Micromachining from a Materials Perspective

Abstract

Mechanical micromachining is a technique that has the potential to become a successful small-scale manufacturing process. If the abilities of macromachining can be scaled down to the microscale, then a versatile manufacturing technique will be created that is capable of processing a wide variety of materials. However, simply scaling down machines is not the solution, reducing the scale from macro to micro presents unique problems that must be overcome that include eliminating tool wear, machining heterogeneous workpiece materials in a uniform manner, and advances in machine tool design. Therefore, a review of the aspects of machining is presented that focuses on the machining process from a materials perspective.

4.1 Machining Theory

In the 1940s, Ernst and Merchant [1] and Merchant [2, 3] developed models for orthogonal cutting, i.e., the situation where the tool and the workpiece are both perpendicular to each other (Fig. 4.1), where undeformed material shears as it passes through a primary shear zone (Fig. 4.2). The simplified model ignores secondary shear that exists between the tool and chip. Earlier work by Piispanen [4] stated that the shearing process of metal is similar to cutting a deck of stacked cards (Fig. 4.3). The cards are inclined at an angle φ that matches the shear angle. As the cards approach the tool they are forced to slide over each other when they encounter the resistive force provided by the tool. There are many assumptions made to reach this point but the analogy helps visualize the cutting process. Merchant [2] began the metal cutting analysis by making the following assumptions.

Fig. 4.1

(a) Orthogonal cutting conditions and (b) oblique cutting conditions [5]

Fig. 4.2

Undeformed material passes through the shear zone [5]

Fig. 4.3

Piispanen’s [4] stacked card analogy of metal cutting

The following analysis is taken from Shaw [5] and is based on Merchant’s approach:

• 1. The tool is perfectly sharp and there is no contact along the clearance face.

• 2. The shear surface is a plane extending upward from the cutting edge.

• 3. The cutting edge is a straight line extending perpendicular to the direction of motion and generates a plane surface as the work moves past it.

• 4. The chip does not flow to either side.

• 5. The depth of cut is constant.

• 6. The width of the tool is greater than the width of the workpiece.

• 7. The work moves relative to the tool with uniform velocity.

• 8. A continuous chip is produced with no built-up edge.

• 9. The shear and normal stresses along the shear plane and tool are uniform.

The forces acting between the chip and tool can be identified and isolated in a free body diagram (Fig. 4.4). At equilibrium the force between the tool and chip R, is equal to the force between the workpiece and chip, R′. These resultant forces can be broken up into their constituent components.

Fig. 4.4

Isolation of forces on a free body diagram [5]

R is broken into forces perpendicular, N _C, and along, F _C, the tool face. R′ can be broken up into two sets of components, horizontal, F _P, and vertical, F _Q, relative to the direction of motion and along, F _S, and perpendicular, N _S, to the shear plane (Fig. 4.5). These forces can be rearranged and applied at the tool tip while being contained within a circle. This is Merchant’s circle of cutting forces and is fundamental to metal cutting theory (Fig. 4.6) [2, 3]. From this circle a range of equations can be generated that describe the cutting process. Only a few forces need to be measured with a dynamometer before these equations can be applied; the equations are listed below where α is angle between the vertical and the tool face called the rake angle and φ is the shear plane angle.

Fig. 4.5

Identification of the constituent components of force [5]

Fig. 4.6

Merchant’s circle of cutting forces [2, 3]

$$ {F}_{\mathrm{S}}={F}_{\mathrm{P}} \cos \varphi -{F}_{\mathrm{Q}} \sin \varphi $$

(4.1)

$$ {N}_{\mathrm{S}}={F}_{\mathrm{Q}} \cos \varphi +{F}_{\mathrm{P}} \sin \varphi $$

(4.2)

$$ {N}_{\mathrm{S}}={F}_{\mathrm{S}} \tan \left(\varphi +\beta -\alpha \right) $$

(4.3)

where β is the friction angle and μ is the coefficient of friction and is given by

$$ \mu = \tan \beta $$

(4.4)

$$ {F}_{\mathrm{C}}={F}_{\mathrm{P}} \sin \alpha +{F}_{\mathrm{Q}} \cos \alpha $$

(4.5)

$$ {N}_{\mathrm{C}}={F}_{\mathrm{P}} \cos \alpha -{F}_{\mathrm{Q}} \sin \alpha $$

(4.6)

$$ \mu =\frac{F_{\mathrm{C}}}{N_{\mathrm{C}}} $$

(4.7)

$$ \mu =\frac{F_{\mathrm{P}} \sin \alpha +{F}_{\mathrm{Q}} \cos \alpha }{F_{\mathrm{P}} \cos \alpha -{F}_{\mathrm{Q}} \sin \alpha } $$

(4.8)

$$ \mu =\frac{F_{\mathrm{Q}}+{F}_{\mathrm{P}} \tan \alpha }{F_{\mathrm{P}}-{F}_{\mathrm{Q}} \tan \alpha } $$

(4.9)

The shear stress τ is given by

$$ \tau =\frac{F_{\mathrm{S}}}{A_{\mathrm{S}}} $$

(4.10)

where

$$ {A}_{\mathrm{S}}=\frac{bt}{ \sin \varphi } $$

(4.11)

Here, b is the width of cut and t is the depth of cut, therefore the shear stress is

$$ \tau =\frac{\left({F}_{\mathrm{P}} \cos \varphi -{F}_{\mathrm{Q}} \sin \varphi \right) \sin \varphi }{bt} $$

(4.12)

Similarly, the normal stress σ is given by

$$ \sigma =\frac{N_{\mathrm{S}}}{A_{\mathrm{S}}} $$

(4.13)

$$ \sigma =\frac{\left({F}_{\mathrm{P}} \sin \varphi +{F}_{\mathrm{Q}} \cos \varphi \right) \sin \varphi }{bt} $$

(4.14)

An equation for φ is still required. It is found experimentally that when certain metals are cut there is no change in density. The subscript C refers to the chip and l is the length of cut.

Therefore,

$$ tbl = {t}_{\mathrm{C}}{b}_{\mathrm{C}}{l}_{\mathrm{C}} $$

(4.15)

It is found experimentally that if b/t ≥ 5, the width of the chip is the same as the workpiece; thus,

$$ \frac{t}{t_{\mathrm{C}}}=\frac{l_{\mathrm{C}}}{l}=r $$

(4.16)

where r is the cutting ratio, or chip thickness ratio

$$ r=\frac{t}{t_{\mathrm{C}}}=\frac{\mathrm{ABsin}\varphi }{\mathrm{ABcos}\left(\varphi -\alpha \right)} $$

(4.17)

where AB refers to the length of the shear plane. Solving for the shear angle φ,

$$ \tan \varphi =\frac{r\; \cos \alpha }{1-r\; \sin \alpha } $$

(4.18)

The work length may be determined by weighing the chip; if the chip weighs w _C and γ′ is the specific weight of the metal, then

$$ l=\frac{w_{\mathrm{C}}}{tb\gamma \hbox{'}} $$

(4.19)

The shear strain γ is given by

$$ \gamma =\frac{ \cos \alpha }{ \sin \varphi \cos \left(\varphi -\alpha \right)} $$

(4.20)

Expressions can also be developed for the cutting velocity V, which is the velocity of the tool relative to the work and directed parallel to F _P. The chip velocity V _C is the velocity of the chip relative to the tool and directed along the tool face. The shear velocity V _S is the velocity of the chip relative to the workpiece and direct along the shear plane.

$$ {V}_{\mathrm{C}}=\frac{ \sin \varphi }{ \cos \left(\varphi -\alpha \right)}V $$

(4.21)

or

$$ {V}_{\mathrm{C}} = rV $$

(4.22)

$$ {V}_{\mathrm{S}}=\frac{ \sin \alpha }{ \cos \left(\varphi -\alpha \right)}V $$

(4.23)

or

$$ {V}_{\mathrm{S}} = \gamma \sin \varphi\;V $$

(4.24)

The shear strain rate γ″ is given by

$$ \gamma "=\frac{ \cos \alpha }{ \cos \left(\varphi -\alpha \right)}\frac{V}{\Delta y} $$

(4.25)

where Δy is the thickness of the shear zone and is often approximated by assuming its value is equal to the spacing between slip planes. It is not always convenient to measure all the quantities required to apply the equations. Therefore, Merchant and Zlatin [6] produced a number of nomographs that can be used to obtain some of these quantities without the need for extensive experiments (Fig. 4.7).

