A thermometer is an instrument that measures temperature by relating it to an observable physical property, such as the volume of a liquid, the electrical resistance of a metal, or the intensity of emitted radiation. Its essential elements are a temperature-sensitive system and a means of converting that system’s response into a numerical indication. These elements may be contained in a single glass instrument or distributed among a probe, electronic circuitry, software, and a remote display.
Thermometers make temperature comparisons reproducible and support experimental science, industrial production, meteorology, and medicine. They do not directly measure heat, which is energy transferred because of a temperature difference. Nor does every thermometer measure precisely the same aspect of a target: an immersed probe responds to its own local temperature, while a radiation thermometer infers temperature from radiation received from a surface.
Thermometry combines instrument design with metrology, the science of measurement. Reliable results depend on the sensing principle, the temperature scale, calibration, and the exchange of energy between instrument and surroundings. The distinction between a sensor’s indication and the temperature actually sought is central to both ordinary measurements and the realization of international temperature standards.
Historical development
Early thermometers developed from the thermoscope, a device that showed changes in temperature without necessarily assigning numbers to them. Instruments associated with Galileo Galilei around the end of the sixteenth century used the expansion and contraction of air to move liquid in a tube. Their operation demonstrated that warming could produce a repeatable mechanical effect, but open air thermoscopes also responded to changes in atmospheric pressure. They therefore could not unambiguously distinguish a temperature change from a pressure change.
During the seventeenth century, instrument makers introduced graduated scales and sealed liquid-filled tubes. Santorio Santorio used graduated thermoscopic devices in physiological investigations. In Florence, instruments developed under Ferdinando II de’ Medici in the mid-seventeenth century employed sealed glass vessels containing alcohol. Sealing separated the working liquid from the atmosphere and removed the most direct source of pressure interference. Nevertheless, instruments differed in their glass, liquid composition, bulb dimensions, and scales, so readings were not automatically comparable between workshops.
Daniel Gabriel Fahrenheit made important advances in reproducible mercury thermometers during the early eighteenth century. Mercury offered a clearly visible column, did not wet clean glass in the same way as many other liquids, and remained liquid across a useful temperature range. Fahrenheit described his temperature scale in 1724. On the modern Fahrenheit scale, the freezing and boiling points of pure water at standard atmospheric pressure are 32 °F and 212 °F. Historical procedures for establishing the scale differed from its later standardized definition.
Anders Celsius proposed a centigrade scale in 1742, originally placing zero at the boiling point of water and 100 at its freezing point. The direction was subsequently reversed in scientific practice, giving the familiar arrangement of 0 °C for freezing and 100 °C for boiling under specified conditions. Neither point is independent of circumstances: boiling temperature is particularly sensitive to pressure, while impurities can alter both freezing and boiling behavior.
Nineteenth-century developments connected thermometry to thermodynamics and introduced electrical measurement methods. Thomas Johann Seebeck discovered the thermoelectric effect in 1821, providing the basis of thermocouples. William Thomson, later Lord Kelvin, proposed an absolute thermodynamic temperature scale in 1848. Platinum resistance thermometry developed through work by investigators including William Siemens and Hugh Longbourne Callendar. In clinical practice, Thomas Clifford Allbutt’s compact thermometer, introduced in the 1860s, made bedside temperature measurement more convenient. Twentieth-century electronics subsequently enabled automated recording, miniature sensors, and noncontact instruments, while international agreements progressively separated the definition of temperature from the behavior of any single working substance.
Temperature scales and primary thermometry
The SI unit of thermodynamic temperature is the kelvin, symbol K. Since 20 May 2019, it has been defined by fixing the numerical value of the Boltzmann constant at exactly
This definition connects temperature to an energy scale rather than to a particular material artifact or fixed-point cell. It belongs to the International System of Units and does not require any one type of thermometer for its practical realization.
Celsius temperature and thermodynamic temperature have the exact relationship
A temperature interval of 1 °C therefore equals an interval of 1 K. The kelvin does not take a degree sign. Fahrenheit temperature is related to Celsius temperature by , where the numerical values are expressed in their respective units. Absolute zero is 0 K, equivalent to −273.15 °C. It is not defined as the complete absence of all microscopic motion; quantum systems may retain zero-point energy.
