Composition, properties and surface structure of tribochemically deposited coatings.

UPTEC Q11002

Examensarbete 30 hp

April 2011

Composition, properties and

surface structure of tribochemically

deposited coatings.

Teknisk- naturvetenskaplig fakultet UTH-enheten Besöksadress: Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0 Postadress: Box 536 751 21 Uppsala Telefon: 018 – 471 30 03 Telefax: 018 – 471 30 00 Hemsida: http://www.teknat.uu.se/student

Abstract

Composition, properties and surface structure of

tribochemically deposited coatings.

Eva-Brita Åkerlund

Five tribochemically deposited coatings on honed cast iron cylinder liner segments has been studied with respect to surface properties, material composition, coating thickness, hardness and friction. Methods like Light Optical Microscopy (LOM), Scanning Electron Microscopy (SEM), Energy Dispersive Spectroscopy (EDS), 3D topography using white light interferometry (VSI) and Electron Spectroscopy for Chemical Analysis (ESCA) were used to study the surface, coating thickness and material composition. Lubrication regimes (friction) were studied using a test set-up in a Lathe. An in-situ scratcher and nano indenter were used to study the hardness. It was found that the honing plateau surface is smoothened by the coating process while the honing scratches are kept more or less intact. The deposited coating thickness is approximately 10-100 nm. The coating is softer than the substrate and shows a butter-like behavior when scratched. Using only sulfur additive in the process fluid gives a smooth surface and an evenly distributed coating. Tungsten additive in the process fluid gives a thicker coating but a more irregular material distribution.

Tungsten additive in the process fluid does not seem to stimulate the formation of WS2, but rather WO3 is formed.

Sammans¨attning, egenskaper och ytstruktur hos

tribokemiskt deponerade bel¨aggningar

Eva-Brita ˚

Akerlund

Bakgrund och problemformulering

Friktionsf¨orluster f¨orekommer i de flesta tekniska system och bidrar till stora on¨odiga kostnader inom t.ex. industriell produktion och i transportsektorn. Ett typiskt ex-empel ¨ar en f¨orbr¨anningsmotor d¨ar stora delar av br¨anslef¨orlusterna (6-9 % av den tillf¨orda energin) sker genom friktion mellan kolv och cylinder. Genom att f¨or¨andra materialets ytegenskaper ¨ar det m¨ojligt att reducera b˚ade energif¨orluster och n¨otning och samtidigt f¨orl¨anga systemets livsl¨angd utan att materialets bulkegenskaper f¨or-¨

andras n¨amnv¨art.

F¨oretaget ANS (Applied Nano Surfaces) utvecklar en teknik f¨or att tribokemiskt1

bel¨agga st˚alytor med bland annat WS2. Metoden g˚ar ut p˚a att ett verktyg av volfram

gnids mot en st˚alyta i en svavelhaltig v¨atska. I idealfallet reagerar svavelatomerna i processv¨atskan med volframatomer fr˚an verktyget och bildar ett ytskikt av WS2 p˚a

st˚alytan.

Tidigare forskning har framf¨or allt fokuserat p˚a sambandet mellan processpara-metrar och erh˚allna tribologiska egenskaper. Kunskapen om bel¨aggningen och vad som verkligen sker i bel¨aggningsprocessen ¨ar fortfarande begr¨ansad och fr˚agor som ”Vad best˚ar bel¨aggningen av f¨or material?” ”Hur tjock ¨ar bel¨aggningen?” ”Vad har bel¨aggningen f¨or h˚ardhet?” ¨ar obesvarade. En stor del av problematiken ¨ar att det tillverkade skiktet uppskattas vara v¨aldigt tunt och dess egenskaper ¨ar sv˚ara att se-parera fr˚an bulkmaterialet. M˚alet med detta arbete ¨ar att f¨ordjupa kunskapen om bel¨aggningen och ¨oka f¨orst˚aelsen f¨or bel¨aggningsprocessen. Sex utvalda bel¨aggningar analyserades med avseende p˚a utseende, material, tjocklek, h˚ardhet och friktionsbete-ende. Arbetet utf¨ordes som en del av ett st¨orre projekt, Wonaco II, ett samarbetspro-jekt mellan ANS, H¨ogskolan i Halmstad, Lule˚a universitet, Scania, Volvo Powertrain och Uppsala universitet.

Analysmetoder

F¨or att studera ytan p˚a bel¨aggningen anv¨andes olika typer av mikroskopi. F¨or en f¨orsta ¨overblick anv¨andes ett ljusoptiskt mikroskop. Ytorna studerades sedan med avseende p˚a topografi och materialsammans¨attning vid h¨ogre f¨orstoring i ett svepe-lektronmikroskop (SEM). F¨or att f˚a ett m˚att p˚a ytj¨amnheten anv¨andes vitljusinter-ferometri, en metod f¨or att m¨ata 3D-topografi.

Materialets sammans¨attning analyserades med tv˚a olika metoder, EDS – r¨ontgen-spektroskopi och ESCA – fotoelektronr¨ontgen-spektroskopi. B˚ada metoderna ger ett spektrum som ¨ar specifikt f¨or varje element i det unders¨okta materialet. Det erh˚allna spektrumet kan ¨aven anv¨andas f¨or att g¨ora en kvantitativ analys. I ESCA kan det ¨aven vara m¨ojligt att se vilka kemiska bindningar som f¨orekommer.

En uppskattning av bel¨aggningarnas tjocklek kan f˚as genom att kombinera infor-mationen fr˚an SEM och EDS med simuleringar av hur elektronerna r¨or sig i provet.

H˚ardhetsm¨atningar utf¨ordes dels f¨or att unders¨oka bel¨aggningens h˚ardhet och dels f¨or att se om substratets egenskaper f¨or¨andrats av bel¨aggningsprocessen. H˚ ardhets-m¨atningarna utf¨ordes med en nanoindenter. En nanoindenter m¨ater den last som kr¨avs f¨or att trycka en trekantig pyramidspets till ett visst djup och ber¨aknar sen h˚ardheten utifr˚an m¨atningarna.

1

Bel¨aggningens kapacitet att ge s¨ankt friktion vid olika hastigheter (sm¨orj-regimer) unders¨oktes i en testuppst¨allning i en svarv med varierande rotationshastighet och j¨amf¨ordes med en obelagd yta.

Analyserade bel¨aggningar

Sex olika bel¨aggningar valdes ut f¨or analys. Fem av bel¨aggningarna ¨ar belagda p˚a henade segment fr˚an insidan av ett cylinderfoder. Hening ¨ar en ytprepareringsmetod som ger j¨amna plat˚aer mellan djupa repor i ett rombformat rutm¨onster, se figur. En av bel¨aggningarna ¨ar belagd utanp˚a ett cylinderfoder. Bel¨aggningarna ¨ar tillverkade med fyra olika processv¨atskor med olika m¨angder svavel- och volfram-additiv. En av processv¨atskorna anv¨andes b˚ade till bel¨aggningen p˚a utsidan av cylinderfodret samt p˚a tv˚a av segmenten med varierande kontakttryck i bel¨aggningsprocessen.

Insidan av ett cylinderfoder och ett belagt cylinderfodersegment.

