Phase transformations in Ti(1-x)Al(x)N coatings

M A S T E R’S T H E S I S

2006:246 CIV

SÉBASTIEN DURAND

Phase Transformations in Ti(1-x)Al(x)N Coating

MASTER OF SCIENCE PROGRAMME Materials Technology • EEIGM

Luleå University of Technology

Department of Applied Physics and Mechanical Engineering Division of Engineering Materials

Phase transformations in Ti _1-x Al _x N coatings

Student: Sébastien Durand Advisor: Hans Söderberg

Luleå University of Technology Division of Engineering Materials

In collaboration with:

SECO Tools AB Year 2006

Abstract:

This project focuses on phase transformations in Ti1-xAlxN coatings and particularly on their thermal aspect. SECO Tools AB Ti_1-xAl_xN films deposited by arc evaporation with 50 and 66 at. % aluminium have been investigated. This report presents an effective way to separate the ceramic coating from its substrate without altering the ceramic. EDX analysis and SEM pictures characterise the composition and the morphology of the coating powder. XRD patterns were used to confirm that the main growing orientation of this material was (002) and also to calculate the lattice parameters (4.20 Å and 4.17 Å) for respectively x = 0.50 and x = 0.66 films. DSC measurements were carried out on both compositions at different heating rates (5, 10, 15, 20 and 30 K/min). Results showed that three exothermic peaks appeared respectively between 600°C - 900°C, 900°C - 1100°C and 1100°C - 1300°C on each curve. It was also proved that the peak temperatures were increasing with increasing heating rates.

Activation energies were calculated for each one of these thermally activated phenomena and the influence of the composition on the material thermal response was also illustrated.

Index:

Introduction………...4

Ti1-xAlxN………5

- Generalities………...5

- Formation process……….5

- Spinodal decomposition………5

Experimental……….8

- Sample preparation………...8

- DSC………..8

- XRD………10

- TTT diagrams………..11

Results……….12

- SEM observation & EDX………...12

- XRD………13

- DSC……….14

• Ti50Al50N………14

• Ti34Al66N………17

Discussion………21

Conclusion………...22

Acknowledgements……….22

References………...23

Appendix……….25

Introduction:

My work can be seen as a continuation of Aldwin Arrigoni’s work [1]. This project also results from a collaboration between the company SECO Tools AB and Luleå University of Technology. Ti1-xAlxN is a metastable ceramic material developed, manufactured and sold by SECO Tools AB as a coating for cutting tools. It exists with different aluminium concentrations which can vary from 0 to 66 at.%.

This titanium-aluminium-nitride coating was designed to replace the conventional TiN coating. The addition of aluminium led to hardness increase and furthermore to improved corrosion resistance. Aluminium reacts with oxygen and forms an impenetrable layer of aluminium oxide, thereby preventing any further oxidation. As a result operational range was extended from 500°C for TiN to 800°C for Ti1-xAlxN.

Yet the phenomena involved in this unexpected hardness increase, and especially its temperature and time dependence, are still not totally understood and established. Their understanding could lead to the improvement of existing coatings or even to the creation of new coatings. In my project I will focus on the phase transformations that occur in this material.

The first step will be to retrieve the coating from the samples provided by the company and to check the success of the separation process. The second step will be the study of the phase transformations by using Differential Scanning Calorimetry (DSC). The third step would be the creation of a TTT diagram for each composition. However, this diagram would be based on heating, starting from room temperature, in contrast to the “normal” diagrams where cooling are carried out.

Ti_1-xAl_xN:

Generalities:

The elaboration of Ti_1-xAl_xN started during the 1980’s. Like most of the other ceramics used for cutting tools application, Ti1-xAlxN most remarkable properties include high comparative hardness, wear resistance, thermal stability, resistance to thermal shocks and corrosion. The colour of the coating is a good indicator for its composition. High aluminium percentage coatings are black, whereas pure TiN has a golden aspect. Optimization of cutting speed, cutting time and operator costs lead to the conclusion that one insert face should have a lifetime of 15 minutes, consequently the whole cutting tool insert generally last 1 hour [1].

Formation process:

As mentioned before, Ti1-xAlxN is a metastable material which means that it should not exist under normal equilibrium conditions. In order to achieve the creation of such temporary stable structure, a very energetic process is required. The most common deposition process for Ti_1-

xAlxN is PVD: Physical Vapour Deposition. PVD processing is carried out in high vacuum (10^-2 to 10^-4 Torr) at temperatures between 150 and 500°C. The high-purity solid coating material (titanium and aluminium) is either evaporated by heat (arc evaporation) or by ion bombardment (magnetron sputtering). Other PVD processes include e.g. ion plating.

During arc evaporation an electrical arc (15 000°C) hits the TiAl compound target and frees metallic ions, electrons and neutrals. A high degree of ionization is achieved. At the same time, a reactive gas (in our case partially ionised nitrogen, N2) is introduced; it forms a compound with the metal vapour and is deposited on the substrate as a thin, highly adherent coating. The parts to be coated are given a negative bias in order to attract the positively charged ions. The result is a very strong mechanical bond between the coating and the tool.In order to obtain a uniform coating thickness, the parts are rotated at constant speed about several axes. The properties of the coating (such as hardness, structure, chemical and temperature resistance, adhesion) can be accurately controlled. Finally annealing treatments enable stress relaxation inside the material [2].

