Wear 274–275 (2012) 229–237

Contents lists available at SciVerse ScienceDirect

Wear
journal homepage: [Link]/locate/wear

Low wear metal sliding electrical contacts at high current density

Nicolas Argibay, W. Gregory Sawyer ∗
Department of Mechanical and Aerospace Engineering, University of Florida, Gainesville, FL 32611 USA

a r t i c l e i n f o a b s t r a c t

Article history: Operation of a low wear (2 × 10−5 mm3 /(N-m)), low contact resistance copper sliding electrical contact
Received 4 January 2011 was demonstrated. The wear rate of a lightly loaded copper–beryllium metal fiber sliding on a polished
Received in revised form 28 August 2011 copper counterface was insensitive to (DC) current density values as great as 440 A/cm2 (in a brush
Accepted 6 September 2011
positive or anodic configuration). Low wear and relatively low friction ( ∼ 0.2 to 0.3) was achieved by
Available online 12 September 2011
operating the contact immersed in a liquid medium consisting of a hydrofluoroether with helium cover
gas, inhibitingoxidationand providing cooling of the contact. Similar experiments performed in liquid
Keywords:
mediums of ultrapure water and dilute (3%) hydrogen peroxide show an order of magnitude increase
Sliding electrical contacts
Wear
in wear rate and provide further insight on the role of electrochemically enhanced oxidation and the
Friction degraded contact resistance and tribological behavior of non-noble sliding electrical contacts in general.
Hydrofluoroether In contrast to high current density slidingin hydrofluoroether, an order of magnitude greater wear rate
Copper was observed for similar sliding conditionsin hydrogen peroxide or water without the aid of externally
supplied electric potential. A conceptual model is proposed correlatingthe rate of brush wear to fatigue
strength and electrochemically enhanced oxidation as a result of high current density transport through
the contact. A mathematical expression was derived to calculate the approximate wear volume of a single
fiber laterally contacting a slip-ring, based on direct measurement of the wear scar geometry.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction eration of waste heat, in both cases in contrast to metal multi-fiber

brushes. Also, monolithic brushes exhibit a limited ability to track
In a 1979 review paper on the state of the art in current collection typically rough slip–ring surfaces, particularly at high speed [2].
technology McNab [1] suggested that construction of metal fiber Kuhlmann–Wilsdorf estimated that typical graphite brushes slid-
brushes is one of the primary technical challenges in the design of ing on metal slip–rings will exhibit an order of magnitude higher
modern current collectors capable of unprecedented performance drop in electrical potential compared to metal–metal sliding elec-
with regard to high current density and low wear rate. The term trical contacts [3].
brush as referring to a sliding electrical contact has its origin in some As described in detail by authors like Holm [4], the same surface
of the first commercial designs of electric motors where current col- films (oxides, moisture, and carbonaceous compounds) responsible
lection was achieved by means of bundled copper fibers resembling for desirable friction and wear behavior of metal sliding electrical
painter’s brushes [1]. Early current collector designs implement- contacts play a detrimental role on contact resistance, in the form
ing metal–metal contacts were plagued with high friction, wear, of film resistance. One of the defining features of metal sliding elec-
and heat generation mainly attributed to electrochemical oxida- trical contacts is that for all practical purposes the overall system
tion and arcing. A paradigm shift in current collection technology performance of a large scale device will be intimately tied to the few
occurred at the end of the 19th century with the invention of the hundred nanometers describing the interface of the sliding contact.
carbon brush. Further improvements in carbon brush technology One of the surprising consequences of this research has been
came with the invention of electrographite and metal-graphite the realization that under the right conditions metal sliding elec-
composite brushes, both of which are still used today in a host of trical contacts are quite robust and capable of sustaining extreme
applications like fuel pumps, power generation turbines, and elec- density of current transport with minimal energy loss. The ability
tric rail transportation. The main disadvantages of carbon brushes to sustain ten or one-hundred fold greater current density than a
remain the low system compliance of monolithic brushes and the carbon brush while maintaining reasonably low wear rates with
inherently high resistance to electrical conduction resulting in gen- relatively inexpensive materials (e.g. copper wire) may facilitate
more compact electrical machinery.
This work presents experimental research highlighting the crit-
∗ Corresponding author. Tel.: +1 352 392 8488; fax: +1 352 392 1071. ical role of interfacial phenomena and electrochemistry on the
E-mail address: wgsawyer@ufl.edu (W.G. Sawyer). performance of self-mated copper sliding electrical contacts, some

0043-1648/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/[Link].2011.09.003
230 N. Argibay, W.G. Sawyer / Wear 274–275 (2012) 229–237

Fig. 1. Photographs of friction testing apparatus (tribometer) illustrating (a) the major components of the tribometer, (b) a close-up of the copper–beryllium fiber, copper
counterface, fluid retention cup, pickup brush (multi-fiber Cu–Be brush at a relatively high contact load), and two capacitance probes used to measure deflection of the flexure
based cantilever and rigidly attached fiber (visible in c) as a calibrated response to frictional and normal components of force; (c) also illustrates the electrical connection to
the single fiber (pin).

