Baker Street Studio貝克街工作室
← Research work

Metal Growth, Two Causes / Research notebook

Can a final weight tell two causes apart?

The same increase in metal mass can hide very different balances of deposition and dissolution.

An intricate gold-coloured metallic tree with visible roots and fine branches inside a sectioned transparent glass vessel.Enlarge ↗
Concept illustration inspired by Newton’s notes on metallic trees and vegetation in glass, Dibner MS 1031 B, f.1r. The gold colour, tree geometry and vessel are artistic choices, not an experiment reconstruction or evidence of making gold.
About this illustration

Concept illustration inspired by Newton’s notes on metallic trees and vegetation in glass, Dibner MS 1031 B, f.1r. The gold colour, tree geometry and vessel are artistic choices, not an experiment reconstruction or evidence of making gold. Source: Newton manuscript, normalized transcription, paragraphs 3–5 and 9.

Metal Growth, Two Causes

From Newton's mineral “vegetation” to the measurements that distinguish formation from loss

A mineral can preserve a final structure without uniquely preserving the route that made it. An electrode can gain the same mass through efficient deposition accompanied by dissolution, or through less efficient deposition with little loss. The common question is an inverse problem: what observations would let us distinguish the histories that produce the same endpoint?

This research began with Newton's language about the growth of metals and minerals. It developed along two connected tracks: a historical examination of what his manuscript actually claims, and modern mathematical models of what can and cannot be inferred from composition and mass measurements. The historical text supplies questions and distinctions; the modern equations are separate models, not equations attributed to Newton.

The account below preserves the substantive progression across seven rounds, including an early mathematical branch, the counterexamples that changed the claims, and two materially different versions of the seventh round. Numerical examples are explicitly identified as synthetic or calculated. No laboratory deposition measurements or natural-mineral dataset were analyzed in these recovered rounds.

1. The question and its vocabulary

“Growth” can name a visible outcome without identifying its mechanism. A branching deposit, a crystalline salt, and a mineral vein may invite comparison, but resemblance alone does not establish common chemistry, kinetics, or origin.

The historical question is what Newton meant by vegetation when discussing metals, minerals, glass vessels, salts, and a latent spirit. The modern question is more narrowly operational: given specified candidate mechanisms, which observations distinguish them?

Several terms organize the research:

The distinction between structural and practical questions becomes important later. A model can be nonidentifiable even with perfect instruments. Conversely, a mathematically identifiable design can be difficult to estimate from noisy measurements.

2. Round 1: one mineral endpoint, several possible histories

Starting question

Could a single terminal chemical indicator uniquely recover the proportions of three possible sources of a mineral sample?

The first mathematical branch deliberately used abstract signals rather than mineral measurements. Three hypothetical sources were assigned signal values 0.2, 0.5, and 0.8. Their nonnegative mixing weights were w₁, w₂, and w₃, with total weight one. The model was:

w₁ + w₂ + w₃ = 1

y₁ = 0.2w₁ + 0.5w₂ + 0.8w₃.

For an observed signal y₁ = 0.5, the recorded calculation exhibited three different histories:

Each produces precisely 0.5. The calculation used exact rational arithmetic rather than relying on rounded floating-point agreement. Its recorded test count was seven.

What the example established

The forward map from history to observation is many-to-one in this example. A shared endpoint cannot select one of these histories. In fact, rewriting the equations gives w₁ = w₃ and w₂ = 1 − 2w₁, so the three examples sit inside a continuous family with 0 ≤ w₁ ≤ 0.5. This algebra simply makes explicit the family already illustrated by the recorded counterexamples.

The attack on an overstrong conclusion

The next step challenged a tempting but invalid generalization: if one signal fails, does that mean formation history is forever unrecoverable?

The attack retained the same three sources and added a second synthetic indicator with source values (0, 1, 0). Its measured value was 0.4. The full system became:

w₁ + w₂ + w₃ = 1

(1/5)w₁ + (1/2)w₂ + (4/5)w₃ = 1/2

w₂ = 2/5.

The coefficient matrix has determinant −3/5 and rank three. The unique solution is:

w₁ = 3/10, w₂ = 2/5, w₃ = 3/10.

