Every measurement anyone has ever trusted is the last link in a chain, and for most of the history of science that chain ended at a piece of metal in a vault. A platinum-iridium bar in a basement outside Paris was, for over a century, the length of one metre by definition — not an approximation of it, the thing itself. If the bar had been damaged, the metre would have changed with it. That arrangement sounds fragile because it was fragile, and the history of scientific instruments and metrology is largely the story of engineering that fragility out of measurement, first by agreeing internationally on what the standards meant, then by building instruments precise enough to compare against them, and finally by discarding the artifacts altogether in favor of constants of nature that cannot be dropped, worn, or contaminated.
This is a history with three real hinge points, each independently verifiable and each solving a different failure mode of the one before it. The 1875 Metre Convention solved the problem of many nations keeping incompatible rulers. The invention of the electron microscope in the early 1930s solved the problem of a fundamental limit on how small an object light itself could resolve. And the 2019 redefinition of the International System of Units solved the problem the original metre bar embodied from the start: that a physical artifact is not actually a fixed reference, only a very good one that everyone had agreed not to look at too closely.
Fact: the treaty that made measurement international
Before 1875, “a metre” or “a kilogram” was whatever a national government’s own reference object said it was, and those objects disagreed with each other by amounts that mattered as trade and science both globalized. The fix was diplomatic before it was technical. On 20 May 1875, seventeen states signed the Metre Convention, an international treaty committing its signatories to “assure the international unification and improvement of the metric system” [1]. The treaty’s signatories included Germany, France, Italy, the United States, Russia, Spain, Switzerland, Austria-Hungary, Belgium, Brazil, Denmark, Peru, Portugal, Sweden and Norway, the Ottoman Empire, and Venezuela — already a genuinely global spread, not a European club [1].
The treaty’s durable output was institutional, not a single artifact: it created the Bureau International des Poids et Mesures (BIPM), a standing international body tasked with keeping measurement standardized worldwide, and it established the governance around it — the General Conference on Weights and Measures (CGPM), which meets periodically to approve changes to the system, and the International Committee for Weights and Measures (CIPM), which oversees the technical work between conferences [1]. That three-layer structure — a treaty organization, a periodic political conference, and a standing technical committee — is still exactly how the SI is governed a century and a half later, and it is worth naming as the real invention of 1875. The prototype bars distributed to member states after the convention were a consequence of this agreement, not its substance; the substance was that for the first time, one organization’s definition of a metre was binding on laboratories in Washington, Tokyo, and Buenos Aires alike.
It is easy to read the Metre Convention as a footnote about weights and measures. It is better read as the moment measurement became explicitly a matter of trust between institutions rather than a private claim by any one nation, and every later development in this history — from interlaboratory comparisons to the constants-based SI — is an extension of that same idea: a measurement is only as good as the chain of agreement standing behind it.
Fact: the metre outgrew its own bar
The platinum-iridium bar that resulted from the 1875 agreement defined the metre by direct comparison until 1960, when the 11th CGPM redefined the metre instead as a fixed number of wavelengths of light from a krypton-86 transition — 1,650,763.73 of them, to be exact [2]. This was already the discipline of preferring a reproducible natural phenomenon over a single guarded object, but krypton-86 interferometry still had a practical ceiling: it could not be realized more accurately than about four parts in a billion [2].
The 1983 redefinition solved that ceiling by inverting the relationship between distance and speed entirely. Rather than measuring how far light travels and calling that a fixed length, the 17th CGPM fixed the speed of light itself — at the value already recommended in 1975, 299,792,458 metres per second — and defined the metre as whatever distance light travels in vacuum in exactly 1/299,792,458 of a second [2]. From 1983 onward, improving the precision of length measurement was no longer a matter of building a better ruler; it was a matter of building better clocks, because a metre had become, by definition, a fixed fraction of a second’s light-travel.
That inversion — trading an artifact for a constant, and a spatial measurement for a temporal one — is the same move the rest of the SI would make with mass, and it did not happen by accident. It happened because krypton interferometry had shown the ceiling every artifact-based standard eventually hits: the instrument used to compare against the standard becomes the limiting factor, not the standard itself.
Fact: seeing past the limit of light
While the metre was being redefined, an entirely separate instrument problem was being solved a few hundred kilometres away in Berlin. Optical microscopes have a real, physical resolution limit set by the wavelength of visible light — no amount of engineering refinement lets a light microscope resolve detail finer than roughly half that wavelength. By the early twentieth century, biologists and materials scientists had a large catalogue of structures they knew existed — viruses, the fine structure of metal grain boundaries, the internal architecture of cells — that sat below that limit and were, for a light microscope, permanently unreachable.
