Quantitative force measurements using frequency modulation atomic force microscopy theoretical foundations

INSTITUTE OF PHYSICS PUBLISHING Nanotechnology 16 (25) S94 S11 NANOTECHNOLOGY doi:1.188/957-4484/16/3/18 Quantitative force measurements using frequency modulation atomic force microscopy theoretical foundations John E Sader 1,4,5,Takayuki Uchihashi 2,MichaelJHiggins 2, Alan Farrell 2,Yoshikazu Nakayama 3 and Suzanne P Jarvis 2 1 Department of Mathematics and Statistics, University of Melbourne, Victoria, 31, Australia 2 SFI Nanoscience Laboratory, Lincoln Place Gate, Trinity College, Dublin 2, Republic of Ireland 3 Department of Physics and Electronics, Osaka Prefecture University, Osaka, Japan E-mail: jsader@unimelb.edu.au Received 22 November 24, in final form 12 December 24 Published 25 January 25 Online at stacks.iop.org/nano/16/s94 Abstract Useofthe atomic force microscope (AFM) in quantitative force measurements inherently requires a theoretical framework enabling conversion of the observed deflection properties of the cantilever to an interaction force. In this paper, the theoretical foundations of using frequency modulation atomic force microscopy (FM-AFM) in quantitative force measurements are examinedandrigorously elucidated, with consideration being given to both conservative and dissipative interactions. This includes a detailed discussion of the underlying assumptions involved in such quantitative force measurements, the presentation of globally valid explicit formulae for evaluation of so-called conservative and dissipative forces, discussion of the origin of these forces, and analysis of the applicability of FM-AFM to quantitative force measurements in liquid. 1. Introduction Quantitative force measurements using the atomic force microscope (AFM) intrinsically require a relation connecting the observed deflection properties of the cantilever to an interaction force. For static measurements [1], this relation is trivial and has facilitated its widespread use. However, the well known jump-into-contact instability of the cantilever can severely restrict the applicability of this technique, particularly in the measurement of short-range attractive forces. Dynamic force measurements [2] overcome this limitation by enabling the use of stiff cantilevers, while simultaneously enhancing sensitivity. However, despite their advantages, such dynamic techniques introduce significant complexity to the interpretation and extraction of quantitative 4 Author to whom any correspondence should be addressed. 5 Work performed while on leave at: SFI Nanoscience Laboratory, Lincoln Place Gate, Trinity College, Dublin 2, Republic of Ireland. force data. In this paper, we address this issue and present a detailed discussion and analysis of the theoretical framework for dynamic force measurements. We specifically focus on frequency modulation atomic force microscopy (FM- AFM) [2, 3] (for a review, see [4]), where a feedback circuit self-excites the cantilever at its resonant frequency, and consider the case where the amplitude of oscillation is kept constant and independent of the tip sample separation. To begin, we elucidate the connection between the observed frequency shift and change in driving force with socalled conservative and dissipative forces. In so doing, we establishthat such terminology may not necessarily reflect the true nature of the force, as is commonly assumed. We also present a detailed discussion of the underlying assumptions involved in interpreting the frequency shift and driving force unequivocally in terms of an interaction force. This poses severe restrictions on measurements where the force experienced by the tip is not a unique function of the tip sample 957-4484/5/394+8$3. 25 IOP Publishing Ltd Printed in the UK S94

Quantitative force measurements using frequency modulation atomic force microscopy theoretical foundations separation, e.g., adhesion and bond breaking measurements, as we shall discuss. Such vital considerations are usually ignored in FM-AFM force measurements. With this fundamental theoretical framework in hand, we then present explicit formulae for both conservative and dissipative forces, that are valid for any amplitude of oscillation. These formulae extend the results given recently by Sader and Jarvis [5] for conservative forces, by enabling both conservative and dissipative forces to be determined. Importantly, these formulae present a significant advance on other theoretical approaches for force extraction [4, 6 8], by permitting the force to be evaluated simply and accurately for any amplitude of oscillation. Finally, we present a detailed discussion of the applicability of FM-AFM to quantitative force measurements in ambient environments (fluids), due to its relevance to agrowingnumber of applications in biology and colloidal science, and include corroborative experimental results [9]. 