The Central Idea
Suppose \(x\) has a characteristic scale \(X_c\).
Define a new variable
\[ x^*=\frac{x}{X_c}. \]Because \(x\) and \(X_c\) have the same units,
\[ [x^*]=1. \]The variable \(x^*\) is dimensionless.
Instead of asking for the absolute value of \(x\), we ask how large \(x\) is relative to its natural scale.
What Does Nondimensional Mean?
A quantity is dimensionless when its units cancel completely.
For example, if both \(x\) and \(L\) are lengths, then
\[ \frac{x}{L} \]has dimension
\[ \frac{L}{L}=1. \]The resulting quantity is a pure number.
Why Nondimensionalize?
Nondimensionalization can:
- reduce the number of parameters;
- make typical variables order one;
- reveal important ratios between competing effects;
- improve comparison between systems of different size;
- clarify which parameters actually govern behavior;
- support approximation and asymptotic reasoning;
- improve numerical conditioning;
- make model structure easier to interpret.
Start with Characteristic Scales
Nondimensionalization begins by choosing characteristic scales.
If position \(x\) has scale \(L_c\) and time \(t\) has scale \(T_c\), define
\[ x^*=\frac{x}{L_c} \]and
\[ t^*=\frac{t}{T_c}. \]These variables measure position and time relative to the natural scales of the model.
Rewrite the Original Variables
From
\[ x^*=\frac{x}{L_c}, \]we can write
\[ x=L_cx^*. \]Similarly,
\[ t=T_ct^*. \]These substitution forms are what we place into the original model.
A Simple Algebraic Example
Consider
\[ y=ax. \]Choose characteristic scales \(X_c\) and \(Y_c\), and define
\[ x=X_cx^* \]and
\[ y=Y_cy^*. \]Substitute:
\[ Y_cy^*=aX_cx^*. \]Divide by \(Y_c\):
\[ y^*=\frac{aX_c}{Y_c}x^*. \]A Dimensionless Parameter Appears
In
\[ y^*=\frac{aX_c}{Y_c}x^*, \]the quantity
\[ \alpha=\frac{aX_c}{Y_c} \]is dimensionless.
The original dimensional parameter \(a\) has been converted into a pure number that measures its effect relative to the chosen scales.
Choosing Scales Can Simplify the Equation
If we choose
\[ Y_c=aX_c, \]then
\[ \frac{aX_c}{Y_c}=1. \]The model becomes
\[ \boxed{y^*=x^*}. \]A suitable scale choice can remove parameters entirely.
Nondimensionalizing a Linear Model
Consider
\[ y=mx+b. \]Set
\[ x=X_cx^* \]and
\[ y=Y_cy^*. \]Then
\[ Y_cy^*=mX_cx^*+b. \]Divide by \(Y_c\):
\[ y^* = \frac{mX_c}{Y_c}x^* + \frac{b}{Y_c}. \]The two dimensionless parameters are
\[ \alpha=\frac{mX_c}{Y_c} \]and
\[ \beta=\frac{b}{Y_c}. \]Dimensionless Variables Preserve Relative Meaning
Suppose
\[ L_c=100\text{ m}. \]If
\[ x=20\text{ m}, \]then
\[ x^*=\frac{20}{100}=0.2. \]If
\[ x=250\text{ m}, \]then
\[ x^*=2.5. \]The dimensionless variable immediately reports the size relative to the system scale.
Dimensionless Time
If \(T_c\) is a characteristic time, define
\[ t^*=\frac{t}{T_c}. \]Then:
- \(t^*=0.1\) is early relative to the characteristic time;
- \(t^*=1\) is one characteristic time;
- \(t^*=10\) is long relative to that time scale.
This interpretation is independent of whether time is measured in seconds, hours, or years.
Exponential Decay Example
Consider
\[ y(t)=y_0e^{-kt}. \]A natural time scale is
\[ T_c=\frac{1}{k}. \]Define
\[ t^*=\frac{t}{T_c}=kt \]and
\[ y^*=\frac{y}{y_0}. \]Then
\[ \boxed{y^*=e^{-t^*}}. \]The original parameters disappear.