Resultat

Bel¨aggningsprocessen verkar ge en helt¨ackande bel¨aggning p˚a plat˚aerna utan att bel¨agga eller fylla i de grova heningsreporna. Processen har en utj¨amnande effekt och ytj¨amnheten hos den belagda ytan (egentligen avvikelsen fr˚an en medelyta) ¨ar i samma storleksordning som den uppskattade skikttjockleken (10-100 nm). Skiktet uppskattas vara mjukare ¨an substratet och l˚agfriktionsegenskaperna f¨orb¨attras vid bel¨aggning j¨amf¨ort med motsvarande obelagd yta. Den bel¨aggning som ¨ar tillverkad med enbart svavel i processv¨atskan ger st¨orre andel svavel i bel¨aggningen och ett j¨amnare men tunnare skikt. ¨Ovriga bel¨aggningar, de som ¨ar tillverkade med varie-rande m¨angd volframadditiv i processv¨atskan, ger tjockare men oj¨amnare skikt och verkar snarare best˚a av volframoxid och en mindre m¨angd svavel. P˚a grund av att bel¨aggningarna ¨ar s˚a tunna ¨ar det sv˚art att hitta n˚agon analysmetod som m¨ater en-bart skiktets egenskaper. Oavsett metod har substratet en viss p˚averkan som ¨ar sv˚ar att bortse ifr˚an. Det g¨or att de flesta m¨atv¨ardena inte g˚ar att anv¨anda som absoluta v¨arden. D¨aremot ¨ar analyserna utf¨orda under samma f¨orh˚allanden s˚a m¨atningarna ¨ar fortfarande inb¨ordes j¨amf¨orbara.

Examensarbete 30 hp p˚a civilingenj¨orsprogrammet Teknisk fysik med materialvetenskap

Contents

2.2 The use of tungsten disulphide as lubricant . . . 6

2.3 ANS coatings . . . 7

2.3.1 Adjacent research and similar methods . . . 7

2.4 Honing . . . 8

2.5 Optical surface profiling . . . 8

2.6 Scanning electron microscopy (SEM) . . . 9

2.6.1 Information depth . . . 9

2.6.2 Atomic number contrast . . . 10

2.6.3 Energy dispersive X-ray spectroscopy (EDS) . . . 11

2.7 Electron spectroscopy for chemical analysis (ESCA) . . . 11

2.8 Nano hardness . . . 11

3 Experimental 12 3.1 Studied coatings . . . 12

3.2 Applied analytical methods . . . 14

3.2.1 Light optical microscopy (LOM) . . . 14

3.2.2 Vertical scanning interferometry (VSI) . . . 14

3.2.3 Scanning electron microscopy (SEM) . . . 14

3.2.4 Energy dispersive X-ray spectroscopy (EDS) . . . 14

3.2.5 Coating thickness from EDS . . . 14

3.2.6 Thickness imaging with BSE . . . 15

3.2.7 Electron spectroscopy for chemical analysis (ESCA) . . . 17

3.3 Hardness characterization . . . 17

3.3.1 In-situ scratching . . . 17

3.3.2 Nano indentation . . . 17

3.4 Lubrication regime study . . . 19

4 Results 20 4.1 Surface overview . . . 20 4.2 Coating composition . . . 29 4.3 Coating thickness . . . 37 4.4 Hardness . . . 37 4.5 Lubrication regimes . . . 42 5 Discussion 44 5.1 Surface overwiew . . . 44 5.2 Coating composition . . . 44 5.3 Coating thickness . . . 45 5.4 Hardness . . . 46 5.5 Lubrication regimes . . . 47 6 Conclusions 48 A Appendix 1 A.1 ESCA spectra . . . 1

A.2 EDS spectra . . . 4

A.3 Hardness . . . 5

1

Introduction

Energy and material losses due to friction are common problems in many technical systems. The science of friction, wear and lubrication is called tribology and is often described as the science of two surfaces in contact moving relative each other. The tribological properties of a system can often be changed by small adjustments of the surface material and structure. It is therefore possible to reduce the energy loss, ma-terial wear and system life time to a large extend without actually affecting the bulk properties. A typical example of a tribological system is the cylinder liner – piston ring contact (fig. 1) in an internal combustion engine. Here, a significant part of the frictional losses of the engine takes place (approximately 6-9% of the total energy supplied) [1]. With the new EU climate goals for CO2 emission, the importance of

improved fuel economy grows, and with it, the need for simple and cheap surface modifying/coating processes.

(a) (b)

Figure 1: (a) Schematic picture of a cylinder liner and piston in an engine. The actual contact is between the cylinder liner and thin rings (piston rings) mounted on the piston as shown in (b) [2].

ANS (Applied Nano Surfaces) has developed a tribo-chemical coating process where a nano-composite of among others WS2 is formed. The principal idea of the process

is that a tungsten tool is rubbed against an iron or steel substrate in the presence of a sulfur-containing process fluid. In the tribological contact between the tool and the substrate a chemical reaction is triggered and in the ideal case a WS2coating will be

of the process does not exist yet. Earlier studies in the field [3, 4, 5] have mainly focused on the correlation between used process parameters and obtained tribological properties. Factors like material composition, material properties and microstructure of the coatings are less investigated.

The goals with this master thesis work are to • investigate the coating properties

• increase the understanding of the deposition process

• connect process parameters and tribological properties with the coating prop-erties

The thesis work was performed as a part of a larger project, Wonaco II, a ProViking financed project between ANS, H¨ogskolan i Halmstad, Lule˚a University, Scania, Volvo Powertrain and Uppsala University.

Six coatings made with varying process parameters were chosen for the investiga-tion. All coatings were made by ANS on cylinder liners from Scania trucks. The coatings were studied with respect to surface properties, material composition, coat-ing thickness, hardness and friction.

2

Theory

2.1

Tribology

Tribology is the science of friction, wear and lubrication, or more specific, the sci-ence of two surfaces in contact moving relative each other. The interface between these surfaces is an extreme environment, much due to the surface topography. No surface is completely flat when magnified enough and the real contact area is much smaller than the geometric contact area. This leads to local pressures, stresses and temperatures that are high enough to deform and modify the surfaces. Elastic and plastic deformation, formation of cracks and phase transitions are typically occurring phenomena. The chemical reactivity is also affected and spontaneous chemical reac-tions can occur at the surface that normally wouldn´t occur in the system, i.e. the sliding contact has a ”catalytic effect”. This type of reaction is called a tribochemical reaction.

2.1.1 Surface contact

Two flat surfaces are in the ideal case completely in contact with each other, and the contact area is the same as the geometric area of the surfaces. Putting a load over the interface give rice to a contact pressure which depends on the applied force (FN)

and the area carrying the load, in this case the geometric area (AN). But no surface

is completely flat and the real contact area is limited to the small contact spots where peaks of the two surfaces are in contact. This area is much smaller than the geometric area, as shown in fig. 2. fi is the load carried by the surface of each contact spot,

ai. The friction and wear of a system is dependent on size, amount and properties of

these contact spots, and hence, understanding the actual surface contact situation is crucial for any tribological system.

Figure 2: The applied force FN over the geometric area AN is divided into several contact

spots ai each with the applied force fi. The real area is the sum of the contact spot area [6,

p. 17, Fig. 2.9].

Even a relatively small load can cause a high contact pressure if there is only a few small contact spots. This leads to elastic and plastic deformation of the surface peaks until the contact spots has grown enough to balance the load, i.e. when the contact pressure has decreased enough to no longer be able to plasticize the surface material. If the surfaces consists of different materials the softer material will be deformed. Increasing the load will not increase the contact pressure once the plasticizing limit is reached, it only increases the real contact area by deforming the material. The contact pressure at the interface is therefore limited by the hardness of the softest material involved and stays constant, independent of the applied load [6, p. 12-22].

The required force to induce a sliding movement and the resistance against sliding are determined by the friction of the system. Again the size, amount and properties of the contact spots are of great importance. By applying a large enough tangential force on the interface, the surfaces start moving relative each other. The surface peaks in contact are exposed to large shear- and deformation forces which finally separates the peaks. As the surfaces moves, ”new” peaks will be brought into contact leading to a continuous change of contact spots. Typically, each contact spot exists 10−6_{− 10}−4 _{s. The work that is used to overcome the force of friction is converted}

to heat (95-98 % for metals) or consumed creating wear fragments, cracks and other defects. The intense heat generation occurring at the contact spot interface causes a quick temperature change called flash temperature. The flash temperature is always close to the melting point of the material. The temperature distribution through the material i shown in figure 3 [6, p. 35-38].

