Low-temperature photoluminescence of oxide-covered single-layer MoS2

0.48 MB·English

by  G. Plechinger

Low-temperature photoluminescence of oxide-covered single-layer MoS
2
G. Plechinger,1 F.-X. Schrettenbrunner,1 J. Eroms,1 D. Weiss,1 C. Schu¨ller,1 and T. Korn1,∗
1Institut fu¨r Experimentelle und Angewandte Physik,
Universita¨t Regensburg, D-93040 Regensburg, Germany
(Dated: January 12, 2012)
We present a photoluminescence study of single-layer MoS2 flakes on SiO2 surfaces. We demon-
strate that the luminescence peak position of flakes prepared from natural MoS2, which varies by
up to 25 meV between individual as-prepared flakes, can be homogenized by annealing in vacuum,
2 whichremovesadsorbatesfromthesurface. WeuseHfO2andAl2O3layerspreparedbyatomiclayer
1 deposition to coversomeof ourflakes. Weclearly observeasuppression ofthelow-energy lumines-
0 cencepeakobservedforas-preparedflakesatlowtemperatures,indicatingthatthispeakoriginates
2 from excitonsboundtosurface adsorbates. Wealso observedifferenttemperature-inducedshifts of

n the luminescence peaks for the oxide-covered flakes. This effect stems from the different thermal
a expansion coefficients of the oxide layers and the MoS2 flakes. It indicates that the single-layer
J MoS2 flakes strongly adhere tothe oxidelayers and are therefore strained.

1
1
The dichalcogenide MoS2 has a layered crystal struc- using ALD. The samples were annealed prior to ALD
◦
] ture, in which adjacent layers are only weakly coupled at 450 C in a 10 percent hydrogen/90 percent nitrogen
i
c by VanderWaalsforces,while intralayercouplingis due atmosphere, then immediately transferred to the ALD
s to covalent bonds. Therefore, flakes containing very few chamber. Bothdielectrics weregrownat250◦ C with al-
-
l or even a single MoS2 layer may easily be prepared on ternatingpulsesoftrimethylaluminum(TMA)andwater
r
t top of SiO2 using the mechanical cleavage method that fortheAl2O3layerandTetrakis(dimethylamino)hafnium
m
is well-established for graphene [1]. While bulk MoS is (TDMAH) and water for the HfO layer, respectively.
2 2
t. an indirect-gap semiconductor, single-layer MoS2 flakes Scanning Raman spectroscopy measurements are per-
a were recently shown to be strong emitters of photolumi- formedatroomtemperature,detailsoftheRamansetup
m nescence [2, 3], indicating a drastic change of the band arepublishedelsewhere[5]. Formicro-PLmeasurements,
- structuretoadirectgap,whichwaspredictedbydensity the samples are mounted in a He-flow cryostat. A mi-
d
functional theory calculations [4]. Recent time-resolved croscope objective is used to focus the excitation laser
n
measurements demonstratedshortphotocarrierlifetimes (continuous-waveexcitation at 532 nm) onto the sample
o
in few- and single-layer MoS [5, 6]. This makes single- and to collect the emitted PL, which is coupled into a
c 2
[ layerMoS2 flakeshighlyinterestingfor potentialelectro- spectrometer and detected using a CCD camera. The

optic devices, e.g.,fastphotodetectors. Forapplications, laserspotisadjustedsuchthatonlythe single-layerpart
3
suchastransistors[7], the MoS mustbe protectedfrom of the investigated flakes is illuminated.
