REPORTS y small-scale convection. The induced relative 3. A D Miall, Geology 20, 787(1992). 21. D. McKenzie, Earth Planet. Sci. Left. 40. 25(1978). sea-level variations are regionally correlatable and N. Christie-Blick, G.S. Mountain, K. G. Miller, Science 2 are synchronous or nearly synchronous over spa 241.596(1988 5. P. ]. Tackley, Nat. Geosci. 1, 157(2008) 23. W. R. Buck, Earth Planet. Sai. Lett. 77 62(1986) tial scales that encompass the dimensions of most 6. F M Richter, /. Geophys. Res. 78, 8735(1973) 24. K G. Miller et al. Science 310, 1293(2005) shelf-slope systems of sedimentary basins, where 7. B. Parsons, D Mckenzie, J Geophys. Res. 83(B9), 4485 25. A. D Miall, The Geology of Stratigraphic Sequences sequences are most readily identifiable. This poses a challenge for the use of sequence stratigraphy in 9. R. Stephenson, C. Beaumont, in Mechanisms of Continental 26. S. Cloetingh, H. McQueen, K. Lambeck, Earth Planet regional and global correlation as well as in the Sc.e75,157(1985) 27. 5. B. Nielsen, E. Thomsen, D L. Hansen, O.R. Clausen construction of global sea-level records. If IPse- (Academic Press, London, 1980), pp. 111-122. Nature435,195(2005) quences are indeed controlled by sublithospheric 10. M. D. Ballmer, J. van Hunen, G. Ito, P ). Tackle 28. R. D Muller, M. Sdrolias, C. Gaina, B. Steinberg convection, they can only be considered locally 11. 5. D. King, l Ritsema, Science 290, 1137(2000) for dating sedimentary successions. Our study thus 12. G. Ranalli, Rheology of the Earth(Chapman HalL, provides an explanation for the enigmatic control 30. We thank M. Huuse, E. Thomsen, N Christie-Blick, and on IP cycles, between the shorter-period varia 13. See supporting data on Science Online three anonymous reviewers for constructive suggestions and tions caused by Milankovich cycles and those 14. 5. L Karato, P. Wu, Science 260, 771(1993) D. L Egholm for providing computational facilitie caused by regional in-plane stress changes and 16.I. Huang, S. Zhong, 1. Van Hunen, I. Geophys. Res. B: Supporting Online Material Solid earth108.2405(2003) wsciencemag. org/cgi/content/ull329/5993/827/DC1 cluding glacio-eustatic sea-level variations. 17. R. Moucha ef al. Earth Planet. Sa. Left. 271. 101 Figs. Sl to S7 18. L L Sloss, GeoL. Soc. Am. BulL 74, 93(1963 es and Notes Muller 1. 0. Catuneanu ef al.. Earth Sci. Rev. 92. 1(2009 ophrys. Res. Lett. 35, L08305(2008). 2. B U. Haq. ]. Hardenbol, P. R Vail, Science 235, 1156 20. R. M. Mitchum, Am. Assoc. Pet. Geol Men. 26, 53 26 March 2010: accepted 2 July 2010 0.1126/ cience.1190115 Three-Dimensional, Flexible spikes(7, 8). For electrical probes, the single elec- trical connection facilitates design and mechan- Nanoscale field-Effect transistors ical insertion into cells, but the requirement of direct ionic and/or electrical junctions between probe tips and cytosol also introduce several lim- as Localized Bioprobes itations. First, the tip size of these probes(0. 2 to 5 um)(9-14) is a compromise between being Bozhi Tian, "Tzahi Cohen-Karni, Quan Qing, Xiaojie Duan, Ping Xie, Charles M. Lieber+t membrane with minimum damage (<5 um)and Nanoelectronic devices offer substantial potential for interrogating biological systems, although nearly sufficiently low (0.2 um)(9, In)so that small integration of a nanoscale field-effect transistor (nanoFEn device at the tip of an acute-angle kinked cellular signals can be discerned from thermal noise. Second, direct exposure of intracellular spe- silicon nanowire, where nanoscale connections are made by the arms of the kinked nanostructure, and cies to extraneous probe surfaces or electrolytes remote multilayer interconnects allow three-dimensional(3D)probe presentation. The acute-angle in probe lumen(9-13), especially for larger glass probe geometry was designed and synthesized by controlling cis versus trans crystal conformations micropipettes, might induce irreversible chang between adjacent kinks, and the nano FET was localized through modulation doping. 3D nanoFET probes to cells and, thus, prevent long-term and non- exhibited conductance and sensitivity in aqueous solution, independent of large mechanical deflections, invasive cellular recordings. Finally, these probe and demonstrated high pH sensitivity. Additionally, 3D nanoprobes modified with phospholipid bilayers techniques are intrinsically passive and are not can enter single cells to allow robust recording of intracellular potentials. capable of built-in signal processing and facile integration with other circuitries, especially given N nowire and nanotube electrical devices essary source(S)and drain) electrical con- the emerging need to enable a cell-machine com- ave been exploited for ultrasensitive nections could move into contact with the cell munication(15) detection of biological markers (D) and and probe within the cell membrane. However, FETs can function in a sub-10-nm-size regime high-resolution extracellular recording from cells minimally invasive insertion of a nanoFET into (16). NanoFETs could function as mechanically (2-5). However, localized and tunable three. the confined 3D space of single cells, or even noninvasive probes capable of entering cells dimensional (3D)sensing and recording with the 3D cellular networks, is a major challenge be. through endocy ways, as can occur with rototypical nanoelectronic device, a nanoscale cause the S and d typically dominate the overall nanoparticles (17). Moreover, when interfacing field-effect transistor(nanoFET)(6), have not device size and define a planar and rigid struc- with cells, the FETs process input/output infor- been demonstrated because almost all examples ture, regardless of whether the nanoFET is on or mation without the need for direct exchange with of these devices are created on planar substrates. suspended above a substrate(5, 6). cellular ions; thus, interfacial impedance and bio- Ideally, rather than force the cell to conform to Existing probes capable of intracellular sensing chemical invasiveness to cells can be minimized. the substrate, a movable nanoFET with the nec- and recording include voltage-sensitive optical In addition, because signals are transduced by dyes(7, 8)and single-terminal glass(9-11)or change in field/potential at well-isolated surfaces, Department of Chemistry and Chemical Biology, Harvard carbon(12-14) microelectrodes. Voltage-sensitive FETs can detect cellular potential (2-5), as well as University, Cambridge, MA02138, USA. School of Engineering and Applied Sciences, Harvard University, Cambridge, Ma dyes can readily be used to interrogate action biological macromolecules(I), and could be 02138,USA tentials with high spatial resolution, but they integrated for potential multiplexed intracellular also have limitations in terms of signal-to-noise measurements. Unfortunately, and as discussed fTo whom corresponden be addressed. E-mail. (S/N)ratio, pharmacological side effects, photo- above, the requirement of two electrical contacts @cmlinis harvard ed toxicity, and difficulty in differentiating single to a FET, the S and D, makes design of 3D probes 13AuguSt2010Vol329ScieNcewww.sciencemag.org