Figure 3: The heat is generated at the peak interfaces and give rice to a flash temperature. The heat is transported away from the interface resulting in an increased average surface temperature that differ from the material base temperature [6, p. 35, Fig. 2.32]

.

2.1.2 Lubrication

Friction between two dry surfaces gives a relatively high friction and is unwanted in many applications. A small amount of lubricating material added at the interface can modify the surface contact properties to a large extend. The most common lubricants are different types of oils, but other liquids, fats, solid- and gaseous materials can also be used. A lubricant can also reduce wear, have a cooling effect, protect from corrosion and give electric isolation. There are three types of lubrication regimes based on two different mechanisms.

Full film lubrication is when the pressure in the lubricant is high enough to sep-arate the two surfaces completely from each other. The applied load is carried and transferred completely by the lubricant and the friction of the system is limited by the viscosity of the lubricant, i.e. the force required to shear the lubrication film.

Boundary lubrication occurs when the pressure in the lubricant is to low to separate the surfaces completely. Thin solid films form at the contact interface, often build up by additives in the lubricant. This prevents direct dry contact and provides a relatively low friction due to the reduction of strong bindings across the interface. The friction is determined by the sliding resistance between the formed films and is independent of the lubricant viscosity. The films can be anything from nanometers to micrometers thick.

Mixed lubrication is the regime where the lubrication mechanism gradually changes from full film lubrication to boundary lubrication. The friction of the system is dependent on both the viscosity of the lubricant and the formation of thin films.

To illustrate the relations between the different lubrication regimes a Stribeck curve can be used. A Stribeck curve shows the friction coefficient µ as a function of (η·ω_P ) where η is the viscosity, ω the sliding velocity and P the applied load. A typical Stribeck curve is shown in figure 4. By changing viscosity, velocity and load the system can be adjusted to wanted lubrication regime. A small load combined with high viscosity and speed is needed to obtain full film lubrication which is wanted in many applications. For some systems it is impossible to obtain full film lubrication and studies and development of boundary lubricated systems are of great importance [6, Chap. 3].

Figure 4: A schematic Stribeck curve. 1 – the area of boundary lubrication. 2 – the area of mixed lubrication. 3 – the area of full film lubrication [7]

.

2.1.3 Tribochemical mechanisms

There are several different mechanisms in a tribological contact that can affect the chemical reactivity. Friction heat increases the temperature locally (flash tempera-ture) and can activate or increase the speed of chemical reactions that normally dont occur in the system. Friction heat is the most important tribochemical factor. An extreme mechanical strain can modify the binding energy between the different atoms and ease breaking the bonds, which stimulates other chemical reactions. Deformation of the surface can expose the underlaying material which often is more reactive than e.g. a covering oxide. Deformation can also induce defects and dislocations in the material increasing the chemical reactivity by increasing the possible diffusion roads for the elements [6, p. 82-83].

2.2

The use of tungsten disulphide as lubricant

W S2 has a lamellae structure, just as graphite and MoS2. It consists of covalent

bonded S-W-S layers held together with van der Waals bonds, as shown in figure 5. Putting a load on WS2 will break the weak van der Waals bonds between the layers

while the covalent bonds within the layers stay intact. This means that moving the layers relative each other is easy and leads to a low shear resistance and a low friction coefficient. WS2 and several other lamellae materials therefore have potential to be

used as a solid lubricant, but it should be pointed out that the lamellae structure itself is not an enough condition for low friction. Graphite is dependent on a moist atmosphere to obtain low friction. WS2 and MoS2 on the other hand, work best

in vacuum or in a dry atmosphere. All three materials have a limited temperature range due to oxidation at high temperatures, but WS2 tolerates higher temperatures

than MoS2and graphite, which makes it an interesting material for high temperature

applications with a dry atmosphere [6, p. 69].

Surface and friction characterization of WS2 as a third body thin film have shown

that WS2 oxidize to WO3 under tribochemical contact in a rolling-sliding Amsler

test. Elemental S and SO−2

4 are also obtained. It is indicated that the decomposition

temperature is lower under sliding conditions than for thermal decomposition. How-ever, as long as the system is kept in a stable friction region the oxidation only has a small influence on the lubricity and the friction increases rapidly only when there is insufficient lubricating material [8].

According to [9] it was pointed out that, in addition to thermal and laser annealing, mechanical rubbing on amorphous WS2 causes the film to crystallize.

Figure 5: Two layers of WS2, the yellow spheres symbolize the sulfur atoms and the turquoise

spheres the tungsten atoms. The atoms are bound with covalent bonds within the layer and van der Waals bonds between the layers. The original image [10] describes MoS2 which has

2.3

ANS coatings

The ANS coating process is based on a tribochemical reaction between a tungsten tool and a sulfur containing process fluid. The tungsten tool is pressed in a sliding motion against a substrate in the presence of a process fluid (se figure 6). This triggers a tribochemical reaction1

and in the ideal case, a film of WS2is formed. The tool has a

smoothening effect due to the applied pressure resulting in some surface deformation.

Figure 6: Schematic setup of the ANS coating process.

2.3.1 Adjacent research and similar methods

The tribochemical reactions used in the ANS coating process were first detected in studies of the sliding contact between a W-DLC coating and steel with a sulfur con-taining EP-additive. It was then suggested that the formed tribo-film consisted of WS2 and FeS and was build up and worn off continuously until the W-DLC coating

(W-source) was completely worn away [3, p. 29]. Other routes on forming WS2 in

a tribochemical reaction by varying tool material as well as EP-additive were also studied.

Burnishing is a surface finish method where the surface is plastically deformed using a roller or a hard ball [11]. The process smoothens the surface and ads a residual compressive stress, improving properties like fatigue strength, hardness, corrosion resistance and wear resistance at a controlled depth of compression [12]. Burnishing can also be used for grain size refinement by cold work[13].

Running in the engine is the mild wear in the initial state of the engine life time. Running in can e.g. smooth surface irregularities and reduce localized pressure be-tween the liner and rings. Running in occurs mainly due to two different wear mecha-nisms, abrasion from small particles originating from inlet air or combustion products and plastic deformation due to the surface irregularities, especially in areas of high localized pressures with insufficient lubrication [14].

1

2.4

Honing

Honing is the surface structuring method traditionally used in preparation of the inner cylinder liner surface. Honing is performed in two steps, coarse honing and plateau honing. A honed surface consists of smooth load-bearing and wear-resistant plateau surfaces separated by deeper honing scratches which work as lubricant oil reservoirs and debris traps [15]. The honing structuring results in a rhombic pattern. A SEM image of a honed surface is shown in figure 7.

Figure 7: SEM image of a honed surface. The topography is enhanced by a 60◦_tilt.

2.5

Optical surface profiling

The topography of a surface can be described in many ways. Common are different types of surface parameters. One parameter is Ra, which is the arithmetic average

deviation from an average level calculated by Ra= Σ |z(x, y)|

n

where z(x, y) is the deviation from the average level at point (x,y) and n is the number of points.

Optical surface profiling is a way for measuring surface parameters and topogra-phy. Here Vertical Scanning Interferometry (VSI) is described. VSI uses white light interference to create an image of a surface. The microscope is similar to a normal optical microscope but contains an interferometer objective. The interferometer splits the beam of light into two separate beams, one is reflected on the sample and one on a reference mirror. The objective is moved along the optical axis causing a phase shift between the beams. When the beams are recombined, the phase shift results in an interference pattern with constructive interference only when the optical paths of the both beams are equal. The registered intensity maxima can then be used to create a topography map of the surface. The vertical resolution is approximately 10 ˚A and the lateral resolution is 0.2 µm at the best depending in objective used. All points on the surface can be registered at the same time, which makes VSI a very fast method [16, Section Optiska ytprofilm¨atare].

