ARTICLE IN PRESS

Tribology International 39 (2006) 15641575 www.elsevier.com/locate/triboint

The feasibility of using electrostatic monitoring to identify diesel lubricant additives and soot contamination interactions by factorial analysis
J.E. Bootha,, K.D. Nelsonb, T.J. Harveya, R.J.K. Wooda, L. Wanga, H.E.G. Powriec, J.G. Martinezb
a School of Engineering Sciences, University of Southampton, Higheld, Southampton. SO17 1BJ, UK Chevron Oronite Company, LLC, 100 Chevron Way, P.O. Box 1627, Richmond, CA 94802-0627, USA c Smiths Aerospace, Information SystemsSouthampton, School Lane, Chandlers Ford, Hampshire, SO53 4YG, UK b

Received 12 October 2005; received in revised form 31 March 2006; accepted 2 April 2006 Available online 8 June 2006

Abstract Electrostatic monitoring is a condition monitoring technique, which has been used for monitoring lubricated sliding contacts. The electrostatic charge is dependent on material wear and charge species in the lubricant (additives and contaminants). This paper presents work carried out on a pin-on-disc (PoD) tribometer to investigate additiveadditive and additivecarbon black interactions. Online electrostatic charge and coefcient of friction (CoF) measurements were recorded. Post-test electro-kinetic sonic amplitude (ESA) measurements were taken and pin and disc material loss was measured using 3D prolometry. Statistical examination of results was conducted using Analysis of Variance (ANOVA) to reveal interactions. The primary conclusions include: primary zinc dialkyldithiophosphate (ZDDP) was found to increase pin wear due to immature antiwear lm formation. Interactions between carbon black and detergent, and carbon black and dispersant, were observed in electrostatic charge data and ESA measurements. A complex between ZDDP and dispersant was highlighted by measured electrostatic charge, ESA, and pin material loss. r 2006 Elsevier Ltd. All rights reserved.
Keywords: Soot; Diesel oil; Electrostatic charge; Oil chemistry; Wear

1. Introduction Present day lubricant chemists are trying to achieve conicting demands of tighter emission regulations, increased fuel economy and extended drain time. The demands on the lubricant formulator are ever increasing and thus the rate at which lubricant formulations are superseded is increasing. The rate of product development and increased number of tests required for approval means that development costs are escalating. It is therefore important that the lubricant formulator extract the maximum amount of information from each test. Lubricant chemists are interested in the interactions between oil additives and try to suppress any antagonistic effects that
Corresponding author. Tel.: +44 23 8059 7667; fax: +44 23 8059 3230.

may be encountered to improve oil formulation. Additive interactions can be related to the charging mechanisms of the additive components within a lubricant. Electrostatic monitoring as an on-line monitoring technique, has been able to detect charges which are correlated with friction and wear [1]. It also has the potential to inform the lubricant chemist about the mobility of charged particles in the lubricant lm such as additives, contaminants (soot, etc.) interactions between additives. Preliminary work investigating the use of electrostatic monitoring for evaluating lubricant performance has been carried out by Wood et al. [2]. 1.1. Aim This paper presents research carried out to investigate additiveadditive and additivecontaminant interactions

E-mail address: jeb301@soton.ac.uk (J.E. Booth). 0301-679X/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.triboint.2006.04.004

ARTICLE IN PRESS
J.E. Booth et al. / Tribology International 39 (2006) 15641575 1565

by measuring on-line electrostatic charge. Of particular interest is the ability of dispersants to prevent carbon black from agglomerating (causing adverse wear) and in knowing whether sufcient anti-wear lms are formed at low temperature. Although the primary function of detergent is not being directly assessed, it is necessary to know if there are any interactions between these three (detergent, dispersant(s) and zinc dialkyldithiophosphate (ZDDP)) common types of additive components. 1.2. Diesel engine lubricant additives and soot contamination To achieve improvements in fuel efciency lower viscosity oils are used, which with the increase in contact loads result in thinner oil lms separating components, giving rise to durability concerns. Tighter emission regulations; specically for diesel engines, has lead to the use of exhaust gas re-circulation systems to reduce NOx emission. Also the deliberate ignition retardation to reduce peak cylinder combustion temperature and therefore reduce NOx production increases the amount of incomplete combustion products, such as soot. Both of these systems increase the amount of soot nding its way into the lubricating oil. The consumer demand for longer service intervals and the increased cost for oil disposal have lead to lengthening of engine oil drain times. Soot levels as high as 5wt%, in certain types of duty cycles, have been noted [3]. Therefore lubricant additives need to minimise the effect of contaminants and thinner oil lm for longer periods of time. This study focuses on interactions between three additive components, namely detergent, dispersant and ZDDP, as well as their interactions with carbon black (a soot surrogate) contamination. Zinc dialkyldithiophosphate is an anti-wear agent, oxidation and corrosion inhibitor. Under mixed lubrication, the two surfaces can intermittently penetrate the thin lubricant lm causing decomposition of ZDDP through asperity contact. ZDDP decomposition products react with metal surfaces and form a tribo-lm, which can be load bearing and reduce metal-to-metal contact. The thickness and composition of this anti-wear lm are directly related to the temperature and the extent of surface rubbing [4]. A fully developed ZDDP tribo-lm formed on a ferrous surface includes three layers (see Fig. 1). Initially ZDDP is reversibly absorbed onto the metal surface at low temperatures. As the temperature increases the ZDDP starts to decompose and sulphur will react with the exposed metal surface to form a thin iron sulphide layer close to the metal surface [5]. Subsequently phosphate forms an amorphous layer of short-chain ortho- and meta-phosphates with minor sulphur incorporation. The phosphate chains become longer towards the surface. This region is sometimes described as phosphated glass in which zinc and iron cations act to stabilise the glass structure. The outermost layer of the anti-wear lm, the phosphate chains contain increasingly more organic ligands, eventually

