Motivation
When a slope depends only on the ratio , enlarging both coordinates by the same factor leaves that slope unchanged. The natural unknown is therefore the ratio, not alone. But differentiating the ratio leaves a factor in the equation. A logarithmic independent variable removes that factor.
This turns a familiar substitution into a complete method. Fixed ratios give straight-line solutions; other ratios satisfy a separable autonomous equation. The transformed interval determines where the original graph exists, and its direction must be interpreted differently on the negative half-axis. The same coordinates also explain a useful integrating factor for homogeneous differential forms.
Definitions
Let the ratio belong to an open interval, and let the slope function be continuously differentiable there. For either sign define the domain
The derivative in (1) is with respect to . A solution is a continuously differentiable graph on a nonempty open interval of positive length, remaining in the indicated domain. The original equation is not defined at . Its slope is unchanged by simultaneous nonzero scaling of both coordinates; positive scaling preserves the chosen signed domain. This meaning of homogeneous does not assert linearity or superposition.
Definition 1.8.1
(Ratio and logarithmic independent variable)
Introduce
The logarithm measures multiplicative scale additively. It is an independent-variable change, not the original physical coordinate with a new label. On the negative half-axis its direction is reversed relative to increasing first coordinate.
For a separate implicit-form version, homogeneous coefficients of degree mean that simultaneous positive scaling multiplies both coefficients by the same power. The degree can be any real number. We will state the domain and nonzero denominator of the integrating-factor construction explicitly.
Theorem/Proposition
Theorem 1.8.2
(Exact equivalence with scalar autonomous flow)
For a fixed sign, transformation (2) gives a one-to-one correspondence between solution graphs of (1) on that signed half-axis and solutions in the ratio interval of
The corresponding intervals are related by the logarithm and signed exponential. Maximal solution intervals within the stated domains correspond under this change of variable. In particular, each zero of the transformed field gives a straight-line solution
Each such straight-line solution has the entire chosen half-axis as its maximal interval in this domain. The graph equation and transformed equation both have local uniqueness under the stated regularity. On the negative half-axis, increasing logarithmic variable means decreasing first coordinate, not increasing it.
Proof
(Differentiate the substitution)
Differentiate the substitution using the actual derivative of the logarithm:
Thus (1) is exactly (3), with no approximation or lost nonzero factor on the chosen half-axis. The same computation proves the converse. Both coordinate maps are smooth diffeomorphisms between their intervals, even though the logarithm reverses orientation on the negative side. An extension of either solution within its domain would pull back to an extension of the other, proving maximality correspondence. The transformed field is continuously differentiable; so is the original slope on its open domain. Ordinary local uniqueness therefore applies to both. Constant transformed solutions give (4).
Theorem 1.8.3
(Separation without losing straight-line solutions)
Choose initial data in the signed domain, and set their logarithmic coordinate and ratio as below. If that ratio is not a zero of the transformed field, let be the connected component of its nonzero set containing it. Define
This function is strictly monotone and has a continuously differentiable inverse on its open interval image. The complete nonconstant solution from that datum is
Its original maximal interval is the connected signed-exponential image of the interval specified in (6). In addition, every zero in (4) must be included. A nonconstant solution cannot reach an interior zero of the transformed field at a finite logarithmic time.
Proof
(Separation on one component of the nonzero set)
On the component, the continuous nonzero field has one sign. Hence the derivative of the integral in (5) is never zero, has one sign, and gives a strictly monotone inverse. Its image is an open interval containing zero. Differentiating the inverse gives (3), while separating any solution remaining in that component gives exactly (6). This proves both construction and uniqueness of the formula there.
Why is staying in the component legitimate? If a nonconstant trajectory reached an interior zero of the field at finite time, uniqueness through that point would make it coincide locally with the constant solution, and continuation of equality would contradict the initial nonzero field. The separation integral explains the same fact quantitatively. Near such a zero, continuous differentiability gives a positive finite bound of the form
The last integral diverges on approach to the zero. Thus a finite endpoint of the separated time interval cannot arise by reaching an interior equilibrium. The ratio is monotone on its component, so it has an extended-real limit at each endpoint. A finite time endpoint instead corresponds to reaching the boundary of the ratio domain or to an unbounded ratio. If an endpoint ratio were still inside the nonzero component, the inverse formula would continue past it; hence the constructed interval is maximal in the ratio domain.