2.6

Scanning electron microscopy (SEM)

In SEM, an electron beam is focused and accelerated towards the sample surface with the acceleration voltage E0. The interaction between the electron beam and the

sample, both elastic and inelastic scattering processes, is used to create an image. The different scattering processes result in several types of signals that can be detected individually (see fig. 8a). In SEM, secondary electrons (SE) and backscatter electrons (BSE) are used for surface imaging. An electron beam is scanned across the sample and the number of outgoing electrons is detected for each spot on the surface. In this way a point-to-point image is build up. The magnification is determined by the selected scanning area and the resolution by the diameter of the electron beam (the smallest possible spot). For compositional analysis characteristic X-rays can be detected.

2.6.1 Information depth

The volume of the sample where scattering occurs is called the interaction volume and is shown in figure 8b, together with the relative information depth from each type of signal. All scattering processes occur in almost the entire volume, but the signals can only ”escape” out of the sample from a certain depth. The size of the interaction volume depends on the electron energy, sample material, beam diameter and sample tilt. The depth of the interaction volume can be approximated using the Kanaya-Okayama range, RKO= 0.0276 · A · E 1_.67 0 Z0_.889 · ρ µm [17, p. 72] (1)

where A is the relative atomic mass in g/mol, ρ the density in g/cm3

and Z the atomic number of the target. The depth of the interactive volume can also be interpreted as the radius of a semicircle centered on the beam impact point which defines the envelope of trajectories in a Monte Carlo simulation [17, p. 73]. The information depth of BSE can be approximated with 0.3 times the Kanaya-Okayama range [17, p. 87].

(a) Electron beam - sample interaction (b) Interaction volume

Figure 8: (a) Schematic picture of electron beam - sample interaction. Typically an SEM uses secondary electrons and backscattered electrons for imaging and X-rays for compositional analysis (b) Interaction volume - showing the information depth for the different types of signals [18]

2.6.2 Atomic number contrast

Atomic number contrast occurs because the electrons are differently scattered by different materials. Atoms with a high atomic number Z (heavy nucleus) has a higher tendency for elastic scattering of the incoming electrons, creating a higher number of BSE. This can be studied with a BSE detector in composition mode, with areas of heavy elements looking brighter and areas of lighter elements looking darker [19]. Backscattering is quantified by the backscatter coefficient η, defined as

η =nBSE nB

where nBSE is the number of BSE and nB is the number of incident beam electrons.

The backscatter coefficient varies as mentioned with the atomic number Z, but also with the acceleration voltage E0. The variation of η with Z can be described with

η = −0.0254 + 0.016 · Z − 1.86 · 10−4_{· Z}2

+ 8.3 · 10−7_{· Z}3

[20, p. 77] (2) A plot of equation (2) is shown in figure 9. For energies lower than 4 keV the backscat-ter coefficient turns non-linear which affects the atomic number contrast at a large extend [21]. For energies E0=4-40 keV, the backscatter coefficient η(E0, Z) can be

described with η(E0, Z) = E0m· C [20, p. 78] (3) where m_{= 0.1382 −} 0.9211√ Z  and C_{= 0.1904 − 0.2235(ln Z) + 0.1292(ln Z)}2_{− 0.01491(ln Z)}3

Plots of equation (3) with E0=4 keV and E0=40 keV are also shown in figure 9.

Figure 9: The backscatter coefficient as a function of atomic number Z from equation (2), which gives rice to the atomic number contrast for BSE imaging. The backscatter coefficients for E0=4 keV and E0=40 keV from equation (3) show the effect when varying acceleration

2.6.3 Energy dispersive X-ray spectroscopy (EDS)

EDS is used for elemental analysis. When the incoming beam electrons interact with the material, electrons of an inner electron shell are excited. Characteristic X-ray photons are generated when the excited electron de-excites by filling the inner vacant position with a valence electron from an outer shell. EDS uses a semiconductor detector where the generated X-rays produce electron-hole pairs, a small electric pulse proportional to the X-ray energy. A spectrum of counted pulses versus X-ray energies from the beam-sample interaction can then be generated. Certain spectral artifacts like Si escape peaks, weak Fe and Cu peaks from the chamber and ”sum peaks” by pulse pile up may occur, giving fake peaks in the spectrum. The peaks from the characteristic X-rays can be used for both qualitative and quantitative analysis of the sample composition. The peaks correspond to the electron binding energy (EB) and

can bee seen as a fingerprint for each element. The intensity for each peak (height of the peak) is proportional to the amount of that specific element. The largest probability for generating an inner shell electron vacancy is when

Ee= 4 · EB (4)

where Eeis the incoming electron energy. So the peak height also varies with Ee[19].

EDS can be used to create element maps of a surface by collecting data only for a set energy window. [17, 19]

2.7

Electron spectroscopy for chemical analysis (ESCA)

Electron Spectroscopy for Chemical Analysis – ESCA is also known as XPS – X-ray Photoelectron Spectroscopy. ESCA is based on the photo-electric effect, using X-rays as the irradiating beam. The number of photo-emmitted electrons for a specific energy is detected giving a spectrum with peaks corresponding to occupied electron orbitals. ESCA can be used for both qualitative and quantitative analysis. ESCA is a very surface sensitive method, the information depth is determined by the inelastic mean free path of the photo-electron, which is typically 0.4-4 nm (2-10 monolayers). Ultra High Vacuum (UHV) is needed to avoid analyzing only the adsorbed molecules on the surface. ESCA is often combined with an ion gun for sputtering, both to remove the adsorbed surface molecules as well as for depth profiling with alternating analysis and sputtering. [22, Chap. 4]

The photo-electric effect in ESCA can be described by the equation EK = hυ − EB− φ

where EK is the measured kinetic energy of the photo-electron, hυ the energy of the

exciting X-ray, EB the binding energy of the electron in the solid and φ the work

function [22, p. 112]. The process is considered to be elastic and by controlling hυ and φ and by accurate measurement of EK the electron binding energies of the

atoms present in the solid can be calculated very precisely. The binding energy is affected by e.g. oxidation state, molecular environment and lattice site, which means that a chemical shift can be observed. The relative peak intensity can be used for quantitative analysis. [22, Chap. 3]

2.8

Nano hardness

Nano hardness can be measured using a nano indenter. Normally a Berkovich diamond pyramid with triangular basis is used as indenter and the hardness can be estimated by measuring applied load as a function of penetration depth during the indentation. The information depth at indentation is approximately 10×(indentation depth). [6]

3

Experimental

Six coatings with varying process parameters have been studied with respect to surface properties, material composition, coating thickness, hardness and lubrication regimes (friction). All coatings were made with the ANS coating method by ANS.

3.1

Studied coatings

The studied coatings and corresponding process parameters are listed in table 1. All coatings were made on cast iron substrates i.e. parts of a cylinder liner from Scania trucks. One of the coatings was deposited on the outside of a cylinder liner (fig. 10b) to obtain a cylindrical symmetry for friction studies in a Lathe. The other five coatings were made on the inside of honed cylinder liner segments, as shown in figure 10a. Overview images of the coated segments are shown in figure 11. Identical process parameters was used for the segments but with varying process fluid. For one of the coating 20% less contact pressure was used.

Table 1: The studied coatings and corresponding additive content of the used process fluid. Coating Process fluid content

ANS-S Sulfur additive

ANS-WS Sulfur and tungsten additive

ANS-4WS Sulfur and 4 times more tungsten additive than in ANS-WS ANS-4WS:80% Sulfur and 4 times more tungsten additive than in ANS-WS,

80% of the contact pressure

ANS-8W No sulfur and 8 times more tungsten additive than in ANS-WS ANS-4WS:cyl Sulfur and 4 times more tungsten additive than in ANS-WS,

(on the outside of cylinder liner)

(a) (b)

Figure 10: (a) A honed cylinder liner and a coated cylinder liner segment. (b) The coated area (darker) on the outside of a cylinder liner.