Thus the second, independent indicator removes the ambiguity within this model. The first result survives in its proper scope: one specified indicator is insufficient. The universal claim that histories can never be inferred does not survive.

Reproduction and checks

The attack used a symbolic matrix calculation. A reproduction then solved the same system using Python Fraction arithmetic and Gauss–Jordan elimination, without using the symbolic matrix library. The final reduced augmented matrix was the identity matrix with right-hand side (3/10, 2/5, 3/10).

The three equations check directly:

This was a different computational method applied in the same research context, not an independent laboratory replication.

Limits and the next question

The signals were not isotope measurements. Real isotope ratios may require concentration weighting; mineral formation need not involve only three sources; alteration can change multiple indicators together. The important next question was therefore whether apparently different measurements are genuinely independent, or are redundant consequences of a common process. That question eventually reappeared in the deposition model.

3. Round 1, historical branch: locating the manuscript and observable quantities

A separate early research package approached the topic through source identification. It located Newton's manuscript Of Natures obvious laws & processes in vegetation, catalogued as ALCH00081, through the Newton Project and the Chymistry of Isaac Newton project. The early card described two unfinished discussions of metal and mineral growth, dated approximately 1670–1675, and recorded a catalogue description of approximately 4,500 words and twelve pages.

It also identified RRUFF as an entry point for Raman, X-ray diffraction, and chemical information about minerals. The proposed bridge was an index with three distinct parts:

historical wording → a candidate modern interpretation → the observations required to test that interpretation.

This was a research map, not yet a completed mapping between Newton's account and a modern mechanism. The package explicitly stated that the manuscript had not yet been closely read paragraph by paragraph. Nor did it report extracting or analyzing a RRUFF sample dataset.

The counterargument was that even a shared final X-ray diffraction phase would not, by itself, establish a unique formation route. This linked the historical branch to the earlier inverse-problem example while leaving the two forms of evidence separate.

These two early branches should both remain visible. The later historical round describes its immediate predecessor as source-location work; that description does not erase the separate first-round mixing calculation and reproduction.

4. Round 2: underground growth, growth in glass, and salt crystallization

Question

When Newton used vegetation, was he merely assigning life to metals, or assembling a broader account of generation that linked processes underground, in glass vessels, and in aggregating crystals?

What was read and distinguished

The second round reported directly reading the normalized transcription of ALCH00081. It separated three contexts:

  1. The outline's third item concerns growth underground.
  2. The fourth concerns growth in glass.
  3. The fifth compares metallic trees with plants and animals.

The reading also identified passages near the second folio that describe salt particles gathering into larger groups, forming crystalline cubes, and crystallizing on sticks and sand. In the round's interpretation, these passages contain descriptions relevant to material accumulation, placed within a seventeenth-century framework of biological and chymical analogy.

That makes the manuscript more specific than the label “mystical living metals” would suggest. It does not make its vocabulary equivalent to modern crystallization theory.

Counter-reading

There were two distinct risks of overinterpretation:

The round rejected both moves. Salt, inferred metallic vapors, underground metal formation, and glass-vessel phenomena need to retain their different evidential roles.

Result and remaining work

The advance was an initial passage-based distinction between underground growth, vessel growth, and salt crystallization. It was not a full critical edition or complete philological study. The next task was to separate observations, analogies, and proposed causes at a finer scale.

5. Round 3: separating Newton's outline from later interpretation

The third round asked whether sharing a term such as plant growth and metal crystallization implied sharing a mechanism.

It returned to the manuscript outline, recording three locatable elements: item 3 on underground growth, item 4 on growth in glass, and item 6 on a latent spirit. It also read the editorial introduction and separated the modern historian's interpretation of a conceptual unifying account from Newton's own outline claims.

The recorded checks were that the three outline locations could be identified and that the primary outline was kept distinct from editorial interpretation. These are documentary checks. They are not experimental confirmations of the outline's proposed mechanisms.

The decisive boundary was semantic: latent spirit belongs to the historical explanatory framework. Renaming it as a measurable modern crystal-growth mechanism would erase the very distinction under investigation.