Fig. 4.7

(a, b). Merchant and Zlatin’s nomographs [6]

Not all machining operations are similar to these idealized cutting conditions, for example, the process of milling. Not only are the contact conditions not orthogonal, they are oblique, but there are multiple cutting points. The surface marks produced by milling can be seen but feel smooth to touch; however, there are peaks and valleys similar to surface roughness. The height of these peaks and troughs or scallops, h, can be determined by

$$ h=\frac{f}{4\left(D/f\right)\pm \left(8n/\pi \right)} $$

(4.26)

where D is the diameter of the cutter, f is the feed per tooth, and n is the number of cutting teeth. Expressions are also available for the undeformed chip thickness and the length of cut. Under certain cutting conditions complications can arise with the assumptions made for ideal cutting. The strain and heat generated at the interface between the chip and tool, called the secondary shear zone, can be so high that part of the chip welds to the tool. This welded material effectively becomes part of the tool that changes its overall geometry and cutting conditions; this is called a built-up edge. In turn this affects the depth of cut and quality of surface finish. With time the built-up edge grows so large that the shear forces are high enough to remove it and another built-up edge begins to form, the process is therefore dynamic.

A secondary effect of a built-up edge is additional heat is generated. At the point where the chip adheres to the tool, its velocity is zero. At the free surface of the chip, the velocity of the chip is at a maximum and in between there is a velocity gradient. This gradient is made possible by internal shear of the chip, which is the source of additional heating. Therefore, the presence of a built-up edge disrupts the cutting process and ideally should be avoided.

In order to predict cutting forces the shear plane angle must be determined; Merchant continued his reasoning and assumed that the shear plane angle would take a value so that the total power consumed would be a minimum. Merchant also reasoned that the shear stress would be influenced by the normal stress; he arrived at an equation that Piispanen [4] had previously independently determined. However, depending on the assumptions made, a variety of equations can be generated for the shear angle. Stabler [7], Lee and Shaffer [8], and Oxley [9] developed such equations. For the conditions likely to be encountered an approach that best matches the assumptions and constants can be selected and acceptable results gained. The failure to generate a widely accepted single model therefore highlights the complexity of the cutting process due to the vast number of variables involved [10, 11]. A comprehensive review of metal cutting at the macroscale has been recently published by Shaw [5].

4.2 High-Speed Machining

High-speed machining is usually defined by spindle speeds between 30,000 and 100,000 revolutions per minute (rpm). Machining at high speed yields advantages such as increased metal removal rate, reduced cutting forces, increased dissipation of heat, and better surface roughness. Schulz [12] discussed the advantages that can only be gained by machining at high speed, which are summarized in Table 4.1. An unwanted effect of high-speed machining is that tool wear rate often increases.

Table 4.1 Summary of the advantages of high-speed machining discussed by Schulz [12]

However, Schulz [12] found that improving the following can extend tool life: (1) Cutting parameters, i.e., machine at an optimized feed rate and cutting speed; (2) Contact conditions, i.e., optimizing tool angles and tool/workpiece contact geometry; (3) Workpiece material, i.e., use an easy-to-machine workpiece where possible; (4) Cutting tool, i.e., ensure sharpness, correct geometry, and isolation from vibrations, also select a cutting tool material that has a low potential for diffusion into the workpiece; (5) Machining strategy, i.e., ensure the tool path does not impart large forces on the tool or workpiece. Schulz [12] found that if these five areas are optimized, then surface roughness produced is often better than roughness operations such as grinding. Eliminating postprocessing decreases lead time and cost providing high-speed machining with an advantage over conventional machining operations. The first application of high-speed machining was to cut aerospace materials such as titanium- and nickel-based alloys; cutting speeds between 30 and 100 m/min were achieved with conventional carbide tools. However, these speeds can be significantly increased if ceramic tools are used. Usui et al. [13] constructed a cutting model based on an energy approach; later Usui and Hirota [14] extended the model to examine chip formation and cutting forces with a single point tool. Finally, Usui et al. [15] investigated thermally activated wear mechanisms for ceramic tools. To establish the tool edge temperature, a 25 μm diameter wire was inserted into a Si₃N₄ tool, which is subsequently sintered and ground until the wire is exposed at the rake face. Experimental work by Kitagawa et al. [16] showed that tool temperature increases with cutting speed (Figs. 4.8 and 4.9) and that rake face temperature is higher than the flank face temperature. Usui [13, 14] suggests that tool wear is triggered by thermally activated processes and to a lesser extent by abrasive processes. Burr height decreases and the quality of surface roughness is found to increase with increasing cutting speed. Usui [13, 14] also observed at a cutting speed of 150 m/min that the temperature was 1,200 °C, which is close to the melting temperature of the workpiece alloy. The dominant wear mechanisms for Al₂O₃ + TiC and Si₃N₄ ceramic tools were found to be notch and crater wear, respectively, with the Al₂O₃ + TiC tool having the lowest wear rate. Kitagawa et al. [16] also found that Taylor’s tool life equation could be used for calculating the life of ceramic tools.

Fig. 4.8

A graph of the work of Kitagawa et al. [16] showing that an increase in cutting speed produces an increase in temperature

Fig. 4.9

A graph of the work of Kitagawa et al. [16] showing the variation of temperature on the tool with increasing cutting speed

Ozel and Altan [17] produced finite element models of high-speed machining by using a variable coefficient of friction to take account of the dynamic cutting situation at the tool–chip interface. The model predicts chip curl that is a result of large strains and thermal expansion. The degree of chip curl is found to be heavily dependent on cutting speed; above 300 m/min chip curl develops early, but if the feed per tooth is increased chip curl begins later. Higher cutting speeds therefore reduce the chip thickness. The model produced by Ozel and Altan [17] predicts cutting forces decrease with increasing cutting speed in the high-speed machining regime. Ozel’s model also predicts a rise in feed rate increases the cutting forces along the tool edge due to an increased chip load. An application of such models is their use in predicting some of the constants required by metal cutting theory.

Traditional approaches to machining yield useful information about the cutting forces but little regarding the stress distribution. A finite element approach can be used to predict stress distribution across the tool. It is shown that critical chip loads are applied when the cutting edge enters and exits the workpiece; stresses during chip formation are not significant. The rake face temperature is predicted to be higher than the primary shear zone temperature and the maximum tool tip temperature is found a short distance from the tool edge. This implies that frictional contact provides a further source of heating as the chip slides over the tool. During high-speed machining heat is generated in the primary shear zone and at the tool–chip interface. For steel, cutting temperatures are predicted to reach between 500 °C and 1,000 °C as shown by Moufki [18]. The coefficient of friction has a significant effect on heat generation but its value is not constant because it depends on dynamic factors such as the rake angle, feed, and cutting speed [19]. Therefore, precisely measuring or predicting the tool–chip interface temperature is a challenging problem.

Montgomery [20] showed that at large sliding velocities a reduction in the coefficient of friction is observed when the normal pressure or sliding velocity is similarly increased. Therefore, despite the large number of models that have been generated using finite element analysis to predict cutting forces, tool wear, and temperature, no existing model adequately accounts for all aspects of cutting and results are not always satisfactory, in particular predictions of temperature. The models are validated by comparing predicted temperatures to directly measured experimental data obtained using thermocouples demonstrated by Groover and Kane [21], infrared sensors demonstrated by Wright and Trent [22], and chip microstructure analysis demonstrated by Fourment et al. [23]. The models are then used to calculate cutting temperature in an attempt to determine if machining problems are likely to be encountered. For example, high temperatures can cause an unacceptable change in the workpiece microstructure or thermal softening of the tool. If either problem is predicted, corrective action can be taken prior to the onset of machining. Although Moufki [18] has shown that thermal softening is not usually sufficient to overcome strain hardening in the workpiece material. Finite element analysis by Kim and Sin [24] can accurately predict the way in which chips form. However, it is difficult to quantify the contributions made by different mechanisms because the situation is dynamic, i.e., the coefficient of friction is dependent on the sliding velocity or normal pressure, which changes as the tool advances (Figs. 4.10 and 4.11).