Proof X-ray
Why the separated interval is the whole maximal interval
Two hypotheses carry the proof. That is continuous and nonzero on the component gives a single sign, hence a strictly monotone with a inverse: this is the construction. Continuous differentiability of is used separately, near an equilibrium at the edge of , to obtain ; the comparison with makes the time needed to reach infinite. That divergent integral is the key move: it is why the straight-line solutions removed by the division are never reached, and why they must be restored separately rather than appearing as endpoints. The argument closes with monotonicity: the ratio has a limit at each end of , and a limit inside would let the inverse formula continue, so every finite endpoint is a genuine boundary of the ratio domain or an unbounded ratio.
An infinite logarithmic endpoint corresponds to or ; a finite one corresponds to a finite nonzero . A finite limit of the graph at zero does not make it a solution of (1) through zero, because that point is excluded from the equation's domain.
Theorem 1.8.4
(An integrating factor for homogeneous forms)
On the positive-coordinate wedge assume two continuously differentiable coefficients are homogeneous of the same real degree:
Write their ratio-coordinate profiles and assume the radial contraction never vanishes:
Since , this is equivalent to the profile denominator being nonzero throughout the ratio interval. It also forces the two original coefficients not to vanish simultaneously, so the potential's level curves are regular.
Then, for any chosen base ratio in the interval, a nowhere-zero integrating factor and a potential for the multiplied form are
The potential is defined on the whole stated wedge, even though it need not be a rectangle in the original coordinates. If a larger original wedge includes a ratio where the radial contraction is zero, that excluded ratio gives a radial straight-line solution of the original implicit form; the factor formula is not defined on that line. Failure of the radial-factor condition is not a general nonexistence theorem for other factors.
Proof
(Logarithmic and ratio coordinates)
Use the logarithmic and ratio coordinates on this positive wedge. Homogeneity expresses the original coefficients through their profiles. Direct differentiation of the coordinate map gives
Division by (7) produces the differential of (8). The profile quotient is continuously differentiable on its interval, so the potential has two continuous derivatives. This explicit construction supplies global exactness on the wedge without invoking a rectangle theorem on a nonrectangular domain. If the radial contraction vanishes at a particular ratio, setting that ratio constant in the undivided form makes the coefficient of the radial tangent zero, proving the straight-line claim.
For the graph equation , take and . Then the radial contraction is . Formula (8) becomes the logarithm minus the separation integral. Thus the integrating factor and autonomous separation are two expressions of the same calculation on nonconstant-ratio regions.
Worked examples
Worked example 1.8.5
(A tangent formula and its maximal endpoints)
Consider the ratio-dependent equation with a positive quadratic slope:
There is no fixed ratio. Integrating the reciprocal quadratic and imposing the datum gives
The argument of the tangent must stay in its particular pole-free interval containing the initial argument. The maximal graph interval is
Both endpoints are finite and positive, and the graph diverges at them. Choosing a later tangent branch does not extend the same finite-valued solution through either pole. The original singular axis has not been reached at either endpoint.
For a nonconstant negative-side example, use the initial point . The same calculation gives
The signed exponential reverses the endpoint order. The graph still has finite blow-up at each endpoint; replacing this interval by a positive-side ordering would be wrong.
Worked example 1.8.6
(The straight lines discarded by separation)
For the quadratic ratio slope, the transformed field and its equilibria are
The two fixed ratios give the zero graph and the unit-slope line. On a nonzero component, partial fractions lead to the remaining family
The formula is used only on a connected interval avoiding a zero denominator. If , its maximal intervals are and ; they are different solution branches. For the denominator never vanishes on the positive half-axis. The value gives the unit-slope line. The zero graph must still be added separately. For the initial point the constant is and the maximal interval is . The graph tends to zero as , but the equation is undefined there. At the other endpoint it diverges. These are two different reasons why the interval ends.