The way past the limit was to stop using light. Electrons, treated as waves, have a much shorter effective wavelength than visible light, and a beam of electrons can be focused with magnetic fields the way a beam of light is focused with glass lenses. Ernst Ruska, a physicist, and Max Knoll, an electrical engineer, both at the University of Berlin, built the first electron microscope in 1931 [7]. That first instrument was a proof of concept rather than a usable tool — its total magnification was only about sixteen times, no better than an ordinary light microscope, and it demonstrated the principle without yet beating the instrument it was meant to replace. The breakthrough came two years later: by 1933, Ruska had built a second-generation column using dual magnetic coils as lenses, arranged by analogy with a compound optical microscope, that reached a magnification around 12,000 times and, for the first time, resolved detail beyond what any light microscope could show [7].
The electron microscope’s history is worth separating carefully from vendor-style retelling. What is verifiable fact is the 1931 prototype, the 1933 instrument’s magnification figures, and Ruska and Knoll’s institutional affiliation [7]. What followed — decades of refinement into transmission and scanning electron microscopy, and eventually cryo-electron microscopy — is a long engineering elaboration of the same principle, not a repeat of the original invention, and this article makes no claim about which later refinement mattered most; that is a separate, contested history of its own.
The connection to metrology is direct rather than metaphorical. An electron microscope is a length-measuring instrument as much as an imaging one — its magnification claim is itself a calibrated quantity, checked against a certified reference structure with known dimensions, the same discipline the rest of this article traces at the level of the metre and the kilogram. An uncalibrated electron microscope produces a picture; a calibrated one produces a measurement.
Fact: the second, tied to a caesium atom instead of the turning Earth
Time measurement went through its own version of the same shift, on almost the same timeline as the metre’s krypton-86 redefinition. Until the mid-twentieth century, the second was defined astronomically — as a fraction of the mean solar day, ultimately tied to Earth’s rotation. But Earth’s rotation is not perfectly uniform, which means a definition of time based on it inherits that irregularity.
Louis Essen and Jack Parry, working at the UK’s National Physical Laboratory, brought the first caesium atomic clock into operation in 1955, with the apparatus first working on 24 May and continuous operation beginning that June [6]. The device was a roughly two-metre horizontal apparatus: a source of caesium atoms at one end, a microwave cavity in the middle used to probe a specific atomic transition frequency, and a detector at the far end [6]. It was accurate to about one millisecond a day — equivalent to about one second of drift every 300 years — which made it roughly thirty times better than the best quartz clocks of the era and some three hundred times better than the most accurate pendulum clocks [6]. On 3 June 1955, the BBC’s time signal incorporated input from Essen’s clock, putting atomic time into public use for the first time [6]. In 1967, the second’s official definition changed to match: rather than a fraction of the solar day, it became a fixed count of periods of the radiation corresponding to a specific caesium-133 transition — the same transition Essen and Parry’s instrument had been built to probe twelve years earlier [6].
This is the same underlying move as the metre’s history: an artifact-adjacent, irregular reference (the rotating Earth) replaced by a reproducible atomic phenomenon that any sufficiently good instrument, anywhere, can in principle reproduce without appeal to a single guarded object.
Analysis: why every one of these redefinitions has the same shape
Read side by side, the metre’s 1960 and 1983 redefinitions, the second’s 1967 redefinition, and — as the next section covers — the kilogram’s 2019 redefinition are not four unrelated stories. They are the same argument applied to four different quantities, and it is worth stating the argument plainly rather than letting it stay implicit in the chronology.
An artifact-based standard has exactly one copy that counts, wherever it happens to be kept, and every other instance is a claim about how well a copy matches that one object. This has two failure modes. First, the artifact itself can drift — through contamination, wear, or simply the slow uncertainty of measuring it against itself over decades — and there is no external check available, because it is the definition, not a measurement of something else. Second, access to the artifact is a bottleneck: a laboratory verifying its own standards ultimately has to trace back to one vault, in one country, guarded by one institution.
A constant-based standard has neither problem. The speed of light, the caesium-133 hyperfine transition frequency, and — since 2019 — the Planck constant do not sit in a vault; they are properties of nature that any suitably equipped laboratory can in principle realize independently, and disagreements between laboratories become genuine experimental questions with error bars, rather than disputes about the condition of a single object. The cost of this move is that it requires much harder instruments — an electron microscope’s precision-scale cousin, an interferometer, a caesium beam apparatus, and eventually a Kibble balance — capable of connecting an abstract constant back to a usable, everyday reading. The 1875 Metre Convention could not have made this move in 1875; the instruments did not yet exist. The story of the twentieth century is those instruments catching up to the ambition the treaty had already set.