Fig. 4.10

The finite element mesh produced by Kim and Sim [24] to model the cutting process

Fig. 4.11

The cutting temperature distribution produced by Kim and Sim’s [24] finite element analysis to model the cutting process

Trent and Wright [25] observed during slow speed machining that the condition of chip sliding is dominant, during high-speed machining the condition of seizure is dominant, and somewhere in between there is a transition between sliding and seizure. Seizure occurs when the apparent area of contact equals the actual area of contact; this is in agreement with the work on friction carried out by Doyle et al. [26]. Seizure takes place when the normal stress exceeds the materials shear strength allowing complete contact between tool and chip. Seizure therefore is a major source of tool wear and is characterized by craters located near to the tool edge. Gekonde and Subramanian [27] predicted that these craters have a maximum depth that correlates to the phase change temperature rather than the maximum temperature as first thought. The phase change temperature is sufficient to cause dislocation generation, which leads to diffusion wear. Gekonde and Subramanian [27] also found that total tool wear is the net results of mechanical and chemical wear. Mechanical wear remains constant and is independent of the cutting speed, whereas chemical wear in the form of diffusion increases with cutting speed because conditions of seizure are more prevalent at higher speed. Therefore, at high speed the contribution to tool wear made by mechanical wear is inconsequential compared to the contribution made by chemical wear. Metal cutting theory is well established but its application can sometimes be difficult, especially when applied to milling; for example, Gygax [28] found the rapid periodic impact of each cutter often makes reading a dynamometer challenging.

Some metal cutting approaches require a number of variables to be measured before the theory can be applied in earnest. Such approaches are accurate but require a great deal of experimental time and effort. Other approaches only measure a few parameters and consequently are less accurate but easier to work with. Rotberg [29] decided the best approaches utilize a large number of sensors, a few experimental results, and idealized cutting conditions to generate a complete range of machining data. These models are useful because they can predict cutting forces and determine what conditions are likely to be encountered prior to machining, e.g., seizure or sliding. Schmitt [30] points out that high-speed milling is not always the most suitable method for machining.

It is usually employed specifically where its attributes are critical for successful machining of the part, e.g., thin ribs in the aerospace industry. Spindles are often selected by their power because this gives them the capability to machine a certain material. One of the main restrictions on spindle speed is the bearings capacity to operate at high speed. Schmitt [30], Weck and Staimer [31], and Ibaraki et al. [32] considered different methods of creating the working envelope. It was found that a hexapod structure is more stable than the construction of a conventional milling machine and produces better results in terms of surface roughness. This is because the hexapod (Fig. 4.12) does not suffer from vibrations caused by jerk (the rate of change of acceleration) that result from the high accelerations and speeds required for high-speed machining.

Fig. 4.12

The hexapod machine structure considered by Schmitt [30]

Moller [33] summarized the special requirements of high-speed spindles. He found that the best designs had no transmission, achieved high speeds with low vibration, were liquid cooled leading to high thermal stability, and had low inertia which allows high accelerations and decelerations. Another significant challenge identified by Moller [33] is the problem of selecting bearings suitable for operation at high speed. It is suggested bearings should be lubricated to reduce friction and aid heat dissipation, however at high speeds ceramic bearings perform better and the use of lubrication is brought into question. Moller [33] also showed the advantages of cutting at high speed include a reduction in force (Fig. 4.13). Cohen and Ronde [34] have proposed that hydrostatic bearings could operate successfully at high speed. A pressurized fluid held in reservoir by localized pockets provides the force to separate two surfaces. The hydrostatic bearing has a load capacity determined by the product of its pocket area and pocket pressure; this places a restriction on the size and shape of the bearing. Optimum rigidity occurs when the gap between both surfaces is a minimum. However a smaller gap produces a greater pressure loss, which results from a velocity profile created across the fluid by the high shaft speed. Additionally hydrostatic bearings eliminate run out, hence they are often used on machine tools that machine optical materials.

Fig. 4.13

Force reduction by using high speed cutting [33]

4.3 Cutting Tool Wear

Establishing the point at which the tool is considered worn is important because after this point machining results are no longer acceptable. If the onset of unacceptable tool wear can be identified, the tool can be changed with no impact upon the part. Tansel et al. [35] applied neural networks to this problem by training neural networks to monitor cutting forces, and based on past failures the critical point of future tool wear could be identified. However, Tansel et al. [35] found this method was only reliable with a soft workpiece. Ingle et al. [36] investigated crater wear, which is a major contributor to overall tool wear, which is composed of chemical and mechanical wear (Fig. 4.14).

Fig. 4.14

The contribution made to overall wear by chemical and mechanical sources [36]

Ingle et al. [36] machined calcium-treated AISI 1045 grade steel with a cemented carbide tool containing 97 % tungsten carbide, 0.4 % Ta(Nb)C, and 2.6 % cobalt. An expression based on the work of Bhattacharyya and Ham [37] and Bhattacharyya et al. [38] was developed to estimate the amount of tungsten transported by diffusion during the cutting time.

$$ W = 1.1284\;{C}_{\mathrm{o}}\cdot f\cdot t\cdot {\left(D/\tau \right)}^{1/2}\left(\mathrm{contact}\ \mathrm{area}\right) $$

(4.27)

where W is the amount of tungsten dissolved, C _o is the equilibrium concentration of W at interface, F is the ferrite volume fraction, t is the cutting time, D is the diffusivity of tungsten in ferrite, and τ is the tool chip contact time.

It was found if a tungsten carbide tool was used to machine steel, a TiN coating decreased the thermodynamic potential for dissolution by six orders of magnitude compared to the uncoated case. Ingle [36] also found that cutting speed affected the cutting temperature; at 50 m/min the temperature was 1,337 K and at 240 m/min the temperature was 1,512 K. A reason contributing to this temperature rise is the tool chip contact area also increases with increased cutting speed. In related work, Subramanian et al. [39] found diffusion is the main source of crater wear when machining AISI 1040 grade steels with a tungsten carbide tool at speeds of 175 m/min and higher. At these high cutting speeds the tribological condition of seizure dominates tool wear and therefore defines tool life. Subramanian et al. [39] found that during chip deformation thermoplastic shear occurs causing localized heating which is sufficient to allow diffusion of the tool material into the chip. Naerheim and Trent [40] observed a concentration gradient of tool material in the chip, specifically W and Co near the chip–tool interface. Subramanian et al. [39] conducted experiments coated and uncoated tools made from K1 cemented tungsten carbide containing 85 %WC, 4 %(TaC/NbC), and 11 %Co. For uncoated and coated tools the amount of tungsten present in the chips is measured by neutron activation analysis. Total tool wear is composed of mechanical wear and chemical wear. A loss of tungsten carbide particles from the workpiece characterizes mechanical wear and a loss of tungsten carbide by dissolution characterizes chemical wear. The initial amount of tungsten in the workpiece was 11 ± 0.5 ppm (parts per million). For the uncoated tungsten carbide tool at 150 m/min the amount of tungsten dissolved in the chip is 1.5 ppm, whereas at 240 m/min this amount is 10.5 ppm. However, the amount of tungsten carbide found in the chip is approximately constant, regardless of the cutting speed, around 0.8 ppm.

To test the effectiveness of a 10 μm thick CVD deposited HfN coating, experiments were performed at cutting speeds of 220 and 240 m/min and the concentrations of tungsten in the chips was found to be 11.3 ± 0.1 and 11.2 ± 10.1 ppm, respectively. It was therefore concluded that at high cutting speeds a thin HfN coating was successful at preventing dissolution of tungsten from the workpiece. Therefore, Subramanian [36] predicted that dissolution is prevented by coating the tool with a barrier layer that has the least thermodynamic potential for dissolution. For example, neutron activation analysis has shown a cemented carbide tool with a CVD deposited HfN coating drastically reduces diffusion. An alternative method to provide a diffusion barrier can be to engineer inclusions into the workpiece, e.g., oxides and sulphides, which can also aid chip formation. During testing of the HfN-coated tools, Subramanian et al. [39] employed a technique suggested by Hastings et al. [41] that was based on a method developed by Boothroyd [42] to determine the chip temperature. Coated tools were found to be 75 °C lower in temperature and regardless of cutting speed it was found that HfN coatings prevented dissolution crater wear.

