The Mack Model

In the quest to fabricate microchips, the central challenge is sculpting circuits a thousand times thinner than a human hair. This feat is achieved through photolithography, a process that relies on selectively dissolving a light-sensitive polymer film, or photoresist. To transform this process from an art into a quantitative science, a robust predictive model is needed to link the resist's chemical state to its physical dissolution speed. Without such a model, controlling the creation of billions of nanoscale transistors would be an impossible guessing game.

This article delves into the Mack model, an elegant and powerful formula that provides this crucial link. It serves as a cornerstone for understanding and simulating photoresist development. In the following sections, you will discover the core principles and chemical mechanisms captured by the model, and then explore its wide-ranging applications and profound interdisciplinary connections. The ​​Principles and Mechanisms​​ section will dissect the model's equation, revealing the physical meaning behind its parameters and its relationship to the cooperative nature of polymer dissolution. Following that, the ​​Applications and Interdisciplinary Connections​​ section will demonstrate how engineers use this model as a predictive tool in semiconductor manufacturing, integrating it into complex simulations to control feature dimensions, manage process variability, and ultimately enable the creation of modern technology.

How do you sculpt something a thousand times thinner than a human hair? You can't use a chisel. Instead, you must master the art of controlled dissolution. In the world of microelectronics, this is the fundamental challenge of photolithography: to etch intricate, nanometer-scale circuits onto a silicon wafer by selectively dissolving a special light-sensitive polymer film called a ​​photoresist​​. The secret lies in a beautiful interplay of light, chemistry, and fluid dynamics, which can be captured by a surprisingly elegant piece of applied physics known as the Mack model.

Imagine the photoresist as a uniform layer of a special kind of paint. Our goal is to make some parts of this paint layer easy to wash away while other parts remain steadfast. We achieve this by changing its local chemistry with a pattern of light.

The key property we control is the local ​​protection fraction​​, a number we can call $P$. At every point $(x,y,z)$ in the film, $P$ represents how "insoluble" the polymer is, typically on a scale from 0 to 1. A high value of $P$ means the polymer is full of chemical "protecting groups" that act like armor, making it resistant to the developer liquid. A low value of $P$ means this armor has been stripped away, leaving the polymer vulnerable to dissolution. The central task of any dissolution model is to answer the question: how does the local dissolution rate, $R$, depend on the local protection fraction, $P$?

Nature provides two main strategies for this microscopic sculpture:

• In a ​​positive photoresist​​, the regions exposed to light become more soluble. During a post-exposure baking step, photogenerated acid molecules act as catalysts, stripping the protective armor from the polymer. Thus, exposure decreases the protection fraction $P$. When we apply the developer, we wash away the pattern that was illuminated.

• In a ​​negative photoresist​​, the opposite happens: exposed regions become less soluble. Here, the photogenerated acid catalyzes a crosslinking reaction, effectively weaving the polymer chains together into a tough, insoluble mesh. Exposure increases the protection fraction $P$. When we develop the resist, the unexposed regions wash away, leaving behind the hardened, light-defined pattern.

For a positive resist, we expect the dissolution rate $R$ to increase as $P$ decreases. For a negative resist, $R$ must decrease as $P$ increases. Let's focus on the more common positive-tone case to build our intuition.

Let's try to invent a formula for the dissolution rate from first principles. What are the essential ingredients? First, there must be a maximum possible speed, $R_{\max}$, at which the resist can dissolve. This occurs when there is zero inhibition ($P=0$) and is ultimately limited by things like how fast developer molecules can be supplied or how quickly dissolved polymer chains can diffuse away. At the other extreme, when the polymer is fully protected ($P \to 1$), it might still dissolve, albeit at a very slow crawl. We can call this non-zero "dark loss" rate $R_{\min}$. The rate $R$ should transition smoothly between these two limits as the protection fraction $P$ changes.

In the 1980s, Chris Mack proposed a brilliant empirical formula that captures this behavior with remarkable grace and utility. The ​​Mack model​​ describes the dissolution rate as a function of an "inhibition measure," $M$. This measure $M$ can represent any species that slows down dissolution, but it's most commonly associated with the protected fraction $P$. The model is given by:

Let's dissect this beautiful equation:

• The terms $R_{\max}$ and $R_{\min}$ are the physical speed limits we just discussed. The equation correctly produces $R(M \to 0) = R_{\max}$ and $R(M \to \infty) = R_{\min}$.

