Chapter map

1. 1. Equation
2. 2. Basis one
3. 3. Basis two
4. 4. Trial field
5. 5. Left boundary
6. 6. Right boundary
7. 7. Interior value
8. 8. Interior residual
9. 9. Boundary residual
10. 10. Forcing
11. 11. Conditioning
12. 12. Conclusion

Visit the How Mathematics Works hub for the wider series and its connections to education, reasoning, science, computing, design and careers.

---

1. Equation

This step deserves a careful pause. Set u second+u=0.

A concrete situation

One-dimensional Helmholtz equation.

What the mathematics does not license

Homogeneous problem. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Record PDE. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

2. Basis one

The useful question is not only how to calculate, but what the calculation assumes. Use cos x.

A concrete situation

Residual exactly zero.

What the mathematics does not license

Differentiate twice. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Record check. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

3. Basis two

A small example makes the mechanism visible. Use sin x.

A concrete situation

Residual exactly zero.

What the mathematics does not license

Differentiate twice. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Record check. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

4. Trial field

This is where modelling choices begin to affect evidence. Write a cos x+b sin x.

A concrete situation

Every combination is Trefftz.

What the mathematics does not license

State interval. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Record field. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

5. Left boundary

The central mathematical idea is concrete. Set u(0)=1.

A concrete situation

a=1.

What the mathematics does not license

Use exact values. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Record equation. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

6. Right boundary

Here the geometry becomes operational. Set u(pi/2)=2.

A concrete situation

b=2.

What the mathematics does not license

Use exact values. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Record equation. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

7. Interior value

This step deserves a careful pause. Evaluate at pi/4.

A concrete situation

3/sqrt(2) about 2.1213.

What the mathematics does not license

Show arithmetic. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Record value. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

8. Interior residual

The useful question is not only how to calculate, but what the calculation assumes. Differentiate twice and add u.

A concrete situation

Zero exactly.

What the mathematics does not license

Analytic teaching field. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Record residual. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

9. Boundary residual

A small example makes the mechanism visible. Re-evaluate both endpoints.

A concrete situation

Both zero.

What the mathematics does not license

Separate from PDE residual. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Record values. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

10. Forcing

This is where modelling choices begin to affect evidence. Add a particular solution when needed.

A concrete situation

Homogeneous basis alone is insufficient.

What the mathematics does not license

State source term. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Record decomposition. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

11. Conditioning

The central mathematical idea is concrete. Add wave directions carefully.

A concrete situation

Redundancy can grow.

What the mathematics does not license

Inspect singular values. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Record metric. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

12. Conclusion

Here the geometry becomes operational. State basis, boundaries and validation.

A concrete situation

Exact for this simple case.

What the mathematics does not license

Complex geometry differs. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Write claim. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

A seven-day practice lab

1. Day 1: define the decision, observation unit, X variables and Y or reference condition.
2. Day 2: reproduce the smallest worked example and check every dimension, unit and sign.
3. Day 3: build a transparent baseline and write what it assumes.
4. Day 4: fit the main method inside a training-only pipeline; record every parameter.
5. Day 5: challenge one assumption with a deliberately difficult dataset or operating regime.
6. Day 6: compare validation metrics with the practical workload created by errors.
7. Day 7: present one supported conclusion, one unresolved uncertainty and one safe next experiment.

Questions for a student, parent or mentor

• Which part of the model is a mathematical definition, and which part is a choice?
• What small calculation would catch a sign, scale or dimension error?
• What evidence comes from training, validation and the untouched test?
• What benign condition could look like a fault or prediction failure?
• How would the answer change for another response, asset, school project or time period?
• What should a human review before acting on the score?

A method-specific audit before deployment

A Trefftz method uses trial functions that satisfy the governing differential equation exactly within each region or element. The unknown coefficients are then chosen to fit boundary conditions, interfaces or a variational statement.

Consider the one-dimensional Helmholtz equation u double prime + u=0 on the interval from 0 to pi/2. The functions cos x and sin x each satisfy this differential equation exactly.