Subramanian [36] also tested the effectiveness of adding inclusions in the workpiece; it was found that they were equally as effective as the HfN coating at cutting speeds between 220 and 240 m/s. However as cutting speed increased, the effectiveness of the inclusion barrier is reduced and diffusion wear increases. The contribution made by mechanical wear to the amount of tungsten in the chips is 0.8 ppm and is independent of cutting speed. However, the contribution of tungsten in the chips made by dissolution increases sharply with increasing cutting speed, e.g., at 150 m/min there is 66 % dissolution wear but at 240 m/min there is 93 % dissolution wear. Subramanian [36] concludes by stating that when machining steels, the solubility of tungsten carbide in austenite is increased at higher temperatures, which results from the condition of seizure. Additionally, the presence of ultrafine grain boundaries provides high diffusivity paths, which increase dissolution wear. If there is no coating, the tool can loose a large amount of tungsten by dissolution to the chips. Sherby et al. [43] found that creep resistance of a metal is higher than its self-diffusion activation energy. Sherby [44] found during high temperature deformation that the formation of subboundaries, grain boundary shear, and fine slip are related to interatomic diffusion. These static diffusion studies by Sherby [44] can help explain the dynamic diffusion present at the tool–chip interface. Gregory [45] studied diffusion between a cemented carbide tool of composition WC 74%, Co 10%, and TiC 16% and a workpiece material of Armco iron of composition C 0.05 %, Mn 0.05 %, S 0.028 %, P 0.01 % and Fe balance. Specimens were diffused under vacuum at temperatures of 1,100, 1,175, 1,250, and 1,325 °C (Figs. 4.15a, b). These results were compared to a tool subjected to orthogonal cutting of the Armco iron at 370 ft/min for 1 h with a feed rate of 0.0015 in./min. Gregory [45] produced graphs charting the composition of the diffusion couples indicate the first stage of diffusion is the outward migration of the cobalt binder phase followed by a build-up of titanium carbide in the reaction zone. The activation energy triggering cobalt diffusion was found to be 83 kcal/mol, which is comparable to 72.9 kcal/mol determined in another study by Suzuoka [46]. After the diffusion phase, a period of stable of wear sets in; this can be considered comparable to the run-in of a tool. Trent [47] suggests the stable period of wear is due to the presence of titanium carbide, which is more difficult to take into solution in steel compared to tungsten carbide. For titanium carbide to leave the tool material a thermal activation energy of 124 kcal/mol is required. However, for titanium carbide to diffuse into the iron–cobalt reaction zone energy levels of 150 kcal/mol are needed. Upon further examinations of the diffusion couples, Gregory [45] decided that wear resistance was also aided by a titanium carbide rich zone that builds up close to the wearing surface. Gregory [45] argued that final wear is explained by the continual diffusion of the titanium carbide zone. The activation energies of its constituents are: 98 kcal/mol for the inward diffusion of iron, 142 kcal/mol for the volume diffusion of iron in tungsten, and 77.5 kcal/mol for the outward diffusion of tungsten during the whole process. These energies point towards grain boundary diffusion mechanisms because Danneberg [48] found that the volume self-diffusion energy of tungsten was between 110 and 121 kcal/mol. Nayak and Cook [49] reviewed some thermally activated methods of tool wear. Nayak and Cook [49] found models attempting to predict thermally triggered wear mechanisms using continuum diffusion theory experience difficulties when the workpiece material and tool material are similar, e.g., cutting steels with high-speed tool steel. This is because the model assumes diffusion is driven by the concentration gradient of the diffusing material; hence there is a problem when this gradient is small. An alternate approach is to assume that vacancy concentration is the mechanism driving diffusion. This approach is successful in describing the wear of cemented tungsten carbide cobalt tools. Another approach is to explain wear as the result of a loss of hardness due outward diffusion of interstitial atoms into the chip, however this method is in poor agreement with experiments conducted. Consideration was given to a number of thermally activated processes that could trigger wear. Characteristics identifying the wear mechanism were determined so that an examination of experimental results would reveal the process responsible. Nayak and Cook [49] conducted a series of experiments and data was obtained using M2 tool steel hardened to HRC 65. Initially wear, temperature, and cutting forces increased rapidly followed by a prolonged steady-state period; eventually catastrophic wear occurs. Wear rate was defined as the rate of increase of crater depth. It was concluded from the best information available; whilst crater temperature distribution is uniform the wear rate is uniform. The size distribution of the wear particles was examined because this can yield information about the wear mechanism. For example, if large wear particles are observed, it is inferred that they result from asperity fracture caused by plastic flow. However, if different wear particles are observed, then plastic flow cannot have occurred and an alternative wear mechanism must be present. There is no evidence that an asperity fracture model can predict wear.

Fig. 4.15

(a, b). Examples of diffusion couples tested by Gregory [45]

A creep model could explain the triggering of wear but the wear rate is calculated at 4 × 10⁻¹² in./s, which is six orders of magnitude less than that observed; this trigger mechanism is therefore rejected. Depletion of interstitial carbon atoms was also investigated; if this mechanism were active, a loss of hardness leading to wear by plastic flow would be observed. Diffusion couples were constructed to test this possibility. 0.18 % C and 0.9 % C steels were heated to temperatures between 1,200 and 1,300 °F for several hours. Carbon diffusion was in the opposite direction to that expected; it was found that high-speed steel was carburized by low and high carbon content steel. Therefore, this mechanism is rejected because the tool should loose and not gain carbon.

Wear by the removal of oxide films is rejected because no change is observed when machining in an oxygen-free environment and there is not enough oxygen in the material to cause the required effect. Nayak and Cook [49] also considered tempering wear. At critical temperatures transition carbides such as W₂C and Mo₂C can form, thus reducing tool hardness due to tempering. The equation $V = 1.4\times {10}^{15}{\mathrm{e}}^{-90,000/RT}\;\mathrm{in}./\mathrm{s}$ can be derived for the tool wear rate, V, which agrees reasonably well with experiments conducted. R is the universal gas constant (cal/mol) and T is the absolute temperature (Kelvin). However, equations predicting other machining parameters generated during the derivation of this equation predict significantly different results to those observed. Atomic wear is also investigated, the energy required for atomic wear is provided thermally rather than mechanically by the applied stresses. Because tool atoms are in a higher energy state than chip atoms, overall migration is from tool to chip where atoms flow to low energy sites in the chip such as vacancies. This mechanism depends on the difference in free energy, the concentration of low energy sites in the chip, the activation energy required for an atom to jump from the tool to the chip, the available thermal activation energy, and the frequency of atomic vibration. If these factors are considered, an equation describing the wear rate V can be developed; thus V is equal to $\upsilon A{\mathrm{e}}^{-Q/RT}$, where υ is the atomic vibration frequency (1/s), a is the atomic spacing (in.), Q is the energy barrier for the atomic jump (kcal/mol), R is the universal gas constant (cal/mol), and T is the absolute temperature (Kelvin). Q and A variables are dependent on the work material. The tool–chip interface is considered similar to a grain boundary interface and therefore the activation energy Q required for an atomic jump is also similar; this is in the order of 40–45 kcal/mol as determined by Leymonie and Lacomb [50]. The atomic vibration frequency is approximately 10¹⁴/s and the atomic spacing approximately 10⁻⁸ in., yielding $v = {10}^6{\mathrm{e}}^{-42,000/RT}\;\mathrm{in}./\mathrm{s}$ which is in excellent agreement when cutting 1,018, 1,045, 4,140, and 10 C–40 C steels. At low cutting speeds mechanical wear is dominant, it can take two forms: abrasion due to ploughing and adhesion due to the formation and subsequent breakage of metallic bonds. These wear mechanisms depend on the hardness ratio between the tool and workpiece, environmental pressure and temperature, the presence of hard constituents and the presence of severe reagents.

A statistical model for the predictions of tool wear can be developed; Davis [51] suggested a Monte Carlo approach for modeling mechanical wear based on the following assumptions. Quanta of energy are added to the rubbing particles all of which leave the same site. The quanta of energy are added to the particles at domains randomly spaced over the surface at moments randomly spaced in time. The quantized energy of a particle over the surface diffuses continuously into the material at a known rate. Finally, the surface energy of a particle causes it to detach as a wear particle. Bhattacharya and Ham [37] extended this approach developing expressions for the width of flank wear due to mechanical wear from abrasive and adhesive sources. The width of the flank wears, H_f (T), at time T is given by

$$ {H}_{\mathrm{f}}(T) = K^{{\prime\prime} }{V}_{\mathrm{C}}{T}^{1-\alpha } $$

(4.28)

where V _c is the cutting velocity, T is the cutting time, α is the clearance angle, and K″ is given by

$$ K^{{\prime\prime} } = 3\ K^{\prime }/2C^{\prime}\left(1-\alpha \right) $$

(4.29)

where C′ is given by

$$ C^{\prime } = {C}^2/{\sigma}^2 $$

(4.30)

where C is a constant, σ is the deviation from the mean, and K′ is given by

$$ K\hbox{'}=\frac{Km}{\rho \left(\frac{ \tan \alpha }{1- \tan \gamma\;\tan \alpha}\right)} $$

(4.31)

where ρ is the density of tool material, m is the mass of the wear particle, γ is the true rake angle, and K is the constant governing decay rate. The height of flank wear, h _f, can also be determined by

$$ {h}_{\mathrm{f}}=\left[\frac{5}{4}\frac{K{V}_{\mathrm{c}}{A}^{3/4}\left(1- \tan \alpha\;\tan \gamma \right)T}{b\; \tan \alpha}\right] $$

(4.32)

where A is a constant based on the tool and workpiece combination and b is the width of cut. This model worked well when applied to experiments conducted by Bhattacharyya et al. [38] and experiments conducted by the internationally recognized tool wear collection body OECD/CIRP. However, a large amount of data for specific tool and workpiece combinations must be collected before this approach can be used. Bhattacharyya et al. [38] point out the reliability of the model depends on the accuracy of the data, making this approach susceptible to compounded errors. It is also inconvenient to collect a large amount of data every time new machining conditions are encountered.

