Rolling and Spinning Friction Characterization of Fine

Particles using Lateral Force Microscopy based Contact

Pushing

∗
BILSAY SÜMER and METIN SITTI

Mechanical Engineering Department, Carnegie Mellon University, Pittsburgh, PA 15213-3890, USA

ABSTRACT. Here, we have utilized Lateral Force Microscopy (LFM) based mechanical pushing of
micro/nano-objects to study adhesion and friction characterization at the micro/nanoscale. Con-
tinuum micro/nano-friction models for particle rolling, spinning and sliding cases are discussed for
general particle-substrate interfaces. A rolling resistance model using the Double-Hertz model is
devoloped for such general interfaces. Using the friction models, the effect of work of adhesion,
effective Young’s modulus, and contact radius at the particle-substrate interface are studied in de-
tail. Combining friction models with experimental particle pushing vertical and lateral force data,
the critical frictional interface parameters such as critical rolling distance and the interfacial shear
strength are measured for a polystyrene particle and glass substrate interface. Results show that
the critical rolling distance varies with the particle radius, and it is measured to be 42 nm, 84 nm,
and 128 nm on average for 5 µm, 10 µm, and 15 µm radius particles, respectively. Next, using the
particle spinning experimental data, the interfacial shear strength of the particle-substrate interface
is measured as 9-15 MPa.

KEYWORDS. Lateral Force Microscopy; micro/nanotribology; micro/nanoscale contact mechanics;

particle friction.
∗
Corresponding author, sitti@cmu.edu

1
1 Introduction

The physical understanding of different types of particle motions, such as sliding, rolling, stick-
slip and spinning, is indispensable since the mode of motion of particles determines the energy
loss and wear in the contacting surfaces. Numerous researchers have studied these motions in
the micro/nanoscale, such as particle attachment and detachment from surfaces [1], aggregation
and dispersion of powders [2], and micro/nano-manipulation operations to achieve two- or three-
dimensional micro/nanoelectromechanical systems [3,4]. Among the possible particle motions, quite
a few works have addressed the object-substrate rolling and spinning friction characterization in
the micro/nanoscale, resulting in a need for more theoretical and experimental work. For example,
using an atomic force microscopy (AFM) probe and by observing the motion change of carbon
nanotubes (CNTs) from sliding-rotating mode to stick-roll mode, the friction anisotropy of CNTs on
graphite surface was explained [5]. Using a needle shaped tool under a scanning electron microscope
(SEM) or a piezoresistive AFM cantilever in ambient environment, different motion behaviors of
spherical objects were characterized [6, 7]. Utilizing a non-contact adhesion measurement, based on
the previous theoretical work of Dominik and Tielens [8], for the first time, the existence of rolling
moment of resistance of an adhesion bond between a microsphere and flat surface was experimentally
demonstrated [9].
This paper focuses on the theoretical and experimental determination of rolling and spinning
friction (resistance) of micro/nanoparticles. If an adhesive elastic spherical particle is brought into
contact with an elastic half-plane, there would be a finite adhesional circular contact area that
can be determined by using continuum elastic theories. However, when an external disturbance,
such as a lateral point force or moment, is applied at the center of the sphere to start the rolling
motion, the contact area changes from the static contact condition and takes an elongated shape
where the rolling contact area is not circular anymore as observed with glass ball experiments on
a polyurethane surface [10] or polydimethylsiloxane (PDMS) hemispheres on a silicon wafer [11] in
the macrodomain. The adhesional contact zone responses to the perturbation such that front side
is adhering and back side is peeling, resulting in a pressure distribution that is asymmetric about

2
the instantaneous center of rotation. Therefore, theoretical analysis of the rolling contact for the 3D
case (sphere) does not lead to an analytical solution due to the nonuniform contact area extension
in contrast to a two dimensional case (cylinder), where an implicit equation is derived describing
the rolling contact area and resistance [12]. Dominik and Tielens [8] proposed an approximate
analytical model to calculate the pressure distribution and resistance moment associated with the
rolling motion. However, their method can only be applied to certain material combinations where
the pressure distribution inside the contact area obeys the Johnson-Kendall-Roberts (JKR) model
and was not experimentally verified. If a lateral point force or moment is applied with some distance
from the center of the sphere, spinning motion of the particle is realized. Spinning motion is the
least investigated among the motion types although the slip mechanism is very similar to sliding
motion.
The ultimate goal of the presented work is to build a reliable method that leads to knowledge
about the mechanical and tribological properties of micro/nanoscale particles. A quasi-static model
of micro/nanoparticle behavior during manipulation is developed and used to explain the experi-
mental results. Previous work of Dominik and Tielens [8] is extended such that a cohesive zone
solution for the rolling moment of resistance has been found by adding the effects of adhesion using
a continuum elastic model, and possible particle motions such as sliding, rolling and rotation in mi-
cro/nanoscale are described by showing several mode diagrams. Mode diagrams have enhanced the
understanding of the manipulation and characterization procedure at the micro/nanoscale. In the
experiments, 5, 10 and 15 µm radius polystyrene (PS) microparticles have been pushed laterally on
a glass substrate and the friction forces are recorded while observing the possible particle motions
and the resistance to rolling and spinning motions are reported.
The organization of the paper is as follows. In Section 2, the kinematics of the transverse
pushing of the micro/nanoparticles is investigated and the critical forces to initiate each motion of
the particle are determined using the equilibrium equations. In Section 3, micro/nanoscale friction
models for rolling, sliding and spinning motion modes of a particle pushed on a flat substrate are
discussed. Here, the upper limits of the resistances for each motion are found for different effective

3
Young’s moduli and work of adhesion cases for varying lateral pushing forces. In Section 4, LFM
based contact pushing experiments for different sizes of PS microparticles on a glass substrate are
reported and these experimental data are used to quantify the particle-substrate friction parameters
such as critical rolling resistance and shear strength. Finally, conclusions and future directions are
reported in Section 5.

