It seems ironic that the one constant in atmosphere heat treating is that the composition of the furnace atmosphere is always changing. As such, we must use measurement and control devices to ensure the proper atmosphere composition necessary to achieve the desired metallurgical quality and mechanical/physical properties throughout the workload. Let’s learn more.

The trend today is to use multiple measurement tools to obtain the most accurate snapshot of the atmosphere in real time. In addition, the industry is rapidly moving toward automated control systems.


Manual or Automatic Control?

In one form or another, atmosphere control has been around for decades, and the control method may be categorized as either manual or automatic. There are benefits and limitations to both (Table 1).

Benefits and limitations of manual and automated control

In a manual control scheme, success depends on the vigilance of the people operating the equipment, relying on their experience coupled with either trial and error or historical records/information. A typical example of this type of control would be as follows.

A furnace operator sets up a load, charges it into the furnace and manually adjusts temperature and atmosphere parameters. As process values approach setpoint, the operator begins periodically checking the atmosphere with sensor devices (e.g., dew-point instruments) and logging values on a heat-treat form of some design.

If process deviation occurs, the operator intervenes and attempts to make adjustments “on the fly” based on experience. At the end of the process, the heat treater checks the load to confirm it does not need to be reworked or scrapped. Essentially, operators learn what works and what does not on a case-by-case basis over time, adding to their knowledge base. For key parameters or data records, a form, a recorder or a printer would provide a paper copy to be filed.

Automated control differs from manual control mainly in that sensors, computers and specially designed software are used during the heat-treatment process to monitor, control and record data about the furnace atmosphere. Load-entry software can even be used to control the timing of a charge and electronically record all load parameters set up by the operator. Output is set and changed by a process controller, such as a loop controller, designed specifically for heat-treatment processes or a PLC.

If the controller detects a process deviation or a condition likely to result in one, it can make adjustments and/or generate a process alarm if operator intervention is needed. Information is delivered through mobile devices to operators, supervisors or even maintenance based on an escalation and profile of the alarm and recipient. Data logging is electronic. Logged data points can be accessed at any point in the future.


Measurement Devices and Sensors

Monitoring and/or controlling the heat-treatment process is accomplished by use of sensors and measurement devices. The most common of these are:

  • Oxygen (carbon) probes
  • Non-dispersive infrared analyzers (NDIR) – single- or multiple-gas
  • Dew-point analyzers
  • Oxygen analyzers
  • Combustion analyzers

Whether neutral or case hardening, annealing or normalizing, a number of variables determine how well a furnace does its job. Throughout the entire cycle it is critical to the process that we control the percentage of carbon dioxide, oxygen and water vapor as well as the ratio of enriching gas (or air) to carrier gas.

For example, surface carbon can be controlled within ±0.05% C during carburizing by measuring one or more of the gases mentioned above and adjusting the addition gases (hydrocarbon and/or air) accordingly.


Oxygen (Carbon) Probes

The oxygen (aka carbon) probe (Fig. 1) is an in-situ device that looks similar to a thermocouple for measuring temperature and typically sits inside the furnace, inside the generator (typically above the catalyst bed or in a separate heated “well” into which the furnace atmosphere is pumped). In whatever location, the oxygen probe measures minute changes in oxygen concentration of the furnace atmosphere.

A difference in partial pressure of oxygen in the furnace atmosphere and the partial pressure of oxygen in the room air induces a voltage across the electrodes in the probe. At any given temperature, there is a known relationship between the probe millivolt output and the oxygen potential of the atmosphere. The oxygen potential can be directly related to the carbon potential. Hence, monitoring the furnace temperature and the probe output can control the carbon potential of the furnace atmosphere.

The oxygen probe uses a conductive ceramic sensor, most often manufactured from zirconium oxide (Zr2O3). Operating range of the probe is normally 650-980˚C (1200-1800˚F). Oxygen probes can be used for a variety of atmosphere compositions, but they need to be calibrated for the specific one in use. They are fast-response devices and subject to contamination by carbon, zinc and certain stop-off paint vapors. When used in carbonitriding applications, the presence of ammonia will shorten the life of the probe.

An oxygen probe in a carburizing atmosphere must incorporate periodic air burnouts (Table 2). The carburizing process in use will determine the burnout frequency.

Recommended burnout frequency

A burnout consists of at least 0.28 m3/hour (10 cfh) of air piped to the burn-off fitting on the head of the probe. Room air or filtered combustion air are most commonly used. It is important not to use compressed air due to water and oil contamination that can damage the probe. The carbon controller should either control the frequency and duration of the burnout or shut off the gas additions in order to prevent excessive gas from compensating for the flow of air to the probe. Burnout flow and duration recommendations vary by manufacturer based on sheath diameter and tip design.

It is good practice never to exceed 90 seconds of air addition at any one time to avoid overheating the tip of the probe. A consistent way to verify a correct burnout is to monitor the millivolts of the carbon controller during the burnout phase. If a proper burnout is taking place, the output will drop below 200 millivolts. This can also vary based upon the circulation in the furnace and the probe placement.

