PRL 98, 216602 (2007)

PHYSICAL REVIEW LETTERS

week ending 25 MAY 2007

End and Central Plasmon Resonances in Linear Atomic Chains

Jun Yan,1,2 Zhe Yuan,3,1 and Shiwu Gao1,2,*
Institute of Physics, Chinese Academy of Sciences, 100080 Beijing, China Department of Physics, Goteborg University, SE-41296, Goteborg, Sweden 3 Department of Applied Physics, Chalmers University of Technology, SE-41296, Goteborg, Sweden (Received 28 January 2007; revised manuscript received 17 April 2007; published 25 May 2007)
2 1

The existence and nature of end and central plasmon resonances in a linear atomic chain, the 1D analog to surface and bulk plasmons in 2D metals, has been predicted by ab initio time-dependent density functional theory. Length dependence of the absorption spectra shows the emergence and development of collectivity of these resonances. It converges to a single resonance in the longitudinal mode, and two transverse resonances, which are localized at the ends and center of the atom chains. These collective modes bridge the gaps, in concept and scale, between the collective excitation of atomic physics and nanoplasmonics. It also outlines a route to atomic-scale engineering of collective excitations.
DOI: 10.1103/PhysRevLett.98.216602 PACS numbers: 72.15.Nj, 71.45.Gm, 73.20.Mf

Collective electron oscillations in nanostructures form localized surface plasmon resonances (LSPR) [17], which exhibit unusual properties in energy, lifetime, optical spectroscopy, and short-time dynamics. Such properties of LSPRs, which are tunable by the shapes and sizes of the structures, are promising for applications in sensing and spectroscopy [1,2], catalytic reactions [8], and biomedical treatment [9]. The quest for new nanostructures of desired purpose and function has been a major ongoing effort in nanoplasmonics and condensed matter physics [10 12]. Various structures with different shapes have been fabricated using electron beam and colloidal lithography [13]. The minimal dimension reachable by lithography has so far been limited by a few tens nanometers. The plasmon excitations in this regime (10 1000 nm) can be qualitatively understood by electrodynamic models like the Mie theory [14]. This letter draws attention to collective excitations of a different class of structures, articial atomic structures assembled with a scanning tunneling microscope (STM). Prototype structures of this kind include, for example, linear atomic chains [15,16], quantum corrals [17], and articial molecular structures [18]. Different from the nanostructures created by lithography and self-assembly, the sizes and shapes of such articial structures are tunable down to the precision of a single atom. Collectivity of electron oscillations at such scales is a fundamental subject of its own interest [19]. It should be strongly inuenced by two competing factors: the local atomic connement and global quantization imposed by the geometric boundary of the systems. The aim of the present study was to nd out whether and how collective plasmon excitation are determined by the structural tunability at atomic scale. As both a conceptual model and a concise numerical study, we have calculated the plasmon excitation of a linear atomic chain of sodium atoms, which can be assembled atom by atom with an STM [16]. Using an ab initio description of the valence electrons, the optical spectra are 0031-9007=07=98(21)=216602(4)

calculated by time-dependent density functional theory (TDDFT). The evolution of absorption spectra, as the chain length was varied, clearly demonstrates the integration of its oscillator strength, which starts from two atoms and accumulates atom by atom. These collective oscillations converge into a single plasmon resonance in the longitudinal mode but splits into two separate transverse modes, one located at the ends (the end or surface plasmon) and the other in the center of the chain (the central or bulk plasmon). In addition, the frequency and polarization of these resonances can be manipulated by attaching a single atom or molecule at different registry to the chains. These atomic plasmons naturally complement and bridge the gaps, in concept and size, between the nanoplasmonics (10 1000 nm) and collective excitation in atomic (a few A) and nuclear (a few fermis) systems [20]. They have also general implications in atomic-scale engineering and manipulations of optical properties atom by atom. All of our calculations have been done by a real-space and real-time TDDFT code, OCTOPUS [21], which treats the Na atoms using norm-conserving pseudopotentials. Localdensity approximation (LDA) for the exchange-correlation is used in both the ground state and excited state calculations. The simulation zone is dened by assigning a sphere around each atom with a radius of 6 A and a uniform mesh . The absorption spectra were obtained by two grid of 0.3 A approaches: 1) real-time propagation of electron wave packet [22] under an impulse excitation, and 2) linearresponse formalism in frequency domain developed by Casida [23]. In the real-time propagation, the electronic wave packets were evolved for typically 10 000 steps with a time step of 0.003 fs. Excitation spectrum is extracted by Fourier transform of the dipole strength. Such a scheme is extensively tested on several well-studied examples, on which reliable absorption spectra have been reproduced. Linear-response calculation in the frequency domain gives essentially the same excitation spectra for all the chains we calculated. Here we demonstrate our results with chains 2007 The American Physical Society