2 Method

To characterize the frictional properties of a fine particle moving on a flat and smooth substrate, an
LFM based lateral particle pushing method is proposed with a protocol such that: 1) Using top-view
optical microscope or AFM tapping mode scanning based images, the particle location is determined
precisely; 2) The AFM probe tip is located above the flat substrate with a height equal to the radius
of the particle (R), and substrate moved with a constant speed V along a line parallel to the y-axis
as illustrated in Figure 1, which passes through the particle center with or without an offset (x0 ); 3)
Due to this substrate motion, the particle is pushed parallel to the y-axis for a prescribed distance,
and probe vertical bending and torsional twisting angle (θ) are measured simultaneously. During
lateral particle pushing, AFM tip is assumed to be in contact with the particle at all times, and the
resulting tip-particle and particle-substrate interaction forces are shown in Figure 2. The objective
is to extract the particle-substrate frictional parameters depending on the particle motion mode,
which could be sliding, rolling, stick-slip, or spinning (rotation). Here, spinning is possible if the
particle is pushed along a line that has a non-zero offset (x0 ) value with the particle centerline.

Figure 1:

The friction of possible particle motion modes will be modeled using continuum micro/nanoscale
friction models, and experimental lateral and vertical force measurements during particle pushing
will be used to quantify the relevant frictional parameters. Approximate analytical models used for
motion modes will be derived depending on the lateral pushing force, x0 , and the tip, particle and
substrate geometrical and material properties. The substrate and the object surfaces are assumed

4
to be very smooth to neglect roughness effects, and the particle-substrate contact mechanics is
assumed to be elastic assuming small and reversible deformations.

Figure 2:

To extract the particle-substrate friction force from the LFM measurements, forces in Figure
2 are solved using a quasi-static dynamic behavior assuming a very slow constant pushing speed.
Taking a rectangular cross section AFM cantilever probe geometry with width (w), thickness (t) and
length (l), where the tip is at the very end of the probe with a height of h, quasi-static equilibrium
force equations can be given as

q
Ft = Fy2 + Fz2

ψ = tan−1 (Fy /Fz )

ft = Ft cos(ψ − φ)

Nt = Ft sin(ψ − φ)

fs = Ft sinψ

Ns = Ft cosψ (1)

where Ft is the reaction force at the tip, which is defined in terms of torsional (Fy ) and normal
forces (Fz ) of the AFM cantilever, subscripts s and t denote the substrate and the AFM probe tip,
respectively, ft and fs are the frictional forces and Nt and Ns are the normal reaction forces at
the tip-particle and particle-substrate interfaces, respectively, ψ is the loading angle, and φ is the
contact angle between the tip and the particle.
During pushing, particle and tip could stick, slide, stick-slip, roll, or spin. Conditions for such
motions are derived using (1). The particle could slide on the substrate if

fs > τs As (2)

5
and the tip could slide on the particle if
ft > τt At (3)

where τs and τt are the shear strengths and As and At are the contact areas for the particle-
substrate and the tip-particle interfaces, respectively, and the sliding friction is assumed as directly
proportional to the area of adhesional contact assuming negligible normal force, single asperity
contact and wearless interfacial sliding. Next, the particle could roll if

(fs + ft ) R > Mmax,s + Mmax,t (4)

or it could spin for a non-zero x0 (x0 ∈ (0, R]) and

Fy x0 > Γmax,s + Γmax,t (5)

where Mmax,s and Mmax,t are the maximum rolling resistances and Γmax,s and Γmax,t are the max-
imum spinning resistances at the tip-particle and particle-substrate interfaces, respectively. The
physical interpretation of these resistances will be discussed in detail in the subsequent sections.
Solving (1) using above conditions gives the critical lateral pushing force conditions for particle
sliding, tip sliding, particle rolling, and particle spinning, respectively as follows:

τ As
Ft ≥
sinψ
τ At
Ft ≥
cos(ψ − φ)
Mmax,s + Mmax,t
Ft ≥
R (sin(ψ − φ) + sinψ)
Γmax,s + Γmax,t
Ft ≥ (6)
x0 sinψ

In the schematic given in Figure 2, and with φ = ψ = π/2 taken from now on, the vertical Fz
force effect on the particle and the tip friction behavior becomes negligible. Thus, Ft = Fy will
be assumed in the friction models, but the friction of the tip-particle interface is included in the

6
experiment section. When Fy exceeds any of the above critical forces, the corresponding motion
behavior is observed. Multiple motion modes can also occur such as sliding with rolling, and rolling
with spinning, if any of the above conditions are obeyed simultaneously.

3 Friction Models

3.1 Rolling Friction Modeling

Assuming pure rolling of an adhesive elastic spherical particle on a rigid half-plane without sliding,
the aim of this section is to find a general expression for the upper limit of the resistance of rolling
moment. The maximum rolling resistance moment is the critical moment that should be exceeded to
start the irreversible rolling process. For a sphere with a radius R on a flat and smooth substrate, the
maximum rolling moment resistances in (6) are modeled using continuum elastic contact mechanics
models. Many aerosol fine particle works [13,14] have used an approximate rolling moment resistance
model such as

Mmax = Fp a (7)

where Mmax is the maximum rolling resistance, Fp is the pull-off force, and a is the equilibrium
contact radius. For a compliant and adhesional interface, the JKR model gives Fp = 1.5πRW ,
where W is the interfacial work of adhesion. However, this model can be used as a first-order
approximation and does not necessarily capture the physics of the problem, since the rolling force
is expected to be much smaller than the detachment force (pull-off force) and the detachment force
never acts at the equilibrium contact area.
More accurate particle rolling resistance models were also developed by calculating the approx-
imate analytical pressure distributions at the rolling interface using the JKR model [8], which is
only accurate for highly adhesional and compliant interface cases, such that :

Mmax ≈ 6πRW ξ (8)