Spinodal decomposition:

“A clustering reaction in a homogeneous, supersaturated solution (solid or liquid) which is unstable against infinitesimal fluctuations in density or composition. The solution therefore separates spontaneously into two phases, starting with small fluctuations and proceeding with a decrease in the Gibbs energy without a nucleation barrier.” [3].

Ti1-xAlxN is affected by spinodal decomposition. It results in the decomposition of the material into two more stable compounds: TiN and AlN. It is a thermally activated phenomenon and should also occur during deposition of the coating. Though it is believed that in that case the ion-surface interaction dominates over the weaker chemical potentials causing spinodal decomposition. However spinodal decomposition is likely to happen when the material reaches high temperatures during annealing or during its use as a cutting tool for example [2].

Fig. 2: Composition profiles drawn at different times during decomposition.

What is the thermodynamical explanation for spinodal decomposition? As mentioned in the definition above, it is linked to Gibbs energy and to be more specific to its second derivative.

Spinodal decomposition occurs spontaneously when ∂²∆G ∂X² <0. The two possible scenarios are presented by the graphic below (Fig. 1). In the first case the curvature is positive, which leads to ∂²∆G ∂X² >0 : nucleation and growth mechanisms dominate. In the second case the curvature is negative and it results in ∂²∆G ∂X² <0 and spinodal decomposition.

Apparently, when ∂²∆G ∂X² <0, the free-energy change is negative for an arbitrarily small fluctuation in composition such that one part of the system gets more concentrated at the expense of another. The system is inherently unstable and phase separation will proceed as illustrated (Fig. 2):

Fig. 1: Variation of Gibbs energy as a function of composition.

Ti_1-xAl_xN_(FCC) TiN_(FCC) + AlN_(FCC) TiN_(FCC)+ AlN_(HCP)

The following theories have been proposed to explain the decomposition occurring in Ti_1-

xAlxN. When spinodal decomposition occurs it is believed that the metastable FCC-Ti1-xAlxN decomposes into FCC-TiN and FCC-AlN. FCC-TiN is stable and will not evolve further in the current temperature range, but this is not the case for FCC-AlN. A second transformation takes place in the material when FCC-AlN transforms into HCP-AlN which is the stable configuration for AlN under the present temperature and pressure conditions.

These structural transformations have an impact on the mechanical properties of the material.

The two cubic structures formed during spinodal decomposition are coherent. Therefore hardness increases as extra stress is required to propagate dislocations through these domains.

The subsequent transformation of FCC-AlN into HCP-AlN is characterised by a volume increase of approximately 20%. Due to this significant volume variation the two coherent phases become incoherent. This lack of coherency leads to greater wear rate and a hardness decrease. As a conclusion spinodal decomposition is a real advantage that Ti1-xAlxN possesses over TiN because it provides Ti1-xAlxN with a hardness increase at high temperature, also known as age hardening (see Fig. 3) [4].

Fig. 3: Effect of annealing temperature on hardness for TiN and Ti₃₄Al₆₆N coatings [4].

Experimental:

Sample preparation:

In order to perform DSC analysis, the coating had to be separated from its substrate. DSC measurements require high purity powder to avoid secondary reactions and the particle size should be as small as possible to offer a better contact with the crucible. The material provided by SECO Tools AB consisted of ten 5x4 cm² thin iron plates coated with approximately 10 µm Ti₅₀Al₅₀N, and another ten identical iron plates coated with approximately 8 µm Ti34Al66N. The separation process necessitated to completely dissolve the iron without damaging the coating.

Concentrated HCl (37%) met these requirements. The plates were left during 2 days in an excess of concentrated acid. “Recycled” chemical bottles were used instead of normal beakers to avoid silicon contamination. After these 48 hours the iron has been completely dissolved by the acid, producing a greenish coloration due to Fe³⁺, and the coating lies at the bottom of the beaker. The acid is poured away and a small quantity of fresh acid is poured inside the beaker in order to perform a final cleaning of the coating particles.

After 2-3 hours the acid is removed again and the coating particles rinsed several times with deionised water to get rid of all the remaining acid and iron. Then the powder is collected and dried inside an oven during 3 hours at 70°C. In some cases the powder is grinded with a mortar if the particles are too big. Finally a composition check is performed with EDX to verify the powder purity. A simple observation with the optical microscope is sometimes enough to get a good idea of the powder cleanness.

An average mass of 60 mg is usually obtained for a single plate of 50/50 alloy, whereas only 35 mg for the 34/66 composition. This difference is explainable by the lower coating thickness for 34/66 as well as its lower density: aluminium (M = 27 g/mol) is lighter than titanium (M = 48 g/mol). It is also possible that the global separation output was lower for the 34/66 composition due to the smaller particles obtained.

