Epitaxial Ge2Sb2Te5 probed by single cycle THz pulses of coherent synchrotronradiationV. Bragaglia, A. Schnegg, R. Calarco, and K. Holldack Citation: Applied Physics Letters 109, 141903 (2016); doi: 10.1063/1.4963889 View online: http://dx.doi.org/10.1063/1.4963889 View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/109/14?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Local structure of epitaxial GeTe and Ge2Sb2Te5 films grown on InAs and Si substrates with (100) and (111)orientations: An x-ray absorption near-edge structure study J. Appl. Phys. 117, 125308 (2015); 10.1063/1.4916529 Coherent phonon modes of crystalline and amorphous Ge2Sb2Te5 thin films: A fingerprint of structure andbonding J. Appl. Phys. 117, 025306 (2015); 10.1063/1.4905617 Low temperature epitaxy of Ge-Sb-Te films on BaF2 (111) by pulsed laser deposition Appl. Phys. Lett. 105, 221908 (2014); 10.1063/1.4903489 Evidence for topological band inversion of the phase change material Ge2Sb2Te5 Appl. Phys. Lett. 103, 243109 (2013); 10.1063/1.4847715 A comparative study on electrical transport properties of thin films of Ge1Sb2Te4 and Ge2Sb2Te5 phase-changematerials J. Appl. Phys. 110, 013703 (2011); 10.1063/1.3603016

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 62.141.165.1 On: Tue, 04 Oct 2016

14:46:34

Epitaxial Ge2Sb2Te5 probed by single cycle THz pulses of coherentsynchrotron radiation

V. Bragaglia,1,a) A. Schnegg,2 R. Calarco,1 and K. Holldack3,b)

1Paul-Drude-Institut f€ur Festk€orperelektronik, Hausvogteiplatz 5-7, 10117 Berlin, Germany2Helmholtz Zentrum Berlin f€ur Materialien und Energie GmbH, Institute for Nanospectroscopy, Kekul�estr. 5,12489 Berlin, Germany3Helmholtz Zentrum Berlin f€ur Materialien und Energie GmbH, Institute for Methods and Instrumentationin Synchrotron Radiation Research, Albert-Einstein-Str. 15, 12489 Berlin, Germany

(Received 13 July 2016; accepted 20 September 2016; published online 4 October 2016)

A THz-probe spectroscopy scheme with laser-induced single cycle pulses of coherent synchrotron

radiation is devised and adapted to reveal the dynamic THz transmittance response in epitaxially

grown phase change materials upon 800 nm fs-laser excitation. Amorphous (a-) and crystalline (c-)

films of the prototypical Ge2Sb2Te5 (GST) alloy are probed with single cycle THz pulses tuned to

the spectral range of the highest absorption contrast at 2 THz. After an initial instantaneous sub-

picosecond (ps) dynamic THz transmittance drop, the response of a-GST in that range is dominated

only by a short recovery time sshort¼ 2 ps of the excited carriers. On the contrary, the behavior of

the c-GST response displays a short decay of 0.85 ps followed by a long one slong¼ 90 ps, sugges-

ting that vacancy layers in an ordered c-GST play a role as dissipation channel for photo-induced

free carriers. Published by AIP Publishing. [http://dx.doi.org/10.1063/1.4963889]

Phase change materials (PCM) are employed in non vol-

atile memory devices due to their fast switching between a

high-resistance amorphous (a-) and a low-resistance crystal-

line (c-) phase.1–4 In electrical memories, the switching pro-

cess is achieved by the application of electrical pulses

(nanosecond range).2,5 The threshold switching, which is

occurring in the a-phase and prior to the a- to c-phase transi-

tion, is governed by electron-phonon coupling and relaxation

processes occurring on the timescale of several hundred fem-

toseconds (fs).6 Such timescale could be addressed by the

ultra-fast electrical read-out schemes, which are particularly

challenging, as conventional electrical voltage driven BIAS

measurements across a PCM are limited by the speed, at

which voltage pulses may be applied and currents may be

measured.

The method of choice to overcome this limitation is an

ultrafast optical-pump/THz-probe detection. THz radiation is

extremely sensitive to both free-carrier absorption and acous-

tic phonons. Despite these facts, only few dynamic THz

studies on polycrystalline GeTe-Sb2Te3 alloys (GST) have

been conducted so far.7,8 On the contrary, several optical-

pump/X-ray-probe can be found in the literature.9,10 In

addition, some of us using the THz beamline11 at Bessy II

reported a huge absorbance contrast between the GST phases

in a specific frequency band.12 Thus, the wide tunability of

the coherent synchrotron radiation (CSR) in the THz range is

highly acceptable. Among many other methods13 to generate

the THz pulses, we selected the laser-induced THz CSR

pulses from free relativistic electrons to achieve tunability in

power and frequency (broad-14 and narrow band15). The radia-

tion is fully transported in vacuum, which is mandatory in the

range of interest here. The current work is devoted to employ

the huge THz transmittance contrast to investigate the

fundamental interactions of the electronic system in sub-

threshold regime for both a- and c-GST.

