RESEARCH ARTICLE is a natural consequence of the longitude of be created by Prometheus and could be M.C Lewis, G.R. Stewart, Astron. J. 120, 3295(2000) observation, and the spreading by Keplerian served in an extended Cassini mission. shear explains also why they all appear 16. R. A. Kolvoord, ]. A Burns, M. R. Showalter, Nature References and notes concentric. In the second study, several clumps were tracked around Saturn, indicating a full 2 C D. Murray. M K Gordon, S M G. winter, Icarus 19. L.N. Pirce. t blished d, 1 (2004) standard deviation of 45 km). The observations 3. M R. Showalter, icarus 171, 356(2004) 4. A.S. Bosh, C B Olkin, R G. French, P D. Nicholson reported here show that the strands, organized 15757(2002) 21. We thank H as a rotating spiral, have a wider range of 5. CC Porco et al., Science 307, 1226(2005) d discussions. We also acknowledge the work of possible that the tracked clumps were only the 8. C D. Murray, S M G. Winter, Nature 380, 139(1996) he CICLOPS operations group in making the obser- brightest ones, naturally located closer to the 9. C D. Murray et al. Nature 437, 1326(20 vations described here possibl core in the spiral model. Showalter, in Supporting Online Material By the end of 2009, Prometheus and the terplanetary Dust, E. Grun, B. A. S. Gustafson, S F .sciencemag org/cgi/content/fulL/310/5752/1300/ F ring will be in a close-encounter configura- tion because of the precession of their orbits M.R.Showalter, J.B. Pollack, M. E Ockert, L. Doyle, Figs. S1 to $3 M. R. Showalter, Science 282, 1099(1998). resulting from Saturns oblateness (2, 8). It is 26 August 2005: accepted 19 October 2005 very probable that additional spirals will then 13. D P. Hamilton, J.A. Burns, Nature 365, 498(1993) REPORTS size, spacing, and number of modulated Encoding Electronic Properties by Synthesis of Axial Modulation al electronics could be facilitated by using nthesis to define the aspects of transistors Doped silicon Nanowires that are currently enabled by lithographic and ion-beam processing, such as feature uniform- ity and controlled doping. For example, the Chen Yang, Zhaohui Zhong, Charles M Liebert high sensitivity of carbon nanotubes to ad- sorbed gases and solid coatings, along with We describe the successful synthesis of modulation-doped silicon nanowires by lithographic patterning, has been exploited in achieving pure axial elongation without radial overcoating during the growth (9, 10). Greater process. Scanning icroscopy shows that the key properties of the mod circuit assembly could be afforded by the abil- ulated structures--including the number, size, and period of the differentially ity to create semiconductor nanowires that are loped regionsare defined in a controllable manner during synthesis, and more niform in shape and that can be doped se- er, that feature sizes to less than 50 nanometers are possible. Electronic devices lectively along their length, in that the for- fabricated with designed modulation-doped nanowire structures demonstrate mation of regions with different electronic synthesis rather than created by conventional lithography-based technip d6s? their potential for lithography-independent address decoders and tunal properties would be intrinsic to nanowire syn- coupled quantum dots in which changes in electronic properties are encoded by hesis and would not require intermediate litho- graphic patterning and/or electrical contacts Many of the wiring steps normally created by A wide range of nanoscale electronic and for encoding key features or information are lithography can be encoded by varying the photonic devices have been made with carbon needed. doping sequence of the nanowires so that the nanotube and nanowire functional elements Modulation of the composition has been only postfabrication lithographic steps would (1-4). Although the nanomaterials are impor- demonstrated recently in relatively simple be those involved in making extemal input and tant for achieving observed functional proper- nanorod and nanowire structures to yield output contacts to individual nanowires ties in these nanodevices, many of the most functional structures (5-8). For example, gold Synthesis of dopant-modulated nanowire critical features have been defined with the use grown on the tips of cadmium selenide structures in which function can be predicted of similar lithographic approaches that drive nanorods provides specific points for self- on the basis of the encoded axial sequence and ultimately limit the planar semiconductor assembly and electrical contact (5). Modu- doping is challenging: It requires effectively industry. The current dependence on lithogra- lation of the dopant or composition of pure axial or one-dimensional (ID) growth phy thus could reduce advantages of these nanowires during synthesis also has been without simultaneous radial or 2D growth nanoscale elements in proposed applications used to define functional p-type/n-type di-(Fig 1A), because even a few atomic layers of and suggests that nonlithographic approaches odes(6) and single quantum dots(8). These dopant deposited on the surface of a nanowire studies show the potential for synthesis to can dominate its overall electronic properties define function without lithography, yet the (ID). In the metal nanocluster-catalyzed vapor level of information and function encoded liquid-solid growth process(3-5), which has Harvard University. Cambridge, MA 02138, USA. in the materials has been very limited. We been widely used to prepare nanowires, the low describe selective dopant modulation dopant must be added exclusively at the nano +To whom correspondence should be addressed. during the growth of silicon nanowires cluster catalyst without reaction and deposition E-mail: cml@cmliris harvarded with essentially complete control over the at the much larger area of the exposed solid 1304 25NovemBer2005Vol310ScieNcewww.sciencemag.org
