for Hadean fields, we next applied a simplifiedpaleointensity method to select zircons. This de-termination consisted of an evaluation of paleo-intensity at a single temperature (565°C) at whichthe characteristic remanent magnetization wasdefined. The 565°C determinations can provide apaleointensity estimate that is within a factor ofapproximately two of the full Thellier-Coe value(fig. S3). Most conservatively, however, we onlyused these data to test for the presence or ab-sence of a geomagnetic field by comparison withthreshold values. Data from 25 of these zirconsmet selection criteria (tables S2 to S4) (11). SHRIMPgeochronological analyses yield 207Pb/206Pb agesbetween ~3.26 and ~4.22 Ga for these samples.These data also lack nonsystematic variations inPb counts, but age data for our two oldest samplesshow some complexity. For our oldest sample(Z565-12; 4.22 Ga), we interpret this complexityas the influence of greenschist grade metamor-phism at ~2.6 Ga (Fig. 2B). For sample Z565-5(4.11 Ga), we cannot exclude the possibility of re-heating between 2.6 and 3.9 Ga. We note thatthis age complexity is not seen in our samplesdating to ~4 Ga and younger (Fig. 2, B to C).Paleointensities recorded by the 565°C deter-minations (Fig. 3) also exceed external field val-ues and therefore suggest the presence of a coredynamo in the Hadean Eon.Before the onset of inner core growth, the geo-

dynamowas likely driven by thermal convection,in which case, the heat flow at the core-mantleboundary (CMB)must have exceeded the adiabat.Recent estimates suggest an adiabatic value around15 TW (14, 15), although some are lower (16).Neither stagnant lid convection, as has beenproposed for the Archean (17), nor the modeledheat transfer across a basal magma ocean (4) isconsistent with an operating dynamo during thisinterval. Advection via heat pipes (18) and/or platetectonics were likely the mainmantle heat trans-fer processes operating on the early Earth.In the absence of a core dynamo field, atmo-

spheric N2 would be susceptible to ionization andremoval by solar wind pickup (1, 19). The presenceof a magnetic field is therefore consistent withthe lack of nitrogen isotopic fractionation deducedby the study of geologic samples of the Archeanatmosphere (20). However, the weakest paleo-intensity values (Fig. 3) are comparable with amagnetic field only ~12% of that of today, whichwould imply magnetopause standoff distancesless than three Earth radii (1). A coronal mass ejec-tion added to this steady-state solar wind couldhave resulted in the pulsing of volatiles and waterfrom Earth’s atmosphere and perhaps implanta-tion of nitrogen on the Moon (21). This in turnargues that Earth’s water budget was initially muchgreater than today—and/or that it was replenishedduring thedeliveryof awater-rich late veneer (22) orduring the tail of the Late Heavy Bombardment(23) at ~3.9 Ga—to account for today’s water in-ventory. In addition to magnetic shielding, totallosses may have been in part mitigated by bottle-necks in transport through the atmosphere and/or an early hydrogen envelope (1, 19), the latterrestricting extreme atmospheric expansion.

Our paleointensity and age data suggest thepresence of a core dynamo >750 million yearsearlier than prior estimates. An early start for thegeodynamo is similar to that of the earlyMartianmagnetic field (24), but the subsequent collapseof the Martian dynamo probably facilitated at-mospheric stripping (25). In contrast, the earlystart and persistence of atmospheric shieldingattendant with the long-lived geodynamo waslikely a key factor in the development of Earth asa habitable planet.

REFERENCES AND NOTES

1. J. A. Tarduno, E. G. Blackman, E. E. Mamajek, Phys. EarthPlanet. Inter. 233, 68–87 (2014).

2. J. A. Tarduno, R. D. Cottrell, M. K. Watkeys, D. Bauch, Nature446, 657–660 (2007).

3. J. A. Tarduno et al., Science 327, 1238–1240 (2010).4. S. Labrosse, J. W. Hernlund, N. Coltice, Nature 450, 866–869

(2007).5. D. J. Dunlop, Ö. Özdemir, Rock Magnetism, Fundamentals

and Frontiers (Cambridge Univ. Press, Cambridge, 1997).6. J. A. Tarduno et al., Science 338, 939–942 (2012).7. W. Compston, R. T. Pidgeon, Nature 321, 766–769 (1986).8. J. W. Valley et al., Nat. Geosci. 7, 219–223 (2014).9. B. Rasmussen, I. R. Fletcher, J. R. Muhling, S. A. Wilde,