Worked example 1.8.7
(A two-coordinate factor from scaling)
On the positive quadrant take the nonexact form
Its profiles are and . The potential in (8), after absorbing a constant, is
Checking both partial derivatives verifies the multiplied form. Its level curves can be labeled by a positive constant without logarithms:
For the datum the constant is . The graph blows up at and tends to as . This form is also separable: another factor depending on both coordinates gives
Its reciprocal-level relation gives the same curves. The example illustrates two compatible methods rather than asserting that the homogeneous factor is the only useful one.
Affine ratios: moving the center or following parallel lines
A slope such as is not a function of , yet it is built from the same ingredients: a quotient of two linear expressions. The classical method for such equations reduces them either to the ratio equation (1) or to a scalar autonomous equation. Which reduction applies is decided by a single determinant, and the reason is a symmetry of the slope field rather than an algebraic accident.
Throughout this section let be an open interval, , and consider
Its domain is the open set of points at which the denominator is nonzero and the quotient lies in . The right-hand side is continuously differentiable on , so (9) has local existence and uniqueness through every point of .
Why a determinant decides the method. If , each affine expression vanishes on a line and the lines meet in exactly one point , and both expressions become linear forms in . Their quotient is unchanged when is multiplied by a nonzero scalar, so the slope field is invariant under dilations centered at : this is the homogeneous situation, with the origin moved. If and the linear parts are not both zero, they are proportional, and both expressions are constant along every line perpendicular to a chosen nonzero coefficient row. The slope field is then invariant under translation along that direction, which is the symmetry of an autonomous equation.
Theorem 1.8.8
(Intersecting lines: translate, then use the ratio)
Suppose , and let be the unique solution of , . Put
Then solves (9) on an interval if and only if solves on . On each half-plane this translated equation is the ratio equation (1) with slope profile , on any open interval of ratios on which and . Consequently both earlier theorems apply there, with .
Proof
(Translate, then divide by )
The translation preserves derivatives, because and the independent variable is only shifted. Substituting , into the two linear expressions removes the constants exactly because lies on both lines; this is where is used. For divide numerator and denominator by to obtain . The map has derivative , so it is a continuously differentiable, strictly monotone map on each interval avoiding its pole. Hence is continuously differentiable on each open ratio interval described in the theorem, which is the regularity hypothesis of (1).
Proposition 1.8.9
(Solutions can cross the vertical line through the center)
In the setting of the intersecting-lines theorem suppose and . Then every point of the line other than the center lies in . If a solution of (9) on one side of has with as , it extends as a solution through into the other half-plane.
Proof
(Continuity of the slope at the line and local uniqueness)
The ratio substitution needs , so the earlier theorems describe solutions only inside the half-planes and . For the homogeneous equation (1) itself the vertical axis is genuinely excluded. For (9) it usually is not. At a point with and , if the quotient in (9) equals . If also , every such point belongs to : only the center itself, where the quotient has the form , is removed from that line.
In that case an endpoint of a ratio-coordinate solution is not an endpoint of the original solution whenever the graph tends to a finite value . Near the limit point the right-hand side of (9) is continuous and bounded, so is bounded and continuous up to ; the graph extends as a function to with , satisfying (9) there by continuity. Local existence and uniqueness through then continue the solution into the other half-plane, and uniqueness identifies the continuation with whatever ratio-coordinate solution occupies that side. For a solution with this finite nonzero limiting value, the maximal interval must therefore include pieces on both sides of . Only when , or , is the line excluded from (9) as well.
Theorem 1.8.10
(Parallel lines: follow the invariant direction)
Let be continuously differentiable on an open set , let be real with , and consider . Put . Then solves this equation on an interval , with , if and only if solves
on the same interval , with values in . Every zero of gives the straight-line solution , and the nonconstant solutions are obtained by separating (11) on a component of .
If in (9) and the two linear parts are not both zero, they are multiples of one nonzero row. When the right-hand side of (9) depends only on and is integrated directly. When , (9) is of this form with
defined on the open set of with and quotient in . If both linear parts vanish, the slope is constant on .