What the Exponential Example Reveals
Models such as
\[ y=10e^{-2t} \]and
\[ y=500e^{-0.01t} \]appear very different in dimensional form.
But after scaling output by \(y_0\) and time by \(1/k\), both become
\[ y^*=e^{-t^*}. \]Nondimensionalization reveals that they share the same underlying shape.
Nondimensionalizing a Differential Equation
Consider
\[ \frac{dy}{dt}=-ky. \]Let
\[ y=Y_cy^* \]and
\[ t=T_ct^*. \]Then
\[ \frac{dy}{dt} = \frac{Y_c}{T_c} \frac{dy^*}{dt^*}. \]Substituting into the Decay Equation
Substitute into
\[ \frac{dy}{dt}=-ky: \] \[ \frac{Y_c}{T_c} \frac{dy^*}{dt^*} = -kY_cy^*. \]Cancel \(Y_c\):
\[ \frac{1}{T_c} \frac{dy^*}{dt^*} = -ky^*. \]Multiply by \(T_c\):
\[ \frac{dy^*}{dt^*} = -kT_cy^*. \]Choose the Natural Time Scale
If
\[ T_c=\frac{1}{k}, \]then
\[ kT_c=1. \]The equation reduces to
\[ \boxed{ \frac{dy^*}{dt^*}=-y^* }. \]The dimensional parameter \(k\) has disappeared from the governing equation.
How Derivatives Transform
If
\[ x=L_cx^* \]and
\[ t=T_ct^*, \]then
\[ \frac{dx}{dt} = \frac{L_c}{T_c} \frac{dx^*}{dt^*}. \]The factor
\[ \frac{L_c}{T_c} \]is the natural speed scale.
Second Derivatives
Differentiating again gives
\[ \frac{d^2x}{dt^2} = \frac{L_c}{T_c^2} \frac{d^2x^*}{dt^{*2}}. \]The prefactor
\[ \frac{L_c}{T_c^2} \]is the characteristic acceleration scale.
This pattern is useful throughout dynamic modeling.
Nondimensionalizing Logistic Growth
Consider
\[ \frac{dP}{dt} = rP\left(1-\frac{P}{K}\right). \]Choose
\[ P_c=K \]and
\[ T_c=\frac{1}{r}. \]Define
\[ p=\frac{P}{K} \]and
\[ \tau=rt. \]Substitute into the Logistic Equation
Since
\[ P=Kp \]and
\[ t=\frac{\tau}{r}, \]we have
\[ \frac{dP}{dt} = Kr\frac{dp}{d\tau}. \]Substituting gives
\[ Kr\frac{dp}{d\tau} = rKp(1-p). \]Cancel \(Kr\):
\[ \boxed{ \frac{dp}{d\tau}=p(1-p) }. \]Why the Logistic Result Matters
The original model contains two parameters:
\[ r \]and
\[ K. \]After nondimensionalization, neither appears explicitly:
\[ \frac{dp}{d\tau}=p(1-p). \]The entire family shares one dimensionless governing equation.
Dimensionless Groups
Nondimensionalization often combines several dimensional parameters into a smaller number of dimensionless groups.
For example, suppose
\[ \Pi=\frac{aL}{b}. \]If the dimensions cancel so that
\[ [\Pi]=1, \]then \(\Pi\) is a dimensionless group.
Such groups often control model regimes more directly than the original dimensional parameters do.
Dimensionless Parameters Compare Effects
Suppose a model contains two competing rates,
\[ r_1 \]and
\[ r_2. \]The ratio
\[ \lambda=\frac{r_1}{r_2} \]is dimensionless when both rates have the same units.
Then:
- \(\lambda\ll1\): the first process is much slower;
- \(\lambda\sim1\): the processes occur on comparable scales;
- \(\lambda\gg1\): the first process is much faster.
Scale Ratios Can Reveal Small Parameters
Suppose
\[ \varepsilon=\frac{h}{L}, \]where both \(h\) and \(L\) are lengths.
If
\[ h\ll L, \]then
\[ \varepsilon\ll1. \]The small dimensionless parameter \(\varepsilon\) may justify simplifying terms multiplied by \(\varepsilon\) or higher powers of it.