is a natural consequence of the longitude of observation, and the spreading by Keplerian shear explains also why they all appear concentric. In the second study, several clumps were tracked around Saturn, indicating a full 90-km range in semimajor axes (with a standard deviation of 45 km). The observations reported here show that the strands, organized as a rotating spiral, have a wider range of semimajor axes (300 km); however, it may be possible that the tracked clumps were only the brightest ones, naturally located closer to the core in the spiral model. By the end of 2009, Prometheus and the F ring will be in a close-encounter configuration because of the precession of their orbits resulting from Saturn’s oblateness (2, 8). It is very probable that additional spirals will then be created by Prometheus and could be observed in an extended Cassini mission. References and Notes 1. B. A. Smith et al., Science 212, 163 (1981). 2. C. D. Murray, M. K. Gordon, S. M. G. Winter, Icarus 129, 304 (1997). 3. M. R. Showalter, Icarus 171, 356 (2004). 4. A. S. Bosh, C. B. Olkin, R. G. French, P. D. Nicholson, Icarus 157, 57 (2002). 5. C. C. Porco et al., Science 307, 1226 (2005). 6. P. D. Nicholson et al., Science 272, 509 (1996). 7. M. R. Showalter, J. A. Burns, Icarus 52, 526 (1982). 8. C. D. Murray, S. M. G. Winter, Nature 380, 139 (1996). 9. C. D. Murray et al., Nature 437, 1326 (2005). 10. J. A. Burns, D. P. Hamilton, M. R. Showalter, in Interplanetary Dust, E. Gru¨n, B. A. S. Gustafson, S. F. Dermott, H. Fechtig, Eds. (Springer-Verlag, Berlin, 2001), pp. 641–725. 11. M. R. Showalter, Science 282, 1099 (1998). 12. M. R. Showalter, J. B. Pollack, M. E. Ockert, L. Doyle, J. B. Dalton, Icarus 100, 394 (1992). 13. D. P. Hamilton, J. A. Burns, Nature 365, 498 (1993). 14. M. C. Lewis, G. R. Stewart, Astron. J. 120, 3295 (2000). 15. S. F. Dermott, Nature 290, 454 (1981). 16. R. A. Kolvoord, J. A. Burns, M. R. Showalter, Nature 345, 695 (1990). 17. J. Ha¨nninen, Icarus 103, 104 (1993). 18. C. C. Porco et al., IAU Circ. 8432, 1 (2004). 19. J. N. Spitale, unpublished data. 20. F. Poulet, B. Sicardy, P. D. Nicholson, E. Karkoschka, J. Caldwell, Icarus 144, 135 (2000). 21. We thank H. Throop, J. Cuzzi, J. Decriem, C. Ferrari, M. Hedman, C. Murray, P. Nicholson, M. Tiscareno, and three anonymous referees for useful comments and discussions. We also acknowledge the work of the CICLOPS operations group in making the observations described here possible. Supporting Online Material www.sciencemag.org/cgi/content/full/310/5752/1300/ DC1 Table S1 Figs. S1 to S3 26 August 2005; accepted 19 October 2005 10.1126/science.1119387 Encoding Electronic Properties by Synthesis of Axial ModulationDoped Silicon Nanowires Chen Yang,1 * Zhaohui Zhong,1 * Charles M. Lieber1,2. We describe the successful synthesis of modulation-doped silicon nanowires by achieving pure axial elongation without radial overcoating during the growth process. Scanning gate microscopy shows that the key properties of the modulated structures—including the number, size, and period of the differentially doped regions—are defined in a controllable manner during synthesis, and moreover, that feature sizes to less than 50 nanometers are possible. Electronic devices fabricated with designed modulation-doped nanowire structures demonstrate their potential for lithography-independent address decoders and tunable, coupled quantum dots in which changes in electronic properties are encoded by synthesis rather than created by conventional lithography-based techniques. A wide range of nanoscale electronic and photonic devices have been made with carbon nanotube and nanowire functional elements (1–4). Although the nanomaterials are important for achieving observed functional properties in these nanodevices, many of the most critical features have been defined with the use of similar lithographic approaches that drive and ultimately limit the planar semiconductor industry. The current dependence on lithography thus could reduce advantages of these nanoscale elements in proposed applications and suggests that nonlithographic approaches for encoding key features or information are needed. Modulation of the composition has been demonstrated recently in relatively simple nanorod and nanowire structures to yield functional structures (5–8). For example, gold grown on the tips of cadmium selenide nanorods provides specific points for selfassembly and electrical contact (5). Modulation of the dopant or composition of nanowires during synthesis also has been used to define functional p-type/n-type diodes (6) and single quantum dots (8). These studies show the potential for synthesis to define function without lithography, yet the level of information and function encoded in the materials has been very limited. We now describe selective dopant modulation during the growth of silicon nanowires with essentially complete control over the size, spacing, and number of modulated regions. Applications of nanowires in conventional electronics could be facilitated by using synthesis to define the aspects of transistors that are currently enabled by lithographic and ion-beam processing, such as feature uniformity and controlled doping. For example, the high sensitivity of carbon nanotubes to adsorbed gases and solid coatings, along with lithographic patterning, has been exploited in transistor structures (9, 10). Greater ease of circuit assembly could be afforded by the ability to create semiconductor nanowires that are uniform in shape and that can be doped selectively along their length, in that the formation of regions with different electronic properties would be intrinsic to nanowire synthesis and would not require intermediate lithographic patterning and/or electrical contacts. Many of the wiring steps normally created by lithography can be encoded by varying the doping sequence of the nanowires so that the only postfabrication lithographic steps would be those involved in making external input and output contacts to individual nanowires. Synthesis of dopant-modulated nanowire structures in which function can be predicted on the basis of the encoded axial sequence of doping is challenging: It requires effectively pure axial or one-dimensional (1D) growth without simultaneous radial or 2D growth (Fig. 1A), because even a few atomic layers of dopant deposited on the surface of a nanowire can dominate its overall electronic properties (11). In the metal nanocluster-catalyzed vaporliquid-solid growth process (3–5), which has been widely used to prepare nanowires, the dopant must be added exclusively at the nanocluster catalyst without reaction and deposition at the much larger area of the exposed solid REPORTS R ESEARCH A RTICLE 1 Department of Chemistry and Chemical Biology, 2 Division of Engineering and Applied Sciences, Harvard University, Cambridge, MA 02138, USA. *These authors contributed equally to this work. .To whom correspondence should be addressed. E-mail: cml@cmliris.harvard.edu 1304 25 NOVEMBER 2005 VOL 310 SCIENCE www.sciencemag.org