Fact: the kilogram was the last artifact standing, and the Kibble balance replaced it
By the early twenty-first century, the kilogram was the only base SI unit still defined by a single physical object: the International Prototype Kilogram, a platinum-iridium cylinder kept at the BIPM outside Paris. Comparisons among it and its official copies over the twentieth century showed relative drifts on the order of tens of micrograms — small, but nonzero, and with no external constant to check the primary artifact against, because it was the definition. The kilogram needed the same move the metre and the second had already made: tie it to a constant instead of an object.
The instrument that made this possible is the Kibble balance, originally called the watt balance when Bryan Kibble invented it in 1975 to improve realizations of the ampere, and renamed in 2017 to credit Kibble after his death in 2016 [5]. A Kibble balance measures the weight of a test mass by balancing it against an electromagnetic force generated by passing a current through a coil in a surrounding magnetic field, then relates that force to a directly measurable electrical power, in watts [5]. The apparatus is run in two modes: a “weighing mode,” in which the current needed to support the test mass’s weight is measured, and a “velocity mode,” in which the same coil is moved through the same field and the induced voltage is measured — and the two measurements combine so that the mechanical unknowns of the magnetic field and coil geometry cancel out algebraically, leaving a relationship purely between electrical quantities and Planck’s constant [5].
Because voltage and resistance can themselves be measured to extraordinary precision using quantum electrical effects (the Josephson effect and the quantum Hall effect), a Kibble balance effectively measures mass in terms of the Planck constant, h. The relationship, in its simplified form, is:
where
Fact: 20 May 2019, four base units, seven constants
On 16 November 2018, representatives of 60 countries meeting at the 26th CGPM in Versailles voted unanimously to redefine the SI [4]. The change took effect on 20 May 2019 — deliberately chosen as World Metrology Day, the anniversary of the original 1875 treaty [4]. Four of the seven base units — the kilogram, the kelvin, the ampere, and the mole — were redefined in terms of fixed values of fundamental constants; the second, the metre, and the candela had already made an equivalent move in earlier decades and needed no further change [4]. NIST describes the full set of seven defining constants as including the speed of light, the elementary electric charge, and the Planck constant, among others, and frames the result as realizing “a 150-year dream of a measurement system based entirely on unchanging fundamental properties of nature” [4].
That framing is worth taking literally rather than as promotional flourish: the “150 years” traces back almost exactly to the 1875 Metre Convention, and the throughline from that treaty to the 2019 vote is continuous institutional custody, not a coincidence of round numbers. The same CGPM that the 1875 treaty created is the body that approved the 2019 change; the same BIPM that the treaty established is the body that now maintains the realizations of the redefined units.
Analysis: a redefinition changes nothing and everything, on purpose
A frequently repeated, and accurate, characterization of the 2019 change is that ordinary users of the SI noticed nothing: a kilogram measured the day after the redefinition weighed the same, within experimental uncertainty, as a kilogram measured the day before. That is not an accident or a coincidence — it is a design constraint the CGPM imposed on itself. The values of the fixed constants were chosen specifically so that the redefined units matched the best available measurements of the old artifact-based units within their stated uncertainties. The redefinition changed who is authoritative when a measurement is disputed and how the base unit can be independently reproduced; it deliberately did not change the practical size of a kilogram, a kelvin, an ampere, or a mole for anyone not operating at the frontier of precision measurement.
What did change, for laboratories capable of building a Kibble balance or an equivalent atomic-counting mass realization, is that the kilogram stopped requiring physical access to one specific cylinder in one specific vault. Multiple national metrology institutes can now realize the kilogram independently and compare results, the same freedom the metre gained in 1983 and the second gained in 1967.
Fact: keeping thousands of laboratories on the same page
A definition fixed at a treaty conference is not, by itself, a working measurement system. The gap between “the kilogram is defined by a fixed value of the Planck constant” and “a factory’s scale reads correctly” is filled by an unglamorous but essential layer: calibration chains and formal interlaboratory comparisons. The mechanism the BIPM uses for this is the CIPM Mutual Recognition Arrangement (CIPM MRA), the framework through which national metrology institutes demonstrate that their measurement standards are equivalent to each other and mutually accept the calibration certificates each issues [9].
