Jump to main content.

Selecting Alternatives to Mercury-Filled Thermometers

Information about the Phase-Out of Mercury Thermometers Used in Industrial and Laboratory Settings


Mercury-filled liquid-in-glass thermometers have a long history of use in a variety of laboratory and industrial applications.  Although these thermometers provide excellent performance, regulations on the sale of mercury thermometers limit their continued availability.  Mercury spilled from broken thermometers also poses an environmental and safety risk.

Alternatives exist for almost all uses of mercury-filled glass thermometers. The information below gives a brief introduction on the selection and use of these alternatives.

Alternatives to Mercury-Filled Glass Thermometers

Resistance Temperature Detector (RTD) or Platinum Resistance Thermometer (PRT)

Currently, almost all RTDs used for accurate thermometry are made with a platinum sensor, and the term platinum resistance thermometer (PRT) generally is synonymous with RTD.  The electrical resistance of the platinum rises as the temperature rises.  The readout converts the measured resistance to indicated temperature using either a standard curve or a calibration function for the particular probe being used.

A properly chosen PRT probe and readout can be used to replace almost all mercury-filled liquid thermometers.  For high-vibration applications within the temperature range -100 °C to 150 °C (-148 °F to 302 °F), PRT sensors formed from a platinum film deposited on a ceramic chip work well.  For applications requiring broader temperature ranges or better uncertainty, we recommend wire-wound PRTs.  In both cases, the sensor typically is mounted in a metal sheath.


Thermistors are an excellent choice for temperature measurements in the range -20 °C to 100 °C (-4 °F to 212 °F).  Thermistors used as thermometers are composed of a blend of metal oxides whose electrical resistance falls as the temperature increases.  The most stable thermistors are sealed with a glass coating.    For general-purpose use, the thermistor and wire leads often are mounted in a protective metal sheath.  Large shocks, such as dropping the probe on the floor, could break the glass coating on the thermistor, leading to increased drift of the sensor.


A thermocouple temperature sensor consists of two dissimilar metals, joined at one end to form the measuring junction.  The two thermocouple wires must extend all the way from the measurement point to the readout.  If intermediate connections are needed, special connectors must be used.  Of all the thermometer types, thermocouples resist shock and vibration the best.  The manufacturing tolerances for thermocouples are relatively large, and readouts add an additional uncertainty. 

Thermocouples may be insulated with ceramic, fiberglass, or polymer insulations.  The insulated thermocouple may be mounted in a metal sheath for additional protection of the sensor.

Thermocouples are a good choice when the desired uncertainty is greater than approximately 1 °C (2 °F), and a mechanically robust or compact sensor is required.

Organic-Liquid-Filled Thermometers

Glass thermometers filled with non-hazardous organic liquids are a good choice when the temperature lies within the range -100 °C to +100 °C (-148 °F to 212 °F), and the desired uncertainty is 0.5 °C (»1 °F)or larger.  A wide variety of organic liquids are used for commercially available thermometers.  The liquid column of an organic-filled thermometer is subject to separation when the thermometer is shipped, used at extreme temperatures, or stored in a non-vertical position.  When an organic-liquid-filled thermometer is subjected to these conditions, the liquid column must be carefully inspected before use.

View a table of organic-liquid-filled thermometer substitutes for mercury-containing thermometers (PDF) (1 pp, 69 KB, about PDF).

Chart of Typical Uncertainties

The charts below give a summary of typical achievable uncertainties or manufacturing tolerances, in units of both degrees Celsius and degrees Fahrenheit. (These charts can assist with the selection of a thermometer, but do not represent the actual uncertainty of any particular thermometer.) With special care, better uncertainties may be obtained.  On the other hand, abuse of the thermometer, long-term use, or use of inferior-quality thermometers can lead to larger uncertainties.  The uncertainties on the charts include allowances for sensor drift and readout uncertainties.

Chart of summary of typical achievable uncertainties or manufacturing tolerances.

Chart of summary of typical achievable uncertainties or manufacturing tolerances.

Top of page

Ensuring Good Measurement Results

1.  Avoid shock to the sensor and readout.  With metal-sheathed thermometers, damage to the sensor will not be apparent from a visual inspection or from a simple operational check.

2.  For thermocouples, avoid kinks in the thermocouple wires, especially in regions where the temperature is changing from one point to another.  For thermocouples used above 150 °C (302 °C), best uncertainties are obtained by using a separate thermocouple for each apparatus, and always using the thermocouple at the same depth into the apparatus.

3.  Unless you are absolutely certain that two probes are interchangeable, never switch the probe used with a readout without updating calibration coefficients.

4.  For readouts that support many types of probes, be absolutely certain that the readout is set to the proper thermometer type (e.g., do not read a type K thermocouple with a readout set for a type J thermocouple).

5.  Do not exceed the recommended temperature limits for the probe and sensor. 

6.  Check the performance of the instrument regularly, following the manufacturer’s recommendations or past history for the device.

Top of page

Differences Between Liquid-in-Glass Thermometers and Electronic Thermometers

Liquid-in-glass thermometers are self contained and self powered.  Glass breakage often indicates mechanical abuse!