by small-scale convection. The induced relative sea-level variations are regionally correlatable and are synchronous or nearly synchronous over spatial scales that encompass the dimensions of most shelf-slope systems of sedimentary basins, where sequences are most readily identifiable. This poses a challenge for the use of sequence stratigraphy in regional and global correlation as well as in the construction of global sea-level records. If IP sequences are indeed controlled by sublithospheric convection, they can only be considered locally for dating sedimentary successions. Our study thus provides an explanation for the enigmatic control on IP cycles, between the shorter-period variations caused by Milankovich cycles and those caused by regional in-plane stress changes and other tectonic and nontectonic contributions, including glacio-eustatic sea-level variations. References and Notes 1. O. Catuneanu et al., Earth Sci. Rev. 92, 1 (2009). 2. B. U. Haq, J. Hardenbol, P. R. Vail, Science 235, 1156 (1987). 3. A. D. Miall, Geology 20, 787 (1992). 4. N. Christie-Blick, G. S. Mountain, K. G. Miller, Science 241, 596 (1988). 5. P. J. Tackley, Nat. Geosci. 1, 157 (2008). 6. F. M. Richter, J. Geophys. Res. 78, 8735 (1973). 7. B. Parsons, D. McKenzie, J. Geophys. Res. 83 (B9), 4485 (1978). 8. W. R. Buck, Nature 313, 775 (1985). 9. R. Stephenson, C. Beaumont, in Mechanisms of Continental Drift and Plate Tectonics, P. A. Davies, K. Runcorn, Eds. (Academic Press, London, 1980), pp. 111–122. 10. M. D. Ballmer, J. van Hunen, G. Ito, P. J. Tackley, T. A. Bianco, Geophys. Res. Lett. 34, L23310 (2007). 11. S. D. King, J. Ritsema, Science 290, 1137 (2000). 12. G. Ranalli, Rheology of the Earth (Chapman & Hall, London, ed. 2, 1995). 13. See supporting data on Science Online. 14. S. I. Karato, P. Wu, Science 260, 771 (1993). 15. S. I. Karato, Geophys. Res. Lett. 19, 2255 (1992). 16. J. Huang, S. Zhong, J. Van Hunen, J. Geophys. Res. B: Solid Earth 108, 2405 (2003). 17. R. Moucha et al., Earth Planet. Sci. Lett. 271, 101 (2008). 18. L. L. Sloss, Geol. Soc. Am. Bull. 74, 93 (1963). 19. S. Spasojević, L. Liu, M. Gurnis, R. D. Müller, Geophys. Res. Lett. 35, L08305 (2008). 20. R. M. Mitchum, Am. Assoc. Pet. Geol. Mem. 26, 53 (1977). 21. D. McKenzie, Earth Planet. Sci. Lett. 40, 25 (1978). 22. C. E. Keen, in Continental Extensional Tectonics (Blackwell Scientific, London, 1987), p. 67. 23. W. R. Buck, Earth Planet. Sci. Lett. 77, 362 (1986). 24. K. G. Miller et al., Science 310, 1293 (2005). 25. A. D. Miall, The Geology of Stratigraphic Sequences (Springer, Berlin, 1997). 26. S. Cloetingh, H. McQueen, K. Lambeck, Earth Planet. Sci. Lett. 75, 157 (1985). 27. S. B. Nielsen, E. Thomsen, D. L. Hansen, O. R. Clausen, Nature 435, 195 (2005). 28. R. D. Müller, M. Sdrolias, C. Gaina, B. Steinberger, C. Heine, Science 319, 1357 (2008). 29. A. Embry, Search and Discovery, www.searchanddiscovery. com/documents/2009/30105embry/ (2009). 30. We thank M. Huuse, E. Thomsen, N. Christie-Blick, and three anonymous reviewers for constructive suggestions and D. L. Egholm for providing computational facilities. Supporting Online Material www.sciencemag.org/cgi/content/full/329/5993/827/DC1 Methods Figs. S1 to S7 Table S1 References Movie S1 26 March 2010; accepted 2 July 2010 10.1126/science.1190115 Three-Dimensional, Flexible Nanoscale Field-Effect Transistors as Localized Bioprobes Bozhi Tian,1 * Tzahi Cohen-Karni,2 * Quan Qing,1 Xiaojie Duan,1 Ping Xie,1 Charles M. Lieber1,2† Nanoelectronic devices offer substantial potential for interrogating biological systems, although nearly all work has focused on planar device designs. We have overcome this limitation through synthetic integration of a nanoscale field-effect transistor (nanoFET) device at the tip of an acute-angle kinked silicon nanowire, where nanoscale connections are made by the arms of the kinked nanostructure, and remote multilayer interconnects allow three-dimensional (3D) probe presentation. The acute-angle probe geometry was designed and synthesized by controlling cis versus trans crystal conformations between adjacent kinks, and the nanoFET was localized through modulation doping. 3D nanoFET probes exhibited conductance and sensitivity in aqueous solution, independent of large mechanical deflections, and demonstrated high pH sensitivity. Additionally, 3D nanoprobes modified with phospholipid bilayers can enter single cells to allow robust recording of intracellular potentials. Nanowire and nanotube electrical devices have been exploited for ultrasensitive detection of biological markers (1) and high-resolution extracellular recording from cells (2–5). However, localized and tunable threedimensional (3D) sensing and recording with the prototypical nanoelectronic device, a nanoscale field-effect transistor (nanoFET) (6), have not been demonstrated because almost all examples of these devices are created on planar substrates. Ideally, rather than force the cell to conform to the substrate, a movable nanoFET with the necessary source (S) and drain (D) electrical connections could move into contact with the cell and probe within the cell membrane. However, minimally invasive insertion of a nanoFET into the confined 3D space of single cells, or even 3D cellular networks, is a major challenge because the S and D typically dominate the overall device size and define a planar and rigid structure, regardless of whether the nanoFET is on or suspended above a substrate (5, 6). Existing probes capable of intracellular sensing and recording include voltage-sensitive optical dyes (7, 8) and single-terminal glass (9–11) or carbon (12–14) microelectrodes. Voltage-sensitive dyes can readily be used to interrogate action potentials with high spatial resolution, but they also have limitations in terms of signal-to-noise (S/N) ratio, pharmacological side effects, phototoxicity, and difficulty in differentiating single spikes (7, 8). For electrical probes, the single electrical connection facilitates design and mechanical insertion into cells, but the requirement