Write the trial field u_h(x)=a cos x+b sin x. Imposing u(0)=1 gives a=1 because cos 0=1 and sin 0=0.

Imposing u(pi/2)=2 gives b=2 because cos(pi/2)=0 and sin(pi/2)=1. The resulting field is u_h=cos x+2 sin x.

At x=pi/4, sin and cos both equal 1/sqrt(2), so u_h=(1+2)/sqrt(2)=3/sqrt(2), approximately 2.1213.

Differentiating twice returns -cos x-2 sin x. Adding u_h gives zero everywhere, so the interior Helmholtz residual is exactly zero for this teaching problem.

Did You Know? Trefftz functions move approximation effort away from the volume equation and towards boundaries and interfaces. This can be especially attractive for wave problems where plane waves already encode oscillation directions.

A Trefftz space must match the operator. Harmonic polynomials suit Laplace problems, plane waves suit constant-coefficient Helmholtz regions, and elasticity needs displacement or stress fields satisfying the relevant equilibrium equations.

Exact satisfaction of a homogeneous PDE does not mean exact satisfaction of boundary data. Too few basis functions, poor directions or difficult geometry can leave significant boundary and interface residuals.

For a problem with volume forcing, one may add a particular solution or decompose the field into particular and homogeneous parts. Ignoring the forcing would solve a different equation.

Hybrid-Trefftz finite elements use internal Trefftz fields and separate boundary frames. Their matrices arise from variational coupling, so element boundaries and inter-element continuity remain central.

Wave bases become challenging at high frequency because direction selection, numerical integration and conditioning interact. Adding many nearly redundant plane waves can make the coefficient system unstable.

Corners, re-entrant regions and material jumps can create local behaviour that a smooth global Trefftz basis represents poorly. Enrichment or domain partitioning may be needed.

Students can verify each basis function by direct differentiation, solve two boundary equations and measure both interior and boundary residuals. This separates what is exact by construction from what is only fitted.

Applications include acoustics, electromagnetics, structural vibration, potential problems and wave scattering. The method is a family of formulations rather than one universal matrix recipe.

A responsible Trefftz report states the PDE, homogeneous and particular parts, Trefftz basis, region partition, boundary or interface formulation, quadrature, conditioning, solver, refinement and validation.

---

Frequently asked questions

What is a Trefftz method?

A method whose trial functions satisfy the governing homogeneous PDE inside each region.

What remains to be approximated?

Boundary conditions, interfaces, forcing decompositions and coefficients.

Are all Trefftz methods meshless?

No. Trefftz ideas can be used globally, by boundary collocation or inside finite elements.

Why are plane waves useful?

They satisfy constant-coefficient wave equations and encode oscillatory directions.

Does zero interior residual prove the whole solution?

No. Boundary, interface, forcing and approximation errors remain.

What must be reported?

PDE, basis, particular solution, regions, coupling, quadrature, conditioning, convergence and validation.

---

Primary source and fact-check note

A. Zieliński, “Many names of the Trefftz method,” Engineering Analysis with Boundary Elements 96, 2018. Definitions and core claims in this article were checked against this primary publication. Calculations shown here were independently recomputed. Examples are educational illustrations; they are not reported industrial performance results.

Useful next reading

• Finite element method
• Method of fundamental solutions source
• RBF collocation
• Computational mathematics hub
• Study and learning methods

Keep the options open. Mathematics education supports clearer questions, transferable problem-solving skills and informed choices across science, engineering, computing, AI, statistics, design and many other careers. It does not guarantee intelligence, income, admission or employment; those outcomes depend on far more than one subject or model.

Method audit: mechanism, algebra and evidence

Basis audit

Verify every trial function satisfies the stated homogeneous PDE.

Boundary audit

Measure boundary and interface residuals separately.

Forcing audit

State the particular solution or decomposition for nonzero sources.

Wave audit

Check directions, quadrature, conditioning and frequency refinement.

Worked evidence notebook: reconstruct every number

Evidence notebook 1: Equation

Set u second+u=0. One-dimensional Helmholtz equation. Keep coordinates, units, matrix entries, signs, basis order and rounding visible. Homogeneous problem. Record PDE. Change one input, recompute the affected mechanism and identify whether the change comes from geometry, basis, boundary data, conditioning, integration, time stepping or physical modelling.

