We have a pretty good idea of what will happen to steel parts if exposed to a defined atmosphere at a given temperature. In order to determine the process parameters, we can use the well-known Lehrer Diagram for a nitriding process, or we might use one of the various Fe-N-C phase diagrams for a nitrocarburizing process.

All of the diagrams mentioned are giving the phase boundary between alpha iron and gamma-prime nitrides as well as the phase boundary toward epsilon (carbo)-nitrides as a function of the nitriding and carburizing potential and the temperature. In addition, we also have to account for the shifting of the phase boundaries, typically given for pure iron, caused by the alloying elements in real parts made from steel. But what will happen to our parts if we encounter deviations between the actual parameters and the set values during control?

This article will show typical measuring errors caused by the technique of the analyzers used and by a faulty reading of the temperature due to temperature deviations throughout the load, and it will explain their influence on the outcome of the heat treatment.

Nitriding and Nitrocarburizing

The goal of a nitriding treatment is to enhance the mechanical and chemical properties of parts by inducing nitrogen into the surface. Depending on the requirements, we are aiming for different types of layers. Figure 1 gives a schematic structure of a nitrided layer.

Starting from the surface, we first observe a thin and very hard ceramic layer that has been formed by the transformation of the base material into iron nitrides. Below this so-called white layer, we find a zone saturated with nitrogen not yet transformed into iron nitrides. This layer is known as the diffusion layer. Within the diffusion layer there are precipitations of non-iron nitrides, which are nitrogen compounds with nitride-forming alloying elements such as chromium, titanium or aluminum. Below this precipitation layer, we find the base material.

For nitrocarburizing, not only nitrogen but also carbon will be induced into the surface. This causes a more rapid growth of the white layer. Nitriding processes are typically aiming for a deep load-bearing diffusion layer with only a thin white layer, whereas nitrocarburizing is used to create corrosion- and abrasion-resistant white layers. Nitriding is typically carried out at temperatures in the range of 480-550°C (896-1022°F) and nitrocarburizing in the range of 570-590°C (1058-1094°F).

Potentials and Process Parameters
Both nitriding and nitrocarburizing can be performed with different processes that are, according to DIN EN 10 052, classified into gaseous, salt, powder and plasma nitriding depending on the nitrogen-bearing medium used. This article will focus on gaseous nitriding, where ammonia is used as the nitrogen source.

The basic nitriding reaction is the catalytic dissociation of the ammonia molecule on the surface of the part.

NH₃ → 1.5H₂ + [N]    (1)  

The effectiveness is defined by the nitriding potential K_N as (Eq. 2):

The phase diagram developed by Lehrer (Fig. 2) displays the phase boundaries in the Fe-N binary system as a function of temperature and nitriding potential.

For nitrocarburizing, a carburizing gas is added to the ammonia. For this reason, there is also a carbon uptake in addition to the nitrogen uptake. The carburizing effect can also be explained by the reactions taking place on the part surface. We distinguish between the Boudouard reaction:  

2 CO → CO₂ + [C]    (3)

with the carburizing potential (Eq. 4; right):             

and the heterogeneous water-gas reaction:  

H₂ + CO → H₂O+ [C]     (5)

with the carburizing potential:  

K_CW = pH₂ • pCO/pH₂O   (6)  

It has to be noted that the two potentials K_CB and K_CW differ in magnitude while having the same nitriding effect. In addition, the heterogeneous water-gas reaction is much faster compared to the Boudouard reaction. These relations have to be considered when picking the control parameters associated with the measurement system used.

The impact of the combined nitriding and carburizing potentials K_N and K_CW on the composition of the white layer has been described by Weissohn[1] in his NICARM Diagram (Fig. 3).

Measuring Systems in Use

Different measuring systems can be used to detect the atmosphere potentials. The nitriding potential can be determined directly by measuring the partial pressures of ammonia and hydrogen in the process atmosphere. Typically, especially in regular nitriding processes, it is sufficient to measure only one of the two components because the other one can be easily derived out of the thermal dissociation of the ammonia.