A primary thermometer determines thermodynamic temperature from a sufficiently well-understood physical relationship without requiring calibration against another thermometer at the temperature being measured. Gas thermometry provides an important example. In the dilute-gas limit, the ideal-gas relationship is
where is pressure, is volume, and is the number of particles. Real gas thermometry requires corrections for interactions between particles, apparatus deformation, and other departures from the ideal model. Historically, measurements made at progressively lower gas densities helped establish thermodynamic temperature.
Modern primary techniques include acoustic gas thermometry, dielectric-constant gas thermometry, Johnson-noise thermometry, and Doppler-broadening thermometry. Acoustic instruments infer temperature from the speed of sound in a gas. Johnson-noise thermometers use the thermal electrical fluctuations of a resistor: in the classical low-frequency regime, the voltage-noise spectral density of a resistance is . Doppler methods obtain temperature from the thermal broadening of spectral lines, linking thermometry with spectroscopy. Each method has a detailed physical model and substantial experimental corrections.
Most practical instruments instead realize an agreed scale through calibration. The International Temperature Scale of 1990, or ITS-90, assigns temperatures to specified equilibrium states and prescribes interpolation instruments and equations. It extends from 0.65 K upward through several overlapping methods. The triple point of water has an assigned ITS-90 temperature of 273.16 K, although since 2019 it no longer defines the kelvin. Practical-scale temperature and thermodynamic temperature are closely related but not identical; their small differences matter in precision research.
Liquid expansion, bimetallic, and pressure thermometers
A liquid-in-glass thermometer consists of a reservoir, usually called the bulb, connected to a narrow capillary. As temperature rises, the liquid generally expands more than the surrounding glass, causing the column to rise. For a limited range, the height change can be approximated by
where is the bulb volume, is the capillary’s cross-sectional area, and and are the volume-expansion coefficients of the liquid and glass. A larger bulb and narrower bore increase sensitivity, but can also increase fragility or response time.
Mercury and colored organic liquids have been widely used as thermometric fluids. Mercury freezes at about −38.8 °C and boils at about 356.7 °C at atmospheric pressure. These phase boundaries constrain its ordinary working range, although specialized pressurized designs can extend the upper limit. Organic liquids permit measurements at lower temperatures but have different expansion behavior, wetting characteristics, and stability limits. The scale of a liquid thermometer must therefore correspond to its particular filling and construction rather than to an assumed universal expansion law.
The immersion condition is part of a liquid-in-glass thermometer’s specification. A total-immersion instrument is calibrated with essentially the whole liquid column at the measured temperature, whereas a partial-immersion instrument is calibrated to a designated depth. If a substantial length of exposed liquid column remains cooler or warmer than intended, the reading changes. Other error sources include nonuniform capillary bore, gradual changes in bulb volume, separation of the liquid column, and parallax when the scale is viewed obliquely. Traditional maximum clinical mercury thermometers include a constriction that retains the maximum reading after removal.
A bimetallic strip joins two metals with different thermal expansion coefficients. Heating bends the composite because one layer tends to lengthen more than the other. Coiling the strip into a spiral or helix converts this bending into rotation of a pointer. Such thermometers require no electrical supply and are common in dial instruments, ovens, and mechanical controls. Their limitations include mechanical hysteresis, friction, aging, and imperfect transfer of heat from the measured medium to the sensing element.
Filled-system pressure thermometers use a bulb connected by a capillary to a pressure-responsive mechanism, commonly a Bourdon tube. A gas-filled system exploits the temperature dependence of gas pressure; a liquid-filled system converts expansion into pressure; a vapor-pressure system uses the saturation pressure associated with liquid–vapor equilibrium. The latter relationship is strongly nonlinear and depends on the working fluid. These instruments can place the display away from the sensing bulb, but ambient temperatures along the capillary and at the indicating mechanism may affect the reading. Compensation mechanisms and restrictions on installation address these influences.
