NEWS & VIEWS RESEARCH suffer from climate-related noise. The authors averaged several AIM events to reduce noise, but although obtaining an average delay is informative, the question remains as to what extent individual DO–AIM pairs might vary in timing. Studies of other Antarctic ice cores8 reveal that the variation of temperature with time during AIM events has geographic variability, and that two phases are typically visible in the AIM profile during warming. They are accompanied by variations in other climate tracers that point to atmospheric-circulation changes that are not synchronous with Greenland events. Buizert et al. looked at atmospheric changes, and observed sea-salt variations during AIM that indicate synchronous changes in sea ice and temperature. Such variations may provide insight into changes in southern freshwater forcing and ocean feedbacks that affect the bipolar see-saw. Establishing the relative timing between see-saw events in the two hemispheres is a big step forward, but the full extent of changes revealed by Antarctic ice cores, including the timing of changes in carbon dioxide level,
remains under-exploited. An integrated understanding of hemispheric climate coupling therefore awaits. Nevertheless, Buizert and colleagues’ findings are particularly compelling in the light of recent indications9 of a contemporary slowing of the AMOC, which has been anticipated. Predicting the global effects of such a change will pivot on our understanding of how the hemispheres communicate. ■ Tas van Ommen is in the Australian Antarctic Division at the Antarctic Climate and Ecosystems CRC, University of Tasmania, Hobart, Tasmania 7050, Australia. e-mail: [email protected] 1. EPICA Community Members. Nature 444, 195–198 (2006). 2. WAIS Divide Project Members. Nature 520, 661–665 (2015). 3. Crowley, T. J. Paleoceanography 7, 489–497 (1992). 4. Broecker, W. S. Paleoceanography 13, 119–121 (1998). 5. Stocker, T. F. & Johnsen, S. J. Paleoceanography 18, 1087 (2003). 6. Pedro, J. B. et al. Clim. Past 7, 671–683 (2011). 7. Morgan, V. et al. Science 297, 1862–1864 (2002). 8. Landais, A. et al. Quat. Sci. Rev. 114, 18-32 (2015). 9. Rahmstorf, S. et al. Nature Clim. Change http:// dx.doi.org/10.1038/nclimate2554 (2015).
MAT ERIALS SCIENCE
Semiconductors grown large and thin Atomically thin layers of semiconductors called transition-metal dichalcogenides have been grown uniformly on the square-centimetre scale — paving the way for the ultimate miniaturization of electronic applications. See Letter p.656 TOBIN J. MARKS & MARK C. HERSAM
T
he ubiquity of electronic devices today derives from the development of semiconductor wafers that have exceptional spatial uniformity. These wafers enable the production of highly integrated circuits, because each of the billions of constituent transistors behaves predictably, with differences between individual devices being among the smallest of any manufactured technology. Meanwhile, the miniaturization of transistors over the years has led researchers to consider the ultimate size limit: atomic-scale electronic devices. This limit has been reached in research laboratories with the fabrication of prototypes from atomically thin semiconducting materials1. But integrated circuits can be made from such devices only if atomically thin materials can be grown uniformly over large areas. On page 656 of this issue, Kang et al.2 report a crucial step in this direction. They have achieved such uniformity on the wafer-scale — several square centimetres — for one of the
most promising classes of two-dimensional semiconductor. Transition-metal dichalcogenides (TMDs) have the general formula MX2, in which M is a metal such as molybdenum (Mo) or tungsten (W), and X can be sulfur, selenium or tellurium. They are semiconducting materials with 2D structures consisting of stacked, three-atom-thick X–M–X monolayers bound largely by interlayer van der Waals forces, roughly analogous to the structure of graphite. The appeal of TMDs has conventionally centred on their bulk forms3, for applications as lubricants, energy-storage materials and catalysts. However, intense interest in the properties of atomically thin, 2D electronic materials such as graphene has extended to MX2 materials, because they offer the possibility of high-performance and mechanically flexible transistors, light detectors, solar cells and light-emitting devices. Most studies that have attempted to fabricate devices from atomically thin TMDs have used mechanically exfoliated forms (samples
