Poorly carburized parts can have devastating consequences. Part hardness and case depth are critical to numerous tooling, automotive and aerospace systems. While diffusion is a well-understood science, the assumptions made in atmosphere carburizing are substantial and can have major impacts on part performance. By understanding the tools used and their limitations, heat treaters can ensure better parts.

Fig. 1. Oxygen sensors are the primary means for controlling carburizing but require assumptions about CO and H2 levels and may react with uncracked CH4 giving false high readings.

The basis for carburization is the diffusion of carbon into the surface of a metal. Originally, carburization involved packing the item in charcoal, ash or iron shavings – a process called “pack carburizing” that is still used for some large parts and in some regions. Today, the most common method for carburizing parts is atmosphere carburizing, where a gaseous source of carbon is used. Vacuum carburizing and plasma carburizing are other ways of diffusing carbon into a part.

First Assumptions Involve Diffusion

The fundamental process by which carbon diffuses into the metal to form a case is expressed using Fick’s Second Law of diffusion. It is in this equation approximations are first introduced.

Fick’s Second Law – and related diffusion equations – starts by reducing the complexity of the math to one dimension and then assumes the metal is a semi-infinite solid in which the surface concentration is held constant.

We then assume uniform alloy composition and grain size, uniform surface conditions and a uniform atmosphere. Good furnace circulation can get one closer to a uniform atmosphere, but surface conditions can vary substantially.

Before we’ve even begun to measure and control the process, we’ve already discovered that the basic science makes assumptions that can significantly impact parts.

Assumption of Equilibrium

The next major assumption undertaken is that the atmosphere in the furnace is in equilibrium. We must do this for almost any sensor to be used. If we measure the gas with an infrared analyzer, dewpoint meter or oxygen sensor, we have to assume that the gas is in equilibrium in order for our measurements to be related to the carbon potential in the furnace since we never measure carbon directly.

In gaseous carburizing, the carbon [C] consumed by the parts is coming out of the carbon-bearing gases carbon monoxide (CO) and methane (CH4). The CO sticking to the part surface will give its oxygen (O) either to hydrogen (H2), building water vapor (H2O) or to another CO creating carbon dioxide (CO2). The CH4 can directly react with the surface and leave H2. The reactions are:

CO+H2[C]+H2O (1)
CO+CO[C]+CO2 (2)
CH4[C]+2 H2 (3)

Unfortunately, these reactions do not have the same speed, so a calculation based on (1) will not exactly match a calculation based on (2) or (3). Actually, reaction 1 is approximately 30 times faster than reaction 2 and 100 times faster than reaction 3. In addition to these three reactions with the parts surface, there is another reaction that takes place in the atmosphere, the so-called water-shift reaction that keeps the components CO, H2, CO2 and H2O in balance.

CO2+H2CO+H2O (4)

With equilibrium we assume equations 1, 2 and 4 are in balance. The direct carburizing effect of (3) is typically negligible as it is simply too slow to have a real impact, except with very high concentrations of uncracked methane in the atmosphere. So, in fact, our atmosphere is constantly being bumped out of equilibrium as we enrich and then balancing out again. The result is that the assumption of equilibrium causes errors in all measurements that assume equilibrium.

Measurement Assumptions - Oxygen Sensors

The predominant means for measuring the “carbon potential” of a carburizing atmosphere is an oxygen sensor (Fig. 1). This is based on the premise that oxygen (O2) and hydrogen (H2) are in equilibrium with water (H2O) and the carbon [C] and H2O are, in turn, in equilibrium with CO and H2 (eq. 1).

If we measure the O2 concentration, we can determine H2O concentration assuming we know the H2 concentration. And, if we further assume a known CO concentration, we can calculate the carbon potential (CP). In fact, this explanation is a bit misleading from the start.

In a typical batch furnace with a 3-foot x 3-foot x 4-foot work zone, one would expect to find about four O2 molecules in the entire furnace while endo is flowing. Far more than 99% of the time the sensor sees exactly zero O2 regardless of endo flow rate, enrichment gas, etc. even though we know enrichment gas increases carbon potential and the oxygen sensor will see this effect. So what exactly does the oxygen sensor sense?