Evidence notebook 2: Basis one

Use cos x. Residual exactly zero. Keep coordinates, units, matrix entries, signs, basis order and rounding visible. Differentiate twice. Record check. Change one input, recompute the affected mechanism and identify whether the change comes from geometry, basis, boundary data, conditioning, integration, time stepping or physical modelling.

Evidence notebook 3: Basis two

Use sin x. Residual exactly zero. Keep coordinates, units, matrix entries, signs, basis order and rounding visible. Differentiate twice. Record check. Change one input, recompute the affected mechanism and identify whether the change comes from geometry, basis, boundary data, conditioning, integration, time stepping or physical modelling.

Evidence notebook 4: Trial field

Write a cos x+b sin x. Every combination is Trefftz. Keep coordinates, units, matrix entries, signs, basis order and rounding visible. State interval. Record field. Change one input, recompute the affected mechanism and identify whether the change comes from geometry, basis, boundary data, conditioning, integration, time stepping or physical modelling.

Evidence notebook 5: Left boundary

Set u(0)=1. a=1. Keep coordinates, units, matrix entries, signs, basis order and rounding visible. Use exact values. Record equation. Change one input, recompute the affected mechanism and identify whether the change comes from geometry, basis, boundary data, conditioning, integration, time stepping or physical modelling.

Evidence notebook 6: Right boundary

Set u(pi/2)=2. b=2. Keep coordinates, units, matrix entries, signs, basis order and rounding visible. Use exact values. Record equation. Change one input, recompute the affected mechanism and identify whether the change comes from geometry, basis, boundary data, conditioning, integration, time stepping or physical modelling.

Evidence notebook 7: Interior value

Evaluate at pi/4. 3/sqrt(2) about 2.1213. Keep coordinates, units, matrix entries, signs, basis order and rounding visible. Show arithmetic. Record value. Change one input, recompute the affected mechanism and identify whether the change comes from geometry, basis, boundary data, conditioning, integration, time stepping or physical modelling.

Evidence notebook 8: Interior residual

Differentiate twice and add u. Zero exactly. Keep coordinates, units, matrix entries, signs, basis order and rounding visible. Analytic teaching field. Record residual. Change one input, recompute the affected mechanism and identify whether the change comes from geometry, basis, boundary data, conditioning, integration, time stepping or physical modelling.

Evidence notebook 9: Boundary residual

Re-evaluate both endpoints. Both zero. Keep coordinates, units, matrix entries, signs, basis order and rounding visible. Separate from PDE residual. Record values. Change one input, recompute the affected mechanism and identify whether the change comes from geometry, basis, boundary data, conditioning, integration, time stepping or physical modelling.

Evidence notebook 10: Forcing

Add a particular solution when needed. Homogeneous basis alone is insufficient. Keep coordinates, units, matrix entries, signs, basis order and rounding visible. State source term. Record decomposition. Change one input, recompute the affected mechanism and identify whether the change comes from geometry, basis, boundary data, conditioning, integration, time stepping or physical modelling.

Evidence notebook 11: Conditioning

Add wave directions carefully. Redundancy can grow. Keep coordinates, units, matrix entries, signs, basis order and rounding visible. Inspect singular values. Record metric. Change one input, recompute the affected mechanism and identify whether the change comes from geometry, basis, boundary data, conditioning, integration, time stepping or physical modelling.

Evidence notebook 12: Conclusion

State basis, boundaries and validation. Exact for this simple case. Keep coordinates, units, matrix entries, signs, basis order and rounding visible. Complex geometry differs. Write claim. Change one input, recompute the affected mechanism and identify whether the change comes from geometry, basis, boundary data, conditioning, integration, time stepping or physical modelling.

Misconceptions, failure modes and recovery

PDE-satisfying means exact solution

Boundaries and interfaces are still approximated. Build a minimal counterexample, restore the missing condition and recalculate before making a broader claim.

