Iterated integral order

1 Definition and basic setup

1.1 Iterated integrals for functions on rectangles

An iterated integral expresses a multiple integral by integrating one coordinate at a time. For a function \(f(x,y)\) defined on a rectangular region \(R=[a,b]\times[c,d]\), a double integral can be written in either order as \[ \int_{c}^{d}\left(\int_{a}^{b} f(x,y)\,dx\right)dy \quad\text{or}\quad \int_{a}^{b}\left(\int_{c}^{d} f(x,y)\,dy\right)dx, \] provided the inner integrals are well-defined and the hypotheses needed for interchange are satisfied. The core question in the “iterated integral order” topic is when these two iterated expressions agree with the same underlying double integral.

For higher multiplicity, the same idea applies: one selects an order (e.g., \(dx\,dy\,dz\)) and integrates successively, treating the remaining variables as parameters during each step.

1.2 Rewriting a multiple integral as an iterated integral

A common starting point is a multiple integral defined over a region with coordinates separated in a convenient way. On rectangles, the conversion between a multiple integral and an iterated integral is straightforward under standard assumptions (such as integrability). On more general regions, one first rewrites the domain in a form that allows one variable to be integrated between explicit bounds depending on the others, producing an iterated integral representation.

This process can be done in more than one way, and different representations correspond to different orders of integration.

1.3 Notation for changing the integration order

Changing order is typically indicated by swapping which variable appears as the outer integral variable. In double integrals this looks like \[ \int \!\!\int_R f(x,y)\,dA \;=\; \int_{c}^{d}\left(\int_{a}^{b} f(x,y)\,dx\right)dy \;=\; \int_{a}^{b}\left(\int_{c}^{d} f(x,y)\,dy\right)dx, \] when the interchange is justified.

In the symbolic level, one also often changes the placement of \(dA\) (area element) versus \(dx\,dy\) or \(dy\,dx\). The mathematics is notational: the real issue is whether the resulting iterated integrals actually converge to the same value and match the intended multiple integral.

1.4 Domains defined by inequalities versus parametrizations

A region \(D\subset \mathbb{R}^2\) (or \(\mathbb{R}^n\)) may be described using inequalities (e.g., \(a\le x\le b\), \(g_1(x)\le y\le g_2(x)\)) or via parametrization. Inequality descriptions are particularly useful for order-swapping because they directly produce limits for inner integrals.

Parametrizations (including polar, cylindrical, or spherical coordinates) can simplify integrands or geometry, but the order of integration then includes the order in which the new coordinates are integrated. Even when the domain is simple in one coordinate system, it may become more intricate in another, affecting how bounds must be written.

2 Order of integration in double integrals

2.1 Typical forms: \(dx\,dy\) versus \(dy\,dx\)

For a region \(D\) where one can write \[ D=\{(x,y): a\le x\le b,\; \phi_1(x)\le y\le \phi_2(x)\}, \] the order \(dy\,dx\) yields \[ \int_{a}^{b}\left(\int_{\phi_1(x)}^{\phi_2(x)} f(x,y)\,dy\right)dx. \] If the same region can also be described as \[ D=\{(x,y): c\le y\le d,\; \psi_1(y)\le x\le \psi_2(y)\}, \] then the order \(dx\,dy\) corresponds to \[ \int_{c}^{d}\left(\int_{\psi_1(y)}^{\psi_2(y)} f(x,y)\,dx\right)dy. \] The “order” issue arises when the interchange is attempted between these two iterated formulas: they can agree, or they can differ if convergence or integrability hypotheses fail.

2.2 Geometric interpretation of order

Geometrically, choosing an order corresponds to choosing families of “slices” parallel to coordinate axes. The inner integral averages \(f\) along one coordinate direction (e.g., fixing \(y\) and integrating over \(x\)), and the outer integral aggregates these slice values over the remaining coordinate.

When swapping order, the slice orientation changes: the computation may still describe the same physical region, but the way contributions are grouped can affect the meaning if the integral is improper or not absolutely integrable.

2.3 Region projection onto coordinate axes

Practical order changes often rely on projections. If projecting \(D\) onto the \(x\)-axis yields an interval \([a,b]\), then one can often express \(y\)-bounds as functions of \(x\). Alternatively, if projecting onto the \(y\)-axis yields \([c,d]\), one obtains \(x\)-bounds as functions of \(y\).

A key diagnostic is whether the region “looks like” a vertical strip (for \(dy\,dx\)) or a horizontal strip (for \(dx\,dy\))—more precisely, whether the boundary can be expressed as single-valued functions of the outer variable over the projected interval.