Order-One Variables
A useful nondimensionalization often produces variables of order one.
Suppose typical \(x\)-values are around
\[ 5000\text{ m}. \]If
\[ X_c=5000\text{ m}, \]then
\[ x^*=\frac{x}{X_c} \]will normally be near \(1\).
This makes scale comparisons more transparent.
Nondimensionalization Is More Than Unit Conversion
Unit conversion changes the unit used to express the same dimensional quantity.
For example,
\[ 1\text{ m}=100\text{ cm}. \]Nondimensionalization instead forms a ratio such as
\[ x^*=\frac{x}{L_c}, \]so that no physical unit remains.
Normalization and Nondimensionalization
The terms normalization and nondimensionalization are related but not always identical.
Normalization broadly means rescaling a quantity to a convenient reference range or magnitude.
Nondimensionalization specifically removes physical dimensions by dividing quantities by compatible scales.
Many nondimensionalizations are also normalizations.
Standardization as a Dimensionless Transformation
In statistics,
\[ z=\frac{x-\mu}{\sigma} \]is dimensionless when \(x\), \(\mu\), and \(\sigma\) have the same units.
The numerator has the same dimension as \(x\), and division by \(\sigma\) removes that dimension.
The resulting \(z\)-score measures deviation relative to a characteristic spread.
Nondimensionalization in Geometry
Suppose a rectangle has length \(L\) and width \(W\).
Instead of treating both dimensions independently, define the aspect ratio
\[ \alpha=\frac{W}{L}. \]The ratio is dimensionless.
Two rectangles of different absolute sizes but the same \(\alpha\) have the same shape.
Nondimensionalization in Finance
Suppose a firm's monthly expense \(E\) is compared with monthly revenue \(R\).
The ratio
\[ \frac{E}{R} \]is dimensionless when both are measured in the same currency.
This ratio expresses expense relative to revenue and can be compared across firms of very different absolute size.
Nondimensionalization in Population Models
If \(P\) is a population and \(K\) a carrying capacity, then
\[ p=\frac{P}{K} \]expresses population as a fraction of capacity.
The interpretation is immediate:
- \(p=0.2\): population is at 20% of capacity;
- \(p=1\): population equals the modeled capacity;
- \(p>1\): population exceeds the nominal capacity.
Nondimensionalization in Transport
If a trip has characteristic distance \(L_c\) and characteristic time \(T_c\), then
\[ x^*=\frac{x}{L_c} \]and
\[ t^*=\frac{t}{T_c} \]allow progress to be compared across routes of different physical length and duration.
Nondimensionalization Supports Similarity
Two systems can be very different in absolute units but equivalent in dimensionless form.
If their dimensionless variables and governing parameter groups match, they may exhibit similar behavior after rescaling.
This is one reason scaled experiments can provide information about larger systems.
Parameter Reduction
Suppose a model originally contains five dimensional parameters.
After nondimensionalization, those parameters may appear only through two independent dimensionless combinations.
Then the effective parameter space is reduced from five quantities to two.
This can make analysis, simulation, and experimentation much more efficient.
Model Regimes
Dimensionless groups often classify model behavior.
Suppose
\[ \lambda=\frac{\text{effect A}}{\text{effect B}}. \]Then:
- \(\lambda\ll1\): effect B dominates;
- \(\lambda\sim1\): both effects matter;
- \(\lambda\gg1\): effect A dominates.
This viewpoint is often more informative than the individual parameter values.
Nondimensionalization and Approximation
Suppose a dimensionless equation contains
\[ y^*=1+\varepsilon x^* \]with
\[ \varepsilon=0.001. \]If \(x^*\) remains order one, then
\[ \varepsilon x^* \]is small relative to the leading term \(1\).
This makes the size of the correction explicit.
Nondimensionalization and Numerical Stability
Numerical methods often behave better when variables have comparable magnitudes.
If one variable is around
\[ 10^{12} \]and another around
\[ 10^{-8}, \]direct computation can be awkward.
