LETTERS

PUBLISHED ONLINE: 5 JULY 2009 | DOI: 10.1038/NNANO.2009.155

Partial sequencing of a single DNA molecule with a scanning tunnelling microscope

Hiroyuki Tanaka* and Tomoji Kawai*
The scanning tunnelling microscope is capable of the real-space imaging and spectroscopy of molecules on an atomic scale. Numerous attempts have been made to use the scanning tunnelling microscope to sequence single DNA molecules, but difculties in preparing samples of long-chain DNA molecules on surfaces, and problems in reproducing results have limited these experiments16. Here, we report single-molecule DNA sequencing with a scanning tunnelling microscope by using an oblique pulse-injection method to deposit the molecules onto a copper surface. First, we show that guanine bases have a distinct electronic state that allows them to be distinguished from the other nucleic acid bases. Then, by comparing data on M13mp18, a single-stranded phage DNA, with a known base sequence7, the electronic ngerprint of guanine bases in the DNA molecule is identied. These results show that it is possible to sequence individual guanine bases in real long-chain DNA molecules with high-resolution scanning tunnelling microscope imaging and spectroscopy. We have previously detected the characteristic electronic states of the uorescent dye molecule uorescein isothiocyanate (FITC) in single-molecule scanning tunnelling microscope (STM) observations and I2V measurements of short strands of DNA and peptide nucleic acid (PNA) labelled with FITC8, but we could not observe peaks in the characteristic electronic states of base molecules due to scatter and the poor signal-to-noise ratio of the IV curve911. Here, by using lock-in detection8, we have successfully detected electronic states that are exclusively characteristic of DNA/PNA. Figure 1 shows typical STM images of single-stranded monomers of FITC-modied PNA (FITCTTGACC) and DNA (FITC TTGGCC), as well as the spectroscopic data for these samples. As can be seen from the wide-area images (Fig. 1a,d), it is difcult to fully determine the base sequences of oligomers due to the existence of other adsorbed materials (thought to be nucleotide monomers, buffers and salts that were not completely eliminated in the purication process) and the diversity of the adsorbed structures of the oligomers themselves. However, by comparing two types of oligomer, it is possible to assign guanine. With FITCTTGACC, only one base is seen to shine in the middle of the base sequence, whereas with FITCTTGGCC there are two, as can be seen from the wide-area image (Fig. 1d) and enlarged image (Fig. 1e). Furthermore, with tunnelling spectroscopy obtained by lock-in detection, it can be seen that the guanine base molecules differ from the other base molecules and the underlying substrate in that they have a characteristic peak in the electronic state in the vicinity of an applied voltage (Vs) of 21.6 V (Fig. 1c,f ). Of the four bases, it is thought that guanine is the easiest to detect because it has the lowest ionization potential12. To the best of our knowledge, there are no previous STM reports where the characteristic state density of base molecules, adsorbed onto a metal substrate, have been detected in this way4,911,13,14. Recent theoretical calculations15 of DNA base molecules adsorbed on a Cu(111) surface suggest that peaks in the density of states capable of being measured by STM should exist for all four types of base molecules and not just guanine. Although the calculated energy of the rst peak below the Fermi energy of the guanine/copper agrees well with our experimental values, the theoretical predictions for the other bases are inconsistent with our STM measurements. It is not understood why this characteristic electronic state occurs only for guanine bases. However, to determine if these results are also correct for real DNA, and if they are effective for sequencing, we analysed long-chain DNA strands. The difculties encountered in the past in sequencing single long-chain DNA molecules with the STM have been due, in part, to the unsuitability of the sample preparation methods2. Although researchers have managed to observe the macro structure of DNA3,4, few studies have been able to make use of the full resolving power of the STM, that is, the clear and reproducible observation of individual nucleotides. Detailed STM imaging, spectroscopy and manipulation studies of molecules are almost always restricted to molecules that can be deposited onto a surface from the gas phase16,17, although alternative sample preparation methods have been examined1,1820. By developing a pulse-injection technique for depositing molecules onto substrates that cannot be vapour deposited (including supramolecular and biological molecules such as DNA), we have previously achieved reliable and repeatable high-resolution STM observations of the helical period of circular double-stranded (plasmid) DNA, and of the individual nucleotides in single-stranded (ss) DNA oligomers and DNA polymers1,5,6. However, this technique suffers from the fact that ssDNA can easily form secondary structures due to intrastrand base-pairing, with the result that sections of sufcient length to allow sequencing are rarely observed6. Most of the DNA stretching methods that have so far been developed for ascertaining the dynamics of DNA and analysing single-molecule genes are either performed in liquid21 or atmospheric conditions22, making them unsuitable for highresolution, ultrahigh-vacuum studies. It was therefore necessary to develop a technique for stretching out ssDNA strands. We have developed23 a method for stretching out and xing single-stranded M13mp18 DNA molecules (which contains 7,249 bases)7 by using the ow effects resulting from the oblique injection8 of a DNA solution onto a substrate using a pulse-injection technique (Fig. 2a). When the DNA solution is injected perpendicularly onto the substrate, no extended DNA is observed (see Supplementary Fig. S1). Figure 2b shows a typical wide-region STM image (400 nm wide) of an ss M13mp18 extended and xed by oblique injection on a Cu(111) surface. Owing to the height differences at steps in the substrate, the contrast is not ideal for recognizing DNA molecules, but it is possible to recognize two DNA strands running from left to right. Figure 3 shows the image obtained over a wider area (800 nm wide). As a result of the wide scan