2. Background theory We now derive the fundamental theoretical framework for the interpretation of FM-AFM force measurements. While aspects of this framework have been discussed previously [4, 1, 11], a rigorous development and discussion of the underlying assumptions and formulae and their implications is yet to appear in the literature. First, we note that in such measurements the cantilever is normally excited at its fundamental resonant frequency ω res only. Consequently, to describe the cantilever motion, it suffices to consider a damped harmonic oscillator with a single degree of freedom, m d2 w dt + b dw + mω 2 res 2 dt w = F int + F drive, (1) where w is the displacement of the cantilever tip from its unperturbed position, m is the effective mass of the cantilever, F int is the interaction force experienced by the tip, F drive is the driving force that excites the cantilever and b and ω res are the damping coefficient and resonant frequency of the cantilever in the absence of an interaction force, respectively. It is assumed that the interaction force is sufficiently weak so that changes in effective stiffness of the cantilever are small, leading to only minute changes in the resonant frequency. Furthermore, the motion of the cantilever is taken to be (approximately) harmonic, irrespective of the non-linear nature of the interaction force. Importantly, these fundamental assumptions are satisfied in practice under standard operating conditions, and are normally assumed in analysis of force measurements using FM-AFM [4]. By employing a self-excitation mechanism, the cantilever tip displacement w is forced to be 9 out of phase with the driving force F drive. Noting that the cantilever motion is (approximately) harmonic then enables the displacement and driving force to be expressed as w = a cos ωt, F drive = F sin ωt, (2) where a is the amplitude of oscillation, and ω is the driving frequency. Since the driving force is 9 out of phase FORCE SURFACE FORCE Figure 1. Schematic diagram showing force experienced by tip during one oscillation cycle. with the cantilever tip displacement 6,itthenfollows that the only permissible frequency of oscillation is the fundamental resonant frequency of the cantilever in the presence of an interaction force. Next, we examine the nature of the interaction force F int. Throughout this section, we only consider the case where the cantilever is oscillating at a fixed distance from the sample,i.e., the distance of closest approach z between tip and sample is held constant. Importantly, F int is a non-linear function of tip sample separation in general, and can differ on the approach and retract portions of the oscillation cycle, as illustrated in figure 1. We therefore decompose the periodic force experienced by the tip into the sum of an even and an odd force, i.e., the even part is the average of the approach and retract force curves, whereas the odd component is obtained by taking the difference between the approach and retract curves and dividing by two; see figure 2. We emphasize that this decomposition is completely general, irrespective of the nature of the force. The physical significance of these even and odd forces, however, can be interpreted by considering the work done W per oscillation cycle of the tip, W = F ds, (3) where F is the vector force experienced by the tip, ds is the elemental position vector in the direction of motion, and the integration is performed over a complete cycle of oscillation. Even component of force The work done on approach of the tip to the sample is equal in magnitude but opposite in sign to the work done on retraction, since the vector force is identical on approach and retract, but the elemental position vector changes sign. Consequently, the net work done per oscillation cycle is zero, and the even force can be formally connected to the conservative component of the force per cycle. 6 For mechanical actuation at the base of the cantilever, which is typically performed, the oscillation of the base is 9 out of phase with the tip displacement. S95

JESaderet al Even FORCE = + Odd Figure 2. Schematic diagram illustrating the decomposition of force experienced by the tip during one oscillation cycle into even and odd components. and approach curves for even force are identical. Odd component of force The work done on approach of the tip to the sample is identical to that done on retraction, since both the vector force and elemental position vector are equal in magnitude but opposite in sign on approach and retract. Consequently, the net work done per oscillation cycle is not zero in general, and the odd force can be formally connected to the dissipative component of the force per cycle. However, we emphasize that such connections to conservative and dissipative forces, although common and appealing [4], can be misleading and ambiguous since the origin of the forces cannot be determined solely by analysing the harmonic motion of the tip in isolation. Indeed, it is entirely possible thatthe conservative force results from a combination of conservative, dissipative and