Routh Array

The stability of a dynamic system, from an aircraft's flight controls to a chemical reactor, is one of its most critical properties. This stability is determined by the roots of its characteristic polynomial, which must all lie in the left-half of the complex plane to ensure disturbances decay over time. However, directly calculating these roots for high-degree polynomials is computationally intensive and often impractical. This article addresses this challenge by introducing a powerful and elegant alternative: the Routh-Hurwitz criterion.

In the following chapters, you will embark on a journey to master this essential tool. The "Principles and Mechanisms" chapter will demystify the construction of the Routh array, a simple tabular method for counting unstable roots without solving for them, and will guide you through handling special cases that reveal deeper insights into system behavior. Subsequently, the "Applications and Interdisciplinary Connections" chapter will demonstrate the criterion's practical power, showcasing how it is used in control engineering to design stable systems and how its fundamental principles apply across various fields of science and physics.

Imagine trying to understand the character of a person. You could perform an exhaustive analysis of their entire life history, a daunting and often impossible task. Or, you could ask a few cleverly chosen questions and, from their answers, deduce a great deal about their nature. In the world of engineering and physics, we face a similar challenge. The "character" of many dynamic systems—be it an aircraft's flight controls, a chemical reactor, or a levitating magnet—is encoded in a mathematical expression called a ​​characteristic polynomial​​. The stability of the entire system, its very ability to not fly apart or crash, depends on the properties of the roots of this polynomial.

For a system to be stable, all of its characteristic roots, which we can call $s$, must have negative real parts. Why? Because the system's response over time often includes terms like $\exp(st)$. If the real part of $s$ is positive, this term grows exponentially towards infinity—a catastrophic failure. If the real part is negative, the term gracefully decays to zero, meaning the system settles down after a disturbance. Roots with zero real parts correspond to sustained oscillations, like a bell ringing forever—a state we call marginal stability. So, the "good" roots live in the left-half of the complex number plane, and the "bad" ones live in the right-half plane.

The obvious approach seems to be finding all the roots of the polynomial and checking them one by one. But for polynomials of higher degrees, this is computationally brutal and often impossible to do by hand. We need a more elegant way. We need to ask the right questions.

Enter the genius of Edward John Routh. In the late 19th century, he developed a stunningly simple procedure that tells us exactly how many "bad" roots a polynomial has, without ever calculating them. The ​​Routh-Hurwitz criterion​​ is like a brilliant accountant who can determine a company's financial health by arranging numbers from the ledger into a special table, rather than by tracking every single transaction. This special table is the ​​Routh array​​.

The construction is wonderfully mechanical. You take the coefficients of your polynomial, say $p(s) = a_n s^n + a_{n-1}s^{n-1} + \dots + a_0$, and write them down.

1. The first row of the array consists of the first, third, fifth, ... coefficients ($a_n, a_{n-2}, a_{n-4}, \dots$).
2. The second row consists of the second, fourth, sixth, ... coefficients ($a_{n-1}, a_{n-3}, a_{n-5}, \dots$).

Let's try this with a system whose characteristic equation is $s^4 + 2s^3 + 8s^2 + 4s + 3 = 0$. The coefficients are $1, 2, 8, 4, 3$.

Our first two rows, labeled by the corresponding power of $s$, are:

(We can pad the second row with a zero for bookkeeping.)

Now, we generate the rest of the table. Each new entry is calculated from a small $2 \times 2$ block from the two rows directly above it. The rule for the first element of the $s^2$ row is: $\frac{(\text{1st element of } s^3 \text{ row}) \times (\text{2nd element of } s^4 \text{ row}) - (\text{1st element of } s^4 \text{ row}) \times (\text{2nd element of } s^3 \text{ row})}{(\text{1st element of } s^3 \text{ row})}$ For our example, this is $\frac{(2)(8) - (1)(4)}{2} = \frac{12}{2} = 6$. The next element in the $s^2$ row is $\frac{(2)(3) - (1)(0)}{2} = 3$. So our array grows:

We continue this process until we reach the $s^0$ row. The full array becomes:

Here is the magic. The number of unstable roots—those in the right-half plane—is simply the number of times the sign changes as you read down the very ​​first column​​. For our example, the first column is $[1, 2, 6, 3, 3]$. Every number is positive. There are ​​zero​​ sign changes. Therefore, there are zero roots in the right-half plane. The system is stable!