(a) ANS-4WS:80%

(b) ANS-4WS

(c) ANS-WS

(d) ANS-S

(e) ANS-8W

3.2

Applied analytical methods

3.2.1 Light optical microscopy (LOM)

An optical microscope was used to get a first overview of the appearance and structure of the coatings.

3.2.2 Vertical scanning interferometry (VSI)

Optical surface profiling was made with Vertical Scanning Interferometry - VSI. A WYKO NT 1100 optical profiling system with the 10x objective and 2x FOV was used. The software WYKO Vision 32 was used to create images of the surfaces and to calculate the Ravalues. Five coated surfaces and an uncoated surface were studied.

Measurements were performed on the honing plateaus at two different spots for each coating.

3.2.3 Scanning electron microscopy (SEM)

A Zeiss LEO440 scanning electron microscope was used to study the surfaces and to obtain an approximation of film thicknesses and compositions. The SEM is equipped with a LaB6 electron gun and an EDAX EDS system. For EDS the software EDAX

Genesis was used. All surfaces were studied with both SE and BSE (COMPO mode) detectors at three different E0(20, 10 and 3 keV). Correlating simulations were made

for complementary information about the information depth. Some of the samples were also studied tilted 60◦ _{for a better topographical view.}

3.2.4 Energy dispersive X-ray spectroscopy (EDS)

EDS was used for qualitative and quantitative analysis of the samples. EDS-mapping was also performed for each detected element, creating an image of the element distri-bution across the surface. In addition, EDS spot measurements were also performed on the honing plateaus.

Due to a software bug, adjusting brightness and contrast at the microscope some-times affected the obtained count-rate and dead time. Therefore, absolute values of Amp time varies when optimizied to obtain a high count-rate at tolerated dead time for each test2

. For qualitative/quantitative analysis a dead time of up to 40% was tolerated and data was collected for approximately 10 minutes. For the element mapping a dead time up to 60% was tolerated and data was collected for approxi-mately 30 min. The element mapping was performed of the detected elements from the qualitative analysis.

The information depths at the used acceleration voltages are much larger than the thickness of the studied coatings, as shown in figure 12. This means that a large part of the detected signal originates from the Fe substrate. An acceleration voltage E0

of 10 keV was used as a compromise between wanting to reduce the Fe background, i.e. decreasing the information depth, and the energy E0 needed for X-ray excitation

calculated from equation (4).

Quantitative analysis was performed using the obtained spectra from the qualita-tive analysis with a build-in function of the Genesis software.

3.2.5 Coating thickness from EDS

A Monte Carlo electron flight simulator was used to obtain an approximation of film thickness. The simulations were made at E0=10 keV and different thicknesses of

W S2 and W O3 on a Fe-substrate were iteratively tested. The simulated spectrum

was compared with the obtained EDS spectrum until the simulation gave a similar

2

(a) 20 keV (b) 10 keV

Figure 12: A Monte Carlo simulation of the interaction volume at 20 keV and 10 keV for a 40 nm WS2 layer on a Fe substrate. The images show electron trajectories in cross section.

The scale is different in the two images, the upper gray areas are the equally thick WS2

layers. The interaction volume in (a) is approximately 30 times the layer thickness and in (b) approximately 10 times. (The interaction volume can be interpreted as the radius of a semicircle centered on the beam impact point which defines the envelope trajectories of the simulation)

spectrum as the real spectrum. Varying film thickness varies the element proportions and hence the relative peak height. The film thickness of the simulated spectrum therefore corresponds to an approximate film thickness of the sample.

3.2.6 Thickness imaging with BSE

To study the film thickness variation, SEM-BSE images at several acceleration volt-ages were used. The basic id´ea is to use the atomic number contrast of each image and construct a thickness image of the surface layer. By varying the acceleration voltage, the information depth is also varied. Images of the same area were made with E0 = 4–20 keV in steps of 2 keV. The images were made on an area marked in

the nano indenter (see section 3.3.2).

To create a thickness image from BSE images at different E0, two parameters were

needed, the BSE information depth and the grayscale intensity corresponding to a specified amount of coating. To get an estimation of these parameters, some approx-imations were made. Following steps were used:

• Monte Carlo simulations were made for each of the elements of interest (W, S, Fe, O, C) at three different E0 (20, 10 and 3 keV).

• Corresponding calculations of the Kanaya-Okayama range from equation (1) were made for the single elements at three different E0(20, 10 and 3 keV).

• The single element simulations were compared to the Kanaya-Okayama range to get a guideline of how the interaction volume depth in any simulation should be defined.

• Monte Carlo simulations were made for both W O3 and W S2 on Fe-substrate

with obtained approximations of sample film thickness from 3.2.5. This was to study the interaction volume depth variations with respect to film thickness and material. An average was calculated and the parameters closest to the average were chosen for the final simulations.

• The final simulations were performed for 40 nm WS2on a Fe-substrate (see fig.

47). 10 000 trajectories were used to obtain an enough statistical significance3

. • The BSE information depth was estimated as 0.3 of the Kanaya-Okayama range or i.e. 0.3 of the estimated interaction volume depth from the simulations (see 2.6.1).

• For each element and used E0 the backscatter coefficient η was calculated using

equation (3). The calculated BSE coefficients are shown in figure 13.

• The BSE images were normalized and the BSE coefficient values for tungsten and carbon were assumed to correspond to grayscale max and min values re-spectively.

• The intensity increase was estimated be linear and an intensity threshold value was set to separate areas containing >15% coating, i.e. 15% of the intensity difference between the values of iron and tungsten.

• The grayscale threshold value was combined with corresponding BSE informa-tion depth for each E0 to create a thickness map of the coating.

The thickness images were made using MatLab. The approximated BSE information depths are attached in appendix A.4. The MatLab script is attached in appendix A.5.

Figure 13: Calculated BSE coefficients for Fe, W, O, S and C as a function of E0, equation

3.2.7 Electron spectroscopy for chemical analysis (ESCA)

Qualitative analysis was performed to examine the presence of elements and chemistry of the surfaces. Quantitative analysis was performed for comparison with the EDS results.

The analysis was performed with an ESCA (PHI Quantum 2000). A coarse energy sweep, in steps of 1 eV, was first made to obtain an overview spectrum and to identify the elements to be analyzed. More accurate energy sweeps, in steps of 0.4 or 0.5 eV, were then made for each of the detected elements. The analysis was performed after 1 min of sputtering. The sputtered area was approximately 2×2 mm and the analysis spot had a diameter of approximately 150 µm. It is not possible to determine the exact position of the analysis spot, but it can be assumed that it covers both honing plateaus and scratches and represents an average over the surface. Two areas on each sample were analyzed. The obtained spectra were compared with elemental reference spectra from [23].

3.3

Hardness characterization

3.3.1 In-situ scratching

A micro-manipulator was mounted in the SEM and used to scratch the coated surfaces with a probe to study their behavior. SE imaging was used both during the scratching as well as afterwards.

3.3.2 Nano indentation

A CSM Ultra Nano Hardness Tester (UNHT 01-03121) was used for nano indentation reference measurements and surface mapping. A MTS Nano Indenter XP was usedR

for cross section measurements. A Berkovich diamond pyramid was used.

Substrate hardness was measured on a polished cross section with indentation depth ∼ 1µm.

If the hardness variation over the surface is due to thickness variation of the coating, a surface mapping of the hardness should correlate to the thickness variation of the BSE image. A honing plateau area was marked with large indentations in the corners. This was to be able to find the same area for BSE-mapping in SEM. After the BSE imaging described in section 3.2.6, a matrix setup of 15 × 11 indentations with 3 µm distance and position marks (four large indentations) at each matrix corner was made. A photo of the setup matrix is shown in figure 14. The indentation mode Progressive Multi Cycle was used with 9 loading-cycles. The indentation depth was set to span from 10 nm to 60 nm with unloading to 5 nm between each loading-cycle. The indentation depth was chosen to result in similar information depths as for the BSE images. Hardness calculations and data processing were made with the indentation program.