of direct ionic and/or electrical junctions between probe tips and cytosol also introduce several limitations. First, the tip size of these probes (~0.2 to 5 mm) (9–14) is a compromise between being small enough to penetrate or rupture the cell membrane with minimum damage (0.2 mm) (9, 11) so that small cellular signals can be discerned from thermal noise. Second, direct exposure of intracellular species to extraneous probe surfaces or electrolytes in probe lumen (9–13), especially for larger glass micropipettes, might induce irreversible changes to cells and, thus, prevent long-term and noninvasive cellular recordings. Finally, these probe techniques are intrinsically passive and are not capable of built-in signal processing and facile integration with other circuitries, especially given the emerging need to enable a cell-machine communication (15). FETs can function in a sub–10-nm-size regime (16). NanoFETs could function as mechanically noninvasive probes capable of entering cells through endocytic pathways, as can occur with nanoparticles (17). Moreover, when interfacing with cells, the FETs process input/output information without the need for direct exchange with cellular ions; thus, interfacial impedance and biochemical invasiveness to cells can be minimized. In addition, because signals are transduced by change in field/potential at well-isolated surfaces, FETs can detect cellular potential (2–5), as well as biological macromolecules (1), and could be integrated for potential multiplexed intracellular measurements. Unfortunately, and as discussed above, the requirement of two electrical contacts to a FET, the S and D, makes design of 3D probes 1 Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA 02138, USA. 2 School 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 830 13 AUGUST 2010 VOL 329 SCIENCE www.sciencemag.org REPORTS
and their minimally invasive insertion into a cell central to our probe-geometry such conformation control is also maintained in or tissue a substantial challenge esentative scanning electron mi- the growth of complex probe structures [fig. Recently, we demonstrated that variation of image of an 80-nm diameter, SIB (9)] reactant pressure during silicon nanowire(SiNw) doubly INW with an intervening seg. The second design consideration made use growth could introduce reproducible 120 kinks ment length(L)of-160 nm between kink units of selective in situ doping during synthesis to (8)and that the junction regions could be doped(Fig. 1 B)shows well-defined cis-linkage and an localize and self-label the nanoscale FET element to create p-n diodes and FETs. We used this overall 60 tip angle( Fig. IA, top). To inves- adjacent to the topologically defined probe tip methodology to create a two-terminal FET probe tigate our ability to synthesize this cis-linkage of(Fig. 1A, pink segments)and to simultaneously that could be inserted into single cells. The kink structural units reproducibly, we analyzed "wire-up"the FEt channel with nanowire S/D growth of kinked SiNWs had to be tailored in their fraction as a function of L in doubly kinked components(Fig 1A, blue segments). As in our the following ways: First, we needed to incor- SiNWS. Notably, the plot of cis/(cis +trans), as studies of single-kinked nanowires(18), we use porate two or three cis-linked kinked units to L was varied from -700 to 50 nm(Fig. IC), heavy n-type doping for the nanowire S/D yield probe-tip angles of 60 or 0, respectively shows that the cis conformation becomes dom- arms and reduced the concentration to light (Fig. IA, top and middle). Because two trans- inant as L decreases. We can selectively synthe- type doping to introduce a short -200-nm region inked units(Fig. IA, bottom) would yield an size cis-linked kinked units in good yield [fig. immediately following the growth of two se- unusable probe tip, the selective synthesis of cis- SIA(19),e., -66% where L- 50 nm, and quential kinks and serving as the"pointlike "FET Fig. 1. Synthesis of kinked SiNW probes (A) A c0.7 Schematics of 60(top) ando° middle)multiply ←S/D→ kinked nanowires and ais and trans (bottom) structures. The blue ate the source/drain S/D and nanoscale fet chan- 150300450600750 L(nm) image of a doubly kinked nanowire with a cis configuration. L is the length of segment between two adjacent kinks (o) cis/(cis +trans) versus L plot Error bars indicate #1 SD from the mean. (D)Transmission electron microscopy image of an ultrathin 60 kinked nanowire. Scale bars, 200 nm(B): 50 nm(D) Fig. 2. 3D kinked nano- wire probes (A) Schemat- ics of device fabrication. SU-8S 8 microribbons [see sup substrate SoM) materials and methods (19) serve as spectively. The dime ions of the lightly de n-type silicon segment (white dots) are -80 by 80 by 200 nm .H and e are the tip height and PDMS orientation, respectively. and s and d designate △H the built-in source and drain connections to the nanoscale FET. (B) SEM D) and bright-field optical 2 6450 2 microscopy (l, IID images of an as- made device. the yellow arrow and pink 6400 74 tar mark the nanoscale ET and SU-8, respectiv ly. ll and Ill are recorded in air and water, respec Deflection (um Time(s) ely. Scale bars, 5 um. (C Device conductance and sensitivity as a function of on of the probe using a micropipette change in ti ( D)Change in potential versus solution pH for a representativ controlled with a miaomanipulat measurements were carried out in pbs 3D nanowit (nset) Experimental scheme. For darity, the pointlike FET ution Error bars indicate +1 SD mean. nset) Experimental scheme. AH, elements are not labeled in the schematics in(o and(D) www.sciencemag.orgScieNceVol32913AuguSt2010 831
