The Tomlinson Model

Friction is a ubiquitous force governing everything from the scrape of a shoe to the precision of a nanodevice, yet its fundamental origins at the atomic scale remain a subject of deep scientific inquiry. How do individual atoms interact to resist motion and dissipate energy? Answering this question requires moving beyond classical laws and developing a new intuition for the nanoscale world. This is the knowledge gap that the simple yet profound Tomlinson model, one of the cornerstones of modern nanotribology, was designed to fill. This article unpacks this powerful conceptual tool. We will first explore the core ​​Principles and Mechanisms​​ of the model, dissecting how it explains the jerky dance of stick-slip motion and the conditions for frictionless sliding. Subsequently, we will venture into its diverse ​​Applications and Interdisciplinary Connections​​, revealing how the model predicts friction in real materials, explains the mystery of superlubricity, and even guides the development of modern artificial intelligence.

Imagine trying to drag a small, heavy ball bearing across a sheet of corrugated iron roofing using a rubber band. If the rubber band is very stiff, the ball will more or less follow your hand smoothly. But if the rubber band is very soft and stretchy, you’ll experience something different. As you pull your hand forward, the rubber band will stretch, but the ball will stay stuck in one of the grooves. It will stick. Then, as the force from the rubber band builds up sufficiently, pop! The ball suddenly jumps over the ridge into the next groove. It slips. This jerky motion—stick, stretch, slip, relax—is the essence of friction at the atomic scale, and a wonderfully simple model, first conceived independently by Prandtl and Tomlinson, captures its very soul.

To understand friction, we don't need to model every single atom in a block of wood sliding on a table. Physics often progresses by finding the absolute simplest picture that still contains the essential truth. For atomic friction, that picture is the ​​Prandtl-Tomlinson model​​.

Let's replace the ball bearing with a single atom at the very tip of a probe, like in an Atomic Force Microscope (AFM). The rubber band becomes a perfect spring with a stiffness we call $k$. The corrugated iron sheet becomes the beautiful, perfectly repeating landscape of a crystal surface. Atoms on a crystal surface are arranged in a periodic lattice, like an endless egg carton. As our tip atom moves across this surface, its potential energy goes up and down, up and down. We can approximate this periodic energy landscape with a simple cosine wave. Let’s say the distance between "grooves" (rows of atoms) is $a$, the lattice constant, and the energy difference between being in a groove and being on a ridge is related to an energy amplitude $U_0$.

Now, we pull the other end of our spring at a steady, slow velocity $v$. At any time $t$, the attachment point of the spring is at position $vt$. Let's say our tip atom is at position $x$. The spring is stretched or compressed by an amount $(x - vt)$. From basic physics, the energy stored in the spring is $\frac{1}{2}k(x-vt)^2$.

The total potential energy of our tip atom is the sum of these two competing energies: the desire to sit in an energy valley on the crystal surface and the pull of the spring. We can write this down in a single, powerful equation:

The first term is the wavy potential of the substrate, and the second is the smooth, parabolic potential of the spring. All the rich behavior of atomic friction—stick-slip, energy loss, and even a state of near-zero friction called superlubricity—is packed inside this simple formula. Our task is to unpack it.

Nature is lazy. It always tries to minimize potential energy. Our tip atom is no different; it will always seek to settle in a valley, a local minimum, of the total potential energy landscape $V(x,t)$.

Here is where the fun begins. The shape of this total landscape is a "dance" between the substrate's periodic wiggles and the spring's overarching parabola. As the spring's attachment point $vt$ moves steadily to the right, it's like we are sliding the parabola to the right, continuously changing the overall landscape.

Let's picture two scenarios:

1. ​​A Stiff Spring​​: If the spring is very stiff (large $k$), its parabolic potential is steep and narrow. When you add the substrate's gentle wiggles to it, the total potential still has only one dominant valley. As the spring pulls, this single valley just shifts smoothly to the right, and the tip atom glides along in it. The motion is smooth.

2. ​​A Soft Spring​​: If the spring is very soft (small $k$), its parabola is wide and shallow. Now, the substrate's wiggles are significant. Adding the two potentials can create a landscape with multiple valleys (local minima), separated by small hills (local maxima).