216602-1

PRL 98, 216602 (2007)

PHYSICAL REVIEW LETTERS

week ending 25 MAY 2007

built from simple metal atoms like Na, calculations for silver atoms (with d electrons) have also been performed, which shows qualitatively similar results. Figure 1 shows the evolution of dipole strength as a function of the chain length, which is measured by the number of atoms Na . The left (a) and right panels (b) correspond to longitudinal (L) and transverse (T) polarizations of dipole excitations, respectively. For a single Na atom (the bottom curve), only one peak at 2.37 eV appears. It is comparable to 2.10 eV in literature [24,25]. It consists of predominantly the 3s ! 3p transition. For the Na dimer, the interatomic distance is taken to be 2.89 A, chosen from the experimental conguration of atom chains on the NiAl(110) surface [16]. The dipole excitation (the second curve from the bottom) splits into two peaks, a longitudinal mode (the L mode) at 2.20 eV and a transverse one (the T mode) at 2.92 eV with comparable strength. These energies are again comparable to absorption peaks 2.092.63 eV [24,25] documented in the literature. The minor offset in energies can be attributed to the difference in the interatomic distances used in the calculations. As more atoms are added to the chain, different evolutions were observed for the L and T modes. Starting from Na3, the L mode, panel (a), redshifts in energy and its intensity increases linearly with the chain length. The linear increase in intensity results from the accumulation of collectivity in the dipole oscillation, as found in a previous model using a conned quasi-one-dimensional electron gas (CQ1DEG) [26]. The redshift of the resonance frequency at increased chain length can be understood by the reduction of the energy gaps involved in the dipole excitation. Similar redshift was found in the CQ1DEG [26]. The self-consistent calculation in this study yields more reliable excitation energies and oscillator strengths.

In particular, the f-sum rule, which was not strictly satised in the CQ1DEG, is now fullled at all lengths. Apart from the lowest dipole mode, multiple excitations (n 2 and 3) of the L mode are also discernible in longer chains, which was not found in CQ1DEG. In the rest of this Letter, we will focus on the T mode, which was not studied before. The evolution of the T modes, panel (b), differs drastically from that of the L mode. The transverse peak of the dimer at 2.92 eV is blueshifted and splits into two branches at Na3 with a low-energy resonance at 3.11 eV and a highenergy band at 4.44 eV. These two peaks gain intensity and become more prominent as the length increases. Figure 2 shows the length dependence of energies (a) and oscillator strength (b) of the transverse dipole modes. For chains shorter than 5 atoms, the transverse modes are dominated by the low-energy resonance, !TE . From Na5, the highenergy branch increases in intensity and forms a welldened resonance, !TC , which is separated from !TE by a nite gap. It suggests that the two modes are nondegenerate in energy and character. In longer chains, the strength of the !TC mode increases almost linearly with the length. In contrast, the intensity of the !TE mode becomes saturated with an integral oscillator strength of about 4 electrons. The strength of the two T modes indicates that both

(a)
Energy (eV)
4
TC

TE

2
L

80

(a)

(b)
L T
TC

(b)
Strength (1/eV)
L TC

60

Strength (1/eV)

TE

40 Na=18 20 Na=10 Na= 4 Na= 1 0 1 2 3 4

20
TE

0 0 4 8 12 16 20

Energy (eV)

Na
FIG. 2. The excitation energy (a) and the dipole strength (b) as a function of the number of atoms Na for the three resonance peaks shown in Fig. 1. They can be assigned to the longitudinal (!L ) and transverse plasmon modes (!TC and !TE ), as discussed in the text.

FIG. 1 (color online). The dipole response (optical absorption) of a linear sodium chain, with variable number of atoms Na , to an impulse excitation, resulting in a momentum change k  1=0:05 A), polarized in the longitudinal (a) and transverse direction (b). The series of spectra corresponds, counting from the bottom, to Na 1, 2, 3, 4, 6, 8, 10, 14, and 18.