7
where ξ is the critical rolling distance ( ≤ ξ ≤ a) and  is the interatomic distance [15]. Critical
rolling distance can be defined as the displacement of the sphere that may roll on the surface before
irreversible rearrangement in the contact area occurs.
The work described in this paper uses the resistance moment derivation method in [8] and
the Double-Hertz (DH) model to derive an analytical rolling resistance model for general particle-
substrate interface cases. Maugis [16] utilized a Dugdale potential to match the Lennard-Jones
potential where a constant adhesional stress acts over a finite radius where there is no real contact.
The Maugis-Dugdale (MD) model is widely accepted and used in micro/nanoscale contact mechanics
problems requiring a transition model. Greenwood and Johnson [17] presented an alternative to
the MD model called the DH model. In the DH model, the difference between two Hertz pressure
distributions is utilized where adhesive tensile stresses act inside and outside the contact area and
compressive stresses act only inside. An advantage of this model over the MD model is that the
shape of the gap outside the contact is described by elementary functions; hence, unlike in the MD
model, no elliptical integrals are involved. Also, the adhesive pressure distribution given by the
DH model includes only elementary functions that simplify the integral calculation in the rolling
resistance whereas the MD model includes inverse trigonometric functions, which should be solved
numerically. Therefore, the DH model is used in this work to derive an analytical expression for the
rolling resistance moment for general particle-substrate interfaces. In the DH model, the pressure
distribution is given by the combination of Hertz pressures along the contact axis (r) for an intimate
contact radius (a) with the adhesional tensile stresses in the radius of (c) as

" 1/2 1/2 #

c2 − r 2 a2 − r 2

P1 (r) = σ0 − r<a (9)
c 2 − a2 c 2 − a2

1/2
c2 − r 2

P2 (r) = σ0 a<r<c (10)
c 2 − a2

where σ0 is defined as the maximum adhesional stress and is given in terms of the work of adhesion

8
(W ) and the interatomic distance () at the closest point as

16 W
σ0 = √ (11)
9 3 

However, there is no direct relationship between the normal load (Fz ) and the real and adhesional
contact radii (a and c) in the DH model unlike the JKR model. Two closed-form equations should
be solved iteratively in order to find (a) and (c) for a given normal load. Defining a parameter
(m = c/a) between two contact radii, the equations are given in dimensionless form as [17]

π ∗ 2 (m + 1)
a∗ µ(m2 − 1)1/2 +a µ −3 =0
2 (m − 1)(m + 2)
2 ∗3 µ ∗2 2 h m − 1 i1/2
Fz∗ − a − a (m + m + 1) =0 (12)
3π 3 m+1

The dimensionless numbers that appear in the above equations are defined as

a
a∗ =
(R2 W/K)1/3
c
c∗ =
(R2 W/K)1/3
Fz
Fz∗ =
(R W/K)1/3
2
 R 1/3
µ = σo (13)
K 2W

where K = [(1 − ν12 )/E1 + (1 − ν22 )/E2 ]−1 represents the equivalent elastic modulus between the
particle and the substrate; E1 and E2 are the Young’s moduli, ν1 and ν2 are the Poisson’s ratios
for the particle and the substrate, respectively; and µ is the Tabor number [18] that represents the
ratio of the elastic displacement of the surfaces at pull-off to the effective range of surface forces.
Using the simulation parameters given in Table 1, the DH, MD and JKR models are compared in
Figure 3 according to the variations of the dimensionless real and adhesional contact radii (a∗ and
c∗ ) to µ in case of a zero normal load. The solution of a∗ for the MD and the DH models match
very close to each other for the full range of µ, while the DH solution of c∗ gives small discrepancy

9
with the MD solution for intermediate and low µ values. Both the DH and MD solutions approach
the JKR solution for large µ values.

Table 1:

Figure 3:

Using the DH model based solution, asymmetric rolling contact geometry in Figure 4(a) is
divided into two contact areas as shown in Figure 4(b) similar to [8] where the leading part is
increased by a ξ value and trailing part is decreased by the same amount and the leading and
trailing edges are assumed to be in the symmetric pressure distribution during the initiation of
rolling motion. The shifted distance is called as critical rolling displacement which can be defined
as a distance of sphere that may roll before irreversible rearrangement in the contact area occurs.
The four radii during the initiation of the rolling motion are defined as a + ξ, c + ξ, a − ξ, and c − ξ.
Moment calculations associated with only one of the contact radii are given below, since others can
be derived in the same way. The moment due to the shift of the left contact area for the real contact
radius a + ξ is given as:

s
3π/2 a+ξ
kσ02
Z Z
π
Mrl = xP1 (r, a + ξ)dA = (a + ξ)4 (14)
π/2 0 8 a2 − δR

where infinitesimal area dA and x can be rewritten as dA = rdrdθ and x = r cos θ in polar
coordinates. Here, δ is the indentation depth and is given in terms of arbitrary parameter (k)
such as:
(1 − k)a2 − kc2
δ= (15)
R

Figure 4:

By calculating the other three radii rolling moment resistances and adding them up, the total
maximum rolling moment resistance is computed as

10
 s
π kσ02
Mmax = Mrl + Mal + Mrr + Mar = 2
[(a + ξ)4 − (c + ξ)4 − (a − ξ)4 + (c − ξ)4 ] (16)
8 a − δR

where subscripts rl, al, rr, and ar represent the real contact radius of the left contact area, the
adhesional contact radius of left contact area, the real contact radius of the right contact area, and
the adhesional contact radius of the right contact area given in Figure 4(b), respectively. Conducting
a polynomial expansion and simplifying the equation by assuming 8ξ 3 a and 8ξ 3 c terms as negligible,
the total maximum rolling moment resistance (Mmax ) and related critical tangential force (Froll )
can be approximated as

(a3 − c3 )
Mmax ≈ πW ξσ0 √
c 2 − a2
πW ξσ0 (a3 − c3 )
Froll ≈ √ (17)
R c 2 − a2

The DH transition model given above is compared with the Dominik and Tielens model in (8) in
Figure 5 using the simulation parameters in Table 1. In the limiting case, where the Tabor number
is larger than 5, both models converge to each other, which shows the validity of the approximation.
For a Tabor number lower than 5, the JKR solution overestimates Froll , showing that the cohesive
zone is responsible for reducing the energy release rate during the rolling motion.