Precambrian Res. 180, 26–46 (2010).10. J. A. Tarduno, R. D. Cottrell, Earth Planet. Sci. Lett. 367,

123–132 (2013).11. Materials and methods are available as supplementary

materials on Science Online.12. M. Hopkins, T. M. Harrison, C. E. Manning,Geology 40, e281 (2012).13. B. Rasmussen, I. R. Fletcher, J. R. Muhling, C. J. Gregory,

S. A. Wilde, Geology 40, e282–e283 (2012) Reply.

14. P. Olson, Science 342, 431–432 (2013).15. H. Gomi et al., Phys. Earth Planet. Inter. 224, 88–103

(2013).16. P. Zhang, R. E. Cohen, K. Haule, Nature 517, 605–607 (2015).17. C. ONeill, Earth Planet. Sci. Lett. 406, 49–58 (2014).18. W. B. Moore, A. A. G. Webb, Nature 501, 501–505 (2013).19. H. Lammer et al., Mon. Not. R. Astron. Soc. 439, 3225–3238

(2014).20. B. Marty, L. Zimmermann, M. Pujol, R. Burgess, P. Philippot,

Science 342, 101–104 (2013).21. M. Ozima et al., Nature 436, 655–659 (2005).22. F. Albarède, Nature 461, 1227–1233 (2009).23. D. de Niem, E. Kührt, A. Morbidelli, U. Motschmann, Icarus 221,

495–507 (2012).24. M. H. Acuña et al., Science 284, 790–793 (1999).25. K. S. Hutchins, B. M. Jakosky, J. G. Luhmann, J. Geophys. Res.

102, 9183–9189 (1997).

ACKNOWLEDGMENTS

We acknowledge the late William Goree for his efforts in developingthe ultrahigh-sensitivity SQUID magnetometer used in thisstudy. We also thank G. Kloc for assistance in sample preparation;T. Pestaj for assistance in the SHRIMP laboratory; and S. Wilde,J. Valley, D. Trail, and C. Manning for helpful discussions. The datapresented in this manuscript are tabulated in the supplementarymaterials. This work was supported by the U.S. NSF (grantsEAR-0619467 and EAR-1015269).

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/349/6247/521/suppl/DC1Materials and MethodsFigs. S1 to S3Tables S1 to S4References (26–47)

12 February 2015; accepted 30 June 201510.1126/science.aaa9114

NANOELECTRONICS

Epitaxial growth of a monolayerWSe2-MoS2 lateral p-n junction withan atomically sharp interfaceMing-Yang Li,1,2* Yumeng Shi,1* Chia-Chin Cheng,2,3 Li-Syuan Lu,4 Yung-Chang Lin,5

Hao-Lin Tang,1 Meng-Lin Tsai,6 Chih-Wei Chu,2 Kung-Hwa Wei,3 Jr-Hau He,6

Wen-Hao Chang,4,7 Kazu Suenaga,5 Lain-Jong Li1†

Two-dimensional transition metal dichalcogenides (TMDCs) such as molybdenum sulfideMoS2 and tungsten sulfide WSe2 have potential applications in electronics because theyexhibit high on-off current ratios and distinctive electro-optical properties. Spatiallyconnected TMDC lateral heterojunctions are key components for constructing monolayerp-n rectifying diodes, light-emitting diodes, photovoltaic devices, and bipolar junctiontransistors. However, such structures are not readily prepared via the layer-stackingtechniques, and direct growth favors the thermodynamically preferred TMDC alloys.We report the two-step epitaxial growth of lateral WSe2-MoS2 heterojunction, where theedge of WSe2 induces the epitaxial MoS2 growth despite a large lattice mismatch.The epitaxial growth process offers a controllable method to obtain lateral heterojunctionwith an atomically sharp interface.

Two-dimensional (2D) transition metaldichalcogenides (2D TMDCs) are of inter-est for electronics applications in that theyoffer tunabilty of several properties, includ-ing the band gap, band offset, carrier den-

sity, and polarity. (The bulk TMDCs have beenknown for a long time and have not evoked sim-

ilar interest.)Heterostructures formedby verticalstacking of different 2D TMDCs have been real-ized via the transfer of their exfoliated or as-grownflakes (1, 2), where their properties are domi-nated by the stacking orientation and interlayercoupling strength. However, lateral heterostruc-tures with edge contacts offer easier band offset