and their minimally invasive insertion into a cell or tissue a substantial challenge. Recently, we demonstrated that variation of reactant pressure during silicon nanowire (SiNW) growth could introduce reproducible 120° kinks (18) and that the junction regions could be doped to create p-n diodes and FETs. We used this methodology to create a two-terminal FET probe that could be inserted into single cells. The growth of kinked SiNWs had to be tailored in the following ways: First, we needed to incorporate two or three cis-linked kinked units to yield probe-tip angles of 60° or 0°, respectively (Fig. 1A, top and middle). Because two translinked units (Fig. 1A, bottom) would yield an unusable probe tip, the selective synthesis of cislinked units is central to our probe-geometry design. A representative scanning electron microscopy (SEM) image of an 80-nm diameter, doubly kinked SiNW with an intervening segment length (L) of ~160 nm between kink units (Fig. 1B) shows well-defined cis-linkage and an overall 60° tip angle (Fig. 1A, top). To investigate our ability to synthesize this cis-linkage of kink structural units reproducibly, we analyzed their fraction as a function of L in doubly kinked SiNWs. Notably, the plot of cis/(cis + trans), as L was varied from ~700 to 50 nm (Fig. 1C), shows that the cis conformation becomes dominant as L decreases. We can selectively synthesize cis-linked kinked units in good yield [fig. S1A (19)], e.g., ~66% where L ~ 50 nm, and such conformation control is also maintained in the growth of complex probe structures [fig. S1B (19)]. The second design consideration made use of selective in situ doping during synthesis to localize and self-label the nanoscale FET element adjacent to the topologically defined probe tip (Fig. 1A, pink segments) and to simultaneously “wire-up” the FET channel with nanowire S/D components (Fig. 1A, blue segments). As in our studies of single-kinked nanowires (18), we used heavy n++-type doping for the nanowire S/D arms and reduced the concentration to light ntype doping to introduce a short ~200-nm region immediately following the growth of two sequential kinks and serving as the “pointlike” FET Fig. 1. Synthesis of kinked SiNW probes. (A) Schematics of 60° (top) and 0° (middle) multiply kinked nanowires and cis (top) and trans (bottom) configurations in nanowire structures. The blue and pink regions designatethe source/drain(S/D) and nanoscale FET channel, respectively. (B) SEM image of a doubly kinked nanowire with a cis configuration. L is the length of segment between two adjacent kinks. (C) cis/(cis + trans) versus L plot. Error bars indicate T1 SD from the mean. (D) Transmission electron microscopy image of an ultrathin 60° kinked nanowire. Scale bars, 200 nm (B); 50 nm (D). Fig. 2. 3D kinked nanowire probes. (A) Schematics of device fabrication. Patterned poly(methyl methacrylate) and SU- 8 microribbons [see supporting online material (SOM) materials and methods (19)] serve as a sacrificial layer and flexible device support, respectively. The dimensions of the lightly doped n-type silicon segment (white dots) are ~80 by 80 by 200 nm3 . H and q are the tip height and orientation, respectively, and S and D designate the built-in source and drain connections to the nanoscale FET. (B) SEM (I) and bright-field optical microscopy (II, III) images of an as-made device. The yellow arrow and pink star mark the nanoscale FET and SU-8, respectively. II and III are recorded in air and water, respectively. Scale bars, 5 mm. (C) Device conductance and sensitivity as a function of deflection of the probe using a micropipette controlled with a micromanipulator. The measurements were carried out in PBS solution. Error bars indicate T1 SD from the mean. (Inset) Experimental scheme. DH, change in tip height. (D) Change in potential versus solution pH for a representative 3D nanowire probe. (Inset) Experimental scheme. For clarity, the pointlike FET elements are not labeled in the schematics in (C) and (D). www.sciencemag.org SCIENCE VOL 329 13 AUGUST 2010 831 REPORTS
REPORTS detector of the overall probe. Scanning gate mi- sitivities of 4 to 8 S/. Similar sensitivities, 4 To further highlight the flexibility and ro- croscopy(SGM) measurements [fig. SI(19) to 8 uS/V, were observed for kinked nanowire bustness of the 3D nanoFET probes, we have showed that nanoscale FETs were integrated at devices fabricated on planar substrates, thus characterized the conductance and sensitivity in the probe tip during overall synthesis indicating that there is no degradation in the bent PBs as a glass micropipette was used to vary We have also examined the size limits of 3D configuration. We note that the sensitivity the tip height(Fig. 2C, inset). Typical data(Fig. these synthetic bioprobes in terms of the overall contribution from the lightly doped nanoFET is 2C) yield a 97% of the total device response. Localized de-- t10-um change in H, which corresponds to and we found that well-defined probe structures tection by the lightly doped region versus the 90% yields with 30 nanoprobe devices per chip. In addition, the probe height and angle(H and 8, Fig. 2A)were ystematically tuned by changing the length and thickness of the free-standing part of the metal interconnects/SU-8 backbone (fig. S3B)(9) We also found that the nanoprobe H and e typi cally increase when submerged in aqueous solu- tion[/e of the device(Fig 2B, I)are 25 um/430 and 38 um/90 in air and water, respectively Fig. 2B, Il and ID. This change is reversible and suggests that the nanoprobe devices an Time(s) intrinsically flexible, and moreover, that the ig. 3. Surface modification and specific orientation could be manipulated chem- Dark purple, light purple, pink, and ry (A) Schematics of nanowire probe entrance into a cell. ically(21). Last, free-standing 3D FET devices denote the phospholipid bilayers, heavily doped nanowire spectively. B)False-color fluorescence image of a lipid- have been stored in air for at least 8 months with- coated nanowire probe. DMPC was c 1% nitrobenzoxadiazole dye-labeled lipids and imaged out appreciable changes in nanoprobe orientations through a 510/21 band-pass filter (O) Differential interference contrast microscopy images (upper panels) and FET sensitivity and 3%a, respectively). and electrical recording (lower panel of an HL-1 cell and 60 kinked nanowire probe as the cell The sensitivity of the 3D nanoscale FET approaches (, contacts and internalizes(ID, and is retracted from (ID the nanoprobe. A pulled-glass probes was characterized in phosphate-buffered ropipette (inner tip diameter-5 um) was used to manipulate and voltage clamp the Hl-1 cell. The saline(PBS) solution(Fig. 2C and fig. S4A). dashed green line corresponds to the micropipette potential Scale bars, 5 um. (D)Electrical recording with Measurements of the conductance versus refer- a 60 kinked nanowire probe without phospholipids surface modification. Green and blue arrows in(O)and ence potential for the 3D probes yielded sen- (D)mark the beginnings of cell penetration and withdrawal, respectively 13AuguSt2010Vol329ScieNcewww.sciencemag.org