The Institute of Scientic and Industrial Research, Osaka University, Mihogaoka, Ibaraki, Osaka 567-0047, Japan. *e-mail: hrtanaka@sanken.osaka-u.ac.jp; kawai@sanken.osaka-u.ac.jp
518
NATURE NANOTECHNOLOGY | VOL 4 | AUGUST 2009 | www.nature.com/naturenanotechnology 2009 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY
a

DOI: 10.1038/NNANO.2009.155

LETTERS
c
5 4 3 2 1 0 1 2 1.5 1 0.5 Cu(111) FITC Thymine Adenine/Cytosine Guanine

FITC T C T G A C 1 nm

dI/dV (a.u.)

5 nm

0 0.5 Vs (V)

1.5

d
1

f
6 Point 1 Point 2 Point 3 Point 4 Point 5 Point 6 Point 7 Point 8 Cu Cu Guanine Guanine Guanine Guanine Guanine Guanine

5 6

8 dI/dV (a.u.) C T C 5 nm 1 nm

5 4 3 2 1 0

FITC 3 4 T

2 1.5 1 0.5

0 0.5 Vs (V)

1.5

Figure 1 | STM imaging and spectroscopy of single-stranded peptide nucleic acid (ssPNA) and DNA adsorbed on a Cu(111) surface. a, A wide-area image of FITC-modied ssPNA (FITCTTGACC; sample voltage 0.2 V, tunnel current 2 pA). A variety of shapes are formed, each measuring $5 nm, and the adsorbed molecules with the brightest contrast at one end are single FITCTTGACC molecules. b, A high-magnication image of ssPNA (0.005 V, 10 pA), and a model of this molecule in vacuum, calculated using Spartan 04, molecular mechanics (MMFF). c, The dI/dV spectra measured from the marker molecules, the individual nucleobases and the Cu(111) substrate. d, A typical wide-area image of FITC-modied DNA (FITCTTGGCC; 21.8 V, 2 pA). It should be noted that, unlike the PNA image, this image was obtained at a negative voltage. As shown by the elliptical dashed lines, there are two adjacent brighter points (nucleobases) at the centre of the molecules. e, A high-magnication image of a single FITCTTGGCC molecule (22 V, 2 pA), and a model of this molecule in vacuum, calculated using Spartan 04, molecular mechanics (MMFF). f, The dI/dV spectra measured from the two adjacent bright nucleobases in d, and the Cu(111) substrate. Points 18 represent the measurement points shown in d.