Proof
(Substitute the invariant combination)
The proof is one line in each direction. If solves the equation, then . Conversely, since , the function is continuously differentiable and . The independent variable is not changed, so the intervals of existence are literally the same; no artificial line is introduced, unlike the translated ratio method. Since (11) is autonomous, maximal solutions in one component of are translates of one another; other solutions in that component are restrictions of these translates, which is the translation symmetry of the slope field read in the coordinate . Why must be abandoned here? When there is no unique center. If both coefficient rows are nonzero, their zero lines can be distinct parallel lines or coincide; in the latter case a common zero can still remove both constants. A zero coefficient row instead gives an affine expression that is constant, with zero set either the whole plane or empty.
Worked example 1.8.11
(A logarithmic spiral that crosses the excluded ratio line)
Solve
Here , and the lines , meet at . With , , the equation becomes , so on
The datum is , , and the ratio interval containing on which is defined is . There , so there is no straight-line solution to restore in this interval. The separation integral (5) is
It increases from (as ) to (as ). Since and , the ratio theorem gives the interval . In polar coordinates , with , the relation becomes . The trajectory lies on the logarithmic spiral .
The right endpoint is genuine. As the graph reaches , which lies on the line where the denominator of (9) vanishes, and the slope tends to there.
The left endpoint is an artifact. There and , so the graph tends to , a point at which the original slope is . By the proposition on crossing the vertical line the solution continues into . To find the continuation, parametrize the spiral for every :
The last expression equals at the point of the spiral, so the curve satisfies (9) wherever it is a graph. The maximal open parameter interval containing on which is ; on it the curve is a graph. That interval contains , which is the initial point . By uniqueness on this curve is the solution, and its maximal interval is
At both ends the graph reaches the line with a finite value and an infinite slope ( on the right, on the left), so neither end can be extended within . Applying the ratio theorem blindly would have reported , missing the whole part of the solution with .
Worked example 1.8.12
(Parallel lines, an invariant line, and a global solution)
Solve
Here , and both linear parts are multiples of . With , equation (11) reads
The zero gives the straight-line solution . Direct substitution confirms this: its slope is , and the right-hand side equals . The initial value lies in the component of , where . Since , separation gives
is strictly increasing on , tends to as , and tends to as . Its inverse is therefore defined for every real , and the solution is on all of . Implicitly, . Differentiating this relation recovers the equation, and it holds at .
The component also explains the qualitative behavior. The solution never meets the denominator line , because it cannot cross the invariant line that separates them. As it approaches the invariant line, and as we have , so . The equilibrium is approached only as , consistent with the divergence of the separation integral near a simple zero established earlier in this note.
Common mistakes / subtle points
Common mistake
(Homogeneous in scale does not imply superposition)
For , the line is a solution. Adding that solution to itself gives . Its derivative is , whereas the right-hand side is . Thus the sum is not a solution. Scale homogeneity of a nonlinear first-order equation and homogeneity of a linear equation are different properties.
Counterexample mode
An attracting ratio need not give a vanishing graph
For a linear slope profile with coefficient strictly between zero and one,
The ratio tends to zero as , but every nonzero graph grows in absolute value. Attraction in the ratio variable describes relative distance from the zero-slope ray, not decay of . Returning from transformed coordinates requires restoring the factor .
Counterexample mode
An undefined radial factor can coexist with a simpler factor
For the form below the radial contraction is identically zero, yet another factor works on the positive half-plane:
Every fixed ratio is a solution in the stated positive half-plane. The autonomous equation has zero field everywhere, so there is no nonconstant-ratio region on which to divide for separation. On this domain the complete family is the straight-line branches with arbitrary finite slope. The excluded vertical axis is not part of this half-plane problem.
These failures have a common source: the ratio and logarithmic coordinates replace a whole ray by one value of , and a scale by one value of .
Concept lensStructural
Rays become equilibrium states
The ratio is a direction label and the logarithm is a scale coordinate. A straight ray becomes one equilibrium of the ratio equation. Motion between rays becomes ordinary scalar flow, and the inverse integral decides how much multiplicative scale that motion consumes. This also explains why a factor singular on an invariant ray can lose a genuine solution if that ray is not restored.
