Chapter map

1. 1. Operator
2. 2. Initial guess
3. 3. Auxiliary operator
4. 4. Controls
5. 5. Deformation
6. 6. Series
7. 7. Coefficients
8. 8. Partial sum
9. 9. Benchmark
10. 10. Residual
11. 11. h study
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. Operator

A small example makes the mechanism visible. Set N[u]=u squared-2.

A concrete situation

Positive-root example.

What the mathematics does not license

Algebraic teaching model. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Record N. 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. Initial guess

This is where modelling choices begin to affect evidence. Use u0=1.

A concrete situation

Close to positive branch.

What the mathematics does not license

Branch matters. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Record u0. 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. Auxiliary operator

The central mathematical idea is concrete. Let L be identity.

A concrete situation

L[0]=0.

What the mathematics does not license

Simple inverse. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Record L. 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. Controls

Here the geometry becomes operational. Use H=1 and h=-.5.

A concrete situation

Auxiliary choices.

What the mathematics does not license

Not physical constants. Evidence becomes stronger when the limitation is tested instead of hidden.

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.

---

5. Deformation

This step deserves a careful pause. (1-q)(Phi-1)=q h(Phi squared-2).

A concrete situation

Zero-order equation.

What the mathematics does not license

Keep signs. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Record formula. 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. Series

The useful question is not only how to calculate, but what the calculation assumes. Phi=1+sum u_m q^m.

A concrete situation

Embedding expansion.

What the mathematics does not license

Convergence at q=1 required. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Record convention. 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. Coefficients

A small example makes the mechanism visible. u1=.5,u2=0,u3=-.125,u4=0.

A concrete situation

Matched by powers.

What the mathematics does not license

Verify algebra. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Record table. 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. Partial sum

This is where modelling choices begin to affect evidence. At q=1 get 1.375.

A concrete situation

Fourth order shown.

What the mathematics does not license

Declare truncation. Keeping this boundary clear makes the result easier to audit and transfer.

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.

---

9. Benchmark

The central mathematical idea is concrete. sqrt2 about 1.41421356.

A concrete situation

Error about .03921356.

What the mathematics does not license

Positive branch. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Record error. 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. Residual

Here the geometry becomes operational. 1.375 squared-2=-.109375.

A concrete situation

Original equation check.

What the mathematics does not license

Not yet tight. 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.

---

11. h study

This step deserves a careful pause. Vary h and order.

A concrete situation

Inspect residual and stability.

What the mathematics does not license

Avoid cherry-picking. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Record curves. 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

The useful question is not only how to calculate, but what the calculation assumes. State all auxiliary choices and evidence.

A concrete situation

Series design needs validation.

What the mathematics does not license

No automatic convergence. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Write limit. 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

The homotopy analysis method constructs a continuous deformation from an initial guess to a solution of a nonlinear operator equation. An embedding parameter generates a series, while an auxiliary linear operator and a convergence-control parameter shape how that series behaves.

Write N[u]=0 and choose an initial guess u0, an auxiliary linear operator L with L[0]=0, an auxiliary function H and a nonzero convergence-control parameter h.

A standard zero-order deformation is (1-q)L[Phi-u0]=q h H N[Phi], where q runs from 0 to 1. At q=0, Phi=u0; if the series converges at q=1, Phi should satisfy the target equation.

For a compact teaching calculation, let N[u]=u squared-2, L be the identity, H=1 and u0=1. Choose h=-0.5 and write Phi=1+u1 q+u2 q squared+u3 q cubed and so on.

Matching coefficients gives u1=0.5, u2=0, u3=-0.125 and u4=0. The fourth-order partial sum at q=1 is 1.375.

The positive root square root 2 is approximately 1.4142135624, so this short series has absolute error about 0.0392135624. More terms or a different admissible h may improve or worsen the approximation.

Substituting 1.375 into u squared-2 gives a residual of -0.109375. This residual, rather than the visual smoothness of the series, is the immediate check on the nonlinear equation.

Did You Know? The convergence-control parameter is not a physical constant in the original model. It belongs to the solution construction and can be studied with h-curves, residual minimisation and convergence tests.

The auxiliary linear operator affects what each deformation equation is easy to solve. It must respect the boundary conditions and have an inverse that is defined on the relevant residuals.

An embedding parameter resembles perturbation bookkeeping, but HAM does not require a naturally small physical parameter. That freedom also creates responsibility: auxiliary choices must be reported and justified.

A flat portion of an h-curve is practical evidence that a computed quantity is less sensitive to h over a range. It is not a proof that the full series converges everywhere.

Multiple solution branches can require different initial guesses or auxiliary operators. One converged series should not be presented as uniqueness evidence.

Students can derive the first coefficient equations, vary h, plot residual against truncation order and compare with Newton iteration for the same root. The comparison separates analytic-series design from local root updates.

Applications include nonlinear waves, fluid mechanics, oscillators and boundary layers. The method matters because it makes convergence design an explicit mathematical object rather than hiding it inside an unspecified expansion.

A responsible report states N, u0, L, H, h, deformation equations, boundary treatment, truncation, residuals, h-selection evidence, branch checks and validation.

The transferable lesson is that freedom in a method is useful only when every choice is recorded and tested. Convergence control is an evidence process, not a decorative parameter.

---

Frequently asked questions

What is the embedding parameter?

A construction variable q that deforms the initial guess at q=0 toward the target equation at q=1.

What is the convergence-control parameter?

An auxiliary nonzero parameter that changes the convergence behaviour of the HAM series.

Is h a physical constant?

No. It belongs to the solution method and must be selected and reported.

Does HAM require a small physical parameter?

No, but convergence at q=1 still needs evidence.

What does an h-curve show?

Sensitivity of a chosen truncated quantity to h; a flat region is practical evidence, not a universal proof.

What must be reported?

Nonlinear and auxiliary operators, initial guess, H, h, deformation equations, boundaries, truncation, residuals and branch checks.

---

Primary source and fact-check note

S. J. Liao, “Homotopy Analysis Method: A New Analytical Technique for Nonlinear Problems,” Communications in Nonlinear Science and Numerical Simulation 2(2), 1997. 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

• Adomian decomposition source
• Quasilinearization source
• Partition of unity method
• Computational mathematics hub
• How Mathematics Works

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

Deformation audit

Check q=0 gives the initial guess and q=1 gives the target operator.

Auxiliary audit

State L, H and h with reasons and boundary compatibility.

Series audit

Derive coefficient equations, truncation order and residuals reproducibly.

Convergence audit

Use h studies, successive orders and an independent benchmark.

Worked evidence notebook: reconstruct every number

Evidence notebook 1: Operator

Set N[u]=u squared-2. Positive-root example. Keep signs, nodes, coefficients, equations, boundary data, truncation order and rounding visible. Algebraic teaching model. Record N. Change one input, recompute the mechanism and identify whether the difference comes from representation, nonlinear solving, convergence, conditioning or validation.

Evidence notebook 2: Initial guess

Use u0=1. Close to positive branch. Keep signs, nodes, coefficients, equations, boundary data, truncation order and rounding visible. Branch matters. Record u0. Change one input, recompute the mechanism and identify whether the difference comes from representation, nonlinear solving, convergence, conditioning or validation.

Evidence notebook 3: Auxiliary operator

Let L be identity. L[0]=0. Keep signs, nodes, coefficients, equations, boundary data, truncation order and rounding visible. Simple inverse. Record L. Change one input, recompute the mechanism and identify whether the difference comes from representation, nonlinear solving, convergence, conditioning or validation.

Evidence notebook 4: Controls

Use H=1 and h=-.5. Auxiliary choices. Keep signs, nodes, coefficients, equations, boundary data, truncation order and rounding visible. Not physical constants. Record values. Change one input, recompute the mechanism and identify whether the difference comes from representation, nonlinear solving, convergence, conditioning or validation.

Evidence notebook 5: Deformation

(1-q)(Phi-1)=q h(Phi squared-2). Zero-order equation. Keep signs, nodes, coefficients, equations, boundary data, truncation order and rounding visible. Keep signs. Record formula. Change one input, recompute the mechanism and identify whether the difference comes from representation, nonlinear solving, convergence, conditioning or validation.

Evidence notebook 6: Series

Phi=1+sum u_m q^m. Embedding expansion. Keep signs, nodes, coefficients, equations, boundary data, truncation order and rounding visible. Convergence at q=1 required. Record convention. Change one input, recompute the mechanism and identify whether the difference comes from representation, nonlinear solving, convergence, conditioning or validation.

Evidence notebook 7: Coefficients

u1=.5,u2=0,u3=-.125,u4=0. Matched by powers. Keep signs, nodes, coefficients, equations, boundary data, truncation order and rounding visible. Verify algebra. Record table. Change one input, recompute the mechanism and identify whether the difference comes from representation, nonlinear solving, convergence, conditioning or validation.

Evidence notebook 8: Partial sum

At q=1 get 1.375. Fourth order shown. Keep signs, nodes, coefficients, equations, boundary data, truncation order and rounding visible. Declare truncation. Record value. Change one input, recompute the mechanism and identify whether the difference comes from representation, nonlinear solving, convergence, conditioning or validation.

Evidence notebook 9: Benchmark

sqrt2 about 1.41421356. Error about .03921356. Keep signs, nodes, coefficients, equations, boundary data, truncation order and rounding visible. Positive branch. Record error. Change one input, recompute the mechanism and identify whether the difference comes from representation, nonlinear solving, convergence, conditioning or validation.

Evidence notebook 10: Residual

1.375 squared-2=-.109375. Original equation check. Keep signs, nodes, coefficients, equations, boundary data, truncation order and rounding visible. Not yet tight. Record residual. Change one input, recompute the mechanism and identify whether the difference comes from representation, nonlinear solving, convergence, conditioning or validation.

Evidence notebook 11: h study

Vary h and order. Inspect residual and stability. Keep signs, nodes, coefficients, equations, boundary data, truncation order and rounding visible. Avoid cherry-picking. Record curves. Change one input, recompute the mechanism and identify whether the difference comes from representation, nonlinear solving, convergence, conditioning or validation.

Evidence notebook 12: Conclusion

State all auxiliary choices and evidence. Series design needs validation. Keep signs, nodes, coefficients, equations, boundary data, truncation order and rounding visible. No automatic convergence. Write limit. Change one input, recompute the mechanism and identify whether the difference comes from representation, nonlinear solving, convergence, conditioning or validation.

Misconceptions, failure modes and recovery

Treating h as fitted physics

The convergence-control parameter belongs to the mathematical construction. Build a small counterexample, repair the omitted condition and rerun the relevant residual or convergence check.

Choosing h from one attractive plot

Residuals and multiple orders must support the selected range. Build a small counterexample, repair the omitted condition and rerun the relevant residual or convergence check.

Forgetting boundary conditions

Every deformation equation needs consistent homogeneous or transformed boundaries. Build a small counterexample, repair the omitted condition and rerun the relevant residual or convergence check.

Assuming q-series convergence

The formal series at q=1 may diverge without suitable choices. Build a small counterexample, repair the omitted condition and rerun the relevant residual or convergence check.

Hiding auxiliary freedom

Different L, H, u0 and h can produce materially different performance. Build a small counterexample, repair the omitted condition and rerun the relevant residual or convergence check.

Claiming all branches

One initial guess and deformation may reach only one solution family. Build a small counterexample, repair the omitted condition and rerun the relevant residual or convergence check.

A responsible eight-step decision path

1. Define target operator

State variables, units, domain, boundaries and target claim.

2. Choose initial guess

Write the mathematical mechanism and its exact promise.

3. Choose auxiliaries

Keep the representation choices and original model separate.

4. Build deformation

Record ordering, signs, equations and tolerances.

5. Derive coefficients

Measure conditioning, convergence or solver behaviour.

6. Select h

Challenge a difficult input or alternative start.

7. Check residual

Repeat with systematic order, mesh or parameter changes.

8. Validate branches

Compare with independent evidence and state limits.

Learner lab: build transferable understanding

Coefficient reconstruction

Expand the teaching deformation through q to the fourth power. Predict first, show the arithmetic, save one diagnostic, compare a baseline and finish with one supported conclusion plus one explicit limit.

h sweep

Repeat coefficients for several negative h values. Predict first, show the arithmetic, save one diagnostic, compare a baseline and finish with one supported conclusion plus one explicit limit.

Order plot

Graph nonlinear residual against truncation order. Predict first, show the arithmetic, save one diagnostic, compare a baseline and finish with one supported conclusion plus one explicit limit.

Branch experiment

Change u0 from positive to negative and observe the root family. Predict first, show the arithmetic, save one diagnostic, compare a baseline and finish with one supported conclusion plus one explicit limit.

Baseline test

Compare HAM partial sums with Newton updates for u squared=2. Predict first, show the arithmetic, save one diagnostic, compare a baseline and finish with one supported conclusion plus one explicit limit.

Communication task

Explain auxiliary freedom as a tested design choice. Predict first, show the arithmetic, save one diagnostic, compare a baseline and finish with one supported conclusion plus one explicit limit.

Deeper transfer: from mathematical structure to trustworthy computation

Operator freedom

A useful L makes each deformation equation solvable and compatible with the desired boundary space.

Residual optimisation

One may choose h by minimising a declared residual norm at a fixed order, then verify that choice at higher orders.

Multiple controls

Some formulations use several convergence-control parameters; extra freedom increases the reporting burden.

Series acceleration

Pade or other resummation can improve a truncated series but creates a new approximation to test.

Open questions

Formal construction, numerical success and rigorous convergence are different levels of evidence.

Career transfer

Nonlinear waves, asymptotic analysis and scientific computing use the same habit of designing and testing representations.

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

For a student

Keep q, h and the physical variables separate in every line.

For a parent

Ask which choices belong to the model and which belong to the solution method.

For a teacher

Use a small algebraic example before a nonlinear differential equation.

For a project

Save deformation equations, h sweeps, order tables and original residuals.

For pathways

Nonlinear analysis, fluids and computational mechanics use representation and convergence reasoning.

For balance

Freedom becomes a strength when the choices are transparent and reproducible.

Scope, evidence and honest limits

What the example proves

One declared deformation yields four visible coefficients and a nonzero checked residual.

What it does not prove

It does not prove that h=-.5 is optimal or that the infinite series converges at q=1.

What software must add

A real study adds h-range selection, higher orders, boundary handling and branch tests.

What to compare

Use Newton, numerical integration or a trusted benchmark under the same model.

What to preserve

Save all auxiliary choices, coefficients, h-curves, residuals and validation data.

Field notes: eight deeper checks before trusting the result

Boundary operator

Choose L so homogeneous deformation corrections satisfy transformed boundaries.

h curve

Plot several orders and relevant quantities, not one cherry-picked scalar.

Residual minimum

Define the norm, grid and order used to choose h.

Coefficient growth

Inspect terms before trusting a visually flat h interval.

Branch tracking

Use alternative u0 or continuation when multiple solutions matter.

Stopping rule

Require residual and order stability at the chosen h.

Portfolio evidence

Publish deformation equations, h curves and benchmarks.

Transfer

Method design teaches that numerical freedom requires explicit evidence.

Further practice: six transferable checks

A coefficient audit

Substitute the truncated q-series back into the deformation equation by powers of q.

A parameter map

Show residual against h and order on the same axes or table.

An operator study

Compare two admissible auxiliary operators on one benchmark.

A branch study

Change the initial guess and record which root or solution family is reached.

A reproducibility table

Record N, L, H, u0, h, order and residual norm.

A communication check

Describe convergence evidence without turning an h-curve into a proof.

Case study: from a classroom mechanism to a defensible computation

From target equation to deformation

Begin with the nonlinear operator and boundary conditions, then choose an admissible initial guess and an auxiliary linear operator whose inverse is manageable. Define the zero-order deformation with q, H and h. Verify its two endpoints algebraically. Only after those checks should the q-series be differentiated to obtain higher-order equations. Every auxiliary choice belongs in the reproducibility record.

Reading the square-root example

The algebraic example keeps L and H simple so the effect of h is visible. Four displayed corrections give 1.375, which is recognisably near square root two but still has a substantial residual. This is useful: the method is not judged by a pleasant number. The original nonlinear equation supplies a direct error signal, and higher orders or alternative h values must earn improvement.

Selecting the convergence control

An h-curve plots a truncated quantity over candidate h values. A stable region can guide selection, but the plotted quantity may look flat while another field component or the residual is not. Use several orders, relevant observables and a declared residual norm. If h is optimised numerically, repeat the check at a higher order and on independent points.

Choosing the auxiliary operator

L determines the equations solved at every order and the homogeneous terms available to meet boundaries. An operator chosen only for easy inversion may represent the target behaviour poorly. A useful study compares at least two admissible choices on a benchmark, accounting for work as well as truncation error. The boundary operator must be carried consistently through the deformation sequence.

Branches and initial guesses

Nonlinear equations may have several solutions. The initial guess and deformation path can favour one branch, while another admissible choice reaches a different one. Use continuation, symmetry arguments or multiple starts when branch structure matters. A small residual confirms one approximate solution under the declared construction; it does not prove that all solutions were found or that the physical interpretation is unique.

What parents can ask

Ask which symbols come from the original problem and which were introduced to solve it. A strong answer separates physical parameters from q and h, explains why h is varied and checks the final residual. That distinction is a valuable general habit: modelling choices, numerical choices and evidence should not be blended into one unexplained output.

What a portfolio should show

Publish the target operator, boundary conditions, initial guess, L, H, h range, deformation equations and coefficient table. Add h-curves for several orders, original-equation residuals and an independent benchmark. If a rational resummation or optimised h is used, compare it with the raw series. This makes convergence control a visible analysis rather than a hidden tuning step.

A careful conclusion

Homotopy analysis matters because it turns the design of a nonlinear series and its convergence into explicit mathematics. That freedom can solve problems without a natural small physical parameter, yet it also permits weak choices. A credible result identifies every auxiliary object, reports the tested h and order range and validates the final field against the target equation and an independent reference.