detector of the overall probe. Scanning gate microscopy (SGM) measurements [fig. S1 (19)] showed that nanoscale FETs were integrated at the probe tip during overall synthesis. We have also examined the size limits of these synthetic bioprobes in terms of the overall nanowire diameter and length L between kinks, and we found that well-defined probe structures are possible for values as small as ~18 and 15 nm, respectively (Fig. 1D and fig. S2) (19). These data show that it is possible to create active semiconductor probes with dimensions smaller than microtubules in cells (20). We next designed an unconventional nanoelectronic-device–fabrication approach that would allow these probes to be used as cellular probes. We made remote electrical interconnects to the S/D nanowire arms on ultrathin SU-8 polymer ribbons above a sacrificial layer (Fig. 2A, top). The interfacial stress between materials (21) was used to bend the probe upward after a final lift-off process (Fig. 2A, bottom); see fig. S3A for fabrication details (19). Our nanoprobes are distinct from previous nanoelectronic devices because (i) the FET channel (Fig. 2A, white dots) and S/D (Fig. 2A, black segments) components are integrated epitaxially at the nanoscale through synthesis (Fig. 1), similar to singlekink structures (18), (ii) the nanoscale FET is free-standing, and (iii) the acute-angle kinkednanowire geometry and extended S/D arms spatially separate the functional nanoscale FET from the bulky interconnects by a distance up to ~30 mm, comparable to the size of single cells, so that the nanoscale interrogation can be realized with minimum interference from macroscopic interconnects. A representative SEM image of one freestanding device (Fig. 2B, I) demonstrates that the 60° kinked probe is intact after fabrication with the two nanowire arm terminals sandwiched between the SU-8 polymer and the metal contacts. We achieved ≥90% yields with ~30 nanoprobe devices per chip. In addition, the probe height and angle (H and q, Fig. 2A) were systematically tuned by changing the length and thickness of the free-standing part of the metal interconnects/SU-8 backbone (fig. S3B) (19). We also found that the nanoprobe H and q typically increase when submerged in aqueous solution [H/q of the device (Fig. 2B, I) are 25 mm/43° and 38 mm/90° in air and water, respectively (Fig. 2B, II and III)]. This change is reversible and suggests that the nanoprobe devices are intrinsically flexible, and moreover, that the specific orientation could be manipulated chemically (21). Last, free-standing 3D FET devices have been stored in air for at least 8 months without appreciable changes in nanoprobe orientations and FET sensitivity (97% of the total device response. Localized detection by the lightly doped region versus the heavily doped S/D nanowire arms is consistent with our SGM measurements on these acuteangle probes (fig. S1) and previous studies of single-kinked nanowires (18). To further highlight the flexibility and robustness of the 3D nanoFET probes, we have characterized the conductance and sensitivity in PBS as a glass micropipette was used to vary the tip height (Fig. 2C, inset). Typical data (Fig. 2C) yield a <20-nS conductance change for a ~T10-mm change in H, which corresponds to <0.31% fluctuation in the total device conductance. Likewise, the device sensitivity remains stable with a maximum change of ~0.15 mS/V or 2.4% variation for this ~T10-mm tip-height change. In addition, repetitive bending does not degrade Fig. 3. Surface modification and cellular entry. (A) Schematics of nanowire probe entrance into a cell. Dark purple, light purple, pink, and blue colors denote the phospholipid bilayers, heavily doped nanowire segments, active sensor segment, and cytosol, respectively. (B) False-color fluorescence image of a lipidcoated nanowire probe. DMPC was doped with 1% nitrobenzoxadiazole dye–labeled lipids and imaged through a 510/21 band-pass filter. (C) Differential interference contrast microscopy images (upper panels) and electrical recording (lower panel) of an HL-1 cell and 60° kinked nanowire probe as the cell approaches (I), contacts and internalizes (II), and is retracted from (III) the nanoprobe. A pulled-glass micropipette (inner tip diameter ~ 5 mm) was used to manipulate and voltage clamp the HL-1 cell. The dashed green line corresponds to the micropipette potential. Scale bars, 5 mm. (D) Electrical recording with a 60° kinked nanowire probe without phospholipids surface modification. Green and blue arrows in (C) and (D) mark the beginnings of cell penetration and withdrawal, respectively. 832 13 AUGUST 2010 VOL 329 SCIENCE www.sciencemag.org REPORTS
the fet performance in these nanowire probes We tested the 间 capabilities of the 3D Nemstian limit. High-resolution recording indi- le Fet performance in different SiNW g the response to vari- cates the capability of resolving changes as small bendin trations suggests the capability of ations ir sbes boy n thin a polydimethylsi- as 0.02 pH units [fig. S4B (19) in this phys- reliable sensing and recording in a flexible and loxane(PDMS) microfluidic channel(Fig. 2D, iologically relevant range of pH tunable 3D manner from single devices, which inset). Stepwise potential increases from 7.5 to To use the 3D nanoFET probes in cells( Fig. could be particularly beneficial for interfacing 6.7 by 0.1 pH units were readily resolved, and 3A), we coated them with phospholipid bilayer with soft and motile biological systems. the sensitivity of 58 mV/pH was near the which can form on a variety of nanostructured inorganic materials(22-24) and also fuse with ell membranes(24). Accordingly, we modified PDMS the negatively charged SiO, surface of the SiNWs extra- intra by fusion with unilamellar vesicles of phosphe lipid bilayers [1, 2-dimyristoyl-sn-glycero- phosphocholine(DMPC)](19, 22). Fluorescence FET microscopy images of dye-labeled DMPC modi- fied probes(Fig. 3B) indicate that the lipid substrate bilayers form a continuous shell on our acute- +S/D angle nanoprobes, and device measureme B show that the lipid surface coating results in 2, and sodium and outward potassium currents, respectively. The letters a to e denote five characteristic phases of a millisecond width(Fig. 4C, I). The peak cardiac intracellular potential, as defined in text. The red-dashed ine is the baseline corresponding to amplitude, shape, and width are similar to extra- intracellular resting state. The cell culture, electronics, and recording details are specified in the SoM cellular recordings made with nanowire devices on materials and methods section (19) substrates(2); moreover, optical images recorded at www.sciencemag.orgScieNceVol32913AuguSt2010 833