DSC:

Differential scanning calorimetry (DSC) is a thermal analysis technique. This technique measures the energy necessary to establish a nearly zero temperature difference between a substance and an inert reference material, both materials being subjected to the same temperature. The temperature of the sample is programmed and may involve heating/cooling at a fixed rate or isotherms. This analysis requires a closed system which means that heat exchanges between the measurement chamber and the external environment have to be minimised as much as possible, like in other calorimetry techniques.

Two main types of DSC systems are commonly used: power-compensation DSC and heat- flux DSC. Both systems give equivalent results. In power-compensation DSC the temperatures of the sample and reference are controlled independently using separate, identical furnaces (Fig. 4). The electric heating power necessary to compensate the difference in heat flow is measured. In heat-flux DSC the sample and reference are connected by a metal disc (high thermal conductivity). The assembly is enclosed in a single furnace (Fig. 5). The

T S R T

R C T

S C

R T T R

T T R

T

H =T − − − = −

∆

temperature difference caused by the difference in heat flow between the sample and the reference is measured [5].

The equipment used during this project corresponds to the heat-flux type. The constantan heating disc is mounted on an alumina rod. The disc has two platforms on which the sample and the reference crucibles are mounted. Chromel and alumel wires are connected to the underside of each platform, thereby creating a thermocouple used to determine the differential temperatures of interest. The main assembly of the DSC cell is enclosed in a cylindrical chamber surrounded by the furnace chamber. A separate thermocouple serves as temperature controller for the programmed heating cycle. An inert gas is flowing inside the sample chamber and another one is used to protect the furnace chamber.

The enthalpy variation ∆H of the sample is equal to the difference between the heat flow to or from the sample Q_s and the heat flow to or from the reference material Q_r: ∆H = Q_S - Q_R. According to the thermal analogy of Ohm’s law: QS x RT = TC - TS and QR x RT = TC - TR

where R_T is the thermal resistance, T_C a constant temperature external to the sample and the reference, T_S the sample temperature and T_R the reference temperature. Combining the previous equations we obtain:

For each of our measurements we used helium inside the chamber and argon as protective gas. We set the quantity of coating powder to 40 mg in order to obtain sufficient signal strength. The sample was placed in alumina crucibles and an alumina lid was always put on top of the crucible to prevent the powder from dispersing into the chamber. Of course the reference was an identical empty crucible and its lid. After each run the crucibles were cleaned by using simple paper to remove the powder. If the previous cleaning was not enough or in case of composition change, the crucibles were boiled in aqua regia (5 parts HCl + 1 part HNO3) during 3 hours, then rinsed, dried, and finally glowed in an alumina furnace tube with oxygen atmosphere at 1500°C during 1 hour.

Because of previous oxidation problems, a procedure had to be created which would solve this problem. At high temperatures the small amount of oxygen present in the reaction chamber reacts with the coating, forming alumina which is recognizable by its white colour.

This unwanted reaction interferes with the signal from the phase transformations within the coating, thereby making it difficult to discern the peaks. A solution to this problem was to totally purge the reaction chamber from oxygen, and this while the sample was inside. This

“baking” procedure basically consisted of a turbopump creating a high vacuum (4.6 x 10^-4 mbar) and a thermal cycle (14 hours at 150°C). Of course the baking temperature was low

Fig. 4: Power-compensation DSC [5] Fig. 5 : Heat-flux DSC [5]

Fig. 8: Schematic view over a crystal and its interaction with radiation, leading to Bragg’s Law [7].

Fig. 6: X-ray diffractometer

XRD:

X-ray diffractometry (XRD) is an experimental technique that uses the fact that X-rays are diffracted by crystals. It is not an imaging technique. X-rays have the correct wavelength to be scattered by the electron cloud of an atom of comparable size. Based on the diffraction pattern obtained from the regular assembly of molecules or atoms in the crystal, the crystalline structure of the sample can be identified. Additional structural information that can be obtained with XRD includes lattice parameters.

Somehow, the X-rays have to be produced and recorded. This is done using a diffractometer.

Many designs of diffractometer are in use, but essentially they are all the same in how they work. A typical X-ray diffractometer is presented below (Fig. 6 and 7). The X-ray tube produces a high intensity X-rays beam and can be moved to hit the crystal with different angles. The sample must have a flat surface; it is positioned in the beam centre. The detector measures the intensity of the scattered radiation [6].