Organic ZDDP decomposition products and undegraded ZDDP replenishes the polyphosphate layer once it has been stripped Film removal

Iron Sulphide (FexSy)

Polyphosphate Glass

Steel

Iron oxides (FexOy)

Fig. 1. Three layers of a fully developed ZDDP antiwear lm.

giving way to a region comprised of organic ZDDP decomposition products and undegraded ZDDP itself. These surface-active compounds form a thin layer that preferentially shears under boundary lubrication conditions. As this layer needs to be constantly replenished, the concentration of ZDDP in the lubricant is critical. Detergents control corrosion and resin build-up in engines. They are metal salts of organic acids that contain an excess base, usually in the form of carbonate. Detergents neutralise inorganic acids from combustion products and organic acid products from lubricant oxidation and thermal decomposition resulting in a decrease in the corrosive tendency of the acids. The neutralisation of acidic blow-by gases is critical to diesel lubricant performances particularly for fuels with high sulphur content (10 and 500 ppm for Europe and US, respectively). The organic proportion of the detergent enables solubilisation of the salts formed by neutralisation and keeps them suspended in the bulk lubricant. Thus oil-insoluble combustion products, such as sludge or soot and oxidation products are dispersed. Detergents are also effective corrosion inhibitors, especially basic detergents [6,7] because they not only neutralise corrosive acidic products but also form surface lms that isolate metal surfaces from corrosive agents [8]. The main action area for detergents in an engine is around the piston, reducing lacquer, carbon and varnish deposits on the pistons as well as preventing pistonring sticking under high-temperature operating conditions. Before 1955 only detergents were used to keep engines clean, this was effective provided that the engine was operated at relatively high temperatures. Low operating temperatures as a result of short-distance/stop-and-go type driving does not allow engine oil temperature to rise sufciently to vapourise water from the oil, thus resulting in highly viscous oil due to severe oil oxidation and a high level of insolubles in the oil. In diesel engines the major contribution to insolubles comes from carbon (soot) and oxidation products of the fuel and oil. More recently

ARTICLE IN PRESS
1566 J.E. Booth et al. / Tribology International 39 (2006) 15641575

European OEMs have identied low-temperature wear as an area of concern [9]. This is a major issue for small diesel engines used in urban environments where start/stop driving is typical. Europe particularly has a large proportion of small urban diesel vehicles. Dispersants control sludge and reduce formation of deposits. They typically consist of a polar group usually oxygen or nitrogen based, and a high molecular weight non-polar group. The polar group associates with sludge or deposits (soot) and the non-polar group keeps the material suspended in the bulk lubricant. The mechanism whereby this occurs can be represented by a basic dispersant molecule attaching itself to an acidic site on a surface and the non-polar groups (hydrocarbon chains) prevent the soot particles from coming together. The stability of the dispersion is inuenced by the molecular weight of the nonpolar group and the level of reactive sites within the polar region. By associating with colloidal particles dispersants prevent them from agglomerating and falling out of suspension. Coalescence is prevented through either stericor electrostatic factors [8,10] thus the charge helps to prevent agglomeration. Dispersants also lower the surface/ interfacial energy of the polar species and reduce their adherence to metal surfaces. Soot is formed due to incomplete combustion of the fuel and contaminates the lubricating oil by travelling past the piston rings [11]. Soot production is mainly a concern in diesel engines, but can occur in direct injection gasoline engines. Soot is made up of hydrocarbon fragments, which form aromatic molecular networks [12]. Unless there are electrical and steric barriers to keep them apart, soot particles will agglomerate as they approach one another. The agglomeration is caused by both van der Waals forces and electrostatic attraction of the charges present on the soot surface [10]. The particles, typically less than 200 nm, which have a tendency to agglomerate into a larger macrostructure, will lead to a signicant rise in oil viscosity, with fuel economy consequences. When aggregates occur on surfaces, such as those of the combustion chamber, soot deposits result. These deposits are soft and aky in texture. Wear of engine components could be due to chemical reactions taking place on the surface (such as corrosion) or abrasive action on the material or anti-wear lm by soot particles. Gautam et al. [13] examined the effects of soot-contaminated engine oil on wear of engine components as well as other oil properties including phosphorus level, dispersant level and sulphonate substrate level. 1.3. Charge particle detection 1.3.1. Electrostatic charge monitoring A condition monitoring technique has been developed based on detecting electrostatic charge generation from deteriorating contacts. Electrostatic sensing was originally developed for detection of debris in the gas path of jet engines and gas turbines [1416]. The fundamental

principle of operation is detection of electrostatic charge associated with the debris present in the gas path. Changes in the charge level can then be related to the presence of increased amounts of debris and hence the onset of gas path component deterioration. The advantage of this technique is that it measures a direct product of the fault, rather than secondary effects, such as increased vibration or temperature exceedance. Thus, it is a useful tool for early detection and progressive monitoring of component deterioration. Collaborative work by the University of Southampton and Smiths Aerospace has investigated the application of electrostatic monitoring to oil systems. It has been demonstrated that this system can detect charge related lubricated contact breakdowns including metal-on-metal contacts and ceramic on steel contacts [1,1721]. For the rst time, this technique has been used to investigate charge changes due to oil chemistry as well as wear in this study. 1.3.2. Electro-kinetic sonic amplitude (ESA) The separation of charge, which exists at the particle liquid interface gives rise to several dynamic phenomena associated with colloidal systems. The driving force for electro-kinetic phenomena is the net charge at the interface between the liquid, which is bound to the particle, surface and the bulk uid. The potential of the interface, known as the plane of shear, is the zeta potential. When an alternating electric eld is applied to a colloidal dispersion, the particles will move in the electric eld because of their net zeta potential. If there is a density difference between the particles and the liquid, this oscillatory motion of the particles will result in the transfer of momentum to the liquid and the development of an acoustic wave. ESA is the pressure amplitude generated by the colloid per unit electric eld strength and is analogous to electrophoretic mobility. 1.4. The use of statistical methods for multiple variables and measured parameters Statistical methods can be used to determine the signicance of variable interactions and variable effects on observed measurements. When experimental costs are high, these methods allow extraction of unbiased information, regarding the factors affecting a variable, from as few observations as possible. Fractionated factorial matrices are used to reduce the number of tests required to evaluate varying order interactions, which is dened as the effect of one component as a result of the level (e.g. presence or absence) of another component. The aim of Optimal Design is to derive unbiased (or least biased) main effects and interactions with a minimum number of observations. The D- and A-optimal design procedures are methods used to select from a list of candidate points (combinations of factors), those points that will extract the maximum amount of information from the experimental region (the n-dimensional space where the