Hardness depth profiling on cross sections was made on a coated and an uncoated sample to see if the coating process has a hardening effect on the iron substrate. If the coating had been thick enough its hardness could also have been tested. This was not the case. The cross sections were polished and a diagonal matrix of 4×25 indentations were setup with 5 µm distance at an angle of 45◦ _{relative the honed/coated surface.}

Figure 14: The setup matrix for surface mapping. The corner triangles are the previously made markings of the area. Only the upper two are visible because focus can’t be obtained in the whole image due to the curvature of the cylinder. Position 1,17,153 and 169 are the larger indentations for marking the position of the matrix.

3.4

Lubrication regime study

The lubrication regimes were studied using a Stribeck curve test performed in a lathe. The tests were performed with motor oil lubricant and a line contact symmetry (fig. 15). Two measurements on the coating and one reference measurement on an uncoated surface were made.

Figure 15: Schematic geometric setup for the lathe.

The rotation velocity was controlled with a LabView program. The velocities were calculated from measured rotation speed. This was done for the two velocities used (v1 and v2) and the acceleration was assumed to be constant. The velocity was first

held constant at v1= 0.014 m/s for 10 min. It was then increased to v2= 0.49 m/s

during 5 min and kept at v2for 10 min. The velocity was decreased back to v1during

5 min and kept at v1 for another 10 min. The velocity as a function of time is shown

in figure 16. The 10 min periods of constant velocity were to see if any running in occurs at the given velocities. The tests were performed with a constant contact force FN = 3 N and the viscosity of the lubricant is considered to be constant.

Figure 16: The velocity as a function of time used to study the lubrication regimes of the coating.

4

Results

4.1

Surface overview

All coated surfaces look smoother than an uncoated surface. LOM images of the studied samples are shown in figure 11 and at higher magnification in figure 17. VSI, as well as LOM, shows a clear difference in surface roughness between a coated and an uncoated surface. Comparing images of coated and uncoated surfaces are shown in figure 18. Note that the surface parameter analysis was made on selected sections of the honing plateaus. Measured values of Ra vary between 50 and 100

nm for the coatings, which should be compared with Ra ≈ 240 nm for the uncoated

surface (see table 2). The coated surface ANS-S has a slightly smoother surface, but there is no large difference between the coatings.

Table 2: Measured Ra values for the different surfaces. The analysis was performed on two

spots for each surface. Note that the analysis was made on a cut-out section of a plateau. The build-in compensation for tilt and cylinder shape was used.

Sample Ra,1 (nm) Ra,2 (nm)

ANS-4WS:80% 97 97 ANS-4WS 81 77 ANS-WS 75 68 ANS-S 43 54 ANS-8W 68 79 Uncoated 245 227

Tilted SEM images of sample ANS-4WS:80%, ANS-WS and ANS-S are shown in figure 19, together with an uncoated surface. It is clearly visible that the plateau surface is affected by the coating process, while the deeper honing scratches are more or less unaffected.

Secondary- and back scatter electron images at three different acceleration voltages (20 keV, 10 keV and 3 keV) are shown for all the samples in figures 21 –25. Figure 20 shows correlating simulations to visualize the effect on the interaction volume when varying acceleration voltage. For a more accurate thickness approximation see section 4.3.

All images show brighter plateaus and darker scratches which could correspond to a tungsten containing coating on the plateaus and a lack of coating in the scratches. The bright fields are irregular for all samples. This could be due to a material com-position variation or a coating thickness variation over the surface. Images of ANS-S (figure 24) show less contrast variation which means that the coating thickness vary less. The black area on coating ANS-4WS:80% (figure 21e) comes from debris on the surface. The striped phenomena occurring in figure 21f and 23f are due to problems with the contrast adjustments at the SEM and has nothing to do with the material composition.

(a) ANS-4WS:80% (b) ANS-4WS

(c) ANS-WS (d) ANS-WS

(e) ANS-8W (f) Uncoated surface

(a) Uncoated sample

(b) Coating ANS-S

(a) Uncoated surface (b) ANS-4WS:80%

(c) ANS-WS (d) ANS-S

Figure 19: Topography enhanced SEM images (SE mode) for three of the coatings and an uncoated surface. The topography is enhanced by tilting the samples 60◦_.

(a) 20 keV Simulation (b) 10 keV Simulation (c) 3 keV Simulation

Figure 20: Monte Carlo simulations of 40 nm WS2 on a Fe substrate, showing the effect of acceleration voltage

(a) 20 keV (SE) (b) 20 keV (BSE)

(c) 10 keV (SE) (d) 10 keV (BSE)

(e) 3 keV (SE) (f) 3 keV (BSE)

Figure 21: SEM images of sample ANS-4WS:80% at three different acceleration voltages. (a),(c) and (e) show SE images, (b), (d) and (f ) show BSE images in COMPO-mode.

(a) 20 keV (SE) (b) 20 keV (BSE)

(c) 10 keV (SE) (d) 10 keV (BSE)

(e) 3 keV (SE) (f) 3 keV (BSE)

Figure 22: SEM images of sample ANS-4WS at three different acceleration voltages. (a),(c) and (e) show SE images, (b), (d) and (f ) show BSE images in COMPO-mode.

(a) 20 keV (SE) (b) 20 keV (BSE)

(c) 10 keV (SE) (d) 10 keV (BSE)

(e) 3 keV (SE) (f) 3 keV (BSE)

Figure 23: SEM images of sample ANS-WS at three different acceleration voltages. (a),(c) and (e) show SE images, (b), (d) and (f ) show BSE images in COMPO-mode.

(a) 20 keV (SE) (b) 20 keV (BSE)

(c) 10 keV (SE) (d) 10 keV (BSE)

(e) 3 keV (SE) (f) 3 keV (BSE)

Figure 24: SEM images of sample ANS-S at three different acceleration voltages. (a),(c) and (e) show SE images, (b), (d) and (f ) show BSE images in COMPO-mode.

(a) 20 keV (SE) (b) 20 keV (BSE)

(c) 10 keV (SE) (d) 10 keV (BSE)

(e) 3 keV (SE) (f) 3 keV (BSE)

Figure 25: SEM images of sample ANS-8W at three different acceleration voltages. (a),(c) and (e) show SE images, (b), (d) and (f ) show BSE images in COMPO-mode.

4.2

Coating composition

The qualitative EDS analysis shows Fe(K,L), W(M), O(K), S(K) and C(K) peaks. Sample ANS-8W shows no sulfur. The qualitative ESCA analysis shows a content of Fe, W, O and C for all analyzed samples (ANS-4WS:80%, ANS-S and ANS-8W). For ANS-4WS:80% and ANS-S sulfur was also detected. The obtained ESCA and EDS spectra are attached in appendix A.1 and A.2. The quantitative analysis from EDS and ESCA are shown in table 3 – 7.

Assuming that:

• carbon is graphite from the cast iron substrate (∼ 3 wt%), • the iron signal originates from the substrate,

• all sulfur is bound to tungsten as W S2,

• all oxygen is bound to tungsten as W O3,

the relative distribution of tungsten compounds can be calculated (see table 3 – 7). For EDS on sample ANS-S there are a O surplus in the calculations, all other analyses show a W surplus. The EDS and ESCA measurements seems to correspond well.

Table 3: Quantitative analysis for coating ANS-4WS:80%. ANS-4WS:80% EDS plateau EDS plateau ESCA

Element average (%At) spot (%At) (%At)

C 17 5 11 O 28 44 39 W 14 29 30 S 2 3 1 Fe 39 19 19 WS2/Wtot 8 6 2 WO3/W_tot 66 52 43 W/Wtot 26 42 54

Table 4: Quantitative analysis for coating ANS-4WS. ANS-4WS EDS plateau EDS plateau Element average (%At) spot (%At)

C 22 19 O 11 31 W 12 36 S 2 8 Fe 54 6 WS2/Wtot 9 11 WO3/Wtot 32 29 W/Wtot 60 59

Table 5: Quantitative analysis for coating ANS-WS. ANS-WS EDS plateau EDS plateau Element average (%At) spot (%At)

C 14 18 O 7 30 W 4 19 S 2 6 Fe 70 27 WS2/W_tot 22 16 WO3/W_tot 53 52 W/Wtot 25 31

Table 6: Quantitative analysis for coating ANS-S.