What does an unstable system look like? Consider a magnetic levitation system with the characteristic polynomial $s^3 + 10s^2 + 31s + 1030 = 0$. Let's build its Routh array.

The first two rows are:

The first element of the $s^1$ row is $\frac{(10)(31) - (1)(1030)}{10} = \frac{310 - 1030}{10} = -72$. The first element of the $s^0$ row is $\frac{(-72)(1030) - (10)(0)}{-72} = 1030$.

The first column is $[1, 10, -72, 1030]$. Let's check the signs:

• From $10$ to $-72$: That's one sign change (positive to negative).
• From $-72$ to $1030$: That's a second sign change (negative to positive).

Two sign changes. This means the polynomial has exactly ​​two​​ roots in the unstable right-half plane. The magnetic levitation system will fail. The Routh array, with a few steps of simple arithmetic, has delivered a verdict of doom.

Sometimes, the mechanical process of building the array hits a snag. This isn't a failure of the method; it's the method telling us something deeper and more specific about the system's character. There are two main "snags."

A Lone Zero in the First Column

What if an element in the first column becomes zero, but the rest of its row is not zero? The formula for the next row involves dividing by this element, and we can't divide by zero. Consider the polynomial $p(s) = s^4 + s^3 + 3s^2 + 3s + 2$.

The first two rows are $[1, 3, 2]$ and $[1, 3, 0]$. The first element of the $s^2$ row is $\frac{(1)(3) - (1)(3)}{1} = 0$. We have a problem.

The fix is wonderfully pragmatic. We replace the zero with a tiny, positive number, which we call $\epsilon$. Think of it as giving the number a little nudge off of zero. The $s^2$ row becomes [$\epsilon$, 2]. Now we continue. The first element of the $s^1$ row becomes $\frac{(\epsilon)(3) - (1)(2)}{\epsilon} = 3 - \frac{2}{\epsilon}$.

The first column of our array now looks like [1, 1, $\epsilon$, $3 - \frac{2}{\epsilon}$, 2]. To interpret this, we imagine what happens as our "nudge" $\epsilon$ becomes infinitesimally small (approaching zero from the positive side). The term $3 - \frac{2}{\epsilon}$ becomes huge and negative. The sequence of signs in the first column is $(+, +, +, -, +)$. We see two sign changes: from $\epsilon$ to the large negative number, and from the large negative number back to $2$. The verdict is clear: two unstable roots. The zero was not a roadblock, but a signpost pointing toward instability.

A Whole Row of Zeros

The second, more profound, special case is when an entire row becomes zero. This happens when the polynomial has a perfect symmetry in its roots, such as pairs on the imaginary axis ($\pm j\omega$) or pairs symmetric about the origin ($\pm \sigma$). These are systems on the very edge of stability, oscillating forever.

When a zero row appears, the Routh array is shouting that it has found a special factor within your polynomial. To find it, you look at the row just above the row of zeros. The coefficients in that row form what is called the ​​auxiliary polynomial​​, $A(s)$. This polynomial will always be "even" (containing only even powers of $s$), and its roots are precisely those symmetric roots that caused the zero row.

For example, for $p(s) = s^3 + s^2 + s + 1$, the Routh array begins:

The next row, $s^1$, would be all zeros. This signals a special case. The auxiliary polynomial, from the $s^2$ row, is $A(s) = 1s^2 + 1$. The roots of $A(s)=0$ are $s = \pm j$, which are roots of our original polynomial and lie on the imaginary axis. The system will oscillate.

To complete the array, we replace the zero row with the coefficients of the derivative of the auxiliary polynomial, $\frac{d}{ds}A(s) = 2s$. So, we use the coefficient '2' to fill in the $s^1$ row and continue. The number of sign changes from that point on tells you about the stability of the other roots.

The Routh array is more than a computational trick; it's a window into the deep structure of polynomials and the systems they describe.

For instance, the polynomial $p(s) = (s+1)^5 = s^5 + 5s^4 + 10s^3 + 10s^2 + 5s + 1$ has all its roots at $s=-1$, so it is profoundly stable. If you run its coefficients through the Routh array, you'll find that all the elements of the first column are positive, yielding zero sign changes, perfectly confirming what we know from its binomial structure.

This also helps us debunk a common myth. A necessary condition for stability is that all polynomial coefficients must be positive. But is it sufficient? No. The polynomial $s^3 + s^2 + 2s + 8$ has all positive coefficients, but its Routh array quickly reveals a sign change, signaling instability. You cannot take shortcuts.

Finally, the theory gives us beautiful, high-level insights. For any real polynomial of an odd degree, it must have at least one real root (because complex roots come in conjugate pairs). For the system to be stable, this real root must be negative. Why must a real root exist? A continuous function that goes from $-\infty$ to $+\infty$ (as any odd-degree polynomial does) must cross the horizontal axis at least once. The Routh criterion is the tool that ensures it crosses on the correct side.

In the end, the Routh array is a testament to the power of mathematical insight. It transforms a difficult, "brute-force" problem of finding roots into a simple, elegant bookkeeping exercise that not only gives the answer but also reveals the underlying character of the system itself.

Now that we have acquainted ourselves with the machinery of the Routh-Hurwitz criterion, we might be tempted to view it as a clever, if somewhat mechanical, mathematical procedure. But to do so would be to miss the forest for the trees. This criterion is not merely a calculation; it is a lens through which we can perceive a fundamental property of the world: stability. Its true power is revealed not in the rows of its array, but in the vast array of questions it allows us to answer, from designing the humble thermostat on your wall to probing the dynamics of physical systems.

Imagine you are an engineer designing the cruise control for a car. Your goal is to create a system that smoothly maintains the desired speed. You introduce a feedback mechanism: a sensor measures the car's actual speed, compares it to the target speed, and tells the engine to accelerate or decelerate accordingly. The "aggressiveness" of this response is governed by a parameter, a gain we might call $K$. If $K$ is too low, the car will be sluggish, taking ages to reach the set speed. If $K$ is too high, the system might overreact, causing the car to lurch back and forth violently. Worse still, a poorly chosen $K$ could make the oscillations grow larger and larger until the system becomes completely unstable.

How do you find the "Goldilocks" range for $K$? You could build a prototype and test it, a costly and potentially dangerous affair. Or, you could use the Routh-Hurwitz criterion. By writing down the mathematical model of the car's engine, the sensors, and the controller, you arrive at a characteristic polynomial where the coefficients depend on your chosen gain $K$. For a typical third-order system, this polynomial might look something like $s^3 + (a+b)s^2 + ab s + K = 0$.

Without building a single component, you can construct the Routh array for this polynomial. The criterion for stability—that all entries in the first column must be positive—translates directly into a set of inequalities for $K$. In this case, it might tell you with unerring certainty that the system is stable if and only if $0 \lt K \lt ab(a+b)$. This is a remarkable result. The abstract algebra has given us a precise, practical design blueprint. It provides a definitive "safe operating area" for the system's parameters. This is the bread and butter of control engineering: using mathematics to predict and guarantee stability before a system is ever built.

But the criterion can do more than just bless a good design; it can also condemn a bad one. Consider the challenge of controlling a "triple integrator" plant, whose transfer function is $G(s) = 1/s^3$. This might model, for instance, the position of an object in space under a constant thrust. An engineer might try to stabilize this with a standard Proportional-Integral (PI) controller. When we write down the characteristic polynomial for this combination, we get something of the form $p(s) = s^4 + K_p s + K_i$.

A seasoned eye notices something is wrong even before building the Routh array: the $s^3$ and $s^2$ terms are missing! The Routh-Hurwitz criterion confirms our suspicion in a rigorous way. The presence of zero coefficients immediately leads to a violation of the stability conditions. The conclusion is stark and absolute: no possible choice of positive controller gains $K_p$ and $K_i$ can make this system stable. The Routh analysis hasn't just told us we have the wrong numbers; it has proven that we are using the wrong approach. This power to rule out entire design strategies is invaluable, saving countless hours of fruitless simulation and testing.

