Motivation
A traveling wave on the whole line can continue indefinitely. A string of finite length cannot: its ends impose additional information. At a fixed endpoint, incident and reflected displacements must cancel. Reflection therefore changes the sign and mirrors the spatial profile of a pulse, while keeping its propagation speed.
Normal modes describe the same motion differently. Each spatial sine profile oscillates at its own frequency, and the initial velocity determines a coefficient divided by that frequency. Writing the resulting infinite sum is easy. Proving that its second derivatives converge, that it attains the data, and that it agrees with the reflected solution are separate tasks.
We derive reflection and mode convergence under explicit regularity assumptions for homogeneous fixed endpoints. Moving supports and general forced boundary problems require additional analysis.
Definitions
Fix a length and wave speed . The fixed-string problem is
Our classical class is on each closed strip , using one-sided derivatives at its boundary. Thus the derivative traces at time zero are part of the regularity statement. Boundary displacement zero implies boundary velocity and acceleration zero. The PDE then forces the corresponding initial compatibility conditions.
Definition 4.2.1
(Odd periodic continuation and mode coefficients)
For a function on vanishing at both endpoints, define its odd continuation on by for positive and for negative , then repeat it with period . In particular it is odd about both endpoints. Write and .
For , set
Here is the sine coefficient of velocity; its contribution to displacement will be , not .
Theorem/Proposition
Theorem 4.2.2
(Reflection constructs the classical fixed-end motion)
Suppose , , and
Then the unique solution in the stated classical class is
It extends to all real times and has the common time period . This need not be its least positive period.
Theorem 4.2.3
(One sufficient theorem for a twice differentiable mode sum)
Suppose more specifically and , with the same endpoint conditions. Then
The series and all partial derivatives of total order at most two converge absolutely and uniformly on every closed strip. Its sum is the reflected solution above. These are sufficient hypotheses, not a claim that less regular data have no meaningful solutions.
Theorem 4.2.4
(Energy is conserved, including the omitted mode energy)
For a classical fixed-end solution, define
Then . This proves uniqueness and energy stability for differences of classical solutions. Under the mode theorem, if is the sum through mode , then
Each identity holds at every time. In particular the truncation energy is under these hypotheses, uniformly in time.
Proof sketch or proof idea
Proof
(Odd reflection, boundary cancellation and recurrence)
At a reflection point, an odd continuation matches function values only if the original value is zero. Its first derivatives match automatically, while its second derivatives have opposite one-sided signs. Thus the endpoint conditions on make globally . For to be , zero endpoint values of suffice; no zero condition on is needed. Periodic joins obey the same checks. The whole-line d'Alembert calculation now gives a wave with the prescribed initial displacement and velocity.
At zero, the two values of cancel, and the integral of the odd over a symmetric interval is zero. Oddness about gives the same cancellation at the other endpoint. This proves the boundary conditions. Since is odd and periodic, its mean over a period is zero. Its primitive is therefore -periodic. Both shifted values and the difference of shifted values are unchanged when time increases by , proving recurrence.
Concept lensGeometric
A reflection folds a straight characteristic back into the string
Continue a characteristic straight through a fictitious neighboring copy of the interval. Folding that copy back into the physical string reverses direction and mirrors the spatial profile. The odd continuation also reverses displacement sign, so incident and reflected values cancel at the fixed end. An asymmetric pulse therefore needs both transformations, not just a translated negative copy.
Proof
(Coefficient decay, termwise differentiation and Fourier uniqueness)
For the mode theorem, integration by parts uses the endpoint conditions to give
Consequently and . Notice that the second formula keeps its endpoint term: our assumptions do not require to vanish there. A spatial or temporal derivative introduces at most one factor proportional to . Since , every term differentiated at most twice is bounded by a constant times . The summable bound proves uniform convergence, including the derivative series. Applying the fundamental theorem of calculus to partial sums and their uniformly convergent derivatives justifies termwise differentiation. Each term solves the PDE and vanishes at the endpoints.