524 31 JULY 2015 • VOL 349 ISSUE 6247 sciencemag.org SCIENCE

RESEARCH | REPORTSon M

arch 24, 2020

http://science.sciencemag.org/

Dow

nloaded from

tuning because the materials are more spatial-ly separated. The direct growth of lateral hetero-junctions is challenging because TMDC alloysare thermodynamically preferred (3). Recently,theMoS2-MoSe2 (4),WS2-WSe2 (4),WS2-MoS2 (2),andMoSe2-WSe2 (5) lateral heterostructures withinteresting optical and electrical properties wereobtained by one-pot synthetic processes. How-ever, the interface regions for these lateral junc-tions are likely alloy structures because all of theprecursors coexist in vapor phases during thegrowth. Such processes only allow the growth ofheterostructures with either different metals orchalcogen, making it difficult to grow p-n hetero-structures such as WSe2-MoS2.We report the controlled epitaxial growth of

WSe2-MoS2 lateral junction, whereWSe2 is grownon substrates through van der Waals epitaxy,

followed by the edge epitaxy ofMoS2 along theWgrowth front. Two-step growth offers precise con-trol to achieve the atomically sharp transition incompositions at the junction. Optical and micro-scopic characterizations revealed the detailedmechanisms for the regrowth (similar to livinggrowth) for the 2D TMDC systems. The 2D lat-eral WSe2-MoS2 heterojunction was synthesizedon c-plane sapphire substrates by sequential chem-ical vapor deposition (CVD) of WSe2 and MoS2(Fig. 1A) (6, 7). To avoid the alloy reaction ob-served in one-pot synthesis, we first preparedsingle-crystalline triangular WSe2 monolayer re-quiring a higher growth temperature (925°C) andthen performed the MoS2 growth at 755°C in aseparate furnace. Aswe previously reported (7), theWSe2 growth proceeds fromWSe2 seeds, followedby van der Waals epitaxy on sapphire. The crucialpoint for successful heterostructure synthesiswithout alloy formation is to control the relativevapor amount of MoO3 and S during the second-step MoS2 growth. The excess in Mo precursorsenhances the MoS2 vertical growth, whereas theexcess in S vapor promotes the formation ofundesired WS2 at the interface (fig. S1).The morphology of in-plane heterostructures

was examined by optical microscopy (OM) andphotoluminescence (PL) and Raman spectros-copies. Figure 1B shows the OM images of thelateralWSe2-MoS2 heterojunctions. All of theWSe2triangles are uniformly surrounded byMoS2, andthe domain for WSe2 and MoS2 can be distin-guished simply by their optical contrast. The lat-

tice constant of WSe2 is 5.53% larger than MoS2(5, 8), which might be one of the factors restrict-ing the growth of MoS2 onto WSe2 basal planes.The Raman and PL spectra in fig. S2 verify thechemical composition of inner WSe2 and outerMoS2 and also reveals the formation of the seam-less WSe2-MoS2 junction.The annular dark field (ADF) image of the

lateral WSe2-MoS2 junction obtained with scan-ning transmission electron microscopy (STEM)revealed that the ADF signal increased with theatomic number (Z) as ~Z1.7 (Fig. 1C) (9). Thus, theW (74), Mo (42), Se (34), and S (16) atoms couldbe distinguished by their intensity. The ADF im-age for another location (Fig. 1D) shows the atomicmodels corresponding to the obtained image.An atomically sharp interface between theWSe2-MoS2 junction was formed, where about 90% ofWatoms are located at the interface bridging totwo pair of Se atoms and one pair of S atoms, asdepicted in Fig. 1E. In addition to the ADF image,the coherent interface is also identified by theelectron energy loss spectroscopy (EELS) mea-surement. The EELS line scan (by monitoringthe EELS spectra change across the hetero-junction) shown in fig. S3 verified an atomicallysharp change. These observations suggest that thegrowth starts from the replacement of Se atomsof WSe2 edge by S atoms.Polarization-resolved second-harmonic gener-

ation (SHG)microscopy is sensitive to the crystalorientation and domain boundaries of surfacelayers (10–16). We used a back-reflection geometry

