J . Phys. Chem. 1990, 94, 4655-4660
4655
Pressure and Resonance Effects in Scanning Tunneling Microscopy of Molecular Adsorbates S. M. Lindsay,*,t 0. F. Sankey,t Y. Li,t C. Herbst,* Department of Physics and Department of Chemistry. Arizona State University, Tempe, Arizona 85287
and A. Rupprecht Arrhenius Laboratory, University of Stockholm, S - 106, Stockholm, Sweden (Received: December 4, 1989; In Final Form: January 22, 1990)
We have investigated an empirical tight-binding model of a tunnel junction which contains a molecule in the gap. Tunnel conduction is enhanced if the molecule has a molecular orbital (MO) with an eigenenergy close to the Fermi energy ( E F ) of the metal. A special case arises when the gap contains an insulator in which a molecule (with a resonant MO) is located equidistant from the metal electrodes. The transmission through this double barrier can approach unity, independent of the thickness ofthe gap. A resonant MO enhances tunnel conduction even when the geometry is far from optimum. There have been many reports of molecular adsorbates which appear to act as conductors when imaged with the scanning tunneling microscope (STM). Images of DNA constitute the largest group, but electrochemical studies of nucleic acids indicate that MO's are about an electronvolt from E F for typical metals, so that resonant tunneling alone will not account for the STM images. However, the pressure in the STM tunnel gap may approach a few gigapascals. The ultraviolet absorption of DNA is red-shifted by -0.5 eV at these pressures, indicating that increased intermolecular interaction has broadened and shifted MO's by at least this much. The STM may operate by adjusting the gap transducer so that an MO is moved by the amount required to establish the set-point current through the resonant tunneling enhancementeffect. The image contrast is a complicated function of the mechanical and electronic properties of the tip, substrate, and adsorbate molecules, and usually cannot be interpreted in terms of height.
Introduction There have been many reports of scanning tunneling microscope (STM) images of organic molecules adsorbed on conducting Quite a few of these images are of DNA molec u l e ~ ' ~while - ~ ~some are of even thicker complexes such as DNA-RecA,26 lipid^,^^-^^ protein^,^^-^^ phage,37 bacterial and tobacco mosaic virus.39 These materials do not conduct electrons and the images cannot be accounted for by tunneling of electrons through an insulating gap with a width equal to that of the complexes.I6 Furthermore, a number of high-resolution images show that the STM does not follow the topography of the molecule.'*J9J2 We focus our attention on DNA because the many published STM studies build a convincing case for an unexpected conduction mechanism, although we realize that a simpler molecular system would be better suited to testing the proposals we make in this paper. Images of DNA and RNA polymers show high contrast variations that are specific to the ~ e q u e n c e , ' ~ J ~ suggesting that the STM may be rather sensitive to the electronic states in the tunnel gap. There is a confusing difference in reports of contrast and values for local barrier height in published studies of DNA.'"' Some of these variations may come from the uncontrolled nature of an interface prepared in air, a difficulty we circumvent by working at an interface under potentiostatic control in an electrochemistry ce11.19-41Nonetheless, we still find that contrast in DNA images can be either negative (the molecule appears lower than the surrounding substrate)l5J6or positive (the molecule appears higher than the substrate).I7J8 These variations appear to correlate with the condition of the STM tip; a blunt tip usually produces negative images of monolayers, while a sharp tip produces positive images. This observation suggests that the pressure in the tunnel gap is important.I9 The negative images can be accounted for as a consequence of the elastic response of the tipsample-substrate system as the material in the tunnel gap is squeezed by the STM However, positive images predominate'7-18.20-22*24-25 and include images of rather thick complexes.I9 Here we propose a theory for these remarkable effects, and offer some experimental data in support of the mechanism we propose.
'Department of Physics, Arizona State University.
*Department of Chemistry, Arizona State University.
0022-3654/90/2094-4655$02.50/0
We study an empirical tight-binding model of a tunnel gap containing a molecule, finding an unexpected enhancement of
(1) Hansma, P. K.; Eilings, V. B.; Marti, 0.;Bracker, 0. Science 1989, 242, 209. (2) Sleator, T.; Tycko, R. Phys. Rev. Lett. 1988, 60, 324. (3) Ohanti, H.; Wilson, R. J.; Mate, C. M. Phys. Rev. Lett. 1988.60, 2398. (4) Foster, J. S.; Frommer, J. E.; Arnett, P. C. Nature 1988, 331, 324.