region and the roughness of the substrate relative to the DNA, the contrast of the DNA is relatively weak (Fig. 3a). To aid visualization, a three-dimensional structure and differential image are also shown in Fig. 3b and c, respectively. An M13mp18 strand with 7,249 bases has a straight-chain length of $5 mm, when calculated at 0.65 nm base21 (ref. 6). Because this exceeded the 3-mm square scanning area of our STM images, it was not possible to image the entire length. Using this extended DNA, and to check whether or not it is possible to assign the individual guanine units, we measured topography images and dI/dV map images over the 100-nm-wide region highlighted in Fig. 2b. In the topography image of Fig. 2c, the individual nucleotides are shown as bright points, which are exceptionally bright in some places. The nucleotides that appeared brightest in the topography image also appear as clear bright points in the dI/dV map of Fig. 2d. It should be noted that dI/dV maps obtained at around the peak Vs make differentiation clearer (see Supplementary Fig. S2). Furthermore, in the spectroscopic data obtained on these bright points, there is a peak in the density of states at Vs 1.6 V (Fig. 4), which is similar to the guanine peak obtained with the oligomers (Fig. 1). For comparison, part of the known base sequence of M13mp18 (ref. 7) is shown in Fig. 2e. The observed image and the known guanine sequence match almost perfectly, illustrating that STM can be used to sequence

the guanine in real DNA. (The possible sequence assignment is shown overlaid on the STM image in Supplementary Fig. S3a.) As control experiments, the same base sequence region of different DNA strands was also investigated. Supplementary Fig. S5d,e shows an STM image and dI/dV map, respectively, of the same base position as shown in Supplementary Fig. 5a,b, but for a different DNA strand. In both the STM image and the dI/dV map of Supplementary Fig. S5d,e, the guanines appear as distinct, bright features. Moreover, they have the same contrast variation as Fig. 5a,b, but the contaminants (the yellow arrows) do not. Further details on how particular segments were searched for and identied are given in the Supplementary Information. To make further advances towards the use of STM as a practical tool for sequencing, we must be able to recognize all four types of base molecule, and achieve greater speed and precision. In general, lock-in detection sacrices temporal resolution for the sake of improved signal-to-noise ratio. In our measurement system, to obtain the dI/dV spectrum, it was necessary to stop the scanning and feedback of the STM probe and then measure the IV characteristic, so that a time ranging from 1 s to 1 min was needed for each point. Furthermore, to obtain the dI/dV map of Fig. 2d required a period of almost 1 h. If STM software and control mechanisms capable of nding chain-shaped polymers such as DNA can be developed, then the time taken to scan parts where the sample is not present can be
519

NATURE NANOTECHNOLOGY | VOL 4 | AUGUST 2009 | www.nature.com/naturenanotechnology 2009 Macmillan Publishers Limited. All rights reserved.

LETTERS
a b

NATURE NANOTECHNOLOGY

DOI: 10.1038/NNANO.2009.155

20 nm

5 nm

5 nm

e
at ggacagact ct t t t act cggt ggcct cact gat t at aaaaacact t ct caagat t ct ggcgt accgt t cct gt ct aaaat ccct t t aat cggcct cct gt t t agct cccgct ct gat t ccaacgaggaaagcacgt t a

Figure 2 | Deposition and STM analysis of single-stranded M13mp18 DNA molecules on a Cu(111) surface. a, Schematic illustration of the oblique injection method. DNA strands tend to align more perpendicular than horizontal to the ow (injection) direction. The Cu(111) substrate is inclined at $458 to the aqueous solution of DNA, which is introduced from a pulse valve. b, Typical wide-area image of M13mp18 (2 V, 5 pA, width 400 nm). Atomic steps in the Cu(111) substrate form a staircase surface structure. In this image, sections of M13mp18 are visualized as linear adsorbed material running from the top left to bottom right. c, An enlarged view of the rectangular region enclosed by the white dashed line in b (2 V, 5 pA, 100 nm). d, A dI/dV map of the same region as in c (1.5 V, 20 pA, 100 nm). To maximize the detection of the density of states of guanine, the measurements were made under slightly lower bias conditions than in c. e, Part of the base sequence of M13mp18 obtained from a databank (the sequence of bases at positions 5322 through 5461). To facilitate comparison with the STM data, the guanine sites are indicated by red characters and are also connected by red arrows to the corresponding parts of the image.