SCIENCE sciencemag.org 31 JULY 2015 • VOL 349 ISSUE 6247 525

1Physical Sciences and Engineering Division, King AbdullahUniversity of Science and Technology, Thuwal, 23955-6900,Kingdom of Saudi Arabia. 2Research Center for AppliedSciences, Academia Sinica, Taipei 10617, Taiwan.3Department of Material Science and Engineering, NationalChiao Tung University, Hsinchu 300, Taiwan. 4Department ofElectrophysics, National Chiao Tung University, Hsinchu,Taiwan. 5National Institute of Advanced Industrial Scienceand Technology (AIST), Tsukuba, Japan. 6Computer,Electrical and Mathematical Sciences and EngineeringDivision, King Abdullah University of Science and Technology,Thuwal, 23955-6900, Kingdom of Saudi Arabia. 7TaiwanConsortium of Emergent Crystalline Materials (TCECM),Ministry of Science and Technology, Taiwan.*These authors contributed equally to this work. †Correspondingauthor. E-mail: lance.li@kaust.edu.sa

Fig. 1. Formation and structure of WSe2-MoS2 heterostructures. (A) Schematic illustration of the sequential growth of the monolayer WSe2-MoS2 in-planeheterostructure. (B) As shown in the optical image, the WSe2 and MoS2 can be distinguished by their optical contrast. (C and D) High-resolution STEM imagestaken from the WSe2-MoS2 in-plane heterostructure. (E) Atomic model showing the interface structure between WSe2 and MoS2.

RESEARCH | REPORTSon M

arch 24, 2020

http://science.sciencemag.org/

Dow

nloaded from

with a linearly polarized pump laser (870 nm)normally incident on a triangular WSe2-MoS2heterostructure sample and detected the SH in-tensities with polarizations parallel (IH) and per-pendicular (IV) to the laser polarization (Fig. 2A)(see fig. S4 for the setup). The total SH intensityItotal (sum of IV and IH) (Fig. 2C), which was gen-erally uniform over the entire WSe2 and MoS2

domains, indicated that the surrounding MoS2was also single crystalline without grain bound-aries. The SH intensity also showed no suppres-sion across the junction, which suggests that theMoS2 grew out from the edges of WSe2 withoutmisorientation. The MoS2 regions showed slightvariations in SH intensity that coincidedwell withthe variations in PL intensity and peak energy

(Fig. 3A, inset). As will be discussed below, weconsider that the SH intensity variations in thesurrounding MoS2 arise from the nonuniformstrain distribution rather than presence of multi-ple grains.To gain more insight into the WSe2-MoS2 het-

erojunction growth, we calculated the crystal ori-entation using q ¼ ð1=3Þtan−1

ffiffiffiffiffiffiffiffiffiffiffiffiIV=IH

p, where q is

the angle between the laser polarization direc-tion and the nearest armchair axis of the sample(13). Themap of q (Fig. 2D) was uniform over theentire WSe2-MoS2 heterostructure, indicatingthat the outgrowing MoS2 was a single crystalwith the same orientation as the innerWSe2. Wehave also performed SHG measurements on aWSe2-MoS2 heterojunction composed of multiplegrains (Fig. 2, B and E). Although these grainshad different orientations (q map, Fig. 2F), theoutgrowingMoS2 followed the orientation of theinner WSe2. These results strongly suggest thatthe outgrowth of MoS2 occurred through epi-taxy off the edge of WSe2 and determined thecrystal orientation, rather than the sapphire sub-strate. Fig. S5 demonstrates that the MoS2 mono-layer also grew out from the prepatterned WSe2monolayer.We noticed that the MoS2 in WSe2-MoS2 het-

erostructures normally exhibits considerable PLenergy differences at different locations (Fig. 3A,inset), where the contour color mapping showsthe spatial distribution of PL energy ranging from1.79 to 1.91 eV. The site with a higher PL energyalways exhibited a higher intensity, as presentedin the PL spectra (Fig. 3A). However, such a largevariation in PLwas not observed in isolatedMoS2monolayers occasionally found on the same sam-ple (fig. S6), so theMoS2 PL variation is related tothe heterostructure formation.We also performedRaman imaging for the MoS2 A1g and E2g fre-quencies (Fig. 3B). Compared with the inset of

526 31 JULY 2015 • VOL 349 ISSUE 6247 sciencemag.org SCIENCE

Fig. 2. Polarization-resolved SHG measurement. (A) The intensity maps of the perpendicular IV andparallel IH components of the SH. The insert shows the optical image. The black double arrow line in-dicates the direction of incident laser polarization. (C and D) Maps of the total intensity Itotal andangle q between the direction of laser polarization and the armchair direction of the sample, respectively.(B) The intensity maps of the IV and IH components of the SH for a multidomain WSe2-MoS2 junction.(E and F) Maps of the Itotal and q of the area shown in (B). The scale bar for all is 5 mm.