REPORTS of the growing nanowire in order to deposition down to the atomic level, we tronic properties associated with modulated le electronic function. Simultaneous ra- local substrate heater (versus a tube dopant concentration, we used scanning gate dial growth with axial elongation, which would reactor) and carried out growth in a obscure modulation of electronic properties H, atmosphere (13), which suppresses the surements, a conducting atomic force mi along the nanowire axis, is likely more com- decomposition of silane and inhibit roscopy probe with an applied voltage (in) mon than recognized (In)and is apparent in tion on the nanowire surface(14, 15 inctions as a spatially localized gate, which extreme cases as visible tapering seen in nano- mission electron microscopy (TEM nables the conductance of the nanowire wire structures modulation-doped n+-nl-nt silicon Studiesof with fabricated source and drain contacts to In nanocluster-catalyzed growth of Si (Fig 1B), where nt and n represent the heavi- be varied and mapped (13). Data acquired nanowires with silane as a reactant, homoge- ly and lightly n-type regions, respectively, from a nt-nl-ht nanowire device with a 555- neous gas-phase decomposition produces re- show that the nanowires prepared in this way nm-long n-type segment (length based on ctive species that can lead to uncontrolled have uniform diameters for lengths of >10 growth time)exhibits localized vari d deleterious radial overcoating of the nano- um High-resolution TEM analysis of the op- conductance between the source and drain wire during elongation (11, 12). To control posite ends of a representative nt-ll-nt silicon 1, C and D). When Vn=-9 V the con- anowire(Fig. 1B) further demonstrate that ductance of this region is reduced, and when the diameters are 17. 4 and 17.1 nm. This 0.3- in=+9 V the conductance is enhanced. The m variation is on the order of a single atomic changes in conductance are consistent wit layer and shows that radial overcoating has the expected depletion or accumulation of these modulation-doped structures. moreover, the -525-nm length of this region To determine whether the n*+ nano- agrees well with that expected of 555 nm. No wires exhibit expected variations in the elec- SGM response was observed from other re t t gions of the nanowire, and is consistent with heavily doped n+ regions connected to the ource and drain in the nt-ll-nt device. These n data thus confirm that our synthetic approad yields spatially and electrically well defined nt-nl-nt nanowire structures We explored limits to these modulated nanowires by preparing and characterizing tructures of the general form n+-(n-n+) where N is the number of repeat units. Rep- sentative Sgm data(Fig. 2, A to C) show .In that N and the repeat spacing can be varied ver a wide range. Specifically, the SGM results show n*-(n-n)x modulated struc 3E tures with N=3, 6, and 8 and repeat spac- ings of 3. 20, 1.55, and 0. 75 um, respectively The observed number of repeats and spacin etween repeats agrees well with that de signed through synthesis (3.12, 1.56, and 0.78 Growth Time(min) these modulated structures, we also prepared structures with variable repeat lengths, where the separation between fixed-length n-type seg ments is varied by the n growth time. The H SgM data show that the designed n= 5 struc- Fig. 1. Synthesis and characterization of dulation-doped nanowires. (A)Illustrations ture has separations of 0.80, 1.15, 1.85, and of (1)pure axial and(2 )simultaneous axial an 3.43 um, which are consistent with the vari- adial growth occurring during gold nanocluster ation in growth times( Fig. 2D). These data and 2. Scalable synthesis of modulation-doped measurements from a number of additional wires. SGM images of n+-(n-n*)N nano- samples(15) have been summarized as a plot deposition of high dopant concentration material wires where (A)N=3,(B)N=6, and (C)N=8, of n-n+ length versus growth time(Fig. 2E) and demonstrate the linear dependence that is odulation-doped silicon nanowire. Scale age of N=5 nanowire, where essential for predictable dopant modulation during synthesi ages recorded at the two ends indicated by red tions are 0.5, 1, 2, and 4 min Scale bars, 1 The potential for scaling n-nlt features to boxes. Scale bar, 10 nm. (C and D)SGM images of(E)Repeat spacing versus growth time at total smaller sizes was also probed by reducing a nt-m-nt modulation-doped nanowire recorded pressure of 320 torr.(F)Growth rate versus the nanowire growth rate. We found that the with a Vip of-9 and +9 v, respectively, and growth pressure. SCM images of n*-(n-n) growth rate decreased linearly to-2.9 nm/s V=1 V. The dark and bright regions cor- growth po of (c)160 torr and(H)80 torr. when the pressure was reduced from 320 to 80 respond to reduced and enhanced conduct- total pr The growth time for each n and n+ region is 15s. torm(Fig. 2F). Notably, reproducible N=3 highlight the positions of source-drain elec cale bars, 100 nm. Error bars in(E)and(F)show structures prepared at 160 and 80 torr ex means±SD. hibited average repeat spacings of 180 nm www.sciencemag.orgSciEnceVol31025NovemBer2005