Electrical resistance, thermocouples, and semiconductor sensors
Electrical thermometers convert a temperature-dependent electrical property into a measured signal. A resistance thermometer uses the variation of electrical resistance with temperature. Platinum is especially important because of its chemical stability and reproducible behavior. A common industrial element, designated Pt100, has a nominal resistance of 100 Ω at 0 °C. A widely used standardized platinum characteristic has an average temperature coefficient of approximately between 0 °C and 100 °C, although precision interpolation uses equations rather than a constant coefficient.
Lead resistance can be significant compared with the change being measured. In a two-wire arrangement, connecting-wire resistance adds directly to the sensor resistance. Three-wire circuits compensate under assumptions about matched leads, while four-wire measurements separate the current-carrying connections from the voltage-sensing connections. The measuring current itself produces Joule heating, warming the sensor above its surroundings. Precision measurements may compare results at different currents and extrapolate toward zero measuring power. Mechanical strain, contamination, and thermal cycling can also alter a platinum element’s calibration.
A thermistor is a temperature-sensitive resistor, usually made from semiconductor materials. Negative-temperature-coefficient thermistors decrease in resistance as temperature rises. A useful approximate model is
where temperatures are in kelvins and is a material-dependent parameter. More accurate conversions often use the Steinhart–Hart equation. Thermistors can be small and highly sensitive over restricted ranges, making them useful in electronic thermometers and equipment monitoring. Their nonlinearity, self-heating, and long-term stability require consideration; high sensitivity alone does not establish high measurement accuracy.
A thermocouple consists of two dissimilar electrical conductors joined in a measurement circuit. Its thermoelectric voltage depends on the temperature distribution and, for homogeneous conductors, can be expressed in terms of measuring and reference junction temperatures:
where and are the materials’ Seebeck coefficients. The signal therefore does not give the measuring-junction temperature unless the reference condition is known. Modern instruments usually measure terminal temperature and apply electronic reference-junction compensation. The reference junction need not be physically cold.
Standardized thermocouple types use specified material combinations. Type K employs nickel-based alloys and is widely used in industrial measurements; types R and S use platinum–rhodium combinations for higher-temperature work. Actual limits depend on atmosphere, wire diameter, insulation, and service duration. Semiconductor integrated-circuit sensors provide another approach, using predictable temperature dependences of junction voltages or related circuit quantities. Their digital interfaces simplify data acquisition, but package heating and thermal coupling remain physical constraints. Thermocouples, resistance probes, and integrated sensors can all provide digital readings; “digital thermometer” describes the readout, not a unique sensing principle.
Radiation thermometers and thermal imaging
A radiation thermometer infers temperature from electromagnetic radiation received from a target. It can operate without touching the object, making it valuable for moving machinery, hot furnaces, fragile surfaces, and electrically hazardous equipment. The physical reference is blackbody radiation. An ideal blackbody’s spectral radiance follows Planck’s law:
where is the Planck constant, is the speed of light, and is wavelength. Increasing temperature changes both the intensity and the spectral distribution of emitted radiation.
Real surfaces generally emit less radiation than a blackbody at the same temperature. Their emissivity depends on material, wavelength, surface finish, viewing direction, and sometimes temperature. For an opaque surface viewed through a nonabsorbing path, the detected radiance can be represented approximately as emitted radiation plus reflected environmental radiation:
where is emissivity, is surface temperature, and represents the surrounding radiance reflected toward the instrument. Low-emissivity polished metals can therefore produce readings strongly influenced by nearby hot or cold objects. An emissivity setting is part of a measurement model, not merely a display adjustment.
Infrared thermometers often operate within selected atmospheric transmission bands, including the approximate 8–14 μm region for many near-ambient applications. Optical pyrometers and short-wavelength radiation thermometers are particularly useful at high temperatures. Ratio or two-color instruments compare signals in two spectral bands. They can reduce certain effects of changing target size, obscuration, or emissivity, but only when the relevant assumptions hold. If emissivity varies differently between the two bands, the ratio can produce a substantial error.