Nature is often more subtle than a simple stable/unstable dichotomy. Sometimes, a system can exist on the very knife's edge of stability, oscillating indefinitely in a state of "marginal stability." The Routh-Hurwitz criterion has a special way of alerting us to this fascinating behavior.

When an entire row of the Routh array becomes zero, the test doesn't simply fail. It signals that the characteristic polynomial has roots that are symmetric about the origin of the complex plane. This often means there is a pair of poles sitting directly on the imaginary axis, of the form $s = \pm j\omega$. These poles correspond to a sustained oscillation at a frequency $\omega$, neither growing nor decaying.

The Routh criterion even gives us a tool to investigate. The "auxiliary polynomial," formed from the row just above the row of zeros, contains these symmetric roots. By solving the auxiliary polynomial, we can find the exact frequency of oscillation. This is how we can predict, for instance, the precise gain $K$ at which a system will begin to hum or vibrate, and what the frequency of that vibration will be.

Sometimes the analysis reveals surprising truths. For a particular system with an open-loop zero in the right-half plane—a feature known to cause control difficulties—the Routh analysis might show that for any positive gain $K$, the system is always unstable. It might only achieve marginal stability for a specific negative gain, a scenario often outside the bounds of a typical design. This kind of profound, non-intuitive insight is where the Routh criterion truly shines.

The concept of stability is not exclusive to the world of engineering. It is a cornerstone of physics, chemistry, and biology. A pendulum has a stable equilibrium hanging downwards and an unstable one balanced perfectly upright. A star is a stable fusion reactor, balanced between gravitational collapse and explosive thermal pressure. The beauty of the Routh-Hurwitz criterion is that it applies wherever this question of stability can be described by linear differential equations.

Consider a physical system described by a set of equations in classical mechanics, for example, the motion of a particle influenced by forces that depend on its position. We can analyze the stability of its equilibrium points (where all forces balance and the particle is at rest). The analysis begins by linearizing the equations of motion around that point, resulting in a matrix that describes the system's local behavior. The stability of the equilibrium is determined by the eigenvalues of this matrix, which are found by solving its characteristic polynomial.

And there it is again—a polynomial in $s$ (or $\lambda$, as physicists often prefer). The problem is transformed. Is the physical equilibrium stable? The question becomes: do all roots of this polynomial have negative real parts? The physicist can now borrow the engineer's tool. By applying the Routh-Hurwitz criterion to this polynomial, they can determine the conditions under which the equilibrium is stable, just as the engineer determined the stability of their cruise control. The mathematics is identical. The physical context is different, but the underlying principle—the signature of stability in the coefficients of a polynomial—is universal.

For all its power, it is crucial to understand what the Routh-Hurwitz criterion tells us and what it does not. It provides a definitive, binary answer to the question of absolute stability: Is the system stable? Yes or no. It draws a sharp line in the sand, separating the stable region of parameters from the unstable one.

However, in the real world, just being on the "safe" side of the line is often not enough. We also need to know how far we are from the edge of the cliff. This is the concept of relative stability. A system that is stable, but very close to the boundary of instability, is fragile. Small changes in its components or environment could easily push it over the edge.

The Routh criterion does not directly measure this robustness. Other tools, like the Bode plot or Nyquist plot, are used for this purpose. For a specific, stable value of the gain $K$, a Bode plot can reveal the system's "gain margin" and "phase margin." These are measures of how much additional gain or phase lag the system can tolerate before it becomes unstable.

So, we have a beautiful division of labor. The Routh-Hurwitz criterion is the mathematician's tool for providing an absolute, algebraic map of the entire stability landscape. The Bode plot is the practicing engineer's tool for surveying a specific point within that stable landscape to assess its robustness. Together, they form a powerful combination, allowing us to not only design systems that are stable but to design systems that are reliably and robustly stable in the face of an uncertain world.