features of which are generally applicable to other non-noble met- transport while immersed in water, resulted in order of mag-
als. Friction, wear, and contact resistance measurements were used nitude larger wear rates. Also, for the case of sliding in water
as a metric of performance for the case of copper sliding elec- with current transport, the presence of current flow immediately
trical contacts. Sliding contacts were immersed in the following led to deterioration in contact resistance, in stark contrast to the
prescribed liquid mediums and cover gases: apparently insensitive contact resistance response of the same
materials operated in an oxidation inhibiting environment under
(1) a dilute (3%) hydrogen peroxide (H2 O2 ) and water solution otherwise equivalent test conditions. This insensitivity provides
exposed to laboratory air, indirect evidence that deterioration in the performance of a slid-
(2) ultrapure water exposed to laboratory air, ing electrical contact is directly correlated to enhanced oxidation,
(3) And a hydrofluoroether with an inert cover gas (helium was a hypothesis first proposed by Boyer et al. [5]. A more in-depth
chosen over heavier and less expensive inert gases due to the investigation attempting to quantify the degree of enhanced oxi-
higher rate of diffusion of He into HFE, minimizing the time dation remains pending. For a more detailed investigation of copper
needed to displace trace amounts of dissolved oxygen as a result oxidation kinetics in dilute hydrogen peroxide see DeNardis et al.
of exposure to lab air during storage). [6]. Details of copper oxidation in water may be found in M. Pour-
baix’s “Atlas of Electrochemical Equilibria in Aqueous Solutions”
These fluids were selected so as to possess certain properties in [7].
common, such as viscosity near room temperature and specific heat This work also presents an alternate approach to the investi-
capacity (the ability to cool the contact due to Joule heating when gation of the tribological performance of metal sliding electrical
transporting 100 s A/cm2 ). By implementing materials like copper contacts. A single fiber contact removes many of the complexities
(with a relatively high thermal diffusivity) it is possible to maintain and uncertainties characteristic of multi-fiber brush experiments
a cool contact while at elevated current density, minimizing the typically found in the literature [1,3,8–14] where rough estimates
role of temperature on oxidation kinetics and highlighting the role must be made on the number of fibers in contact, affecting the
of electrochemical processes on the tribological performance of the real distribution of mechanical and electrical load and confounding
contact. analyses of tribological performance. In the method described here
More specifically, while maintaining otherwise similar mate- the wear scar is directly observed and its area quantified, making
rial and environmental properties it was shown that a copper analysis of nominal contact stress and current density. Similarly,
sliding electrical contact exhibits wear rates and contact resis- this method makes it possible to measure wear after sliding hours
tances that are insensitive to the density of current transport up rather than weeks, typically a necessity in larger scale experiments
to values as high as 400 A/cm2 for sliding distances exceeding due to the inherently low wear rates and high uncertainty (due
10,000 cycles (∼1000 m, corresponding to about 24 h of sustained to factors such as splaying of multi-fiber brushes) [9]. A mathe-
operation) when sliding in an oxidation inhibiting environment matical expression was derived to estimate the wear volume of a
(in this case a hydrofluoroether). It was also demonstrated that metal fiber laterally contacting a slip-ring, based on direct observa-
enhanced oxidation, achieved chemically using a dilute hydro- tion and measurement of the wear area. The ability to prescribe an
gen peroxide solution or electrochemically as a result of current environment, and to immerse the contact, also adds a greater level
 N. Argibay, W.G. Sawyer / Wear 274–275 (2012) 229–237 231

Fig. 2. (a) A photograph of the front view of the metal fiber mounted to the cantilever and an illustration of the fiber geometry and wear volume, an example of (b) 3D
micrographs and (c) 2D micrographs acquired by scanning white light interferometer microscopy used to estimate the semi-major radius (a) of the elliptical wear scar, used
in the calculation of the wear volume.