We still must identify the initial sums with and . Absolute convergence makes the sine sums continuous and permits computing their Fourier coefficients term by term. Their odd periodic extensions have exactly the coefficients of and , by sine orthogonality. To see why equal coefficients identify continuous periodic functions, rescale the period to and use the finite kernel
At multiples of the period its value is the continuous limiting value . Expanding the finite square shows its integral is and that convolution with it is a finite weighted sum of Fourier coefficients. The kernel is nonnegative; outside an angular neighborhood of zero, it is bounded by a constant divided by . Uniform continuity and a near/far split therefore prove that its normalized convolution tends uniformly to every continuous periodic function. If all coefficients of a continuous difference are zero, every such convolution is zero, hence the difference is zero. This proves the needed Fourier uniqueness without presuming convergence of arbitrary ordinary Fourier sums.
Thus the initial displacement and velocity series are exactly their data. At time zero, differentiating the velocity contribution gives in each mode. Uniform convergence of derivatives ensures their correct traces. The mode sum belongs to the classical class and agrees with the reflection construction once uniqueness is proved.
Proof X-ray
Where the extra regularity is spent in the mode theorem
The hypotheses are used in the two coefficient formulas. Four integrations by parts move four derivatives onto : the cosine boundary terms vanish because and vanish at both ends, and the sine boundary terms vanish because . Hence . For only is assumed, so the third integration leaves and is only . The key move is the frequency divisor: the displacement coefficient is , and two derivatives cost at most , leaving the summable majorant . Had the velocity coefficients entered undivided, the same count would give only . Convergence alone yields a classical solution with unidentified traces; the argument closes when Fourier uniqueness matches those traces with and , and energy uniqueness matches the sum with the reflected solution.
Proof
(Energy identity, uniqueness and the truncation bound)
Differentiate the energy, substitute the equation and integrate by parts:
The boundary term vanishes because the endpoints are fixed for all time, so there. For a difference with zero initial data, zero energy forces both spatial and temporal derivatives to vanish; its zero endpoint value then forces the difference itself to vanish. This proof requires the derivative traces in the stated class, rather than just continuity of displacement.
For a finite mode sum, orthogonality of sines and cosines gives the displayed energy formula; each scalar oscillator has constant squared velocity plus squared frequency times displacement. Uniform convergence of the first derivative series passes finite-sum energies to the limit. The same argument applies to the tail, proving the exact truncation identity. The summands are , whose tail is . Also, since a zero-endpoint difference satisfies , Cauchy–Schwarz yields
Thus mode truncation gives a uniform displacement bound for all time as well. No damping factor was used; the omitted energy remains present rather than decaying away.
Worked examples
Worked example 4.2.5
(Displacement and velocity have different mode coefficients)
Take and . The solution is
At time zero, differentiating the second term supplies precisely the velocity coefficient. Omitting its frequency divisor changes the prescribed velocity. Both modes recur after the common period, while their individual periods can be shorter.
Worked example 4.2.6
(A reflected pulse has reversed sign and a mirrored profile)
Let with , and set , . A convenient expression is
The sum is locally finite: each support has width less than , and the translates within each family repeat every . At initial time only the first term with zero index contributes inside the string, giving the stated data. Reindexing at each endpoint proves cancellation, and each summand is a traveling wave, so uniqueness verifies the expression. Before the pulse reaches the right end it moves to the right unchanged. When , only the reflected pulse remains inside and it equals , moving left. Its speed is unchanged; its displacement is sign-reversed and its spatial profile is mirrored. Only for a symmetric pulse can the result also be described as a translated negative copy without the mirroring qualification.
Worked example 4.2.7
(Small height need not mean small energy)
For an integer , take and . The displacement stays bounded by , but
independent of the frequency index. The derivative of the short wavelength compensates for the small amplitude. This does not contradict energy stability: the energy norm measures velocity and slope, not just the largest displacement.