energy gaining processes over different parts of the oscillation cycle, even though the net energy lost or gained is zero over a complete cycle 7.Importantly, a formal connection to conservative and dissipative forces can only be made if the precise nature and origin of the forces involved is known. Therefore, contrary to convention, it is preferable to refer to these forces simply as even and odd in general, which shall henceforth be adopted. With this discussion in hand, the governing equations for the even and odd forces are then obtained by decomposing the interaction force, F int = F even + F odd. (4) 7 Such a situation can occur, for example, when the sample and/or tip surfaces are deformable, and the interaction involves both truly conservative and dissipative forces. As an illustrative example, consider the interaction between two charged deformable surfaces with fluid imbedded between them. Here, the hydrodynamic (dissipative) force can affect the separation between the surfaces, and hence the electrostatic (conservative) force. This results in the even and odd components of the force not being identical to the conservative (electrostatic) and dissipative (hydrodynamic) forces, respectively. Thus, even though the net energy lost per cycle by the even force is zero, referring to it as the conservative force is misleading and invalid. Substituting equation (4) into equation (1), and performing a Fourier analysis, then gives the required results. The Fourier cosine seriesgives the governing equation for the even force F even, ( ) ω 1 1 T = F even (z + a + w(t)) cos ωt dt, (5) ω res ak T whereas the Fourier sine series gives the governing equation for the odd force F odd, F + abω = 2 T T F odd (z + a + w(t)) sin ωt dt, (6) where ω is the change in resonant frequency 8, T is the oscillation period, z is the distance of closest approach between tip and sample, k is the dynamic spring constant of the cantilever 9,andF is the magnitude of the driving force. We note that equation (5) was originally derived by Giessibl [13] using an alternate but equivalent approach, whereas the well known relation for the average power dissipated per cycle [8] can be recovered directly from equation (6). To enable determination of the interaction force (both even and odd), it will be necessary that all forces are uniquely defined for a given tip sample separation. This condition is 8 The following definitions are used for changes in resonant frequency, generalized damping coefficient, and driving force: ω = ω res + ω, Ɣ = b + Ɣ, F = F + F. 9 Strictly, the dynamic spring constant is probed in dynamic force measurements, since the cantilever is oscillated at its fundamental resonant frequency. Importantly, the deflection function of a cantilever loaded by astatic force applied to its end-tip is similar but slightly different to the deflection function of a cantilever oscillating at its fundamental resonant frequency. Hence, the dynamic spring constant of a cantilever oscillating in its fundamental mode is similar to its static spring constant; it typically differs from the static spring constant normally quoted by a few per cent [12]. Clearly, if higher order modes of the cantilever are used, then the dynamic spring constant will differ greatly from the static spring constant. S96

Quantitative force measurements using frequency modulation atomic force microscopy theoretical foundations Odd = Odd X Even F odd = Γ ( z, a, ω,w) ẇ(t) Figure 3. Schematic diagram illustrating decomposition of odd force into the product of velocity and generalized damping coefficient. and approach curves for generalized damping coefficient are identical. satisfied by the even force F even,but not by the odd force F odd ;seefigure 2. This limitation can be overcome by further decomposing the odd force into the product of an additional odd and even function. For the additional odd function, we choose the velocity of the cantilever tip ẇ(t), asillustrated in figure 3. This generates what is commonly referred to as a generalized damping coefficient Ɣ(z, a, ω, w(t)) as the corresponding even function. Importantly, this decomposition is formally exact and rigorous, irrespective of the nature of F odd,andleads to the following exact expression: F odd = Ɣ(z, a,ω,w(t))ẇ(t), (7) with the generalized damping coefficient being uniquely defined for a given tip sample separation, provided the distance of closest approach z is held fixed, as has been assumed in this section. Note that the generalized damping coefficient can also depend on both the oscillation amplitude and frequency of oscillation. Equations (5) (7), and changing variables of integration, then gives the required governing equations for both the even force F even and the generalized damping coefficient Ɣ: ω = 1 1 u F even (z + a(1+u)) ω res πak du, (8a) 1 1 u 2 F F ω = 2 1 Ɣ(z + a(1+u)) 1 u 2 du, (8b) ω res πb 1 where F is the driving force in the absence of an interaction force. Ɣ is the change in the generalized damping coefficient resulting from the interaction, and F is the change in the driving force; both these values are with reference to their unperturbed values, i.e., in the absence of any interaction force (see footnote 8). 