of variability and control in the continuing effort to understand the ZYGO NewView 5032) to evaluate the surface finish. The average
complex interactions that define these types of contacts. (Ra ) and root-mean-square (Rq ) surface roughness were calculated
based on measured values for three arbitrarily chosen representa-
tive scans (365 × 365 ␮m2 areas) of the running surface for each
2. Description of experiments counterface, and were in all cases Ra ∼ 15 nm and Rq ∼ 20 ±10 nm,
respectively. Testing was promptly initiated after imaging. Tests
2.1. Tribometer description and sample preparation were periodically interrupted to image the fiber and assess the vol-
ume loss as a function of sliding distance, requiring a removal from
Friction experiments were performed on a custom built fric- the prescribed environment lasting approximately 15 min.
tion testing apparatus (tribometer) shown in Fig. 1. The tribometer
operates in a pin-on-disk configuration. The pin consisted of a sin- 2.2. Force measurement and friction coefficient
gle bent copper–beryllium fiber (UNS C17200, full hard from Little
Falls Alloys, USA) with a nominal diameter (2r) of 120 ␮m (see Fig. 2 The pin (metal fiber) was coiled and epoxied to a polymer cylin-
for illustration of geometry). The bend radius (R) varied in the range der with a small hole for the fiber, made of polyetheretherketone
of 2.95–3.05 mm due to spring-back of the coiled fiber. The counter- (PEEK), meant to electrically insulate the fiber from the tribome-
face was the flat end of a polished 20 mm diameter, 30 mm long cop- ter frame. Refer to Fig. 1(c) for illustration. The PEEK cylinder is
per cylinder (UNS C11000 from McMaster-Carr, USA). The copper epoxied to the tip of a cantilevered double-leaf flexure. Contact
cylinder was press-fit onto a polymer cup intended to hold a liquid between pin and counterface is achieved by vertically displacing
volume of 12.5 cm3 , submerging the counterface and bottom half of the instrumented head, consisting of the flexure cantilever and two
the fiber. The polymer cup was machined from virgin polytetrafluo- capacitance probes. Friction coefficients are calculated from deflec-
roethylene (PTFE) rod stock. The counterface was mounted directly tions of the cantilever tip as a response to normal and traction forces
to the shaft of a stepper drive motor (Schneider Electric Motion, on the pin, measured using two capacitance probes. A detailed
USA, MDrive 17). A fixture consisting of a pair of aluminum disks sits description of force measurement and application using capaci-
between the counterface and motor shaft, with a thin foam spacer tance probes and leaf-flexure cantilevers can be found in [15–17].
and three equally spaced screws along the circumference fastening Based on the displacement measurement limits of the capacitance
the two disks together. The fixture was adjusted manually using probes the range of measurable normal and traction forces was on
the three screws until the motor shaft and counterface axes were the order of ±2 mN, with an optimal resolution (based on a 15 nm
aligned, eliminating load fluctuations due to disk runout. Runout probe resolution and neglecting other sources of error) of ∼0.6 ␮N.
resulted in a normal force fluctuation approximately ±14% of the In practice, typical fluctuations in load measurement when the pin
target load (e.g. for a 0.50 mN normal force the load fluctuated sinu- was brought into contact with a stationary disk were on the order
soidally from 0.43 to 0.57 mN). Prior to every test, the running sur- of 5 ␮N.
face of the copper counterface was wet sanded with progressively
finer sandpaper down to 1000 grit, followed by wet lapping with 2.3. Electrical current source and setup
1 ␮m Al2 O3 –water slurry on a lapping felt pad. The polished cylin-
ders were promptly rinsed with laboratory grade isopropyl alcohol In some experiments a direct (DC) electrical current was applied
and imaged using a scanning white light interferometer (SWLI; through the sliding contact. The metal fiber was always configured
232 N. Argibay, W.G. Sawyer / Wear 274–275 (2012) 229–237