v 2
theenvironmentbyalarge-bandgapdielectriclayer,such First, we discuss the characterization of the MoS
7 2
4 as Al2O3 or HfO2. flakes. Figure 1 (a) shows an optical micrograph of a
7 Here, we present scanning Raman and temperature- MoS2 flake: thicker areas of the flake are opaque with
3 dependent photoluminescence (PL) measurements of a grey-white colour,thinner areas are distinguishable by
2. single-layerMoS2flakes,someofwhichhavebeencovered different shades of blue, while the SiO2 layer has a uni-
1 byoxidelayersusingatomiclayerdeposition(ALD).Our form, purple color. A small region of the flake depicted
1 samples are prepared from natural MoS , and deposited in Fig. 1 (a) is imaged using AFM (Fig. 1 (b)). Areas
2
1 onaSiliconwafercoveredwitha300nmthickSiO layer of uniform height are clearly seen, and the step heights
: 2
v (thermal oxide) using the transparent tape liftoff tech- extractedfromAFMtracesallowustoidentify theareas
Xi nique. Foreaseoforientationonthewafer,metalmarker marked’1L’ and ’3L’as single andtriple layersof MoS2.
structuresarelithographicallydefinedpriortodeposition Scanning Raman measurements were performed on the
ar of MoS2. First characterization of the deposited flakes same flake. Figure 1 (c) depicts a false color image in
is performed with an optical microscope, and few-layer which the intensity of the MoS2 E12g mode (in-plane op-
flakes are readily identifiable because of the phase shift tical vibration of Mo and S atoms) is color-coded. The
induced bythe thin flakes[8, 9]ontopofthe SiO layer. Ramanintensitymapstheflakeshape,withalargeinten-
2
Selected flakes are analyzed in more detail using atomic sityfortheintermediate-thicknessareaonthetoppartof
force microscopy (AFM), PL and scanning Raman mea- theflake. Furtheranalysisispossibleduetothethickness
surements. For deposition of oxide layers, we first iden- dependence of the MoS2 Raman modes [10]: while the
tifywaferpiecescontainingsingle-layerflakes,thencover A1g mode increases in frequency with the flake thickness,
the whole wafer piece with about 15 nm HfO2 or Al2O3 the E12g anomalously softens due to increased dielectric
screening [11], so that the difference of the mode fre-
quencies is characteristic for a certain thickness. Figure
1 (d) shows this frequency difference, ∆ω, as a function
∗Electronicaddress: [email protected] of position, the image area is the same as in Fig. 1 (c).
2
ω (cm-1)
10 10 diff
(a) (5b µ) m1L3L Positiony(µm)-505 (c) Hig111112223344567811111222334Ramanintensity013570371628655601357037160053485746528301805348685760h................388552106900541054278506984 Positiony(µm)-505 (d) 11122222227890123456..........00000000000000000000
-10 Low -10 27.00
-10 -5 0 5 10 -10 -5 0 5 10
Positionx(µm) Positionx(µm)
FIG. 1: (a) Optical micrograph of a MoS2 flake on a SiO2/Si wafer. The orange square indicates the area of the AFM scan
shown in (b). (c) False color image showing the intensity of the Raman E12g mode. The scan area is similar to the AFM scan
in (b). (d) False color image showing the peak frequency difference ∆ω of the Raman A1g and E12g modes. The scan area is
identical to (c).
Tthheeulpowpeerrppaarrttsohfotwhes aflalakreghears∆∆ωωo=f 2138ccmm−−11(3(1lalayyeersr)),, norm.) (a) as-prepareDd orm.)(b) annealeDd
w∆hωile=th25e lcamrg−e1, tahsicekxppeocrtteiodnfoorf tbhuelkfl-alikkeeoflnaktehse.leSfutbhsaes- nsity ( C sity (n C
qobuseenrtvePdLomnetahseusrienmgelen-tlsayceornpfiarrmt othfathtestflraokneg.PL is only PL Inte AB L Inten BA
P
Next, we discuss the PL of single-layer flakes at room
1.7 1.8 1.9 1.7 1.8 1.9
temperature. Figure 2 (a) shows the room-temperature Energy (eV) Energ y (eV)
PLspectrafor4differentMoS2 flakespreparedusingthe (c) 4 K m.) (d) 240 K
same natural MoS2 source. We note that the PL peak nor E
position varies significantly from flake to flake, with a y (
standard deviation of 10 meV and an average value of A nsit B (Al2O3)
EmnePeanlet=sd,1itn8h2ev1ascamumueVmpl.easAtwfateetrreemtmhpioseurfiantrtusetrdesieonrfiae1s5c0or◦yfoCPstLfaotrma3en0adsmuairnne--- sity (norm.) E PL Inte F 1(H.7fO2) 1.8 1.9
updteeemrsfo.ornmSsuterbdasteewqiutthehnaottutPthLbermePaeLkaisnpugeraetkmheepnovtsasictouiofuntmhb,ee(csFaommig.eesfl2ma(kboer)se), PL Inten B (Al2O3) 11os. (eV)..8980(eE)nergy (eV) H AafslO2-Op2r3ep.