3. Fundamental requirements Force measurements using FM-AFM arenormally performed by varying the distance of closest approach z, andmonitoring the change in driving force and resonant frequency. To evaluate the interaction force giving rise to these changes, equations (8a) and(8b) mustbeinverted. However, this is only possible if F even and Ɣ are unique functions of the absolute tip sample separation (z + a + w(t)), irrespective of the distance of closest approach z. In cases where this condition is violated, then inverting equation (8) to determine the interaction force is not justified. This vital condition is often ignored in measurements, and can lead to erroneous results. A specific case where this condition is violated occurs when bonds are created and broken at different tip sample separations, e.g., adhesion measurements. In such a case, the breaking of bonds can lead to a reduced force as the distance of closest approach z is increased, even though the same absolute tip sample separation is probed. Conversely, a case where this fundamental condition can be satisfied occurs in the measurement of long-range forces, e.g., vdw and hydrodynamic forces. Also fundamental to the inversion of equation (8), as detailed in the following section, is the requirement that the interaction force is zero when the distance of closest approach is infinite. This condition is normally satisfied, but there do exist situations where it does not hold when interpreting measurements, requiring appropriate modification of the theory, e.g., see [14]. 4. Determination of forces In this section, we solve equations (8a) and(8b) toobtain analytical formulae enabling the even force F even and change in generalized damping coefficient Ɣ to be easily and S97

JESaderet al explicitly evaluated from the measured change in drive force and resonant frequency. This follows the methodology developed by Sader and Jarvis in [5], where the corresponding formula for the even force F even was derived. The explicit formulae presented here for both even and odd forces are universally valid for any oscillation amplitude. To begin, we note that equation (8b) can be transformed into the same form as equation (8a) using integration by parts. This leads to the following equivalent formula for equation (8b): = 1 1 u B(z + a(1+u)) du, (9) πab 1 1 u 2 where = F ω, (1a) F ω res B(x) = 2 x Ɣ(x) dx. (1b) We therefore focus on the general solution of equation (9), since this also gives the solution to equation (8a). Following the analysis in [5], we first express the unknown function B as B(z) = A(λ) exp( λz) dλ, (11) where A(λ) is formally the inverse Laplace transform of B(z). This implicitly requires that B(z) approach zero as z, which is a fundamental condition ofsolution (as discussed above), and places no other restriction on B(z). Substituting equation (11) into (9) then gives (z) = 1 A(λ)T (λa) exp( λz) dλ, (12) ab where T (x) = I 1 (x) exp( x), and I n (x) is the modified Bessel function of thefirstkind of order n [15]. Importantly, the governing equations for B(z) and (z) presented in equations (11) and (12) differ only by the function T (λa) in their respective integrands. This feature immediately enables inversion of equation (9), leading to the following exact analytical solution for B(z): B(z) = ba 2πi c+i c i (ẑ) exp( λ[z ẑ]) dẑ dλ, (13) T (λa) where c is a real constant. However, equation (13) is of little value in practice, since it necessitates numerical evaluation of the inverse Laplace transform, which can pose a formidable challenge. We therefore construct an approximate representation for T (x) by evaluating its asymptotic limits as x andx,andusing a Padéapproximant to connect these two limits: T (x) = x 2 ( 1+ 1 8 x + π 2 x 3 2 ) 1. (14) This approximate equation gives an excellent representation for T (x), seefigure 4; the error exhibited by equation (7) is less than 5% for all values of x, andisexact in the limits as x andx. T(x).2.15.1.5.1.1 1 1 1 1 1 x Figure 4. Comparison of the function T (x) (solid curve) and its approximation, equation (14) (dashed curve). Substituting equation (14) into equation (13), and using the properties of Riemann Liouville fractional calculus [16], then enables equation (13) to be evaluated explicitly, leading to the following result: ( ) a 1 2 B(z) = 2b 1+ 8 (t) π(t z) z a 3 2 d (t) dt. (15) 2(t z) dt Substituting equation (1b) intoequation (15), and applying equation (15) to equation (1a), then gives the required explicit formulae for the even force and the change in the generalized damping coefficient: ( ) a 1 2 F even (z) = 2k 1+ 8 (t) π(t z) z a 3 2 d (t) dt, 2(t z) dt Ɣ(z) = b ( ) a 1 2 1+ z 8 (t) π(t z) z (16a) a 3 2 d (t) dt, (16b) 2(t z) dt where (z) = ω(z), (17a) ω res (z) = F (z) F ω(z) ω res. (17b) Equations (16a) and(16b) aretheresults we seek, enabling the even force F even and generalized damping coefficient Ɣ to be evaluated explicitly from the measured frequency shift and change in driving force. We emphasize that these formulae are valid for all amplitudes of oscillation, provided the fundamental requirements detailed in section 3 are satisfied. We now demonstrate the accuracy and validity of these formulae by presenting a simulated experiment. Note that equations (16a) and(16b) must exhibit similar accuracy, since they are derived using the same formalism. Hence, we focus our discussion on equation (16a) forthe even force, noting that an identical conclusion holds for equation (16b). The assessment presented is identical to that given in [5], and is S98

Quantitative force measurements using frequency modulation atomic force microscopy theoretical foundations F even ().75 Arbitrary Amplitude Formula.5.25.25.5 1 2 3 4 5 Figure 5. Actual (solidcurve) and recovered (dashedcurve) Lennard-Jones force laws using equation (16a). Amplitudes of oscillation used are a/l =.1,.3, 1, 3, and 1. only briefly summarized here for completeness. The reader is therefore referred to [5] for a more detailed comparison and discussion. To assess the accuracy of equation (16a), we first specify a force law, evaluate the resulting frequency shift using equation (8a), and then recover the force law using equation (16a). The accuracy and validity of equation (16a) can then be assessed by comparing the original and recovered force laws. We choose a Lennard-Jones force law [5], consisting of a long-range attractive component and a shortranged repulsive force: ( l 4 F even (z) = G 3z 1 ), (18) 6 z 2 where G is a constant, and l is the separation where the attractive force is maximum which sets a natural length scale for the force law. A comparison showing the original and recovered force laws is given in figure 5. Results are presented covering the entire spectrum spanning from the case where the oscillation amplitude is small in comparison to the length scale l to the situation where the amplitude is large compared to l. From figure 5 it is strikingly evident that equation (16a) gives excellent accuracy, with the original and recovered force laws being virtually identical, irrespective of the amplitude chosen. Note that the maximum error of 5% occurs at intermediate amplitudes. This is expected, since the accuracy of equation (16a) (andequation (16b)) is dictated solely by the approximation for the function T (x). Consequently, equation (16a) (andequation (16b)) can be used with confidence to determine the even force F even and generalized damping coefficient Ɣ, regardless of their nature and the oscillation amplitude used. 5. Operation in liquids We now examine the applicability of the above theoretical formalism to measurements conducted in ambient (fluid) environments. This section significantly extends the discussion presented in [9] with regard to the effect of fluid on cantilever dynamics. Throughout the above analysis, it has been implicitly assumed that the effective mass of the cantilever is independent of the interaction force. While this is certainly true for operation in vacuum, measurements performed in ambient or liquid environments require careful consideration, since hydrodynamic forces imposed on an oscillating cantilever contain both dissipative and inertial components [17]. Hydrodynamic dissipative forces pose no problem, since they contribute only to the generalized damping coefficient. However, inertial forces add directly to the effective mass of the cantilever, thus modifying its resonant frequency. Therefore, unless accounted for, such inertial loading can lead to discrepancies in the measured interaction force, since the frequency shift is interpreted solely as a change in effective cantilever stiffness in the theoretical formalism. It is well known that the resonant frequency of a cantilever immersed in fluid can differ significantly from its value in vacuum [17, 18]. For a cantilever not in proximity to a surface, calculations and measurements [17, 18] show that for typical AFM cantilevers immersed in air the resonant frequency is reduced by several per cent, whereas the quality factor is reduced by several orders of magnitude from its corresponding value in vacuum. The effect is greatly enhanced in liquid, with the resonant frequency and quality factor decreasing by a further order of magnitude in some cases [18] from their values in air. However, the critical issue for FM-AFM force measurements is the change in inertial loading as the cantilever approaches the surface, since changes in dissipation are inherently accounted for in the theoretical formalism, as discussed above. For the formalism presented in sections 2 4 to be valid, the following conditions must be satisfied: (a) the resonant frequency must be independent of the tip sample separation; (b) inertial loading must be independent of frequency, otherwise a change in frequency due