as the positive element (anode). Current was supplied using a digi- Table 1
Summary of steady-state experimental results. Uncertainties in wear rate measure-
tal multimeter and power supply (Keithley 2400 Sourcemeter) in a
ments (Kvol ) are shown in Fig. 3, and standard deviations in friction coefficients ()
constant voltage output mode (current is allowed to vary as a func- and contact resistances (Rc ) are shown in Fig. 4.
tion of the changing contact resistance). The line resistance of the
Liquid/cover gas Kvol (mm3 /(N-m)) J (A/cm2 )a  Rc ()
circuit was calibrated to be approximately 100 . Reported con-
tact resistance values are calculated using Ohm’s Law, based on the 3% H2 O2 /lab air 2.9 × 10−4 0 0.49 –
measurements of current flow through and voltage drop across the Ultrapure water/lab air 2.6 × 10−4 265 0.1–0.4b 68
Ultrapure water/lab air 1.2 × 10−4 0 0.46 –
contact. The voltage drop is measured at the two soldered connec-
Hydrofluoroether/helium 2.2 × 10−5 88 0.15 26.5c
tions near the contact, one near the end of the pick-up brush and Hydrofluoroether/helium 1.8 × 10−5 0 0.31 –
one near the metal fiber (visible in Fig. 1(c) where the current deliv- Hydrofluoroether/helium 1.7 × 10−5 442 0.27 8.2
ery and voltage tap wires are soldered to the coiled copper wire in a
Nominal current density (J) is calculated based on the cross-sectional area of the
series with the metal fiber). This measurement type is generally 120 ␮m diameter metal fiber (∼1.13 × 10−4 cm2 ).
b
referred to as a four-point probe, where two wires deliver electri- Steady-state friction was not achieved, the values were highly variable during
cal current by means of a power supply (the circuit through which periods of applied electric potential.
c
A faulty pick-up brush resulted in poor contact for this experiment (the stan-
the current flow is being measured) and two wire taps are placed as
dard deviation in the steady-state contact resistance Rc was consequently high,
close to the sliding contact as possible so as to measure mainly the  ∼ 20.7 ), which may also be correlated to the relatively low measured friction
voltage drop through the sliding contact (minimizing the contribu- coefficient for this instance.
tion of bulk material). A pick-up brush is necessary to complete the
electrical circuit through the rotating slip-ring. This setup implies 3. Results
that a part of the contact resistance will be due to the bulk resis-
tance of the copper wire coil, the length of copper–beryllium 3.1. Metal fiber wear rate
fiber, and the length of the multi-fiber copper pick-up brush. The
contribution to contact resistance is estimated to be less than Wear was assessed by measuring the volume removal rate on
about 0.5 , a negligible fraction of the reported values for contact the metal fiber as a function of sliding distance. The specific wear
resistance. rate (Kvol in mm3 /(N-m), Eq. (1)) or volumetric wear rate is reported
for these experiments.
Vloss
2.4. Environment: liquid medium and cover gas Kvol = (mm3 /N-m) (1)
Fn s
For the experiments described here the sliding contact was The specific wear rate is the ratio of volume loss (Vloss ) and the
operated immersed in a prescribed liquid medium. The tribome- product of sliding distance (s) and applied normal force (Fn ). The
ter was operated either exposed to laboratory air (approximately average normal force value of 0.5 mN was used to calculate the wear
35% relative humidity and 21 ◦ C) or inside a gas tight chamber rate. The sliding distance is calculated based on the calibrated RPM
with a continuous flow of high purity, laboratory grade helium. and wear track radius (a nominal wear track radius of 8 mm and a
Three different liquid mediums were tested: (1) ultrapure (18 M- linear sliding speed of 10 mm/s were used, upon a freshly prepared
cm) water, (2) a 3% hydrogen peroxide and water solution, and copper counterface, for all experiments). The measurement of wear
(3) a hydrofluoroether engineered fluid (3MTM , NovecTM 7500). A volume on the toroid fiber is based on the geometry of the wear
helium cover gas was only used for case (3) where it was desirable scar, evaluated using SWLI microscopy micrographs such as the
to minimize oxidation of the metal fiber and counterface due to example shown in Fig. 2. An expression was derived for calculating
oxygen and water dissolved in the HFE during storage and expo- the approximate wear volume on the surface of a thin diameter (2r)
sure to lab air. The possibility of hydroplaning of the immersed fiber bent into a circular arc (a toroid) of measured radius (R). The
metal fiber at 10 mm/s sliding speed was negligible; observed fric- worn volume is effectively the segment of a toroid, producing an
tion coefficients are indicative of sliding in the boundary lubrication elliptical flat spot on the side of the metal fiber with a measured
regime. semi-major radius (a). The detailed derivation of the expression
for the volume of a toroidal segment (Eq. (2)) is provided in the
appendix.
2.5. Data acquisition 
a4 r
Vloss ∼
= (2)
Data was acquired using a desktop PC with custom LabVIEW 4R R
software and a National Instruments DAQ card (PCI-6229). Ana- The metal fibers had a variation in bend radius (R) between 2.95 and
log voltages are supplied by the capacitance probe controller as 3.05 mm which was recorded for each experiment, and a nominal
inputs and are supplied to the stepper motor controller as out- fiber radius (r) of 60 ␮m. Wear rate data is shown in Fig. 3. Refer
put via an NI SCB-68 connector block. The analog signals sent to Table 1 for a summary of test results. The following is a list of
from the capacitance probe controller for each probe is acquired notable observations pertaining to the wear data of Fig. 3.
at 10 kHz and averaged into a single data point per second.
Each data point is plotted and saved to a tab delimited text 1. The wear rate of the metal fiber sliding immersed in hydroflu-
file, and corresponds to the average of 10,000 discreet measure- oroether was approximately 2 × 10−5 mm3 /(N-m) and was
ments. The resolution of the acquisition card is 16-bit, so for a insensitive to the sustained passage of DC electrical current
±10 V range of input there is a maximum theoretical resolution through the junction up to a density of 442 A/cm2 based on the
of 0.3 ␮V. For the known stiffness of the cantilevers used in this metal fiber cross-sectional area (or more than 2300 A/cm2 based
study this corresponds to an effective normal force resolution on the worn contact area).
limit of 61 nN, or several orders of magnitude smaller than the 2. The wear rate of the metal fiber sliding immersed in 3%
measured values. The custom software calculates friction coef- H2 O2 was an order of magnitude greater, at approximately
ficient based on the capacitance probe signals, plots and saves 3 × 10−4 mm3 /(N-m).
data, and communicates with the motor to set RPM and start/stop 3. The wear rate of the metal fiber sliding immersed in ultra-
motion. pure water without electric current flow was also an order of
 N. Argibay, W.G. Sawyer / Wear 274–275 (2012) 229–237 233

Fig. 3. Plot of volume loss of the toroid metal fiber (the plot ordinate) as a function of the product of applied normal load and sliding distance, the sliding distance, and the
sliding time for seven experiments; the table on the bottom right of the plot shows operating parameters common for all experiments (note that where a range of values
is shown these are not due to uncertainty but are measured differences or fluctuations about the average); the table on the top right summarizes the specific wear rates
(based on the change in volume between intervals of sliding) for the various environments and current densities (based on the nominal fiber cross-section and the measured
steady-state DC current average); the current density values shown next to each data point for the hydrofluoroether experiments were based on the measured area of the
wear scar on the fiber at that point in the experiment.