216602-2

PRL 98, 216602 (2007)

PHYSICAL REVIEW LETTERS

week ending 25 MAY 2007

are collective in character. Although there is no measurement available for these collective resonances in atomic chains, qualitative length dependence (the blueshift and enhancement) were indeed observed in nanoparticle arrays [27,28]. In the experiment, it was however impossible to identify the number of the T modes due to the energy resolution and the small number of particles (up to 7) assembled in the experiment. To elucidate the nature of the T modes, the induced charge response has been analyzed in real-time propagation. Figure 3 shows the Fourier transform of the induced density, obtained from time-evolution, in a plane crossing the chains at the plasmon frequencies for Na 3 (row a), 8 (row b), and 18 (row c), respectively. The density proles for both modes are dipole like in character and differ in their spatial distributions. The !TE mode is localized at the two ends, while !TC is distributed in the center of the chains. This difference is already discernible, despite strong mixing, for short chains as Na 3 (row a), becomes clearer at Na 8 (b), and is completely separated in Na 18 (c). This charge distribution is consistent with the energy and strength of the two modes shown in Fig. 2, where the strength of !TC starts to increase linearly in the long chain limit. It is interesting to note that the charge oscillation of the central mode has nearly vanishing am-

plitudes on the 6 atoms at the two ends. For the !TE mode, the intensity and charge distribution does not change signicantly for chains longer than 8 atoms, and are completely localized on the 6 end atoms (3 on each side). Such a spatial separation in the transverse excitation between !TE and !TC resembles the density distribution of the surface and bulk plasmons of solids and thin lms: where the surface (end) plasmons are lower in energy than the bulk (central) plasmon. The induced charge oscillations are localized at the surfaces (the interior region) of the solids. The existence and nature of two transverse plasmon resonances has in general two different origins: (i) the variation of electron potential on different atoms along the chains, which is mainly responsible for the energy splitting; and (ii) the bonding and interaction between electrons residing on different atoms. The latter would mainly determine the collectivity and strength of the end mode. In order to further differentiate these two effects, extensive calculations have been done for sodium chains with variable interatomic distances by elongating or con tracting the distances for up to 1 A. In the range of distances calculated, the strength of the end mode was found to change only slightly (by less than 10%). Calculations for other atomic chains like Ag and K have also been performed, which veried the general existence of the end and central modes. The strength of the end mode

TE

TC
0.01 0.002

Strength (1/eV)

(a)

20

(a)
10

Na8 Ag-Na8

-0.01

-0.002 0.01

(b)

0.01

0 2 3 4 5

(b)
-0.01 -0.01 0.01

Energy (eV)
0.012

(c)
0.003

(c)

0.01

-0.01

-0.01

-0.012

-0.003

FIG. 3 (color online). Fourier transform of the induced density for Na 3 (a), Na 8 (b), and Na 18 (c) at the two transverse plasmon resonances !TE (the middle column) and !TC (the right column). The black squares specify the position of the atoms. The small panels to the left show the ground state electron density.

FIG. 4 (color online). (a) The dipole strength of the Na8 atom chain (black curve) modied by attaching a single silver atom at one end, forming an Ag-Na8 chain (red curve). The end mode reduces in its intensity (b), and splits into a high-energy mode at the Ag side (c).

216602-3

PRL 98, 216602 (2007)

PHYSICAL REVIEW LETTERS

week ending 25 MAY 2007

was found to depend more sensitively on the atomic species, with about 6 electrons for Ag chains, and 2 for K. This trend in the strength of the end mode is in line with the bonding order (interaction) among the three elements, with Ag > Na > K, which is justiable both in the chains and in their solids. Stronger electronic coupling favors larger strength in the end mode. In the light of potential application of the chain plasmon resonances and their mediated processes, it is desirable to control such atomic plasmons in both energy and space, using, for example, atomic manipulation with an STM. To explore the tunability of these chain plasmon resonances, we calculated the plasmon modes by attaching a single atom at different positions of the chains. Figure 4(a) shows the dipole strength of the Na8 chain (black) with a single silver atom attached at one end, forming an Ag-Na8 (red) chain. The bond-length of Ag-Na was still 2.89 A. The end mode at 3.26 eV decreases in intensity, and splits into a high-energy mode at 4.5 eV, which localizes at the silver end of the Ag-Na8 chain (c). In the meantime, the central mode of Na8 at 3.75 eV increases in its intensity due to the fact that one more atom is added to the chain. Splitting of the end mode results from the broken symmetry in the chains and the coupling to the attached Ag atom. Figure 4 demonstrates the sensitivity and tunability of collective excitation of the chains by modifying the atomic environment of the chains. End electronic states have recently been observed in linear gold chains by scanning tunneling spectroscopy [29]. It is our belief that the collective plasmon resonances reported in this work should be observable [30,31] in future experiments. In addition, the ultimate chemical and geometric sensitivities of these plasmon resonances should nd unique and profound applications in atomic-scale engineering of optical spectroscopy and associated dynamical processes. The splitting between the end and central modes, for example, makes it possible to selectively induce charge excitations in different spaces, selective control of charge localization and transport, local eld enhancement and chemical reactions, and plasmon-medium interaction at the atomic scale. The authors acknowledge Peter Nordlander and Bengt Kasemo for discussions and comments. This work has been supported by the 100-Talent Program of Chinese Academy of Sciences, the ATOMICS and PhotoNano consortium of SSF, and the Institutional Grant Programme of STINT.