Figure 5:

3.2 Sliding and Spinning Friction Modeling

By applying a lateral point force with φ = π/2, the particle can slide depending on the applied
normal and lateral forces and adhesion at the particle-substrate interface. Assuming a single asperity

11
contact, sliding friction changes with the real contact area [19, 20] such as

Fslide = τ A (18)

where A = πa2 is the contact area, τ is the interfacial shear strength, and a is the real contact
radius. a is computed using the DH model in this work. Assuming a pure sliding case, microslip is
not considered, which could occur in the static friction phase.
For a perfect crystal, the theoretical shear strength τ is approximated as a constant value of
G/30 [21], which is shown to be the case for defect-free nanoscale contacts in the AFM experiments
[19], where G = 2G1 G2 /(G1 + G2 ) is the effective shear modulus given in terms of the shear moduli
(G1 , G2 ) of the particle and the substrate, respectively, defined as Gi = Ei /[2(1 + νi )]. However,
surface force apparatus (SFA) experiments, which have typically micrometer scale contact areas,
give τ again constant but much smaller than the AFM experiments, i.e. τ = G/1290. This shows
that the shear strength is scale dependent, which has been investigated using a micromechanical
dislocation model [22]. A single-dislocation-assisted (SDA) slip model is proposed for a wide range
of contact radii from 10 nm to 10 µm. Although no experimental evidence is presented so far, the
proposed theory fills the gap between the AFM and SFA experiments, and establishes the frictional
stress as a varying function of contact size. According to the SDA slip model, for very small contact
radii (a < 20 nm), τ is taken as G/43 using the previous AFM experimental results [19]. When
the contact size increases, shear stress is decreased until a second constant region is encountered,
where τ levels to G/1290 when a is larger than around 40 µm as observed in SFA experiments [20].
Therefore, the first region is the result of concurrent slip without the aid of dislocation motion; the
second region is the result of SDA slip; and the third region is the result of multiple-dislocation-
cooperated (MDC) slip. Thus, τ , which is a function of a, can be defined as

G/43 a < 20 nm





τ (a) = G10N (a/b)M 20 nm < a < 40 µm (19)



 G/1290

a > 40 µm

12
where M = tan−1 [(G/43 − G/1290)/(8 ∗ 104 b − 28b)], N = 28b, and b = 0.5 nm is the Burgers
vector [23].
In the pure spinning motion case, there would be a relative shear motion between the particle
and the substrate as in the case of sliding motion. However, particle rotates about its own center
without a real displacement in contrast to sliding motion. Therefore, the slip mechanism discussed
above is valid, and the maximum spinning moment resistance and lateral force required to spin the
particle for a given x0 value can be written as

Za
2π
Γmax = 2πr2 τ (a)dr = τ (a)a3
3
0
2π
Fspin = τ (a)a3 (20)
3x0

where Fspin is the critical lateral force to spin the particle (Γmax = Fspin x0 ).

3.3 Mode Diagrams

In this section, rolling, sliding and spinning friction models are used to create mode diagrams
showing the regions of possible motion of a micro/nanoparticle for given lateral pushing forces (Fy ),
effective Young’s moduli (K), and interfacial work of adhesion (W ). Here, it is assumed that a
lateral point force, Fy , is applied at the mid-plane of a sphere (ψ = φ = π/2) with a negligible
normal force component. Thus, Ft = Fy is taken from now on, where the friction between the
AFM probe tip and particle is assumed to be negligible. This is a reasonable assumption since
the order of magnitude for the tip-particle interface friction values is far smaller than the values of
the particle-substrate interface. The simulation parameters are given in Table 1 for a polystyrene
particle and a smooth glass substrate interface case since the experiments were conducted for these
materials. When the total force Fy obeys the condition Fy > Froll , Fy > Fslide , or Fy > Fspin , the
particle rolls, slides, or spins, respectively. If more than one conditions is fulfilled simultaneously,
the particle would show all of those motion behaviors such as rolling with sliding and spinning with
rolling.

13
 The diagram in Figure 6 shows the possible motion modes of a polystyrene particle for different
critical rolling distance assumptions where  < ξ < a. In the plots, the lower and upper limits of
the critical rolling distance, i.e., ξ =  and ξ = a, respectively, are taken to set a minimum and
maximum value for the rolling resistances, respectively. For each figure, there is a minimum lateral
force required to push the particle, otherwise the particle sticks to the surface. For this particular
configuration of the polystyrene particle and glass substrate, pure rolling is the most dominant
mode of motion and there are only small regions that are shown in Figure 6(c) where the particle
starts to slide first before rolling with increasing Fy .

Figure 6:

Rolling and sliding could be observed simultaneously, if Fy is higher than both Froll and Fslide
critical force limits. However, if Fy is applied with an offset x0 to the mid-plane of the particle and
Fy > Fspin , the particle would also spin (rotate). In the extreme cases, where x0 = R and ξ = ,
rolling is the first motion mode for a low Fy following a combined spinning and rolling motion as
seen in Figure 6. If Fy is increased further, the particle can roll, rotate and slide simultaneously
leading to a very complicated motion. However, in the extreme cases where x0 = R and ξ = a, first
the spinning motion occurs, followed by rolling and sliding motions with higher Fy .
Similarly, the effect of the interface properties, such as work of adhesion (W ) and equivalent
elastic modulus (K) on the motion of any particle on any flat and smooth surface are investigated
by keeping the radius of the particle fixed and varying the other parameters. K is varied from 5 GPa
to 100 GPa, while W is varied from 0.01 J/m2 to 1 J/m2 . In Figure 7, x0 = 0 and ξ = a are taken,
and the critical lateral forces required to roll and slide a particle are plotted in 3D for four different
particle radii. Sliding motion is observed first for nanoparticles with about 100 nm diameter with
W > 0.35 J/m2 for any K. For stiff nanoparticles with W < 0.35 J/m2 , rolling is observed first.
On the other hand, for much larger microparticles (e.g., R = 15 µm), rolling dominates the initial
motion at low lateral forces; only for highly adhesional (W > 0.3 J/m2 ) and relatively soft (K < 60
GPa) interfaces, sliding happens first.
When x0 = R, the spinning behavior of a particle with R = 5 µm is investigated in Figure 8

14
 Figure 7:

together with the sliding and rolling motions. For the whole range of W and K parameters, the
spinning resistance is much lower than the two other motions. Thus, spinning is observed first for
microparticles in general. However, when x0 → 0, spinning resistance can become the same as
or larger than the rolling and sliding resistances; thus rolling with spinning and rolling with both
spinning and sliding could be possible at a critical x0 .