to an interaction force will result in a change in effective mass. First, we note that the hydrodynamic length scale for an AFM cantilever is given by its width [17]. Consequently, as a cantilever approaches the surface its inertial hydrodynamic loading will only be significantly affected if the cantilever sample separation is less than or comparable to the cantilever width. This property is demonstrated in figure 6(a), which shows the change in resonant frequency as the cantilever approaches a surface measured using FM-AFM. As expected, the length scale for the change in resonant frequency is given by the cantilever width. Consequently, if the tip sample separation is much smaller than the cantilever width, then a change in separation of the order of the tip sample distance will have little effect on the resonant frequency. This feature is demonstrated in figure 6(b), which presents results for the resonant frequency as a function of separation when the cantilever is in close proximity to the surface. Note that the resonant frequency is independent of separation in this case. Therefore, for practical FM-AFM force measurements where the tip is by necessity in close proximity to the surface, and the change in separation is much smaller than the cantilever width, condition (a) above is satisfied. Next, we note that inertial hydrodynamic loading is frequency dependent [17]. However, this dependence is weak and varies approximately as the square root of the oscillation frequency. Therefore, this variation can be neglected to leading S99

JESaderet al (a) 15 f (Hz) 2 25 2 4 6 8 1 12 14 (b) Separation ( µm) 15 ω ( ω 1 3 ) res (a) (b) 4 3 2 1 1 2.2 a = 2., 3.9, 7.2 nm 1 2 3 4 5 6 Separation (nm) a = 2., 3.9, 7.2 nm f (Hz) 2 25 Force (nn).1.1.5.1.15.2 Separation ( µm) Figure 6. Measurement of change in resonant frequency of cantilever immersed in water as a function of separation. The magnetic particle is glued onto the backside of the cantilever to enable magnetic activation [2]. The resonant frequency of the cantilever (Nanosensors EFM cantilever) in water is 13 12 Hz. Cantilever dimensions: length 225 µm, width 28 µm, tip height 13 µm. Maximum travelbythepiezo is 14 µm. Frequency shift is relative to the value when the cantilever is far from the surface. (a) Frequency shift over the last 14 µm beforecontact; (b) frequency shift over the last 2 nm before contact. order since it introduces an uncertainty of 1% which 1 is comparable to other uncertainties, e.g., errors in spring constant calibration. Condition (b) above is therefore also satisfied. Hence, for the above formalism to be valid, the specified resonant frequency ω res in equation (1) is the value when there is no interaction force F int between tip and sample, and the cantilever is in close proximity to the surface, i.e., when the tip sample separation is much smaller than the tip height and cantilever width, e.g., the resonant frequency shown in figure 6(b). If the resonant frequency far from the surface or in a different fluid medium (such as air) is used for the unperturbed value ω res,thenthis can lead to significant errors. To demonstrate the validity of using FM-AFM for quantitative measurements in liquid, we present results of measurements obtained in octamethylcyclotetrasiloxane (OMCTS), which exhibits short-range ordering (and hence forces) on confinement. Importantly, the force is expected to be a unique function of the tip sample separation in this system, since no bonds are formed or broken throughout the oscillation cycle of the cantilever, thus satisfying the fundamental conditions stipulated in section 3. These results appear in [9], and are presented here for completeness. 1 The inertial loading of AFM cantilevers is only weakly affected by proximity to surfaces, [19]. Frequency dependence of inertial loading can then be calculated using results for an unbounded fluid [17], leading to an uncertainty of 1%..2.5 1 1.5 2 2.5 3 Separation (nm) Figure 7. Experimental force measurements in OMCTS, as described in section 5. (a) Measured relative frequency shifts as a function of separation and oscillation amplitude; (b) recovered force versus separationcurve using data in (a) as a functionof separation and oscillation amplitude. a = 2.nm(solid curve), a = 3.9nm (dashed curve), a = 7.2nm(dotted curve). The reader is therefore referred to [9] for a more complete description and comparison. Measurements were performed against a freshly cleaved highly orientedpyrolytic graphite surface. To minimize hydrodynamic effects of the bulk standard AFM tip, and hence the required driving force, we used a multi-walled carbon nanotube as the probing tip. The carbon nanotube tip was made by attaching a nanotube onto a standard AFM cantilever tip (Nanosensors EFM cantilever) with a spring constant of 3 N m 1. A magnetic particle was glued onto the backside of the cantilever to enable magnetic activation [2]. The resonant frequency of the cantilever (with magnetic particle attached) was 28.19 