(5) Hubacek, J. S.; Brokenbrough, R. T.; Gammie, G.; Skala, S. L.; Lyding, J. W.; Latten, J. L.; Shapley, J. R. J . Microsc. 1988, 152, 221. (6) Spang, J. K.; Mizes, H. A.; LaComb, L. J.; Dovek, M. M.; Frommer, J. E.; Foster, J. S. Nature 1989, 338, 137. (7) Feng, L.; Hu, C. Z.; Andrade, J. D. J . Microsc. 1988, 152, 811. (8) Reneker, D. F.; Howell, B. F. J . Vac. Sci. Technol. 1988, A6, 553. (9) Albrecht, T. R.; Dovek, M. M.; Lang, P.; Quate, C. F.; Kuan, S. W. J.; Frank, C. W.; Pease, R. F. W. J. Appl. Phys. 1988, 64, 1178. (IO) Rabe, J. P.; Sano, M.; Batchelder, D.; Kalatchev, A. A. J . Microsc. 1988, 152, 573. (1 1) Mizutani, W.; Shigeno, M.; Saito, K.; Morito, N.; Yoshioka, T.; Ono, M.; Kajimura, K. J. Microsc. 1988, 152, 547. (12) Binnig, G.; Rohrer, H. In Trends in Physics; Janta, J., Pantoflicek, J., Eds.; European Physical Society: The Hague, 1984; pp 38-46. (13) Travaglini, G.;Rohrer, Amrein, M.; Gross, H. Surf Sci. 1987, 181, 380. (14) Lindsay, S. M.; Barris, B. J. Vac. Sci. Technol. 1988, Ab, 544. (15) Barris, B.; Knipping, U.; Lindsay, S. M.; Nagahara, L.; Thundat, T. Biopolymers 1988, 27, 1691. (16) Lindsay, S. M.; Thundat, T.; Nagahara, L. J . Microsc. 1988, 152, 213. (17) Lindsay, S. M.; Thundat, T.; Nagahara, L.; Knipping, U.; Rill, R. L. Science 1989, 244, 1063. (18) Lindsay, S. M.; Nagahara, L. A.; Thundat, T.; Oden, P. J . Biomol. Strucr. Dyn. 1989, 7, 289. (19) Lindsay, S . M. E M S A Bull. 1989, 19:2, 60. (20) Beebe, T. P.; Wilson, T. E.; Ogletree, D. F.; Kape, J. E.; Balhorn, R.; Salmeron, M. B.; Seikhaus, W.J. Science 1989, 243, 370. (21) Lee, G.; Arscott, P. G.;Bloomfield, V. A,; Evans, D. F. Science 1989, 244, 475. (22) Arscott, P. G.;Lee, G.;Bloomfield, V. A.; Evans, D. F. Nature 1989, 339, 484. (23) Keller, D.; Bustamante, C.; Keller, R. Proc. Natl. Acad. Sci. U.S.A. 1989,86, 5356. (24) Cricenti, A.; Selci, S.;Felici, A. C.; Generosi, R.; Gori, E.; Djaczenko, W.; Chiarotti, G.Science 1989, 244, 1226. (25) Dunlap, D. D.; Bustamante, C. Nature 1989, 342, 204. (26) Amrein, M.; Durr, R.; Stasiak, A,; Gross, H.; Travaglini, G. Science 1989, 243, 1708.
0 1990 American Chemical Society
4656 The Journal of Physical Chemistry, Vol. 94, No. 11, 1990
METAL
GAP
METAL
(A)
METAL
GAP
MOLECULE
GAP
METAL
(B)
Lindsay et al. barrier models which assume the existence of a “conducting state” in the gap. has dealt with the problem of single atom imaging using the Bardeen formulation of the problem.46 We want to develop the theory to include the case of rather stronger interactions because it appears that the tip pushes into an adorba ate.'^,^^ We also want to deal with the problem of extended molecular orbitals. This is quite straightforward, at least in one dimension, using the empirical tight-binding (ETB) appr~ach.~’ We show an ETB model of a simple metal-vacuum-metal tunnel junction in Figure 1A (-t is the hopping matrix element, to is the on-site energy, a is the lattice constant, and -T is the hopping integral across the gap). In the weak coupling limit (small T ) we recover the result of a calculation carried out with the Bardeen approach46as follows: With Bloch wave functions of the form exp(ikla), where 1 is a site index, and energies, E ( k ) , that satisfy the Schriidinger equation for the metal, the Schriidinger equation, ( H - E ) / $ ) = 0, for an electron incident on the barrier from the left, yields the following matrix equations
Figure I . Tight-binding model of (A) a tunnel gap and (B) a molecule in a tunnel gap. The metal (lattice constant a ) is composed of atoms of on-site energy co connected by hopping matrix elements -1. The gaps are represented by - T , - T ~ ,and - T ~ . The molecule is composed of atoms of on-site energies cI...cN connected by hopping matrix elements -7l...-T,+l. The tip pushes into insulating molecules, so the gaps are not empty; they represent parts of the system with eigenenergies far from the Fermi level.
conduction if the molecule is located near the middle of the tunnel gap and has an orbital in resonance with the metallic Fermi level. Electrochemical data suggest that the required energy resonance is unlikely to occur in the unperturbed molecule, but we have taken optical absorption spectra of DNA over the range of pressures that might be encountered in the STM tunnel gap, finding that spectral features change by -0.5 eV. MO’s must move by at least this much relative to the Fermi level. We conclude that this pressure shift (which is adjustable by the STM servo) may bring electronic levels into resonance. The contrast will then reflect the changes in pressure that are required to bring various states into resonance as the tip is scanned over the aggregate on the substrate. It will not normally be possible to interpret the z transducer motion in terms of the height of the molecules in the gap.