Counterexample mode
An even reflection enforces the wrong endpoint condition
Start with and zero velocity, but reflect the initial profile evenly at zero. For , the resulting whole-line formula has endpoint value . It fails the fixed-end boundary condition, although the interior equation is satisfied. Even reflection instead encodes a zero endpoint derivative; the sign of an image is determined by the boundary condition.
The sign of the image is not the only requirement: the corners must also be compatible.
Counterexample mode
Zero endpoint displacement is not every classical compatibility condition
Take and . Both displacement endpoints vanish, but . A solution in the stated closed-strip class would have from its fixed endpoint and from the PDE and continuous derivative traces. This is impossible. Reflections or series can still describe a less regular motion, but they cannot create the missing corner regularity by assertion.
Comparison: a free string and the missing zero mode
A different endpoint condition changes both the reflection and the spectrum. Consider with . In the small-slope string model these conditions prescribe zero transverse endpoint force. They do not require the endpoint displacement to vanish. We keep , and take real data , with
Extend and evenly about each endpoint, with period , calling the extensions and . The zero first derivatives make and : at a join, the second derivatives of the even extension of agree automatically. The whole-line formula
is consequently and has initial data . Evenness about each endpoint makes even about that endpoint for each time, hence there. This constructs a classical solution; unlike odd reflection, even reflection preserves the displacement sign.
To see the modes, solve with . Integration by parts gives for a nonzero eigenfunction. At , the endpoint conditions leave the constant function. For , they give scalar multiples of with , . Thus finite-mode data with cosine coefficients produce
where
The constant spatial mode obeys , so its contribution is ; dividing by its zero frequency is meaningless. For general data the same expression is a formal infinite series unless convergence and its derivatives are separately justified. The reflection construction above already proves existence under (1), without claiming such a convergence theorem.
The energy is conserved because . For two solutions with the same initial displacement and velocity, their difference has zero energy. Both derivatives therefore vanish; the difference is a constant in space and time, and its zero initial displacement makes that constant zero. Integrating the PDE in space also gives
Thus the mean displacement can move at constant speed even while energy is conserved. In particular the full solution need not be periodic.
Worked example 4.2.8
(A drifting mean with conserved energy)
For and , (3) gives
Its spatial derivative vanishes at both ends and its initial velocity is . The mean is ; the nonconstant mode oscillates. The zero mode carries kinetic energy , and the oscillating mode carries , so their sum stays fixed.
Quick check
(With and for the free string, find the motion and its energy. Does zero energy force the displacement to be zero?)
Recall what the zero mode contributes.
Solution · Answer
The solution is and . Zero energy forces zero velocity and slope, not zero displacement. The prescribed initial value selects the constant .
One fixed end and one free end
Suppose the left end is held at zero displacement while the right end is free of transverse force. In the same small-slope model, this means
The right endpoint can move; its slope, not its displacement, must vanish. Neither the fixed–fixed sine frequencies nor the free–free cosine frequencies automatically satisfy both conditions. We find the allowed modes and use them for finite-mode data, so no infinite sum has to be differentiated.
Write a nonzero separated solution as and choose the separation constant so that . The boundary conditions are and . Multiplying the eigenvalue equation by and integrating gives
because . Thus is real and nonnegative. If , then ; the conditions force . For , the left condition leaves and the right condition requires . Consequently
There is no zero-frequency mode: a nonzero constant is allowed at the free end but forbidden at the fixed end. The half-integer frequencies make the slope zero at , while the sine makes displacement zero at .
If and , the unique classical solution is
The product-to-sum identity gives for and for . Thus the coefficients of finite-mode data are and . Direct differentiation verifies the PDE, both boundary conditions and both initial conditions in (6).