For the generation of THz pulses in the storage ring

and optical excitation of the sample, a 800 nm fs-laser

(COHERENT Legend Elite PRO; 1.8 mJ pulse energy and

45 fs FWHM-pulse length) routinely operated at 6 kHz repe-

tition rate was employed.14 THz probe pulses are emitted

from a laser excited relativistic electron bunch according to

Fig. 1(a). THz generation is accomplished by a resonant

FIG. 1. Experimental setup of the optical-pump/THz-probe experiment (a)

at BESSY II: THz-CSR is emitted by a slice in the electron bunch (blue

ellipsoid) in the storage ring after 800 nm laser energy modulation in a wig-

gler. Optical and THz pulses are indicated by orange and wider red electric

field traces, respectively. The resulting THz pulse is used to probe changes

in a sample exposed to the same laser pulse. (b) A sub-ps time resolution is

derived from a THz-THz cross-correlation experiment in a 0.1 mm GaP

(110) sample (b) and the measured arrival time histogram of such pump-

probe delay scans is depicted in (c).

a)Electronic mail: bragaglia@pdi-berlin.deb)Electronic mail: karsten.holldack@helmholtz-berlin.de

0003-6951/2016/109(14)/141903/4/$30.00 Published by AIP Publishing.109, 141903-1

APPLIED PHYSICS LETTERS 109, 141903 (2016)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 62.141.165.1 On: Tue, 04 Oct 2016

14:46:34

energy modulation of a slice of the 1.72 GeV electron bunch

along its propagation in a wiggler. The first harmonic radia-

tion of the latter is tuned to the laser wavelength of

800 nm.16 After passing dispersive elements along its track,

the laser-excited electrons move longitudinally out of their

initial excitation region (the slice), leaving a dip in charge

density behind. This dip (with side lobes)17 gives rise to the

intense ultrashort single cycle THz pulses at subsequent

bending magnet sources in our case at the THz-beamline.

Using a crosscorrelation method depicted in Fig. 1(b), a sig-

nal proportional to the field transient as well as it’s temporal

arrival time jitter in the time domain (Fig. 1(c)) was deter-

mined. The length of the dip in the electron bunch deter-

mines the frequency spectrum of the emitted THz pulse (see

Fig. 2(a)) and via the Fourier limit also the minimum pulse

length (inset in Fig. 2(a)). In addition, much weaker but

broad band incoherent synchrotron radiation, approaching

also the THz range, is emitted by the electron bunch.

For time-slicing with a laser pulse length of rL (here

given in root mean square, rms) in a storage ring, the average

power spectrum Pcoh (x) of a laser-induced THz pulse on a

bunch of length rB as plotted normalized to the total

incoherent emission Pinc (x) from the storage ring in Fig. 1

is given by

Pcoh xð Þ ¼ f

e

rLISB

rB�R

� �2 �L

Itot

Pinc xð Þ; (1)

where the form factor f is the normalized square of the

Fourier transform of the longitudinal shape of the density

modulation along the electron bunch and �L¼ 6 kHz and

�R¼ 1.25 MHz are the repetition frequencies of laser and

bunch, respectively. The bunch currents are ISB¼ 4 mA for

the sliced bunch and Itot¼ 300 mA total current at BESSY II;

e is the elementary charge. By inserting the relevant values,

one finds that Pcoh (x) from one single slice exceeds the

average power from all other bunches.

This is confirmed by the measurement in Fig. 2(a),

where the laser induced Pcoh and reference spectra Pinc taken

at time-averaged detection are simply divided one by the

other. The observed peak is at around 70 cm�1 (i.e., at about

2 THz), but it can be tuned by changing the laser parameters

(chirp) and detuning the so called free electron laser (FEL)

resonance condition16 given by the wiggler gap setting,

as shown in Fig. 2(a). This allows one to tune the probe

spectrum to exactly match the broad minimum in the static

frequency response of c-GST, which becomes an obvious

comparing of Figs. 2(b) and 2(c).