the FET performance in these nanowire probes. Finally, the stable FET performance in different bending configurations suggests the capability of reliable sensing and recording in a flexible and tunable 3D manner from single devices, which could be particularly beneficial for interfacing with soft and motile biological systems. We tested the sensing capabilities of the 3D SiNW probes by recording the response to variations in solution pH within a polydimethylsiloxane (PDMS) microfluidic channel (Fig. 2D, inset). Stepwise potential increases from 7.5 to 6.7 by 0.1 pH units were readily resolved, and the sensitivity of ~58 mV/pH was near the Nernstian limit. High-resolution recording indicates the capability of resolving changes as small as 0.02 pH units [fig. S4B (19)] in this physiologically relevant range of pH. To use the 3D nanoFET probes in cells (Fig. 3A), we coated them with phospholipid bilayers, which can form on a variety of nanostructured inorganic materials (22–24) and also fuse with cell membranes (24). Accordingly, we modified the negatively charged SiO2 surface of the SiNWs by fusion with unilamellar vesicles of phospholipid bilayers [1,2-dimyristoyl-sn-glycero-3- phosphocholine (DMPC)] (19, 22). Fluorescence microscopy images of dye-labeled DMPC modified probes (Fig. 3B) indicate that the lipid bilayers form a continuous shell on our acuteangle nanoprobes, and device measurements show that the lipid surface coating results in <1% changes in both the nanoFET conductance and sensitivity. We then monitored the calibrated potential change of phospholipid-modified nanoFET probe while an isolated HL-1 cell (25) was moved into contact and then away from the nanoprobe using a glass micropipette under microscopy visualization (Fig. 3C, top). The micropipette was also used to clamp the intracellular potential at –50 mV (11). Notably, measurement of the potential versus time from the nanoFET probe shows a sharp ~52-mV drop within 250 ms after cell-to-tip contact. While the nanoprobe tip is within the cell, the recorded potential maintains a relatively constant value of ~–46 mV and then returns to baseline when the cell was detached from the nanowire probe end. NanoFET probes of similar sensitivity that were not coated with a phospholipid bilayer modification exhibited only baseline fluctuations (<T1 mV), as the HL-1 cell was brought into contact and then retracted (Fig. 3D). These results suggest that the biochemical state of the nanowire probe surfaces (26) is critical for assisting access to the intracellular region, possibly through membrane fusion (Fig. 3A) (24), and is distinct from larger, more rigid probes commonly used for intracellular electrical recording. We have also investigated the formation of intracellular interfaces between our 3D nanoFET probes and spontaneously firing electrogenic cells. Embryonic chicken cardiomyocytes were cultured on PDMS substrates (2) and then positioned to place individual cells over phospholipid bilayer– modified vertical (q = 90°) nanoprobes within a cell-perfusion chamber (2, 4), as shown schematically in Fig. 4A. Representative conductanceversus-time data recorded from a 3D nanoFET probe initially in gentle contact with a spontaneously beating cardiomyocyte cell (fig. S5A) (19) showed a sequence of distinct features (Fig. 4B). Initially, we observed regularly spaced spikes with a frequency of ~2.3 Hz, consistent with the beating cardiomyocyte (Fig. 4B, I). The peaks have a potential change of ~3 to 5 mV, a S/N ratio ≥ 2, and a submillisecond width (Fig. 4C, I). The peak amplitude, shape, and width are similar to extracellular recordings made with nanowire devices on substrates (2); moreover, optical images recorded at Fig. 4. Electrical recording from beating cardiomyocytes. (A) Schematics of cellular recording from the cardiomyocyte monolayer on PDMS (left) and highlight of extracellular (middle) and intracellular (right) nanowire/cell interfaces. The cell membrane and nanowire lipid coatings are marked with purple lines. (B) Electrical recording from beating cardiomyocytes: (i) extracellular recording, (ii) transition from extracellular to intracellular recordings during cellular entrance, and (iii) steady-state intracellular recording. Green and pink stars denote the peak positions of intracellular and extracellular signal components, respectively. The red-dashed boxes indicate regions selected for (C). (C) Zoom-in signals from the corresponding red-dashed square regions in (B). Blue and orange stars designate features that are possibly associated with inward sodium and outward potassium currents, respectively. The letters a to e denote five characteristic phases of a cardiac intracellular potential, as defined in text. The red-dashed line is the baseline corresponding to intracellular resting state. The cell culture, electronics, and recording details are specified in the SOM materials and methods section (19). www.sciencemag.org SCIENCE VOL 329 13 AUGUST 2010 833 REPORTS
REPORTS the same time [fig. S5A(19) are consistent with chanical invasiveness; and the nanoFETs have high 18. B. 2. Tian, P Xie, T ). Kempa, D. C Bell, C M Lieber, extracellular signals. After a relatively brief (40-s)period of extra- 19. Materials and methods are available as supporting cellular signals, we observed several pronounced References and notes D. A. Giljohann, C.A. Mirkin, Nature 462, 461(2009). 20. C. Conde, A. Caceres, Nat Rev. Neurosci. 10, 319(2009) changes in recorded signals(Fig. 4, B and C, I 2. T. Cohen-Kani, B P. Timko, L. E Weiss, C.M. Lieb . G Leong et al, Proc Not. Acad Sci. USA 106, 703 (2009). and Iln without application of extemal force to Proc. Natl. Acad. Sci. U.S.A. 106, 7309(2009) 22. N Misra et al, Proc. Natl. Acad. Sci. U.S.A. 106, 13780 the PDMS/cell support. Specifically, the initial 3. J. F. Eschermann et al, Appl. Phys. Lett. 95,083703 extracellular signals gradually disappeared(Fig. 4. 0.Qing et al, Proc. Natt. Acad. Sc U.S.A 107, 1882 23. X.]. Zhou, ]. M. Moran-MirabaL, H. G. Craighead, P L McEuen, Nat Nanotechnol. 2, 185(20 4, B and C, Il, pink stars). There was a con- V. Chernomordik, M. M. Kozlov, Nat. Struct. Mol. BioL. comitant decrease in baseline potential, and new 5. 1.Heller, W.T. T Smaal, S.G. Lemay. C.Dekker, Small 5, 15.675(2008) peaks emerged that had an opposite sign, similar 25. w. C Claycomb ef al., Proc. Natl. Acad. Sci. U.S.A 95 6. W. Lu, C M. Lieber, Nat. Mater. 6, 841 (200 2979(1998 requency, much greater amplitude, and longer A Grinvald, R. Hildesheim, Nat. Rev. Neurosci. 5. 874 26. B D Almquist, N A Melosh, Proc. NatL Acad. Sa. USA duration(Fig. 4B, I, green stars). These new 07,5815(2010 peaks, which are coincident with cardiomyocyte 8 M Scanziani, M Hausser, Nature 461, 930(2009) 27.D.M. Bers Nature415,198(2002) 4B, I with an average calibrated peak ampli- 9. We thank G. Yellen, w. C. Claycomb, B. P tude of 80 mv and duration of 200 ms. The E Neher, Anu. Rev. Physiol. 