An Empirical Tight-Binding Theory of Tunneling via a Molecule Garcia and Garcia have proposed that disorder in a molecular adsorbate modifies electronic properties so as to permit conduct i o ~ ~Keller . ~ ~ et al.43and Salmeron et al.” have used simple (27)Fuchs. H.; Schrepp, W.; Rohrer, H. Surf. Sci. 1987,181,391. (28)Smith, D. P.E.; Bryant, A.; Quate, C. F.; Rabe, J. P.; Gerber, Ch.; Swalen, J. D. Proc. N a f l . Acad. Sci. U.S.A. 1987,84, 969. (29)Horber, J . K. H.; Lang, L. A.; Hansch. T. W.; Heckl, W. M.; Mowald, H. Chem. Phys. Left. 1988,145, 15 1. (30)Fuchs. H.Phys. Scr. 1988,38, 264. (31)Engel, A.; Stemmer, A.; Aebi, U. Proceedings ofthe 47th Meeting of the EMSA; Bailey, G. W., Ed.; San Francisco Press: San Francisco, 1989; p 12. (32)Wu, X. L.;Lieber, C. M. J. Phys. Chem. 1988,92,5556. (33)Voelker, M. A.; Hameroff, S . R.; He, J. D.; Dereniak, E. L.; McCuskey, R. S.; Schneiker, C . W.; Chvapil, T. A,; Bell, L. B. J. Microsc. 1988,152,551. (34)Feng, L.; Hu, C. Z.; Andrade, J. D. J. Colloid Interface Sci. 1988, 126,650. (35)Welland, M. F.;Miles, M. J.; Lambert, N.; Morris, V. J.; Coombs, J. H.; Pethica, J. B. Int. J. Bioi. Macromol. 1989,11, 29. (36)Edstrom, R. D.; Meinke, M. H.; Yang, X.;Evans, D. F. Biochemistry 1989,28,4939. (37)Baro, A. M.; Miranda, R.; Alaman, J.; Garcia, N.; Binnig, G.; Rohrer, H.; Gerber, Ch.; Carracosa, J. L. Nature 1985,315, 253. (38)Dahn, D. C.;Watanabe, M. 0.; Blackford, B. L.; Jerico, M. H.; Beveridge, T. J. J. Var. Sci. Technol. 1988, A6. 548. (39)Mantovani, J. G.; Allison, D. P.; Warmack, R. J.; Ferrell, T. L.; Ford, J. R.; Manos, R. E.; Thompson, J. R.; Reddick, B. B.; Jacobson, K. B. J. Microsc., submitted for publication. (40)Contrasts and barrier heights in DNA images are surveyed in the discussion by: Stemmer. A.; Hefti, A.; Aebi, U.; Engel, A. Ultramicroscopy 1989,30, 263. (41)Thundat, T.;Nagahara, L. A.; Oden, P.; Lindsay, S. M. J. Var. Sci. Technol. 1990. A8. 645.
where a, and a, are the amplitudes of transmitted and reflected waves at the barrier. For small T the solution is la,/2= 4 ( ~ / r )cos2 ~ ka With weak overlap we have the usual result for r2
-
exp[ -2( %)lild]
(2) T
(3)
where 4 is the work function, m the electron mass, and d the tunnel gap. We can apply this formalism to the general problem of a molecule in the gap (with no restriction on T ) using the model illustrated in Figure 1B. We have solved this problem (Lindsay, S. M.; Sankey, 0. F., manuscript in preparation) but can illustrate the important features more simply with the case of a two-atom molecule made from atoms of on-site energies c, and c2, coupled by the matrix element -7,. The gaps at the left and right are represented by the Hamiltonian matrix elements T~ and 71. These are not empty gaps, because the tip will push into an insulator on the surface. Rather, they represent those parts of the system with states far removed from the Fermi level. The Schrodinger equation now yields the following matrix equations relating the reflection and transmission amplitudes to the wave function amplitudes, b, and b2, on the atoms of the molecule
w h e r e A = E ( k ) - to + t exp(ika). This can be inverted exactly, yielding a transmission coefficient
(5)
(42)Garcia, R.; Garcia, N . Proc. Int. School Solid State Phys., Erice, in press. (43) Keller, D.; Bustamante, C.; Keller, R. Proc. Narl. Acad. Sci. U.S.A. 1989,86,5356. (44)Salmeron, M.; Beebe, T. P.; Odriozola, J.; Wilson, T.; Ogletree, D. F.; Siekhaus, W. J. Vac. Sci. Technol., in press. (45)Lang, N. D. Comments Condens. Mafter. Phys. 1989,14, 253. (46)Bardeen, J. Phys. Rev. Lett. 1961,6, 57. (47)Ziman, J. M. Principles ofthe Theory of Solids, 2nd ed.;Cambridge University Press: Cambridge, 1972;pp 91-96.