The optical field of view also matters. A target must adequately fill the instrument’s effective measurement area, including contributions from optical scattering beyond a nominal spot boundary. Distance-to-spot specifications describe spatial sampling but do not remove all size-of-source effects. Windows, smoke, flames, and atmospheric gases may absorb or emit radiation along the optical path. Ordinary window glass is not transparent across the long-wave infrared band used by many handheld thermometers, so such an instrument directed at a glass pane normally measures a radiometric response dominated by the pane rather than an object behind it.
A thermal imaging camera extends radiation thermometry to an array of detector elements. It produces a spatial map of radiometric measurements, often displayed with artificial colors. Each pixel is subject to calibration, emissivity, reflection, focus, and spatial-resolution limitations. A thermal image maps surface radiation; it does not directly reveal internal temperature or measure air temperature between the camera and target. Its quantitative interpretation depends on the same radiation-transfer principles as a single-point instrument.
Thermal coupling, response time, and measurement disturbance
A contact thermometer measures the temperature of its sensing element. That temperature represents the target only when heat exchange and thermal coupling make the two sufficiently close. The equilibrium basis of thermometry is straightforward, but real measurements frequently involve competing heat flows: conduction from the target, conduction along wires or a protective stem, convection to surrounding fluid, and radiative exchange with nearby surfaces. A probe can reach a stable reading while remaining at a temperature different from the intended target.
A simplified dynamic model treats the sensor as a body with uniform temperature , heat capacity , and thermal conductance to a constant-temperature environment :
The characteristic response time is . Following a sudden change in environmental temperature, the sensor’s departure from the final value decays as
After one time constant, the sensor has completed about 63.2% of the total change; after three, about 95%. These figures describe a first-order model, not every thermometer. Multilayer probes and instruments with substantial internal temperature gradients may show several time constants.
Response time is not solely a property of the sensor. A probe may respond much faster in vigorously stirred water than in still air because the heat-transfer coefficient differs. Protective sheaths, thermowells, and coatings improve mechanical or chemical resistance but add thermal resistance and heat capacity. Electronic filtering and display averaging introduce further delay. An instrument can therefore have a rapidly responding sensing element but a slowly updating indicated value.
Immersion depth is particularly important in fluids. If too much of the probe remains outside the measured medium, conduction along the stem can pull the sensing element toward ambient temperature. On solid surfaces, imperfect contact, roughness, oxide layers, and contact pressure influence heat transfer. Attaching a sensor can itself change the surface’s heat flow, especially on thin components or small electronic devices. Fine thermocouple wires reduce this disturbance but can be more fragile and susceptible to chemical change.
Dynamic measurements also face spatial limitations. A large probe averages conditions over its active volume and may smooth a steep gradient. A moving probe can lag behind the actual temperature profile, displacing apparent features. Correcting such effects requires a valid model and information about response under the actual measurement conditions; a single time constant measured elsewhere may be insufficient. These issues explain why higher display resolution cannot compensate for poor sensor placement or inadequate thermal contact. The measurement problem includes the interaction between instrument and object, not just the conversion of an electrical signal into degrees.
Calibration, traceability, and uncertainty
Calibration establishes the relationship between an instrument’s indications and reference values under specified conditions. It is distinct from adjustment: calibration can characterize an error without changing the instrument, whereas adjustment changes its response. A thermometer may be calibrated by comparison with a reference thermometer in a stable bath or furnace, or at reproducible fixed points. The result applies to the stated instrument configuration, measurement conditions, and temperature range.
Fixed points exploit reproducible phase transitions. ITS-90 uses specified states of substances including water, gallium, indium, tin, zinc, aluminium, silver, gold, and copper. Their usefulness depends on material purity, cell construction, and the establishment of appropriate equilibrium conditions. The gallium melting point, for example, has an assigned ITS-90 value of 29.7646 °C. A fixed-point realization is not simply an arbitrary sample beginning to melt: impurity effects, pressure corrections, heat flow, and the measurement position within the cell can matter at high precision.