Crystals are three dimensional ordered structures than can be described as a repetition of identical unit cells. The unit cell is made up of the smallest possible volume which is representative of the entire crystal. The dimensions of a unit cell can be described with 3 edge lengths (a,b,c) known as lattice parameters and 3 angles (alpha, beta, gamma). The most common crystalline structures for metals are FCC, BCC (both characterized by a = b = c and α = β = γ = 90°) and HCP (a = b ≠ c and α = β = 90° and γ = 120°). The 3D location of atoms within a unit cell can be listed as their x, y, z Cartesian coordinates. The unit cell is determined from the diffraction of the x-rays using the Bragg equation (Fig. 8):

Fig. 7: Sample chamber and detector

) (h² k² l² d

a= × + +

In this equation n is the number of wavelengths (integer), λ the wavelength (Å), d the distance between two planes of atoms (Å) and θ the beam incidence angle (rad). Once the diffraction pattern analysed and its peaks associated with the correct crystal planes, it becomes possible to calculate the lattice parameters. The different crystal planes are defined by their Miller indices. The Miller indices of a crystal plane are constructed by determining the intersection of the plane with the x, y, z axes and then taking the reciprocals, by convention written as (hkl) [7]. For cubic structures:

TTT diagrams:

Time-temperature-transformation (TTT) diagrams are mostly used in metallurgy. This diagram is a plot of temperature versus time for a specific material. It is used to determine when transformations begin and end for an isothermal (constant temperature) or dynamic heat treatment. In steel metallurgy, these diagrams are very important to estimate which phases will be obtained (austenite, pearlite, bainite, martensite) for a given cooling rate and how long the transformation is going to last (Fig. 9).

Even for very different materials, most TTT curves have a common nose shape for medium temperatures. Left curves indicate the start of a transformation and right curves represent the finish of a transformation. A large number of measurements is obviously necessary to get an accurate diagram. The DSC curves should provide the starting and ending temperature for each transformation. Then knowing the heating rate used during the measurement it is possible to calculate the transformation time. Temperature and time can then be plotted on the diagram and transformation curves extrapolated from the experimental points. Usually TTT diagrams are read and used starting at high temperature with a subsequent cooling. In this project they are built and used the opposite way, hence starting at room temperature with a subsequent heating.

Fig. 9: Typical steel TTT diagram, showing four micro-constituents.

Results:

SEM observation & EDX:

An attempt to measure the 50/50 coating thickness gave us the following result: 6.16 µm (Fig.

10). Though the picture was not taken exactly from above and with a better perspective a value closer to 10 µm should result.

During the separation process we did get some particles with a roll shape (Fig. 11); even the

“flat-looking” ones were slightly curled as well. The natural tendency for the coating to curl is linked to the compressive residual stresses inside the material. Valleys are also visible on the picture and seem to indicate that the coating also would like to curl in the perpendicular direction. The amount of rolls we obtained was not constant for every batch and the reasons for this non-reproducibility are still at a hypothetical state.

The factors which may influence this phenomenon could be the following: coating thickness, material composition and acid quantity. For the same composition it has been observed a tendency to obtain more rolls for lower coating thickness. Deposition of a thicker coating induces more residual stresses inside the coating. This results in a higher tendency to fall apart during the separation process. A bigger amount of rolls has been observed for the 50/50 composition than for the 34/66 composition.

EDX analyses were also performed to check if the powder composition was matching the one announced by the company. A first analysis was carried out on a WC sample coated with Ti50Al50N. The expected elements (Ti, Al and N) were positively identified in the surface spectrum though the atomic percentage did not correspond exactly to our expectations (see Appendix 1). The percentage of Ti was very superior to 50% and the Al percentage much lower. Discussion about the EDX sensibility led to the conclusion that it was much more sensible to high atomic mass elements (such as Ti) than light elements (Al and N).

The second analysis made on the powder (Appendix 2) confirms this hypothesis. The results show the presence of Ti, Al, N as well as C, Si and Fe. The traces of Si and Fe are due to the separation process: Fe from the substrate and Si from the beakers. Their percentage remains

Fig. 10: SEM picture showing the coating cross-section.

Fig. 11: SEM picture of coating roll.

negligible. It is unsure if C could not be assimilated to N because of their very close atomic number or it could come from the carbon tape used to keep the powder in place. In this analysis the atomic percentage of Ti is still exaggeratedly high and the Al percentage too low.

In our case EDX appears to be very reliable qualitatively but not so reliable quantitatively.

XRD:

Two samples coated with approximately 10 µm Ti50Al50N and 8 µm Ti34Al66N respectively were analysed with XRD. In both cases the substrate was WC and their surface dimensions were 1.2 x 1.2 cm². The X-ray source produces Cu K_α radiations with a wavelength of 1.54186 Å. This K_α wavelength is actually composed by 2 parts of K_α1 = 1.54060 Å and 1 part of Kα2 = 1.54439 Å. The analysis was performed in Philips 4-axis MRD with 2θ values ranging from 30° to 100°. Both XRD patterns can be found in Appendix 3 and 4. The data was then analysed with XRD tables to identify the peaks. No Ti_1-xAl_xN tables could be found so regular TiN and AlN tables were used instead.

Most peaks on these curves actually belong to the substrate WC. All the interesting peaks correspond to cubic Ti1-xAlxN and no peaks matched the hexagonal one. The angles values for Ti50Al50N are almost identical to those of pure TiN though slightly shifted to higher values.

The most important peak is Ti₅₀Al₅₀N (002) with the peak maximum at 2θ = 43.1°. Other noticeable peaks include (111) at 37.0°, (022) at 62.7°, (311) at 75.3° and the two replicate (222) at 78.5° and (004) at 94.8°. When comparing the Ti50Al50N and the Ti34Al66N results as shown in Fig. 12, many similarities can be noticed.