Fig. 3. Photoluminescence and Raman measurements. (A) PL spectra of MoS2 and its cor-responding spatial modulation as shown in the inset contour color map. (B) Spatial maps of the A1g

and E2g peak position. (C) The extracted MoS2 Raman spectra from the WSe2-MoS2 junction, where the colored spectra corresponds to the inset contour colormap in (A). (D and E) The maps showing PL intensity and the spatial modulation of photon energy for a selected heterojunction. (F) Spectra 1 to 9 werecollected from the location marked in (D).

RESEARCH | REPORTSon M

arch 24, 2020

http://science.sciencemag.org/

Dow

nloaded from

Fig. 3A, the location with a higher PL intensityand energy also exhibited higher A1g and E2g

frequencies.According to previous studies on the strain-

dependent E2g and A1g Raman modes in MoS2monolayer (17–20), we conclude that both the PLand Raman variations reflect the local strain dis-tribution in the MoS2. The frequency upshift(downshift) of both Raman modes is associatedwith a compressive (tensile) strain. The spectrafrom the isolated MoS2 and the three corners ofthe triangular heterostructure have an identicalPL energy of 1.86 eV, near that of 1.82 ± 0.02 eVfromunstrainedMoS2 (19). For simplicity, assum-ing the isolated MoS2 is nearly strain free, wecould then map out the relative strain onMoS2;the tensile area is colored with green, cyan, andblack, and the compressive area is colored withred and yellow in the inset of Fig. 3A. The repre-sentative Raman spectra under tensile and com-pressive strain (Fig. 3C) show that theMoS2 withthe lowest PL energy (1.79 eV) was frequentlyobserved from the area with a tensile strain. Asthe compressive strain increases, the PL peakshifted to a higher energy with a pronounced in-tensity increase. Symmetry breaking of the crys-

tal (19, 20) broke the degeneracy in the MoS2 E2g

Raman mode (subpeaks E′– and E′+) for the as-grown TMDC monolayer (black curve).The strain variation likely originated from the

lattice mismatch betweenMoS2 andWSe2. Figure3A shows that the strain of MoS2 adjacent toWSe2 was tensile (cyan color) and then graduallychanged to strain free (blue color), particularlyat the corners. To balance the strain built uponthe MoS2, someMoS2 areas exhibited compres-sive strain (red color). We estimated the strain(relative to the as-grown isolated MoS2) in theMoS2 region of the MoS2-WSe2 heterostructureto be 1.59 ± 0.25% for largest tensile strain and1.1 ± 0.18% for largest compressive strain, basedon the reported linear PL energy shift rate of45 meV/% strain (17–22). Such a large strain dif-ference induced by the lateral heterostructurecould indicate the possibility of using monolayerTMDCs for straintronics (23).In the inner triangular WSe2, the PL spectra

showed a prominent direct band-gap emission at~1.63 eV. By contrast to the outer MoS2 region,the PL energy and intensity in the WSe2 regiondid not show pronounced variations (Fig. 3, Dand E). Additionally, the Raman frequencies of

the WSe2 region were also relatively unchanged(fig. S7). Interestingly, Fig. 3D shows that the PLemission from the heterojunction interface wasstronger and the PL enhancement was localizedat the WSe2 side of the lateral interface. The PLspectra taken fromthepointsmarkedas 1 through9 in Fig. 3D are displayed in Fig. 3F, where twocharacteristic peaks—1.62 eV for WSe2 (points1 to 3) and 1.85 eV for MoS2 (points 7 to 9)—wereobserved, respectively. The line scan was per-formed at a selected MoS2 area nearly free ofstrain. Noticeably, the PL spectrum for WSe2 ad-jacent to the heterojunction (point 3) showed anarrower and stronger peak at 1.62 eV. The WSe2-MoS2 vertical contact forms a type II bandalignment (1, 24, 25), with an interband tran-sition peak at ~1.59 eV. However, we did notdetect any appreciable interband transition inPL measurements.Our observation differs from that for the