Scaling both variables near \(1\) can improve interpretation and sometimes improve numerical conditioning.
Nondimensionalization Does Not Remove Meaning
Dimensionless variables may appear abstract, but their interpretation is usually direct.
If
\[ x^*=0.4, \]then \(x\) is 40% of the chosen characteristic scale.
If
\[ t^*=3, \]then three characteristic times have elapsed.
The dimensional meaning can always be recovered by rescaling.
Recovering the Original Variables
If
\[ x^*=\frac{x}{X_c}, \]then
\[ \boxed{x=X_cx^*}. \]If
\[ t^*=\frac{t}{T_c}, \]then
\[ \boxed{t=T_ct^*}. \]Nondimensionalization is reversible when the scales are known.
Common Error: Dividing by an Incompatible Scale
If \(x\) is a length, then the denominator in
\[ x^*=\frac{x}{X_c} \]must also be a length.
Dividing a length by a time produces speed, not a dimensionless variable.
Common Error: Forgetting to Transform Derivatives
If
\[ x=L_cx^* \]and
\[ t=T_ct^*, \]then
\[ \frac{dx}{dt} \neq \frac{dx^*}{dt^*}. \]The correct transformation is
\[ \boxed{ \frac{dx}{dt} = \frac{L_c}{T_c} \frac{dx^*}{dt^*} }. \]Common Error: Scaling Only One Side of an Equation
Every occurrence of a dimensional variable must be replaced consistently.
If
\[ x=L_cx^*, \]then all appearances of \(x\) in the model must use that substitution.
Partial substitution can create inconsistent equations.
Common Error: Choosing Scales That Leave Huge Numbers
A mathematically valid scale is not always a useful one.
If typical values of \(x\) are around
\[ 10^6, \]choosing
\[ X_c=1 \]leaves
\[ x^*\sim10^6. \]A scale near the actual system magnitude would usually be more informative.
Common Error: Assuming the Dimensionless Form Is Unique
Different valid scale choices can produce different dimensionless forms.
One choice may emphasize initial conditions, another equilibrium behavior, and another numerical convenience.
Nondimensionalization is therefore partly a modeling decision.
A Reliable Nondimensionalization Procedure
- Write the dimensional model clearly.
- Identify the units and dimensions of all variables and parameters.
- Choose characteristic scales for the main variables.
- Define dimensionless variables such as \(x^*=x/X_c\).
- Rewrite each original variable in terms of its scale and dimensionless variable.
- Transform derivatives and integrals carefully when present.
- Substitute into the original equation.
- Divide through by a convenient common scale.
- Identify remaining dimensionless parameter groups.
- Choose scales strategically when they can eliminate parameters.
- Check that every remaining term is dimensionless.
- Interpret the dimensionless variables and groups.
Example: Full Algebraic Workflow
Consider
\[ y=ax+b. \]Choose scales \(X_c\) and \(Y_c\):
\[ x=X_cx^*, \qquad y=Y_cy^*. \]Substitute:
\[ Y_cy^* = aX_cx^*+b. \]Divide by \(Y_c\):
\[ y^* = \frac{aX_c}{Y_c}x^* + \frac{b}{Y_c}. \]Define
\[ \alpha=\frac{aX_c}{Y_c}, \qquad \beta=\frac{b}{Y_c}. \]The dimensionless model is
\[ \boxed{y^*=\alpha x^*+\beta}. \]Example: Full Differential Workflow
Consider
\[ \frac{dx}{dt}=-kx. \]Choose
\[ X_c=x_0 \]and
\[ T_c=\frac{1}{k}. \]Define
\[ x=X_cx^*, \qquad t=T_ct^*. \]Then
\[ \frac{dx}{dt} = \frac{X_c}{T_c} \frac{dx^*}{dt^*}. \]Substitute:
\[ \frac{X_c}{T_c} \frac{dx^*}{dt^*} = -kX_cx^*. \]Cancel \(X_c\):
\[ \frac{dx^*}{dt^*} = -kT_cx^*. \]Since \(T_c=1/k\),
\[ \boxed{ \frac{dx^*}{dt^*}=-x^* }. \]Practice 1: Form a Dimensionless Length
If
\[ x=15\text{ m} \]and
\[ L_c=60\text{ m}, \]find
\[ x^*=\frac{x}{L_c}. \]Practice 2: Recover the Dimensional Variable
If
\[ x^*=2.5 \]and
\[ L_c=8\text{ m}, \]find \(x\).