2.4 Boundary and limit selection for each order

Selecting limits requires careful attention to the boundary curves or surfaces that determine the region. In double integrals, one must ensure:

• The inner variable ranges exactly between the lower and upper boundary for each fixed outer variable.
• The outer variable ranges over the full projection interval of the region onto its axis.
• The bounds are consistent with any piecewise structure of the region (e.g., when the top boundary changes formula at a particular \(x\)-value).

When the region is non-rectangular but still regular (e.g., defined by inequalities that remain functions), both orders are often possible. When the region is more complex, one order may require splitting into subregions to maintain correct bounds.

2.1.1 Fubini-type conditions for equality

Interchanging integration order is governed by Fubini-type results: under suitable conditions, the multiple integral equals both iterated integrals.

2.1.1.1 Absolute integrability and measurable functions

A common sufficient hypothesis is absolute integrability: \[

\int\!\!\int_D	f(x,y)	\,dA < \infty.

\] Under standard measurability assumptions, this implies both iterated integrals exist and are equal, and they coincide with the value of the double integral. Intuitively, absolute integrability prevents cancellations from hiding divergence, ensuring that the order in which one groups contributions does not alter the outcome.

2.1.1.2 Integrability without absolute values (conditional cases)

If \(\int\!\!\int_D	f	\) diverges, equality may fail. In some settings, conditional convergence can still allow an interchange, but additional structure must be verified. Often, one must check that the iterated integrals converge and that the limiting process defining the multiple integral agrees with the iterated limits.

Examples of conditional behavior demonstrate that two iterated limits can each exist yet not match the intended multiple integral, or match each other but not be stable under variations of truncation. The theory therefore distinguishes between the existence of each iterated integral and the existence of the corresponding multiple integral.

2.1.1.3 When equalities can fail (pathologies)

Order interchange can fail in several ways:

• One iterated integral diverges while the other converges.
• Both iterated integrals converge but to different values.
• Each iterated integral converges, yet the double integral defined via symmetric truncation or via measure-theoretic integration differs (or the multiple integral is undefined).

Such failures are typically linked to singularities (where \(f\) becomes unbounded), non-absolute convergence, or boundary-related improper behavior.

3 Order of integration in higher dimensions

3.1 Triple integrals and permutation of orders

For a triple integral over a region \(D\subset\mathbb{R}^3\), there are six possible orders corresponding to permutations of \((x,y,z)\). For example, \[ \int\!\!\int\!\!\int_D f(x,y,z)\,dV \] may be expressed as an iterated integral with one of the orders \(dx\,dy\,dz\), \(dx\,dz\,dy\), \(dy\,dx\,dz\), and so on.

Swapping order is conceptually the same as in two dimensions but becomes more delicate because each swap may introduce or remove dependence in bounds on the newly outer variables. Moreover, singularities can occur along lines or surfaces, affecting the validity of interchange.

3.2 Systematic enumeration of integration orders

When bounds are available in multiple forms, one can systematically enumerate all orders compatible with correct slicing:

1. Identify a description of the region suitable for one variable being inner and another outer.
2. Derive bounds from that description.
3. Attempt alternative descriptions by projecting onto the axes corresponding to the intended outer variables.
4. If the region is not representable by single-valued boundary functions for the chosen projection, split into subregions and apply the construction piecewise.

This approach ensures that each iterated integral matches exactly the same geometric region, even when the order changes.

3.3 Symmetry considerations and repeated application

Symmetry of the region or of the integrand can reduce the number of computations. If \(f\) and \(D\) are invariant under swapping coordinates, many orders yield identical expressions. Even without full symmetry, repeated use of legitimate interchange steps can simplify evaluation: for example, swapping \(x\) and \(y\) first and later swapping \(y\) and \(z\), provided each interchange is justified under the relevant integrability conditions.

3.4 Extension to \(n\)-fold integrals

The general principle extends to \(n\)-fold integrals: there are \(n!\) iterated orders. Under appropriate integrability hypotheses on \(f\) over \(D\), iterated integrals in any order coincide with the same value of the \(n\)-fold integral.

Without such hypotheses, the order dependence can intensify. Improper integrals and singularities may interact with the geometry of the region in ways that make some orders convergent while others fail.

4 Theoretical results governing order changes

4.1 Fubini’s theorem (interchanging order)

Fubini’s theorem provides the central justification for swapping integration order. In its standard form, if \(f\) is integrable on a product domain (or more generally on a measurable set with appropriate structure), then:

• the iterated integrals exist (almost everywhere in the relevant sense),
• the iterated integral equals the multiple integral,
• and switching the order does not change the value.