surface of the growing nanowire in order to encode electronic function. Simultaneous radial growth with axial elongation, which would obscure modulation of electronic properties along the nanowire axis, is likely more common than recognized (11) and is apparent in extreme cases as visible tapering seen in nanowire structures. In nanocluster-catalyzed growth of Si nanowires with silane as a reactant, homogeneous gas-phase decomposition produces reactive species that can lead to uncontrolled and deleterious radial overcoating of the nanowire during elongation (11, 12). To control radial deposition down to the atomic level, we used a local substrate heater (versus a tube furnace reactor) and carried out growth in a H2 atmosphere (13), which suppresses the decomposition of silane and inhibits deposition on the nanowire surface (14, 15). Transmission electron microscopy (TEM) studies of modulation-doped nþ-n-nþ silicon nanowires (Fig. 1B), where nþ and n represent the heavily and lightly n-type regions, respectively, show that the nanowires prepared in this way have uniform diameters for lengths of 910 mm. High-resolution TEM analysis of the opposite ends of a representative nþ-n-nþ silicon nanowire (Fig. 1B) further demonstrate that the diameters are 17.4 and 17.1 nm. This 0.3- nm variation is on the order of a single atomic layer and shows that radial overcoating has been effectively eliminated during growth of these modulation-doped structures. To determine whether the nþ-n-nþ nanowires exhibit expected variations in the electronic properties associated with modulated dopant concentration, we used scanning gate microscopy (SGM) (6, 13, 16). In these measurements, a conducting atomic force microscopy probe with an applied voltage (Vtip) functions as a spatially localized gate, which enables the conductance of the nanowire with fabricated source and drain contacts to be varied and mapped (13). Data acquired from a nþ-n-nþ nanowire device with a 555- nm-long n-type segment (length based on growth time) exhibits localized variations in conductance between the source and drain (Fig. 1, C and D). When Vtip 0 –9 V the conductance of this region is reduced, and when Vtip 0 þ9 V the conductance is enhanced. The changes in conductance are consistent with the expected depletion or accumulation of carriers in the lightly doped n-type region, and moreover, the È525-nm length of this region agrees well with that expected of 555 nm. No SGM response was observed from other regions of the nanowire, and is consistent with heavily doped nþ regions connected to the source and drain in the nþ-n-nþ device. These data thus confirm that our synthetic approach yields spatially and electrically well defined nþ-n-nþ nanowire structures. We explored limits to these modulated nanowires by preparing and characterizing structures of the general form nþ-(n-nþ)N, where N is the number of repeat units. Representative SGM data (Fig. 2, A to C) show that N and the repeat spacing can be varied over a wide range. Specifically, the SGM results show nþ-(n-nþ)N modulated structures with N 0 3, 6, and 8 and repeat spacings of 3.20, 1.55, and 0.75 mm, respectively. The observed number of repeats and spacing between repeats agrees well with that designed through synthesis (3.12, 1.56, and 0.78 mm, respectively). To further explore the synthetic control of these modulated structures, we also prepared structures with variable repeat lengths, where the separation between fixed-length n-type segments is varied by the nþ growth time. The SGM data show that the designed N 0 5 structure has separations of 0.80, 1.15, 1.85, and 3.43 mm, which are consistent with the variation in growth times (Fig. 2D). These data and measurements from a number of additional samples (915) have been summarized as a plot of n-nþ length versus growth time (Fig. 2E) and demonstrate the linear dependence that is essential for predictable dopant modulation during synthesis. The potential for scaling n-nþ features to smaller sizes was also probed by reducing the nanowire growth rate. We found that the growth rate decreased linearly to È2.9 nm/s when the pressure was reduced from 320 to 80 torr (Fig. 2F). Notably, reproducible N 0 3 structures prepared at 160 and 80 torr exhibited average repeat spacings of 180 nm Fig. 1. Synthesis and characterization of modulation-doped nanowires. (A) Illustrations of (1) pure axial and (2) simultaneous axial and radial growth occurring during gold nanocluster (yellow) catalyzed nanowire synthesis. Simultaneous radial growth (2) leads to undesirable deposition of high dopant concentration material over the entire nanowire. (B) (Top) Schematic and low-resolution TEM image of a representative nþ- n-nþ modulation-doped silicon nanowire. Scale bar, 500 nm. (Bottom) High-resolution TEM images recorded at the two ends indicated by red boxes. Scale bar, 10 nm. (C and D) SGM images of a nþ-n-nþ modulation-doped nanowire recorded with a Vtip of –9 and þ9 V, respectively, and Vsd 0 1 V. The dark and bright regions correspond to reduced and enhanced conductance, respectively. The white dashed lines highlight the positions of source-drain electrodes and nanowire. Scale bar, 1 mm. Fig. 2. Scalable synthesis of modulation-doped nanowires. SGM images of nþ-(n-nþ)N nanowires where (A) N 0 3, (B) N 0 6, and (C) N 0 8, and the growth times for the n/nþ regions are 1/3, 1/1, and 0.5/0.5 min, respectively. (D) SGM image of N 0 5 nanowire, where the growth time for the n regions is 0.5 min, and nþ sections are 0.5, 1, 2, and 4 min. Scale bars, 1 mm. (E) Repeat spacing versus growth time at total pressure of 320 torr. (F) Growth rate versus growth pressure. SGM images of nþ-(n-nþ)3 modulation-doped nanowires synthesized with total pressures of (G) 160 torr and (H) 80 torr. The growth time for each n and nþ region is 15 s. Scale bars, 100 nm. Error bars in (E) and (F) show means T SD. www.sciencemag.org SCIENCE VOL 310 25 NOVEMBER 2005 1305 R EPORTS