4.4 Tool Coatings

In environments where tools experience high wear forces and extreme pressure, sintered tungsten carbide tools are used. However, ultrahard materials such as silicon alloys, abrasive materials, synthetic materials, and composites can only be machined with diamond-coated tools. Faure et al. [52] discussed the main issues involving diamond coatings.

Thin uncoated tool edges are susceptible to rounding and therefore must be protected. Coating the tool with a hard material such diamond does not offer sufficient protection because there is a sharp hardness gradient between the tool and its coating. Upon impact this gradient causes the coating to break therefore exposing tool material to the usual wear mechanisms. This effect can be reduced if an interlayer is introduced with hardness between the tool material and diamond, e.g., TiC/TiN. Such an interlayer reduces the severity of the hardness gradient and enhances adhesion between the diamond and tool. Diamond is an ideal coating material; it has high hardness, high wear resistance, and is chemically inert. However it is difficult to coat steels, Ni alloys, cemented carbides, and alloys containing transition metals with diamond. It is possible to coat WC–Co, but the cobalt must be etched away or a diffusion barrier introduced [53].

Faure et al. [52] applied a TiCl₄ precursor to the tool before depositing an interlayer in a CVD reactor at a temperature between 900 and 1,000 °C. One percent methane in hydrogen gas is introduced at a flux of 2,000 sccm at a pressure of 20 mbar to a CVD chamber that has a tantalum filament, graphite work holders, and a working surface area of 200 cm². These conditions allow a diamond film to grow on the tool. The tool material coated was a 6 % Co tungsten carbide. A cobalt etch was undertaken before deposition of a TiN interlayer and a TiN/TiC multilayer. Seeding with a solution of diamond micrograins was performed prior to deposition of the diamond coating. The coating effectiveness was assessed by a “Revetest” test device and drilling experiments. The “Revetest” device applies a constant force to an indenter during constant velocity displacement of the sample and critical loads are identified by an acoustic signal (Fig. 4.16a, b). Drilling tests were performed at 69,000 rpm at 3.5 m/min with a 1 mm diameter drill. The critical load causing coating failure is a function of the substrate hardness and the strength of adhesion between coating and tool. It is common for a soft substrate to deform plastically while its coating does not because it has a high Young’s modulus. The coating alone bears the load and fails thus exposing tool material to wear mechanisms (Fig. 4.17). Diamond coatings with a TiN interlayer (Fig. 4.18) can withstand ten times more force compared to diamond coatings without the interlayer; adhesion is not necessarily better but toughness is improved. Both cases with and without an interlayer offer a significant improvement over tools with no coating where rounding of the tool edge is observed after drilling only one hole (Fig. 4.19).

Fig. 4.16

(a, b). Scratch tests performed by Faure et al. [52] to determine the adhesion of diamond coatings

Fig. 4.17

Fracture of the coating [52]

Fig. 4.18

A drill produced by Faure et al. [52] with an interlayer and a diamond coating before testing

Fig. 4.19

An uncoated drill after drilling one hole [52]

If a TiC/TiN/TiC multilayer is applied, large forces, around 100 N, are required to break the surface. It was found that thicker interlayers improved adhesion between the tool and diamond coating and interlayers with a Young’s modulus between that of the tool and diamond coating produced better machining results. Uncoated tools were able to drill 10,000 holes and coated tools drilled 20,000 holes before they were considered worn. Uncoated tools experienced rounding of the cutting edge.

Bell [54] categorized tool materials into three basic groups: high-speed steels, cemented carbides, and ceramic and superhard materials including alumina-based composites, sialons, diamond, and cubic boron nitride. In addition, the surface properties of these materials can be modified with the coatings varying in thickness between 10⁻¹ and 10⁴ μm (Fig. 4.20). Tool wear can be defined by the following description, “A cutting tool is considered to have failed when it has worn sufficiently that dimensional tolerance or surface roughness are impaired or when there is catastrophic tool failure or impending catastrophic tool failure.” To prevent failure, Mills [55] suggests surface modification techniques such as coating should be employed to improve tool performance. PVD has the capability to deposit wear-resistant ceramic layers on high-speed tool steels. Although CVD can be used PVD appears to be advantageous in terms of coating adhesion and producing coatings with compressive stresses. It is increasingly being discovered that cubic boron nitride is an effective tool coating. Different machining processes are characterized by different wear mechanisms (Fig. 4.21) and the choice of tool coating should be selected to offer the best protection for a particular set of machining conditions; e.g., the combination of wear rate, bearing pressure, and tool material (Fig. 4.22).

Fig. 4.20

Thickness of surface layers and their method of manufacture [54]

Fig. 4.21

Different wear mechanisms on a cutting tool [55]

Fig. 4.22

Normalized wear rate, ka, plotted against bearing pressure for a range of tool materials [55]

Tool coatings can modify contact conditions, which alters the coefficient of friction, in turn this alters heat generation and heat flow; there are four main types of coating. The first category is titanium-based coatings such as TiAlN; other elements are added to improve hardness and oxidation resistance. Titanium-based coatings are popular because they provide a wide range of average protection properties, have good adhesion, and are relatively easy and quick to coat. The second category is ceramic-based coatings, e.g., Al₂O₃, except for this example ceramics have good thermal properties and excellent resistance to wear but are difficult to deposit. The third category is superhard coatings such as CVD diamond. The fourth category is solid lubricant coatings such as amorphous metal carbon, Me-C:H. Combinations of these coatings can give the best wear resistance; for example, a recent development has been to take superhard coatings and deposit low friction MoS₂ or pure carbon on their surfaces. Kubaschewski and Alcock [56] concluded that to prevent the onset of diffusion the enthalpy of the coating must be as negative as possible to increase the temperature at which diffusion is triggered. From this point of view most carbide coating materials such as TiC, HfC, and ZrC are more suitable for cutting steel than tungsten carbide, similarly for the nitrides except CrN up to a temperature of 1,500 °C.

The technique used to coat the tool can affect its performance; CVD requires high temperatures that can have an annealing effect on the tool. This affects the tool’s toughness and rupture strength because a brittle η phase (Co _X W _Y C _Z )² forms. A standard CVD process operating at 1,100 °C can reduce the materials strength by 30 %. PVD techniques such as evaporation, sputtering, and ion plating usually take place between 200 and 500 °C avoiding such problems. Eventually, multilayer coatings were developed. Klocke and Krieg [57] summarized their three main advantages: (1) Multilayer coatings have better adhesion to the tool. This is because the adhesion between the tool, intermediate layer, and coating is better than the adhesion between the tool and coating alone; for example, the presence of TiC in a CVD TiC-Al₂O₃-TiN coating system; (2) Multilayer coatings have improved mechanical properties such as hardness and toughness. For example, the advantages of a (Ti, Al)N layer can be combined with advantages of a TiN layer (good bonding and high toughness) by building up alternate layers of each coating. Layering is also thought to provide a barrier to resist crack propagation; and (3) Each level of a multilayer coating can provide a different function. For example, a coating could consist of a mid-layer with high thermal stability and a top layer with high hardness to produce a coating that maintains its hardness at high temperatures. It is possible for coatings to protect the substrate from heat if they are good insulators, have low thermal conductivity, and have a low coefficient of heat transfer. An example of such an improvement is a TiN-NaCl multilayer, which has hardness that is 1.6 times greater than a single TiN layer. (Al, Ti)N-coated tools have been compared to TiN and uncoated tools when milling at 600 m/min. The (Al, Ti)N-coated tools exhibit much less flank wear, which correlates to a higher hardness, 2720 HV for (Al, Ti)N vs. 1930 HV for TiN and an improved oxidation temperature, 840 °C for (Al, Ti)N vs. 620 °C for TiN. The improved performance of (Al, Ti)N-coated tools is due to their ability to maintain a higher hardness at elevated temperatures.