WS2-MoS2 heterojunction in a previous report(2), where a broader PL peak with an interme-diate band gap energy, identified as the interbandtransition, was observed. Meanwhile, it was re-ported that the edges of WS2 monolayers exhibitextraordinary PL intensity. (26) Our STEM re-sults showed that interfacial W is bonded toSe and S, respectively, from each side, and theinterface structure is similar to the reportedWS2edges. We performed a separate study on gasphase sulfurization of isolated WSe2 monolayertriangles and found that the PL emission fromWSe2 edgewas largely enhanced after edge sulfuri-zation (fig. S8). The enhancement of PL at theinterface could be related to the chemical compo-sition change in the junction.To study the depletion region of the atomically

sharp WSe2-MoS2 heterojunction, we used scan-ning Kelvin probemicroscopy (SKPM) to directlyextract the spatial distribution of the surface po-tentials. Figure 4A and its inset show the SKPMand atomic force microscopy (AFM) images forthe junction, respectively. The color contrast inthe SKPM image between the WSe2 and MoS2regions revealed the distinct potential differenceacross the junction. SKPM allows the measure-ment of depletion width, but the actual built-inpotential difference is not accurate because it isstrongly affected by the surface adsorbates. Theline profile in fig. S9 reveals that the junctiondepletion width is ~320 nm. The value agreeswith that of the 100 to 500 nm estimated fromthe depletion approximation of solving Poisson’sequation based on the assumptions of no free car-riers and constant dopant concentration in the de-pletion region (see the supplementary materials).To investigate the electrical properties, the as-

grownWSe2-MoS2 heterojunction was transferredonto a SiO2/Si substrate, and two contact metals,Pd andTi/Au, were deposited onWSe2 andMoS2,respectively (see the supplementarymaterials fordetails). Figure 4B shows the OM image of theWSe2-MoS2 heterojunction, and Fig. 4C shows thecurrent-voltage (I-V) curves of the heterojunctionwithout (black) and with (red) white light illumi-nation. The characteristic curve exhibits goodrectification character, with a threshold voltage

SCIENCE sciencemag.org 31 JULY 2015 • VOL 349 ISSUE 6247 527

Fig. 4. Electrical properties of the devices based on heterojunction. (A) SKPM image showing thesurface potential distribution for the WSe2-MoS2 heterostructure. Inset shows the AFM height images,and no obvious height difference is observed from two sides of the junction. (B) Optical image for theWSe2-MoS2 p-n junction device. Scale bar, 4 mm. (C) Electrical transport curves (I versus V) with andwithout light exposure (1 mW/cm2), showing the presence of a p-n junction and photovoltaic effect.(D) Electrical transport curves for Ti-contacted MoS2 and Pd-contacted WSe2 separately.

RESEARCH | REPORTSon M

arch 24, 2020

http://science.sciencemag.org/

Dow

nloaded from

at about 0.9 V under forward bias (fig. S10). Aphotovoltaic effect with an open-circuit voltageVoc of 0.22 V and short-circuit current Isc of 7.7 pAunder white light illumination (power densityEw of 1 mW/cm2) is shown in the inset of Fig. 4C.The nearly symmetric I-V curves andbarely photo-voltaic effect for individualWSe2 (contacted withPd) and MoS2 (contacted with Ti/Au) in Fig. 4Dcorroborate that the p-n junction from the hetero-structure is predominant, rather than the smallSchottky barriers between metal and TMDCs.We calculated the power conversion efficiency

(PCE) of the device with the photon-to-electronconversion equation, PCE= ISCVOCFF/EWAC,whereFF is the fill factor and AC is the effective areawith energy conversion. The FF of 0.39 was ex-tracted from the inset of Fig. 4C. The small FFmight result from the high equivalent series re-sistance of the intrinsic TMDC layers. We esti-mated the maximum AC by considering both thedepletion area of the junction and the adjacentdiffusion area of each TMDC layers, giving riseto a maximum area about 32 mm2 (see the sup-plementary materials for details). The calculatedPCE is at least 0.2%, comparable with the few-layer MoS2 vertical p-n junction (27) and mono-layer lateral WSe2 p-n junction (28). The p-njunction of WSe2-MoS2 is further corroboratedwith the results from another device (fig. S11),where the carrier transport is clearly through theheterojunction interface.The presence of depletion width (320 nm),

rectifying behaviors, photoresponses, and photo-voltaic effects confirms the intrinsic p-n junctionproperties for lateral WSe2-MoS2. The realizationof controlled-edge epitaxy of TMDC opens thedoor to construct other monolayer componentsfor future monolayer electronics.