The next question was more precise than “Was Newton right?” It was how to classify individual claims about metallic trees and salts on folios 1r–2r, and how to detect a translation that silently imports a modern mechanism.

6. Round 4: a five-part audit of historical claims

The fourth round built a five-entry correspondence between a manuscript location, the type of claim, and a modern misreading to avoid. The following preserves its classifications while paraphrasing the manuscript wording.

1. Shared laws of vegetation

At folio 1r, outline item 2, Newton proposes that metals vegetate according to the same laws. The audit classifies this as a hypothesis. It is not evidence establishing biological growth in metals according to a modern definition of life.

2. Growth in glass

Outline item 4 names vegetation in a glass vessel. The audit treats this as a heading in a historical research program. A heading does not supply measured data that uniquely identify the origins of ores.

3. Latent spirit

Outline item 6 attributes vegetation to a latent spirit. The audit classifies this as a speculative mechanistic premise. It does not translate directly into a modern variable, reaction pathway, or kinetic law.

4. Metallic trees underground

Outline item 9 extends a manner of growth to metallic trees in the earth. The audit classifies the move as analogy or inference. Similar branching appearances do not establish identical kinetics between vessel deposits and geological veins.

5. Salt and stony substance

The salt passages discuss a hard, stony, transparent substance. The audit places this within natural-historical theory. It does not treat the statement as a completed account of modern silicate-mineral genesis.

What the audit did and did not decide

The round recorded that all five entries had primary-page locations and that claim types were distinguished. Classification nevertheless involves interpretation by editors and researchers. The manuscript outline cannot be converted into a record of five independently verified experiments merely by assigning each entry a category.

The resulting direction was to compare historically described appearances with modern electrodeposition, corrosion, and precipitation through observable quantities, rather than through a shared word alone.

7. Round 5: turning “growth” into a charge-to-mass prediction

Model choice

The fifth round selected a modern, operational relationship: accumulated charge and the ideal deposited mass of copper reduced from Cu(II). It used Faraday's law, with the simplifying assumption that all modeled charge contributes to this deposition reaction.

For charge Q, molar mass W, charge number z, Faraday constant F, and efficiency η:

m = ηWQ/(zF).

For constant current I over duration t, Q = It. The recorded inputs were:

The ideal deposited mass was:

m = 0.19758236387982903 mg, approximately 0.198 mg.

The accompanying code checked that the result lay between 0.19 and 0.21 mg. This is a theoretical charge-to-mass calculation, not a balance reading from an electrodeposition experiment.

Why an input control matters

The model gives an explicit relationship to controlled charge. Setting the modeled deposition input to zero removes that deposition term. Passive salt precipitation need not depend on externally imposed current, so this provides a possible way to discriminate candidate processes rather than relying on appearance.

The scope of that statement matters. It concerns the specified externally driven deposition model. It does not establish that all metal formation must stop when an external power supply is off, or that unmodeled reactions cannot occur.

The attack and next step

If current efficiency is below one, or material dissolves or corrodes after deposition, the ideal mass is no longer the net amount retained. Nucleation, morphology, and transport are also not determined by the charge-to-mass equation alone.

The round proposed comparing a modern dataset containing matched current, mass, morphology, and deviations from the simple model. A 2024 electrodeposition study was recorded as the literature entry point, but the round's 0.198 mg figure was its own illustrative calculation; it was not a result extracted from that paper.

8. Round 6: deposition efficiency and dissolution can hide each other

LESS DEPOSITION, NO LOSS

60% − 0% = 60%

MORE DEPOSITION, SOME LOSS

80% − 20% = 60%

Synthetic copper example. Calculated values, not laboratory measurements. Every percentage uses the ideal deposited mass as its denominator.

Replacing ideal mass with net mass

The sixth round retained the same ideal mass m₀ = 0.19758236387982903 mg and wrote:

m_net = ηm₀ − L,

where L is the mass lost through dissolution.

Two synthetic cases give the same retained amount:

The final decimal difference is floating-point rounding. The formula makes the two cases equal: 0.60m₀ = 0.80m₀ − 0.20m₀. The recorded code asserted agreement within 10⁻¹³ mg.