50 Years Ago In an article in a recent issue of Minerva … on “The President’s Science Advisers”, Dr. P. H. Abelson discusses the use which the Presidents of the United States have made of the service of a ‘Science Adviser’ since 1957. Dr. Abelson is concerned with the part which the Science Adviser and his staff have actually played, including his relation to the Office of Science and Technology created in March 1962, and more particularly he directs attention to some of the limitations of the system … Dr. Abelson goes so far as to maintain that the Science Adviser and his staff have failed to address themselves to many major problems which might be expected to fall within the Science Adviser’s responsibility. Instead he believes they have been occupied with many relatively trivial problems, and the consequent exclusion of such questions as whether in the United States the present allocations of money and manpower has led to some discontent with the advisory system. From Nature 1 May 1965
100 Years Ago The great consumption of petrol as a motor fuel, which last year, in spite of the disturbing element of war, rose to the enormous volume of 120 million gallons in England, and to nearly ten times that amount in America, has led to the attempt being made to add to the natural supply by the so called “cracking” of the heavy residual oils left after the petrol and the lamp oil have been distilled off from the crude oil … The term “cracking” is one of those delightful Americanisms which express so exactly the meaning we wish to impart that it has been adopted universally. From Nature 29April 1915 3 0 A P R I L 2 0 1 5 | VO L 5 2 0 | NAT U R E | 6 3 1
© 2015 Macmillan Publishers Limited. All rights reserved
RESEARCH NEWS & VIEWS
+ 550 °C, H2 26 hours S(CH2CH3)2
Monolayer MS2
M(CO)6
Figure 1 | Large-area growth of atomically thin, layered semiconductors. a, Kang et al.2 report the fabrication of electronically uniform monolayers of semiconductors, MS2 (M is molybdenum or tungsten), on the squarecentimetre scale on silicon wafers coated with silica (SiO2). The process
prepared by stripping off a few atomic layers from the bulk material, for example by using self-adhesive tape)4. Alternatively, irregularly shaped monolayers have been used, or fragments a few layers thick on electrically insulating dielectric substrates4. The next step towards large-scale manufacturing will require practical pathways for coating wafer-scale areas with structurally and electronically uniform MX2 monolayers on diverse substrates. With this goal in mind, the methods of chemical vapour deposition (CVD) and metal–organic chemical vapour deposition (MOCVD) offer the ability to produce largearea coatings without the need for the expensive vacuum equipment required for other approaches. These techniques typically use volatile precursors to deliver the chemicals required to coat heated substrates, sometimes with extra energy provided in the form of a plasma or light. The practicality of CVD and MOCVD processes has led to their use on a vast scale for growing coatings on substrates as diverse as continuous glass sheets, drill bits and optoelectronic devices. So far, MX2 growth using these techniques has focused largely on thick, low-friction coatings, often with toxic and corrosive MF6 precursors5,6 (F is fluorine) and highly toxic hydrogen sulfide (H2S) as a source of sulfur. Related approaches include aerosol-assisted MOCVD7 and passing H2S or sulfur vapour over hot metallic films or metal oxides (MO3) pre-deposited on a substrate8,9. However, it seems unlikely that structurally and electronically uniform MX2 monolayers can be reliably grown over large areas by any of the above approaches. By contrast, Kang et al. describe an MOCVD process for growing MoS2 and WS2 mono layers that have excellent spatial uniformity on silicon wafers coated with silica (SiO2). The process uses commercially available, volatile and relatively safe metal hexacarbonyl compounds, Mo(CO) 6 and W(CO) 6, as precursors of molybdenum and tungsten, respectively, and diethylsulfide (S(CH2CH3)2) as a sulfur precursor (Fig. 1). The authors also added hydrogen gas to the precursor stream
involves exposing heated wafers to precursor compounds — M(CO)6 and diethylsulfide (S(CH2CH3)2) — at high temperatures in the presence of hydrogen gas. Sulfur, yellow; metal atoms, blue; oxygen, red; carbon, dark grey; hydrogen, light grey. Wafers are not shown.