magnitude higher than the case of hydrofluoroether, and was of the pin and counterface to lab air during imaging (allowing nat-
sensitive to the addition of an electrical current load through the ural oxidation to occur on otherwise oxide free surfaces operating
contact. Upon applying a 5 V DC supply to the circuit (consisting in hydrofluoroether). A nearly order of magnitude higher contact
of a 100  line resistance and the contact resistance, in series) resistance was observed after prolonged operation in water. This is
the wear rate was comparable. Although a sharp increase in in contrast to operation in hydrofluoroether. Friction coefficients in
contact resistance from 10  to 68  was observed, along hydrofluoroether were approximately 0.27 (with a standard devi-
with a decrease in friction coefficient, it was not possible to ation of 0.06), and in oxidative mediums (ultrapure water and 3%
determine if there was a change in wear rate during the period H2 O2 ) approximately 0.46 (with a standard deviation of 0.03).
of sliding with current due to the relatively high uncertainty
in measurement of the wear scar area (the contact area and
reflectivity were irregular, in contrast to well defined, flat, more 3.3. Quantifying wear rate uncertainty
reflective wear scars observed after sliding in HFE). The friction
coefficient showed a gradual recovery to its original value after The uncertainty in wear rate u(Kvol ) is a function of the uncer-
a period of sliding without current. tainties in measured volume loss u(Vloss ), normal force u(Fn ), and
sliding distance u(s). As described in greater detail by Burris et al.
[18] and Argibay et al. [9] the uncertainty in volume is typically
3.2. Friction and contact resistance the primary source of uncertainty in calculation of specific wear
rates for these experimental setups. Eq. (4) is an expression of the
Friction coefficient and contact resistance results are shown in uncertainty in specific wear rate. The derivation is similar to that
Fig. 4 and are again summarized in Table 1. The plots in Fig. 4 appearing in Burris et al. [18] and Schmitz et al. [19], and assumes
represent full data sets. For a linear sliding speed of 10 mm/s the negligible uncertainty in the measurement of radii r and R.
total duration of a test lasting 100/200/1000 m of sliding corre-
sponds to 2.8/5.5/27.8 h of test duration. Average friction coefficient   
a3 r
and contact resistance values shown in Fig. 4 are calculated for u(Vloss ) ∼
= u(a) (3)
periods of relatively steady-state operation, as illustrated in the R R
figure.
For the case of sliding while immersed in ultrapure water the A conservative approximation of the uncertainty in measurement
friction coefficient gradually increases during the first 100 m of slid- of the wear scar semi-major radius u(a) ∼ 5 ␮m is assumed for all
ing (∼2.8 h). Upon applying a DC current through the contact there measurements. This approximation is based on the error in esti-
is a sharp drop in friction coefficient and increase in contact resis- mating the boundary of the wear scar on a topographical white
tance, attributed to an electrochemical enhanced process at the light interference micrograph. For the metal fiber under considera-
interface. Contact resistance at the onset of electrical current flow tion (r = 60 ␮m, R = 3.0 mm) and a value for a of 100 ␮m, the volume
was similar in magnitude to the steady-state value observed for the loss based on Eq. (2) is Vloss ∼ 3.7 × 10−6 mm3 and the uncertainty
case of hydrofluoroether particularly following the brief exposure in volume loss based on Eq. (3) is u(Vloss ) ∼ 0.74 × 10−6 mm3 .
234 N. Argibay, W.G. Sawyer / Wear 274–275 (2012) 229–237

Fig. 4. Representative data sets for friction coefficient and contact resistance (Ohms) vs sliding distance (m) for four experiments with the common theme of using a single
toroid shaped Cu–Be fiber as the pin (120 ␮m diameter, ∼3 mm bend radius) sliding on a polished rotating Cu counterface (average roughness ∼ 15 nm), with a linear sliding
speed of 10 mm/s (wear track radius ∼ 8 mm, counterface rotation speed 12 rpm), and an average normal load of 0.53 mN; the sliding contact was immersed in the liquid
medium stated above each plot, and where indicated an DC electrical current was applied through the contact using a constant voltage of 5 V through a total circuit resistance
consisting of a 100  line resistance + the contact resistance (Rc ).