REPORTS Fig 2G)and 90 nm(Fig. 2H), respectively, sistor is easily turned off by gate 2(1 V), 3B, we took advantage of stochastic end-to-end where the average section length in each whereas the other gates produce only sma owire alignment to produce distinct codes structure was only 90 and 45 nm, respec- conductance changes. Because the three gates from a single type of nanowire. Indeed, the tively. These values do not represent a lower are fabricated at the same time in a parallel general case of a stochastic addressing wit limit of the feature size. Because dopant dif- process, the observed address selectivity is in- modulation-doped nanowires has been ana fusion is negligible at the growth tempera- trinsic to the differences in dopant concentration lyzed and shown to be an efficient approach ture(17), it should be possible to reduce the of n and n+ sections of the Si nanowire and is for addressing dense nanoscale arrays; that is, it feature size to at least that of the diameter of distinct from lithographically defined steps requires-22In(M) nanowires for addressing N the nanowire (6), which for molecular-scale previously used (19) to create differential re- lines when N is large(20) Si nanowires (18)is 3 to 5 nm. sponses in a specific nanowire region. Control of the size and separation of The substantial differences in nanowire We have extended this basic approach to modulation-doped regions also enables syn- conductance resulting from locally gating n arrays(Fig. 3B). Two metal top gates, Inl thesis to define quantum dot(QD) structures, and nt segments can be used as a general and In2, were deposited directly on two in which the Fermi level offset caused by method for addressing individual nanowires modulation-doped silicon nanowires config. variations in dopant concentration produces in an array when the nanowires have dif- ured as outputs, Outl and Out2 Conductance potential barriers confining the Qd(Fig 4A) ferent n/nt sequences or codes. To test this versus applied gate voltage (.)data measured Conductance versus v and bias voltage(l irst fabricated nt-n-nt Si nanowire at the four cross points show that OutI and studies of n*t-n-nopt-l-nt modulation-doped with three equally spaced top Out2 can only be tumed off by Inl and In2, silicon nanowires--in which the lightly doped metal gates (Fig. 3A). The nanowire tran- respectively (fig. SI), and thus, these inputs can n-type regions define barriers for a variable selectively address Outl and Out2(Fig. 3B); length QD, nop+--reveal well-defined dia- A800 that is, the array functions as an address de- mond structures(Fig. 4B); the single-period coder circuit for multiplexing and demultiplex- diamond shows that transport occurs through ing signals. a key point in our approach is that a single QD structure(21). Notably, the lithography is only used to define a regular geometry-dependent gate capacitance, Ce, de- array of microscale gate wires and is not termined from these data, 23.5 af, agrees well needed to create a specific address code at the with the value, 24. 1 aF, calculated from the nanoscale as in previous work(19). Because -500-nm QD size determined by SGM imag- synthesis is used to define the code required for ing(Fig. 4B, inset). In addition, current anoscale addressing(not lithography ) we call versus v data for this nanowire and a this a"lithography-independent" addressing structure in which the nont section is reduced scheme. In the demonstration example of Fig. by half to -250 nm(Fig. 4C) show single- Vg (v) Fig. 4. QD structures 多燃 QD structure confined by two n-type barriers 0.15 0.20 Out2 0.2 conduction band (E) fset of the nt and n 之 sections induce tun- 出三区 40202 Out1 is <u, the char D ergy (21). The red X indicates blockage of Out2 charge transport. (B) 0.10 0.140.16 n+ device. the blue Fig 3. Modulation-doped address decoder. (A)/ w values of alav versus v, (d= 1 v) measured for and the red regions 136 (inset) correspond to blue, black, and red curves, ues, the red cold espectively. The native silicon oxide was used a gate dielectric with Au metal gates.(Inset Scanning electron microscopy(SEM) image of the device. Scale bar, 1 um.( B) SEM image of a (i wwwww The middle nop+ and two n sections w for 3 and 0.5 2-by-2 decoder configured using two modulation- 0.15 0. 21 min, respectively, at 80 Out2) and two Au metal gates, which were age of the same device. deposited over a uniform Si,N, dielectric as in- Scale bar, 200 nm.(C)I-v, data taken at 1.5 in (B)(blue curve) and another device (In1 and In2). Scale bar, 1 um. Plots of input with the nop+t section grown for 1.5 blue) and output(red voltages for the 2-by-2 double-QD structure with variable-width n the two QDs(Right)/ data oder. Supply voltage is-2 V recorded at 1.5 K on three devices with n, sections grown for 15, 10, and 5 s(top to bottom) 1306 25NovemBer2005Vol310ScieNcewww.sciencemag.org
(Fig. 2G) and 90 nm (Fig. 2H), respectively, where the average section length in each structure was only 90 and 45 nm, respectively. These values do not represent a lower limit of the feature size. Because dopant diffusion is negligible at the growth temperature (17), it should be possible to reduce the feature size to at least that of the diameter of the nanowire (6), which for molecular-scale Si nanowires (18) is 3 to 5 nm. The substantial differences in nanowire conductance resulting from locally gating n and nþ segments can be used as a general method for addressing individual nanowires in an array when the nanowires have different n/nþ sequences or codes. To test this idea, we first fabricated nþ-n-nþ Si nanowire transistors with three equally spaced top metal gates (Fig. 3A). The nanowire transistor is easily turned off by gate 2 (–1 V), whereas the other gates produce only small conductance changes. Because the three gates are fabricated at the same time in a parallel process, the observed address selectivity is intrinsic to the differences in dopant concentration of n and nþ sections of the Si nanowire and is distinct from lithographically defined steps previously used (19) to create differential responses in a specific nanowire region. We have extended this basic approach to arrays (Fig. 3B). Two metal top gates, In1 and In2, were deposited directly on two