ARTICLE IN PRESS
J.E. Booth et al. / Tribology International 39 (2006) 15641575 1567

model is applicable). D-optimal design maximises the determinant D of a matrix, which indicates factor effects are maximally independent of each other. A-optimal designs also seeks independence, but by maximising the diagonal elements of the matrix, while minimising the offdiagonal elements. Searching for the best design is not an exact method, but rather an algorithmic procedure that employs D- and Aoptimal criteria along with the model (e.g. fractionated factorial) required to t to the data and the number of tests desired to nd the best design. Analysis of Variance (ANOVA) is used to uncover the interactive nature of reality, as manifested in high-order interactions of factors for design procedures described above. ANOVA is a statistical method used to determine whether two or more means are different. This procedure tries to identify sources of variability from one or more factors and uses these variances to decide whether the means are signicantly different. Further details of the statistical methods used can be found in [22]. 2. Experimental procedure Fig. 2 shows a schematic of the pin-on-disc (PoD) wear test rig with the associated instrumentation used. The electrostatic probe, to monitor charge, was positioned 901 away from the pin/disc contact and 0.5 mm above the disc wear track, a strain gauge was installed to monitor the contact friction force, and a tachometer to monitor the disc rotating speed. A data acquisition system was used to collect the electrostatic charge and coefcient of friction (CoF) signals in real-time. Suitable signal conditioning was applied to the sensors. 2.1. Test conditions and procedure PoD tribometer described above was used to simulate the wear of diesel engine valve train components. Experiments were carried out under ambient conditions (temperature 1523 1C, relative humidity 2060%) at a sliding speed of 5 m/s and load of 30 N (2.05 GPa initial
LVDT Load (N) B.S. 534A99 (En31) Pin Lubrication Pin Sensor 1 Button Type Wear Site Electrostatic Sensors (10 mm ) (a) B.S. 534A99 (En31) Disc
90

contact pressure). Hertzian contact stress within the valve train in the range 1.72.07 GPa has been reported for low emission diesel engines [23]. Thus the PoD conditions are mildly accelerated compared with typical valve-train entrainment velocities and contact pressure. Tests were carried out with a bearing steel ball loaded against a bearing steel disc, lubricated by different blends of oils containing various additive combinations in a basestock. The half factorial test matrix is shown in Table 3. The properties of the test material BS534A99 (pin and disc) are given in Tables 1 and 2, these are similar properties to the valve train components of interest. Bearing steels rather than cast iron have been used by other researchers looking at the fundamentals of diesel lubricant interactions with an iron-based surface [24]. Homogeneous bearing steels are used to enhance the reproducibility and repeatability of experiments; both of which are integral to validate statistical analysis. The test started with rotating the disc without contacting the pin for two minutes in order to record background signals before the oil was sprayed onto the disc using a pneumatic spray at a rate of 120 ml/h. Five minutes later, once the disc surface was fully lubricated, the pin was brought into contact with nominal initial load. This load was then ramped up to the maximum load of 30 N, using a hydraulic loading system, over a period of about 12.5 min. This was considered sufcient running-in and once the maximum load was reached the tests were run for 1 h. The test pins were ultrasonically cleaned in a mixture of acetone and ethyl acetate for 15 min before testing. The

Table 1 Elemental composition of steels Material Composition BS534A99 (En 31) 0.951.10% C, 0.100.35% Si, 0.400.70% Mn, 1.201.60 Cr, Remainder Fe

Table 2 Details of test conditions Ball Material Hardness (Hv30) Elastic modulus, E (GPa) Poissons ratio, n Density, r (kg/m3) Dimensions (mm) Roughness, Rq (mm) Absolute viscosity @ 30 1C (cP) Pressureviscosity index (1/Pa) Oil ow rate (ml/h) Oil temperature (1C) RPM Pure sliding speed (m/s) Load (N) Contact pressure (GPa) BS534A99 (En 31) 640 210 0.3 7.80 103 6 0.09 $133 $9 108 $120 1621 Variable 5 30 2.05 100 10 0.02 Disc

Force Transducer

220

(b)

Fig. 2. (a) Schematic of the PoD test set-up and (b) position of electrostatic sensor.

ARTICLE IN PRESS
1568 J.E. Booth et al. / Tribology International 39 (2006) 15641575

disc surfaces were also cleaned with the same solvent using cotton wool balls before testing. 2.2. Oil chemistry and test matrix This paper investigates the interactions between 4 additives and carbon black. Oil blends were prepared from baseline formulations that contained 7.3 wt% viscosity index improver concentrate (ethylene/propylene copolymer) with an API group I base oil. The four additives included one detergent, two dispersants and one type of ZDDP.