Practice and solutions
Quick check
(Reduce the equation below on each signed half-axis, solve it, and explain how letting increase toward on the negative half-axis compares with increasing logarithmic time.)
Use the same ratio and logarithm on both halves; keep the orientation sign visible.
Solution · Answer
The transformed equation is linear, with solutions
The arbitrary constant absorbs the fixed sign and includes zero. Its value can be chosen independently on the two half-axes. On the negative half-axis, sends to , not . A phase-line statement for increasing logarithmic time must therefore be reversed when discussing that original-coordinate direction. The formula solves the equation separately on either half-axis; the original quotient equation remains undefined at zero.
Quick check
(Find the explicit solution and maximal positive interval for the positive-quadratic ratio equation with the initial data below.)
Determine the initial tangent argument and the two neighboring poles.
Solution · Answer
The initial ratio is , so . Hence
At the initial point the tangent term gives ratio one, as required. The adjacent poles give exactly these endpoints. Both are positive finite blow-up locations, so no finite-valued extension through them is possible.
Quick check
(Find the radial factor and a potential for the homogeneous form below wherever the radial denominator is nonzero, on the half-plane . Recover the straight lines excluded by that factor directly from the original form.)
The radial contraction is a quadratic. Compare its differential with the original form.
Solution · Answer
The quadratic and its differential give
Use this factor on each connected slope interval where the quadratic is nonzero. Its zero-ratio lines are
Substituting their constant slopes into the original form gives zero, so both are genuine implicit solution curves excluded by the factor. The unmultiplied form is already exact, with potential half the quadratic; the new logarithmic potential only relabels its nonzero levels. It must not be used to discard the zero level.
This is an implicit-form argument. At ratio one the second coefficient vanishes, so dividing to form a slope would leave the regular graph framework. The potential still describes a regular curve there, with a vertical tangent rather than a differentiable graph .
Quick check
(Explain why finite-time arrival at a fixed ratio was excluded by the smoothness assumption. For the continuous transformed field below, construct a solution starting at a positive ratio that reaches zero in finite logarithmic time, then remains zero. Translate its hitting location to a value of .)
Separate only before the hitting time and check differentiability at the join.
Solution · Answer
The positive branch and its continuation are
The one-sided transformed derivatives at the join are both zero, so this is a continuously differentiable solution. Returning by (2) gives a graph joining the zero ray at the finite positive location shown. A locally Lipschitz field would make the reciprocal travel integral diverge by the estimate in the theorem. Here the positive-side travel integral is finite, and the slope profile is not continuously differentiable:
The hitting solution and the constant solution coincide after the hitting time but have different pasts. Thus backward uniqueness fails there. The transformed inward field is decreasing, so forward uniqueness still holds: the derivative of the squared difference of two transformed solutions is nonpositive, which forces solutions with the same initial value to agree for future time. The smooth theorem uses uniqueness through a point in both time directions, a stronger property absent here.
Quick check
(Solve the equation below with the given initial value and find its maximal interval. Which reduction for affine ratios applies?)
Identify the linear combination on which the right-hand side depends.
Solution · Answer
The right-hand side depends only on , so the parallel-line theorem applies with and . Equation (11) is , whose field has no zeros, so there is no straight-line solution. The initial value is , and separation gives . Because has image ,
Both endpoints are finite blow-up points, since there. Continuing past them with another branch of the tangent would give a different solution, not an extension of this one.
Quick check
(Reduce the equation below by translation, solve it, and decide whether solutions can cross the vertical line through the center. Compare with the spiral example.)
Locate the center first, then test whether the original equation is defined on the vertical line through it.
Solution · Answer
Here , and the center is . With and the equation becomes , so and . The fixed ratio gives the line . The other solutions are . Absorbing the sign into the constant on each half-plane gives
and direct substitution verifies the equation. Now , so the denominator of the original equation vanishes on the whole line . Every solution has maximal interval or , and every one of them tends to the center as . A graph through the center built from different constants on the two sides is not a solution, because the equation is undefined at . In the spiral example , so the line minus the center belonged to the domain. Here it does not, and that is the whole difference.