modulation-doped silicon nanowires configured as outputs, Out1 and Out2. Conductance versus applied gate voltage (Vg) data measured at the four cross points show that Out1 and Out2 can only be turned off by In1 and In2, respectively (fig. S1), and thus, these inputs can selectively address Out1 and Out2 (Fig. 3B); that is, the array functions as an address decoder circuit for multiplexing and demultiplexing signals. A key point in our approach is that lithography is only used to define a regular array of microscale gate wires and is not needed to create a specific address code at the nanoscale as in previous work (19). Because synthesis is used to define the code required for nanoscale addressing (not lithography), we call this a Blithography-independent[ addressing scheme. In the demonstration example of Fig. 3B, we took advantage of stochastic end-to-end nanowire alignment to produce distinct codes from a single type of nanowire. Indeed, the general case of a stochastic addressing with modulation-doped nanowires has been analyzed and shown to be an efficient approach for addressing dense nanoscale arrays; that is, it requires È2.2ln(N) nanowires for addressing N lines when N is large (20). Control of the size and separation of modulation-doped regions also enables synthesis to define quantum dot (QD) structures, in which the Fermi level offset caused by variations in dopant concentration produces potential barriers confining the QD (Fig. 4A). Conductance versus Vg and bias voltage (Vsd) studies of nþ-n-nQDþ-n-nþ modulation-doped silicon nanowires—in which the lightly doped n-type regions define barriers for a variable length QD, nQDþ—reveal well-defined diamond structures (Fig. 4B); the single-period diamond shows that transport occurs through a single QD structure (21). Notably, the geometry-dependent gate capacitance, Cg , determined from these data, 23.5 aF, agrees well with the value, 24.1 aF, calculated from the È500-nm QD size determined by SGM imaging (Fig. 4B, inset). In addition, current (I) versus Vg data for this nanowire and a structure in which the nQDþ section is reduced by half to È250 nm (Fig. 4C) show singleFig. 4. QD structures defined by synthesis. (A) Schematic of nþ QD structure confined by two n-type barriers within a modulationdoped nanowire. The conduction band (Ec ) offset of the nþ and n sections induce tunneling barriers with Coulomb blockade phenomenon observed when thermal energy is ¡U, the charging energy (21). The red X indicates blockage of charge transport. (B) Plot of ¯I/¯Vsd versus Vsd and Vg recorded at 1.5 K on a nþ-n-nQDþ- n-nþ device. The blue regions correspond to low values of ¯I/¯Vsd, and the red regions correspond to high values; the red color corresponds to 1.8 mS. The middle nQDþ and two n sections were grown for 3 and 0.5 min, respectively, at 80 torr. (Inset) SGM image of the same device. Scale bar, 200 nm. (C) I – Vg data taken at 1.5 K on the device in (B) (blue curve) and another device with the nQDþ section grown for 1.5 min (red curve). Vsd 0 0.2 mV. A.U., arbitrary units. (D) Coupled double-QD structure with variable-width n2 section between the two QDs. (Right) I – Vg data recorded at 1.5 K on three devices with n2 sections grown for 15, 10, and 5 s (top to bottom). Fig. 3. Modulation-doped address decoder. (A) I versus Vg (Vsd 0 1 V) measured for an nþ-n-nþ silicon nanowire device, where gates 1, 2, and 3 (inset) correspond to blue, black, and red curves, respectively. The native silicon oxide was used as a gate dielectric with Au metal gates. (Inset) Scanning electron microscopy (SEM) image of the device. Scale bar, 1 mm. (B) SEM image of a 2-by-2 decoder configured using two modulationdoped silicon nanowires as outputs (Out1 and Out2) and two Au metal gates, which were deposited over a uniform Si3N4 dielectric as inputs (In1 and In2). Scale bar, 1 mm. Plots of input (blue) and output (red) voltages for the 2-by-2 decoder. Supply voltage is –2 V. 1306 25 NOVEMBER 2005 VOL 310 SCIENCE www.sciencemag.org R EPORTS
REPORTS period oscillations in both devices 11.A B. Greytak, L J. C M the period Al. is approximately 4176(200 fining barriers for an InAs QD(8) the 250- versus 500-nm QD. Bec 12.L」 Lawhon.Ms.Gt D. Wang, C M. Lieber, 23. inversely proportional to the gate capacitance, 13. Mat Al. e/Co, and QD size, this comparise naterial on Science Online shows that the true size of the confined QD 4.J MJasinski, S.M.Gates, Acc.Chem.Res24,9 compared with individual QD capacitances(23). can be controlled in a predictable manner in 15. R.T. White, R L. Espino-Rios, D.S. A Ring 25. N Mason, M.J. Biercuk, C M. Marcus, Science 303, these modulation-doped nanowires (22) The potential of our approach for encoding 17. P. M. Fahey, P. B. Griffin, ].D. Plummer, Rev. Mod. Fuhrer, M. T. Bjork, L Samuelson, Nano coupled quantum structures has been explored 27. We thank H. Park and A Dehon for discussion. C.M. L. in modulation-doped silicon na res that have 18. Y Wu et al, Let.4.433(2004) acknowledges support of this work by the Defense structures of the form nt-n -nlont-nls-non+. 9. Z. Zhong D. Wang Y. Cui, M. w. Bockrath, CM Lieber, Science 302, 1377(2003 n-nt, where n, are fixed-width tunnel bar- 20. A DeHon, P. Lincoln,] E Savage, IEEE Trans.Nano- upporting Online Material riers that weakly couple the structure to source and drain electrodes. and n. is a variable-width 21. LP. Kouwenhoven et al., in Proceedings of Advanced DC1 terials and Methods arier that couples the two QDs(Fig. 4D, left L P. Kouwenhoven, G. Schon, Eds.(Kluwer, Fig S1 panel). The I-V data recorded from rep- 22. Recent studies have demonstrated the formation 12 Augu 4 October 2005 ferent n, barrier widths coupling the QDs of single quantum dots in InP/inAs composition- 10.1126/science. 