*To whom correspondence should be addressed. Electronic address: shiwu.gao@physics.gu.se [1] S. Nie and S. R. Emory, Science 275, 1102 (1997). [2] H. Xu, E. J. Bjerneld, M. Kall, and L. Borjesson, Phys. Rev. Lett. 83, 4357 (1999).

[3] E. Prodan, C. Radloff, N. J. Halas, and P. Nordlander, Science 302, 419 (2003). [4] C. Sonnichsen, T. Franzl, T. Wilk, G. von Plesson, J. Feldmann, O. Wilson, and P. Mulvaney, Phys. Rev. Lett. 88, 077402 (2002). [5] K. Li, M. I. Stockman, and D. J. Bergman, Phys. Rev. Lett. 91, 227402 (2003). [6] N. Fang, H. Lee, C. Sun, and X. Zhang, Science 308, 534 (2005). [7] P.-A. Hervieux and J.-Y. Bigot, Phys. Rev. Lett. 92, 197402 (2004). [8] A. T. Bell, Science 299, 1688 (2003). [9] L. R. Hirsch, R. J. Stafford, J. A. Bankson, S. R. Sershen, B. Rivera, R. E. Price, J. D. Hazle, N. J. Halas, and J. L. West, Proc. Natl. Acad. Sci. U.S.A. 100, 13549 (2003). [10] W. L. Barnes, A. Dereux, and T. W. Ebbesen, Nature (London) 424, 824 (2003). [11] J. B. Pendry, L. Martin-Moreno, and F. J. Garcia-Vidal, Science 305, 847 (2004). [12] K. L. Kelly, E. Coronado, L. L. Zhao, and G. C. Schatz, J. Phys. Chem. B 107, 668 (2003). [13] P. Hanarp, M. Kall, and D. S. Sutherland, J. Phys. Chem. B 107, 5768 (2003). [14] U. Kreibig and M. Vollmer, Optical Properties of Metal Clusters (Springer, Berlin, 1995). [15] N. Nilius, T. M. Wallis, and W. Ho, Science 297, 1853 (2002). [16] G. V. Nazin, X. H. Qiu, and W. Ho, Phys. Rev. Lett. 90, 216110 (2003). [17] M. F. Crommie, C. P. Lutz, and D. M. Eigler, Science 262, 218 (1993). [18] G. V. Nazin, X. H. Qiu, and W. Ho, Science 302, 77 (2003). [19] R. Huber, F. Tauser, A. Brodschelm, M. Bichler, G. Abstreiter, and A. Leitenstorfer, Nature (London) 414, 286 (2001). [20] G. F. Bertsch, Oscillations in Finite Quantum Systems (Cambridge University Press, Cambridge, England, 1994). [21] M. A. L. Marques, A. Castro, G. F. Bertsch, and A. Rubio, Comput. Phys. Commun. 151, 60 (2003). [22] K. Yabana and G. F. Bertsch, Phys. Rev. B 54, 4484 (1996). [23] C. Jamorski, M. E. Casida, and D. R. Salahub, J. Chem. Phys. 104, 5134 (1996). [24] I. Vasiliev, S. Ogut, and J. R. Chelikowsky, Phys. Rev. Lett. 82, 1919 (1999). [25] S. Kummel, K. Andrae, and P.-G. Reinhard, Appl. Phys. B 73, 293 (2001). [26] S. W. Gao and Z. Yuan, Phys. Rev. B 72, 121406 (2005). [27] S. A. Maier, P. G. Kik, and H. A. Atwater, Appl. Phys. Lett. 81, 1714 (2002). [28] Q.-H. Wei, K.-H. Su, S. Durant, and X. Zhang, Nano Lett. 4, 1067 (2004). [29] J. N. Crain and D. T. Pierce, Science 307, 703 (2005). [30] T. Nagao, S. Yaginuma, T. Inaoka, and T. Sakurai, Phys. Rev. Lett. 97, 116802 (2006). [31] P. Segovia, D. Purdle, M. Hengsberger, and Y. Baer, Nature (London) 402, 504 (1999).

216602-4