Ti50Al

50N

Int (a.u)

30 35 40 45 50 55 60 65 70 75 80 85 90 95 100

2θ

Ti34Al

66N

37.0° 43.1° 62.7° 75.3°

The peak positions of the 34/66 composition matches almost exactly the 50/50 composition angles which were given above. Though it appears the 34/66 peaks are located at slightly superior angles, probably because of the slightly smaller lattice parameter of this composition.

Considering the peak heights it seems that the 50/50 composition renders a better signal due to the higher amount of Ti. It possesses a higher electron concentration than Al and therefore is more likely to diffract the X-rays. The average lattice parameter for Ti50Al50N and Ti34Al66N was calculated by using the equations presented in the experimental part and the result obtained is respectively a = 4.196 ±0.010 Å (Tab. 1) and a = 4.171 ±0.011 Å (Tab. 2).

Unit cell 2θ (°) d (Å) a (Å)

(111) 37,0 2,430 4,208

(002) 43,1 2,099 4,198

(022) 62,7 1,482 4,191

(311) 75,3 1,262 4,186

Average 4,196 S. Deviation 0,010

DSC:

In order to plot TTT diagrams for each composition, DSC measurements were carried out with a Netzsch STA 449 until 1400°C at different heating rates: 5, 10, 15, 20 and 30 K/min.

Due to equipment limitations the 30 K/min heating rate could only be maintained until 1000°C, then it had to be decreased to 20 K/min. The sample mass evolution was plotted as well to check for possible oxidation.

• Ti50Al50N:

The first measurements started with Ti50Al50N. A typical curve is provided below (Fig. 13).

This curve was obtained for a 20 K/min heating rate. The blue curve displays the DSC signal and the green curve the mass evolution. Generally these measurements were characterised by negligible mass increase (low oxidation) and by exothermic peaks looking the following way:

- a 1^st wide peak with medium height within the range 600°C - 800°C

- a 2^nd narrow peak with low height, quite discrete, within the range 800°C - 1000°C - a 3^rd wide peak with high height within the range 1100°C - 1300°C

Unit cell 2θ (°) d (Å) a (Å)

(111) 37,5 2,398 4,154

(002) 43,3 2,090 4,179

(022) 63,0 1,475 4,173

(311) 75,5 1,259 4,176

Average 4,171 S. Deviation 0,011 Tab. 1: Calculation of Ti₅₀Al₅₀N

lattice parameter

Tab. 2: Calculation of Ti₃₄Al₆₆N lattice parameter

200 400 600 800 1000 1200

-1.2 -1.0 -0.8 -0.6 -0.4 -0.2 DSC /(mW/mg)

98 99 100 101 102

TG /%

Peak: 740.6 °C

Peak: 1254.0 °C

[4.1]

[4.1]

↑ exo

The raw DSC curves were then processed. First the signal baseline was approximated with polynomial functions and subtracted to make the peaks more visible (Fig. 14).

Afterwards the curves were fitted with Origin 7.5 by using Gaussian multi-peaks (Fig. 15).

Fitting parameters included minimum enthalpy value y₀, peak maximum temperature x_c, width of peak at half maximum w_x and peak area A_x. Much iteration was necessary to make the software find the three peaks. Sometimes parameters had to be fixed and set free again after the software started to converge to the three peaks.

-0,1 -0,05 0 0,05 0,1 0,15 0,2 0,25

0 200 400 600 800 1000 1200 1400 1600

DSC (mW/mg)

DSC signal Peak 1 Peak 2 Peak 3

Fig. 14: Ti50Al50N DSC curve (20 K/min) with baseline subtracted.

Once the fitting procedure completed, the temperature for each peak maximum and for each heating rate could be determined and plotted (Tab. 3 and Fig. 16). Blue values in the table indicate uncertain values which had to be estimated by another way than Origin and red values have been affected by a heating rate change due to equipment limitations. The stars on Figure 16 represent the values obtained by Hörling et al. [8] for a heating rate of 50 K/min.

Ti50Al50N Variations

Heating rate (°C/min)

1st peak (°C)

2nd peak (°C)

3rd peak (°C)

1st peak (°C)

2nd peak (°C)

3rd peak (°C)

5 711 884 1189 15 15 15

10 721 935 1218 2 1 1

15 756 955 1259 2 4 1

20 741 967 1268 15 4 1

30 787 971 1265 2 1 1

Unfortunately the determination of the starting and ending point of each peak did not give any coherent result. Different criteria were tried but none gave anything satisfactory. The main problem for this determination is that these extremis are highly dependant on the chosen baseline. A close baseline would make the peaks look narrower and with far baselines the peaks appear wider. Fortunately the peak maximums were not so affected by this problem.