Figure 8:

4 Experiments

4.1 Sample Preparation

Plain glass microslides (Fisher Scientific, Pittsburgh, PA, USA) were used as substrates. They
were kept approximately 5 minutes in acetone, methanol and deionized (DI) water filled cups in an
ultrasonic bath cleaner in the given sequence. They were dried using a nitrogen gun. By using a
micropipette, a monodispersed polystyrene particle dispersion (Alfa Aesar, Ward Hill, MA, USA)
diluted with DI water were scattered on a glass slide randomly with a low density, which uses the
surface tension based particle self-assembly during water evaporation and dried in air.

4.2 Experimental Conditions

Experiments were conducted in ambient conditions. Ambient environmental parameters such as

temperature and humidity were measured as 19 C 0 and 30−32%, respectively. Polystyrene particles
were selected since they are highly hydrophobic [24]. Thus, the water meniscus layer is not expected
to form between the tip-particle and the particle-substrate interfaces. The material of the AFM
probe tip was silicon. Since there is always a natural oxide layer with a nanometer scale thickness in
ambient conditions for silicon, tip-particle and particle-substrate interfaces have the same physical

15
and chemical properties. Measured from the AFM images, root mean square (rms) roughness the
glass slides was around 0.17 nm, suggesting that roughness effect on the friction forces could be
neglected. The effect of wear on friction was also neglected. It was assumed that the natures of
the friction between the tip and the particle, and also between the particle and the substrate were
predominantly adhesional due to the smooth surfaces and low normal forces. Finally, particles
were pushed with the same constant velocity (0.1 µm/s) in all experiments. Recent observations
indicate that dry friction increases with increased velocity [25], and the frictional properties of
a micro/nanoparticle heavily depend on the velocity. At high velocities, dynamic effects such as
inertial and damping forces should be taken into account whereas at very low speeds stick-slip
behavior of the particle is observed. Therefore, the substrate moves with a constant and low speed
that satisfies the quasi-static assumption in the force models.

4.3 Experimental Setup

An AFM system (Autoprobe CP-II; Veeco, Santa Barbara, CA, USA) with a top-view optical
microscope imaging system was used to monitor and manipulate the particles. In this system, the
AFM probe was fixed and the substrate was moved with a constant speed. A 25 µm diameter
polystyrene particle attached AFM probe (Novascan, Ames, IA, USA) with a stiffness of 7.75 N/m
was used to measure W between a glass and silicon oxide interface experimentally as approximately
0.1 J/m2 from the pull-off force measurements. For stable particle manipulation, a tapping mode
AFM probe (ATEC-NC; Nanosensors , Neuchatel, Switzerland) with a stiffness of around 45 N/m
and a tip radius of less than 10 nm was used. This probe has a tetrahedral tip that protrudes from
the very end of the probe, which enables direct top-view imaging of the AFM probe tip. During
contact pushing experiments, the AFM was operated in constant height mode where the force
feedback control was disabled, and the normal and lateral forces were recorded during the particle
pushing operation. Here, the substrate alignment is critical. To take care of possible alignment
errors, contact height of at least three points on the glass substrate were measured through the
AFM force-distance curves to calibrate the alignment of the substrate with respect to the xyz piezo

16
scanner motion axes.
The dimensions of the AFM cantilever were measured from the scanning electron microscope
(SEM) images where its length, width and tip height were measured as 151.9 µm, 45.1 µm and 17.2
µm, respectively. Since it was difficult to measure the cantilever thickness from SEM images, the
normal bending stiffness of the probe were calibrated using the Sader’s method [26] as 46.3 N/m
for measuring the vertical bending forces. Moreover, lateral force measurements require calibration
of the torsional stiffness of the AFM probe and lateral deflection sensitivity which are detailed in
the Appendix.

4.4 Experimental Results

LFM based contact pushing of polystyrene microparticles with radii of 5, 10 and 15 µm was per-
formed to measure the critical rolling and spinning forces. These diameters of particles were selected
to enable top-view optical microscope imaging in real-time while pushing the particles where the
Autoprobe CP-II optical microscope imaging resolution is limited to image only this size of diam-
eters reliably. In addition to the top-view optical microscope images, lateral and vertical forces on
the AFM probe during particle pushing were recorded to compute the critical frictional parameters.
As a first step, the critical rolling distances of polystyrene microparticles with 5 µm, 10 µm and
15 µm radii are characterized and sample lateral forces during rolling motion are given in Figure
9(a). A set of 15 pure rolling based pushing measurements were conducted using the same particles
for the three radii. Using these vertical (Fz ) and lateral (Fy ) experimental pushing force data and
the proposed maximum rolling resistance model in (17), ξ values were computed for the three radii
as 42.2 ± 5.9 nm, 83.5 ± 6.3 nm and 128.5 ± 13.6 nm, respectively. The resulting rolling and
spinning resistance values, in terms of lateral force values, are presented in Table 2. These values
are observed to be in the expected range of  < ξ < a. From these measurements, pull-off forces
and contact radii for the three radii are computed as 2.4 µN , 4.7 µN and 7.1 µN and 0.21 µm, 0.33
µm, and 0.43 µm, respectively, using the DH model discussed before. As expected, the rolling force
is much smaller than the detachment force and the critical rolling distance value is in the range of

17
 a a

ξ∈ ,
5 3
for the given polystyrene-glass interface and as the particle radius increases the critical
rolling distance approaches the real contact radius (a). Apart from the visual feedback observed
during the motion, sliding experiments were done using the PS attached AFM probes in order to
prove that these experiments were the result of pure rolling motion of the spheres. The experimental
procedure is detailed in the Appendix. The forces recorded in the sliding experiments were much
higher than the forces recorded during the rolling experiments proving that rolling experiments were
the result of the pure rolling motion of the spheres.