In general, electronic thermometers cost more than mercury-filled thermometers of comparable accuracy.  Readouts for electronic thermometers are easy to read, and easy to integrate into an automated data system.  These advantages can reduce reading errors and operational costs for electronic thermometers relative to liquid-in-glass thermometers.  Damage to the sensor for an electronic thermometer often is not visually apparent.

Both types of thermometers require regular validation or recalibration. 

Top of page

Special Conditions of Use

In some standards applications, liquid-in-glass thermometers may be used in a manner that is highly reproducible, but that does not indicate true temperature.  Switching to an alternative in these cases can alter a measurement bias in an unpredictable way.

Examples of this effect include:

1. The original liquid-in-glass thermometer was used at an incorrect immersion.  Liquid-in-glass thermometers may be either partial immersion or total-immersion types.  Partial immersion thermometers have a line around the circumference of the thermometer and/or a printed immersion depth on the back (e.g., 76 MM for 76 millimeters immersion).  Thermometers with no indicated depth are the total immersion type.  When a partial-immersion thermometer is used, the bottom of the thermometer up to the immersion line should be exposed to the temperature being measured, with the remainder of the thermometer exposed to ambient conditions.  When a total immersion thermometer is used, the bulb and the entire portion of the stem containing liquid, except for the last 1 cm, are exposed to the temperature being measured.  If the thermometer is not used in this manner, the thermometer immersion is incorrect.  In practice, incorrect immersion is not a significant problem if the measured temperature is within 20 °C (36 °F) of ambient temperature.

2. The liquid-in-glass thermometer was used for a test where the temperature is not stabilized before a reading.

3.  The body being measured has a temperature that is not uniform.

If any of these conditions hold, the reading of a liquid-in-glass thermometer may not correspond to the reading of an alternative thermometer, even if both thermometers are perfectly accurate.  We recommend, first, that the alternative thermometer be carefully specified in construction.  Second, the readings of a calibrated liquid-in-glass thermometer under these conditions of use should be compared with the readings of a calibrated alternative sensor to identify any measurement bias.

Top of page

Selection Flow Chart

The selection process described in this document also can be described by a flow chart.  Begin with Step 1, and enter Step 2 at the indicated point.

Flow chart of the selection process for accessing the usage of liquid in glass thermometer.Flow chart of the selection process for indetifying an alternative with appropriate properties.

Top of page

Frequently Asked Questions

How much does an electronic thermometer cost?

Price varies with performance.  A digital readout with a thermistor or PRT sensor will cost approximately $200 for a tolerance of ±0.2 °C (±0.4 °F).  Thermometers with higher or lower accuracy are available at proportionately higher or lower costs.

Do I need to have my thermometer calibrated?  How often?

There is no general rule stating whether thermometers require calibration. Thermometers may require calibration because:

See the documents on Maintaining Traceability {Insert hyperlink to WA_4-14_Task 3 Maintaining Traceability.doc} to NIST for a discussion of how often to calibrate a thermometer. 

What is the difference between accuracy, tolerance, and uncertainty?

Graph of manufacturing tolerance.

“Accuracy” is a common term, but it is not well defined.  In its most common usage, “accuracy” is the manufacturer’s guarantee that the instrument will give the correct answer to within the stated accuracy.  In this sense, the word “accuracy” is equivalent to a manufacturing tolerance.  A tolerance is the allowed variation in some property of the thermometer sensor, as shown below.

Sensors manufactured to meet a tolerance will be interchangeable to within that tolerance, at least when the sensor is new.  Remember that thermometers may drift with use. How long a thermometer will meet the manufacturer’s accuracy statement depends on the thermometer type and its use!

In the field of metrology, the possible error of a measurement is given as the “measurement uncertainty”.   The language of uncertainty expresses results in terms of probability.  A calibration report might say that at a temperature of 100 °C, a thermometer gave a reading of 99.3 °C with an expanded uncertainty (k = 2) of 0.4 °C.  This language is approximately equivalent to saying:
At a temperature of 100 °C, we obtained a reading of 99.3 °C on your thermometer.  There is a 95 % likelihood that at a true temperature of 100 °C, your thermometer would read between 98.9 °C and 99.7 °C.
Those temperature limits are calculated by adding or subtracting the uncertainty from the measured value:  98.9 °C = 99.3 °C – 0.4 °C, or 99.7 °C = 99.3 °C + 0.4 °C.  Different values of k (called the coverage factor) correspond to different levels of likelihood, or confidence.

Top of page

Learning More

National Institute of Standards and Technology Mercury Thermometer Alternatives website

General references include:
1. J. Nicholas and D. R. White, Traceable Temperatures (Wiley, 2001).
2. Handbook of Temperature Measurement (Vols 1-3), ed. Robin E. Bentley (Springer, 1998).

An additional reference for thermocouples is:
3. ASTM Manual on the Use of Thermocouples in Temperature Measurement, MNL-12 (ASTM, West Conshohocken, PA, 1993).

A discussion of particular issues for replacing liquid-in-glass thermometers for standards work is found in:
4. Dean C. Ripple and Gregory F. Strouse, “Selection of Alternatives to Liquid-in-Glass Thermometers,” J. ASTM International 2, JAI13404 (2005).

Top of page

Local Navigation

Jump to main content.