to remove carbonaceous deposits that form during the growth process. In this way, Kang and colleagues prepared MoS2 monolayers across square-centimetre areas, and used them to make arrays of microscopic devices called field-effect transistors with a 99% device yield — only 2 out of 200 transistors characterized by the authors did not conduct. The electron mobility of the devices at room temperature was 30 square centi metres per volt per second — good for TMDs — and the mobility was minimally dependent on the dimensions of the transistors or their location on the wafer. The authors went on to demonstrate that silica can be deposited on the TMD monolayer after the first MOCVD growth step, and that the MOCVD process can then be repeated, resulting in multiple electrically isolated layers of atomically thin TMDs. Three-dimensional electronic architectures can thus be made, offering the enticing possibility of fabricating ultrahigh-density circuits — something that has not been possible using conventional silicon electronics. Although the new findings constitute a substantial advance for atomically thin semiconductors, major issues must still be addressed before practical applications are possible. For example, the optimum growth conditions require a temperature of 550 °C to be maintained for 26 hours. This is too high a temperature to be used with currently available, flexible plastic substrates, and so a lower-temperature process is required, or a method for transferring large-area, atomically thin layers of TMDs from the growth substrate without introducing contamination, defects or wrinkles. The long growth time will probably also be an issue for high-throughput manufacturing. Although the observed electron mobilities are respectable for TMDs, they are still more than ten times smaller than those of bulk crystalline silicon. Further research is also needed to find ways of ‘doping’ the TMDs — for example, by adding trace quantities of impurities — to control the type and concentration of charge carriers. Finally, the threshold voltage of the devices must be tuned to allow the development of low-power
6 3 2 | NAT U R E | VO L 5 2 0 | 3 0 A P R I L 2 0 1 5
© 2015 Macmillan Publishers Limited. All rights reserved
electronic architectures, such as the widely used complementary metal-oxide–semiconductor (CMOS) technology. The authors’ 2D semiconductors open up opportunities for making devices and circuits that are far beyond conventional field-effect transistors. For example, because monolayer TMDs are atomically thin, they do not fully screen electric fields applied perpendicular to the monolayer; this property might allow the development of ‘gate-tunable’ heterojunction diodes for application in high-speed communications circuits10. Moreover, extended defects, such as grain boundaries in monolayer TMDs, can be manipulated using applied voltages. This might lead to the production of gate-tunable memristor devices, which are promising building blocks for ‘non-volatile’ computer memory and for neuromorphic (brain-like) circuit architectures11. Widespread access to uniform, large-area monolayer TMDs will help to accelerate progress in these emerging areas, allowing the full potential of 2D semiconductors to be quickly explored. ■ Tobin J. Marks and Mark C. Hersam are in the Department of Materials Science and Engineering, the Department of Chemistry and the Materials Research Center, Northwestern University, Evanston, Illinois 60208, USA. e-mails: [email protected]; [email protected] 1. Jariwala, D., Sangwan, V. K., Lauhon, L. J., Marks, T. J. & Hersam, M. C. ACS Nano 8, 1102–1120 (2014). 2. Kang, K. et al. Nature 520, 656–660 (2015). 3. Chhowalla, M. et al. Nature Chem. 5, 263–275 (2013). 4. Li, H., Wu, J., Yin, Z. & Zhang, H. Acc. Chem. Res. 47, 1067–1075 (2014). 5. Scharf, T. W., Prasad, S. V., Mayer, T. M., Goeke, R. S. & Dugger M. T. J. Mater. Res. 19, 3443–3446 (2004). 6. Doll, G. L., Mensah, B. A., Mohseni, H. & Scharf, T. W. J. Therm. Spray Tech. 19, 510–516 (2010). 7. McCain, M. N., He, B., Sanati, J., Wang, Q. J. & Marks, T. J. Chem. Mater. 20, 5438–5443 (2008). 8. Lee, Y.-H. et al. Adv. Mater. 24, 2320–2325 (2012). 9. Song, I. et al. Angew. Chem. Int. Edn 53, 1266–1269 (2014). 10. Jariwala, D. et al. Proc. Natl Acad. Sci. USA 110, 18076–18080 (2013). 11. Sangwan, V. K. et al. Nature Nanotechnol. 10, 10.1038/nnano.2015.56 (2015).