three or more factor interactions, however, it was decided that this number of interactions was not a key concern at this early research stage. The original 16 blends were designed so that the factor effects are maximally independent of each other (A- and D-optimality), which are shown in Table 3 as the test matrix. 2.3. On-line signal processing The real time data, including COF and rms charge were averaged for the 60 min duration to assess correlation over the whole test period. The data were also averaged for every 5 min interval (described as interval electrostatic charge hereafter) to assess correlations of dynamic realtime features. All average data were statistically assessed at the completion of the test matrix. 2.4. Off-line analysis 2.4.1. Surface prolometry The volume loss for the pin and disc were measured by TaiCaan 3D laser Prolometry. The volume loss for each tribo-couple was then inserted into Eq. (1) to calculate the specic wear rate (SWR): SWR VLmm3 ; F N SDm (1)

 

Raven 1040 carbon black was used as a surrogate for diesel engine soot and blended to 2 wt%. Raven 1040 has a primary aggregate particle of size 28 nm, and average agglomerated particle size just under 300 nm. The detergent chosen was an overbased sulphurised calcium phenate, which was blended to 50 milli-molar calcium (about 2 wt%). The two dispersants studied were both post treated conventional succinimides made from polyisobutene. Dispersant 1 was a bis-succinimide, which has one polar head per molecule and dispersant 2 was a polymeric succinimide, which has about 10 polar heads per molecule. Both dispersants were blended to 8 wt%. However, when the two dispersants were used together, they were blended at an equal percent of 8 wt% in total. A primary ZDDP was blended to 0.1 wt% phosphorus.

where VL is the volume loss of the disc or pin, F is the load and SD is the sliding distance. 2.4.2. Electro-kinetic sonic amplitude (ESA) A Matek Instruments MBS 8000 system was used to measure the ESA of the 16 oils. The instrument was congured for non-aqueous measurement, and the phase was referenced to use diesel engine oil that contained approximately 6 wt% of soot by thermogravimetric analysis. Ten measurements were performed for each sample and the results averaged. All phase measurements were within 151 of the reference. 2.4.3. X-ray photoelectron spectroscopy (XPS) Post-test XPS analysis was carried out on three worn pin samples by Evans Analytical Group (California, USA) using a PHI Quantum 2000 instrument. The X-ray beam has a spot size of 5 mm across and 95% of the analysed signal originates from a depth of $50100 A.

(Further details regarding the structures of the chemicals can be found in [2,25]). Oils containing carbon black were homogenised for 1 min using a rotor stator type-high shear mixer, model T25 Basic Dispenser w/S25N-18G, manufactured by IKA Laboratory Analytical and Processing Equipment, and then the oil was degassed in a vacuum oven for 15 min at about 55 1C. The non-carbon black containing blends were stirred for 10 min followed by 60 min in an ultrasonic bath at 30 1C. All blends were ultrasonically shaken for 30 min prior to testing on the PoD tribometer. Standard analysis would require all 5 variables to be separated by running 32 tests. For the work presented in this paper a fractionated (half) factorial matrix 2(51) (16 test oils) allowed the main effects and 2-factor interactions to be evaluated by ANOVA. The half factorial matrix described above has the disadvantage of insensitivity to
Table 3 Factorial matrix oil blends Oil run Detergent Dispersant 1 Dispersant 2 ZDDP Carbon black 1 2 | | | 3 4 | | | | 5 6 | | | | | | 7

8 | | |

10 |

11 |

12 | |

13

14 |

15 | | |

16 | | | | |

| |

| |

| | |

| |

ARTICLE IN PRESS
J.E. Booth et al. / Tribology International 39 (2006) 15641575 1569

3. Results The results for the measured on-line and off-line parameters are shown graphically in Figs. 35. ANOVA was carried out on both the on-line and off-line parameters. Only a few blend component interactions were detected for the 60 min electrostatic charge average. Electrostatic monitoring is sensitive to dynamic events and thus when ANOVA was carried out for each 5 min interval of the 60 min period, interactions were found, and in some cases persisted for the duration of the test. The statistical analysis yielded many interactions and therefore narrowing down these statistical results was necessary. This was achieved by focusing on the electrostatic results and calculating the signicance of such interactions; p-values. 3.1. p-valuesreduced model It is important to identify the most signicant effect(s) by a single factor or by two factor interactions. The residual error is required for estimating the probability of signicance. The source of variability is separated to determine whether the factor(s) or interaction affects a response. A model with only the most probable signicant effects was introduced to overcome insufcient degrees of freedom for error estimating required by a full model (i.e. randomised repeat tests). Extra degrees of freedom created by this model were used to estimate the error. The probability of errors, i.e. p-value, is used to indicate the signicance of the interactions on the variable and its associated response. For p-values less than 0.05, the interactions between the factors are considered signicant because only 5% of the interaction could be explained by other factors. The most interesting interactions with the lowest p-values discovered by ANOVA are listed in Tables 4 and 5.

4. Discussions A marginal correlation between ESA and 60 min average surface charge detected by an electrostatic sensor was suggested by a relatively high p-value of 0.15. Although both are charge related techniques, ESA is a measurement of the pre-existing charge in the oil, whereas the electrostatic sensor detects the total charge as a result of tribological activity, which is likely to be a combination of charging mechanisms. Thus the correlation is marginal rather than signicant. 4.1. Carbon black interactions with other additives and the effects on charge responses Carbon black increased the electrostatic charge detected and ESA measured (see Tables 4 and 5). Carbon black particles are known to be chargeable. In oil containing

0.10 0.09 0.08 0.07 0.06 0.05 Run 1 Run 2 Run 3 Run 4 Run 5 Run 6 Run 7 Run 8 Run 9 Run 10 Run 11 Run 12 Run 13 Run 14 Run 15 Run 16

Coefficient of Friction

Lubricant Blends
Fig. 4. Bar chart of the 60 min COF average.

Fig. 3. Bar chart of the 60 min average electrostatic charge values and the off-line ESA measurements.