STM of Molecular Adsorbates
The Journal of Physical Chemistry, Vol. 94, No. 11, 1990 4651
where Dois the determinant of 4 X 4 matrix in eq 4. The 2 X 2 block in the center of this matrix is simply E - HM,where H , is the Hamiltonian of the unperturbed molecule which has eigenvalues E? and E?. The solution is simplified if TL and rRare small, so that perturbation theory can be used to second order to calculate new eigenvalues for the molecule. The result of such a calculation is
7R21Ay2(R)12+ rL21Ay2(L)12
(Er2- E ( k ) where Ar2(L) and Ay2(R) are amplitudes of the unperturbed electronic molecular wave functions on the left (L) and right (R) atoms of the molecular levels 1 and 2. This expression has maxima near E ( k ) = EY2. The resonances are nearly Lorentzian in form with a width, I‘, given by
r
= I-(t sin ~ u ) ( T ~ ~ ~ A+ ~rR21Ay2(R)12)/t21 ~(L)I~ (7)
This shows that the tunneling is enhanced strongly when a molecular orbital (MO) is in resonance with the energy of an electron in the metal. The width of the resonance depends upon the relative strength of the coupling of the molecule to the metals ( q / t , T R / t ) . The magnitude of the enhancement is rather surprising. This may be seen by considering eq 6 near resonances, for example, when E ( k ) = E?. One term in the product in the denominator of six becomes E? - EY (this difference being large compared to rLor 7R) while the other term consists of the wave-amplitude-weighted sum of the squares of rLand rR. We can rewrite E? - E? in terms of the atomic energies and the molecular hopping integral to obtain the following expression for transmission at a resonance: btI2€(k).€,M
-2
0 Energy
Figure 2. Plots of transmission versus electron energy for (A) a fouratom molecule with twostates in the conduction band and rL = zR = 0.11, (B) a five-atom molecule with only one state in band with its wave amplitude largely confined to the center atom, and (C) as in (A) but with T L = 4rR = 0.1f. The broken lines show the transmission through simple barriers of widths equal to the sum of the gaps each side of the molecule for (A) and (B) (dashed line) and for (C) (dotted line).
I ‘PI
1
\ \
REDUCED STATE FERM
EFERENCE
METAL
-
(48) Bohm, D. Quanrutn Theory; Prentice-Hall; Englewood Cliffs, NJ, 1951; pp 283-295. (49) Esaki, L. lEEE J . Quanr. Electronics 1986, QE-22, 161 I . (50) Shewchuk, P.; Chapin, P. C.; Coleman, P. D.; Kopp, W.; Fischer R.; M o r k g , H. Appl. Phys. Let?. 1985, 46, 508. (51).Kane, E . 0. In Tunneling Phenomenu in Solids; Burstein. E.. Lundqvist, S.,Eds.; Plenum: New York, 1969; p I . (52) Duke, C. B.; Alferieff, M. E. J . Chem. Phys. 1967, 46, 923
LEVEL
SUBSTRATE
u
We see that if 7L equals T R and #(L) = A?(R), the transmission becomes independent of rLor rR. In other words, as long as quantum coherence is maintained, a resonant state (of appropriate symmetry) in the middle of a tunnel gap can mediate tunneling over arbitrarily large gaps. This result can also be obtained by considering the problem of plane wave tunneling through two identical barriers which contain a resonant states4* The consequent transport of electrons through large insulating regions is the basis of quantum-well electronic devices,49 and fabricated structures of 100 A dimension manifest these effects at room temperature.% Results similar to eq 8 have been obtained by Kanesl in an analysis of an asymmetric double well, and by Duke and Alferieff in a study of resonant enhancement of field emission.s2 Equation 8 suggests that the peak transmission could approach unity, but many factors can lower it. Nonetheless, the effect can be so big that a significant enhancement occurs for a range of parameter values. We now illustrate this point with some models which we have solved numerically. Our first model is a four-atom molecule for which z1 = 0, c2 = -4, e3 = 0, and t4 = -4 eV. We choose r 1 = 72 = i 3= t = 1 eV, leading to a conduction bandwidth of f2 eV around eo (taken as 0 eV) and we set r L = f R = 0.1 eV. There are two MO’s in the band near 0 eV and two below the band near -4 eV. We plot
/
/ EMPTY STATE
=
(8)
VACUUM
I
1 1
L‘..,”
FILLED STATE OXIDIZED STATE
” -
MOLECULE METAL GAP GAP
ELECTROCHEMISTRY
Illustrating the relationship between states of the unperturbed molecule, the metal, and the electrochemically reacted molecule. The levels are discussed in the text (they are shown here with no external bias app:ied to the tunnel junction). The -gaps” represent part of the system with eigenstates far from EF. Figure 3.
the transmission coefficient (la,I2)calculated as a function of incident electron energy in Figure 2 (curve A). The transmission approaches unity for the highest MO. A measure of the enhancement of tunnel conduction is given by comparing the transmission to that expected across a gap equal to the sum of the left and right gaps alone (see eq 2). For this case, lat12 = 47L2?R2/t2= 4 X IO4 (dashed line in Figure 2). The finite width and overlap of the two states leads to enhancement over an energy range of 1 eV. Over this range, the molecule enhances tunnel current to more than could be obtained by removing it and closing the gap by the length of the molecule. Our second model is a five-atom molecule for which el = t2 = t4 = tS = -4 eV, t3 = -1 eV, and the other parameters are as above. In this case, only one state lies in the band. Even though the resonant MO is largely confined to the center atom, a strong enhancement of the tunneling is still found to occur over a small energy range (curve B in Figure 2). Finally, we consider a model similar to that of Figure 2A, but with unequal barriers ( 7 L = 0.1, rR = 0.025),corresponding to a difference of 1.4 in the exponent of eq 3, for example. The transmission (curve C in Figure 2) is reduced by the square of the reduction in rR, but because the transmission through the equivalent single gap is reduced by this factor, there is still a considerable relative enhancement compared to the result 4 7 L 2 r R 2 / t 2 without the molecule (shown as the dotted line). The range of relative enhancement is reduced only because the line widths sharpen somewhat as r R 0.