The theorem’s hypotheses typically involve measurability and integrability (often expressed via absolute integrability), ensuring that both the inner and outer integrals can be interpreted consistently.

4.2 Tonelli’s theorem (nonnegative integrands)

Tonelli’s theorem handles nonnegative functions \(f\ge 0\). For such integrands, one can form iterated integrals without requiring absolute integrability in advance; the iterated expressions may evaluate to \(+\infty\), and interchange is still valid: \[ \int\!\!\int_D f\,dA = \int \left(\int f\,dx\right)dy = \int \left(\int f\,dy\right)dx, \] with the understanding that all quantities live in \([0,\infty]\).

Tonelli’s framework is particularly useful for proving inequalities and for studying improper integrals where monotone convergence can be applied.

4.3 Improper integrals and additional hypotheses

Improper integrals arise when the domain is unbounded or when \(f\) has singularities inside the region. One common construction truncates the region (e.g., excluding neighborhoods of singular points or restricting to bounded sets) and then takes a limit. Interchange of order must be compatible with this limiting process.

Thus, additional hypotheses are often required beyond mere pointwise definition:

• absolute integrability on truncated sets plus control of the limit,
• dominance by an integrable function,
• or conditions that ensure the convergence is independent of the order of truncation.

Without such control, two different regularizations can produce different answers even when each iterated integral is computed in a formally similar way.

4.4 Measure-theoretic framing of iterated integration

A measure-theoretic perspective clarifies why pathologies occur and what theorems actually guarantee. In that framing:

• A multiple integral is defined as an integral with respect to product measure.
• Iterated integrals correspond to integrating a conditional slice function (integrate along one coordinate, then integrate the resulting slice values).

Fubini’s theorem asserts that, for integrable functions, the slice-wise integral can be integrated again to recover the product-measure integral, and this is independent of the chosen slicing order. When the function is not integrable (or only conditionally integrable), the slice integrals can behave differently, explaining order sensitivity.

5 Techniques for computing iterated integrals in different orders

5.1 Changing order by re-describing the region

Changing order typically begins with a re-description of the region so that the new outer variable has straightforward bounds. One identifies which projection onto an axis produces the simplest interval and then solves for the other variable using the boundary equations.

This method works well when the boundary curves can be solved explicitly and when for each outer-variable value there is a unique inner interval.

5.2 Splitting regions into subregions

Some regions cannot be expressed globally with simple single-valued bounds for a given order. In such cases, the region is partitioned into pieces where the desired inequality representation holds. After computing iterated integrals on each subregion, the results are added.

This technique is especially common for regions bounded by curves that intersect or for shapes like polygons or regions under piecewise-defined functions.

5.3 Handling piecewise-defined integrands

If the integrand changes formula across a curve (for example, \(f\) equals one expression in one part of the region and another expression elsewhere), swapping order still requires care. The iterated integral must reflect the same piecewise structure in the new coordinates.

Often, the boundary in the integrand’s definition aligns with boundaries of the region slices; if not, additional splitting may be required to keep the inner integral formula consistent across the slices.

5.4 Using symmetry and coordinate transformations

Symmetry can reduce computational effort when evaluating iterated integrals in different orders. If the integrand is even or odd in a variable and the region is symmetric with respect to that variable, certain slice integrals vanish or pair up.

Coordinate transformations also change what “order” means: after a transformation, the iterated integrals are with respect to new variables, and the domain must be re-expressed in those variables. When the Jacobian determinant is included correctly, transformations can turn an order-exchange problem into a simpler computation.

6 Examples and instructive computations

6.1 Rectangular domains with smooth integrands

For \(f(x,y)\) continuous on a rectangle \( [a,b]\times[c,d]\), both iterated orders are valid and equal. A typical computation proceeds by:

1. integrating \(f\) with respect to the inner variable,
2. then integrating the result with respect to the outer variable.

Doing the same in the opposite order yields the same value, illustrating the general principle under standard regularity.

Smoothness removes singularities and ensures absolute integrability on bounded sets, so order interchange is safe.

6.2 Non-rectangular regions (triangles, annular sectors, etc.)

Consider a triangular region described by inequalities such as \(0\le x\le 1\) and \(0\le y\le 1-x\). Integrating in \(dy\,dx\) produces bounds that depend on \(x\). Swapping to \(dx\,dy\) requires rewriting the triangle as \(0\le y\le 1\) and \(0\le x\le 1-y\).