Clearly, the oxygen sensor isn’t measuring those four O2 molecules. Instead, it is measuring a process called “reduction” on the outer electrode of the oxygen sensor. Reduction refers to a strong chemical reaction whereby chemicals react with O2. In a carburizing furnace, the electrochemical force of reduction literally pulls oxygen out of the sensor (from the reference air) to react with gases inside the furnace. The reduction process is what generates the sensor’s millivolt signal that is used, along with temperature, to calculate carbon potential.

In an endo-based carburizing atmosphere, there are three primary reducing species: hydrogen (H2), carbon monoxide (CO) and methane (CH4). In effect, ignoring temperature dependence we can write the mV output of an oxygen sensor as equation 5:

mV[02]=f(H2,CO,CH4)    (5)

So, a change in any of these components has an impact on the millivolt output of an oxygen sensor even if that change doesn’t necessarily impact carburization. For example, if we fill a furnace with nothing but H2, we will read a maximum millivolt signal (about 1300 mV) even though there is no carbon present to harden the load.

So how is it that an oxygen sensor can be used to measure and control carburizing? In one word: assumptions.

For oxygen-sensor-based carbon control, we assume a known and constant CO value (we typically assume 20% when using endogas made from natural gas and air), and we assume the H2 and CH4 levels are constant so that any changes in the sensor millivolts are strictly caused by the 20% CO. If the H2 and CH4 do not change and the CO holds at a constant 20%, we will have a good measure of carbon potential once we correct the baseline reading using shim analysis or cut parts.

Unfortunately, CH4 is a common enriching gas and both CO and H2 levels change when CH4 is added to the furnace. CH4 is oxidized by CO2 and H2O with only the CO shift affecting carbon potential. To make matters worse, oxygen probes react catalytically with CH4, which generates a mV signal that is completely independent of carbon potential – though some probes use designs to minimize this effect.

Even the assumption that CO is constant is actually not correct because the atmosphere itself is not in equilibrium. The biggest problem for the CO is the load itself. At the beginning of a process the CO might drop by 1-2%, whereas at the end of a process the CO can go up by ~ 3%.

Fortunately, for most applications a regular check of the probe output to other measurements such as dewpoint, NDIR or, preferably, shim can minimize the error introduced by the assumptions we make when using oxygen probes. Be careful when using such “process factors” that it is done at the more critical points in the process. A correction done during a boost stage will result in a far less accurate diffuse stage.

Some instrumentation and software make use of more sophisticated atmosphere models that take these CO changes into account.

Fig. 2. Dewpoint analyzers are commonly used to check oxygen-probe readings, but they require that assumptions be made about the furnace atmosphere.

Measurement Assumptions - Dewpoint Meters

Dewpoint, or the temperature at which condensation occurs, has long been used as a measure of carburizing atmosphere. Originally, dewpoint was determined by flowing furnace gas across a mirror that was chilled slowly until condensation was observed. The temperature at which condensation was observed was the gas’ dewpoint. In practice, most modern instruments (Fig. 2) used today measure the relative humidity using either radioactive decay or dielectric sensors and convert this to the traditional dewpoint number. Dewpoint can then be converted – via well-established tables – to carbon potential.

As with probes, the use of dewpoint is based on the equilibrium of CO and H2 with [C] and H2O (eq. 1). If we know the H2O (inferred from the dewpoint measurement) and we know the CO and H2 levels, we can calculate the available carbon.

As in the case of the oxygen sensor, assumptions about the gas must be made. We must know the CO and H2 values to relate dewpoint to carbon potential. Further complicating things is the chemical degradation of the capacitive-type dewpoint sensors. Essentially, these sensors are dielectric capacitors with their insides exposed to the gas, and they slowly change with time, making regular calibration critically important.

In a bizarre reversal of this measurement, oxygen sensors are often used to measure carbon potential (as described in the previous paragraph) inside an endo generator, and then this is converted back to dewpoint.

Fig. 3. Both fixed and portable (inset) NDIR analyzers measure multiple gas components. As long as the furnace is at equilibrium, they provide the best indirect measure of carbon potential.

Measurement Assumptions - NDIR

Non-dispersive infrared (NDIR or “IR” shown in Fig. 3) has become a more common way of correcting oxygen-probe readings with both fixed-automatic and portable models available. NDIR measurement of carbon potential, like probes and dewpoint meters, is based on inference and not direct measurement of carbon. NDIR meters typically measure both CO and CO2 (and in some cases CH4) and use these gas concentrations to determine the carbon potential.