ANS-S EDS plateau EDS plateau ESCA

Element average (%At) spot (%At) (%At)

C 15 0.1 16 O 6 5 7 W 3 4 16 S 5 7 16 Fe 71 83 45 WS2/W_tot 89 86 51 WO3/W_tot 11 14 49 W/Wtot 0 0 0

Table 7: Quantitative analysis for coating ANS-8W.

ANS-8W EDS plateau EDS plateau ESCA

Element average (%At) spot (%At) (%At)

C 23 26 28 O 15 20 21 W 18 26 36 Fe 45 28 15 WO3/W_tot 28 26 20 W/Wtot 72 74 80

The images from EDS mapping of sample ANS-4WS:80% are shown in figure 26. It is easily seen that W and O occur in the same area. S only gives a vague signal that is hard to distinguish from the background noise, but it could be anticipated to be in the same area as W and O. The area of high carbon intensity could be either debris on the surface or graphite from the cast iron. The iron image is approximately the inverse to the W image, which indicates that the Fe signal mostly originates from the substrate.

The images from EDS mapping of sample ANS-4WS are shown in figure 27. The C signal appears in the honing scratch or at small spots which indicates that it has its origin in substrate graphite or from rests of the process fluid. The higher O intensity in the scratch could be e.g. iron oxide or process fluid rests. The W, O and S images show higher intensities in the same areas and are inverse to the Fe image which indicates a coating of W, O and S on a Fe substrate. It could also be that Fe is mixed up in the coating.

The images from EDS mapping of sample ANS-WS are shown in figure 28. All elements but iron are hard to distinguish from the background noise. C appears strongest in the honing scratches. W and Fe are again inverse to each other.

The images from EDS mapping of sample ANS-S are shown in figure 29. O and C occur mainly in the honing scratch. W, S and Fe images are similar, lacking intensities in the scratch but to different extend. For Fe this could be due to overshadowing from O and C. W and S are probably only present on the plateau. This sample differs from the others in that there are mainly W and S on the plateaus and less O. This is also confirmed by the quantitative analysis in previous section.

The images from EDS mapping of sample ANS-8W are shown in figure 30. As for the other coatings there are high W and O intensities at the plateaus and high Fe and C intensities in the honing scratches.

(a) O (b) C

(c) W (d) Fe

(e) S (f) 10 keV (BSE)

Figure 26: EDS mapping of sample ANS-4WS:80% and the same area pictured with BSE. The mapping is performed at E0=10 keV.

(a) O (b) C

(c) W (d) Fe

(e) S (f) 10 keV (BSE)

Figure 27: EDS mapping of sample ANS-4WS and the same area pictured with BSE. The mapping is performed at E0=10 keV.

(a) O (b) C

(c) W (d) Fe

(e) S (f) 10 keV (BSE)

Figure 28: EDS mapping of sample ANS-WS and the same area pictured with BSE. The mapping is performed at E0=10 keV.

(a) O (b) C

(c) W (d) Fe

(e) S (f) 10 keV (BSE)

Figure 29: EDS mapping of sample ANS-S and the same area pictured with BSE. The mapping is performed at E0=10 keV.

(a) O (b) C

(c) W (d) Fe

(e) 10 keV (BSE)

Figure 30: EDS mapping of sample ANS-8W and the same area pictured with BSE. The mapping is performed at E0=10 keV.

4.3

Coating thickness

The comparison between Monte Carlo simulations and EDS-spectra give a coating thickness variation from 20 to 100 nm. The obtained approximations of the coating thickness are shown in table 8. The obtained thickness varies for each coating de-pending on if the EDS measurement was performed on a larger surface or at a single spot. This indicates that the thickness varies over the surface. For ANS-S there is less variation, which indicates a more homogeneous thickness distribution. The created BSE thickness images are shown in figure 31. Calculated mean thickness from the BSE thickness images are shown in table 8.

Table 8: Approximated coating thickness from simulation and BSE-topography. The approx-imations are based on an assumption that the coating either consists of only a WS2 or a

WO3 layer.

Average over ANS- ANS-4WS ANS-WS ANS-S ANS-8W

the surface 4WS:80% WO3 (nm) 30 25 20 20 40 WS2 (nm) 40 35 20 20 50 Spot measurement WO3 (nm) 60 80 45 20 50 WS2 (nm) 100 60 20 70 BSE thickness 36 6 map (nm)

4.4

Hardness

The uncoated substrate hardness was 4.7(±0.6) GPa. Calculated average for mea-sured nano hardness as a function of load-cycle number from the Progressive Multi Cycle measurements are shown in figure 46. Each color represents average and stan-dard deviation for each sample. There is no large difference between the two coatings but they clearly differ from the uncoated surface. The hardness as function of pene-tration depth for all measurements are attached in appendix A.3.

In-situ scratching was performed on three of the samples, 4WS:80%, ANS-WS and ANS-S, as shown in figure 33. All three coatings had a similar butter-like ploughing behavior.

Two thirds of the measurements on ANS-4WS:80% failed due to problem with the indenter and could not be used for mapping.

Created maps from ANS-S are shown in figure 34. The grayscale are relative within the image and not representing absolute hardness values, this means that the images can´t be directly compared.

Hardness as a function of depth under the surface (from cross section measure-ments) for coating ANS-S and the uncoated reference are shown in figure 32b. Cal-culated average values are marked with lines. No difference between the two samples can be distinguished. Average of all measurements are 5.9 GPa for both coated and uncoated cylinder liner segments.

(a) ANS-4WS:80%

(a)

(b)

Figure 32: (a) Measured hardness average and standard deviation for ANS-4WS:80% (green), ANS-S (blue) and an uncoated surface (red). The loading cycle represent indentation depth from 10 to 60 nm. (b) Hardness as a function of depth under the surface (from cross section measurements) for coated and uncoated sample. Calculated average values are marked with lines.

(a) ANS-4WS:80% (b) ANS-4WS:80%

(c) ANS-WS (d) ANS-S

Figure 33: SEM images of in-situ scratching for (a) the used tip on sample ANS-4WS:80%, (b) the scratch on ANS-4WS:80%, (c) the scratch on sample ANS-WS and (d) the tip and scratch on ANS-S.

(a) Cycle 1 (b) Cycle 2 (c) Cycle 3

(d) Cycle 4 (e) Cycle 5 (f) Cycle 6

(g) Cycle 7 (h) Cycle 8 (i) Cycle 9

Figure 34: Created hardness maps of sample ANS-S. Note that the grayscale are relative within each image although absolute values are comparable.

4.5

Lubrication regimes

Stibeck curve tests were performed on coating ANS-4WS:cyl. The obtained Stribeck curve is shown in figure 36. Full film lubrication occurs at lower velocities with a coated surface. The friction coefficient for boundary lubrication can not be seen due to the too high initial velocity. The periods of constant velocity v1 give a slight

running-in effect as shown in figure 35a. The running-in effect is more significant for the reference than the coating. The periods of constant velocity v2 show a slight

increase of friction coefficient (see figure 35b).

(a) v1

(b) v2

Figure 35: Running in effect of friction coefficient at Stribeck curve test. Data was not registered for part of the measurement in (b) due to software problem.

Figure 36: Stribeck curve for coating ANS-4WS:cyl and reference. Contact force and viscosity is considered to be constant so the friction coefficient can be plotted as a function of velocity.