More complex shapes, such as regions involving annular sectors in polar coordinates, illustrate that order dependence is usually avoided only after the region is properly expressed. If one tries to use inappropriate global bounds, the iterated integrals may cover the wrong geometric set.

6.3 Singular integrands and integrability checks

Suppose \(f(x,y)=\frac{1}{\sqrt{x^2+y^2}}\) on a region containing the origin. The integrand is singular at a point, so integrability must be checked. On many planar regions, such singularities may still yield a finite integral due to how the area element behaves near the singularity. In other cases, the divergence is strong enough that the integral becomes improper.

Determining whether order interchange is permitted is then tied to whether the relevant integrability conditions hold (often via comparison tests or estimating \(	f	\) near singularities), not merely to formal manipulation of bounds.

6.4 Comparing numerical outcomes across orders

In computation-oriented settings, one may approximate iterated integrals by numerical quadrature in different orders. When hypotheses hold (e.g., absolute integrability), the approximations converge toward the same value as resolution increases. When the integral is improper or only conditionally convergent, numerical values can drift with truncation strategy, and the apparent “agreement” may be accidental.

Such comparisons are often used as a diagnostic: if two orders produce stable matching approximations under refinement, that suggests well-behaved integrability; if not, it signals a need to revisit convergence assumptions.

7 Common pitfalls and diagnostic criteria

7.1 Mistakes in limits when switching order

A frequent error is incorrect bounds after swapping order—especially when the region requires splitting or when boundary functions must be solved carefully for the new inner variable. A practical diagnostic is to draw the region and verify that the chosen inequality description matches the same set of points for every outer-variable value.

If the bounds “look plausible” but the slices omit or double-count portions of the region, the resulting iterated integral will not represent the intended multiple integral.

7.2 Confusing absolute and conditional convergence

Another pitfall is assuming that because each iterated integral converges, order interchange is automatically valid. Conditional convergence is subtle: cancellations may depend on how the truncation is performed. A diagnostic step is to check whether \(\int\!\!\int	f	\) is finite or to apply the appropriate theorem’s hypotheses rather than relying on informal convergence judgments.

7.3 Singularities at boundaries and how they affect interchange

Singular behavior near the boundary can be as problematic as interior singularities. For example, a function may be integrable in some slice directions but not in others if the singular set aligns with the integration paths induced by the order. Diagnostic tools include:

• estimating the integrand’s growth near the singular region,
• using comparison with known integrable singularities,
• and verifying that the resulting truncated integrals converge in a stable way.

7.4 Misapplication of theorems outside their conditions

Errors occur when a theorem is invoked without meeting its assumptions (e.g., assuming Fubini-style interchange for a function that is not integrable). A diagnostic criterion is to identify the theorem’s requirement—measurability, integrability, nonnegativity, or control of improper limits—and then check it against the function and domain at hand.

When in doubt, it is safer to compute using the order that makes the region and integrand most regular, while explicitly checking convergence for that order, and then re-evaluating interchange assumptions.

8 Connections and further topics

8.1 Relations to changing variables in multiple integration

Changing variables and changing integration order are related but distinct operations. A change of variables typically transforms both the integrand and the region via a Jacobian, after which one may again swap orders if the new iterated representations are compatible. The order-swap theorems apply in the transformed setting as well, but only when the transformed integrand meets the relevant integrability conditions.

8.2 Links to distributions and generalized function viewpoints

In generalized function theory, one sometimes manipulates integrals involving objects like delta distributions. Order issues can reappear because distributional products and regularizations require careful definitions. While the present topic focuses on classical integrals, the broader connection highlights that “interchange” is not purely syntactic; it depends on how the integral is defined.

8.3 Iterated integrals in probability (expectations as iterated integrals)

In probability, the expectation of a joint function \(g(X,Y)\) can be written as a double integral. Under appropriate integrability conditions (finite expectation), one can often interchange the order of integration and compute via conditional expectations. This makes the order-of-integration discussion directly relevant to modeling: different computational routes correspond to different conditional averaging procedures.

8.4 Applications to solving differential equations via integral representations

Integral representations of solutions to differential equations (such as via Green’s functions or convolution formulas) often lead to expressions where order of integration must be swapped during derivations. Justification relies on interchange results to ensure that moving integrals past derivatives or rearranging multiple integrals does not invalidate the formula. In these applications, integrability and decay conditions play the role of ensuring that order changes are legitimate.