Table 2:

Figure 9:

In the experimental lateral force data analysis, we are interested in the critical force value (peak
value) that corresponds to a sharp drop in the data, which shows the initiation of the rolling motion.
If the AFM probe tip is withdrawn after this critical rolling distance, the force value is expected to
drop back to zero or a near-zero value. However, since the particle is forced to move continuously
on the surface after the motion starts, we hypothesize that the fluctuations in the lateral force data
are related to the continuous peeling and adhering of trailing and leading edges in the contact area,
respectively. Thus, the increase in the amplitude of the fluctuations in the force data in Figure 9(a)
for larger particles could be due to the increased tip-particle contact area.
For the characterization of the spinning resistance, microparticles were pushed with different x0
values close to their radius values. For two different x0 values, spinning lateral forces are shown in
Figure 10 for a 10 µm radius particle. As expected from (20), critical lateral force increases with
decreasing x0 . For characterizing the polystyrene particle and glass interface spinning resistance,
a set of 10 measurements with different x0 values were conducted for each particle radius using
the same particles. From these measurements, interfacial shear strength values were computed as
15.8 ± 1.38 M P a, 11.4 ± 0.27 M P a and 9.3 ± 0.25 M P a for particles having 5 µm, 10 µm and 15
µm radii, respectively. Since G = 2.71 GP a for the polystyrene-glass interface as given in Table
1, τ values were observed to be in the range of G/170 to G/290, which is in the intermediate

18
regime in (19). Also, a values are hundreds of nanometers, which also match with the intermediate
regime condition for the contact radii in (19). Moreover, it was observed that as the contact size
of the particle was increased, the shear strength values were decreased showing the scaling effect as
predicted with the theory.

Figure 10:

The pushing point of AFM tip on the particle should be carefully aligned. Any misalignment
from the center mid-plane could cause a loss of contact at the particle-tip interface due to a complex
rolling with spinning motion of the particle. Such a particle contact loss sample behavior is displayed
in Figure 11. In this particular experiment, a 10 µm diameter particle was pushed with a non-zero
mid-center offset value of 625 nm (x0 = 625 nm), which resulted in a rolling with spinning behavior.
Using the proposed models for a 5 µm radius particle with x0 = 625 nm, the critical force values
to initiate rolling and spinning motions are computed to be very close to each other as 75.3 and
74.1 nN, respectively. Therefore, x0 = 0 and x0 = R should be applied for pure rolling and spinning
behaviors, respectively, to avoid a complex rolling with spinning motion. Identifying and modeling
such complex motions in the dynamic regime are particularly important for the particle micro/nano-
assembly applications, and it is a future work.

Figure 11:

5 Conclusion

A LFM based contact pushing method has been proposed to characterize the tribological properties
of micro/nanoscale particle and flat substrate interfaces. Continuum micro/nano-friction models
for particle rolling, spinning and sliding cases have been devoloped for particle and flat substrate
interfaces. Using these models, the effect of particle-substrate interface work of adhesion, effective
Young’s modulus, and contact radius have been studied. If the particle is pushed from the centerline
of the particle with no offset, mode diagrams show that microscale particles, in most of the cases,

19
first start to roll and then slide at higher lateral forces while nanoscale particles tend to start to
slide first and then start to roll. By pushing particles with an offset of particle radius, it has been
shown that the spinning resistance is much smaller than the rolling resistance in most cases and the
sliding friction for microscale particles while it can get almost negligible at the nanoscale since it
scales with a3 . These trends might change completely depending on the specific particle-substrate
case, though they can still be predicted precisely using the given friction models. Combining the
friction models with the experimental pushing vertical and lateral force data, the critical frictional
interface parameters, such as the critical rolling distance and the interfacial shear strength have been
measured for a polystyrene and glass interface case. The critical rolling distance changes with the
particle radius, and it has been measured to be 42 nm, 84 nm, and 128 nm on average for 5 µm, 10
µm, and 15 µm radius particles, respectively. Next, using the particle spinning experimental data,
the interfacial shear strength of the particle-substrate interface has been measured as 9 − 15 M P a,
which matches with previous studies. This characterization method could be used to understand
and control the frictional behavior of micro/nanoparticles for 2D precision particle assembly (for
prototyping particle based micro/nanodevices and fabrication templates [27]), particle based solid
lubrication, particle removal, and aerosol applications.

Acknowledgement

This work is supported by the NSF CAREER award program (IIS − 0448042). The authors thank
Burak Aksak for his help with AFM, Cagdas D. Onal for his valuable comments on modeling and
Afshin Tafazzoli for his early works.

20
Appendix

A. Sliding Experiments

Besides the possible particle pushing modes such as sliding, rolling, rotation, sticking, stick-slip
motion, the sliding case is the most investigated one due to the unsophisticated measurement tech-
niques. Using conventional AFM probe tips or gluing a particle to the tipless AFM cantilever (colloid
probes), sliding shear strength of almost any material can be determined [28]. In the experiments,
PS attached AFM cantilever utilized which also used for the work of adhesion calculation. The
particle probe is forced to slide on the glass substrate with a negligible normal load while recording
the LFM voltage signal in friction loop. To derive friction loop, lateral force is plotted against the
lateral sample displacement as the piezoelectric stage moves perpendicular to the cantilever long
axis in a lateral direction, first one way then the other, as shown in Figure 12. At the beginning of
each trace, the shearing force increases until the static friction force is overcome followed by a com-
plete sliding of the particle in the kinetic friction regime. After 10 successful sliding experiments,
the pure sliding force value is observed to be 1112 ± 92 nN which is much higher than the rolling
and spinning critical forces.

Figure 12:

B. Calibration Procedures

Several lateral force calibration methods have been proposed in the literature such as the optical
geometry method [29], static friction method [30], vertical lever method [31] and wedge method [32].
The basic difficulty in the lateral force calibration is the high stiffness value in contrast to normal
force calibration causing a challenge to determine the lateral deflection sensitivity. For example, in
the static friction method, friction loop is obtained experimentally by conducting lateral scanning of
the AFM cantilever on a hard substrate in LFM mode and the slope of the static friction is taken as
equal to the lateral sensitivity. While this approach may give satisfactory results for sphere attached

21
probes (colloidal probes), in a typical AFM tip and substrate experiment, the contact stiffness of
the interface is comparable with the lateral stiffness of the cantilever. Although the contact stiffness
can be calculated using the continuum elastic theories and subtracted from the measured sensitivity,
this approach makes it an indirect method for the calibration. A new technique called diamagnetic
lateral f orce calibrator is proposed by Li et al. [33]. In this method, a pyrolytic graphite sheet
levitated by a strong magnetic field is used as a reference spring to apply a known force on the AFM
cantilever-tip assembly and by recording the output signals the force constants can be obtained as
a system response. The basic advantage of this system is the cross-talk effect (coupling between
normal and lateral force signals) elimination and easy setup. In our setup, the lateral stiffness value
of the pyrolytic graphite sheet was found to be 34.6 mN/m giving force constants of αll = 690.8
nN/V and αnl = −122.8 nN/V for the AFM cantilever used in the rolling and spinning experiments.
The force constants of particle attached probes were αll = 120.3 nN/V and αnl = −11.8 nN/V .