5

Discussion

5.1

Surface overwiew

Both LOM, VSI and tilted SEM images show that the surface is clearly smoothened by the coating process. This is probably a combined effect of running-in of the substrate surface and smearing out the coating material. The estimated coating thicknesses are of the same order of magnitude as the measured Ra values. The BSE images also

indicate a material variation over the surfaces. It is not clear if the material variation is mostly due to the surface roughness or due to the underlying honing structure.

5.2

Coating composition

Both ESCA and EDS show peaks for the same elements, W, S, O, C and Fe. Unfor-tunately, most of the chemical shifts involved in ESCA are hard to distinguish from each other. Some information could still be obtained and are as follows.

• The chemical shifts between WS2, FeS and S for the sulfur peak can not be

distinguished from each other.

• The oxygen peak position is typical for metal-oxides, but it is not possible to distinguish between tungsten- and iron oxides.

• The tungsten spectrum from sample ANS-4WS:80% show peaks for both WO3

and W. This could also be possible for ANS-8W, but no clear WO3 peaks are

visible in the spectrum.

• The W4f7/2 peak is supposed to be higher than the W4f5/2 peak, this is not the case for sample ANS-4WS:80% and ANS-8W. This could be due to overlapping from the oxide peaks. This supports the idea that ANS-8W also contains WO3.

• The carbon peak for sample ANS-S shows a small hump to the right of the peak and could be due to some formation of WC. If this is really the case can not be confirmed or disproved by the tungsten peak since WC shift for tungsten are smaller than the resolution.

• The obtained Fe-spectrum shows an iron content for all samples. A shift for iron oxides are clearly visible for sample ANS-4WS:80% and possible for sample ANS-8W.

The assumptions made when calculating the tungsten compound distribution could neither be confirmed or disproved (section 4.2). For sample ANS-4WS:80% and prob-ably ANS-8W the presence of WO3 and iron oxide are confirmed. The presence of Fe

shown by surface sensitive ESCA analysis could indicate that Fe is mixed up in the coating. But there is no way to determine the exact analysis position in ESCA and hence, there is no guarantee that only a honing plateau was analyzed. This means that the Fe signal could as well come from a honing scratch. It could also be that the coating is not completely covering the substrate even at the plateaus.

In further testing it would be interesting to perform ESCA spectroscopy at a higher energy resolution to be able to separate the different compounds from each other. For this, reference spectra on pure elements/compounds from the same ma-chine would also be of interest. With better positioning it could also be possible to focus the analysis spot on a honing plateau and avoid the deep scratches.

In modern EDS systems, like the one used here, an automatic ZAF-correction is performed for the quantitative analysis. The ZAF-correction compensates for the dif-ference in back scattering (Z-factor), absorption (A-factor) and fluorescence (F-factor)

are homogeneously distributed through the sample. This, in addition to the uncertain quality of the calibration makes the quantitative EDS analysis an approximation, and can only be used in a relative comparison between the samples and not as absolute values.

Initial sputtering is performed to avoid the influence of adsorbed surface mole-cules in ESCA. Sulfur has a high gas pressure and is preferentially sputtered away which affects the element proportions. All samples are treated in the same way so the quantitative ESCA analysis can still be used in a relative comparison between the samples with the notion that the sulfur content is higher than it seems.

The carbon content of the analyzed samples is approximately 3 %wt, well cor-responding to the carbon content for cast iron (2-4% wt). The calculated tungsten distributions correspond well between ESCA and EDS analyses. The coatings made with tungsten additive in the process fluid (ANS-4WS:80%, ANS-4WS, ANS-WS, ANS-8W) show a higher content of tungsten and oxygen, while the coating made with only sulfur additive in the process fluid (ANS-S) shows much less oxygen and more sulfur. It seems as if the addition of a tungsten additive in the process fluid does not stimulate the formation of WS2.

EDS mapping shows that tungsten, sulfur and oxygen are mainly present on the plateaus and carbon and iron occur mainly in the scratches. This is confirming the idea that the iron and carbon signals have their origin in the substrate and that the coating mainly consists of tungsten, oxygen and sulfur. The absolute intensities can not be used to compare the coatings due to the software problem described in section 3.2.4. Again it is clear that the coatings made with tungsten additive in the process fluid contain more W and O, and the coating made with only sulfur additive in the process fluid contains more sulfur and less oxygen.

5.3

Coating thickness

The obtained approximations of coating thickness are based on a visual comparison between simulated and measured EDS spectra. The comparisons are made with respect to the relative hight difference of iron and tungsten peaks. The simulations are based on a given coating thickness for a given material with defined density. Simulations were performed for WO3 and WS2 but, as described in the previous

section, the coating composition and distribution is far from the ideal case. It should however indicate the coating thickness order of magnitude.

The coating thickness seems to vary both over the surface as well as between the coatings. This confirms the idea that the irregular pattern in the BSE images are due to a material variation, or indirect a thickness variation over the surface.

The generated BSE thickness images show a thickness and material distribution well corresponding to the BSE images and approximated thickness values. The thick-ness imaging is made based on approximations and should therefore only be considered as a coarse estimation. For more accurate imaging a lot more factors needs to be taken into account and the method could be much more developed. These first images are more a test of the idea that BSE images of different acceleration voltages could be combined with depth information than an actual developed imaging technique.

5.4

Hardness

Due to the thin coating thickness it is impossible to know the magnitude of the sub-strate influence when scratching. The observed soft behavior is similar to that of iron in the same situation. But no brittle behavior is visible so it can still be assumed that the coating is relatively soft.

Using the Progressive Multi Cycle mode in the nano indenter gives a large varia-tion of the measured values. The calculated standard deviavaria-tion for each loading-cycle is ∼ 30%, both for samples and reference and seems to be independent of the number of measured indentations4

. This could be due to the fact that the measurements are performed on a cylindrical surface, but since the same diversion was obtained also for the polished cross section reference, the geometry should be of less importance. It could also be due to a variation of coating thickness over the surface since the hardness could also be affected by the thickness variation. But this still does not explain the large standard deviation for the reference sample. The calculations of average value and standard deviation are done with respect to indentation cycle and not with respect to actual indentation depth. Absolute values of indentation depth could not be used as input, only the number of cycles in a given interval. The 9 cycles from 10 to 60 nm used here give an indentation depth with a standard deviation of ∼2 nm for each cycle. This could be the reason for the large deviation. It would be interesting to test the Progresive Multi Cycle test on a calibration sample to see if the deviation comes from the instrument or if it is related to the material. Cast iron is not homogeneous and could also be a contributing factor to the varying results.

Some problems occurred during use of the nano indenter UNHT and several of the measurements failed completely5

. The machine is in need of a relatively flat and aligned surface to be able to perform accurate measurements. Measuring on the inside of a cylinder therefore could lead to complications, but still does not explain why two thirds of the measurements failed for sample ANS-4WS:80% when all but three measurements succeeded for sample ANS-S. The first measurement set on the polished reference cross-section also failed completely and that is definitely not due to sample curvature. The origin of this intrument problem has not yet been located. Problem also occurred with the positioning of the indentations. Larger visible marks were made to be able to see the actual matrix position (figure 37). The black crosses show the supposed position, the red crosses show the actual position of origo and indentation markings. The position accuracy is supposed to be ± 0.25 µm, but here the diversion is of a magnitude of 4–10 µm. Due to the machinery problem, large deviations and low resolution it is hard to obtain any type of 2D information from the made hardness mapping. Some information can still be obtained by plotting the hardness as a function of indentation depth (see figure 46). The honed uncoated surface was to rough for the nano indenter and the reference measurements were per-formed on a polished cross-section. This should also be considered when comparing the measurements.

The cross-section measurements are performed in a different nano indenter and with different settings than the substrate reference measurements. For a better compari-son, new reference measurements should be performed with the same settings as the cross-section measurements and in the same nano indenter.

4

Using all 160 indentations from sample ANS-S for calculation gives a similar standard deviation as the 33 reference measurements.

5