This is a mechanistic ambiguity, not merely poor measurement precision. Even a perfect measurement of the shared net mass cannot choose between these two parameter combinations.

A separate issue: resolving small differences

If dissolution is absent, increasing efficiency from 60% to 90% changes net mass by:

(0.90 − 0.60)m₀ = 0.05927470916394872 mg.

The round compared this with a 0.1 mg display increment. Such an increment is coarser than the predicted difference and does not support reliable separation of these cases from a single rounded reading. This resolution observation is separate from the structural nonuniqueness above; instrument resolution, accuracy, and uncertainty are not interchangeable.

What the model omits

The round did not model overpotential, mass transport, nucleation, concentration changes, or morphology. It also did not establish that the chosen 1 mA condition corresponds to any experiment described by Newton.

A proposed measurement that required correction

The sixth-round evidence proposed measuring solution-phase copper concentration alongside deposited mass to separate efficiency from dissolution. The seventh round tested that proposal and found that, under a closed-cell mass balance, it can fail. The correction is part of the result, rather than an omission to be hidden.

9. Round 7-A: two measurements can contain only one piece of information

Testing the proposed extra observation

Assume a closed cell in which copper moves only between the solution and the deposit. Let D denote gross deposition and R redissolution, expressed in the same copper-mass units.

Electrode mass gain is y₁ = D − R.

Solution copper depletion, after converting concentration and volume into mass, is y₂ = D − R.

The observation matrix is:

A = [[1, −1], [1, −1]].

Its rank is one. The second row repeats the first. It may provide a useful mass-balance consistency check, but it does not separately identify D and R in this stipulated system.

This directly revises the sixth-round measurement suggestion. More instruments do not automatically provide more independent mechanistic information.

Introducing time structure

The archived seventh-round package instead introduced an on/off input and a minimal dynamic model:

dM/dt = ηq(t) − kM.

Here M is deposited mass, q(t) is the ideal deposition mass rate corresponding to the imposed input, η is constant current efficiency, and k is a first-order dissolution coefficient with units of inverse time.

The assumptions are:

The last two are physical assumptions to check, not consequences of the algebra.

Solving the forward model

For an on duration t_on and k > 0:

M_on = ηq[1 − exp(−kt_on)]/k.

After a further off duration Δt:

M_off = M_on exp(−kΔt).

The off interval isolates the modeled decay term because the deposition input is set to zero.

Recovering the parameters

For positive M_on and M_off and nonzero Δt:

k = ln(M_on/M_off)/Δt.

Substituting this value into the on-interval equation gives:

η = M_on k / {q[1 − exp(−kt_on)]}.

At k = 0 the apparent division by zero is removed by the continuous limit:

η = M_on/(qt_on).

These formulas show why the temporal intervention adds information that the duplicate conservation measurement did not.

The archived synthetic recovery test

The code used q = 1, t_on = 10, Δt = 3, η = 0.65, and k = 0.12 in consistent illustrative units. It generated:

The test required both recovered parameters to agree with the supplied parameters within 10⁻¹². The example does not assign these masses to milligrams or the durations to seconds, and it is not a continuation of the 1 mA, 600 s laboratory-scale scenario. Its role is to test algebraic recovery from exact synthetic data.

The counterexample to unqualified use

Interrupting current can itself change dissolution kinetics. If k_off differs from k_on, the mass ratio estimates the off-period coefficient; substituting it as the on-period coefficient can bias the inferred efficiency.

Thus the round obtained a conditional identification method, not a universal recipe. The next question is what additional time point or genuinely different observation is needed when the on and off coefficients are allowed to differ. The recovered package did not complete that extended model or fit public experimental data.

10. Round 7-B: a separate constant-dissolution model

Why this version must remain separate

A standalone seventh-round research card describes a different model from the archived package. It uses a fixed dissolution rate, whereas Round 7-A uses dissolution proportional to the amount already deposited. Both cards identify the same sixth-round parent. The available record does not establish that one formally supersedes the other.