References

[1] P. Vainshtein, G. Ziskind, M. Fichman, and C. Gutfinger, Phys. Rev. Lett. 78, 551–554 (1997).

[2] A. D. Zimon, Adhesion of Dust and Powder. 2nd edition, Kluwer Acedemic , New York (1982).

[3] C. D. Onal and M. Sitti, IEEE Trans. Control System Technol. 15, 842–852 (2007).

[4] M. Sitti and H. Hashimoto, IEEE/ASME Trans. Mechatronics 5, 199-211 (2000).

[5] M. R. Falvo, R. M. Taylor III, A. Helser, V. Chi, F. P. Brooks Jr, S. Washburn, and R.
Superfine, Nature 397, 236–238 (1999).

[6] S. Saito, H. T. Miyazaki, T. Sato, and K. Takahashi, J. Appl. Phys. 92, 5140–5149 (2002).

[7] M. Sitti, IEEE/ASME Trans. Mechatronics 9, 343-349 (2004).

[8] C. Dominik and A. G. G. M. Tielens, Astrophys. J. 480, 647–673 (1997).

[9] M. D. Murthy Peri and C. Cetinkaya, Philosophical Magazine 85, 1347–1357 (2005).

22
[10] M. Barquins, D. Maugis, J. Blouet, and R. Courtel, Wear 51, 378–384 (1978).

[11] H. She and M. K. Chaudhury, Langmuir 16, 622 -625 (2000).

[12] M. Barquins, J. Natural Rubber Res. 5, 199-210 (1990).

[13] H. C. Wang, Aerosol Sci. Technol. 13, 386–393 (1990).

[14] M. Soltani and G. Ahmadi, J. Adhesion Sci. Technol. 8, 763–785 (1994).

[15] L. O. Heim, J. Blum, M. Preuss, and H. J. Butt, Phys. Rev. Letters 83, 3328–3331 (1999).

[16] D. Maugis, J. Colloid Interface Sci. 150, 243–269 (1992).

[17] J. A. Greenwood and K. L. Johnson, J. Phys. D: Appl. Phys. 31, 3279–3290 (1998).

[18] D. Tabor, J. Colloid Interface Sci. 58, 2–13 (1977).

[19] R. W. Carpick, D. F. Ogletree, and M. Salmeron, Appl. Phys. Lett. 70, 1548–1550 (1997).

[20] A. M Homola, J. Israelachvili, P. M. McGuiggan, and M. L. Gee, Wear 136, 65–83 (1990).

[21] A.H. Cottrell, Introduction to the Modern Theory of Metals. Institute of Metals, London
(1988).

[22] J. A. Hurtoda and K. S. Kim, Proc. R. Soc. Lond. A 455, 3363–3384, (1999).

[23] K. L. Johnson, Proc. Instn. Mech. Engrs. 214, 1–10 (2000).

[24] J. A. S. Cleaver and L. Looi, Powder Technol. 174, 34–37 (2007).

[25] H. Yoshizawa, C. Y. Lung, and J. Israelachvili, J. Phys. Chem. 97, 4128–4140 (1993).

[26] J. E. Sader, J. W. M. Chon, and P. Mulvaney, Rev. Scient. Instrum. 70, 3967–3969 (1999).

[27] A. Tafazzoli, C. M. Cheng, C. Pawashe, E. K. Sabo, L. Trofin, M. Sitti, and P. R. LeDuc, J.

Mater. Res. 22, 1601–1608 (2007).

23
[28] E. Gnecco, R. Bennewitz, T. Gyalog, and E. Meyer, Wear 192, 141–150 (1996).

[29] E. Liu, B. Blanpain, and J. P. Celis, J. Phys. : Condensed Matter 13, 619–642 (2001).

[30] R. G. Cain, M. G. Reitsma, S. Biggs, and N. W. Page., Rev. Scient. Instrum. 72, 3304–3312
(2001).

[31] S. Ecke, R. Raiteri, E. Bonaccurso, C. Reiner, H. J. Deiseroth, and H. J. Butt, Rev. Scient.
Instrum. 72, 4164–4170 (2001).

[32] D. F. Ogletree, R. W. Carpick, and M. Salmeron, Rev. Scient. Instrum. 67, 3298–3306 (1996).

[33] Q. Li, K. S. Kim, and A. Rydberg, Rev. Scient. Instrum. 77, 1–13 (2006).

[34] L. E. Nielsen and R.F. Landel, Mechanical Properties of Polymers and Composites. 2nd
Edition, CRC Press, Boca Raton, FL, (1994).

[35] K. Zeng, K. Breder, D. J. Rowcliffe, and C. Herrström, J. Mater. Sci. 27, 3789-3792 (1992).

24
Table 1: Simulation parameters used in the contact mechanics analysis for a polystyrene (PS)
particle and glass substrate interface.
Parameter Symbol Value Units Reference
Elastic Modulus (PS) E1 3.80 GP a [34]
Elastic Modulus (Glass) E2 73.40 GP a [35]
Shear Modulus (PS) G1 1.42 GP a
Shear Modulus (Glass) G2 31.40 GP a
Poisson’s Ratio (PS) ν1 0.34 - [34]
Poisson’s Ratio (Glass) ν2 0.17 - [35]
Work of Adhesion W 0.1 J/m2
Contact Angle φ π/2 rad.
Loading Angle θ π/2 rad.

25
Table 2: Experimental rolling and spinning resistance results for different radii (R) particles.
R (µm) Froll (nN ) ξ (nm) a Fspin (nN ) τ (M P a) b
5 78.2 ± 14.1 42.2 ± 5.9 64.2 ± 5.1 15.8 ± 1.38
10 157.3 ± 12.2 83.5 ± 6.3 87.9 ± 1.2 11.4 ± 0.27
15 254.2 ± 25.3 128.5 ± 13.6 105.2 ± 1.4 9.3 ± 0.25
a
The critical rolling distance.
b
Interfacial shear strength.