46, 455(1984 amplitude, sign, and duration are near those re-11 Patch Clamping: An Intro Guide to I. Dvir for help with experiments and data analys rted for whole-cell patch clamp recordings from cardiomyocytes(27, 28); thus, we conclude 12. R. M. Wightman, Science 311, 1570(2006). ngineering Faculty Fellow(NSSEFF) award ( N00244-09-1- that these data represent a transition to steady- 13.A G. Ewing, T G Strein, YY Lau, Acc. Chem. Res. 25, 0078), and the McKnight Foundation Neuroscience award. state intracellular recording(Fig. 4A, right) with 0(1992) the 3D nanowire probe 14. M G. Schrlau, N.J.Dun, H H. Bau, ACS Nano 3.563 Supporting Online Material Detailed analysis of the latter steady-state ncemag. org/cgifcontent/full329/5993/830/DC1 eaks(Fig. 4C, I shows five characteristic 15. ]. P. Donoghue, Nat. Neurosci. 5(suppL), 1085 Materials and Methods 2002 of a cardiac intracellular potential (27, 28), M. leong, B. Doris, ]. Kedzierski, K Rim, M. Yang eferences including(a)resting state, (b)rapid depolarization Science306,2057(2004) 10 May 2010: accepted 7 July 2010 (c) plateau, (d) rapid repolarization, and(e) hy- 17. M. Ferrari Nat Rev. Cancer 5, 161(2005) perpolarization. In addition, a sharp transient peak (blue star) and the notch (orange star) possibly associated with the inward sodium and outward potassium currents(28)can be resolved. Optical Terrestrial Gross Carbon dioxide intracellular peaks (ig. SSB)showed the kinked Uptake: Global Distribution and nanowire p region of the cell (19). When the PDMS/cell Covariation with Climate kinked nanowire devices, the intracellular peak disappeared,but they reappeared when the cell Christian Beer, Markus Reichstein, Enrico Tomelleri, Philippe Ciais, Martin Jung, 1 substrate was brought back into gentle contact Nuno Carvalhais 1,3 Christian Rodenbeck, M Altaf Arain, Dennis Baldocchi,6 with the device. This process could be repeated Gordon B Bonan, Alberte Bondeau, Alessandro Cescatti, "Gitta Lasslop, Anders Lindroth.10 multiple times without degradation in the rec- Mark Lomas, Sebastiaan Luyssaert, Hank Margolis, Keith W.oleson, devices were bent into a configuration with angle F lan Woodward,Dario Papale 0<-50o with respect to the substrate, or when kinked nanowire devices were fabricated on Terrestrial gross primary production(GPP)is the largest global CO2 flux driving several ecosystem planar substrates, we could record only extra- functions. We provide an observation-based estimate of this flux at 123+8 petagrams of carbon per cellular signals. These results confirm that elec- year (Pg C year )using eddy covariance flux data and various diagnostic models. Tropical forests and trical recording arises from the highly localized, savannahs account for 60%o. GPP over 40yo of the vegetated land is associated with precipitation pointlike nanoFET near the probe tip, which( State-of-the-art process-oriented biosphere models used for climate predictions exhibit a large tween-model variation of GPPs latitudinal patterns and show higher spatial correlations between imultaneously records both extracellular and GPP and precipitation, suggesting the existence of missing processes or feedback mechanisms which intracellular signals as the nanoFET spans the attenuate the vegetation response to climate. Our estimates of spatially distributed GPP and its cell membrane, and (i)records only intracellular covariation with climate can help improve coupled climate-carbon cycle process models. signals when fully inside the cell. Additional work remains to develop this new restrial plants fix carbon dioxide(co2) the major processes controlling land-atmosphere ynthetic nanoprobe as a routine tool like the as organic compounds through photo- CO2 exchange, providing the capacity of terres- patch-clamp micropipette (10, In), although we synthesis, a carbon (C) flux also known trial ecosystems to partly offset anthropogenic believe that there are already clear advantages: at the ecosystem level as gross primary produc- CO2 emissions. Electrical recording with kinked nanowire tion(GPP). Terrestrial GPP is the largest global Although photosynthesis at the leaf and can- probes is relatively simple without the need for carbon flux, and it drives several ecosystem func- opy level are quite well understood, only tentative resistance or capacitance compensation (9, In): tions, such as respiration and growth GPP thus observation-based estimates of global terrestrial the nanoprobes are chemically less invasive than contributes to human welfare because it is the GPP have been possible so far. Plant- and stand- pipettes, as there is no solution exchange, the basis for food, fiber, and wood production. In level GPP has previously been calculated as two small size and biomimetic coating minimizes me- addition, GPP, along with respiration, is one of times biomass production(1, 2), with substantial 834 13AuguSt2010Vol329ScieNcewww.sciencemag.org
the same time [fig. S5A (19)] are consistent with extracellular signals. After a relatively brief (~40-s) period of extracellular signals, we observed several pronounced changes in recorded signals (Fig. 4, B and C, II and III) without application of external force to the PDMS/cell support. Specifically, the initial extracellular signals gradually disappeared (Fig. 4, B and C, II, pink stars). There was a concomitant decrease in baseline potential, and new peaks emerged that had an opposite sign, similar frequency, much greater amplitude, and longer duration (Fig. 4B, II, green stars). These new peaks, which are coincident with cardiomyocyte cell beating, rapidly reached a steady state (Fig. 4B, III) with an average calibrated peak amplitude of ~80 mV and duration of ~200 ms. The amplitude, sign, and duration are near those reported for whole-cell patch clamp recordings from cardiomyocytes (27, 28); thus, we conclude that these data represent a transition to steadystate intracellular recording (Fig. 4A, right) with the 3D nanowire probe. Detailed analysis of the latter steady-state peaks (Fig. 4C, III) shows five characteristic phases of a cardiac intracellular potential (27, 28), including (a) resting state, (b) rapid depolarization, (c) plateau, (d) rapid repolarization, and (e) hyperpolarization. In addition, a sharp transient peak (blue star) and the notch (orange star) possibly associated with the inward sodium and outward potassium currents (28) can be resolved. Optical images recorded at the same time as these intracellular peaks (fig. S5B) showed