REFERENCES AND NOTES

1. M.-H. Chiu et al., ACS Nano 8, 9649–9656 (2014).2. Y. Gong et al., Nat. Mater. 13, 1135–1142 (2014).3. S.-H. Su et al., Small 10, 2589–2594 (2014).4. X. Duan et al., Nat. Nanotechnol. 9, 1024–1030 (2014).5. C. Huang et al., Nat. Mater. 13, 1096–1101 (2014).6. Y.-H. Lee et al., Adv. Mater. 24, 2320–2325 (2012).7. J.-K. Huang et al., ACS Nano 8, 923–930 (2014).8. Y. Shi et al., Nano Lett. 12, 2784–2791 (2012).9. O. L. Krivanek et al., Nature 464, 571–574 (2010).10. Y. R. Shen, Annu. Rev. Phys. Chem. 40, 327–350

(1989).11. J. F. McGilp, D. Weaire. C. H. Patterson, Eds., Epioptics:

Linear and Nonlinear Optical Spectroscopy of Surfaces andInterfaces (Springer-Verlag GmbH, Berlin, 1995).

12. G. A. Reider, T. F. Heinz, in Photonic Probes of Surfaces:Electromagnetic Waves, P. Halevi, Eds. (Elsevier ScienceLimited, Amsterdam, 1995), vol. 2, chap. 9.

13. N. Kumar et al., Phys. Rev. B 87, 161403 (2013).14. H. Zeng et al., Sci. Rep. 3, 1608 (2013).15. Y. Li et al., Nano Lett. 13, 3329–3333 (2013).16. W.-T. Hsu et al., ACS Nano 8, 2951–2958 (2014).17. Y. Y. Hui et al., ACS Nano 7, 7126–7131 (2013).18. L. Yang et al., Sci. Rep. 4, 5649 (2014).19. H. J. Conley et al., Nano Lett. 13, 3626–3630 (2013).20. C. R. Zhu et al., Phys. Rev. B 88, 121301 (2013).21. A. Castellanos-Gomez et al., Nano Lett. 13, 5361–5366

(2013).22. Q. Ji et al., Nano Lett. 13, 3870–3877 (2013).23. T. Heine, Acc. Chem. Res. 48, 65–72 (2015).24. H. Fang et al., Proc. Natl. Acad. Sci. U.S.A. 111, 6198–6202

(2014).25. M.-H. Chiu et al., Determination of band alignment in the single

layer MoS2/WSe2 heterojunction. Nature Commun. (2015);available at http://arxiv.org/abs/1406.5137).

26. H. R. Gutiérrez et al., Nano Lett. 13, 3447–3454 (2013).27. H.-M. Li et al., Nat. Commun. 6, 6564 (2015).28. A. Pospischil, M. M. Furchi, T. Mueller, Nat. Nanotechnol. 9,

257–261 (2014).

ACKNOWLEDGMENTS

L.-J.L. acknowledges support from King Abdullah University ofScience and Technology (Saudi Arabia), Ministry of Science andTechnology (MOST) and Taiwan Consortium of EmergentCrystalline Materials (TCECM), Academia Sinica (Taiwan), andAOARD-134137 (USA). Y.-C.L. and K.S. acknowledge support fromthe Japan Science and Technology Agency research acceleration

program. W.-H.C. acknowledges the support from TCECM, MOST ofTaiwan under grant NSC102-2119-M-009-002-MY3 and the Centerfor Interdisciplinary Science of Nation Chiao Tung University.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/349/6247/524/suppl/DC1Materials and MethodsFigs. S1 to S11References (29–38)

30 April 2015; accepted 24 June 201510.1126/science.aab4097

FOREST ECOLOGY

Pervasive drought legacies in forestecosystems and their implicationsfor carbon cycle modelsW. R. L. Anderegg,1,2* C. Schwalm,3,4 F. Biondi,5 J. J. Camarero,6 G. Koch,3 M. Litvak,7

K. Ogle,8 J. D. Shaw,9 E. Shevliakova,10 A. P. Williams,11 A. Wolf,1 E. Ziaco,5 S. Pacala1

The impacts of climate extremes on terrestrial ecosystems are poorly understoodbut important for predicting carbon cycle feedbacks to climate change. Coupledclimate–carbon cycle models typically assume that vegetation recovery from extremedrought is immediate and complete, which conflicts with the understanding of basicplant physiology. We examined the recovery of stem growth in trees after severe droughtat 1338 forest sites across the globe, comprising 49,339 site-years, and compared theresults with simulated recovery in climate-vegetation models. We found pervasive andsubstantial “legacy effects” of reduced growth and incomplete recovery for 1 to 4 yearsafter severe drought. Legacy effects were most prevalent in dry ecosystems, amongPinaceae, and among species with low hydraulic safety margins. In contrast, limited or nolegacy effects after drought were simulated by current climate-vegetation models. Ourresults highlight hysteresis in ecosystem-level carbon cycling and delayed recovery fromclimate extremes.