It has been noted that in some extreme cases of machining highly abrasive materials, e.g., hypereutectic Al-Si alloys where the Si particles are highly abrasive, variations of Ti-based coatings do not improve the tool life. In such cases only the hardness of diamond coatings can improve the abrasion resistance and therefore prolong tool life. Quinto et al. [58] investigated coatings deposited by CVD and PVD techniques on two alloys A and B (Figs. 4.23, 4.24, 4.25, and 4.26). Coatings with an Al content tend to perform better regardless of the application, coating process, or chemical content of other coating constituents. This is because abrasive resistance, oxidation resistance, and hardness are all improved. Thermal relief experienced by the substrate is of special interest in terms of volume effects like fatigue and diffusion. Examples of coatings offering this relief are PVD (Ti, Al)N and CVD TiC-Al₂O₃-TiN. Quinto et al. [58] found that PVD coatings outperform CVD coatings that outperform uncoated tools (Figs. 4.27, 4.28, and 4.29).

Fig. 4.23

CVD deposited TiN on alloy A [58]

Fig. 4.24

CVD deposited TiN on alloy B [58]

Fig. 4.25

PVD deposited TiN on alloy A [58]

Fig. 4.26

PVD deposited TiN on alloy B [58]

Fig. 4.27

Appearance of an uncoated insert at the end of tool life after machining 1045 AISI steel [58]

Fig. 4.28

Appearance of an TiN CVD-coated insert at the end of tool life after machining 1045 AISI steel [58]

Fig. 4.29

Appearance of an TiN PVD-coated insert at the end of tool life after machining 1045 AISI steel [58]

Dry machining of steels in the range of 55–62 HRC at 15,000–25,000 rpm generates cutting temperatures of 1,000 °C and the tool must be protected from oxidation wear. Previous work by Munz et al. [59] has shown that above 800 °C diffusion of stainless steel is triggered and cavities begin to form between the substrate and coating, although this can be prevented by adding 1 % yttrium. One method used by Constable et al. [60] to analyze the integrity of coatings is Raman microscopy. It has been used to study wear, wear debris, stress, oxidation, and the structure of single layer PVD coatings; it is now being used to study multilayer PVD coatings. Raman microscopy works by analyzing photon interactions with the lattice to determine the characteristic vibrations of the molecules. Thus the process is sensitive to lattice rearrangement that can occur as a result of local changes in material properties due to high cutting temperatures. Constable et al. [61] demonstrated the usefulness of Raman microscopy when a PVD combined cathodic arc/unbalanced magnetron deposition system was used to coat high-speed steel and stainless steel for abrasion tests. The coatings had a thickness between 2.5 and 4 μm with a surface roughness of 0.02–0.03 μm Ra (roughness average). Polycrystalline corundum with a hardness of 1900 HVN and Ra of 0.2 μm was brought into contact with the PVD coating. A constant load of 5 N was applied and the relative sliding speed between the two surfaces was 10 cm/s. A 25 mW HeNe laser with an excitation wavelength of 632.8 nm was used to obtain Raman results. The PVD coating consisted of a 1.5 μm thick TiN base layer with alternating layers of TiCN, 0.4 μm thick and TiN, 0.6 μm thick; the total thickness was 3.9 μm. The expected wear debris was rutile, TiO₂, which yields a peak at 426 cm⁻¹; however peaks were also detected at 512 and 602 cm⁻¹ indicating anastase debris. This would suggest the contact temperature was lower than expected, indicating the coefficient of friction was also lower than expected. Deeming et al. [62] investigated the effect different coatings have on delaying the onset of oxidation. During high-speed machining, temperatures regularly exceed 900 °C. Deeming found a TiN coating delayed oxidation until 500 °C, a TiAlN coating delayed oxidation until 700 °C and a multilayered system delayed oxidation formation until 950 °C. The final coating tested was TiAlCrYN, prior to deposition Cr is added to etch the tool thereby achieving the smoothest, most strongly adhered, dense coatings possible. The addition of yttrium increases wear at low temperatures; however at higher temperatures yttrium causes maximum wear to occur at 600 °C and minimal wear to occur at 900 °C. Without the addition of yttrium the wear rate continually increases with temperature. Deeming et al. [62] suggest yttrium diffuses into the grain boundaries and at high temperatures there is some stress relaxation.

Heat-treated TiAlCrYN therefore has lower internal compressive stresses than regular TiAlCrYN. TiAlCrYN also has a lower coefficient of friction with increasing temperature compared to TiN (Fig. 4.30) and also has a lower wear rate at elevated temperatures (Fig. 4.31). A problem with PVD coatings identified by Creasey et al. [63] is when ion etching is used to evaporate target materials; subsequent deposition by magnetron sputtering can lead to the formation of droplets which adhere badly to the surface and cause weaknesses in the coating. This is particularly problematic when depositing TiAlN. It has been shown by Munz et al. [64] the melting temperature of the cathode material influences the number and size of these droplets. Munz compared TiAl and Cu targets, TiAl had a higher melting point and produced a droplet density of 100 × 10³/mm² compared to a droplet density of 60 × 10³/mm² for the lower melting point Cu. Comparing droplet size, up to 1 μm TiAl had twice the number of droplets as Cu. Between 2 and 4 μm the number of droplets was the same and between 4 and 8 μm more Cu droplets were detected. It was expected Cu would produce more particles for all sizes of droplet, but due to its high vapor pressure Cu can escape the surface and is therefore not deposited. It is noted higher melting point targets exhibit spherical droplets suggesting they have solidified before contact with the surface and are subsequently included between the layers. Lower melting point targets show none spherical droplets. Generally higher melting point targets reduce the number and size of defects when depositing TiAlN with UBM. An alternate to the Ti-based coatings is CrN _x , which can be deposited at temperatures as low as 200 °C, the oxidation temperature for this coating is 700 °C. There are a number of ways to deposit this coating in addition to magnetron sputtering. Gahlin et al. [65] have demonstrated cathodic arc deposition and Wang and Oki [66] have demonstrated low voltage beam evaporation deposition. The pressure of the N₂ gas defines which of the different phases result, Cr, Cr-N solid solution, Cr₂N, Cr₂N + CrN, or CrN. Cr₂N has the hardest coating between 1,700 and 2,400 HV. Hurkmans et al. [67] deposited the coatings using a combined steered arc/unbalanced magnetron-arc bond sputtering (ABS) technique. Ferritic stainless steel (Cr18) and high-speed steel (M2) were etched either by inert Ar ion or Cr metal ion, and then polished to a Ra of 0.01 μm. The coating temperature was a constant 250 °C and the thickness a constant 3 μm. It has been found low adhesion problems can be overcome by using the ABS technique and conducting a pre-etch at 1,200 eV Cr prior to UBM film deposition; this can more than double the adhesive force from 25 to 50 N. Both TiAlN and CrN coatings exhibit good properties for dry high temperature machining; Wadsworth et al. [68] reasoned that a multilayer composed of these materials would produce an optimal coating for preventing tool wear; such a coating can be produced using PVD techniques. Wadsworth deposited coatings with and ABS system at 450 °C and a base pressure of 1 × 10⁻⁵ mbar. The targets were three Ti_0.5Al_0.5 plates and a Cr plate, the substrates were various sized sheets of 304 austenitic stainless steel and 10 mm diameter 10 mm long cylinders of heat-treated M2 high-speed steel (HRC ~ 62), polished to an Ra of 0.01 μm. A 0.2 μm thick TiAlN base layer was deposited prior to the TiAlN-CrN multilayer; the final coating thickness was 3.5 μm.

Fig. 4.30

Comparison of the coefficient of friction produced by TiN and TiAlCrYN coatings [62]

Fig. 4.31

Comparison of the wear rate produced by TiN and TiAlCrYN coatings at elevated temperatures [62]

A series of coatings were tested at elevated temperatures. It was found that up to 700 °C the coatings were unaffected, with the hardness dropping from an initial 3,500–2,800 Knoop hardness (HK_0.01) after 16 h.