ARTICLE IN PRESS
1570 J.E. Booth et al. / Tribology International 39 (2006) 15641575
Ball SWR

3.5x10-15 3.0x10-15 Ball SWR [m3/Nm] 2.5x10-15 2.0x10-15 1.5x10-15 1.0x10-15 5.0x10-15 0.0 Run 10 Run 11 Run 12

Disc SWR

2.5x10-13 2.0x10-13 1.5x10-13 1.0x10-13 5.0x10-13 0.0 Disc SWR [m3/Nm]

Run 13

Run 14

Run 15

Lubricant Blends
Fig. 5. Bar chart of disc and ball SWR.

Table 4 Statistical results from ANOVA for On-line parameters Factor Carbon black Dispersant 2 without detergent ZDDP with dispersant 1 Carbon black without detergent Carbon black without dispersant 2 Variable Average charge Interval charge Interval charge Interval charge Interval charge Response p-value Increase Decrease Increase Increase Increase 0.13 o0.05 o0.05 o0.05 o0.05

Table 5 Statistical results from ANOVA for post-test parameters Factor ZDDP Carbon black with ZDDP Dispersant 2 without dispersant1 Carbon black ZDDP without dispersant 2 Detergent with dispersant 1 ZDDP with dispersant 1 Variable Ball SWR Disc SWR Disc SWR ESA ESA ESA ESA Response Increase Decrease Decrease Increase Increase Decrease Increase p-value 0.06 0.01 0.04 0.01 0.04 0.09 0.12

2 wt% of carbon black, considering an average diameter of 300 nm for spherical carbon black agglomerates, there are 7 1012 carbon black particles in one ml of oil sample. Agglomerated particles, if large, could retain a high enough localised charge to be detected by the electrostatic sensor. Fig. 6 illustrates the statistical nding that carbon black with detergent will reduce interval electrostatic charge and carbon black with dispersant 2 will also reduce interval electrostatic charge. The statistical analysis indicates that both dispersant 2 and detergent reduce the effect that carbon black contributes towards increasing electrostatic charge; dispersant 1 does not have this effect. For dispersed carbon black to acquire a charge, dispersant and/or detergent must de-adsorb from the surface taking a charge

with it and leaving an equal but opposite charge on the aforementioned surface. Typically, dispersant molecules perform this function, and as they are usually amine-based they acquire a positive charge from the formation of an acidbase pair at the soot surface. Thus the dispersants in this study will effect negative surface charge formation on dispersed carbon black. Relative to dispersant 2, dispersant 1 is more surface active, and as there are fewer polar regions per non-polar hydrocarbon tailsthere are 2 hydrocarbon tails per polar region in dispersant 1it is more likely to de-adsorb from dispersed carbon black. Dispersant 2 may bind very tightly to a surface; there are about 10 polar regions per molecule, and the ratio of polar to non-polar regions is about 1:1. For these reasons, dispersant 2 does not support charge formation as well as dispersant 1. Detergents generally employ organic acid-based surfactants. These soaps are negatively charged. Consequently colloidal particles associated with detergents may obtain a positive surface charge. As there may be both positive and negative charges present when dispersant and detergent are used, these positively charged particles serve to offset, or neutralise the measured static charge. 4.2. Specimen wear Fig. 7(ad) shows 3D topographical ball scar maps grouped in relation to the presence or absence of ZDDP and carbon black, which have historically had the greatest impact on wear performance. The four maps are representative of the four different ball scar shapes seen for all 16 tests, but with varying severity. For oils without either ZDDP or carbon black (e.g. run 5, see Fig. 7(a)) the ball SWR was small and conformal. Oils containing carbon black, but no ZDDP exhibit slightly greater (abrasive) ball wear, the SWR for run 9 (see Fig. 7(b)) is double that of

Run 16

Run 1

Run 2

Run 3

Run 4

Run 5

Run 6

Run 7

Run 8

Run 9

ARTICLE IN PRESS
J.E. Booth et al. / Tribology International 39 (2006) 15641575 1571

0.18 0.16 0.14 0.12 QRMS / pC QRMS / pC 0

(a)

0.18 0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 5 10 15 20 25 30 35 40 45 50 55 60

(b)

0.10 0.08 0.06 0.04 0.02

10

15

20 25

30

35

40

45 50

55 60

Time / mins

Time / Mins

Fig. 6. Real-time electrostatic charge for (a) Run 9 (carbon black) (b) Run 14 (detergent, dispersant 2, carbon black).

Fig. 7. (a) Run 5 (dispersant 2), (b) Run 9 (carbon black), (c) Run 16 (detergent, dispersant 1, dispersant 2, carbon black, ZDDP) and (d) Run 1 (ZDDP).

run 5. For oils which contain ZDDP and carbon black the ball shoulders are severely worn leaving a Mohawk type feature in the middle of the ball (see Fig. 7(c)). A

comparison of run 16 to run 9 shows 14 times greater ball SWR for run 16. All oils containing both ZDDP and carbon black exhibit this Mohawk feature; but with varying

ARTICLE IN PRESS
1572 J.E. Booth et al. / Tribology International 39 (2006) 15641575