4. Discussion temperature, hinting at the importance of effective cooling and

frictional heating due to high contact pressure and/or sliding speed.
4.1. Role of oxidation on the wear rate of metal sliding electrical The observed trends in contact resistance measured for the
contacts copper systems evaluated in this study, contrasting between mea-
surements of copper sliding electrical contacts operating immersed
Previous studies by Reichner et al. [12,13] and Boyer et al. [5,20], in water or a hydrofluoroether, appears to support the proposed
among others, describe a phenomenon observed in metal sliding role of oxidation kinetics on the wear rate of sliding metal con-
electrical contacts with a high density of DC electrical current trans- tacts constructed from non-noble materials. There was negligible
port where the positive (anodic) brush wears at a higher rate than change in the contact resistance of copper sliding electrical contacts
the negative (cathodic) brush while operating in humidified inert sliding in an oxidation inhibiting, thermally conductive, electrically
cover gases. Boyer et al. [5] proposed a theory based on a modifi- insulating medium (a hydrofluoroether with a helium cover gas).
cation of the Cabrera–Mott oxidation model [21,22] whereby the A two-fold increase in the wear rate and a gradual increase in con-
electric field through the copper oxide surface films, generated tact resistance with time were observed for the contact immersed
due to large current driving potentials, may be sufficient to sur- in water when an electric potential was applied to the contact
pass the chemical potential driving ion diffusion through the oxide (constant 5 VDC through a total circuit resistance consisting of
layer, resulting in an enhanced (or inhibited, depending on polarity) a 100  line resistance + the measured contact resistance, result-
growth rate of the oxide. Boyer et al. postulate [5] that for a positive ing in a maximum of 30–50 mA of current flow, depending on the
brush (anodic) the oxidation growth rate is enhanced, and for a neg- environment). Contacts immersed in dilute (3%) hydrogen perox-
ative brush (cathodic) it is inhibited, often by as much as an order of ide solution, a medium known to enhance copper oxidation [6],
magnitude with respect to the natural, chemical oxidation rate in a wore at a rate approximately equal (or slightly greater) to the case
given oxidizing environment. The model is sensitive to localized of immersion in ultra-pure water.
 N. Argibay, W.G. Sawyer / Wear 274–275 (2012) 229–237 235

4.2. Proposed correlation between fatigue strength and

delamination wear

Previous published [8,9,23] and unpublished research by the

authors implementing multi-fiber copper and copper–beryllium
brush sliding electrical contacts in humid carbon dioxide environ-
ments revealed that the wear morphology of these systems was
predominately flake-like and of a relatively large aspect ratio (typ-
ical debris flake length of 100 ␮m and thickness of 1–10 ␮m). This
type of debris morphology is documented for copper systems in
Argibay et al. [8,9] and for silver multi-fiber brushes in Brown et al.
[24]. Copper debris gathered from the anode brush from an exper-
iment using copper multi-fiber brushes at high current density
(180 A/cm2 ) similar to the procedure described in [9] was analyzed
by I.L. Singer (Naval Research Lab). X-ray diffraction (XRD) revealed
that the flake-like debris was predominately copper, rather than
copper oxide (Cu2 O), implying that it is not simply a matter of
enhanced oxidation resulting in an increased rate of copper oxide
growth followed by film delamination on the positive brush (and
equally inhibited rate on the negative brush). That is, the wear
debris is not simply delamination of copper oxide films. In fact,
oxide films are typically on the order of 1–10 nm in thickness [5,21]
whereas the copper debris flakes were on the order of 1–10 ␮m
in thickness, indicating that the depth at which delamination was
occurring surpassed the depth of oxide by an order of magnitude.
Argibay et al. [8] observed an order of magnitude reduction in wear
rate at high current density (200 A/cm2 ) by replacing copper multi-
fiber brushes with higher fatigue strength, higher yield strength
copper–beryllium multi-fiber brushes, an indirect indication of the
importance of fatigue strength.
The authors propose a conceptual model (illustrated in Fig. 5)
suggesting a correlation between the modified oxidation rate
mechanism proposed by Boyer et al. [5], the fatigue strength of the
surface materials, and the wear rate of non-noble metal contacts at
high current density in an oxidative medium. In a 1973 publication
Fig. 5. Illustration of a proposed conceptual model for wear of lightly loaded metal
Nazarenko et al. [25] provided experimental evidence of what they sliding electrical contacts correlating (a) the model of delamination wear (adapted
referred to as dislocation trapping under an increasingly thicker from Suh et al. [31]), (b) a reduction in surface fatigue strength due to surface films
oxide film on a zinc single crystal due to cyclic frictional stress. or oxides as reported by Buckley [26] and Duquette [29], and (c) the phenomenon
of dislocation trapping under increasingly thicker surface films or oxides as pro-
They showed by etching and microscopy that there is an increase
posed by Shen et al. [30] and Nazarenko et al. [25], which are correlated to the
in dislocation density under the friction surface with increasingly enhanced/inhibited electrochemical oxidation model proposed by Boyer [5].
thicker oxide films. Buckley [26] observed that relatively ductile
gold substrates exhibited surface fatigue cracks due to cyclic fric-
tional stress with the addition of a gold halide surface film. Wang of aluminum specimens with varying degrees of prescribed surface
et al. [27] reported a 15–30 fold difference in the fatigue strength film thicknesses. The link between dislocation trapping and wear
of copper when experiments were performed in air or high vac- may exist in the theory of delamination wear proposed and devel-
uum (with higher fatigue strength in vacuum). Thompson et al. oped by Suh et al. [31,32], Fleming et al. [33,34], and Jahanmir et al.
[28] showed that the fatigue life of copper single and poly crys- [35,36] in the 1970s. The delamination theory of wear provides a
tals fatigued in dry nitrogen was approximately 10 fold greater mechanism for the nucleation of subsurface dislocations as a func-
than when in air. They also found that the fatigue strength of cop- tion of frictional stress near a metal friction surface. They also show
per specimens could not be extended by annealing (in vacuum) that the depth of maximum shear stress due to contact stress in slid-
alone prior to failure through fatigue, though removal of the top ing tends to occur at a depth typically an order of magnitude greater
few micrometers (where the majority of damage was observed) than the thickness of a metal oxide, which may explain the observed
through chemical etching elongated the fatigue life of copper speci- thickness of wear debris formed from a fatigue process during slid-
mens effectively indefinitely. Duquette [29] performed fatigue tests ing exacerbated or alleviated by the electrochemical rate of oxide
on nickel single crystals in vacuum, dry air, and humid air and formation. Although there are numerous competing mechanisms
similarly reported a substantial decrease in fatigue strength while affecting the wear behavior of copper sliding electrical contacts at
operating in an oxidative environment. Duquette suggests that the varying degrees of current density there is ample evidence sug-
source of this phenomenon may be explained in a conceptual model gesting that the correlation between fatigue strength, dislocation
proposed by Shen et al. [30], wherein dislocations nucleating near trapping, and enhanced oxidation may, in some instances, accu-
a metal surface due to cyclic stress are inhibited from propagating rately describe the leading pathway to wear particle formation.
and relieving at the surface under increasingly thicker oxide films. It is proposed that for lightly loaded metal sliding electrical con-
With increasingly thicker oxides, Shen also suggests a mechanism tacts there is a possible correlation between surface fatigue wear
of dislocation trapping near the metal–oxide interface. An increase and enhanced electrochemical oxidation. The rate of delamination
in dislocation density near the surface may correlate to a greater (surface fatigue wear) is exacerbated by the presence of thicker
rate of sub-surface crack nucleation and propagation. Shen’s model oxides. The electrochemically enhanced/inhibited rate of oxide
was based in part on his own experimental data from fatigue testing growth may be coupled to the rate of delamination by a process
236 N. Argibay, W.G. Sawyer / Wear 274–275 (2012) 229–237