However if the coatings were held only 50 °C higher for the same length of time the hardness dropped to 475 HK_0.01. The reduction in hardness is thought to coincide with a breakdown in the multilayers and oxidation of the exposed surface. The addition of Cr however reduces the onset and rate of oxidation. Smith et al. [69] coated various substrates and performed a series of tests. The two coatings tested were TiAl and TiAlN with a Cr etch. The growth defects previously discussed were highlighted in the TiAl coating (Fig. 4.32) while the Cr etched TiAlN coatings was defect free (Fig. 4.33). The substrates coated were 30 mm diameter and 5 mm thick cylinders made from M2 high-speed steel (HRC ~ 65) with an Ra = 0.005 μm, 25 × 25 × 0.8 mm bright annealed 304 stainless steel plates and 8 mm diameter HSS twist drills. The drills are of particular interest and were initially tested at 835 rpm and a feed rate of 0.28 mm/rev, for each subsequent test the spindle speed was incrementally increased. It was found TiAlN-coated drills outperformed commercially available drills. This improvement was partially attributed to the smoother surface observed on the TiAlN coating, which was enhanced further with a Cr etch. Petrov et al. [70] also experienced the droplet formation described by Munz during the coating procedure. Polycrystalline Ti _l−x−y Al _x N alloys and Ti _l−x Al _x N/Ti _l−y Nb _y N multilayers exhibiting smooth flat layers up to a total film thickness of 3 μm were grown on ferritic b.c.c. stainless steel substrates at temperatures around 450 °C using unbalanced magnetron sputtering and cathodic arc deposition. Petrov et al. [70] prepared the substrate for final deposition by performing cathodic arc etch using a Ti_0.5Al_0.5 or Ti_0.85Nb_0.15 target. Ti_0.5Al_0.5 has a lower melting point and results in large droplet formation, which decreases the integrity of the film. Ti_0.85Nb_0.15 however has a higher melting point and produces less droplets that are much smaller in size. Under UBM deposition conditions the films exhibited columnar growth with a compressive stress around 3 GPa and around 1.5 % trapped Ar. For a thickness of 3 μm peak-to-peak roughness was observed at 50 nm. Using a UBM/CA deposition technique a thickness of 3 μm peak-to-peak roughness was observed at only 5 nm. Higher compressive stresses were observed, 9 GPa with an Ar concentration of 0.5 %. Further materials that would be suitable for coating applications are niobium and tantalum because they are extremely stable; but they are difficult to deposit using conventional methods. However an ion beam assisted deposition technique has developed that takes place at 250 eV and leads to niobium coatings that are fully dense. Salagean et al. [71] also deposited such coatings using an arc and unbalanced magnetron-sputtering cathode equipped with an Nb target; base pressure 5 × 10⁻⁵, 10 kW and a 6A coil at 400 °C. It was found prior etching treatment defined the quality of the new surface as well as the voltage bias during deposition (Figs. 4.34, 4.35, and 4.36). Salagean et al. [71] found niobium whose melting point is 2,468 °C could be deposited on a substrate with a temperature of 400 °C. Donohue et al. [72] preferred TiAlN to TiN coatings due to their greater resistance of oxidation (Figs. 4.37 and 4.38). Donohue et al. [72] also found Ti _{l−x−y−z} Al _x Cr _y Y _z N deposited by magnetron sputtering offered even greater oxidation resistance due to the additional alloying elements. Substrates coated were M2 tool steel with Ra of 10 nm and cold-rolled 304 austenitic stainless steel. Prior to coating the samples were Cr ion etched at −1.2 kV for 20 min at a pressure of 6 × 10⁻⁴ mbar to reduce droplet formation. A 0.2 μm thick Ti _l−x−y Al _x Cr _y N base layer was then deposited, three separate coatings were then grown on top of this base layer to a thickness of 3 μm; they were Ti _{l−x−y−z} Al _x Cr _y Y _z N, Ti _l−x−y Al _x Cr _y N, and Ti _l−x Al _x N. In all cases the bias was −75 V. The exact compositions were Ti_0.43Al_0.52Cr_0.03Y_0.02N, Ti_0.44Al_0.53Cr_0.03N, and Ti_0.46Al_0.54N. The surface roughness for all films had Ra values between 0.038 and 0.04 μm. The hardness of the coatings was Ti_0.43Al_0.52Cr_0.03Y_0.02N = HK_0.025 = 2,700 kg mm², Ti_0.44Al_0.53Cr_0.03N = HK_0.025 = 2,400 kg mm², and Ti_0.46Al_0.54N = HK_0.025 = 2,400 kg mm². Donohue et al. [72] also investigated the oxidation temperature, for TiN it was 600 °C, for Ti_0.46Al_0.54N it was 870 °C, for Ti_0.44Al_0.53Cr_0.03N it was 920 °C, and for Ti_0.43Al_0.52Cr_0.03Y_0.02N it was 950 °C. He therefore concluded the extra alloying elements were significant. Annealing of the Ti_0.44Al_0.53Cr_0.03N at 950 °C for 1 h showed significant oxidation with micron sized voids however annealing of the same conditions of the Ti_0.43Al_0.52Cr_0.03Y_0.02N shows no effect, highlighting the importance of yttrium.

Fig. 4.32

Cross section showing growth defects in a TiAl coating [69]

Fig. 4.33

Cross section of a Cr etched TiAlN coating [69]

Fig. 4.34

Coatings deposited at −25 V bias [71]

Fig. 4.35

Coatings deposited at −75 V bias [71]

Fig. 4.36

Coatings deposited at −125 V bias [71]

Fig. 4.37

Thermo-gravimetric oxidation rate measurements in air during a linear temperature ramp at 1 °C min⁻¹ [72]

Fig. 4.38

Thermo-gravimetric oxidation rate measurements in air during an isothermal anneal at 900 °C [72]

4.5 Micromachining

Inamura et al. [73] found it difficult to apply finite element analysis to micromachining processes around 1 nm (Fig. 4.39); they found that a more successful approach is to use molecular dynamics simulations. Molecular dynamics works by establishing the position of each atom by resolving Newtonian dynamics for the system. Such simulations highlight the need for sharp tools because dull tools are predicted to generate a large shear area (some area either side of the defined shear plane) and this leads to significant work hardening of the material. It has been shown faster cutting speeds produce thinner chips and sharp tools operated at high cutting speeds impart low forces to the tool. However, Kim and Moon [74] found that when considering the cutting zone (Fig. 4.40), if the tool becomes blunt the forces generated by a blunt tool at high speed are significantly higher than forces produced by a blunt or sharp tool at low speed. Therefore, if the full benefits of microcutting are to be utilized according to predictions made by molecular dynamics simulations of chip flow (Fig. 4.41) the cutting tool must be sharp and cutting speeds must be high.

Fig. 4.39

Finite element mesh to simulate micromachining

Fig. 4.40

Molecular dynamics simulation of the cutting zone [74]

Fig. 4.41

(a) Cutting copper with a depth of cut of 1 nm at 1000 m/s, chip thickness is 6.69 nm; (b) cutting copper with a depth of cut of 1 nm at 1000 m/s, chip thickness is 7.69 nm; (c) cutting copper with a depth of cut of 1 nm at 1000 m/s, chip thickness is 9.09 nm; and (d) cutting copper with a depth of cut of 1 nm at 1000 m/s, chip thickness is 10.79 nm. Chip formation predicted by molecular dynamics simulation [74]

Burr formation is observed at the macroscale and can contribute up to 30 % of the time and cost it takes to produce a part. Burr formation is also observed at the microscale, but Gillespie [75] discovered macroscale burr removal techniques could not be applied because dimensional inaccuracies and residual stresses may be induced. Smaller burrs are more difficult to deal with; therefore investigations into burr formation and burr reduction have been undertaken. Gillespie and Blotter [76] stated there are three generally accepted burr formation mechanisms: lateral deformation, chip bending, and chip tearing.

Most microburr formation studies are at a disadvantage by operating well below the recommended cutting speed, for example, machining aluminum with a 50 μm tool demands a cutting speed of 105 m/min which would require a spindle speed of 670,000 rpm. At the macroscale, once formed, burrs can be quantified; for example, burr height and burr width can be measured. However, at the microscale such techniques are difficult to employ; if they are employed then Kim [77] found accuracy and repeatability are lost. Work by Ko and Dornfield [78] described a three-stage process for the burr, initiation, development, and formation. They developed a model for the burr formation mechanism that worked well for ductile materials such as Al and Cu. However, at this time burr formation is not well understood but correlations between machining parameters can identify key variables that help reduce burr size; for example, observations of burrs (Fig. 4.42) by Lee and Dornfeld [79] indicate that up-milling generally creates smaller burrs than down-milling and these observations along the length of a track led to the categorization of different burr types (Fig. 4.43).

Fig. 4.42

An exit burr [79]

Fig. 4.43

Categorization of burr types [79]

It cannot be assumed that if milling is scaled down to the microlevel, then machining characteristics will scale by the same amount. During macromachining the feed per tooth is larger than the cutting edge radius; however during micromachining the feed per tooth is equal to or less than the cutting edge radius. Conventional milling tools have a slenderness ratio that prevents bending whereas microtools are susceptible to bending because their diameters are only a few hundred microns. During macroscale cutting the leading tool edge encounters bulk material and therefore avoids contact with hard particles; in comparison, a microcutting edge encounters individual grains ensuring contact with hard particles.