widths. The Mohawk width for run 16 (see Fig. 7(c)) is 220 mm, this is larger than the initial maximum Hertzian contact pressure diameter of 167 mm. For all other oils which contain ZDDP and no carbon black, the ball scar is approximately at, with small ridges and has the greatest ball SWR, a fact shown in the statistical analysis (see Table 5). Run 1 (see Fig. 7(d)) has the highest ball SWR out of the 16 test oils, it has 4 times more wear than run 9 and over 100 times more wear than run 5. This effect of ZDDP to exacerbate wear is contrary to what is generally thought of as an antiwear additive; in following sections this negative effect is termed pro-wear. The wear scar shapes; in particular those oils containing carbon black and ZDDP, can be explained by elastohydrodynamic theory. The surface of a point contact, even under small loads, will deform, initially elastically. Although these deformations may be very small they can play a vital role in promoting the generation of a uid lm to separate two surfaces in relative motion. In sliding elastohydrodynamic point contacts, deformation of the (originally) spherical surface will cause a constriction in the oil lm with an associated pressure spike, towards the downstream end and around the side of the contact, forming a horse-shoe shaped constraint [26]. This pressure spike can be considerably higher than the maximum Hertzian contact pressure. Optical interference patterns [27] have conrmed that the contours of constant lm thickness follow the contours of constant pressure [28], with the areas of minimum lm thickness occurring downstream and around the side (the horse-shoe) of the contact [29]. Wear for run 1 started around the area of maximum pressure/minimum lm thickness and progresses to the centre of the ball, attening the spherical surface. For oils containing ZDDP and other factors (additives and/or contaminants) the pro-wear effect of ZDDP is slowed preventing the progression of wear from the ball shoulders to the centre. The following sections discuss reasons behind the pro-wear effect and how this is minimised due to interactions with other additives, in particular dispersants. 4.2.1. ZDDPs pro-wear effect XPS was conducted on three worn ball surfaces and an unusual tribo-lm was found for those surfaces run with oils containing ZDDP. Table 6 shows the elemental compositions of these lms.

The results show low concentrations of zinc, sulphur and phosphorous compared to other ZDDP antiwear lms [30]. The low phosphorous level especially relative to sulphur indicates that a minimal amorphous-polyphosphate layer had formed. This layer is critical for an effective ZDDP antiwear lm because it shears preferentially to the metal component under boundary lubrication conditions. This amorphous layer is constantly replenished by the thin top layer of organic decomposition products and undergraded ZDDP. The pro-wear effect of ZDDP is due to the formation of an under-developed anti-wear lm. Primary ZDDP has high thermal stability and the tests were run at room temperature, which limited the decomposition of primary ZDDP, thus prevented the formation of a fully developed antiwear lm. However, the reaction of sulphur within ZDDP, with iron to form iron-sulphide is less temperature dependent. Active sulphur has a signicant inuence on promoting wear [3]. The thin phosphate layer is easily stripped away leaving an iron sulphide layer, which is also readily sheared leaving a nascent metal surface (see Fig. 8). This nascent metal surface is able to form iron oxide, which reacts with sulphur to form iron sulphide. This cyclic repeats to promote ball wear. 4.2.2. Interactions between ZDDP and dispersants The SWR for run 1 and run 7 clearly shows that dispersants reduces the pro-wear effect of ZDDP. The ZDDPdispersant interaction is thought to be a combination of three mechanisms: increasing the thermal stability

Film removal Small Phosphate layer

Iron Sulphide (FexSy)

Steel

Iron oxides (FexOy)

Fig. 8. Schematic diagram of proposed under-developed lm which leads to a pro-wear characteristicZDDP reacts with the surface to form initial stage of tribolm but constant removal of this iron rich lm will promote wear (Tribochemical wear).

Table 6 XPS analysis Ball specimen Run 5 Run 7 Run16 (centre) Run 16 (side) C 40.2 39.8 35.5 34.8 N 2.0 1.7 0.4 0.7 O 40.9 40.9 44.1 43.4 Na 0.8 0.7 0.9 0.8 Si 1.8 1.3 0.9 P 0.9 0.2 0.4 S 0.4 0.5 0.6 Cl 0.6 0.4 0.4 0.8 K 0.5 Ca 1.1 0.8 0.6 0.5 Fe 10.3 10.0 13.2 13.2 Zn 1.0 2.5 3.6 3.1 Pb 0.9 0.7 0.7 0.8

ARTICLE IN PRESS
J.E. Booth et al. / Tribology International 39 (2006) 15641575
+ + + + + D1 D1 D1 + + Energy Required for D2 + D2 D2 + D1 D1 + + + De-adsorption D1 D2 D1 + D2 D1 ZDDP +D1 ZDDP ZDDP ZDDP D2 + ZDDP ZDDP +D1 ZDDP ZDDP ZDDP ZDDP D1+ ZDDP ZDDP + D1 D2 ZDDP D1+ ZDDP ZDDP + D1 + D1 D1 D2 + D1 D2 D2 D1 + + ZDDP + + + + ZDDP ZDDP ZDDP ZDDP ZDDP ZDDP ZDDP ZDDP ZDDP ZDDP ZDDP ZDDP

1573

Steel d1

d1>>d2 d2

Steel

(a)

(b)

Fig. 9. Schematic of the (a) dispersant 1 and (b) dispersant 2 interaction with ZDDP which leads to a reduction in the pro-wear effect of ZDDP.

of ZDDP, suspension of ZDDP and ZDDP decomposition products, and competition for surface sites. Harrison et al. [31] reported formation of a complex between ZDDP and succinimide polyamine. The basic nitrogen from the dispersant was shown to form a stable complex with ZDDP. It has been demonstrated that the ZDDPamine complex makes the ZDDP more resistant to thermal degradation by retarding the rate of peroxide decomposition [32]. Usually a reduction in ZDDP decomposition products would reduce the ability of ZDDP to form an effective lm to minimise wear. However, under the test conditions in this study a reduction in the amount of ZDDP decomposition products, especially sulphur containing, is benecial as it minimises the pro-wear effect. The presence or concentration of sulphur is thought to be the cause of the pro-wear effect. Apart from reacting with ZDDP, the dispersants are likely to suspend undegraded ZDDP and ZDDP decomposition products. The signicance of the latter is that suspension prevents sulphur containing decomposition products from reacting with the steel surface, further reducing the pro-wear effect of ZDDP. The prevention of decomposed ZDDP from reaching the steel surface is also enhanced by surface competition between the dispersant molecules and ZDDP decomposition products. The more adsorbed dispersant molecules on the steel surface, the fewer ZDDP decomposition products can reach the surface. This effect is discussed in Section 4.2.3. Fig. 9 shows a simplication of three mechanisms associated with ZDDPdispersant interaction. The interaction between ZDDP and both dispersants is also shown by electrostatic charge and ESA and is thought to be related to the same interaction found for ball SWR.