Fig. 6. Geometry of a toroid segment (the shaded volume) corresponding to the wear geometry observed on the metal fibers; the image illustrates coordinate axes and
variable definitions implemented in the derivation of an approximate expression for the volume of a toroid section.

of dislocation trapping. The electrochemically enhanced/inhibited as 450 A/cm2 based on the nominal fiber cross-sectional area (or
rate of oxide growth is correlated to the asymmetry in wear rate greater than 2300 A/cm2 based on the measured wear scar area).
observed in previous published studies of high current density
multi-fiber brushes as a function of brush polarity (anode or pos- Acknowledgements
itive brushes wearing faster). Consequently, for copper sliding
electrical contacts operating immersed in an oxidation inhibiting The authors would like to thank Arwed Uecker (Mersen Co.) for
hydrofluoroether liquid medium (where the metal fiber was in an fruitful technical discussions. The authors would also like to thank
anode configuration, the case of enhanced wear) there should not Irwin L. Singer (Naval Research Lab) for characterization of copper
be and was not observed to be a strong correlation between current wear debris, providing invaluable insight into the mechanism of
density and oxide growth rate (and consequently also wear rate). wear debris formation. This work was funded in part by a grant from
Mersen Co. (previously Carbone-Lorraine) and in part by the Office
of Naval Research (ONR) under grant number N00014-09-1-0584.
5. Conclusions