The sample is excited with the same 800 nm laser pulse

that generated the THz pulse in the wiggler. To detect the

fs-laser-induced dynamic changes, the probe pulse was

sampled using digital boxcar averaging (UHFLI, Z€urich

Instruments) of the THz signal (with a 4.2 K bolometer) trig-

gered to 3 kHz, a phase-locked sub-harmonic of the laser

repetition frequency (6 kHz) (see Fig. 1(a)). Since the pump

pulse of 100 fs pulse duration has to pass a chopper wheel

that blocks it every second shot, the pumped and unpumped

sample transmittance can be acquired in separate channels

but simultaneously. The delay between laser and THz pulse

is varied by an optical delay stage in the laser path. The time

resolution of the experiment is given by the pulse length of

the THz pulse rather than the laser. It is determined not only

by the initial sub-ps density modulation (inset in Fig. 2(a))

on the electron bunch but also by the lengthening effects

given by two z-cut quartz windows (in vacuum) that separate

the THz beamline from the storage ring vacuum system. An

upper limit of 0.7 ps [FWHM] was obtained in Fig. 1(b)

using a GaP (110) single crystal in transmittance as cross-

correlator to probe a signal proportional to the THz electric

field transient ~ETHz. The gray shaded area depicts the corre-

sponding intensity envelope of j~ETHzj2.

For the static measurements shown in Fig. 2(a), the used

a- and c-GST films of 60 nm thickness and Ge2Sb2Te5 nomi-

nal composition were grown by molecular beam epitaxy on

highly insulating Si wafers, ensuring highly textured and

ordered material,18 as compared to conventionally used poly-

crystalline GST. The details on growth and XRD characteri-

zation of a-19 and c-GST18 can be found elsewhere.18,19 A

highly resistive bare Si wafer is used as a reference. For

c-GST, the 60 nm thickness resulted to be an ideal compro-

mise in order to have high quality sample and good pump/

probe signal detection in the dynamic experiment. For

FIG. 2. THz spectra of the probe pulse measured for different gap settings

of the wiggler (modulator, U139) (a) with corresponding sub-ps electron

density modulations from simulations (inset). Measured transmission of

the amorphous (orange) and crystalline (blue) as grown GST films (b).

Measured intensity spectrum of the THz probe pulse chosen for the current

study (c). All measurements are obtained from the FTIR spectroscopy with a

4.2 K Si-bolometer at the Bruker IFS 125HR spectrometer with 6 lm Mylar

beamsplitter. The probe pulse spectra in (a) and (c) were normalized to the

incoherent background spectrum, as emitted from all the other bunches in

the laser-OFF case.

141903-2 Bragaglia et al. Appl. Phys. Lett. 109, 141903 (2016)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 62.141.165.1 On: Tue, 04 Oct 2016

14:46:34

a-GST instead, a 60 nm thick film was grown on quartz sub-

strate, ensuring the total transparence of the substrate to both

laser-pump and THz-probe. All the samples were capped

with 35 nm ZnS-SiO2 to prevent oxidation.

The THz transmittance measurements taken with the

Fourier transform infrared (FTIR) spectroscopy on a- and

c-GST films are shown in Fig. 2(b). The spectra are normal-

ized to the Si substrate. The a-GST film (orange curve)

shows almost no absorption owing to its low concentration

of free carriers and low mobility, which is in agreement with

the insulating nature of the amorphous state. On the contrary,

the crystalline state (blue curve) shows a strong absorption

over the whole spectrum with a broad feature centered at

�70 cm�1, resulting in an �60% difference in transmittance.

In a-GST, carriers are localized20–22 while in c-GST delocal-

ized electrons allows for the conduction.21,23 Such strong

absorption is in accordance with low temperature electrical

measurements, displaying a metallic behavior with a low

DC-resistivity value of 10�3 X cm.18

Upon laser pumping, a signal increase is expected when

full switching from c- to a-GST is induced (see orange arrow

in Fig. 2(b). For pulse energies too low to drive the c- to

a-GST phase transition (sub-threshold pumping), the trans-

mittance must instead drop (blue arrow), as excess photo-

induced carriers lead to additional absorption. Pumping

a-GST can only cause a transmittance drop due to either

photo-excited carriers or a- to c-GST transition (opposite

direction of orange arrow in Fig. 2(b)). A dedicated study on

transmittance/absorbance change upon GST phase transition

can be found elsewhere.12

Fig. 2(c) shows the THz spectrum resonant to the pho-

non mode chosen for the dynamic optical-pump/THz-probe

measurements. Corresponding THz time traces for the c- and

a-phases at a fluence of 1.5 mJ/cm�1 are shown in Fig. 3

(blue and orange symbols, respectively). The THz transmit-

tance curves are normalized to the unpumped case. Both the

phases show a sub-ps drop of the signal at t¼ 0, which is

attributed to the photo-injection of carriers, on a time scale

much shorter than the temporal resolution of the experiment

(�0.7 ps; see the grey peak at time zero in Fig. 3).