50 nm

Figure 3 | STM imaging of M13mp18 DNA extended and xed on a Cu(111) surface. a, A typical wide-area image of extended M13mp18 (22 V, 5 pA, width 800 nm, height 10 nm). A large number of steps are present in the Cu(111) substrate. Compared with the contrast of this image (10 nm), the height of ssDNA (0.20.3 nm) is small, making it difcult to recognize the DNA from the raw data. b, The topographic data of a is shown here converted into pseudo-three-dimensional form with WSxM imageprocessing software (www.nanotec.es/) to make the DNA easier to recognize. c, The topographic data of a is shown here in a differentiated image, converted with TIFS image-processing software (www2.big.or.jp/ $osamu/TIFS/), to make the DNA easier to recognize. The ends of the extended DNA strand in the images are indicated by the red arrows.

greatly reduced. Savings in cost and time can also be made if sequencing is performed from the topographic image alone without using a lock-in amplier, but it would still be necessary to use a method for identifying contamination.
520

A method that is relatively fast and easy to implement involves comparing the STM bias dependence. That is, if guanine can be distinguished by comparing two images obtained at a value of Vs that is much lower than the peak in the density of states for guanine (for example, 22 V) and in fact at a value close to 0 V, then it should be possible to perform sequencing at high speed without the need for spectroscopy. As shown in Fig. 5, comparing images obtained at different bias voltages helps distinguish between DNA base molecules and contaminants: irregular points of brightness that do not vary with the bias voltage (yellow arrows in Fig. 5a) are contaminants. The undulating pattern surrounding the DNA chain in Fig. 5a is observed at low bias conditions (Vs 0.1 V), and is thus a standing wave resulting from the scattering of surface electrons24. Low bias conditions result in higher resolution (clearer image) because the tip comes closer to the sample. Compared with the image obtained with a low bias voltage, some of the nucleotides in the high bias image (Vs 22 V) are brighter and thus correspond to guanine. To verify this, we compared the results with the sequence extracted from a databank (at positions 29113011) as shown in Fig. 5c, which can conrm if the guanine base molecules are completely matched either individually or in groups (g, gg, ggg). Unfortunately, there was insufcient resolution to state categorically that the numbers of nucleotides other than guanine that are separated by the guanine patterns were fully matched. We did, however, verify a complete match with regard to the pattern of guanine sequences. A possible sequence assignment overlaid on the STM image to aid comparison is shown in Supplementary Fig. S3b. It can thus be seen that it is possible to recognize the guanine pattern with few errors simply by obtaining a pair of STM topographic images in this way (preferably dual bias mode). (See Supplementary Fig. S4 for further examples.)

NATURE NANOTECHNOLOGY | VOL 4 | AUGUST 2009 | www.nature.com/naturenanotechnology 2009 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY
a
1 2 3 4 5 6

DOI: 10.1038/NNANO.2009.155

LETTERS

8 9

b
5 nm

3.5 3 2.5 2 dI/dV (a.u.) 1.5 1 0.5 0 0.5 2 Area 1 Area 2 Area 3 Area 4 Area 5 Area 6 Area 7 Area 8 Area 9

By closely examining the sections of the STM images of Fig. 5 where the nucleotides are neatly arranged, it appears as though the cytosine units are smaller than the thymine units (white arrows). However, the apparent size of the nucleotide units in the STM images are thought to be largely dependent on the orientation of the base molecule, so the cytosine units may not always appear to be smaller than the thymine units. It should also be noted that the measured spectroscopy of ssDNA is not the same as for double-stranded DNA. In conclusion, by developing a method for extending and xing DNA strands, we have taken a step towards the realization of electronic-based single-molecule DNA sequencing. Of the four bases, we were able to precisely identify guanine because the STM is able to pick up on the characteristics of its electronic state, which is largely independent of the adsorption structure. If vibrational spectroscopy is performed using inelastic electron tunnelling spectroscopy16, it should be possible to identify all of the base molecules. Furthermore, because STM can select a specic position of interest along a DNA strand, the technique could have a unique advantage in analysing, for example, single nucleotide polymorphisms.