1118798 (Fig. 4D, right panel) demonstrate several key points. First, the device with the largest barrier exhibits a single coulomb oscillation Super-Compressible Foamlike period that yields a capacitance consistent with the size of each individual QD deter Carbon Nanotube films mined from SGM measurements. This result shows qualitatively that the two QDs are weakly coupled, and moreover, have sizes Anyuan Cao, Pamela L. Dickrell, W. Gregory Sawyer, that are similar. Second, the data from the Mehrdad N. Ghasemi-Nejhad, Pulickel M. Ajayan* device with an intermediate-width n. barrier exhibits a splitting of each of the Coulomb We report that freestanding films of vertically aligned carbon nanotubes oscillation peaks into doublets, which is the exhibit super-compressible foamlike behavior. Under compression, the nano signature of enhanced tunneling conductan tubes collectively form zigzag buckles that can fully unfold to their original between the QDs(23, 24). This observation length upon load release. Compared with conventional low-density flexible agrees with previous studies(23, 25, 26)where foams, the nanotube films show much higher compressive strength, recovery coupled dots were defined by lithographically rate, and sag nd the open-cell nature of the nanotube arrays gives patterned gate electrodes. Last, as the barrier excellent breat The nanotube films present a class of open-cell foam width is reduced further, a single Coulomb of well-arranged one-dimensional units(nanotube oscillation period is again observed, although struts). The lightweight, highly resilient nanotube films may be useful as the capacitance shows that the effective QD Ing gs size is twice that of the individual n that is, the structures are fully delocalized Structural foams(1, 2) have a variety of ap- modulus(1). Metallic(e.g, Al) foams have These studies demonstrate the ability to plications in modern society such as in er compressive strength than polymeric thesize coupled QDs within nanowires, construction, energy dissipation, cushioning, foams, but the plastic deformation of cell struc- here the interaction between quantum struc- and packaging. Mechanical strength(com- tures results in little resilience upon load re- tures is defined by synthesis not lithography. pressive stress) and compressibility(strain) lease(5). The elastic segments(struts)between More generally, this work demonstrates the are two important factors that determine the adjacent cells form the architecture of a foam, potential of encoding functional information performance and applications of foams; how- and it is the bending and buckling of these struts that allows the foam to be compres le believe will open up opportunities for con- nature. Increasing the volume of the cells(i.e, the property of a strut( determined by its com- tional and quantum electronic devices and the void area)in a foam results in higher position, geometry, and dimension) dictates the ircuits in the future ompressibility(up to 75%)but causes rap- compressive behavior(6, 7) idly decreasing strength(2-4). For the foam A carbon nanotube(8, 9) is perhaps the best Reference and notes a fixed chemical composition, its modulus strut to make ultralight yet strong foams, con- P.LMcEuen, M.s. Fuhrer, H. Park, IEEE Trans. (Ep decreases with increasing relative cell vol- ume(o)as E= CE(I-o, where C is a low density, and high elasticity (10). In par constant(close to unity) and e is the cell edge ticular, the nanotube exhibits extreme struc- 4. L Samuelson et al. Phys. E 25, 313(2004). ural flexibility(10-12)and can be repeatedly Science 304, 1787(2004). Popov, R Cost, U. Banin. Department of Mechanical Engineering. University nt through large angles and strains without ructural failure(13). The ability of nanotubes 6.M.5. Gudiksen, L J. Lauhon, i Wang D. C Smith, ment of Mechanical and Aerospace Engineering, to adopt and switch between various buckled 9. C. Zhou L, Kong E. Yenilmel, H Dai, Science 290, Rensselaer Polytechnic Institute, Troy, NY 12180 use morphologies makes them capable of accom- 7. M. T Bjork et nent of Materials Science and modating and sustaining large local strains " To whom correspondence should be addressed. while maintaining structural integrity (14, 15 10. V. Derycke, R MarteL, J. Appenzeller, Ph. Avouris, mail: anyuan@hawaii. edu(AC): ajayan @rpi.edu We show that vertically aligned nano- wsciencemag. org SCIENCE VOL 310 25 NOVEMBER 2005 1307
period oscillations in both devices, although the period DVg is approximately doubled in the 250- versus 500-nm QD. Because DVg is inversely proportional to the gate capacitance, DVg 0 e/Cg, and QD size, this comparison shows that the true size of the confined QD can be controlled in a predictable manner in these modulation-doped nanowires (22). The potential of our approach for encoding coupled quantum structures has been explored in modulation-doped silicon nanowires that have structures of the form nþ-n1-nQDþ-n2-nQDþ- n1-nþ, where n1 are fixed-width tunnel barriers that weakly couple the structure to source and drain electrodes, and n2 is a variable-width barrier that couples the two QDs (Fig. 4D, left panel). The I – Vg data recorded from representative nanowire devices with three different n2 barrier widths coupling the QDs (Fig. 4D, right panel) demonstrate several key points. First, the device with the largest barrier exhibits a single Coulomb oscillation period that yields a capacitance consistent with the size of each individual QD determined from SGM measurements. This result shows qualitatively that the two QDs are weakly coupled, and moreover, have sizes that are similar. Second, the data from the device with an intermediate-width n2 barrier exhibits a splitting of each of the Coulomb oscillation peaks into doublets, which is the signature of enhanced tunneling conductance between the QDs (23, 24). This observation agrees with previous studies (23, 25, 26) where coupled dots were defined by lithographically patterned gate electrodes. Last, as the barrier width is reduced further, a single Coulomb oscillation period is again observed, although the capacitance shows that the effective QD size is twice that of the individual nQDþ regions; that is, the structures are fully delocalized. These studies demonstrate the ability to synthesize coupled QDs within nanowires, where the interaction between quantum structures is defined by synthesis not lithography. More generally, this work demonstrates the potential of encoding functional information into nanostructures during synthesis, which we believe will open up opportunities for conventional and quantum electronic devices and circuits in the future. Reference and Notes 1. P. L. McEuen, M. S. Fuhrer, H. Park, IEEE Trans. Nanotechnology 1, 78 (2002). 