Fig. 10 illustrates the statistical nding that ZDDP with dispersant 1 increases interval electrostatic charge. The interaction between ZDDP and dispersant 2 is shown by the statistical analysis to be only signicant for ESA. However, online electrostatic data shown in Fig. 10(b) indicates that there maybe a marginal reduction in electrostatic charge due to the presence of dispersant 2. This follows the marginal correlation between ESA and electrostatic charge discussed earlier. The suspension of ZDDP and ZDDP decomposition products by dispersant 2 and the corresponding reduction in charge, maybe due to the reasons discussed for the carbon blackDispersant 2 interaction. The tenacious bond formed between the 10 polar regions of dispersant 2 and ZDDP requires more energy to de-adsorb dispersant 2 from ZDDP than dispersant 1 from ZDDP. Thus the suspension of ZDDP by dispersant 2 is more stable hence the ESA measurement is lower. Dispersant 1 in the presence of ZDDP however, increased the charge detected by both electrostatic sensor and ESA. Dispersant 1 molecules can form multiple layers, but are more easily de-adsorbed; especially at the extremities. Therefore dispersant 1 has the ability to form a higher localised charge even if the retention of this charge is comparatively unstable. Fig. 10(b) exhibits oscillating charge features and interval charge was used to identify the transient nature of a wearing contact. However, the statistical analysis does not take into account the oscillating characteristic that occurs in a few experiments, neither does it assess if the oscillations have different periods. Future research work is required to investigate the effect of single additives on such transient electrostatic signals.

ARTICLE IN PRESS
1574 J.E. Booth et al. / Tribology International 39 (2006) 15641575

0.026 0.024 0.022 QRMS / pC QRMS / pC 0.020 0.018 0.016 0.014 0.012 0
(a)

0.026 0.024 0.022 0.020 0.018 0.016 0.014 0.012 5 10 15 20 25 30 35 Time / mins 40 45 50 55 60
(b)

10

15

20

25

30

35

40

45

50

55

60

Time / Mins

Fig. 10. Real-time electrostatic charge for (a) Run 4 (Detergent, Dispersant 1, ZDDP) (b) Run 6 (Detergent, Dispersant 2, ZDDP).

4.2.3. Effect of dispersant on wear For the oils which did not contain ZDDP, it appears that the additives in particular dispersants provided a lm which minimised the wear on the ball (see Fig. 7(a)). Both dispersants employed in this study are expected to form surface lms capable of small wear reductions. The surface packing efciency of the two dispersants is very different. A molecule of dispersant 1 will have two hydrocarbon tails standing out from the surface. Each of these tails will occupy space at the surface based on the area occupied by the polar region, and the volume lled by the motion of the hydrocarbon tail in the oil medium. The packing efciency of dispersant 1 is limited by the volume required for each molecule. Dispersant 2, on the other hand, contains hydrocarbon tails linked together through each polar region; the distance between the anchor points for the hydrocarbon tails is of the order of several carboncarbon single bonds. The molecule is expected to form comb-like adsorbed layers [33]. The close proximity of the hydrocarbon tails serves to thicken the surfactant layer in the contact (relative to dispersant 1) and reduce wear. When both dispersants were employed, the anti-wear effect of dispersant 2 was not observed (see Table 5); clearly, competition for the surface favours the lower molecular weight and more mobile dispersant 1. 5. Conclusions A ve factor test matrix 2(51) was carried out to investigate the interactions between different factors on measured parameters including: charge, wear and friction. The ve factors include four additives (detergent, dispersant 1, dispersant 2 and ZDDP) and one oil contaminant (carbon black a soot surrogate). Analysis of Variance (ANOVA) revealed single factor effects and two factor interactions on the measured parameters of electrostatic charge and ESA and specimen SWR. The presence of carbon black was found to increase average electrostatic charge and ESA. Where steric or

electrical barriers are insufcient to keep carbon black particles dispersed, agglomeration will occur; these large particles could retain a high enough charge to be detected by the electrostatic sensor. Dispersant and detergents can also charge carbon black through de-adsorption. Both dispersant 2 and detergent in the presence of carbon black were found to reduce the interval electrostatic charge. Both dispersant 2 and detergent are known to bind tightly to the carbon black surface; minimising charge formation through de-adsorption. ANOVA revealed that ball wear was increased in the presence of ZDDP. Investigations using XPS analysis identied that under the test conditions used in this study, the presence of primary ZDDP did not generate a fully developed antiwear lm. Instead, formation and stripping of iron sulphide promoted ball wear. This paper has demonstrated the use of electrostatic sensors and ESA to detect changes in charge associated with the presence of additives and a contaminant, as well as detecting binary interactions on a PoD tribometer. Electrostatic sensors that correlate or show anti-correlation with off-line electrokinetic measurements such as ESA, could be extremely useful for lubricant performance evaluation. There is potential for electrostatic sensors and ESA to be used in a test programme to monitor the interaction between additives and contaminants, as contaminant loading varies during red engine tests. By comparing the pre-existing charge in the lubricant measured by ESA with the electrostatic charge detected from the trib-contact, further understanding of additive and contaminant interactions could be gained. Acknowledgements The authors would like to thank Dr. S. Stults, Dr. E.S. Yamaguchi and Dr. J.J. Harrison from Chevron Oronite Company for nancial and technical assistance; Dr. V. Palekar from Chevron Global Lubricants; Tai Caan technologies for use of a 3D laser prolometer, and

ARTICLE IN PRESS
J.E. Booth et al. / Tribology International 39 (2006) 15641575 1575