Any oscillatory basis fits any wave

The operator and wavenumber must match. Build a minimal counterexample, restore the missing condition and recalculate before making a broader claim.

Forcing can be ignored

Homogeneous bases solve a different equation. Build a minimal counterexample, restore the missing condition and recalculate before making a broader claim.

More plane waves always help

Redundancy can ruin conditioning. Build a minimal counterexample, restore the missing condition and recalculate before making a broader claim.

Zero volume residual proves accuracy

Boundary residual may dominate. Build a minimal counterexample, restore the missing condition and recalculate before making a broader claim.

Corners need no special treatment

Local singular behaviour may require enrichment. Build a minimal counterexample, restore the missing condition and recalculate before making a broader claim.

Every Trefftz method is meshless

Hybrid finite elements also use Trefftz fields. Build a minimal counterexample, restore the missing condition and recalculate before making a broader claim.

One frequency test proves robustness

Resolution must follow wavenumber. Build a minimal counterexample, restore the missing condition and recalculate before making a broader claim.

A responsible eight-step decision path

1. Define PDE

State variables, units, domain and target claim.

2. Choose Trefftz basis

Write the exact algebra and verify its defining promise.

3. Add particular field

Separate local approximation from the full governing model.

4. Fit boundaries

Record matrices, ordering and boundary conventions.

5. Couple interfaces

Measure residual, conditioning and conservation as appropriate.

6. Inspect conditioning

Challenge corners, difficult data and perturbed geometry.

7. Refine frequency resolution

Repeat with a systematic resolution or parameter change.

8. Validate physics

Compare with independent evidence and state the limit.

Learner lab: build transferable understanding

Basis verification

Differentiate sin and cos twice. Predict first, show the arithmetic, record one diagnostic, compare a baseline and finish with one supported conclusion plus one limit.

Boundary solve

Recover a=1 and b=2. Predict first, show the arithmetic, record one diagnostic, compare a baseline and finish with one supported conclusion plus one limit.

Residual ledger

Separate volume and boundary errors. Predict first, show the arithmetic, record one diagnostic, compare a baseline and finish with one supported conclusion plus one limit.

Forcing

Add a particular solution. Predict first, show the arithmetic, record one diagnostic, compare a baseline and finish with one supported conclusion plus one limit.

Plane waves

Vary directions in two dimensions. Predict first, show the arithmetic, record one diagnostic, compare a baseline and finish with one supported conclusion plus one limit.

Interface

Couple two constant-coefficient regions. Predict first, show the arithmetic, record one diagnostic, compare a baseline and finish with one supported conclusion plus one limit.

Conditioning

Add nearly duplicate basis functions. Predict first, show the arithmetic, record one diagnostic, compare a baseline and finish with one supported conclusion plus one limit.

Frequency

Increase wavenumber and basis size. Predict first, show the arithmetic, record one diagnostic, compare a baseline and finish with one supported conclusion plus one limit.

Student explanation

Explain exact interior behaviour. Predict first, show the arithmetic, record one diagnostic, compare a baseline and finish with one supported conclusion plus one limit.

Parent explanation

Relate prepared puzzle pieces to edge fitting. Predict first, show the arithmetic, record one diagnostic, compare a baseline and finish with one supported conclusion plus one limit.

Deeper transfer: from mathematical structure to trustworthy computation

Local Trefftz spaces

Elementwise Trefftz functions can satisfy a PDE inside each cell while numerical fluxes or hybrid traces connect the cells.

Wave resolution

A plane-wave basis carries phase information that ordinary polynomials represent inefficiently, but directions and counts must match geometry and frequency.

Particular fields

A known particular solution can remove volume forcing before the homogeneous Trefftz expansion. Approximate particular solutions introduce their own error ledger.

Variational coupling

Hybrid and discontinuous formulations derive stable boundary or interface forms rather than simply collocating every condition.

Evanescent modes

Near interfaces and corners, decaying modes may be as important as propagating plane waves. A direction-only basis can miss them.

Dispersion

Wave computations should measure phase and amplitude error, not only a pointwise residual.