2. H. Dai, Acc. Chem. Res. 35, 1035 (2002). 3. C. M. Lieber, Mater. Res. Soc. Bull. 28, 486 (2003). 4. L. Samuelson et al., Phys. E 25, 313 (2004). 5. T. Mokari, E. Rothenberg, I. Popov, R. Costi, U. Banin, Science 304, 1787 (2004). 6. M. S. Gudiksen, L. J. Lauhon, J. Wang, D. C. Smith, C. M. Lieber, Nature 415, 617 (2002). 7. M. T. Bjork et al., Appl. Phys. Lett. 80, 1058 (2002). 8. M. T. Bjork et al., Nano Lett. 4, 1621 (2004). 9. C. Zhou, J. Kong, E. Yenilmez, H. Dai, Science 290, 1552 (2000). 10. V. Derycke, R. Martel, J. Appenzeller, Ph. Avouris, Nano Lett. 1, 453 (2001). 11. A. B. Greytak, L. J. Lauhon, M. S. Gudiksen, C. M. Lieber, Appl. Phys. Lett. 84, 4176 (2004). 12. L. J. Lauhon, M. S. Gudiksen, D. Wang, C. M. Lieber, Nature 420, 57 (2002). 13. Materials and methods are available as supporting material on Science Online. 14. J. M. Jasinski, S. M. Gates, Acc. Chem. Res. 24, 9 (1991). 15. R. T. White, R. L. Espino-Rios, D. S. Rogers, M. A. Ring, H. E. O’Neal, Int. J. Chem. Kinet. 17, 1029 (1985). 16. A. Bachtold et al., Phys. Rev. Lett. 84, 6082 (2000). 17. P. M. Fahey, P. B. Griffin, J. D. Plummer, Rev. Mod. Phys. 61, 289 (1989). 18. Y. Wu et al., Nano Lett. 4, 433 (2004). 19. Z. Zhong, D. Wang, Y. Cui, M. W. Bockrath, C. M. Lieber, Science 302, 1377 (2003). 20. A. DeHon, P. Lincoln, J. E. Savage, IEEE Trans. Nanotechnology 2, 165 (2003). 21. L. P. Kouwenhoven et al., in Proceedings of Advanced Study Institute on Mesoscopic Electron Transport, L. L. Sohn, L. P. Kouwenhoven, G. Scho¨n, Eds. (Kluwer, Dordecht, Netherlands, 1997). 22. Recent studies have demonstrated the formation of single quantum dots in InP/InAs compositionmodulated nanowires, where InP is used to produce confining barriers for an InAs QD (8). 23. F. R. Waugh et al., Phys. Rev. B 53, 1413 (1996). 24. The tunneling conductance for the strongly coupled QDs was estimated to be 0.9 G0, where G0 0 2e2/h, e is the elementary charge, and h is Planck’s constant, assuming that interdot capacitance is negligible compared with individual QD capacitances (23). 25. N. Mason, M. J. Biercuk, C. M. Marcus, Science 303, 655 (2004). 26. C. Fasth, A. Fuhrer, M. T. Bjork, L. Samuelson, Nano Lett. 5, 1487 (2005). 27. We thank H. Park and A. DeHon for discussion. C.M.L. acknowledges support of this work by the Defense Advanced Research Projects Agency. Supporting Online Material www.sciencemag.org/cgi/content/full/310/5752/1304/ DC1 Materials and Methods Fig. S1 12 August 2005; accepted 24 October 2005 10.1126/science.1118798 Super-Compressible Foamlike Carbon Nanotube Films Anyuan Cao,1 * Pamela L. Dickrell,2 W. Gregory Sawyer,2 Mehrdad N. Ghasemi-Nejhad,1 Pulickel M. Ajayan3 * We report that freestanding films of vertically aligned carbon nanotubes exhibit super-compressible foamlike behavior. Under compression, the nanotubes collectively form zigzag buckles that can fully unfold to their original length upon load release. Compared with conventional low-density flexible foams, the nanotube films show much higher compressive strength, recovery rate, and sag factor, and the open-cell nature of the nanotube arrays gives excellent breathability. The nanotube films present a class of open-cell foam structures, consisting of well-arranged one-dimensional units (nanotube struts). The lightweight, highly resilient nanotube films may be useful as compliant and energy-absorbing coatings. Structural foams (1, 2) have a variety of applications in modern society such as in construction, energy dissipation, cushioning, and packaging. Mechanical strength (compressive stress) and compressibility (strain) are two important factors that determine the performance and applications of foams; however, these two properties are of opposing nature. Increasing the volume of the cells (i.e., the void area) in a foam results in higher compressibility (up to 75%) but causes rapidly decreasing strength (2–4). For the foam at a fixed chemical composition, its modulus (Ef ) decreases with increasing relative cell volume (f) as Ef 0 CE(1 – f)2, where C is a constant (close to unity) and E is the cell edge modulus (1). Metallic (e.g., Al) foams have higher compressive strength than polymeric foams, but the plastic deformation of cell structures results in little resilience upon load release (5). The elastic segments (struts) between adjacent cells form the architecture of a foam, and it is the bending and buckling of these struts that allows the foam to be compressed; the property of a strut (determined by its composition, geometry, and dimension) dictates the compressive behavior (6, 7). A carbon nanotube (8, 9) is perhaps the best strut to make ultralight yet strong foams, considering its exceptional mechanical strength, low density, and high elasticity (10). In particular, the nanotube exhibits extreme structural flexibility (10–12) and can be repeatedly bent through large angles and strains without structural failure (13). The ability of nanotubes to adopt and switch between various buckled morphologies makes them capable of accommodating and sustaining large local strains while maintaining structural integrity (14, 15). We show that vertically aligned nanotubes (16) form a highly resilient open-cell 1 Department of Mechanical Engineering, University of Hawaii at Manoa, Honolulu, HI 96822, USA. 2 Department of Mechanical and Aerospace Engineering, University of Florida, Gainesville, FL 32611, USA. 3 Department of Materials Science and Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180, USA. *To whom correspondence should be addressed. E-mail: anyuan@hawaii.edu (A.C.); ajayan@rpi.edu (P.M.A.) www.sciencemag.org SCIENCE VOL 310 25 NOVEMBER 2005 1307 R EPORTS