RAEng for support through their International Travel Grant scheme. References
[1] Tasbaz OD, Wood RJK, Browne M, Powrie HEG, Denuault G. Electrostatic monitoring of oil lubricated sliding point contacts for early detection of scufng. Wear 1999;230:8697. [2] Wood RJK, Penchaliah R, Wang L, Harvey TJ, Nelson K, Yamaguchi ES, et al., Electrostatic monitoring of the effects of carbon black on lubricated steel/steel sliding contacts. In: Proceedings of the 31st Leeds-Lyon symposium on tribology, 2004. [3] Rudnick LR, editor. Lubricant additives chemistry and applications. New York: Marcel Dekker; 2003. p. 398. [4] Fujita H, Spikes HA. The formation of zinc dithiophosphate antiwear lms. Proc Inst Mech Eng Part J 2004;218(4):26577. [5] Shiga M,Hirano K, Matsushits M. Methods for preparing overbased lubricating oil additives. US Patent 4,057,504 (8/11/77). [6] King LE. Basic alkali metal sulfonate dispersions, process for their preparation, and lubricants containing same. US Patent 5,037,565 (6/8/91). [7] Koch P, Di Serio A. Compounds useful as detergent additives for lubricants and lubricating compositions. US Patent 5,021,174 (4/6/ 91). [8] Rizvi SQA. Additives and additive chemistry, ASTM MNL 37 Fuels and lubricants handbook: technology, properties, performance, and testing. Additives and Additive Chemistry, 2005, p. 199248. [9] Mortier RM, Orszulik ST, editors. Chemistry and technology of lubricants. second ed. London: Blackie Academic & Professional; 2003. [10] Smiechowski MF, Lvovich VF. Characterization of non-aqueous dispersions of carbon black nanoparticles by electrochemical impedance spectroscopy. J Electroanal Chem 2005;577:6778. [11] Patterson, Henein NA. Emissions from combustion engines and their control. Ann Arbor, MI: Ann Arbor Science Publication; 1972. [12] Kornbrekke RE, Patrzyk-Semanik P, Kirchner-Jean T, Raguz MG, Bardasz EA. Understanding Soot Mediated Oil Thickening Part 6: Base oil effect. SAE 982665. [13] Gautam M, Chitoor K, Durbha M, Summers JC. Effect of diesel soot contaminated oil on engine wearinvestigation of novel oil formulations. Tribol Int 1999;32:68799. [14] Powrie HEG, McNicholas K. Gas path condition monitoring during accelerated mission testing of a demonstrator engine. AIAA Paper 97-2904, 1997. [15] Nurse J, Petch C, Fisher CE. Engine gas path integrity monitoring. Aerotech 94, Paper C470/6/052, 1994. [16] Powrie HEG, Fisher CE. Engine health monitoring: towards total prognostics. In: Proceedings of the international conference on condition monitoring proceedings, 1999. p. 17590.

[17] Powrie HEG, McNicholas K. Gas path condition monitoring during accelerated mission testing of a demonstrator engine, In: Proceedings of the 33rd AIAA/ASME/SAE/ASEE joint propulsion conference & exhibit, 1997. [18] Tasbaz OD, Wood RJK, Browne M, Powrie HEG, Denuault G. Electrostatic monitoring of oil lubricated sliding point contacts for early detection of scufng. Wear 1999;230(1):8697. [19] Wang L, Wood RJK, Harvey TJ, Morris S, Powrie HEG, Care I. Wear performance of oil lubricated silicon nitride sliding against various bearing steels. Wear 2003;255:65768. [20] Wang L, Wood RJK, Care I, Powrie HEG. Electrostatic wear sensing of ceramicssteel lubricated contacts. In: Proceedings of the 30th Leeds-Lyon symposium on tribology, 2003. p. 71119. [21] Wang L, Wood RJK, Harvey TJ, Morris S, Powrie HEG, Care I. Feasibility of using electrostatic monitoring for oil lubricated ceramic to steel sliding contacts. In: Proceedings of the 29th Leeds-Lyon symposium on tribology, 2002. [22] Box GE, Hunter WG, Hunter JS. Statistics for experimenters: an introduction to design, data analysis, and model building. New York: Wiley; 1978 0-471-09315-7. [23] McGeehan JA, Ryason PR. Preventing catastrophic Camshaft Lobe failures in low emission diesel engines. SAE 2000-01-2949, 2000. [24] Yamaguchi ES, Ryason PR, Yeh SW, Hansen TP. Boundary lm formation by ZnDTPs and detergents using ECR. Tribol Trans 1998;41(2):26272. [25] Vartanian PF. The chemistry of modern petroleum product additives. J Chem Edu 1991;68:1015. [26] Jacobson B. Regimes of elasto-hydrodynamic lubrication. J Mech Eng Sci 1991;12(1):916. [27] Foord CA, Wedeven LD, Westlake FJ, Cameron A. Optical elastohydrodynamics. Proc Inst Mech Eng 196970;184:487506. [28] Evans HP, Snidle RW. The EHL of point contacts at heavy loads. Proc R Soc 1982;A382:1839. [29] Chittenden RJ, Dowson D, Dunn JF, Taylor CM. EHL lm thickness in concentrated contacts. Proc R Soc 1985;A387 24569 and 27194. [30] De Barros MI, Bouchet J, Raoult I, Le Mogne Th, Martin JM, Kasrai M, et al. Friction reduction by metal suldes in boundary lubrication studied by XPS and XANES analyses. Wear 2003;254: 86370. [31] Harrison PG, Brown P, McManus J. 31P NMR study of the interaction of a commercial succinimide-type lubricating oil dispersant with zinc(II) bis(O,O0 -di-iso-butyldithiophosphate). Wear 1992;156:3459. [32] Inoue K, Watanabe H. Interactions of engine oil additives. ASLE Trans 1983;26:18999. [33] Chevalier Y, Dubois-Clochard M- C, Durand J- P, Delfort B, Gateau P, Barre L, et al. Adsorption of poly(isobutenylsuccinimide) dispersants at a solid-hydrocarbon interface. Prog Colloid Polym Sci 2001;118:1104.