The results of Ikawa et al. [80] suggest there is a critical minimum depth of cut, below which chips do not form. The analysis of Yuan et al. [81] indicates chip formation is not possible if the depth of cut is less than 20–40 % of the cutting edge radius. Micromachining at 80,000 rpm produces chips similar to those created by macroscale machining, where chip curl and helix effects are observed [82]. Kim also observes if the feed rate is too low a chip is not necessarily formed by each revolution of the tool. This anomaly can be demonstrated by calculation; if the feed rate is low enough, the volume of material removed is predicted to be greater than the volume of chips created. Thus some rotations that were assumed to create a chip could not have done so. This effect can also be demonstrated by experimental evidence. Sutherland and Babin [83] found that feed, or machining marks, are separated by a spacing equal to the maximum uncut chip thickness. Results show at small feeds per tooth the distance between feed marks is larger than the uncut chip thickness indicating no chip has been formed. Kim [82] concludes that a tool rotation without the formation of a chip is due to the combined effects between the ratio of cutting edge radius to feed per tooth and the lack of rigidity tool of the tool.

Work by Ikawa et al. [84] and Mizumoto et al. [85] showed single point diamond turning can machine surface roughnesses to a tolerance of 1 nm; the critical parameter was found to be repeatability of the depth of cut. It is useful to model the cutting process using a molecular dynamics simulation approach. Ikawa and Mizumoto [84] found that the behavior and flow of the material can be predicted by the laws governing interactions between neighboring atoms. Ikawa’s simulations [85] suggest the depth of cut and cutting edge radius are critical parameters that determine chip formation (Fig. 4.44). The velocities of these atoms can be calculated for each time step.

Fig. 4.44

Chip formation in relation to the depth of cut and cutting edge radius [80]. (a) Cutting edge radius: 0.5 nm, (b) cutting edge radius: 1.5 nm, and (c) cutting edge radius: 5.0 nm

$$ {v}_i^{\prime }={v}_i\sqrt{\frac{NkT}{{\displaystyle {\sum}_i\frac{m_i{v_i}^2}{2}}}} $$

(4.33)

where v _i is the atom velocity, N is the number of atoms, k is Boltzmann’s constant, T is temperature, and m is the atomic mass. By dividing the process into small intervals it is possible to compute the position of each atom, in this way material flow during the chip formation process can be predicted. Because a molecular dynamics simulation does not account for electron behavior temperature predictions are unreliable. Temperature gradients are overestimated; therefore an adjustment is required to predict heat flow more accurately. The adjustment process is conducted using the following steps, to convert the atom velocity to atom temperature T _i .

$$ {T}_i=\frac{m{v_i}^2}{2k} $$

(4.34)

The scaled temperature is given by

$$ {T}_i^{\prime }={T}^i-\left({T}_i-{T}_m\right)\beta $$

(4.35)

where T _m is the average temperature of the atoms and is given by

$$ {T}_m=\frac{1}{N}{\displaystyle {\sum}_i{T}_i} $$

(4.36)

Eventually forces on the tool tip and the cutting forces can be calculated. Shimada et al. [86] compared their predicted results with experimentally determined values for a copper workpiece and found them to be accurate. If the strain applied to the workpiece by the tool is large enough, forced lattice rearrangement will occur therefore generating dislocations. As the tool advances more dislocations form at the tool–chip interface and if enough join at the primary shear zone then a chip is formed. After the tool has finished cutting, dislocations that penetrated the workpiece migrate out towards the surface because the lattice can relax. This phenomenon can be observed as atomic sized steps on the surface, which represent the best surface roughness possible [87]. Shimada also showed machining accuracy can be defined by the minimum undeformed chip thickness, below a depth of 0.2 nm simulations have shown chips will not form for a tool edge radius of 5 nm, although this depth varies with tool size and material. In macroscale machining the tool edge sees bulk properties of the workpiece; however in micromachining the tool edge sees features of the material matrix such as grain boundaries. Shimada et al. [88] used molecular dynamic models to simulate this interaction. Simulations running cutting speeds at 2,000 m/s show the kinetic energy imparted to the workpiece is far greater than the cohesive energy of the workpiece. Other molecular dynamic simulations have identified four stages of the cutting process: (1) Compression of the work material ahead of the tool; (2) Chip formation; (3) Side flow; and (4) Subsurface deformation of the workpiece. Komanduri et al. [89] conducted simulations that help highlight differences between macro- and microscale cutting, for example, a volume change is observed when machining silicon. A pressure induced phase change modifies the structure from cubic to body centered tetragonal resulting in a 23 % denser chip.

Usually silicon is brittle but if more electrons are available in the conduction band (usually at high temperatures) it can behave more like a metal. It has also been reported that there are no dislocations found in silicon substrates, this is an attributed effect of cutting at high speed. Significant differences were found between the machining characteristics of aluminum and silicon; most of these were attributed to the variation in ductility between the two materials. Aluminum chips form due to plastic deformation along their preferred crystallographic planes. The mechanism for plastic deformation of silicon is similar when it is being machined or extruded; this is due to the phase transformation from cubic to body centered tetragonal. This phase change also causes surface and subsurface material below the chip to become denser.

The evidence for phase transformation is that silicon, which is usually brittle, acts in a more ductile manor; this is only made possible by a pressure induced phase change. Komanduri et al. [89] also suggested subsurface damage, although very small, is inherent in micromachining, however the subsurface damage was found to decrease as the aspect ratio of the tool decreases. The subsurface densification was shown to decrease with an increase in rake angle and increase with an increasing aspect ratio. Side flow was predicted to decrease with increasing width of cut and increase with an increasing depth of cut. Low cutting forces are produced at the microscale therefore smaller machine structures can achieve the damping and stiffness characteristics required for successful machining. The smaller footprint allows a greater number of machines per floor area yielding a greater throughput and making the technique suitable for mass production. Vogler et al. [90] have also examined the differences between macro- and microscale machining. Vogler et al. [90] observe the tool edge and workpiece material grains become comparable in size. The tool edge radius becomes a similar size to the uncut chip thickness; this results in large ploughing forces. The tool’s slenderness ratio reaches a point where tool stiffness is reduced. Vogler’s force prediction model accounted for different grains such as pearlite and ferrite by examining their individual machining characteristics and incorporating them into his model, machining characteristics of the bulk material can then be predicted. It was found cutting edge conditions had a large effect on machining forces. A worn tool edge can produce 300 % more cutting force and results in poor surface roughness and increased burr formation. Cutting conditions of 120,000 rpm, a feed per tooth of 4 μm and a depth of cut of 100 μm produced a peak cutting force of 3 N. Separation between feed marks was 0.004 mm and the associated waviness wavelength was 0.02 mm. A molecular dynamics model was constructed and it was found when machining aluminum with a diamond tool chips can only form if the minimum depth of cut is approximately half the cutting tool radius. If the feed per tooth is less than the tools edge radius, the maximum uncut chip thickness does not coincide with the feed per tooth, indicating several rotations without the production of a chip. The chip forms when the critical depth of cut is reached.

4.6 Research Directions

The study of micro- and nanomachining from a materials perspective has demonstrated a new outlook on the problems associated with machining with conventional cutting tools at these scales. The economies of scale can certainly be gained if a number of emerging problems can be solved. One dominant problem that occurs at the microscale is the bending of the cutting tool and the inability to cut chips at low speeds. Therefore, special attention must be applied to constructing stiffer tools and to prevent the rounding of the tool edge when the tool is initially straight. This may be overcome by coating small cutting tools with nanocrystalline diamond that has many cutting points, which are in contact with the workpiece even when tool bending takes place.

The incompatibility between diamond and ferrous materials can be overcome by coating the nanocrystalline diamond with a thin coating of a compound that has the least thermodynamic potential for dissolution. This direction will give rise to using multilayered coatings that have beneficial advantages such as thermally conducting heat away from the zone of cutting and reducing the generation of frictional heat by using a carbon-based soft layer lubricating coating. In terms of the construction of meso machine tools, future research directions include designing spindles that rotate the cutting tool at extremely high speeds that use thin layer diamond coatings on bearing surfaces. Air turbine spindles with integrated gas bearings are the possible solution to achieving extremely high spindle speeds. In addition to using small-scale machine tools, machine tools must be axisymmetric in construction so that mechanical and thermal disturbances are minimized.