Practice 3: Dimensionless Time
A characteristic time is \(4\text{ h}\). What is the dimensionless time corresponding to \(10\text{ h}\)?
Practice 4: Exponential Decay Scale
For
\[ y=y_0e^{-0.2t}, \]with \(t\) measured in days, find the natural characteristic time.
Practice 5: Dimensionless Exponential Model
Using
\[ y^*=\frac{y}{y_0} \]and
\[ t^*=kt, \]rewrite
\[ y=y_0e^{-kt}. \]Practice 6: Dimensionless Population
A population is \(30{,}000\) and carrying capacity is \(50{,}000\). Find
\[ p=\frac{P}{K}. \]Practice 7: Rate Ratio
Two processes have rates
\[ r_1=0.02\text{ s}^{-1} \]and
\[ r_2=0.5\text{ s}^{-1}. \]Find
\[ \lambda=\frac{r_1}{r_2} \]and interpret it.
Practice 8: Transform a Derivative
If
\[ x=L_cx^* \]and
\[ t=T_ct^*, \]write \(dx/dt\) in terms of \(dx^*/dt^*\).
Practice 9: Aspect Ratio
A plate has thickness \(2\text{ mm}\) and length \(500\text{ mm}\). Find the dimensionless thickness-to-length ratio.
Practice 10: Why Scale?
Typical values of a variable are around \(2\times10^6\). Explain why a characteristic scale of \(10^6\) is generally more useful than a scale of \(1\).
Practice Answers
Practice 1
\[ x^* = \frac{15}{60} = \boxed{0.25}. \]Practice 2
\[ x = L_cx^* = (8\text{ m})(2.5) = \boxed{20\text{ m}}. \]Practice 3
\[ t^* = \frac{10\text{ h}}{4\text{ h}} = \boxed{2.5}. \]Practice 4
\[ T_c = \frac{1}{0.2\text{ day}^{-1}} = \boxed{5\text{ days}}. \]Practice 5
Divide by \(y_0\) and use \(t^*=kt\):
\[ \boxed{y^*=e^{-t^*}}. \]Practice 6
\[ p = \frac{30{,}000}{50{,}000} = \boxed{0.6}. \]The population is 60% of carrying capacity.
Practice 7
\[ \lambda = \frac{0.02}{0.5} = \boxed{0.04}. \]The first rate is much smaller than the second, so the first process is much slower on the corresponding time scale.
Practice 8
\[ \boxed{ \frac{dx}{dt} = \frac{L_c}{T_c} \frac{dx^*}{dt^*} }. \]Practice 9
\[ \varepsilon = \frac{2}{500} = \boxed{0.004}. \]The thickness is much smaller than the length.
Practice 10
If
\[ X_c=10^6, \]then typical values satisfy
\[ x^*\sim2, \]which is order one. Using \(X_c=1\) would leave values around \(2\times10^6\), providing little simplification.
What to Remember
Nondimensionalization rewrites a model in terms of pure numerical ratios.
- choose characteristic scales for the important variables;
- form dimensionless variables such as \(x^*=x/X_c\);
- rewrite the original variables in terms of scales and dimensionless variables;
- transform derivatives carefully;
- substitute and divide through by convenient scales;
- identify dimensionless parameter groups;
- choose scales strategically to simplify the model;
- interpret dimensionless variables as quantities relative to natural scales;
- use small and large dimensionless groups to identify model regimes.
The central modeling habit is: replace absolute size with size relative to the system whenever that comparison reveals structure.
Where We Go Next
Nondimensionalization shows how characteristic scales can remove units and expose the ratios that govern a model. The next lesson, Scaling Laws, develops relationships that describe how one quantity changes when another changes in size, including power-law behavior, geometric scaling, proportionality, and the consequences of changing system scale.