Methods
The substrate used was a Cu(111) surface, cleaned to be atomically at by argon ion sputtering at 773 K in ultrahigh vacuum1. Single-stranded PNA (FITCTTGACC) and DNA (FITCTTGGCC) oligomers were purchased from FASMAC, Japan. FITC (uorescein isothiocyanate) is a derivative of uorescein used as a uorophore. M13mp18 ssDNA was purchased from Bayou Biolabs and was subjected to dialysis (Slide-A-Lyzer MINI Dialysis Unit, 10,000-molecular-weight-cut-off type; Rockford) against water to remove excess buffer solution and salt. Oligos and M13 mp18 DNA were respectively dissolved in water at concentrations of 2 mmol l21 and 0.5 nmol l21 and were deposited on clean Cu(111) surfaces using the oblique pulseinjection method at room temperature, shown in Fig. 2a (ref. 23). The distance from the pulse valve to the substrate was 50 mm. The solution was injected towards the substrate when the valve was opened for 1.5 ms. When the DNA solution was injected perpendicularly onto the substrate, no extended DNA was observed in atomic force microscopy images (see Supplementary Fig. S1a,c). When the DNA solution was injected at a slanting angle such as 458 onto the substrate, extended DNA was observed (see Supplementary Fig. S1b,d; Fig. 2b). In these images, the injected solution passed from up to down. DNA strands tend to align more perpendicular to than horizontal to the ow direction. In addition to the extended DNAs, short DNA fragments (,1 mm) were found that were not linearly extended. Although the detailed mechanism is still being studied, it is thought that the oblique ow contributes to this extension. We used two low-temperature STM systems, one (LTSTM, Omicron GmbH) for FITCTTGACC and M13mp18 DNA, and the other (USM-1200S2N1, Unisoku) for FITCTTGGCC. Both STM chambers (below 1.0 1028 Pa) were cooled by liquid nitrogen, and the observation temperature was 80 K. The tunnelling current for topographic images was set to less than 10 pA to avoid decomposition of the DNA molecules or changes in the scanning. Tunnelling spectra of dI/dV (refs 25, 26) were simultaneously measured using a lock-in amplier system with current measurement. Sinusoidal modulation of 12 kHz in frequency and 0.1 V r.m.s. in

1.5

0.5

0 Vs (V)

0.5

1.5

Figure 4 | Partial topographic image, dI /dV map and dI /dV spectra of M13mp18 DNA. a, Topographic image of M13mp18 taken with an STM (21.4 V, 5 pA, horizontal width 50 nm). b, dI/dV map measured at the same time as the topographic image: the use of a lock-in amplier means that the contrast in this image is less affected by thermal drift than the topographic image. Brighter regions in this image correspond to a larger density of states. The guanine units are seen as bright spots. Because the dI/dV spectrum was obtained while capturing the STM image, the resulting image consists of discontinuous narrow strips. c, The dI/dV spectra measured in the areas indicated by points 19 in the topographic image. At each point, the average of multiple dI/dV spectra was obtained. To suppress the effects of thermal drift, the dI/dV spectra are measured in a shorter time interval than in Fig. 1, so there is more noise and greater variation in each spectrum.