p
uniform, with a standard deviation of 1.5 meV and an 1ak .86
aovbesrearvgaetivoanluseasoffoEllPow=s: 1s8in1c5emtheeV.sinWgele-ilnatyeerrprfleatktehseaset F (HfO2) 1PL pe.84
1.82
the substrate surface are exposed to ambient conditions,
water vapor and other adsorbates are likely to form on 1.7 1.8 1.9 0 50 100150200250
Energy (eV) Temp (K)
top of the flakes, leading to a change of the dielectric
environment (as compared to air/vacuum) on top of the FIG.2: (a)Room-temperaturePLspectraof4differentMoS2
flakes after preparation. (b) Room-temperature PL spectra
flakes, which in turn influences the exciton binding en-
for the same flakes shown in (a), after annealing the flakes
ergy due to screening of the electron-hole Coulomb in- ◦
in vacuum at 150 C for 30 minutes. The dashed lines in
teraction. Furtherspectralshiftcanarisefromthe Stark
both graphs indicate the range of the PL peak positions of
effect, whichcan be causedby chargedimpurities onthe
theflakes. (c)PLspectraof4differentMoS2 flakesmeasured
flake surface. The PL peak position therefore shifts, and at4K.SamplesAandEarenotcoveredbyoxidelayers. (d)
themagnitudeoftheshiftdependsonthecoverageofthe PLspectrafor3oftheflakesshownin(a)measuredat240K.
individual flake. These adsorbates are partially removed Thedashed lines in both graphsindicate thehigh-energy PL
from the flake surface by the annealing process. The peak positions of the flakes. (e) PL peak position for the 3
slightdifferencesinthePLlineshapebetweenindividual flakesshown in (d) as a function of temperature.
flakes,even after annealing, indicate inhomogeneousline
broadening due to some remaining adsorbates.
Finally, we discuss the low-temperature PL experi- peaks, a broad, low-energy peak, and a more narrow
ments. Figure2(c)showsPLspectraoftwoas-prepared high-energy peak. By contrast, the two oxide-covered
samplesandsamplescoveredwitheitherAl O orHfO , flakes only show the high-energy peak, with a weakly
2 3 2
measured at 4 K. We clearly observe that for both as- pronounced low-energy shoulder. We may therefore in-
prepared samples, the PL spectra show two distinct fer thatthe low-energypeak observedin the as-prepared
3
flakes stems from excitons bound to surface impurities, gap. Since the as-prepared flakes show a larger shift of
and that the oxide layers effectively suppress the for- the bandgap with T than the oxide-covered flakes, we
mation of such impurity-bound excitons. This indicates mayinferthatthethermalexpansioncoefficientforMoS
2
thattheannealingprocessandtheheatingofthesamples inthetemperaturerangebetween240Kand4Kislarger
in the ALD chamber effectively remove the majority of thanthatofthestackedAl O /HfO -SiO sytem,result-
2 3 2 2
the surface impurities. We note that the peak positions ing in tensile strain during cooling. We note that the
for the two oxide-covered samples are shifted to higher room-temperature thermal expansion coefficent of ther-
(Al O ) or lower (HfO ) energy as compared to the as- mal oxide SiO ,α = 0.24× 10−6/K [12] is signifi-
2 3 2 2 SiO2
prepared flakes. If dielectric screening of the Coulomb cantlylowerthanthatofeitherofthetwoALD-deposited
interaction were the cause of the observed peak shifts, oxides [13] and probably dominates the thermal expan-
we would expect a blueshift of the PL, which should be sion of the stacked system. The strain-induced bandgap
largerforHfO (ǫ ≈20)thanforAl O (ǫ ≈10). Ase- shift also indicates that the MoS flakes strongly adhere
2 r 2 3 r 2
riesoftemperature-dependentPLmeasurementsdemon- totheoxideiftheyaresandwiched betweentopandbot-
strates that the temperature-induced PL-peak shift is tom layers.