This is the origin of stick-slip. Our tip atom starts in one of these valleys. As we pull the spring, the whole landscape tilts. The valley our atom is in becomes shallower and shallower, while the next one over gets deeper. At a critical point, our atom's valley disappears entirely, merging with a nearby hill in what mathematicians call a ​​saddle-node bifurcation​​. The atom is now on an unstable slope and immediately slides—slips—down into the next available deep valley. The spring relaxes, the force drops, and the process begins anew. This is the atomic "pop" we talked about earlier.

This raises a crucial question: what exactly do we mean by "stiff" or "soft"? It must be relative to something. That "something" is the curvature of the substrate potential. The spring, with its stiffness $k$, tries to make the total potential landscape convex (always curving up). The substrate potential, at the top of its energy barriers, has a negative curvature that tries to create dimples and multiple valleys.

Smooth sliding is guaranteed if the spring's stiffness is always greater than the most negative curvature the substrate can muster. This defines a ​​critical stiffness​​, $k_c$.

• If $k \ge k_c$, the spring wins. The total potential is always convex, there's only one valley, and the tip slides smoothly. This frictionless-like state, arising from the structure of the system, is a form of ​​structural superlubricity​​.

• If $k \lt k_c$, the substrate can win. Multiple valleys can form, and the motion is characterized by stick-slip.

For our simple sinusoidal potential, the maximum negative curvature of the substrate occurs right at the top of the energy peaks. A straightforward calculation shows that this critical stiffness is:

This beautiful result tells us that a stronger atomic interaction (larger $U_0$) or a more finely corrugated surface (smaller $a$) makes it harder to achieve smooth sliding—you need a stiffer spring. The transition between these two regimes is governed by the dimensionless ratio $\eta = k/k_c$. As this ratio approaches 1 from below, the stick-slip events become smaller and smaller, vanishing completely at the threshold.

When stick-slip occurs, it leaves behind several tell-tale signatures.

First, there's a maximum force the substrate can exert before the tip lets go. This is the ​​static friction force​​. Intuitively, this happens where the energy landscape of the substrate is steepest. For our cosine potential, the force is the derivative of the potential, and its maximum value is easily found to be $F_s = \frac{2\pi U_0}{a}$. The spring must be stretched to exert this force before a slip can occur.

Second, the process is not reversible. If you pull the tip forward and then backward, the force you measure does not retrace its steps. It forms a ​​hysteresis loop​​. The area enclosed by this loop on a force-vs-displacement graph has a profound meaning: it is the energy dissipated as heat during one full stick-slip cycle. This, in a nutshell, is friction—the conversion of mechanical work into disordered thermal energy. The Tomlinson model allows us to calculate this dissipated energy precisely from its fundamental parameters.

Third, what is the distance of the "slip"? In the idealized limit of a very, very soft spring ($k \to 0$), the tip sticks for as long as it can and then slips to the most convenient nearby spot. Given the perfect periodicity of the crystal, this spot is in the next potential well over. The slip distance, in this elegant limit, is precisely one lattice constant, $a$.

Our story so far has taken place in a perfectly cold, quiet world ($T=0$). But in reality, atoms are constantly jiggling due to thermal energy. This jiggling provides a way for our tip atom to "escape" its potential valley even before the valley mechanically disappears. It can get a lucky kick of energy and hop over the barrier.

This is a process of ​​thermally activated escape​​, famously described by Kramers' theory. The rate of these hops, $\Gamma$, depends exponentially on the height of the energy barrier, $\Delta E$, and the temperature, $T$, roughly like $\exp(-\Delta E / k_B T)$, where $k_B$ is Boltzmann's constant. As the spring pulls and the barrier gets lower, the probability of a thermal hop per second skyrockets.

This has a crucial consequence: friction is no longer just a static property but becomes dependent on velocity and temperature. If you pull slowly, you give the tip more time to thermally hop over a relatively high barrier. If you pull quickly, the tip is forced to climb higher up the barrier before it either hops or the barrier disappears mechanically. This dynamic interplay means that the average friction force we measure depends on how fast we pull—a phenomenon observed in countless experiments, all beautifully rationalized by adding the last missing ingredient, temperature, to our simple mechanical model.

Now that we have grappled with the inner workings of the Tomlinson model, we might be tempted to sit back and admire the elegance of our little theoretical machine—a single point, pulled by a spring, skating over a cosmic washboard. But to do so would be to miss the real adventure. The true delight of a good physical model is not in its own sterile perfection, but in its power to illuminate the world outside the theory, to connect seemingly disparate phenomena, and to guide our hands as we build and explore. The Tomlinson model, for all its simplicity, is a master key that unlocks doors in fields from materials science to mechanical engineering and even computer science. So, let's take this key and see what we can open.

Imagine you are an explorer of the atomic realm, equipped with an Atomic Force Microscope (AFM). Your instrument's tip is the finger you use to touch and feel the landscape of atoms. When you drag this tip across a surface, what will you feel? Will it be a smooth glide or a jerky, staccato dance? The Tomlinson model answers this directly. As we saw, the behavior depends on a competition between the stiffness of your instrument, $k$, and the ruggedness of the atomic terrain, encapsulated in a critical stiffness, $k_c$. This critical stiffness is determined by the corrugation energy $U_0$ and the lattice spacing $a$ of the surface, through the beautifully simple relation $k_c = \frac{4\pi^2 U_0}{a^2}$.

If your cantilever is "soft" ($k \lt k_c$), it will bend and store energy as it drags the tip over an atom, only to release it in a sudden "slip" as the tip jumps to the next potential well. This is stick-slip motion, the atomic heartbeat of friction. If your cantilever is "stiff" ($k \ge k_c$), it can muscle the tip over the potential barriers smoothly, with no instabilities. By calculating $k_c$ for a given surface, we can predict the very nature of the friction we expect to measure, guiding the design of our experiments before we even turn the machine on.

But real surfaces are not just one-dimensional lines. Crystalline materials have a structure, a grain, like a piece of wood. It is easier to sand wood along the grain than against it. The same is true at the atomic scale, and the Tomlinson model, when extended to two dimensions, explains why. A surface might have different lattice spacings, $a$ and $b$, and different corrugation energies, $U_x$ and $U_y$, along different directions. The model predicts that the friction you feel will depend on the direction you are scanning. The critical stiffness itself becomes anisotropic, a function of the scan angle $\varphi$, a discovery that elevates friction from a simple scalar quantity to something with directional character. When scanning at an oblique angle on a rectangular lattice, the tip's path is not straight but becomes a microscopic staircase, a sequence of slips along the $x$-direction and slips along the $y$-direction, whose relative frequency is precisely determined by the scan angle and the lattice constants. This anisotropy is not a mere curiosity; it is a fundamental property of friction on ordered surfaces.

Our idealized washboard is a perfect, infinite crystal. Real surfaces are messier. They have edges, atomic vacancies, and steps—like cliffs one atom high. These are not just minor blemishes; they are often the main characters in the story of friction and wear. The Tomlinson model provides a brilliant framework for understanding their role. We can model a defect as a local modification to the potential landscape, a "speed bump" superimposed on the regular periodic potential.

Consider the edge of a graphene flake. The carbon atoms at the edge are undercoordinated—they have "dangling bonds" craving to interact with their surroundings. This enhanced chemical reactivity translates directly into a larger local corrugation amplitude, a stickier potential well $U_0$. A larger $U_0$ means a higher local energy barrier and a higher critical stiffness $k_c$. Both effects make it easier for the stick-slip condition to be met, leading to a dramatic increase in local friction. These defects become hotspots of energy dissipation.

This local increase in friction has profound consequences. It is at these very hotspots that surfaces begin to die. We can introduce a "wear initiation threshold"—a critical force beyond which atoms are permanently displaced or bonds are broken. A perfect, pristine terrace might exhibit a peak friction force well below this threshold. But when the AFM tip encounters a defect, the local friction force can spike. If this spike exceeds the wear threshold, a wear event is nucleated. The Tomlinson model, by showing us how local atomic structure dictates local force, allows us to predict the birth of wear, a process of immense technological importance for the longevity of any mechanical device.

After discussing how defects can increase friction, it is natural to ask: can we ever get rid of it? The surprising answer is yes, sometimes spectacularly so. This state of ultra-low friction is called structural superlubricity, and a multi-particle version of the Tomlinson model provides the key to understanding it.

Imagine not one tip, but an entire sheet of atoms sliding over another. If the two atomic lattices are perfectly aligned and have the same spacing (a commensurate interface), all atoms feel the same force at the same time. The total force is simply the force on one atom multiplied by the number of atoms, $N$. The friction is huge.

But what if the lattices are "incommensurate"—mismatched in their spacing or twisted at an angle? This is like trying to stack two different types of egg cartons. The bumps of one layer will randomly fall on the bumps and in the hollows of the layer below. The lateral force on one atom pushing to the right will be canceled by the force on another atom pushing to the left. The total force on the slider is the sum of a vast number of quasi-randomly phased forces. Much like a random walk, the total force does not grow in proportion to the number of atoms $N$, but only as its square root, $\sqrt{N}$. Since the contact area grows as $N$, the friction force per unit area scales as $N^{-1/2}$. For a sufficiently large contact, the static friction effectively vanishes! This is structural superlubricity. It is not about eliminating the atomic-scale potential, but about using statistics to cancel its effect. This beautiful principle is at the heart of the ultra-low friction seen in twisted 2D materials like graphene, which form moiré superlattices that provide the perfect stage for this force cancellation.

Here we face a puzzle. The everyday friction we feel, governed by Amontons' laws, states that the friction force is directly proportional to the normal load, $F \propto L$. But if we apply our model to a single, elastic nanoscale contact (like an AFM tip), we find something different. The real contact area, according to Hertzian mechanics, scales as $A \propto L^{2/3}$. If friction is proportional to this real contact area, then friction should scale as $F \propto L^{2/3}$. This sublinear relationship violates Amontons' law! Does our model fail so spectacularly when we try to scale it up?

The resolution is one of the triumphs of modern tribology, and it lies in appreciating the difference between a single nanoscale tip and a macroscopic object. No macroscopic surface is truly flat. On a microscopic level, it is a rugged landscape of mountains and valleys. When two such surfaces touch, they make contact only at the tips of the highest "asperities," or mountains. As we push them together with a greater normal load $L$, these asperities deform, and more of them come into contact. For a wide range of materials, especially those where asperities deform plastically, the result is that the total real area of contact becomes, to a very good approximation, directly proportional to the load, $A_{\text{real}} \propto L$.

If we assume a constant interfacial shear strength $\tau$, the friction force is $F = \tau A_{\text{real}}$, and thus $F \propto L$. Amontons' law is recovered! The law is not a fundamental law of nature in the same vein as Newton's laws; it is an emergent statistical property of a large number of micro-contacts. The Tomlinson model's failure to predict Amontons' law for a single asperity is not a failure at all; it is a profound insight, revealing that the rules of friction are different at the nanoscale and helping us understand how our familiar macroscopic laws arise from a complex, multiasperity reality.

The journey of our simple model does not end with a bridge to the macroscopic world; it continues into the digital frontier. Predicting friction in complex, dynamic systems is computationally expensive. Here, the Tomlinson model finds a striking new role as a guide for artificial intelligence.

When building a machine learning model to predict friction, one could naively feed it raw parameters like temperature $T$ and velocity $v$. But the Tomlinson model, especially when including thermal effects, tells us the relationship is more subtle. It predicts that in the thermally activated regime, friction depends not on $v$ itself, but on its logarithm, $\ln(v)$. It predicts that near the athermal limit, the behavior is governed by a characteristic power law. It tells us that what matters is not temperature alone, but the dimensionless ratio of thermal energy to corrugation energy, $\frac{k_B T}{U_0}$.

Armed with this physical insight, we can perform "feature engineering," creating input features for an AI model that are physically meaningful, such as $\ln(v)$ and $(\frac{k_B T}{U_0})[\ln(\frac{v_*}{v})]^{2/3}$. By teaching the machine the language of the underlying physics, we create far more powerful and accurate predictive models. The Tomlinson model becomes the theoretical foundation upon which data-driven science is built, allowing us to create "gray-box" models that blend physical laws with the flexibility of machine learning.

From the tip of an AFM to the heart of an AI, the Tomlinson model proves to be more than just an elegant abstraction. It is a lens through which we can view the rich and complex world of friction, revealing the deep unity between the atomic dance of a single atom and the grand, emergent laws that govern our everyday world.