The THz dynamic signal for a-GST practically fully

recovers within 7 ps following a single exponential decay

with a recovery time of sshort¼ 2 ps. In c-GST for the same

applied sub-threshold fluence, more carriers are excited in

accordance with its metallic behavior. In addition, the THz-

transmittance can be fitted by a double exponential decay

convoluted with a Heavyside function that takes the time res-

olution of r¼ 0.3 ps (0.7 ps FWHM) into account. From the

fit, two recovery times are extracted: a short one and a long

one. The short one with a recovery time constant sshort

¼ 0.85 ps is faster than what was previously reported in the

literature.8 The long one bears a recovery time of slong

¼ 90 ps and indicates a slow dissipation of the remaining

excess carriers.

Within the short carrier recombination time, it is rea-

sonable to assume an interplay between different fast

recombination pathways. These may contain electron-

phonon scattering as well as recombination via traps and

defects. In particular, traps play a fundamental role in a-

GST due to its band structure.2,6 Possible fast recombina-

tion pathways for both phases through Raman and IR active

phonons are expected, as the measurements are performed

resonant to the phonon spectral frequency range.12

Discrepancies in the short time scales between the present

work and Ref. 8 are attributed to the improved epitaxial

sample quality. The long recovery for the crystalline phase

was not observed before in the polycrystalline GST sam-

ples.8 In addition, based on the previous studies12,18 that

show the structure and electronic properties of epitaxially

grown c-GST, we speculate that the longer recovery might

be linked to the presence of highly ordered vacancy layers.

The latter could represent another recombination channel

for photoexcited carriers that cannot recombine through

alternative faster path ways.

In conclusion, we have shown that 800 nm optical-

pump and THz-probe measurements by tunable THz-CSR

pulses in combination with static THz-FTIR spectroscopy

provide a powerful tool to study the ultrafast light induced

conductivity changes in PCM materials. Benefiting from

the properties of a THz-CSR based set-up, we could tune

the probe pulse to a specific energy window, thereby pro-

viding ultimate sensitivity to the transient transmittance

changes in the ps-regime. Measuring the THz transmit-

tance following a 100 fs optical pulse, we revealed that at a

fluence of 1.5 mJ/cm�1 a decrease in transmittance, indi-

cating a strong free carrier absorption, is observed for both

a- and c-GST. Furthermore, no ultrafast amorphization of

c-GST is evidenced (increased transmittance expected). As

deposited, a-GST is dominated only by a short recovery

time of excited carriers sshort¼ 2 ps, as opposed to c-GST

that exhibits a bi-exponential decay: a short one with

sshort¼ 0.85 ps followed by a long one with slong¼ 90 ps, a

clear indication of the fact that for epitaxially grown crys-

talline structures at least two channels for carrier dissipa-

tion are present. The longer recovery time, which is not

observed in the polycrystalline c-GST, might be due to the

presence of vacancies ordered into layers, which strongly

affect the electronic transport properties of c-GST.12,18 To

complete the picture about dynamics in GST upon laser

excitation, a combination of ultrafast THz and X-ray dif-

fraction experiments is well suited to separate the elec-

tronic and structural contributions.

FIG. 3. THz transmission change of the a- (orange triangles) and c- (blue

squares) as grown Ge2Sb2Te5 films. Fits of the experimental data are super-

imposed. Measurements are done with short delay scans (15 ps) to resolve

the instantaneous excitation and the short recovery time sshort. In the inset,

the c-GST transmission change is shown for delay times up to 120 ps reveal-

ing the longer recovery time slong.

141903-3 Bragaglia et al. Appl. Phys. Lett. 109, 141903 (2016)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 62.141.165.1 On: Tue, 04 Oct 2016

14:46:34

We thank R. Mitzner for scientific support, C. Weniger

and D. Ponwitz for the operation of the laser systems and the

cryo facilities, S. Behnke and C. Stemmler for technical

support, and T. Flissikowski for carefully reading the

manuscript. This work was partially supported by the

European Union within the FP7 project PASTRY (GA

317746) and partly by the Leibniz Gemeinschaft within the

Leibniz Competition on a project titled “Epitaxial phase

change superlattices designed for investigation of non-

thermal switching.”