the kinked nanowire probe tips in a possible intracellular region of the cell (19). When the PDMS/cell substrate was mechanically retracted from the 3D kinked nanowire devices, the intracellular peaks disappeared, but they reappeared when the cell substrate was brought back into gentle contact with the device. This process could be repeated multiple times without degradation in the recorded signal. When vertical 3D nanoprobe devices were bent into a configuration with angle q < ~50° with respect to the substrate, or when kinked nanowire devices were fabricated on planar substrates, we could record only extracellular signals. These results confirm that electrical recording arises from the highly localized, pointlike nanoFET near the probe tip, which (i) initially records only extracellular potential, (ii) simultaneously records both extracellular and intracellular signals as the nanoFET spans the cell membrane, and (iii) records only intracellular signals when fully inside the cell. Additional work remains to develop this new synthetic nanoprobe as a routine tool like the patch-clamp micropipette (10, 11), although we believe that there are already clear advantages: Electrical recording with kinked nanowire probes is relatively simple without the need for resistance or capacitance compensation (9, 11); the nanoprobes are chemically less invasive than pipettes, as there is no solution exchange; the small size and biomimetic coating minimizes mechanical invasiveness; and the nanoFETs have high spatial and temporal resolution for recording. References and Notes 1. D. A. Giljohann, C. A. Mirkin, Nature 462, 461 (2009). 2. T. Cohen-Karni, B. P. Timko, L. E. Weiss, C. M. Lieber, Proc. Natl. Acad. Sci. U.S.A. 106, 7309 (2009). 3. J. F. Eschermann et al., Appl. Phys. Lett. 95, 083703 (2009). 4. Q. Qing et al., Proc. Natl. Acad. Sci. U.S.A. 107, 1882 (2010). 5. I. Heller, W. T. T. Smaal, S. G. Lemay, C. Dekker, Small 5, 2528 (2009). 6. W. Lu, C. M. Lieber, Nat. Mater. 6, 841 (2007). 7. A. Grinvald, R. Hildesheim, Nat. Rev. Neurosci. 5, 874 (2004). 8. M. Scanziani, M. Häusser, Nature 461, 930 (2009). 9. R. D. Purves, Microelectrode Methods for Intracellular Recording and Ionophoresis (Academic Press, London, 1981). 10. B. Sakmann, E. Neher, Annu. Rev. Physiol. 46, 455 (1984). 11. A. Molleman, Patch Clamping: An Introductory Guide to Patch Clamp Electrophysiology (Wiley, Chichester, UK, 2003). 12. R. M. Wightman, Science 311, 1570 (2006). 13. A. G. Ewing, T. G. Strein, Y. Y. Lau, Acc. Chem. Res. 25, 440 (1992). 14. M. G. Schrlau, N. J. Dun, H. H. Bau, ACS Nano 3, 563 (2009). 15. J. P. Donoghue, Nat. Neurosci. 5 (suppl.), 1085 (2002). 16. M. Ieong, B. Doris, J. Kedzierski, K. Rim, M. Yang, Science 306, 2057 (2004). 17. M. Ferrari, Nat. Rev. Cancer 5, 161 (2005). 18. B. Z. Tian, P. Xie, T. J. Kempa, D. C. Bell, C. M. Lieber, Nat. Nanotechnol. 4, 824 (2009). 19. Materials and methods are available as supporting material on Science Online. 20. C. Conde, A. Cáceres, Nat. Rev. Neurosci. 10, 319 (2009). 21. T. G. Leong et al., Proc. Natl. Acad. Sci. U.S.A. 106, 703 (2009). 22. N. Misra et al., Proc. Natl. Acad. Sci. U.S.A. 106, 13780 (2009). 23. X. J. Zhou, J. M. Moran-Mirabal, H. G. Craighead, P. L. McEuen, Nat. Nanotechnol. 2, 185 (2007). 24. L. V. Chernomordik, M. M. Kozlov, Nat. Struct. Mol. Biol. 15, 675 (2008). 25. W. C. Claycomb et al., Proc. Natl. Acad. Sci. U.S.A. 95, 2979 (1998). 26. B. D. Almquist, N. A. Melosh, Proc. Natl. Acad. Sci. U.S.A. 107, 5815 (2010). 27. D. M. Bers, Nature 415, 198 (2002). 28. D. P. Zipes, J. Jalife, Cardiac Electrophysiology: From Cell to Bedside (Saunders, Philadelphia, ed. 2, 2009). 29. We thank G. Yellen, W. C. Claycomb, B. P. Bean, P. T. Ellinor, G. H. Yu, D. Casanova, B. P. Timko, and T. Dvir for help with experiments and data analysis. C.M.L. acknowledges support from a NIH Director’s Pioneer Award (5DP1OD003900), a National Security Science and Engineering Faculty Fellow (NSSEFF) award (N00244-09-1- 0078), and the McKnight Foundation Neuroscience award. Supporting Online Material www.sciencemag.org/cgi/content/full/329/5993/830/DC1 Materials and Methods Figs. S1 to S5 References 10 May 2010; accepted 7 July 2010 10.1126/science.1192033 Terrestrial Gross Carbon Dioxide Uptake: Global Distribution and Covariation with Climate Christian Beer,1 * Markus Reichstein,1 Enrico Tomelleri,1 Philippe Ciais,2 Martin Jung,1 Nuno Carvalhais,1,3 Christian Rödenbeck,4 M. Altaf Arain,5 Dennis Baldocchi,6 Gordon B. Bonan,7 Alberte Bondeau,8 Alessandro Cescatti,9 Gitta Lasslop,1 Anders Lindroth,10 Mark Lomas,11 Sebastiaan Luyssaert,12 Hank Margolis,13 Keith W. Oleson,7 Olivier Roupsard,14,15 Elmar Veenendaal,16 Nicolas Viovy,2 Christopher Williams,17 F. Ian Woodward,11 Dario Papale18 Terrestrial gross primary production (GPP) is the largest global CO2 flux driving several ecosystem functions. We provide an observation-based estimate of this flux at 123 T 8 petagrams of carbon per year (Pg C year−1 ) using eddy covariance flux data and various diagnostic models. Tropical forests and savannahs account for 60%. GPP over 40% of the vegetated land is associated with precipitation. State-of-the-art process-oriented biosphere models used for climate predictions exhibit a large between-model variation of GPP’s latitudinal patterns and show higher spatial correlations between GPP and precipitation, suggesting the existence of missing processes or feedback mechanisms which attenuate the vegetation response to climate. Our estimates of spatially distributed GPP and its covariation with climate can help improve coupled climate–carbon cycle process models. Terrestrial plants fix carbon dioxide (CO2) as organic compounds through photosynthesis, a carbon (C) flux also known at the ecosystem level as gross primary production (GPP). Terrestrial GPP is the largest global carbon flux, and it drives several ecosystem functions, such as respiration and growth. GPP thus contributes to human welfare because it is the basis for food, fiber, and wood production. In addition, GPP, along with respiration, is one of the major processes controlling land-atmosphere CO2 exchange, providing the capacity of terrestrial ecosystems to partly offset anthropogenic CO2 emissions. Although photosynthesis at the leaf and canopy level are quite well understood, only tentative observation-based estimates of global terrestrial GPP have been possible so far. Plant- and standlevel GPP has previously been calculated as two times biomass production (1, 2), with substantial 834 13 AUGUST 2010 VOL 329 SCIENCE www.sciencemag.org REPORTS