• $M_c$ is the ​​characteristic inhibition level​​. It represents the concentration of the inhibitor at which the dissolution rate is exactly halfway between $R_{\max}$ and $R_{\min}$. You can think of it as the "tipping point" of the resist's chemistry—a fundamental fingerprint of the material. If we identify the inhibitor $M$ with the protected fraction $P$, then $M_c$ is simply the protected fraction at which the rate is halved.

• And then there is the exponent $n$. This parameter, often called the ​​contrast​​ or ​​selectivity​​, governs the steepness of the transition from fast to slow dissolution. It's not just a mathematical curve-fitting parameter; it holds the secret to one of the deepest physical concepts in resist chemistry: cooperativity.

What does the exponent $n$ truly represent? Imagine trying to pull a long, heavy carpet off the floor. If one person pulls on a single thread, not much will happen. But if several people grab the edge of the carpet and pull together, it moves easily. The dissolution of a long polymer chain is much the same. A single deprotected site on a polymer might not be enough for a developer molecule to get a good "grip." For a chain to be plucked out of the resist matrix, the developer might need to attack several nearby sites simultaneously. This phenomenon is called ​​cooperative inhibition​​.

If a successful dissolution event requires $n$ neighboring sites to be simultaneously available (unprotected), then the probability of finding such a "dissolution motif" scales with the unprotected fraction, $U = 1-P$, raised to the power of $n$. This gives a physical basis for the rate being proportional to $U^n$. The exponent $n$ in the Mack model captures this same cooperative effect, governing the sharpness of the dissolution curve. The exponent $n$, therefore, tells us the degree of teamwork required for dissolution. A resist with $n=1$ is non-cooperative; each site acts independently. A resist with $n=10$ is one where dissolution is a highly coordinated event.

Now, let's push this idea to its extreme. What if $n$ is enormous, say $n \to \infty$? The smooth S-shaped curve of the Mack model steepens into a vertical cliff. The dissolution rate becomes a digital switch: it's either ON (at speed $R_{\max}$) or OFF (at speed $R_{\min}$), flipping at the critical threshold $M = M_c$. This extreme case is often called the ​​notch model​​ or, more generally, a ​​threshold model​​.

This reveals a profound trade-off at the heart of semiconductor manufacturing. A high value of $n$ is the chipmaker's dream, as it produces features with perfectly sharp, vertical sidewalls. But it is also a nightmare for process control. Such a resist is exquisitely sensitive. A tiny, accidental fluctuation in the exposure energy could shift the local inhibitor concentration just enough to cross the threshold, causing the dissolution rate to jump from nearly zero to its maximum value. This amplifies the smallest process variations into potentially catastrophic changes in the final circuit dimensions, shrinking the all-important ​​process latitude​​. The quest for better resists is, in many ways, a quest to optimize this balance between high contrast and robust stability.

The Mack model provides a powerful link between the chemical state of the resist ($P$) and the resulting dissolution speed ($R$). But to truly appreciate its place in the world, we must see it not as an isolated formula, but as one crucial link in a chain of physical processes.

First, where does the intricate, three-dimensional map of the protection fraction $P(x,y,z)$ come from? It is the final masterpiece of the preceding exposure and baking steps. During exposure, photons create a latent image of acid molecules. During the subsequent bake, these acid molecules embark on a random walk (diffusion) while simultaneously acting as potent catalysts, cleaving protecting groups wherever they go. This entire prequel can be described by its own set of reaction-diffusion partial differential equations. The solution to these equations provides the initial protection map, $P(x,y,z)$, that serves as the input to the Mack model for the development simulation.

Second, is the Mack model's specific mathematical form the only one possible? Not necessarily. In certain regimes, it can be approximated by simpler functions. For instance, in the limit of very low protection ($P \ll P_c$) and low cooperativity ($n=1$), the Mack model's behavior can be approximated by a simple linear function of the protection fraction $P$. This is a beautiful lesson in modeling: different mathematical descriptions can converge in specific physical limits. However, the full Mack model's ability to incorporate a non-zero minimum rate $R_{\min}$ and a tunable cooperativity $n$ makes it far more versatile and physically realistic for a wide range of resist systems.

Finally, we must always ask: when does our model break? The Mack model is fundamentally a ​​surface reaction-limited​​ model. It assumes the chemical reactions happening at the resist-developer interface are the slowest, rate-determining step. But what if the "road" to the worksite gets congested? What if the developer molecules are consumed at the surface so fast that they can't be replenished quickly enough from the bulk liquid? This is a ​​transport-limited​​ regime. We can quantify this by comparing the characteristic time for diffusion to the characteristic time for reaction. When diffusion is the bottleneck (a condition described by a large ​​Damköhler number​​), the overall rate is no longer governed by the elegant chemistry of the Mack model, but by the plodding pace of diffusion. The observed rate becomes less sensitive to the underlying chemistry, effectively "smearing out" the high contrast that the resist was designed to have.

Furthermore, the developer liquid isn't always pure. It may contain additives, such as inhibitors designed to adsorb onto the resist surface to fine-tune performance. To account for this, the Mack model can be extended. We can layer another set of kinetic equations on top of it to describe the dynamic balance of these additives adsorbing, desorbing, and being swept away by the dissolving interface.

This is the true beauty of the Mack model. It is not just a static formula, but a dynamic and extensible framework. It provides a clear, intuitive, and physically-grounded description of a complex process, while also defining the boundaries of its own validity and inviting extensions that connect it to an even wider universe of physical and chemical phenomena. It transforms the messy art of dissolving polymers into a quantitative science.

Having understood the principles of the Mack model, one might be tempted to see it as a neat but narrow piece of chemical physics. Nothing could be further from the truth. In reality, this model is not an isolated island; it is a critical bridge, a Rosetta Stone that translates the esoteric language of photochemistry into the practical grammar of silicon engineering. It serves as a vital gear in the vast, intricate machinery of Technology Computer-Aided Design (TCAD), the collection of software tools that allows us to design and manufacture the microchips powering our world before a single wafer is ever processed. Let us embark on a journey to see how this elegant piece of theory finds its power in the real world, connecting disciplines and enabling the marvels of modern technology.

At its most fundamental level, the Mack model is an engineer's crystal ball. In the high-stakes world of semiconductor manufacturing, where a single production run costs millions of dollars, the ability to predict the outcome of a process is paramount. If we know the chemical state of the resist after exposure—the pattern of protected and deprotected molecules—the Mack model allows us to answer a simple but crucial question: how long should we leave the wafer in the developer bath? By integrating the local dissolution rate, $R(P(z))$, from the surface down to the substrate, we can calculate the total time required for the developer to "eat through" the resist and clear the intended pattern. This calculation is not just an academic exercise; it is the basis for setting process recipes, and it can be adapted to handle real-world complications like the "notch effect," where the development rate is mysteriously inhibited in a thin layer right at the surface.

But prediction is only half the story. True mastery comes from control. Imagine the goal is to create a transistor gate with perfectly vertical sidewalls. A sloping wall would lead to an ill-defined gate length and a poorly performing transistor. What does it take to achieve this nanoscopic perfection? The Mack model provides a stunningly simple and profound answer. For the development front to move straight down like a piston, its speed must be constant at every depth. This means the dissolution rate, $R(P(z))$, must be independent of $z$. Since the rate is a strictly increasing function of the deprotection level, this implies a necessary and sufficient condition: the protection profile itself, $P(z)$, must be uniform from top to bottom. That is, we must have $\frac{dP}{dz} = 0$.

This insight is a powerful guide for the process engineer. It transforms the problem. The question is no longer "How do we get vertical walls?" but "How do we create a uniform protection profile?" The focus shifts to the preceding step, the Post-Exposure Bake (PEB), where photo-generated acid molecules diffuse and drive the deprotection reaction. To achieve a uniform profile, the engineer must encourage the acid molecules to spread out evenly, turning to the fundamental knobs of diffusion: temperature and time. A higher bake temperature or a longer bake time allows the acid to diffuse further, smoothing out the initial top-heavy concentration gradient and leading to the desired uniform chemical state. The Mack model, in this sense, acts as a compass, pointing engineers toward the correct process adjustments.

This predictive power also allows us to define the robustness of our process. How much can the exposure dose from the laser fluctuate before the feature we are printing becomes unacceptably wide or narrow? This is the question of "exposure latitude." By linking the dose to the chemical state, and the chemical state to the dissolution rate via the Mack model, we can calculate the sensitivity of the final Critical Dimension (CD) to tiny variations in dose. This sensitivity, it turns out, is directly proportional to the slope of the Mack model's rate curve. A very steep rate curve (a high-contrast resist) produces sharp features but is exquisitely sensitive to noise, whereas a gentler curve is more forgiving. The model allows us to quantify this trade-off, defining a "process window" within which we can operate reliably. This is not just modeling; this is the scientific underpinning of mass production.

The simple picture of a developer bath uniformly dissolving a resist film is, of course, an idealization. The real world is far more complex, and it is here that the Mack model shows its true versatility as a building block for more sophisticated, multi-physics simulations.

Consider the challenge of developing a deep, narrow trench. We can picture the developer molecules as tiny workers trying to get to their job site at the bottom of the trench. As they travel down, some are consumed by the reaction at the sidewalls. In a very narrow trench, this consumption can be significant, leading to a depletion of the developer concentration at the bottom. Since the dissolution rate predicted by the Mack model is proportional to the local concentration of reactants, the development slows down deep inside the trench. This phenomenon, known as the "loading effect" or "aspect-ratio dependent etching," can be modeled by coupling the Mack model's chemical kinetics with Fick's laws of mass transport. The result is a diffusion-reaction model that correctly predicts that the bottom of a high-aspect-ratio feature will clear more slowly than the top.

This ability to be integrated into larger physical frameworks extends to the third dimension. The sharp, 90-degree corners of a designed contact hole are a fiction of our CAD software. In reality, the developing resist front rounds these corners. Why? One reason is that the chemical gradient of the deprotection field is sharpest at these corners, which can locally accelerate the dissolution reaction. By augmenting the Mack model's speed with a term sensitive to this gradient, simulators can capture this effect. Advanced simulation tools then solve the Eikonal equation, $\left\| \nabla T \right\| = 1/S(\mathbf{x})$, where $S(\mathbf{x})$ is the local speed from our enhanced Mack model. This equation, which has deep roots in optics, describes the arrival time $T$ of the development front at every point $\mathbf{x}$ in the resist. By finding the surface of points where the arrival time equals the total development time, we can precisely predict the final, three-dimensional, rounded shape of the contact hole.

Perhaps the most profound application of the Mack model framework is in understanding and predicting the inevitable randomness of the nanoworld. The edges of a transistor gate are not perfectly smooth; they exhibit a nanoscale jaggedness known as Line-Edge Roughness (LER). This roughness arises from the discrete nature of matter and energy: the random arrival of individual photons, the random walk of a finite number of acid molecules.

How does the process design fight this randomness? The answer, once again, lies in the underlying physics that the Mack model helps describe. A fluctuation in the local chemical state will cause the edge position to jitter. The magnitude of this jitter, however, is not constant. It is inversely proportional to the slope of the latent image at the feature edge. Imagine trying to define a position on a gentle hill versus on the edge of a steep cliff. A small random push will move your position much more on the gentle hill than on the cliff. Similarly, a process that creates a sharp chemical "cliff" at the feature edge will be more robust to the random "pushes" from molecular-scale noise. The parameters of the Mack model, which define the contrast and sharpness of the dissolution process, are therefore direct levers for controlling the final geometric variability.

This connection provides the final, crucial link in the chain: from chemistry to device performance. The geometric roughness of a gate (LER) and the fluctuations in its average width (Line-Width Roughness, or LWR) directly cause fluctuations in the transistor's electrical properties, like its switching speed and leakage current. Using the principles of error propagation, the statistical properties of LWR—which are governed by the resist chemistry and development kinetics—can be translated into the statistical variance of the transistor's performance.

In this light, the Mack model is transformed. It is no longer just a deterministic equation but a statistical tool. It helps us understand how the randomness inherent in the quantum and molecular world propagates up the scales to affect the performance and reliability of the final chip. This is the heart of TCAD's mission: to build a causal chain from fundamental physics to final product yield. And in that chain, the simple, elegant kinetics of the Mack model serve as an indispensable link, connecting the dance of molecules in a developer bath to the symphony of logic in a microprocessor. The model's parameters are not abstract constants; they are the tuning knobs that engineers use to calibrate their simulations, "teaching" the software the unique personality of their specific process line, and verifying its consistency across different patterns and geometries. It is a beautiful testament to the power of physics to unify the seemingly disparate worlds of chemistry, engineering, and statistics.