remains under-exploited. An integrated understanding of hemispheric climate coupling therefore awaits. Nevertheless, Buizert and colleagues’ findings are particularly compelling in the light of recent indications9 of a contemporary slowing of the AMOC, which has been anticipated. Predicting the global effects of such a change will pivot on our understanding of how the hemispheres communicate. ■ Tas van Ommen is in the Australian Antarctic Division at the Antarctic Climate and Ecosystems CRC, University of Tasmania, Hobart, Tasmania 7050, Australia. e-mail: [email protected] 1. EPICA Community Members. Nature 444, 195–198 (2006). 2. WAIS Divide Project Members. Nature 520, 661–665 (2015). 3. Crowley, T. J. Paleoceanography 7, 489–497 (1992). 4. Broecker, W. S. Paleoceanography 13, 119–121 (1998). 5. Stocker, T. F. & Johnsen, S. J. Paleoceanography 18, 1087 (2003). 6. Pedro, J. B. et al. Clim. Past 7, 671–683 (2011). 7. Morgan, V. et al. Science 297, 1862–1864 (2002). 8. Landais, A. et al. Quat. Sci. Rev. 114, 18-32 (2015). 9. Rahmstorf, S. et al. Nature Clim. Change http:// dx.doi.org/10.1038/nclimate2554 (2015).
MAT ERIALS SCIENCE
Semiconductors grown large and thin Atomically thin layers of semiconductors called transition-metal dichalcogenides have been grown uniformly on the square-centimetre scale — paving the way for the ultimate miniaturization of electronic applications. See Letter p.656 TOBIN J. MARKS & MARK C. HERSAM
T
he ubiquity of electronic devices today derives from the development of semiconductor wafers that have exceptional spatial uniformity. These wafers enable the production of highly integrated circuits, because each of the billions of constituent transistors behaves predictably, with differences between individual devices being among the smallest of any manufactured technology. Meanwhile, the miniaturization of transistors over the years has led researchers to consider the ultimate size limit: atomic-scale electronic devices. This limit has been reached in research laboratories with the fabrication of prototypes from atomically thin semiconducting materials1. But integrated circuits can be made from such devices only if atomically thin materials can be grown uniformly over large areas. On page 656 of this issue, Kang et al.2 report a crucial step in this direction. They have achieved such uniformity on the wafer-scale — several square centimetres — for one of the
most promising classes of two-dimensional semiconductor. Transition-metal dichalcogenides (TMDs) have the general formula MX2, in which M is a metal such as molybdenum (Mo) or tungsten (W), and X can be sulfur, selenium or tellurium. They are semiconducting materials with 2D structures consisting of stacked, three-atom-thick X–M–X monolayers bound largely by interlayer van der Waals forces, roughly analogous to the structure of graphite. The appeal of TMDs has conventionally centred on their bulk forms3, for applications as lubricants, energy-storage materials and catalysts. However, intense interest in the properties of atomically thin, 2D electronic materials such as graphene has extended to MX2 materials, because they offer the possibility of high-performance and mechanically flexible transistors, light detectors, solar cells and light-emitting devices. Most studies that have attempted to fabricate devices from atomically thin TMDs have used mechanically exfoliated forms (samples
50 Years Ago In an article in a recent issue of Minerva … on “The President’s Science Advisers”, Dr. P. H. Abelson discusses the use which the Presidents of the United States have made of the service of a ‘Science Adviser’ since 1957. Dr. Abelson is concerned with the part which the Science Adviser and his staff have actually played, including his relation to the Office of Science and Technology created in March 1962, and more particularly he directs attention to some of the limitations of the system … Dr. Abelson goes so far as to maintain that the Science Adviser and his staff have failed to address themselves to many major problems which might be expected to fall within the Science Adviser’s responsibility. Instead he believes they have been occupied with many relatively trivial problems, and the consequent exclusion of such questions as whether in the United States the present allocations of money and manpower has led to some discontent with the advisory system. From Nature 1 May 1965