NH₃ → 0.5N₂ + 1.5H₂     (7)  

If, besides ammonia and pre-dissociated ammonia, nitrogen is added to the process atmosphere, we either have to know the inlet gas flows or we have to measure both hydrogen and ammonia.

In oxynitriding processes, air is added, causing a reaction of the oxygen with hydrogen. This creates water steam that shifts the percentages between the gas components, and the nitrogen added with the air causes a dilution. Therefore, we also have to know the inlet gases, or we need to measure the water content of the process gas.

For nitrocarburizing with CO, CO₂, Endogas or Exogas, the bound oxygen is added with the carbon, and, in the case of Endo- or Exogas, hydrogen and nitrogen will also be injected into the furnace. This will establish the water-gas equilibrium:

H₂ + CO₂ → H₂O+CO   (8)

with the thermodynamic equilibrium constant (Eq. 9; right):              

To calculate the carburizing potential(s), we also have to determine CO or CO and CO₂. This can be done by a direct measurement of the gas component(s) or by measuring the inlet mass balances.

The measuring systems regarded in the following consist of:

• An H₂ analyzer using a measurement of the thermal conductivity of the process gas
• An infrared analyzer to measure ammonia, as an alternative to the H₂ analyzer or as an addition
• An additional oxygen probe
• A CO-CO₂ infrared analyzer

Typical Errors when Measuring Hydrogen
When using an analyzer that derives the hydrogen content of a gas mixture by measuring the thermal conductivity, there are three effects that influence the quality of the measurement:

• Temperature stability – The thermal conductivities of the various components in the sampling gas change in a different way when exposed to a shift in temperature.
• Pressure stability – Basically, the thermal conductivity of gases is stable in a wide range of pressure, but there are still little deviations.
• Viscosity – The thermal conductivity of a gas mixture is not represented by the sum of the thermal conductivities of the gas components but is curved by viscosity of the gas mixture.

On top of this, there are the built--in failures of a measuring system caused by the design and the sensor system used in the instrument, notably:

• Resolution, linearity and drift of the analog circuit
• Thermal stability of the sensor and/or longtime drift
• Deviations in the sampling gas flow

But there is one error that is even worse. Analyzers of this type are calibrated on a binary gas mixture, typically on percentage hydrogen in nitrogen (%H₂:N₂). When measuring the process gas of a nitriding or nitrocarburizing process, we measure a mixture of hydrogen, nitrogen, ammonia and additional gases like carbon monoxide, carbon dioxide and water vapor. This causes a notable deviation of the interpreted H₂ content to the real hydrogen percentage.

If we have a closer look at an analyzer that is within a higher cost range, the manufacturer states for a measuring range of 0-100% H₂:N₂:  

Output signal variations: < ±0.75% of the lowest possible measuring range                     

Zero drift: < 1% per week of the lowest possible measuring range                     

Repeatability: < 1% of the selected measuring range  

Linearity deviations: < ±1% of the selected measuring range

This translates into an uncertainty of ±1.54% H₂:N₂ absolute. In addition, we have to account for other influences such as environmental temperature, sampling gas flow and pressure, power-supply voltage and a zero offset caused by other gases besides nitrogen and hydrogen. Figure 4 gives an example how the total uncertainty of an instrument is calculated.

Exaggerating, if we use this analyzer outside of a temperature and air-pressure-controlled chamber, we will have different readings during summertime and wintertime. The offset caused by ammonia is shown in Fig. 5, and the deviations of a measurement of a nitrocarburizing atmosphere are shown in Fig. 6.

Typical Errors when Measuring Ammonia
When measuring ammonia with an infrared analyzer, we have to account for some fundamental problems in addition to the deviations given by the mechanical and electrical design. For an IR measurement, there is a linear dependency of the measured value on pressure. Therefore, the pressure either has to be stabilized or measured and compensated for. The biggest error can be expected by the cross sensitivity to water vapor. The comparably small absorption lines of the ammonia are placed in a comb of water lines, and an accurate detection of the ammonia lines can be done only by applying high-tech and high-cost efforts. For this reason, we typically use a simple trick – the sampling gas is dried before being passed through the analyzer. In other words, the water is taken out. But this also causes two errors in the measured values:

• As the water content is taken out, it will shift the percentages of the other gases in the sample.
• If the water is filtered out by condensation, it will react with ammonia to form ammonium hydroxide (Eq. 10). In the presence of CO₂, it will form ammonium bicarbonate (Eq. 11).