A possible side effect of extended burnout duration is oxidation of the tip of the sensor. This problem can manifest itself in higher-than-normal millivolt values over the remaining life of the sensor, which will require a lower CO-factor setting for the same calculation of carbon potential. Consideration should be made for the duration of the burnout based upon the carbon level in the furnace.

The current trend is to use oxygen probes in combination with three-gas analysis (CO, CO2, CH4) equipment (Fig. 2) to calibrate the probe to a known CO value and to monitor the amount of free hydrocarbon gas in the furnace atmosphere.


Measuring an Endothermic Atmosphere in a Furnace[5]

Carbon (oxygen) probes, shim stock or carbon analyzers and dew point are typically used to measure the carbon potential of the endothermic atmosphere in a heat-treatment furnace. An infrared analyzer can also be used. The CO2 value – similar to water vapor and oxygen – varies based on furnace temperature and carbon potential but will typically lie between 0.25 and 0.50%. Multiplying the CO2 value by 100 can roughly approximate the dew point (in ˚F). Charts and tables can also be used and are available from most manufacturers.

If a radiant tube is leaky in a gas-fired furnace, the leak (oxygen and moisture) will dilute the atmosphere carbon potential in the heat chamber when the furnace goes to high fire. As a result, the carbon controller will compensate by calling for more enriching gas to increase the carbon potential coincidental with high fire. A “saw-tooth” trend appears on the recorder. Setting the control to low fire will eliminate the saw tooth if the tube is cracked and leaking.

Oxygen probes and infrared analyzers are often used in conjunction when monitoring and controlling a nitrogen-methanol atmosphere. Using NDIR, values between 14-28% CO have been reported (target 20%). CO2 values typically range from 0.10-0.80%, and free methane can be as high as 8% depending on the carbon setpoint. All testing should occur without hydrocarbon (e.g., natural or propane) gas additions. For nitrogen/methanol systems, the CO percentage is largely determined by furnace temperature and the relative flows of nitrogen and methanol.


Measuring an Endothermic Atmosphere in a Generator[5]

Another popular trend, especially in the control of endothermic gas generators, is to use an oxygen probe in combination with a dew-point analyzer.

Using an oxygen sensor to measure and control the dew point in an endothermic generator is done either in-situ, in a thermal well or by use of an oxygen probe with an integral thermal well/sheath arrangement. The millivolt signal can be converted to a dew-point value or, in some of the more advanced dew-point control systems, by calculation. This calculation uses the millivolts generated by the probe, the hydrogen factor of the controlling instrument and the temperature of the oxygen sensor. The temperature is required for the calculation, but the dew point of the gas is not temperature-dependent.

Oxygen sensors are often run at 1040˚C (1900˚F), although they will provide the same reading when operating at 815˚C (1500˚F). The generator gas exiting the retort is sent through a heat exchanger to freeze the composition. As long as the sensor is accurately measuring the millivolts of the gas, the temperature of the sensor can be as low as 593˚C (1100˚F).

Changes in sensor location have occurred over the years. Initially, a sensor would be mounted on the top of a retort in an air-cooled, fabricated fixture to measure the oxygen. Then a ceramic reheat well mounted through the sidewall of the generator was used. Now a modified sheath with an integral reheat well makes the installation much easier. The sheath and the integral well are aluminized prior to assembly so that the nickel in the sheath material does not react with the endothermic gas. This is especially important between 705˚C and 480˚C (1300˚F and 900˚F) where, over time, the endothermic reaction will reverse if nickel is present.

Lambda-style probes like those used in an automobile engine are also used in generator applications. While less expensive, they often lack the long-term stability of the zirconia technology and must be re-oxidized to avoid drift. 


Next Time: In Part Two we learn more about measuring and controlling furnace atmospheres.



  1. Herring, D. H., “Selection and Use of Furnace Atmospheres,” Industrial Heating Webinar, 2011
  2. Herring, D. H., “Understanding Furnace Atmospheres, Atmosphere Operation and Atmosphere Safety,” Heat Treating Hints, Vol. 1 No. 7
  3. Thompson, Stephen, “A Practical Approach to Controlling Gas Nitriding and Ferric Nitrocarburizing (FNC) Processes,” Industrial Heating, December 2013
  4. Oakes, James and Jeremy R. Merritt, “Automated Control of Heat Treating Processes: Technology, Data Acquisition, Maintenance and Productivity Gains,” Industrial Heating, August 2014
  5. Fincken, Robert, “Protective Atmospheres, Measurement Technologies and Troubleshooting Tools,” Conference Proceedings, Furnaces North America, 2014
  6. “Process Automation,” Super Systems, Inc., white paper
  7. Donald Bowe, lead engineer, Air Products & Chemicals Inc., (, technical and editorial review, private correspondence
  8. James Oakes, vice president business development, Super Systems, Inc. (, technical and editorial review, private correspondence
  9. Thomas Philips (retired), Air Products & Chemicals (, private correspondence