100 Years Ago The great consumption of petrol as a motor fuel, which last year, in spite of the disturbing element of war, rose to the enormous volume of 120 million gallons in England, and to nearly ten times that amount in America, has led to the attempt being made to add to the natural supply by the so called “cracking” of the heavy residual oils left after the petrol and the lamp oil have been distilled off from the crude oil … The term “cracking” is one of those delightful Americanisms which express so exactly the meaning we wish to impart that it has been adopted universally. From Nature 29April 1915 3 0 A P R I L 2 0 1 5 | VO L 5 2 0 | NAT U R E | 6 3 1
© 2015 Macmillan Publishers Limited. All rights reserved
RESEARCH NEWS & VIEWS
+ 550 °C, H2 26 hours S(CH2CH3)2
Monolayer MS2
M(CO)6
Figure 1 | Large-area growth of atomically thin, layered semiconductors. a, Kang et al.2 report the fabrication of electronically uniform monolayers of semiconductors, MS2 (M is molybdenum or tungsten), on the squarecentimetre scale on silicon wafers coated with silica (SiO2). The process
prepared by stripping off a few atomic layers from the bulk material, for example by using self-adhesive tape)4. Alternatively, irregularly shaped monolayers have been used, or fragments a few layers thick on electrically insulating dielectric substrates4. The next step towards large-scale manufacturing will require practical pathways for coating wafer-scale areas with structurally and electronically uniform MX2 monolayers on diverse substrates. With this goal in mind, the methods of chemical vapour deposition (CVD) and metal–organic chemical vapour deposition (MOCVD) offer the ability to produce largearea coatings without the need for the expensive vacuum equipment required for other approaches. These techniques typically use volatile precursors to deliver the chemicals required to coat heated substrates, sometimes with extra energy provided in the form of a plasma or light. The practicality of CVD and MOCVD processes has led to their use on a vast scale for growing coatings on substrates as diverse as continuous glass sheets, drill bits and optoelectronic devices. So far, MX2 growth using these techniques has focused largely on thick, low-friction coatings, often with toxic and corrosive MF6 precursors5,6 (F is fluorine) and highly toxic hydrogen sulfide (H2S) as a source of sulfur. Related approaches include aerosol-assisted MOCVD7 and passing H2S or sulfur vapour over hot metallic films or metal oxides (MO3) pre-deposited on a substrate8,9. However, it seems unlikely that structurally and electronically uniform MX2 monolayers can be reliably grown over large areas by any of the above approaches. By contrast, Kang et al. describe an MOCVD process for growing MoS2 and WS2 mono layers that have excellent spatial uniformity on silicon wafers coated with silica (SiO2). The process uses commercially available, volatile and relatively safe metal hexacarbonyl compounds, Mo(CO) 6 and W(CO) 6, as precursors of molybdenum and tungsten, respectively, and diethylsulfide (S(CH2CH3)2) as a sulfur precursor (Fig. 1). The authors also added hydrogen gas to the precursor stream
involves exposing heated wafers to precursor compounds — M(CO)6 and diethylsulfide (S(CH2CH3)2) — at high temperatures in the presence of hydrogen gas. Sulfur, yellow; metal atoms, blue; oxygen, red; carbon, dark grey; hydrogen, light grey. Wafers are not shown.