5 nm t t t g t t cg g c t a t c t gc t t a c t t t t c t t aaaaagg g c t t c g g t a a g a t a g c t a t t gc t a t t t c a t t g t t t c t t g c t c t t a t t a t t g g g c t a a c t c a

Figure 5 | Bias voltage dependence on the STM imaging of M13mp18 DNA on a Cu(111) surface. a,b, Raw data obtained with the same tip on the same 70 nm 9 nm area: Vs 0.1 V (a), and Vs 22 V (b). The ripple pattern on the surface is a standing wave in the electrons of the Cu(111) surface24. c, Part of the known base sequence of M13mp18 (the sequence of bases at positions 2911 to 3011)7. To facilitate comparison with the STM data, the guanine sites are indicated by red characters and are also connected by red arrows to the corresponding parts of the image. Some of the cytosine units are indicated by blue characters and the corresponding nucleotides in the STM images (a,b) are indicated by white arrows. The yellow arrows in the STM images indicate contaminants.
NATURE NANOTECHNOLOGY | VOL 4 | AUGUST 2009 | www.nature.com/naturenanotechnology 2009 Macmillan Publishers Limited. All rights reserved.

521

LETTERS

NATURE NANOTECHNOLOGY

DOI: 10.1038/NNANO.2009.155

amplitude was superimposed on the sample bias voltage. The dI/dV map was measured with lock-in detection of the a.c. tunnelling current by modulating the sample bias (0.1 V r.m.s., 1 kHz) while keeping the feedback loop active. It should be mentioned that a high electric eld between the tip and sample can result in bending, migration, diffusion, decomposition, evaporation, and so on, even at cryogenic temperatures. So, wide-gap molecules, such as base molecules, are expected to be difcult to analyse, in contrast to dye molecules such as FITC, which have molecular orbitals near the Fermi level. Fortunately, guanine has a peak at around 1.6 V. If this was much lower (for example, at 3 V), the contrast in STM/STS would become weak and the junction unstable, and hence more difcult to measure.

Received 2 June 2008; accepted 3 June 2009; published online 5 July 2009

References
1. Tanaka, H. & Kawai, T. Scanning tunneling microscopy imaging and manipulation of DNA oligomer adsorbed on Cu(111) surfaces by a pulse injection method. J. Vac. Sci. Technol. B 15, 602604 (1997). 2. Clemmer, C. R. & Beebe, T. P. Jr. Graphite: a mimic for DNA and other biomolecules in scanning tunneling microscope studies. Science 251, 640642 (1991). 3. Guckenberger, R. et al. Scanning tunneling microscopy of insulators and biological specimens based on lateral conductivity of ultrathin water lms. Science 266, 15381540 (1994). 4. Shapir, E. et al. Electronic structure of single DNA molecules resolved by transverse scanning tunnelling spectroscopy. Nature Mater. 7, 6874 (2008). 5. Tanaka, H., Hamai, C., Kanno, T. & Kawai, T. High-resolution scanning tunneling microscopy imaging of DNA molecules on Cu(111) surfaces. Surf. Sci. 432, L611L616 (1999). 6. Tanaka, H. & Kawai, T. Visualization of detailed structures within DNA. Surf. Sci. 539, L531L536 (2003). 7. Ebright, R., Dong, Q. & Messing, J. Corrected nucleotide sequence of M13mp18 gene III. Gene 114, 8183 (1992). 8. Tanaka, H. & Kawai, T. High-resolution scanning tunneling microscopy and dI/dV map studies of peptide nucleic acid and uorescein isothiocyanate. Appl. Surf. Sci. 252, 54745476 (2006). 9. Nishimura, M., Tanaka, H. & Kawai, T. Structure of linear double-stranded deoxyribonucleic acid adsorbed on Cu(111) surfaces: a low-temperature scanning tunneling microscopy study. Jpn J. Appl. Phys. 41, 75107511 (2002). 10. Nojima, Y., Tanaka, H., Yoshida, Y. & Kawai, T. High-resolution scanning tunneling microscopy and spectroscopy studies of deoxyribonucleic acid and uorescein isothiocyanate. Jpn J. Appl. Phys. 43, 55265527 (2004). 11. Yoshida, Y., Nojima, Y., Tanaka, H. & Kawai, T. Scanning tunneling spectroscopy of single-strand deoxyribonucleic acid for sequencing. J. Vac. Sci. Technol. B 25, 242246 (2007). 12. Yang, X., Wang, X. B., Vorpagel, E. R. & Wang, L. S. Direct experimental observation of the low ionization potentials of guanine in free oligonucleotides by using photoelectron spectroscopy. Proc. Natl Acad. Sci. USA 101, 1758817592 (2004).