NH₃ + H₂O → NH₄OH (10)

NH₃ + H₂O + CO₂ → NH₄HCO₃   (11)  

These effects result in a false measurement in the range of ±2% absolute compared to the real ammonia content in the process gas.  

Typical Errors when Measuring with an Oxygen Probe
The measuring principle of an oxygen probe is based on the effect that zirconium dioxide is permeable for oxygen ions at temperatures above 350°C (662°F). If the two sides of a zirconia element are exposed to different oxygen partial pressures, there will be ionization and then a diffusion of the ionized oxygen atoms from the side with the higher oxygen partial pressure to the side with the lower oxygen partial pressure. In this way, equilibrium is established where the charge displacement created by the ions corresponds to the gradient in the oxygen partial pressures on the two sides of the element. The charge displacement can be measured as a voltage. If the oxygen partial pressure on one side of the element is known (reference), the oxygen partial pressure on the other side of the element (sample gas) can be calculated out of the measured cell voltage and the cell temperature (Fig. 7).

The relation between partial-pressure gradient, temperature and expected voltage is given by the Nernst Equation (Eq. 12). R is the universal gas constant, F is the Faraday constant and T is temperature in Kelvin. If the reference is operated with air, p⁰O₂ has to be set to 0.209.

The reading will be affected by the following:

• Zero drift – Especially when using Lambda probes, we have to take into account that such instruments typically show 12-50 mV when measuring air instead of the expected 0 mV.
• Thermo voltage – If an in-situ probe is using different metals for the inner and the outer electrode used to pick up the voltage, there will be an error that changes with temperature.
• Catalysis – At the tip of an oxygen probe there might be a catalytic dissociation of the ammonia. This will cause higher voltages than expected.
• Unknown, sometimes unstable cell temperature – When using probes that are not equipped with an internal thermocouple, the Nernst Equation has to be solved with an estimated temperature. Even if this temperature has been determined by a calibration, it might change during the process and lead to wrong assumptions in the measured partial pressures.
• Measurement current too high – Oxygen probes operate like a voltage source. If the analog input circuit of the measuring instrument has an input resistance that is too low, the measurement current will exceed the current supplied by the ion diffusion. This effect causes too low voltage readings.
• Electron conduction – Zirconium dioxide has the electrical property of an NTC resistor. The higher the temperature, the more the cell will get conductive, causing an electron current that decreases the expected Nernst Voltage.

Table 1 shows the effect of measurement errors on the determined oxygen partial pressure. The deviations do not look impressive if compared with the absolute magnitude of the measured partial pressures, but in a nitriding or nitrocarburizing process, the oxygen partial pressure is used to calculate the amount of water vapor. It is also used to calculate the partial-pressure ratio between CO and CO₂ in the furnace atmosphere. Water vapor and hydrogen are in relation to oxygen following the reaction:

H₂ + ½ O₂ → H₂O      (13)  

and establishing the thermodynamic equilibrium:

K = pH₂O / pH₂ • p^0.5O₂ (14)

With respect to the potentials K_N and K_CB this might lead to considerable deviations, shown in Tables 2 and 3.

Typical Errors when Measuring CO and CO₂
The measurement of CO and CO₂ is typically performed by using an infrared analyzer, just like the measurement of ammonia. Therefore, we have to account for the same inbuilt errors based on the general design. In addition, we have to ensure that the analyzer is ammonia-resistant. The measuring range has to match the conditions of a nitrocarburizing atmosphere.

The accuracy or, better to say, the uncertainty of the measurement will easily be in a range of several percent of the measuring range. On top of the uncertainty of the analyzer, we might encounter a shift from CO to CO₂ and vice versa as, together with H₂ and H₂O, the gas will try to establish the water-gas equilibrium at the sample gas temperature. This prefers higher CO₂ and lower CO percentages in the furnace. This effect will end in big deviations in the determined carburizing potential.