Comparison calibration places the instrument under test and a reference in a sufficiently uniform temperature environment. Stirred liquid baths provide effective heat transfer over suitable ranges; dry-block calibrators provide convenient heated wells but require attention to axial and radial gradients, loading, and immersion. Furnace comparisons extend calibration to higher temperatures. Radiation thermometers are calibrated using sources with known radiance, commonly cavity blackbodies whose geometry and surface properties produce high effective emissivity.
Metrological traceability connects a measurement result to a stated reference through a documented, unbroken chain of calibrations, each contributing uncertainty. Traceability belongs to the result and its supporting process, rather than being an unconditional attribute of a sensor. A calibrated probe used with a different readout, inappropriate extension cable, or unsuitable installation may not produce a result covered by its calibration. Calibration intervals likewise depend on stability, usage, required uncertainty, and evidence from previous measurements.
An uncertainty budget identifies significant contributions such as reference uncertainty, calibration-model residuals, repeatability, resolution, drift, thermal gradients, immersion effects, and readout errors. For approximately independent input quantities, a common propagation expression is
where . Correlated inputs require additional covariance terms. Expanded measurement uncertainty is often reported as , accompanied by the coverage factor and relevant interpretation. Neither a fine digital increment nor a manufacturer’s accuracy specification substitutes for an uncertainty evaluation. A display resolving 0.001 °C may still have a much larger uncertainty because of calibration, drift, or installation effects.
Applications and specialized measurement environments
In meteorology, the intended quantity is usually air temperature under specified exposure conditions. A thermometer exposed directly to sunlight can become warmer than the surrounding air because it absorbs radiation. Weather stations therefore use radiation shields, often with natural or forced ventilation, to reduce radiative heating while permitting air exchange. Sensor height, nearby buildings, ground cover, ventilation, and instrument changes can all affect observations. Long climate records depend not only on thermometer calibration but also on documented exposure and observing practices.
Clinical thermometers measure at particular anatomical sites or infer a temperature from surface radiation. Oral, axillary, rectal, tympanic, and forehead measurements are not interchangeable physical observations. They sample tissues or surfaces with different relationships to internal body temperature and may respond differently to ambient conditions and measurement timing. Electronic clinical instruments may use a predictive algorithm to estimate an endpoint before full thermal equilibrium is reached. The displayed number can consequently include a model-based estimate rather than only the instantaneous sensor temperature. Device validation must address the complete measurement method, including any algorithm.
Industrial thermometry encounters pressure, vibration, corrosive chemicals, strong electrical interference, and large temperature gradients. A thermowell separates a probe from a process fluid, permitting replacement without opening the process containment, but it also changes the thermal response and can experience flow-induced vibration. Furnace measurements may favor thermocouples or radiation thermometers, whereas tightly controlled laboratory baths often use resistance thermometers. In food processing, sterilization, and chemical manufacturing, the location of a sensor can be as consequential as its calibration because a vessel or product may contain persistent temperature differences.
At cryogenic temperatures, ordinary room-temperature sensor relationships may become insensitive or unsuitable. Specialized resistance sensors, diode thermometers, vapor-pressure methods, and magnetic thermometry cover different low-temperature ranges. Below 1 K, the Provisional Low Temperature Scale of 2000 uses the melting pressure of helium-3 over the range from 0.9 mK to 1 K. Magnetic methods exploit temperature-dependent magnetic properties, while noise thermometry connects fluctuations to temperature. These techniques require careful control of parasitic heat leaks: even small electrical measuring powers can appreciably warm a tiny, weakly coupled sample.
Miniaturized thermometry addresses electronic devices, microfluidic systems, and microscopic environments. Methods include lithographically fabricated resistance elements, fluorescence-based probes, Raman measurements, and sensors based on temperature-dependent quantum properties. Their spatial resolution may be excellent, but interpretation can be difficult when electrons, lattice vibrations, and surrounding material are not in mutual equilibrium. A single local temperature may then fail to describe every subsystem. In such measurements, the reported quantity must correspond to the physical population or response actually sampled—for example, an electronic noise temperature or a temperature inferred from a particular spectroscopic transition.