khz in air. When immersed in OMCTS and in close proximity to the surface (5 nm tip sample separation), a resonant frequency of 19.54 khz was measured, and used in equation (1). Interestingly, this system shows little variation and structure in the driving force as a function ofseparation [9], andhence we only focus on changes in resonant frequency as a function of separation. Figure 7(a) shows frequency shift measurements obtained using a range of different oscillation amplitudes. As previously observed for measurements in vacuum [4], note that the frequency shift decreases as the oscillation amplitude increases. This is expected, since the cantilever tip experiences aforce over a smaller fraction of its oscillation cycle as the amplitude increases, hence decreasing the change in effective stiffness of the cantilever. Nonetheless, force curves obtained from each of these individual frequency shift measurements using equation (1a) are in good agreement; see figure 7(b). S1

Alossinsensitivity for the highest amplitude measurement (a = 7.9 nm) for separations greater than 3 nm is responsible for the apparent discrepancy with measurements at lower amplitudes. When the results given in figure 7(b) are scaled by the radius of the tip, force gradient values of the order of 1 mn m 1 are obtained which are also in excellent agreement with previous independent measurements. These results demonstrate the validity of the FM-AFM technique for quantitative force measurements in liquid, and when taken with the known capabilities of the method in vacuum [4] establish FM-AFM as a universal technique for quantitative force measurements regardless of the environment. 6. Conclusions The theoretical foundations for quantitative force measurements using FM-AFM have been rigorously elucidated with specific regard to the underlying assumptions. In so doing, we established that decomposition of the interaction force into conservative and dissipative components is only valid provided strict conditions are satisfied. Use of such terminology in general can be misleading and ambiguous, and alternative terminology was proposed. We also discussed the fundamental assumptions necessary for conversion of the observed deflection properties of the cantilever into an interaction force. Importantly, cases where these assumptions are violated exist in practice, requiring modification of the existing theoretical framework or restricting its use, e.g., in adhesion measurements or measurements where bonds are formed or broken leadingtoa non-unique force for a given tip sample separation. The theoretical formalism presented in [5] for even forces wasextended to encompass both even and odd forces, thus enabling access to the complete dynamic interaction force experienced by the tip. Finally a discussion of the validity of FM-AFM in quantitative force measurements in liquid was presented, establishing that the theoretical framework developed for operation in vacuum is also applicable to operation in ambient or liquid environments, provided certain modifications are implemented. Quantitative force measurements using frequency modulation atomic force microscopy theoretical foundations Acknowledgments This research was supported by a Science Foundation Ireland research grant (1/PI.2/C33), the Human Frontier Science Program, the Particulate Fluids Processing Centre of the Australian Research Council and the ARC Grants Scheme. JES acknowledges the award of a SFI Walton Visitor Award (2/W/M28). This research was conducted while JES was on leave at Trinity College Dublin. References [1] Burnham N A and Colton R J 1989 J. Vac. Sci. Technol. A 7 296 [2] Albrecht T R, Grutter P, Horne D and Rugar D 1991 J. Appl. Phys. 69 668 [3] Durig U, Zuger O and Stalder A 1992 J. Appl. Phys. 72 1778 [4] Garcia R and Perez R 22 Surf. Sci. Rep. 47 197 [5] Sader J E and Jarvis S P 24 Appl. Phys. Lett. 84 181 [6] Durig U 2 Appl. Phys. Lett. 76 123 [7] Giessibl F J 21 Appl. Phys. Lett. 78 123 [8] Gotsmann B, Seidel C, Anczykowski B and Fuchs H 1999 Phys. Rev. B 6 1151 [9] Uchihashi T, Higgins M J, Yasuda S, Jarvis S P, Akita S, Nakayama Y and Sader J E 24 Appl. Phys. Lett. 85 3575 [1] Durig U 2 New J. Phys. 2 5.1 [11] Holscher H, Gotsmann B, Allers W, Schwarz U D, Fuchs H and Wiesendanger R 21 Phys. Rev. B 64 7542 [12] Sader J E 22 Encyclopaedia of Surface and Colloid Science ed A Hubbard (New York: Dekker) pp 846 56 [13] Giessibl F J 1997 Phys. Rev. B 56 161 [14] Higgins M J, Riener C, Uchihashi T, Sader J E, McKendry R and Jarvis S P 25 Nanotechnology 16 S85 9 [15] Abramowitz M and Stegun I A 1975 Handbook of Mathematical Functions (New York: Dover) [16] Samko S G, Kilbas A A and Marichev O I 22 Fractional Integrals and Derivatives (London: Taylor and Francis) [17] Sader J E 1998 J. Appl. Phys. 84 64 [18] Chon J W M, Mulvaney P and Sader J E 2 J. Appl. Phys. 87 3978 [19] Green C P and Sader J E 24 in preparation [2] Jarvis S P, Uchihashi T, Ishida T, Tokumoto H and Nakayama Y 2 J. Phys. Chem. B 26 691 S11