In a nitrocarburizing process at 580°C (1076°F) and an inlet gas mixture of 90% ammonia and 10% CO₂ with a controlled nitriding potential of K_N=1 and a set carburizing potential of K_CB=0.16, a shift of 0.5% from CO to CO₂ will change the measured K_CB to 0.10.

Typical Errors when Measuring Temperature
When measuring temperature, we first have to look at the specification of the components used. Thermocouples as well as measuring instruments have classified maximum deviations.

Assuming that in nitriding and nitrocarburizing furnaces the typical thermocouple in use would be a type-K, the maximum measuring error is ±1.5°C, according DIN IEC 584 – class 1. If the thermocouple is connected to a high-precision measuring instrument of class 0.1, we have to allow for another 0.1% of the measured temperature. At 580°C, this adds up to an uncertainty of ±2°C absolute.

Next are obviously effects like long-time drift but also deviations at the measurement of the reference temperature at the terminal block of the instrument.  

Independent of the quality of the measuring system, we have to account for remarkable deviations between the parts in the load. AMS 2750D allows for a maximum deviation of ±3°C throughout the hot zone in a furnace, using a class-1 industrial furnace.

Impact on the Result of a Treatment

How much do such errors influence the result of a real treatment? We will first look at a nitriding process, with the nitriding potential controlled and using a hydrogen analyzer to measure the potential. Equation 7 gives the way to calculate residual NH₃ from the measured H₂ percentage. K_N will be derived using Eq. 2.

The confidence limits can be estimated as shown in Eq. 15.  

E₁ is the error in the measurement of ammonia, and E₂ is the error in the measurement of hydrogen. As the H₂ is produced by the thermal dissociation of ammonia, E₁ can be expressed as a function of E₂ .

E₁ = 4/3 E₂       (16)  

Figure 8 displays the relative deviation of the determined nitriding potential compared to the real nitriding potential as a function of measuring errors in hydrogen. As both errors, E₁ and E₂ practically come to a doubled error – too low hydrogen reading will automatically cause too high ammonia reading and vice versa. It might be an advantage to use an additional ammonia analyzer but not necessarily.

When applying the allowed-for measuring errors on the Lehrer Diagram, we come to fuzzy phase boundaries. Figure 9 shows the impact of an error of ±1% hydrogen and ±5°C.

Consequently, when performing processes where the nitriding potential has to be controlled close to one of the phase boundaries, this fuzzy range has to be taken into account. The same applies for nitrocarburizing processes. To clarify this effect, Fig. 10 displays a section within the three-dimensional temperature/-nitriding-potential/carburizing-potential diagram ensuring the formation of an epsilon white layer. We can see that with increasing temperature we have to increase the carburizing potential to stay within the control window, while the nitriding potential is almost unaffected.

Generally speaking, for nitrocarburizing processes, it is best not to control potentials close to the phase boundaries but rather try to find an operating setpoint that will create the desired structure over a wide tolerance band.

Summary

When using the available instrumentation to measure and control nitriding and nitrocarburizing atmospheres, we might observe remarkable deviations from the desired outcome. Besides obvious reasons like a passivation that has not been removed completely by the cleaning process, these deviations might also be caused by the measuring and control system or by temperature deviations within the furnace.

The article explains how relatively small errors within the process chain add up and might therefore lead to unexpected results of a heat treatment. For this reason, it is essential to choose process parameters that account for the allowed uncertainties of the equipment used. Those errors will add up and end in a fuzzy range around the phase boundaries of the Fe-N and Fe-N-C diagrams that are used to determine the atmosphere potential setpoints.

The parameters finally used to control the treatment should be chosen in such a way that a sufficient safety distance is maintained. IH

For more information: Contact Dipl.-Ing. (FH) Karl-Michael Winter, PROCESS-ELECTRONIC GmbH, a member of United Process Controls, Heiningen; tel: +49 7161 94 888 0; e-mail: km.winter@process-electronic.com; web: www.group-upc.com