to remove carbonaceous deposits that form during the growth process. In this way, Kang and colleagues prepared MoS2 monolayers across square-centimetre areas, and used them to make arrays of microscopic devices called field-effect transistors with a 99% device yield — only 2 out of 200 transistors characterized by the authors did not conduct. The electron mobility of the devices at room temperature was 30 square centi metres per volt per second — good for TMDs — and the mobility was minimally dependent on the dimensions of the transistors or their location on the wafer. The authors went on to demonstrate that silica can be deposited on the TMD monolayer after the first MOCVD growth step, and that the MOCVD process can then be repeated, resulting in multiple electrically isolated layers of atomically thin TMDs. Three-dimensional electronic architectures can thus be made, offering the enticing possibility of fabricating ultrahigh-density circuits — something that has not been possible using conventional silicon electronics. Although the new findings constitute a substantial advance for atomically thin semiconductors, major issues must still be addressed before practical applications are possible. For example, the optimum growth conditions require a temperature of 550 °C to be maintained for 26 hours. This is too high a temperature to be used with currently available, flexible plastic substrates, and so a lower-temperature process is required, or a method for transferring large-area, atomically thin layers of TMDs from the growth substrate without introducing contamination, defects or wrinkles. The long growth time will probably also be an issue for high-throughput manufacturing. Although the observed electron mobilities are respectable for TMDs, they are still more than ten times smaller than those of bulk crystalline silicon. Further research is also needed to find ways of ‘doping’ the TMDs — for example, by adding trace quantities of impurities — to control the type and concentration of charge carriers. Finally, the threshold voltage of the devices must be tuned to allow the development of low-power
6 3 2 | NAT U R E | VO L 5 2 0 | 3 0 A P R I L 2 0 1 5
© 2015 Macmillan Publishers Limited. All rights reserved
electronic architectures, such as the widely used complementary metal-oxide–semiconductor (CMOS) technology. The authors’ 2D semiconductors open up opportunities for making devices and circuits that are far beyond conventional field-effect transistors. For example, because monolayer TMDs are atomically thin, they do not fully screen electric fields applied perpendicular to the monolayer; this property might allow the development of ‘gate-tunable’ heterojunction diodes for application in high-speed communications circuits10. Moreover, extended defects, such as grain boundaries in monolayer TMDs, can be manipulated using applied voltages. This might lead to the production of gate-tunable memristor devices, which are promising building blocks for ‘non-volatile’ computer memory and for neuromorphic (brain-like) circuit architectures11. Widespread access to uniform, large-area monolayer TMDs will help to accelerate progress in these emerging areas, allowing the full potential of 2D semiconductors to be quickly explored. ■ Tobin J. Marks and Mark C. Hersam are in the Department of Materials Science and Engineering, the Department of Chemistry and the Materials Research Center, Northwestern University, Evanston, Illinois 60208, USA. e-mails: [email protected]; [email protected] 1. Jariwala, D., Sangwan, V. K., Lauhon, L. J., Marks, T. J. & Hersam, M. C. ACS Nano 8, 1102–1120 (2014). 2. Kang, K. et al. Nature 520, 656–660 (2015). 3. Chhowalla, M. et al. Nature Chem. 5, 263–275 (2013). 4. Li, H., Wu, J., Yin, Z. & Zhang, H. Acc. Chem. Res. 47, 1067–1075 (2014). 5. Scharf, T. W., Prasad, S. V., Mayer, T. M., Goeke, R. S. & Dugger M. T. J. Mater. Res. 19, 3443–3446 (2004). 6. Doll, G. L., Mensah, B. A., Mohseni, H. & Scharf, T. W. J. Therm. Spray Tech. 19, 510–516 (2010). 7. McCain, M. N., He, B., Sanati, J., Wang, Q. J. & Marks, T. J. Chem. Mater. 20, 5438–5443 (2008). 8. Lee, Y.-H. et al. Adv. Mater. 24, 2320–2325 (2012). 9. Song, I. et al. Angew. Chem. Int. Edn 53, 1266–1269 (2014). 10. Jariwala, D. et al. Proc. Natl Acad. Sci. USA 110, 18076–18080 (2013). 11. Sangwan, V. K. et al. Nature Nanotechnol. 10, 10.1038/nnano.2015.56 (2015).