Preconditioning

Basis orthogonalisation, scaling and suitable traces can reduce the ill conditioning created by many oscillatory functions.

Career connection

Trefftz work combines differential equations, waves, variational methods and high-performance linear algebra.

Student and parent guidance: learning the method without losing the meaning

A student’s first route

Verify sin and cos by differentiation, then solve the two boundary equations by hand.

A parent’s useful question

Ask what the basis satisfies exactly and which conditions are still fitted.

A small coding exercise

Create a Helmholtz basis, evaluate volume residuals and solve a least-squares boundary fit.

An engineering exercise

Compare polynomial and plane-wave bases for a simple acoustic cavity at increasing frequency.

A verification habit

Record PDE, boundary, interface, forcing and phase errors separately.

A career pathway

Waves, PDEs, variational methods and computation support acoustics, electromagnetics and vibration careers.

Scope, evidence and honest limits

What the worked example proves

The chosen sine and cosine span the exact solution of one simple homogeneous one-dimensional problem.

What a software result must add

A wave computation adds geometry, frequency, basis directions, interfaces, quadrature, conditioning and phase validation.

What to compare

Compare boundary and field errors against an analytical or independently refined solution across frequencies.

What to preserve

Save basis definitions, matrices, residuals, singular values, frequency sweeps and corrected calculations.

A final communication check

Explain the method once to a student and once to a parent. Preserve the mechanism, arithmetic, evidence and limit while changing only the vocabulary.

Field notes: ten deeper checks before trusting the result

A two-dimensional basis

For the Helmholtz equation, plane waves exp(i k d dot x) satisfy the homogeneous PDE when d is a unit direction. Real sine and cosine combinations give equivalent real bases, but the chosen directions control angular resolution.

Interface continuity

Across two materials, a field may require continuity of value and a scaled normal flux. Enforcing only one condition can leave a visually smooth but physically incomplete solution, so keep separate interface residuals.

Domain decomposition

Partitioning a difficult domain lets each region use a tailored Trefftz space. The price is a larger interface system whose stability depends on traces, normals, flux definitions and the coupling formulation.

High-frequency counting

As wavelength shrinks, the basis and geometric representation must resolve more oscillations. A fixed number of directions does not remain adequate merely because every plane wave still satisfies the PDE exactly.

Complex arithmetic

Wave bases often use complex exponentials. Real physical fields may be obtained from real parts or paired coefficients, but matrix norms, boundary data and energy calculations must use a consistent complex convention.

Numerical integration

A basis can satisfy the PDE analytically while boundary integrals are evaluated numerically. Curved boundaries and rapidly oscillating traces need sufficient quadrature; otherwise integration error defeats the exact interior property.

Nullspaces and resonances

Interior resonance can make a boundary formulation singular or nearly singular at special frequencies. Detecting it requires spectral or singular-value evidence, not a larger coefficient tolerance.

Model versus discretisation

A perfectly converged acoustic Trefftz solution can still be wrong if wall impedance, material speed or source description is wrong. Separate mathematical discretisation checks from physical parameter validation.

A useful portfolio task

Solve the teaching interval, a rectangle with plane waves and a two-region interface problem. For each, publish PDE residual, boundary residual, coefficient condition number and refinement evidence.

Transfer to learning

Trefftz thinking teaches students to choose representations that already obey part of a problem. This reduces one kind of error while making the remaining constraints more visible and easier to audit.

A frequency-resolution record

For every wavenumber, record wavelength, region diameter, basis count, direction set, quadrature order, singular values and field error. Repeating the same table across a frequency sweep shows whether accuracy loss comes from insufficient angular resolution, boundary integration, resonance or coefficient sensitivity. This evidence is more useful than saying that the method works well for waves, because it identifies the range in which the chosen basis and coupling are actually reliable.

An honest stopping rule

Stop enriching the basis when field and boundary errors have stabilised at the required tolerance and further functions mainly increase coefficient sensitivity. State the tolerance, validation points and arithmetic precision. This makes the final basis size a reproducible decision rather than a convenient number selected after looking at the plot.