-
-
4658
The Journal of Physical Chemistry, Vol. 94, No. 11, 1990
Location of Molecular Orbitals in DNA A simplified diagram of the location of molecular and metallic states is shown in Figure 3. The spaces (TL,TR) represent those parts of the gap that contain atoms with levels far from EF(e.g., oxide layers on the metals, the sugar-phosphate backbone in DNA, etc.). Typical tunnel currents are -IOio electrons s-l with each transition occurring in s so a MO that mediates tunneling does not alter its charge state.53 We require a measure of the energies A&, AE, that separate the Fermi level from the nearest states of the molecule in the high electric field at the interface. The states of the neutral molecule that accept electrons from an electrode (states that are reduced at a potential AERED) or donate electrons to the electrode (states that are oxidized at a potential LEoX)are closely related. They differ from the states relevant to the tunnel problem by the amount of the environmental relaxation caused by charging and by the intrinsic difference in electronic structure between the neutral and ionized molecule (AERELAX) but they are similar in that the molecule is located on a metal surface in an electric field of V / A in both cases. The organic bases are the main electroactive groups in DNA. Adenine and cytosine are reduced on mercury54and graphite55 at about -1.5 V vs the saturated calomel electrode (SCE). Oxidation occurs on graphite at 1.2 V (vs SCE) for adenine, and 0.9 V (vs SCE) for guanine.56 The potential for zero charge vs SCE (A&) is typically -4.5 V,57but AERELAY is unknown, although it could be several electron volt^.^^ Even if a state for one base was near an accidental resonance, it is clear that some of the other bases would be separated from EF (A&, or AE,) by at least an electronvolt.
Lindsay et al. 3
h
0 0
-
-
Pressure and Molecular Orbitals in the STM Gap It is clear that the STM tip pushes into a surface covered with an insulating a d s ~ r b a t e . ~Contact ~ J ~ ~forces ~ ~ ~of ~ N have been measured even in a clean, ultra-high-vacuum STM.6i In these experiments, pressure in the gap (- 1 GPa) was estimated by taking the resolution as a measure of the contact area. However, as Mamin et al. have pointed out,60elastic interactions probably occur over a much larger area than that of the tunneling asperity. The tunnel junction, and the associated substrate deformation, may be imaged directly by using an STM located in a transmission electron microscope.62 Contact forces and areas vary over a surface and vary with the tip geometry (as does STM c ~ n t r a s t ' ~ Estimates ). of junction pressure range from 0.1 to 10 GPa (Spence, J. C. H., personal communication). When the material in the STM gap is in the form of a dense molecular aggregate, its macroscopic properties at high pressure may reflect the changes that occur in the STM. In this case, gigapascal pressure may be adequate to bring about the required resonances, and we have used optical methods as an indirect probe of the effects of pressure on the MO's.
s.2
-: 4
Y
a
1
300
400
350
4
h(nm) Figure 4. Solid lines are the edges of the UV absorbance peak for solid, hydrated DNA at various pressures. The absorbance is plotted between its value in the relatively flat part of the spectrum between 450 and 500 nm (A450) and 3 times A450. Curve A is a two-Gaussian fit to the published data at ambient pressure, plotted in the same way, while curve B is calculated for the same parameters, but with the peak energy lowered and its width increased by the amounts reported by Okamoto and Drik a ~ n e rfor ~ ~adenosine at 10 GPa.
k
71.0
4
-
(53) We are considering transitions that occur at the Fermi level, so the
occupancy of a state is not relevant. Although we describe the energy matching as "resonance", the lifetime of the state in the gap should not affect these time scales a great deal: this may be seen by considering the process in a time-dependent formalism where the quantities h / t or h / r represent uhopping" times. For an estimate of the transit time for tunneling, see: Hartman, T. E. J. Appl. Phys. 1962,33, 3427. (54) Valenta, P.; Niirnberg, H. W.; Klahre, P. Bioelecirochem. Bioenerg. 1974,I , 487. (55) ThCvenot, D. J . Eleciroanal. Chem. Interfacial Elertrochem. 1913, 46. 89. (56) Brabec, V. Bioelerrrochem. Bioenerg. 1983,1 1 , 245. (57) Bard, A. J.; Faulkner, L. R. Electrochemical Methods, Fundamentals and Applicarions; Plenum: New York, 1980. (58) If the oxidized and reduced states correspond to an allowed optical transition (-5 eV) then the difference between the relaxation in the upper and lower states is -2.7 eV for adenine. For a discussion of the relationship
between these states see: Drickamer, H. G.; Frank, C. W.; Schlichter, C. P. Proc. Natl. Acad. Sci. U.S.A. 1912,69, 993. (59) Coombs, J. H.; Pethica, J. 8. IBM J . Res. Deu. 1986,30, 455. (60) Mamin, H. J.; Ganz, E.; Abraham, D. W.; Thompson, R. E.; Clarke, J. Phys. Rev. 1986,834. 9015. (61) Durig. U.; Gimzewski, J. K.; Pohl, D. W. Phys. Reu. Lett. 1986,57,
2403. (62) Kuwabara, M.; Lo, W.; Spence, J. C. H. J . Vac. Sci. Technol. 1989, A 7 , 2145
3.7
0
2
4
6 P (GPa)
8
1
0
.,
Figure 5. Plot of the energy (Eu)at which the absorbance becomes three times A450for hydrated D N A (0, pressure increasing; 0 , pressure decreasing) and for dry D N A (0, pressure increasing; pressure decreasing). The circled letters are calculated from Okamoto and Drickamer's data67 for (A) adenosine, (C) cytidine, and (T) thymidine. The lower curve shows the variation of normalized volume (V(P))which fitted the measured peak shifts for these nucleosides using eq 9. The upper curve is the corresponding variation in dipole energy calculated with eq 10.