26
 Photodiode
Laser

zp z

tip x
y
V microparticle

substrate

Figure 1: Schematic of the LFM based particle lateral pushing and friction characterization setup:
an AFM probe tip is used to push a micro/nanoparticle on a flat and smooth substrate in lateral
y direction while measuring the corresponding vertical and torsional probe deflections. A reflected
light type optical microscope is used for the real-time top-view images of the probe tip and the
particle with a micrometer scale resolution.

27
 Fy Fz

y f xo
Rotation
Гt+Гs
z Top view
t
t

t Rolling
s t
Microparticle
z
s
y
Sliding

s
d Substrate

Front view

Figure 2: Schematic of the tip-particle and particle-substrate interaction forces during the lateral
particle pushing operation with a constant speed of V . φ = ψ = π/2 is taken to apply only a lateral
point force on the particle during pushing in the friction modeling section. Particle can stick, slide,
roll, spin, or stick-slip (Inset image shows the spinning case where the particle is pushed with an
offset x0 from the particle centerline.) depending on the lateral pushing force, x0 , particle size, and
tip, particle and substrate material and interfacial adhesion properties.

14

12

a* (DH)
10 a* (MD)
c* (DH)
c* (MD)
8
a *, c *

a*= c* (JKR)

0
0 1 2 3 4 5
Tabor number (µ)

Figure 3: The simulated variation of a∗ and c∗ with the Tabor number (µ) according to the DH,
MD and JKR models in case of a zero normal load.

28
 Front View
R R
Fy

Cohesive Zone O
O

a a
Top View

c c
ξ

Static Contact Pure Rolling Contact

(a)

a-ξ
a+ξ
c-ξ
c+ξ
ξ

(b)

Figure 4: a) Top and front views of a sphere in static and rolling contact on a flat surface. The
sphere is rotated in clockwise direction and elongated shape of the rolling contact with the critical
rolling distance (ξ) is shown in the top view of the rolling contact schematic. The solid lines indicate
the edges of the real contact area and the dotted lines indicate the edges of the adhesional contact
area. b) Top view of the approximated shape of the rolling contact.

29
 1

0.9
Transition Model
Dominik and Tielens Model
0.8

0.7
Fy

0.6

0.5

0.4

0 1 2 3 4 5 6 7 8
Tabor number (µ)

Figure 5: Comparison of the DH transition model (17) with the Dominik and Tielens model in [8]
for a zero normal load where the DH solution is normalized by the JKR solution using the simulation
parameters in Table 1.

30
 3
3 x 10
−3

Sliding
−3
x 10
8
Sliding
2.5 Rolling
2.5
6
Rolling 3
Spinning
F (µN)

Fy (µN)
4
2
y

2
2 0
2 1

0
F ( N)

F ( N)
1 2 3 4 0 1 2 3 4
1.5 R (µm) −3
x 10
1.5 R (µm)
y

y
1
1
0.5
0.5
0
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
R ( m) R ( m)
(a) (b)

3 3
x 10
−3
Sliding
0.02
Sliding
2.5
20
Rolling
2.5 0.015
Rolling 15
Spinning
F (µN)

Fy (µN)
0.01 10
y

5
2 0.005
2 0
0
0 0.01 0.02 0.03 0.04 −5
Fy ( N)

R (µm)
Fy( N)

0 0.01 0.02 0.03 0.04 0.05

1.5 1.5 R (µm)

1 1

0.5 0.5

0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
R ( m) R ( m)

(c) (d)

Figure 6: Simulated rolling, sliding, and spinning particle motion mode diagrams for various particle
radii: a) The particle is pushed at the mid-plane (ψ = φ = π/2) and rolling occurs for the whole
region where the critical distance is taken as the atomic distance (ζ = ). b) The particle is pushed
with some offset (xo = R) and spinning, sliding and rolling lines are shown (ξ = ). c) The critical
distance is taken as the real contact radius (ξ = a) and rolling is the dominant part for the most of
the region and transition occurs at about 200 nm and the motion of the particle is depicted in the
figure being pushed from the mid-plane of the sphere. d) The critical distance is taken as the real
contact radius (ξ = a), and the motion of the particle is shown in the figure pushed with an offset
(xo = R).

31
 (a) (b)

(c) (d)

Figure 7: 3D particle motion mode diagrams showing the possible regions of sliding and rolling
motions for a) 50 nm, b) 0.5 µm, c) 5 µm, and d) 15 µm particle radii. K is the effective Young’s
moduli and W is the interfacial work of adhesion. Here, x0 = 0 and ξ = a.

32
 (a)

(b)

Figure 8: a) A 3D particle motion mode diagram showing the possible regions of sliding, rolling
and spinning motions for a particle with 5 µm radius. K is the effective Young’s moduli and W is
the interfacial work of adhesion. Here, x0 = R. b) The spinning 3D critical lateral force diagram is
shown for the same particle.
33
 50

R=5 m
0 R = 10 m
R = 15 m

−50

−100
Fy (nN)

−150

−200

−250

−300
0 5 10 15 20 25
y-position ( m)

(a)

(b)

Figure 9: a) Experimental lateral force data during rolling a 5 µm, 10 µm and 15 µm radius
polystyrene particle. (The AFM probe is fixed and the substrate with the particle is moved with
a constant speed.) The particles are pushed approximately 10 µm distance. b) Top-view optical
microscope images before (left image) and after (right image) the pushing operation for a 10 µm
radius particle to show the traveled distance.

34
 20

−20
xo = 9.2 µm

−40 xo = 8.5 µm
Fy (nN)

−60

−80

−100

−120
5 10 15 20 25
y-position (µm)

Figure 10: The experimental spinning lateral force data for a 10 µm radius particle for two different
offset (x0 ) values.

20

-20

-40

-60

-80

-100

-120

-140

-160

-180
0 5 10 15 20 25 30 35
y-position (µm)

Figure 11: Sample experimental lateral force data where the microparticle loses the contact with
the AFM probe tip due to a non-zero x0 value initially (x0 = 625 nm).

35
 Kinetic Friction Phase
(Complete Sliding)
Fy (μN)

Static Friction Phase

y-position (nm)

Figure 12: Sample friction loop is shown for a PS particle attached probe while particle slides on a
glass surface where the normal force is negligible.

36