also very different for flakes with and without oxide lay- In conclusion, we have investigated single-layer MoS
2
ers (Fig. 2 (e)): the largest shift in the temperature by photoluminescence as a function of temperature for
range from 4 K to 240 K is observedfor the as-prepared as-prepared, annealed and oxide-covered flakes. We
flakes (∆E = 55 meV), while HfO -(∆E = 36 meV) observe that the PL peak position for different as-
2
and Al O -covered (∆E = 34 meV) flakes show signifi- preparedflakescanbehomogenizedbyannealing. Inlow-
2 3
cantly smaller shifts, so that at 240 K (Fig. 2 (d)), both temperature measurements, we observe two PL peaks
oxide-covered flakes are blue-shifted with respect to the for the as-prepared flakes, while the lower-energy peak
as-prepared flakes. We may understand these observa- is suppressed for the flakes covered with oxide layers.
tions as follows: the ALD deposition of the oxide layers The as-prepared flakes exhibit a stronger temperature-
◦
occurs at 250 C, and during the subsequent cooling of dependent band gap shift than the oxide-covered flakes,
the samples to room temperature and below, the MoS indicatingthattheyarestrainedduetothedifferentther-
2
flakesarestrainedduetothedifferentthermalexpansion mal expansion coefficients of the oxides and the MoS .
2
coefficientsoftheMoS flakeandthetopandbottomox- FinancialsupportbytheDFGviaSFB689andGRK1570
2
idelayers,resultinginstrain-inducedchangesoftheband is gratefully acknowledged.
[1] K.S.Novoselov,D.Jiang,F.Schedin,T.J.Booth,V.V. [8] A. Castellanos-Gomez, N. Agra¨ıt, and G. Rubio-
Khotkevich,S.V.Morozov,andA.K.Geim,Proc.Natl. Bollinger, Applied Physics Letters 96, 213116 (2010).
Acad.Sci. U.S.A. 102, 10451 (2005). [9] M. M. Benameur, B. Radisavljevic, J. S. Hron, S. Sa-
[2] K. F. Mak, C. Lee, J. Hone, J. Shan, and T. F. Heinz, hoo, H. Berger, and A. Kis, Nanotechnology 22, 125706
Phys.Rev.Lett. 105, 136805 (2010). (2011).
[3] A. Splendiani, L. Sun, Y. Zhang, T. Li, J. Kim, C.-Y. [10] C. Lee, H. Yan, L. E. Brus, T. F. Heinz, J. Hone, and
Chim, G. Galli, and F. Wang, Nano Letters 10, 1271 S. Ryu,ACS Nano 4, 2695 (2010).
(2010). [11] A. Molina-S´anchez and L. Wirtz, Phys. Rev. B 84,
[4] S. Leb`egue and O. Eriksson, Phys. Rev. B 79, 115409 155413 (2011).
(2009). [12] C. Tsou, Y.-S. Huang, H.-C. Li, and T.-H. Lai, Sensors
[5] T. Korn, S. Heydrich, M. Hirmer, J. Schmutzler, and and Materials 17, 441 (2005).
C. Schu¨ller, Applied Physics Letters 99, 102109 (2011). [13] Y. Fei, Mineral Physics and Crystallography: A Hand-
[6] R.Wang, B. A.Ruzicka, N.Kumar, M. Z. Bellus, H.-Y. book of Physical Constants (American Geophysical
Chiu, and H. Zhao, http://arxiv.org/abs/1110.6643v1. Union, 1995), chap. Thermal expansion, pp.29–44.
[7] B.Radisavljevic,A.Radenovic,J.Brivio,V.Giacometti,
and A. Kis, Naturenanotech. 6, 147 (2011).