The technical backbone of the CIPM MRA is the Key Comparison Database, or KCDB, which as of recent BIPM figures covers more than 1,500 scientific comparisons, roughly two-thirds of them formal “key comparisons” between national metrology institutes and about a third supplementary comparisons [8]. Participating laboratories must operate under a quality system consistent with ISO/IEC 17025, and their claimed calibration and measurement capabilities are peer-reviewed, taking into account each laboratory’s demonstrated performance in these comparisons [8]. In practice, a key comparison works by circulating a single reference artifact — a mass standard, a voltage reference, a length gauge — among participating national laboratories in sequence, each measuring it independently and reporting a result with its own stated uncertainty; the spread and consistency of those results is the actual empirical evidence that “one kilogram” means the same thing in Washington, Braunschweig, and Tsukuba.
This is where the abstraction of a constants-based SI meets the discipline that predates it: a calibration is only meaningful if its uncertainty is stated and its traceability is documented back through an unbroken chain of comparisons to a primary realization. Reproducibility, in this sense, is not a virtue metrologists profess — it is the entire measurable output of the system. A laboratory’s calibration and measurement capability claim is worth exactly as much as its performance record in the KCDB says it is, nothing more.
Scenario: what an unbroken chain looks like end to end, and where it can still break
Consider a concrete, illustrative chain, described here as a scenario rather than a specific documented case: a hospital’s blood-analysis instrument reports a patient’s glucose concentration. That reading is traceable, in principle, through a sequence of calibrations — the hospital’s instrument checked against a reference material from an accredited calibration laboratory, that laboratory’s reference checked against a national metrology institute’s primary standard, and that standard’s realization ultimately checked against the SI’s defining constants through the kind of atomic and quantum-electrical measurements described above. Each link in that chain carries its own uncertainty budget, and the uncertainties compound rather than cancel, so the hospital’s reported number is less certain than the national institute’s primary realization by a knowable, quantifiable amount.
The place this chain most often breaks in practice is not at the level of fundamental constants — it is in the middle, at calibration intervals that lapse, reference materials that degrade faster than assumed, or a laboratory’s quality system drifting out of the condition it was accredited under. The CIPM MRA’s ongoing key comparisons exist precisely to catch this kind of drift statistically, by re-checking whether a laboratory’s results are still consistent with its peers, rather than assuming a single accreditation event holds indefinitely. This scenario is offered as an illustration of how the system is meant to function, not as a claim about any specific institution’s actual practice.
Prediction: where the next redefinition pressure will come from
A reasonable, horizon-bounded prediction: the next serious pressure for an SI-level change is more likely to come from the second than from any other unit, on a horizon of the next one to two decades. Optical atomic clocks, which use visible-light transitions rather than caesium’s microwave transition, have already demonstrated stabilities and systematic uncertainties exceeding the caesium standard that currently defines the second by a wide margin in laboratory demonstrations. The assumption underlying this prediction is that at least one class of optical clock reaches a level of international reproducibility, through the same kind of key-comparison process described above, that the metrology community judges good enough to serve as a primary realization rather than a secondary one.
Observable indicators that would support this prediction: a CIPM Consultative Committee for Time and Frequency formally recommending a specific optical transition as a new primary second-defining standard ahead of a CGPM vote, and multiple independent national metrology institutes publishing key-comparison results for that transition that agree within the target uncertainty. The prediction would be disconfirmed, or at least substantially delayed, if optical clock comparisons across institutes continue to show systematic disagreements larger than caesium’s own uncertainty, or if no consensus emerges on which of several competing optical transitions should serve as the new reference — a genuine, unresolved technical disagreement in the field at present, and one this article does not attempt to resolve in either direction.
The instrument is the argument
The thread connecting a 1875 treaty, a 1931 electron microscope, a 1955 caesium clock, and a 2019 vote in Versailles is not really about units at all. It is about what counts as a fact in an experimental science: a number is only as trustworthy as the documented chain connecting it back to something the whole world has agreed to treat as fixed, with every step of that chain carrying an honestly stated uncertainty. The Metre Convention built the institutions to make that agreement international. The electron microscope and the caesium clock are examples of instruments built specifically to reach past a previous instrument’s fundamental limit. The Kibble balance and the 2019 redefinition are the moment the last physical artifact in the system was retired in favor of the constants it had always been an imperfect stand-in for. None of this was inevitable, and none of it is finished — the KCDB’s ongoing comparisons and the live debate over optical clocks are the same nineteenth-century argument, still being carried out one calibration at a time.