Pressure Dependence of the UV Absorbance in DNA We have investigated the absorbance of DNA films in a diamond anvil cell (DAC).63 The optical quality of our diamonds was inadequate for polarization measurements, and they did not transmit wavelengths below 300 nm. We could only study the long-wavelength tail of the 258-nm absorbance peak.@ The DAC was illuminated with a xenon arc lamp, using a nearly parallel cf/60) beam in order to minimize aberrations caused by diamond dispersion. These aberrations cause run-to-run variations of -20% if the cell is moved (for pressure adjustment, for example). The transmitted light was analyzed with a SPEX 1401 monochromator, and normalized by using reference spectra obtained with a 4:l methanol/ethanol mixture in the DAC. We used wet-spun films of calf thymus DNA.65 These were either equilibrated at 88% relative humidity (rh) prior to loading (resulting in a water content of 1 g per gram of DNA66) or they were loaded dry. The wet Jayaraman, A. Rev. Mod. Phys. 1983,55,65. Tinoco, 1. J . Am. Chem. Soc. 1960,82, 4785. Rupprecht, A.; Forslind, B. Biochim. Biophys. Acta 1970,204, 304. Lindsay, S. M.; Lee, S. A.; Powell, J. W.; Weidlich, T.; DeMarco, C.; Lewen, G.;Tao, N. J.; Rupprecht, A. Biopolymers 1988, 27, 1015. (63) (64) (65) (66)
STM of Molecular Adsorbates films were slightly swollen and filled the gasket, while the dry (20 r m thick) films had to be folded several times. The wet films maintained hydrostaticity to 12 GPa, whereas the dry films developed pressure gradients of 2 GPa at 6 GPa (as determined by ruby chip fluore~cence~~). The films distort plastically and elastically, but these changes appear to be compensated by densification, because the data show little hysteresis. We have scaled the spectra to cover a range from the absorbance in the nearly flat region from 450-500 nm (A,,!) to 3 times A450, in order to compensate for the run-to-run variations. Some typical spectra for the wet films are shown in Figure 4. Their interpretation is complicated because they reflect both peaks shifts and width increases. However, Okamoto and D r i ~ k a m e rhave ~ ~ made a careful study of the absorption spectra of adenosine, cytidine, and thymidine over a similar pressure range. We compare their results with ours as follows: We have made a two-Gaussian fit to the DNA spectrum published by Voet et This fit is then used to match our 0 GPa spectrum by adjusting its overall height (curve A in Figure 4). The corresponding absorbance at 258 nm (as calculated from the fit) is in line with estimates based on the known film thickness. We then adjust the 258-nm wavelength maximum and its width by the amounts appropriate to the pressure shifts for the nucleosides as reported by Okamoto and D r i ~ k a m e r .All ~~ other parameters are left unchanged. Curve B in Figure 4 is calculated by using data for adenosine at I O GPa (a summary of fits at several pressures is given in Figure 5). The agreement with our spectra indicates that, to within the limited accuracy of our data, the shape of the absorbance spectra of DNA under pressure may be calculated as a simple sum of the spectra for the nucleosides at the same pressure. This result holds for solid B-DNA at ambient pressure69 and, to a lesser extent, for solid Z-DNA at ambient pressure.70 We summarize our results in Figure 5 where we plot the energy, E3A,at which the absorbance reaches 3 times A4%versus pressure. The scatter on the data points reflects the variation in spectra from run-to-run (open points are increasing pressure and solid points are decreasing pressure). The data for dry DNA (squares) do not differ dramatically from those for wet DNA (circles). In the wet DNA, the bases are stacked in the A geometry. In the dry DNA, they are disordered, so the broad agreement suggests that solid-state effects which depend on order (such as excitons7I) do not dominate the spectra. We have used the data of Okamoto and Drickamer to simulate spectra for the nucleosides, so that we could calculate an E3A equivalent to the quantity we have measured for DNA. The results are shown as the circled letters in Figure 5. The red shift of the edge of the DNA absorption peak is mainly accounted for by the red shift and broadening of the adenosine absorption. The main point here is that, whatever the details, the red shift implies that a state (or states) must have moved toward the Fermi level by an amount comparable to (or more than) the observed shift (of -0.5 eV). We would like a quantitative estimate of the shift in eigenenergies. Although the shape of the 258-nm absorbance peak does not appear to change as the nucleosides are incorporated into a polymer (the strength does&), the spectrum is quite complex, being composed of several transitions in each base, each of which is perturbed differently on incorporation into the polymer; this process has been analyzed quantitatively for a Z-DNA crystal.70 Therefore, the implications of the optical data for the pressure dependence of the M O s cannot be worked out without a detailed model. Nonetheless, we can make some simple estimates: If a particular state possesses a dipole moment, then, in the simplest case of an isotropic geometry like that of solute dipole embedded in a solvent, the dielectric theory72gives for the optical transition (67) Okamoto, B. Y.; Drickamer, H. G.J . Chem. Phys. 1974, 61, 2878. (68) Voet, D.; Gratzer, W. 9.; Cox, R. A,; Doty, P. Biopolymers 1963, I , 193. (69) Priore, D. R. C.; Allen, F. S.Biopolymers 1979, 18, 1809. (70) Ho,P. S.;Zhou, G.; Clark, L. B. Biopolymers, submitted for publication. (71) Kasha, M. Reo. Mod. Phys. 1959, 31, 162. (72) McCrae, E. G. J . Chem. Phys. 1957, 61, 562.