Knowing the peak maximum temperature and the associated heating rate, it became possible to calculate the transformation activation energies. According to Hörling et al. [8], it can be determined by plotting ln(1/T² x dT/dt) versus 1/(k_B x T) where T is the peak temperature

Tab. 3: Ti50Al50N peak maximum temperatures and variations

600 700 800 900 1000 1100 1200 1300 1400

0 50 100 150 200 250

Time (min)

Temperature (°C)

1st peak 2nd peak 3rd peak

Hörling et al. 50 K/min

Fig. 16: Ti50Al50N “TTT diagram”. Black stars correspond to values from Hörling et al [8].

(K), dT/dt the heating rate and k_B Boltzmann’s constant. The plotted points should be aligned and the slope of this line should be the activation energy (Fig. 17). It seems that the result dispersion is quite high for the first peak, so the activation energy for this transformation should be considered carefully. The slopes reveal E_a = 1.7 eV, 2.1 eV and 3.3 eV respectively for the 1^st, 2^nd and 3^rd phenomena.

• Ti₃₄Al₆₆N:

The same procedures and measurements were repeated for Ti₃₄Al₆₆N. A typical curve obtained with the same heating rate than the previous one (20 K/min) is presented below (Fig.

18). The blue curve still displays the DSC signal and the green curve the mass evolution.

Unfortunately this series of experiments commonly featured a sample mass increase revealing oxidation even though the baking procedure was applied as before. This problem was equipment-related; the presence of air being spotted around 800°C for other samples as well, and could never be completely resolved. It results that the peak amplitude was decreased, especially for the 3^rd one. Sources [4] and [8] confirm that the 3^rd peak should be the most important in amplitude, though in our case it seems that the 2^nd one is usually the highest. For this composition the peaks generally look the following way:

- a 1^st wide peak with medium height within the range 600°C - 800°C - a 2^nd narrow peak with high height within the range 900°C - 1100°C - a 3^rd wide peak with low height within the range 1100°C - 1300°C

y = -3,2914x + 13,183 R² = 0,8973

y = -2,0964x + 8,4109 R² = 0,911

y = -1,6737x + 7,8676 R² = 0,7814

-15 -14 -13 -12 -11 -10 -9

6,0 7,0 8,0 9,0 10,0 11,0 12,0 13,0

1/(k_B.T) (eV^-1) ln[(1/T2)(dT/dt)]

Peak 1 Peak 2 Peak 3

Fig. 17: Ti₅₀Al₅₀N activation energies

The peak maximum temperatures were calculated by Origin after the same fitting procedure and the values are in Table 4. Once again blue values in the table indicate uncertain values and red values have been affected by a heating rate. Ti₃₄Al₆₆N films have been widely studied by Hörling et al. [8] and therefore result comparisons are possible (Fig. 19).

Ti34Al66N Variations

Heating rate (°C/min)

1st peak (°C)

2nd peak (°C)

3rd peak (°C)

1st peak (°C)

2nd peak (°C)

3rd peak (°C)

5 662 980 1135 4 3 3

10 703 1007 1178 15 6 4

15 698 1001 1202 1 1 1

20 723 1027 1216 4 2 12

30 732 1024 1205 2 1 3

Fig. 18: Ti₃₄Al₆₆N DSC curve obtained for a 20 K/min heating rate.

200 400 600 800 1000 1200

Temperature /°C

-0.20 -0.15 -0.10 -0.05 0.00 0.05 DSC /(mW/mg)

98 99 100 101 102 103

TG /%

Peak: 720.3 °C

Peak: 1018.8 °C

Peak: 1216.9 °C

[1.1]

[1.1]

↑ exo

Tab. 4: Ti34Al66N peak maximum temperatures and variations

600 700 800 900 1000 1100 1200 1300

0,0 50,0 100,0 150,0 200,0 250,0

Tim e (m in)

Temperature (°C)

1st peak 2nd peak 3rd peak Hörling 1st Hörling 2nd Hörling 3rd

Puissance (Hörling 1st)

Fig. 19: Ti₃₄Al₆₆N “TTT-diagram” with Hörling et al. values.

When we compare our results (continuous lines) with Hörling’s ones (discontinued lines) a few comments can be made. Our values are always slightly higher though as they run in parallel to each other, they basically indicate the same trend. The incertitude on our points is quite low and two points marked with a red circle should be located at slightly higher temperatures if the heating rate could have been maintained. We also decided to calculate the activation energies for this composition. The same graph as before was plotted with the different temperatures and heating rates (Fig. 20).

Our values are plotted with full lines and Hörling’s values with discontinued lines like in the previous graph. If we compare our results with those we got for the other composition, it seems we get very similar results except for the second phenomena. For the 34/66 composition the equations reveal Ea = 1.8 eV (1.7 eV for 50/50), Ea = 4.4 eV (2.1 eV) and Ea

= 3.3 eV (3.3 eV) respectively for the 1^st, 2^nd and 3^rd phenomena. It is hard to explain why the results are so close for the 1^st and 3^rd peak and so different for the 2^nd. Now if we compare our results to Hörling’s results it appears that the 3^rd one is very similar and the two others are not, even if they are in the same magnitude order.