1S. R. Ovshinsky, Phys. Rev. Lett. 21, 1450 (1968).2S. Raoux, W. Welnic, and D. Ielmini, Chem. Rev. 110, 240 (2010).3M. Wuttig and N. Yamada, Nat. Mater. 6, 824 (2007).4J. E. Boschker, M. Boniardi, A. Redaelli, H. Riechert, and R. Calarco,

Appl. Phys. Lett. 106, 023117 (2015).5A. L. Lacaita, Solid-State Electron. 50, 24 (2006).6D. Ielmini, Phys. Rev. B 78, 035308 (2008).7M. J. Shu, P. Zalden, F. Chen, B. Weems, I. Chatzakis, F. Xiong, R.

Jeyasingh, M. C. Hoffmann, E. Pop, H.-S. P. Wong, M. Wuttig, and A. M.

Lindenberg, Appl. Phys. Lett. 104, 251907 (2014).8M. J. Shu, I. Chatzakis, Y. Kuo, P. Zalden, and A. M. Lindenberg, Appl.

Phys. Lett. 102, 201903 (2013).9K. V. Mitrofanov, P. Fons, K. Makino, R. Terashima, T. Shimada, A. V.

Kolobov, J. Tominaga, V. Bragaglia, A. Giussani, R. Calarco, H. Riechert,

T. Sato, T. Katayama, K. Ogawa, Y. Togashi, M. Yabashi, S. Wall, D.

Brewe, and M. Hase, Sci. Rep. 6, 20633 (2016).

10P. Fons, P. Rodenbach, K. V. Mitrofanov, A. V. Kolobov, J. Tominaga, R.

Shayduk, A. Giussani, R. Calarco, M. Hanke, H. Riechert, R. E. Simpson,

and M. Hase, Phys. Rev. B 90, 094305 (2014).11K. Holldack and A. Schnegg, J. Large Scale Facil. 2, A51 (2016).12V. Bragaglia, K. Holldack, J. E. Boschker, F. Arciprete, E. Zallo, T.

Flissikowski, and R. Calarco, Sci. Rep. 6, 28560 (2016).13X.-C. Zhang and J. Xu, Introduction to THz Wave Photonics (Springer,

Berlin, Heidelberg, New York, 2010), p. 246.14K. Holldack, S. Kahn, R. Mitzner, and T. Quast, Phys. Rev. Lett. 96,

054801 (2006).15S. Bielawski, C. Evain, T. Hara, M. Hosaka, M. Katoh, S. Kimura,

A. Mochihashi, M. Shimada, C. Szwaj, T. Takahashi, and Y. Takashima,

Nat. Phys. 4, 390 (2008).16A. Zholents and K. Holldack, in Proceedings of the Free Electron Laser

Conference FEL 2006, Berlin, Germany (2006), p. 725.17J. M. Byrd, Z. Hao, M. C. Martin, D. S. Robin, F. Sannibale, R. W.

Schoenlein, A. A. Zholents, and M. S. Zolotorev, Phys. Rev. Lett. 96,

164801 (2006).18V. Bragaglia, F. Arciprete, W. Zhang, A. M. Mio, E. Zallo, K. Perumal, A.

Giussani, S. Cecchi, J. E. Boschker, H. Riechert, S. Privitera, E. Rimini, R.

Mazzarello, and R. Calarco, Sci. Rep. 6, 23843 (2016).19V. Bragaglia, B. Jenichen, A. Giussani, K. Perumal, H. Riechert, K.

Perumal, and R. Calarco, J. Appl. Phys. 116, 054913 (2014).20K. Shportko, S. Kremers, M. Woda, D. Lencer, J. Robertson, and M.

Wuttig, Nat. Mater. 7, 653 (2008).21T. Siegrist, P. Jost, H. Volker, M. Woda, P. Merkelbach, C.

Schlockermann, and M. Wuttig, Nat. Mater. 10, 202 (2011).22B. Huang and L. Robertson, Phys. Rev. B 81, 081204 (2010).23W. Zhang, A. Thiess, P. Zalden, R. Zeller, P. H. Dederichs, J.-Y. Raty, M.

Wuttig, S. Bluegel, and R. Mazzarello, Nat. Mater. 11, 952 (2012).

141903-4 Bragaglia et al. Appl. Phys. Lett. 109, 141903 (2016)

Reuse of AIP Publishing content is subject to the terms at: https://publishing.aip.org/authors/rights-and-permissions. Download to IP: 62.141.165.1 On: Tue, 04 Oct 2016

14:46:34