Anthropogenic climate change is projectedto alter both climate mean conditions andclimate variability, leading tomore frequentand/or intense climate extremes such asheat waves and severe drought (1). Increas-

ing variability is likely to profoundly affect eco-systems, as many ecological processes are more

sensitive to climate extremes than to changes inmean states (2–4). In turn, the impacts of theseextremes can have major effects on ecosystem-level carbon cycling, feeding back to accelerate orlimit climate change. The 2003 European heatwave, for example, led to the development of astrong anomalous carbon source, reversing 4years of carbon uptake by terrestrial ecosystemson a continental scale (5).Forest ecosystems store nearly half of the

carbon found in terrestrial ecosystems (6), butthe fate of forests under climate change and withincreasing climate extremes remains uncertainand controversial. Whereas some studies con-tend that large regions of forest are poised onthe verge of collapse into an alternate state (7–9),others suggest that forests are relatively resilientand likely to experience only modest changes(10–12). The sensitivity of forests to climate ex-tremes has become apparent in global patternsof widespread forest mortality (13), which high-light the possibility that the forest carbon sinkcould be weakened or could even transition rap-idly to a carbon source in some regions (13–15).Thus, the response of forest growth and mortal-ity to extreme drought and heat constitutes a

528 31 JULY 2015 • VOL 349 ISSUE 6247 sciencemag.org SCIENCE

1Department of Ecology and Evolutionary Biology, PrincetonUniversity, Princeton, NJ 08544, USA. 2Department ofBiology, University of Utah, Salt Lake City, UT 84112, USA.3Center for Ecosystem Science and Society, NorthernArizona University, Flagstaff, AZ 86011, USA. 4School ofEarth Sciences and Environmental Sustainability, NorthernArizona University, Flagstaff, AZ 86011, USA. 5DendroLab andGraduate Program of Ecology, Evolution, and ConservationBiology, University of Nevada–Reno, Reno, NV 89557, USA.6Instituto Pirenaico de Ecología, Consejo Superior deInvestigaciones Científicas, Avda. Montañana 1005, 50192Zaragoza, Spain. 7Department of Biology, University of NewMexico, Albuquerque, NM 87131, USA. 8School of LifeSciences, Arizona State University, Tempe, AZ 85287-4501,USA. 9Rocky Mountain Research Station, U.S. Forest Service,Ogden, UT 84401, USA. 10National Oceanic and AtmosphericAdministration (NOAA) Geophysical Fluid DynamicsLaboratory, Princeton, NJ 08540, USA. 11Lamont-DohertyEarth Observatory of Columbia University, 61 Route 9W,Palisades, NY 10964, USA.*Corresponding author. E-mail: anderegg@princeton.edu

RESEARCH | REPORTSon M

arch 24, 2020

http://science.sciencemag.org/

Dow

nloaded from

interface lateral p-n junction with an atomically sharp2-MoS2Epitaxial growth of a monolayer WSe

Kung-Hwa Wei, Jr-Hau He, Wen-Hao Chang, Kazu Suenaga and Lain-Jong LiMing-Yang Li, Yumeng Shi, Chia-Chin Cheng, Li-Syuan Lu, Yung-Chang Lin, Hao-Lin Tang, Meng-Lin Tsai, Chih-Wei Chu,

DOI: 10.1126/science.aab4097 (6247), 524-528.349Science 

, this issue p. 524Sciencetungsten selenide with an atomically sharp boundarystacking the layers, which make a weak junction, but by growing molybdenum sulfide on the edge of a triangle ofp-n heterojunctions between two of these materials, molybdenum sulfide and tungsten selenide. They did this not by

succeeded in makinget al.difficult to make a good voltage-biased (p-n) junction between different TMDC materials. Li have band gaps. Transistors based on these materials can have high ratios of ''on'' to ''off'' currents. However, it is often ''leaky''; that is, they are hard to turn off. Related transition metal dichalcogenides (TMDCs) such as molybdenum sulfideare inherently nanoscale and can carry high currents. However, graphene has no band gap and the transistors are

Two-dimensional materials such as graphene are attractive materials for making smaller transistors because theyElectronic junctions on edge

ARTICLE TOOLS http://science.sciencemag.org/content/349/6247/524

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2015/07/29/349.6247.524.DC1

REFERENCES

http://science.sciencemag.org/content/349/6247/524#BIBLThis article cites 35 articles, 1 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Copyright © 2015, American Association for the Advancement of Science

on March 24, 2020

http://science.sciencem

ag.org/D

ownloaded from