Like other sensors, NDIR assumes the gas is in equilibrium. But unlike the other sensors, it does not require assumptions about either CO or CO2 since both are measured directly. As long as the process is actually in equilibrium, NDIR offers the best possible indirect measurement.

Errors arise in NDIR from things that affect equilibrium such as load surface area and CH4 concentration. Large concentrations of CH4 soak up CO2 and H2O, generate CO and H2 and push the furnace out of equilibrium, making equation 3 invalid. And since the oxidation with H2O is faster, this results in false low carbon readings.

Measurement Assumptions - Sampling

One large source of error for any extractive measurement, such as dewpoint meters or NDIR, is sampling. In any extractive measurement, there is an implied assumption that the gas being “seen” by the dewpoint sensor or NDIR analyzer is the same as the gas in the furnace. This is often not the case. Obviously, any leaks in a sampling system will introduce large errors, but even assuming a leak-free sample system there are many sources for potential errors.

Many users of extractive systems will utilize a burn-off port on an oxygen sensor to pull a sample from the furnace. This essentially uses a probe sheath as a sample port. A typical probe sheath is both highly catalytic (high nickel content reacts with CH4) and dirty (soot deposit react with O2). Both of these factors strongly influence the resulting measurement. Finally, CO2 and H2 readily react to form CO and H2O (water-shift reaction), making any extractive measurement suspect.

Fig. 4. Modern electromagnetic shim analyzers provide direct measurement of actual carbon without any assumptions and without the hassle and expense of traditional calorimetric systems.

Shims - Use Them

Shims are a non-equilibrium sensor. It might sound strange to call a shim a sensor, but that’s essentially what they are. More importantly, they measure the actual carbon uptake in a furnace directly. They soak up the carbon right next to the load and make no inference based on equilibrium. A properly carburized shim will tell the real carbon potential in the furnace.

Unfortunately, shims are, simply put, a pain. One must typically follow careful procedures in handling and use a very expensive calorimeter to get a reliable and accurate measure of the shim carbon content. As a result, fewer shops use shims than ever, instead choosing to rely on dewpoint or IR with all of their inherent assumptions about the furnace gas composition and equilibrium conditions.

Fortunately, new electromagnetic analyzers (Fig. 4) are now available that can determine the carbon content of a shim without special handling and without the use of a calorimeter. In these analyzers, the electromagnetic properties of the carburized shim are measured. Since electromagnetic behavior is a function of carbon content and crystallographic structure, the carbon content can be precisely determined as long as the quench/cooling method is constant (same microstructure) and the same shim thickness is used.

With such analyzers, handling is no longer an issue (no sensitivity to oil from fingers, etc.), and measurements take only seconds on a carburized shim. The results can then be used to adjust and correct the controlling probe or more regular IR reading.


Carburizing is a non-equilibrium process that is routinely controlled using sensors that are subject to substantial errors. These sensors are then corrected using analyzers and sensors that have their own substantial problems in that they rely on assumptions about gas composition and/or equilibrium of the gas.

Shim-stock analysis, while traditionally cumbersome, is the best means for correcting or adjusting the controlling sensor and the backup analyzers. Traditional obstacles to shim analysis, such as time, handling and cost, are overcome by using electromagnetic shim analysis.

With shim measurements, both oxygen probes and their backup sensors – dewpoint and IR – can be made more accurate and reliable, resulting in better parts and overall higher quality.IH

For more information:Contact Eric Boltz at Marathon Monitors Inc, 8904 Beckett, West Chester, OH 45069; tel: 513-772-1000; e-mail: eboltz@marathonsensors.com; or Karl-Michael Winter at Process-Electronic, Dürnauer Weg 30, D-73092, Heiningen, Germany; e-mail: km.winter@process-electronic.com.

United Process Controls is a group comprised of the following companies: Furnace Control Corp., Marathon Monitors Inc., Nitrex Metal, Process-Electronic & Waukee Engineering and can be found on the web at: www.group-upc.com www.group-upc.com.

Additional related information may be found by searching for these (and other) key words/terms via BNP Media SEARCH at www.industrialheating.com: carburization, diffusion, vacuum carburizing, plasma carburizing, dewpoint, oxygen sensor, NDIR, shim