We gratefully acknowledge financial support from the
mamoto, T. Yamamoto, et al., Applied Physics Letters
VILLUM foundation, The Danish Council for Indepen-
104, 161103 (2014).
dent Research / Technology and Production Sciences,
[20] I. Gierz, S. Link, U. Starke, and A. Cavalleri, Faraday
the Swiss National Science Foundation (NSF), EPSRC, Discussions (2014.
TheRoyalSocietyandtheItalianMinistryofUniversity [21] L. Bignardi, T. Haarlammert, C. Winter, M. Montag-
and Research (Grants No. FIRBRBAP045JF2 and No. nese, P. H. M. van Loosdrecht, E. Voloshina, P. Rudolf,
FIRB-RBAP06AWK3). Access to the Artemis Facility and H. Zacharias, Phys. Rev. B 89, 075405 (2014).
[22] S. Ulstrup, J. C. Johannsen, F. Cilento, J. A. Miwa,
was funded by STFC. Work in Chemnitz was supported
8
A. Crepaldi, M. Zacchigna, C. Cacho, R. Chapman, [37] T. Li, L. Luo, M. Hupalo, J. Zhang, M. C. Tringides,
E.Springate,S.Mammadov,etal.,Phys.Rev.Lett.112, J.Schmalian,andJ.Wang,Phys.Rev.Lett.108,167401
0257401 (2014. (2012).
[23] I. Gierz, M. Mitrano, J. C. Petersen, C. Cacho, I. C. E. [38] S. Hellmann, K. Rossnagel, M. Marczynski-Bu¨hlow, and
Turcu,E.Springate,A.Sto¨hr,A.Ko¨hler,U.Starke,and L. Kipp, Physical Review B 79, 035402 (2009).
A. Cavalleri, arXiv:1409.0211 (2014. [39] S. Yang, J. A. Sobota, P. S. Kirchmann, and Z. Shen,
[24] A. Stange, C. Sohrt, T. Rohwer, S. Hellmann, G. Ro- Applied Physics A 116, 85 (2013).
hde, L. Kipp, K. Rossnagel, and M. Bauer, EPJ Web of [40] T. Kampfrath, L. Perfetti, F. Schapper, C. Frischkorn,
Conferences 41, 04022 (2013). and M. Wolf, Phys. Rev. Lett. 95, 187403 (2005).
[25] C. Riedl, C. Coletti, T. Iwasaki, A. A. Zakharov, and [41] M. Breusing, C. Ropers, and T. Elsaesser, Phys. Rev.
U.Starke,PhysicalReviewLetters103,246804(pages4) Lett. 102, 086809 (2009).
(2009). [42] C. H. Lui, K. F. Mak, J. Shan, and T. F. Heinz, Phys.
[26] F.Speck, J.Jobst, F.Fromm, M.Ostler, D.Waldmann, Rev. Lett. 105, 127404 (2010).
M. Hundhausen, H. B. Weber, and T. Seyller, Applied [43] J. C. W. Song, M. S. Rudner, C. M. Marcus, and L. S.
Physics Letters 99, 122106 (pages 3) (2011). Levitov, Nano Letters 11, 4688 (2011).
[27] R. Larciprete, S. Ulstrup, P. Lacovig, M. Dalmiglio, [44] T.WinzerandE.Mali´c,Phys.Rev.B85,241404(2012).
M. Bianchi, F. Mazzola, L. Hornekær, F. Orlando, [45] J. C. Johannsen, S. Ulstrup, A. Crepaldi, F. Cilento,
A.Baraldi,P.Hofmann,etal.,ACSNano6,9551(2012). M. Zacchigna, J. A. Miwa, C. Cacho, R. Chapman,
[28] S. Ulstrup, M. Andersen, M. Bianchi, L. Barreto, E. Springate, F. Fromm, et al., To be published (2014).
B. Hammer, L. Hornekær, and P. Hofmann, 2D Mate- [46] S. Butscher, F. Milde, M. Hirtschulz, E. Malic, and
rials 1, 025002 (2014. A. Knorr, Applied Physics Letters 91, 203103 (pages 3)
[29] C. Cacho, J. Petersen, I. Gierz, H. Liu, S. Kaiser, (2007).
R. Chapman, E. Turcu, A. Cavalleri, and E. Springate, [47] H.Wang,J.H.Strait,P.A.George,S.Shivaraman,V.B.
in 19th International Conference on Ultrafast Phenom- Shields, M. Chandrashekhar, J. Hwang, F. Rana, M. G.
ena (Optical Society of America, 2014). Spencer, C. S. Ruiz-Vargas, et al., Applied Physics Let-
[30] F. Frassetto, C. Cacho, C. A. Froud, I. C. E. Turcu, ters 96, 081917 (pages 3) (2010).
P. Villoresi, W. A. Bryan, E. Springate, and L. Poletto, [48] R. Bistritzer and A. H. MacDonald, Phys. Rev. Lett.
Opt. Express 19, 19169 (2011). 102, 206410 (2009).
[31] A. Crepaldi, F. Cilento, B. Ressel, C. Cacho, J. C. Jo- [49] J. C. W. Song, M. Y. Reizer, and L. S. Levitov, Phys.
hannsen,M.Zacchigna,H.Berger,P.Bugnon,C.Grazi- Rev. Lett. 109, 106602 (2012).
oli, I. C. E. Turcu, et al., Physical Review B 88, [50] A. C. Betz, S. H. Jhang, E. Pallecchi, R. Ferreira,
121404(R) (2013). G. Feve, J.-M. Berroir, and B. Placais, Nat Phys 9, 109
[32] S. V. Hoffmann, C. Søndergaard, C. Schultz, Z. Li, (2013).
and P. Hofmann, Nuclear Instruments and Methods in [51] M.W.Graham,S.-F.Shi,D.C.Ralph,J.Park,andP.L.
Physics Research, A 523, 441 (2004). McEuen, Nat Phys 9, 103 (2012).
[33] E. L. Shirley, L. J. Terminello, A. Santoni, and F. J. [52] F. Rana, Phys. Rev. B 76, 155431 (2007.
Himpsel, Phys. Rev. B 51, 13614 (1995). [53] I. Pletikosic, M. Kralj, P. Pervan, R. Brako, J. Coraux,
[34] M. Mucha-Kruczynski, O. Tsyplyatyev, A. Grishin, A. T. N’Diaye, C. Busse, and T. Michely, Physical Re-
E.McCann,V.I.Fal’ko,A.Bostwick,andE.Rotenberg, view Letters 102, 056808 (2009).
Physical Review B 77, 195403 (pages 12) (2008). [54] M.Kralj,I.Pletikosi´c,M.Petrovi´c,P.Pervan,M.Milun,
[35] A. Crepaldi, B. Ressel, F. Cilento, M. Zacchigna, A. T. N’Diaye, C. Busse, T. Michely, J. Fujii, and
C. Grazioli, H. Berger, P. Bugnon, K. Kern, M. Grioni, I. Vobornik, Phys. Rev. B 84, 075427 (2011).
and F. Parmigiani, Phys. Rev. B 86, 205133 (2012). [55] A. Zangwill, Physics at Surfaces (Cambridge University
[36] S.Ulstrup,J.C.Johannsen,M.Grioni,andP.Hofmann, Press, 1988).
Review of Scientific Instruments 85, 013907 (2014.