They are therefore preserved here as parallel seventh-round versions. Their rate parameters must not be interchanged: a fixed mass-loss rate and a first-order coefficient have different dimensions and imply different trajectories.

The standalone card's substantive result

The standalone card reports writing net mass as efficiency times passed charge, with the appropriate charge-to-mass conversion, minus a constant dissolution rate times elapsed time. It reports a symbolic determinant, an independent linear solve, and an error-propagation check.

Its central finding is that repeated weighing at the same average current does not identify efficiency and constant dissolution separately. Two intervals with different average currents can identify them under the constant-parameter assumptions.

Making the reported design condition explicit

The following algebra spells out the model described by that card; it is an explanatory reconstruction, not a recovered numerical test file.

Let c = W/(zF) be the ideal copper mass per unit charge in the chosen units. For two intervals, let Q₁ and Q₂ be passed charge, t₁ and t₂ their durations, m₁ and m₂ their net mass increments, and r a constant dissolution mass rate:

m₁ = ηcQ₁ − rt₁

m₂ = ηcQ₂ − rt₂.

The design matrix is [[cQ₁, −t₁], [cQ₂, −t₂]]. Its determinant is:

c(Q₂t₁ − Q₁t₂).

For positive durations, this is nonzero precisely when Q₁/t₁ differs from Q₂/t₂: the intervals must have different average currents. Merely waiting longer at the same current scales the same equation rather than adding an independent one.

Writing average currents Iᵢ = Qᵢ/tᵢ and average net growth rates gᵢ = mᵢ/tᵢ gives the same result transparently:

g₁ = ηcI₁ − r

g₂ = ηcI₂ − r

η = (g₂ − g₁)/[c(I₂ − I₁)]

r = ηcI₁ − g₁.

The recorded attack and validation proposal

If efficiency or dissolution changes with current, a two-constant model is inadequate. An unknown weighing offset adds an unknown parameter when the observations are absolute masses; the exact consequence depends on whether increments or absolute readings are modeled.

The card therefore proposed reserving a third interval as a check not used in fitting. Two points can exactly determine two parameters without testing whether the mechanism is correct. A held-out third observation tests a further prediction of the fitted model.

The card's supporting numerical artifact was named but was not recovered with the material used for this reconstruction. Its exact synthetic inputs, uncertainty assumptions, and error-propagation outputs are therefore not reproduced as verified results here. The model, reported design conclusion, and counterarguments are preserved; numerical details have not been invented to fill the gap.

11. What changed, what remains open

Across the rounds, progress came from narrowing claims and improving the question:

The immediate research gap is not to repeat the same mass example with new numbers. It is to connect a clearly selected model to matched observations and test the assumption that makes the inverse problem solvable.

For the first-order model, the open issue is the relation between on- and off-period dissolution. For the constant-rate model, it is whether parameters remain constant across the changed-current intervals, and how measurement offsets and uncertainty affect the design. For both, a same-experiment record of current, timing, retained mass, solution composition, and relevant conditions could allow direct model checking; the existence of multiple columns alone would not guarantee independent information.

The historical branch also remains open. Its recovered work is a passage-level reading and classification of selected manuscript material, not a complete study of Newton's mineral theory across manuscripts. That work can continue in its own right without requiring every modern model to return to a historical analogy.

12. Appendix: equations, checks, and public research references

Equation and unit guide

What each recovered check actually tests

The first-round rational examples test equality of synthetic weighted signals. The determinant and row reduction test uniqueness within a specified linear system. The historical checks test document locations and classification, not physical truth. The fifth-round check tests a charge-to-mass calculation. The sixth-round assertion tests equality of two modeled net masses. The archived seventh-round rank check tests redundancy of a closed-system observation matrix, and its recovery check tests inversion of noiseless synthetic trajectories.

The original rounds used exploration followed by methodological counterargument in a single research context. These checks do not become independent experimental replications merely because separate cards or algorithms recorded them.

Public references retained from the original research

These are the public source entries recorded in the recovered rounds. They distinguish historical primary material, editorial interpretation, mineral-data entry points, and modern methodological context. They are not presented as newly reviewed sources or as external endorsements of the synthetic results.