For uniqueness, let be the difference of two real solutions on a closed time strip with identical data, and define
The flux at vanishes because differentiating gives ; that at vanishes because . The initial difference has zero energy, so . The fixed boundary then gives . This proves uniqueness in the stated class; energy conservation alone does not construct a solution for arbitrary data.
Worked example 4.2.9
(Two mixed modes, one with initial velocity)
Take and . Equation (6) gives
At the first bracket equals and its derivative is ; the second bracket equals and its derivative is . Both spatial modes vanish at and have zero derivative at . The energy is constant, and evaluating at gives
since .
The mixed sine functions are complete in . To see this, assume the standard sine system is complete in . Extend by reflection about : set for and for . The th sine mode on changes by under , so the even modes have zero inner product with and each odd mode has twice the corresponding inner product with . If is orthogonal to all mixed sine modes, then is orthogonal to the entire standard sine system and hence is zero in ; therefore in . This proves completeness. Completeness permits an expansion of general data, but it does not by itself justify differentiating the wave series twice or attaining boundary data pointwise.
Quick check
(For the fixed–free string, let and for a constant . Find the motion and its energy.)
Only one oscillator is excited, and it starts from zero displacement.
Solution · Answer
Here and , with all other coefficients zero. Therefore . At , and throughout the interval. Thus , which remains constant.
Common mistakes
Common mistake
(Treating an infinite sum of modes as a classical solution)
An infinite sum of solutions need not be twice differentiable. In the wave equation, time does not supply the exponentially decaying high-frequency factor available for heat flow. The coefficient and derivative estimates therefore do real work. Our explicit sufficient theorem does not settle optimal regularity, energy-space existence or every nonsmooth plucked-string profile.
Common mistake
(Reading too much into periods and signs)
Do not confuse a common recurrence period with the least period of a particular motion: a single mode already recurs after . Nor is a reflected displacement sign a sign of energy dissipation.
Common mistake
(Using energy conservation as an existence proof)
Energy conservation proves uniqueness only after boundary cancellation and the regularity needed for integration by parts have been checked. It is not itself the construction of a solution.
Summary
Odd periodic reflection constructs the classical fixed-end motion under precise corner conditions. With stronger regularity, integration by parts gives coefficient decay sufficient for a twice differentiable sine expansion. Fourier uniqueness identifies the data, while conserved energy identifies the solution and quantifies the exact energy omitted by truncation. Reflections and modes are two verified descriptions of the same undamped motion.
Exercises
Quick check
(With zero displacement and initial velocity , find the motion, its energy and its least positive period when .)
The coefficient of displacement contains the reciprocal angular frequency.
Solution
The solution is , with . For nonzero , its least positive time period is . The full-string common period still works, but is twice as large for this particular mode.
Quick check
(For initial displacement and zero velocity, verify the reflection compatibility, calculate the initial energy and give an all-time displacement bound.)
Differentiate the polynomial before applying the energy estimate.
Solution
The cubic factors make both and vanish at both endpoints; all required regularity holds. Since ,
The bound uses conserved energy and need not be the exact maximum height. It is valid at every time without computing any individual Fourier coefficient.
Quick check
(Suppose and . Give an explicit tail-energy estimate in terms of , , , and , then bound the displacement error.)
Compare the tail of the sixth-power series with an integral.
Solution
For , the decreasing sixth-power function gives . Hence
The error bound is uniform in time. It describes a convergent classical approximation under the stated coefficient bounds; it does not say that discarded modes physically lose energy.
Quick check
(Under the mode theorem, prove . Explain why a motion containing only even-indexed modes has the shorter common period .)
Compare the sine profile at and , and shift each oscillator by half the full recurrence time.
Solution
Shifting time by multiplies the oscillator in mode by , while . Thus
Uniform convergence permits these operations on the sum. For even indices the time multiplier is one, so is already a period. A nonzero first displacement mode with zero velocity changes sign at that time, showing why the shorter period is not universal.