It is also possible to compare two “TTT-diagram” type curves obtained for different compositions but for the same heating rate. An example of this referred as Fig. 21 is given below. The heating rate for both compositions was 15 K/min. Such curves can also be found in Hörling et al. work [8]. Some trends are characteristic of these curves. First all peaks are exothermic. The 1^st 34/66 peak always appears before the 50/50 equivalent and the enthalpy measured is always lower. Secondly the enthalpies recorded for the 2^nd and 3^rd peaks increase with the amount of aluminium. According to [8] the 2^nd peak should be much weaker than the 3^rd one. Our 50/50 curve confirms this trend, though our 34/66 does not but this is probably due to the oxidation problem mentioned before.

y = -3,2806x + 14,09 R² = 0,8303

y = -4,4005x + 28,135 R² = 0,8191

y = -1,7872x + 10,067 R² = 0,8981 y = -3,6317x + 17,593

R² = 0,989

y = -2,903x + 15,539 R² = 0,9856

y = -2,79x + 22,45 R² = 0,9796

-14 -13,5 -13 -12,5 -12 -11,5 -11 -10,5 -10 -9,5 -9

6,0 7,0 8,0 9,0 10,0 11,0 12,0 13,0

1/(k_B.T) (eV^-1) ln(1/T2.dT/dt)

Peak 1 Peak 2 Peak 3 Hörling Peak 1 Hörling Peak 2 Hörling Peak 3 Linéaire (Peak 3)

Fig. 20: Ti₃₄Al₆₆N activation energies

Discontinued lines added for comparison are values from Hörling et al. [8].

Finally it is also necessary to mention that measurements were also carried out for both compositions at a very slow heating rate (2 K/min). We decided not integrate this data to the previous one because the results did not seem plausible. Due to the very slow heating rate the prolonged time at a high temperature produced more oxidation than compared to other heating rates, independently of composition. The maximum peak temperatures for Ti50Al50N were 766°C, 935°C and 1168°C. For Ti₃₄Al₆₆N only a single peak was spotted and it appeared at 815°C. In both cases it does not seem natural that these temperatures are higher than those reported for the 5 K/min heating rate. The general evolution of the peak position is actually the opposite way: the peak temperature increases with increasing heating rate. If these values were to be confirmed it would generate a new trendline on the TTT-diagram.

Fig. 21: Enthalpy vs. Temperature for 50/50 and 34/66 compositions (15 K/min).

Discussion:

XRD results show that Ti_1-xAl_xN grows along its preferred orientation (002). Other minor growth directions include (111), (022) and (311). The intensity of the XRD signal is linked to the concentration of titanium. Because of their higher electron density, rich titanium alloys diffract the X-rays more efficiently than aluminium-rich alloys. The calculation of Ti₅₀Al₅₀N and Ti34Al66N lattice parameters (respectively 4.20 Å and 4.17 Å) is coherent with the values found in other publications. According to reference [2], the lattice parameter of pure TiN would be 4.25 Å. The author specifies that his value is 0.01 Å larger than the reference value stated in references [9] and [10]. The lattice parameter of Ti_1-xAl_xN decreases with increasing amount of Al and this is explainable by the simple fact that Al is a smaller atom than Ti. As a consequence the 2θ angles associated with Ti34Al66N peaks are slightly higher to those associated with the corresponding Ti50Al50N peaks.

DSC measurements show that the phase transformations in Ti1-xAlxN can be characterized by three exothermic peaks. Concerning a 4^th peak mentioned in reference [4] its presence is possible though it could never be isolated or put in a prominent position. Reference [8] reports that Ti1-xAlxN DSC analysis usually feature a 1^st peak with medium height within the range 600°C - 900°C, a 2^nd peak with low height, quite discrete, within the range 900°C - 1100°C and a 3^rd peak with high height within the range 1100°C - 1300°C. Our Ti₅₀Al₅₀N DSC measurements match this description, though this was not the case for Ti₃₄Al₆₆N where a very important 2^nd peak and a smaller 3^rd peak appeared. This difference can partially be explained by the oxidation phenomena which occurred regularly during the Ti34Al66N analysis. For the same heating rate Ti₃₄Al₆₆N films show a more exothermic response than Ti₅₀Al₅₀N films.

The results for x = 0.50 and x = 0.66 Ti1-xAlxN films show that the thermically activated transformations occur at higher temperatures for increasing heating rates. Though it remains unclear how these transformation curves are going to evolve (slow- or fast-changing). The highest heating rate that was reached was 30 K/min and complementary analysis at higher rates would have been necessary to determine this evolution. The expected peak broadening with increasing heating rates could not be directly observed even though it seems possible with our data. Further comparisons with Hörling et al work [8] show that the peak temperatures we obtained were always a little bit higher though the plotted curves followed each other and globally indicated the same trend. These differences are probably due to different experimental parameters. Hörling used a Netzsch STA 409 equipment with Ar gas and 30 mg samples; we used a Netzsch STA 449 equipment with He gas and 40 mg samples.

Conclusion:

This project was a successful attempt to learn more about phase transformations in Ti_1-xAl_xN.