The Journal of Physical Chemistry, Vol. 94, No. 11, 1990 4659 energy
where n(P) is the optical refractive index at pressure P, AP"(0) is the transition energy in the gas phase, p(P) is the density, and Cis a parameter which depends on quantities such as the oscillator strength and molecular dimensions. If we assume that C is a constant, we can obtain good fits (not shown here) to the data of Okamoto and D r i ~ k a m e for r ~ ~the wavelength of the peak for the three nucleosides as a function of pressure by using eq 9. To do this, we have followed their method in taking the average of values for the moduli and pressure derivatives of the moduli for 15 organic 1nateriaIs.7~ These are used in a Mumaghan equation7, to calculate density as a function of pressure. We then use the Lorentz-Lorenz relation to calculate n(P) given a value of 1.55 for DNA at ambient pressure.'s To the extent that this procedure works in describing the transition energy (a good fit is difficult to avoid given the number of fitting parameters), the calculated optical properties at pressure can be used to estimate the changes in absolute dipole energy of a state by using the O n ~ a g e formula r~~ E(P) = E(1) 2n2 + 1 2n2(P) 1
+
where E(1) is the dipole energy at one atmosphere and n the refractive index at 1 atm. In a solid-state picture, dipole-dipole interactions lead to an exciton-like band7' with a width that increases with decreasing base-base separation (rij)as riY3. In the simplest case, the broadening will follow the specific volume. We illustrate the scale of both these effects (for the P-Vdata that we used to fit Okamoto and Drickamer's experiment^^^) with the solid lines in Figure 5. The upper curve is calculated by using eq 10, and the lower curve plots the normalized molar volume. Thus a dipolar state with an initial energy of 5 eV and width of 0.5 eV might shift 1 eV and broaden 0.1 eV as the result of a 10 GPa pressure change. Conclusions A resonant MO in the tunnel gap will serve both to conduct
tunnel current and to enhance tunnel conduction over its value in the equivalent single gap. Resonant states are not normally available, but a MO might be moved into resonance near the Fermi level by the pressure generated as the tip is advanced. Enhancement is easier to obtain if the molecule has several MO's within a small (- 1 eV) range, as is the case for the DNA bases.70 Contrast would then arise from the variation in the required pressure as the tunneling part of the junction scans over various MO's in the aggregate at the electrode surface. The magnitude of the contrast variations will depend in detail on the mechanical properties of the tip and substrate, because they will dictate the z motion required for a given change in p r e s ~ u r e . ' ~This - ~ ~ z- ~ motion could be much larger than real topographical variations in the molecule. We have observed large alternations in apparent contrast over a DNA polymer with an alternating sequence,'* while Spong et al. see large enhancements over phenol rings! The many reports of STM images of DNA12-24have motivated this work, but our ideas need to be tested on simpler molecules to see if image interpretation can be simplified: for example, if the pressure dependence of the various MO's were broadly similar, then the relative apparent height over various MO's would correlate with their oxidation or reduction potentials. In the event that pressure broadening were small enough, the current-voltage characteristics over a molecular aggregate might show structure characteristic of the molecules. In the more interesting case of a heterogeneous monolayer of molecules, the "pressure" broadening and shifting (73) Vaidya, S.N.; Kennedy, G. C. J . Chem. Phys. 1971, 55,987. (74) Okomoto, B. Y . ;Drikamer, H. G. J . Chem. Phys. 1974, 61, 2870. (75) Weidlich, T.; Lindsay, S. M.; Rupprecht, A. Biopolymers 1987, 26, 439. (76) Onsager, L. J . Am. Chem. SOC.1936, 58, 1486.
J . Phys. Chem. 1990, 94, 4660-4665
4660
would occur because of direct interaction with the tip and substrate (Q and 7R in our model). If the consequent broadening is not too large, it might be possible to obtain chemical information on a nanometer scale. Acknowledgment. We are grateful to George Wolf, Roland Hanson, and Tom Moore for the loan of equipment, and to John
Page, Tom Thundat, Larry Nagahara, Rick Oden, and Lin Oliver for useful discussions and advice. Roger Wartell encouraged this work by pointing out the contrast variations in STM images of DNA. Pui S. Ho and Leigh Clarke shared their work with us prior to its publication. We received financial support from the NSF (BBS8615653) the O N R (N00014-87-K-0478) and the office of the Vice President for Research at ASU.
Molecular Dynamics Simulation of a Deoxydinucleoside-Drug Intercalation Complex: dCpG/Proflavin S. Swaminathan, D. L. Beveridge,* Chemistry Department, Wesleyan University, Middletown. Connecticut 06457
and H. M. Berman Institute f o r Cancer Research, Fox Chase Cancer Center, 7701 Burholme Avenue, Philadelphia, Pennsylvania 19111 (Received: August 11, 1989; In Final Form: January 2, 1990)
A molecular dynamics (MD) study of the dCpG/proflavin crystal hydrate based on the GROMOS force field and simulation
protocol is reported. The objectives of the study are (a) to determine the extent to which the theoretical calculations account for the dynamical motion of the complex and the organization of the water network and (b) to investigate the nature of the intermolecular hydrogen bonding network and determine the extent of interconnections on a static and dynamical basis. The results show the GROMOS energy function to have a well-defined energy minimum with a root-mean-square deviation of 0.19 A for the complex and 0.38 A for the water. The hydrogen bond network in the energy-minimized water structure is not as fully connected as was indicated by the 3.2-A crystallographic contacts. MD calculations were performed on a system of two unit cells (eight asymmetric units) for 25 ps at a temperature of 300 K. The dynamical behavior of the complex was reasonable, even to the point of reproducing mixed sugar puckers. The water structure was examined in terms of a hydrogen bond density representation. Here, in a dynamical sense and symmetry averaged over asymmetric units, the water network was found to be fully hydrogen bonded and consistent with 3.2 8, contact representation. The contact representation is thus a good indication of the time-averaged hydrogen bond network.
Introduction
The crystal structure of 2:2 complex of the dinucleoside dCpG and the simple intercalator proflavin reported in a series of papers by Neidle, Beman, and Shieh’-*provided the first example of a nucleic acid system with highly structured hydration. The dCpG/proflavin crystal has subsequently proved to be of considerable interest from the point of view of both structural studies and theoretical calculations. The system provided considerable detail on the conformational flexibility of the nucleic acid at the intercalation site and dynamics of the intercalator. However, the nature of the water network proposed by crystallographers has been questioned3 with regard to whether the virtal bonds drawn at “contact” distances of 3.2 A accurately represent the underlying intermolecular hydrogen bond network or not. The dCpG/proflavin crystal, as a well-determined, intermediate-sized system, provides a prototypical case for evaluating the performance of molecular simulations on nucleic acids and a basis for investigating the sensitivity of results to the choice of force field, potential function truncations, and other characteristics and assumptions in the calculations as well as assumed system size. We describe herein further theoretical studies on the dCpG/ proflavin crystal hydrate based on both energy minimization (EM) calculations and molecular dynamics (MD) simulations and investigate further the issues raised above. A comparison of the 0 K EM structure with the instantaneous (snapshot) and diffu( I ) Shieh, H. S.;Berman, H. M.; Dabrow, M.; Neidle, S. Nucleic Acids Res. 1980, 8, 85. (2) Neidle, S . ; Berman, H. M.; Shieh, S. Nature 1980, 288, 129. (3) Savage, H. In Water Science Reuiews; Franks, F., Ed.; Cambridge University Press: Cambridge, U.K., 1986; Vol. 1.
0022-3654/90/2094-4660$02.50/0
sionally averaged structures, termed by Eisenberg and Kauzmann4 the “I” and “D” structures, respectively, permits the investigation of the dynamical as well as static aspects of structure in this system. This differentiation has proved to be crucial in understanding the nature of the hydrogen bond network in the structured water.
Background The dCpG/proflavin complex is shown in Figure 1. The crystal consists of stacked intercalated duplexes. The general organization of water in the crystal is shown in Figure 2. The observed hydration network is shown in Figures 3 and 4. The major groove cavity is filled with water molecules arranged in pentagonal arrays, and the minor groove features a flat polygonal disk of waters. In the arrays there are five edge-linked pentagons, four of which consist solely of waters while the fifth consists of four water molecules and a phosphate oxygen. These pentagons are anchored to the complex via the heteroatoms on the bases and the proflavines. Previous theoretical studies on nucleic acid crystal hydrates based on molecular simulation have been carried out on dCpG/proflavin and related systems. Mezei et aLs carried out a Monte Carlo (MC) simulation on the water in the crystallographic unit cell, with the nucleic acid adduct atoms fixed in their X-ray positions. Here 108 water molecules were included, a number chosen on the basis of independent density measurements (4) Eisenberg, D.; Kanzmann, W. Structure and Properties of Water; Oxford University Press: New York, 1969. ( 5 ) Mezei, M.: Beveridge, D. L.; Berman, H. M.; Goodfellow, j. m.;Finney, J . L.; Neidle, S . J . Biomol. Srruct. Dyn. 1983, 1 , 287.
0 1990 American Chemical Society