13. Ohshiro, T. & Umezawa, Y. Complementary base-pair-facilitated electron tunneling for electrically pinpointing complementary nucleobases. Proc. Natl Acad. Sci. USA, 103, 1014 (2006). 14. Xu, M., Endres, R. G. & Arakawa, Y. The electronic properties of DNA bases. Small 3, 15391543 (2007). 15. Kilina, S. et al. Electronic properties of DNA base molecules adsorbed on a metallic surface. J. Phys. Chem. C 111, 1454114551 (2007). 16. Ho, W. Single-molecule chemistry. J. Chem. Phys. 117, 1103311061 (2002). 17. Katano, S., Kim, Y., Hori, M., Trenary, M. & Kawai, M. Reversible control of hydrogenation of a single molecule. Science 316, 18831886 (2007). 18. Yamada, T., Ge, M., Shinohara, H., Kimura K. & Mashiko, S. Study on the interaction between different solute molecules in a molecular beam produced by the spray-jet technique: an application to dendrimer/dye system Chem. Phys. Lett. 379, 458465 (2003). der, H. J. et al. Processing of giant graphene molecules by soft-landing mass 19. Ra spectrometry. Nature Mater. 5, 276280 (2006). 20. Laracuente, A. R., Baker, L. A. & Whitman, L. J. UHV characterization of ambient-dosed hydrogen-terminated Si(001). Surf. Sci. 602, 38 (2008). 21. Bensimon, A. et al. Alignment and sensitive detection of DNA by a moving interface. Science 265, 20962098 (1994). 22. Fang, X. & Reneker, D. H. DNA bers by electrospinning, J. Macromol. Sci. B 36, 169173 (1997). 23. Tanaka, H. & Kawai, T. Japanese patent Kokoku 2005-46665 (2005). 24. Crommie, M. F., Lutz, C. P. & Eigler, D. M. Imaging standing waves in a two-dimensional electron gas. Nature 363, 524527 (1993). 25. Bonnell, D. A. (ed.) Scanning Tunneling Microscopy and Spectroscopy: Theory, Techniques, and Applications (Wiley, 1993). 26. Scudiero, L., Barlow, D. E., Mazur, U. & Hipps K. W. Scanning tunneling microscopy, orbital-mediated tunneling spectroscopy and ultraviolet photoelectron spectroscopy of metal(II) tetraphenylporphyrins deposited from vapor. J. Am. Chem. Soc. 123, 40734080 (2001).

Acknowledgements
The authors thank A. V. Balatsky and S. Kilina for valuable comments and correction of the manuscript, and H. Sakamoto and K. Suzuki for their technical assistance. This work was nancially supported by CREST, JST (Japan Science and Technology Agency) and NEDO (the New Energy and Industrial Technology Development Organization). H.T. is grateful for a Grant-in-Aid for Scientic Research for the Encouragement of Young Scientists (B) (no.14750236) from the Japan Society for the Promotion of Science (JSPS).

Author contributions
H.T. and T.K. conceived and designed the experiments. H.T. performed the experiments and analysed the data. H.T. wrote the paper. H.T. and T.K. discussed the results.

Additional information
Supplementary information accompanies this paper at www.nature.com/ naturenanotechnology. Reprints and permission information is available online at http://npg. nature.com/reprintsandpermissions/. Correspondence and requests for materials should be addressed to H.T. and T.K.

522

NATURE NANOTECHNOLOGY | VOL 4 | AUGUST 2009 | www.nature.com/naturenanotechnology 2009 Macmillan Publishers Limited. All rights reserved.