The first step was to separate the coating from its iron substrate with concentrated HCl and to make it into a powder that could be used for DSC measurements. EDX analysis confirms that sufficient purity was reached and SEM pictures testify that coating pieces have the tendency to curl up due to residual stresses. XRD patterns enabled to determine the main orientation of the coating as well as the lattice parameters for both x = 0.50 and x = 0.66 films.

During the second step DSC measurements were carried out on Ti₅₀Al₅₀N and Ti₃₄Al₆₆N at different heating rates. Results showed that three exothermic peaks appeared on each curve and that the peak temperatures were increasing with increasing heating rates. Activation energies were calculated for each one of these thermally activated phenomena. The influence of the composition on the material thermal response has also been illustrated. Unfortunately it didn’t become possible to draw complete TTT-diagrams or to study the width variations of the peaks because these widths could not be measured accurately.

Future work could focus on these peak width studies or include DSC measurements at higher heating rates to complete the data already gathered. It could also be of interest to perform similar analysis on other related coatings such as Ti_1-xSi_xN to check if similar behaviours could be pointed out.

Acknowledgements:

Special thanks go to Hans Söderberg and Magnus Odén for their advices and their constant support. Tommy Larsson, Jacob Sjölén and Lennart Karlsson from SECO Tools AB are gratefully acknowledged for providing the coating samples and for making this project possible. The author would also like to thank Jonny Grahn for the SEM pictures and EDX analysis.

References:

[1] “Transformations Kinetics in Ti1-xAlxN thin films”

by A. Arrigoni, Report, Div. of Eng. Mat., Luleå University of Technology, 2005 [2] “Thermal stability of arc evaporated high aluminum-content Ti1-xAlxN thin films”

by A. Hörling, L.Hultman, M.Odén, J.Sjölén and L.Karlsson, J. Vac. Sci. Tech. A, 20, 1815, 2002

[3] IUPAC Compendium of Chemical Terminology 2nd Edition, 1997 [4] “Self-organized nanostructures in the Ti–Al–N system”

by P. H. Mayrhofer, A. Hörling, J. Sjölén, L. Karlsson, T. Larsson, C. Mitterer, L. Hultman, Appl. Phys. Lett., 83, 2049, 2003

[5] “Differential Scanning Calorimetry”

by H. K. D. H. Bhadeshia, University of Cambridge

[6] http://www.chm.bris.ac.uk/webprojects2000/dedmonds/Dave's%20Documents/chemistry [7] http://hyperphysics.phy-astr.gsu.edu/hbase/quantum/bragg.html

[8] “Phase Transformations in Ti1-xAlxN Thin Films”

by A. Hörling, L.Hultman, M.Odén, P. H. Mayrhofer, L. Karlsson , C. Mitterer, M. J.

Frederick and G. Ramanath, unpublished

[9] Powder Diffraction File, JCPDS International Center for Diffraction Data, Swarthmore, PA, 1992; TiN (38-1420), AlN (25-1495)

[10] L. Karlsson, A. Hörling, M. P. Johansson, L. Hultman, and G. Ramanath, Acta. Mater., 50, 5103, 2002

Appendix:

Appendix 1:

Project 1 08/02/2006 10:38:04

Sample: Sample 2 Type: Default ID: 50_50

Element App Intensity Weight% Weight% Atomic%

Conc. Corrn. Sigma

N K 0.58 0.5852 2.95 0.61 8.87

Al K 2.63 0.9335 8.44 0.36 13.18

Ti K 28.94 0.9775 88.61 0.67 77.95

Totals 100.00 100.00

Appendix 2:

Element App Intensity Weight% Weight% Atomic%

Conc. Corrn. Sigma

C K 0.42 0.7432 2.18 0.40 7.15

N K 0.47 0.5330 3.38 0.66 9.50

Al K 2.19 0.9366 9.04 0.33 13.19

Si K 0.03 0.9568 0.11 0.11 0.15

Ti K 21.27 0.9696 84.72 0.90 69.61

Fe K 0.13 0.8923 0.56 0.56 0.39

Totals 100.00 100.00

Project 1 07/02/2006 10:35:46

Project: Project 1 Owner: jongra Site: Site of Interest 1

Sample: Sample 1 Type: Default ID: Powder

30 35 40 45 50 55 60 65 70 75 80 85 90 95 100

TiAlN (311) WC (200)

WC (102)

TiAlN (022)

WC (201)

WC (112) TiAlN (004) WC (111)

WC (002) WC (110) WC (101)

TiAlN (002)

TiAlN (111)

Ti50Al

50N

WC (100)

WC (001)

Int (a.u)

2θ

30 35 40 45 50 55 60 65 70 75 80 85 90 95 100

TiAlN (022)

TiAlN (311)

WC (200) WC (112)

Ti34Al

66N

WC (201) WC (102)

WC (111)

WC (002) WC (110) WC (101)

TiAlN (002)

TiAlN (111) WC (100)

WC (001)

Int (a.u)

2θ

Appendix 3:

Appendix 4: