Evidence for macroscopic performance and limits

A useful evidence package names the product

Report fibre length, diameter or linear density, processing state, sampling and test conditions. Include uncertainty and distributions. Tensile, electrical and thermal properties need their respective measurements; excellence in one does not establish all three.

A nanotube, a manufactured strand and a polymer-infiltrated composite are different tested objects. Tube contacts can slip, some paths may remain poorly loaded, and a matrix can carry force and alter constraints. Electrical and thermal paths also depend on contacts and structure. There is no general rule converting a high tensile result into a conductivity, or a transport result into a fatigue life. Identify the architecture before deciding which property a number describes.

Behabtu and colleagues reported manufactured, wet-spun CNT fibres and several measured functions in 2013. That is evidence at a macroscopic fibre level. A separate CNT-fibre/epoxy study reported a hierarchical structure in which polymer entered the fibre architecture, with transport measurements on composite specimens. Such studies motivate testing each actual structure and function; the examples below are entirely invented and reproduce neither study’s data nor processing.

Use R = F_max/μ₀ for the specimen’s specific tensile strength, where F_max is its resolved maximum tensile force and μ₀ is reference mass per length. A tex is one gram per kilometre; R is reported here in N/tex. It is a density-normalised tensile endpoint, not work absorbed or the capacity of a complete cable. Specific strength and the scope of a reported result develops that normalisation.

Let c denote axial electrical conductivity in MS/m, where 1 MS/m = 10⁶ siemens per metre. Conductivity is not the conductance of a whole wire, which also depends on geometry. Thermal conductivity k, in W/(m·K), describes heat transport associated with a temperature gradient. These three quantities need their respective observations, stated specimen basis and measurement state.

Keep each property attached to its specimen and state

Consider two fictional preparation routes, A and B. Six separated, prespecified test pieces come from one selected one-metre segment of each route, from one preparation run per route. All tensile gauges are 20.0 mm long and have supplied μ₀ = 0.200 tex. Their complete valid force records resolve F_max; the table is a summary, not a claim that a few isolated points alone locate every maximum. Reference length, mass basis, conditioning, gripping assessment and gauge-strain-rate procedure are matched.

For each piece, c is measured axially while unloaded at 23 °C in the stated dry condition, before its destructive tensile test under the same temperature and conditioning. The electrical procedure is supplied as verified not to alter the subsequent tensile response. Its measurement model and gross cross-sectional convention are common to the two routes. The IDs therefore pair two defined results on the same piece; they do not say c was measured during rupture, or that it remained unchanged under load.

For A1, F_max = 0.680 N and μ₀ = 0.200 tex give R = 3.40 N/tex, paired with c = 2.00 MS/m. A6 instead has R = 1.60 N/tex and c = 5.40 MS/m. Reporting “3.40 N/tex and 5.40 MS/m” as if these were one tested piece would construct a specimen absent from the record. A claim about paired functions needs the matching IDs as well as two columns of attractive numbers.

These six observations do not establish the distribution of an entire production population. They describe sampled gauges from one selected segment per route, not six independent preparation runs or a direct continuous test of the whole metre. Manufacturing control, yield and reproducibility develops that broader production question. Retain where pieces came from and how they were chosen; more digits do not repair selection bias.

Two separate achievements need not occur together

Invent a record-comparison benchmark: a piece must have R ≥ 2.50 N/tex and its prior unloaded c ≥ 3.50 MS/m, with equality included. These thresholds are chosen teaching criteria, not a standard, a product qualification or a requirement that both functions persist simultaneously during service. The first scatter plot places both values for each piece at one point. Points are not joined into a response curve.

Route A has three pieces meeting the R criterion and a different three meeting the c criterion. None meets both. Route B has B1, B2 and B3 meeting both central-value criteria, while B4–B6 do not. This does not make B universally superior: A has larger individual maxima, and a different requirement could favour different records. It answers this defined combined question on these tested pieces.

Route A’s means are R̄ = 2.50 N/tex and c̄ = 3.70 MS/m, so both means meet the respective thresholds even though no A piece does. Route B’s means are 2.55 N/tex and 3.45 MS/m: its conductivity mean misses the criterion, although three B pieces meet both. A pair of means is not the joint distribution of individual results, and a mean point need not be an actual specimen.

If only A’s two separate value lists survived, pairing would be missing. Keeping its R list but reversing the c list would pair its first three high-R pieces with its three high-c results, producing three combined central-value matches. Every separate mean, range and maximum would remain the same. The actual IDs rule out that alternative for this record. A plot of corresponding values can reveal association; it does not prove that increasing one property causes the other to change.

Now suppose thermal conductivities 80, 100 and 120 W/(m·K) are recorded on three other matched companion pieces from route B. Their mean is 100 W/(m·K), but none is identified as B1–B6. This adds thermal evidence for those companions, not a measured triple of R, c and k for any listed tensile piece. A justified sampling/model relation or appropriately paired testing would be needed for a new joint claim. Retain the thermal achievement rather than discarding it or attaching it to an untested individual.

Variation and measurement limits answer different questions

Observed spread is variation among the reported specimen results. A has R from 1.60 to 3.40 N/tex and c from 2.00 to 5.40 MS/m; B has R from 2.30 to 2.80 and c from 3.20 to 3.70 in the same units. A narrower set of six results is not by itself a complete uncertainty estimate, a guaranteed population range or proof of repeatable manufacture.

For a separate exact arithmetic model, stipulate that each piece’s true R and c lie in the closed rectangle extending ±0.10 N/tex and ±0.10 MS/m around its reported pair. Every combination in that rectangle is permitted by this supplied model, with no probability assigned. These are invented admissible bounds, not standard uncertainties, confidence intervals, or a real uncertainty budget obtained from the table’s scatter. Real measurement results require justified uncertainties and correlations appropriate to their measurement models.

Under this rectangle model, the combined criteria are established only if both lower limits meet their thresholds. They are excluded if either upper limit is below its threshold. Otherwise the evidence leaves the classification unresolved. This preserves an undecided region instead of forcing every central value into a definite claim.

B1 has lower limits 2.70 N/tex and 3.60 MS/m, and B2 has 2.60 and 3.50; both establish the stated criteria within the model. B3’s centre meets both, but its c can be as low as 3.40, so its classification is unresolved. B4’s centre misses c, but its rectangle reaches c = 3.50 with R ≥ 2.50, so it too is unresolved. B5’s c upper limit is only 3.40, and B6’s R upper limit is only 2.40: both are excluded.

All six A rectangles are excluded: the first three remain too low in c, and the last three remain too low in R. Thus the fictional bounds establish two B pieces, leave two unresolved and exclude two, while establishing none of A. These counts concern the supplied tested records and model, not a population probability or a usable-production yield. The second figure explains why a central-value match and a bounded conclusion can differ.

A whole-composite record is not a direct constituent measurement

Adding a matrix changes both the tested architecture and the inference problem. Consider a separate invented composite tensile record with no bending or acceleration. At a given common section, the total axial force is F_C = F_M + F_R, the sum carried by matrix and reinforcement. Force balance is general for these two defined parts, but identifying each contribution requires more information than the total force.

For a deliberately restricted estimate, suppose matrix and reinforcement have common gauge strain, and the matrix inside the composite follows a separately measured reference force–strain response at that strain. The reference has the same matrix cross-sectional amount as the composite, the same gauge, temperature and history, and no processing change. Assume no extra confinement or interface effect changes that matrix response. These assumptions are supplied for this teaching calculation; they are not established merely by testing a free matrix coupon.

The table then permits F_R = F_C − F_M,ref. At ε = 1.00%, the conditional reinforcement contribution is 12.0 − 4.0 = 8.0 N. At the resolved composite maximum at ε = 2.00%, it is 18.0 − 5.0 = 13.0 N. That is a force carried by all reinforcement at that state, not its intrinsic tensile strength, the force in every tube, or a new isolated-fibre breaking-force measurement.

The separate reference matrix reaches its own maximum of 6.0 N at 4.00% strain. Subtracting that maximum from the composite maximum would give 12.0 N, but it mixes two strain states. Maxima of separate records cannot generally be subtracted to identify simultaneous phase contributions. Moreover, if the matrix-in-composite assumption fails, even the 13.0 N subtraction is not identified by the supplied measurements. A changed interface, constraint or matrix state is a competing explanation, not automatically a loss of tube strength.

A discriminating next observation could measure phase strain or matrix response in the embedded state, test a matched architecture/control, and document damage or slip at the same loading frame. Compare a model prediction with measurements of the same quantity, structure and state, including its uncertainty. Agreement at one point can test that comparison; it does not validate every mechanism, load history or eventual failure mode.

Compare progress without changing levels

A study can demonstrate improved macroscopic behaviour without reaching ideal microscopic predictions. State both the achievement and its range. Reproducible scale, durability, interfaces and integration remain testable engineering questions rather than consequences of a headline material name.

An accurate achievement statement for the invented dossier is: these sampled, conditioned macroscopic pieces have the reported paired tensile endpoint and prior unloaded conductivity; their combined classification depends on whether central values or the stated bound model are being used. The separate companion thermal measurements and model-dependent composite load account add different evidence. None supplies conductivity during rupture, unmeasured fatigue or continuous metre-scale capacity.

The next test should resolve a specific ambiguity. Pair thermal testing with identified pieces if the claim concerns a triple of functions; observe transport during a stated mechanical load if it concerns coexisting loaded performance; test the embedded phase assumption if it concerns constituent load. Broader production coverage and connected-component assessment remain distinct tasks. Preserving an achievement and naming the next discriminating measurement makes a stronger scientific conclusion than combining unrelated best numbers.

Worked examples

Try the reasoning

1. Can one property be inferred from another impressive value?

   Show the answer and reasoning

   Not without a validated relation.

   The mechanisms and measurements differ.

2. Why compare distributions?

   Show the answer and reasoning

   Practical output varies.

   A single extreme value does not describe the produced population.

3. Recover R for B3 from its 0.520 N maximum force and μ₀ = 0.200 tex. Does its paired c = 3.50 MS/m mean conductivity was measured at that maximum load?

   Show the answer and reasoning

   R = 2.60 N/tex; no.

   0.520/0.200 = 2.60. The matched ID links the two results, but c was measured in the prior unloaded state. It cannot be relabelled as transport under the destructive tensile maximum.

4. Change only the central-value benchmark to R ≥ 3.00 N/tex, with no c requirement. Which listed route has matches? Does that contradict the earlier comparison?

   Show the answer and reasoning

   A1–A3 match; no B piece does. There is no contradiction.

   The earlier combined question required c as well. Requirements and endpoints define a comparison; B’s three matches under one chosen benchmark did not establish universal superiority.

5. In a new set of six identified pieces, four meet a mechanical criterion and five meet an electrical criterion, but their matching IDs are lost. What smallest and largest number could meet both?

   Show the answer and reasoning

   At least three and at most four of the six, with the exact count unresolved.

   The two groups have nine memberships among six pieces, so at least three memberships must overlap. At most all four mechanically qualifying pieces can also qualify electrically. Constructing either arrangement shows why the separate counts do not identify the exact joint count.

6. A new record has R = 2.55 N/tex and c = 3.65 MS/m, with the same ±0.10 rectangle model. Classify it under the original combined benchmark. Then reduce only the R bound to ±0.04 N/tex.

   Show the answer and reasoning

   Initially unresolved; with the narrower supplied R bound it is established under the model.

   Initially R can be 2.45, below 2.50, while c is at least 3.55. With ±0.04, R is at least 2.51 and c remains at least 3.55, so every permitted pair meets both. A narrower justified bound, rather than more displayed digits, changes the inference.

7. The three separate thermal companions have k = 80, 100 and 120 W/(m·K). What is their mean? Does it prove B1 has k ≥ 100 W/(m·K)?

   Show the answer and reasoning

   The mean is 100 W/(m·K); B1’s thermal result remains unmeasured in this dossier.

   The arithmetic is (80 + 100 + 120)/3 = 100. The thermal pieces do not have B1’s ID, and no validated specimen-to-specimen prediction is supplied. Retain the companion results without inventing a joint individual measurement.

8. For a new matched composite frame at ε = 1.50%, total force is 16.0 N and the phase-equivalent reference matrix force is 4.5 N. Under the stated model find F_R. What if embedded-phase evidence instead establishes 6.0 N matrix force at that same frame?

   Show the answer and reasoning

   11.5 N under the reference-response assumption; 10.0 N using the new embedded-phase evidence.

   Force balance gives 16.0 − 4.5 = 11.5 N, or 16.0 − 6.0 = 10.0 N. The revised attribution does not change the measured 16.0 N total or itself show damage to the nanotubes. The partition depended on the matrix information.

9. A report combines one route’s best tensile result before treatment with its best conductivity after a treatment that changes mass and structure. What additional link is missing?

   Show the answer and reasoning

   A matched specimen/state relation supporting the combined claim.

   A treatment label cannot make the before- and after-state values coexist. Keep the reference mass/geometry, architecture and measurement state with each result. Paired tests or a justified transformation model are needed, not the two maxima alone.

10. Six separated 20.0 mm gauges from a selected one-metre segment give valid tensile results. Is continuous metre-long tensile performance or a whole production population established?

    Show the answer and reasoning

    No. The measured scale and sampling unit remain those tested gauges from that selected segment.

    Longer spans include untested regions, end locations and potential different failure paths. Pieces from one preparation run are not independent demonstrations across runs. A new length or production claim needs an appropriate matched test and sampling design rather than multiplication of gauge results.

What to carry forward

• Measure each claimed function at the right scale.
• Keep achievements and untested requirements explicit.
• A combined specimen claim needs paired properties at explicitly defined measurement states.
• Separate specimen variation, measurement uncertainty and an invented bound model.
• Whole-composite force is observed; phase attribution can depend on a same-state model.
• Choose the next measurement to resolve the specific missing correspondence, scale or mechanism.