The experimental research summarized in this paper exam- Appendix A. Appendix

ines the role of environment on the friction, wear, and contact
resistance characteristics of self-mated copper sliding electrical This section presents the derivation of an approximate solution
contacts. More specifically, the experiments were designed to for the volume of a toroid segment used in the calculation of worn
remove some of the complexities typically encountered in macro- volume. Fig. 6 illustrates the geometry, variables, and coordinate
scopic scale experiments with multi-fiber metal brushes. Arguably axes considered. The trigonometry identities of A.1 are used as the
the most notable conclusion derived from the data is the observed starting point (shorthand notation will be used for references to
insensitivity of brush wear rate to current density when operating equations; for example, Eq. (A.1) will be referred to simply as (A.1)).
in an oxidation inhibiting fluorinated solvent (a hydrofluoroether). 2
R2 = (R − ıs ) + s2 R − ıs = R cos (A.1)
This is in contrast to a previously established correlation between
the wear and enhanced electrochemical oxidation of copper (and A Taylor series expansion of the cosine term is implemented.
other non-noble metals) in oxidative environments (humid CO2 , Only the first two terms are considered, the higher order terms
humid N2 , etc.) above a threshold current density [5,9,12,13,37]. being neglected based on the assumption of small angles.
The fact that the phenomenon of enhanced wear was not
2
observed in an oxidation inhibiting medium (operating the con- cos ∼
=1− (A.2)
tact immersed in liquid hydrofluoroether with an He cover gas) 2!
supports the theory proposed by Boyer et al. [5] that it is, for a (A.3) are derived from (A.1) and (A.2) for the range of values
positive (anode) brush, the enhanced rate of oxidation due to the 0 ≤ s ≤ a.
applied electric field across insulating oxide films (needed to drive
current flow through the contact) that causes a higher wear rate at s s2
∼
= ıs ∼
= (A.3)
high enough current density. The insensitivity of wear rate to cur- R 2R
rent density for these copper sliding electrical contacts at relatively (A.4) correlates the term hs and the parameters a, s, and R.
high current density in a non-oxidizing environment suggests that
by inhibiting oxidation the film resistance and thus the applied a2 − s 2
hs + ıs = d → hs = (A.4)
electric field across the film (the voltage drop) must remain low. 2R
The possibility of achieving sustained current densities
An expression for the cross-sectional area As corresponding to
(400+ A/cm2 ) while maintaining relatively low, current density
the worn portion of a slice of the toroid segment is given in (A.5).
insensitive wear ratesthrough a lightly loaded copper sliding elec-
Note that the small angle approximation is applied once again such
trical contact was demonstrated. This behavior was achieved by
that the cross-section is assumed to be circular for all values of s,
immersion of the sliding contact in a liquid medium that inhibited
neglecting the fact that with increasing s the circular cross-section
cold-welding of the nominally oxide free copper surfaces, pro-
will become an ellipse.
vided effective cooling (mitigating Joule heating), and prevented
oxidation. This system exhibited an effective decoupling of current As ϕ 1
= r 2 − hs r cosϕ (A.5)
density from metal fiber wear rate for current densities as high 2 2 2
 N. Argibay, W.G. Sawyer / Wear 274–275 (2012) 229–237 237

Following the same procedure used to derive (A.3), the expressions [10] D. Kuhlmann-Wilsdorf, What role for contact spots and dislocations in friction
of (A.6) are found, again assuming small angles for ϕ. and wear? Wear 200 (1996) 8–29.
[11] I.R. McNab, J.L. Johnson, High-current brushes. Part III. Performance evaluation
b b2 for sintererd silver-graphite grades, IEEE Transactions on Components, Hybrids,
ϕ∼
= hs ∼
= (A.6) and Manufacturing Technology CHMT-2 (1979) 84–89.
r 2r [12] P. Reichner, Metallic brushes for extreme high current applications, IEEE Trans-
Then the expression (A.5) may be reduced as in (A.7). Note that actions on Components, Hybrids, and Manufacturing Technology CHMT-3
(1980) 21–25.
this approximate expression for the area As is not a function of [13] P. Reichner, High current tests of metal fiber brushes, IEEE Transactions on
semi-minor radius b. Components, Hybrids, and Manufacturing Technology CHMT-4 (1981) 2–4.

As ∼ hs 
[14] P.G. Slade (Ed.), Electrical Contacts: Principles and Applications, Marcel Dekker,
As ∼ b3 New York, 1999.
= → = 2hs r (A.7)
2 4r 2 2 [15] H. Liu, B. Bhushan, Adhesion and friction studies of microelectromechanical
systems/nanoelectromechanical systems materials using a novel microtri-
An expression for the approximate volume of the toroid segment boapparatus, in: Papers from the 49th International Symposium of the
is then calculated by integration of (A.7), with the aid of (A.4), as American Vacuum Society, 2003.
shown in (A.8). [16] A.C. Rennie, P.L. Dickrell, W.G. Sawyer, Friction coefficient of soft contact lenses:
 a 
measurements and modeling, Tribology Letters 18 (2005) 499–504.
[17] P.L. Dickrell, S.B. Sinnott, D.W. Hahn, N.R. Raravikar, L.S. Schadler, P.M. Ajayan,
3 a4 r
Vloss ∼
=2 As ds → Vloss ∼
= (A.8) W.G. Sawyer, Frictional anisotropy of oriented carbon nanotube surfaces, Tri-
s=0
16 R R bology Letters 18 (2005) 59–62.
[18] D. Burris, W. Sawyer, Measurement uncertainties in wear rates, Tribology Let-
The small angle approximation results in an under-estimation ters 36 (2009) 81–87.
in volume calculation. This error is evident and quantifiable for the [19] T.L. Schmitz, J.E. Action, J.C. Ziegert, W.G. Sawyer, The difficulty of measuring
low friction: uncertainty analysis for friction coefficient measurements, Journal
special case of where r = R (a spherical cap). Comparing a closed- of Tribology-Transactions of the Asme 127 (2005) 673–678.
form solution for the volume of a spherical cap and the approximate [20] L. Boyer, J.P. Chabrerie, J. Saint-Michel, Low wear metallic fibre brushes, Wear
volume calculation (A.8) for the special case of r = R, the adjusted 78 (1982) 59–68.
[21] N. Cabrera, N.F. Mott, Theory of the oxidation of metals, Reports on Progress in
expression (A.9) is shown to achieve greater precision. Physics 12 (1947) 163–184.
 [22] N.F. Mott, The theory of the formation of protective oxide films on metals-III,
 a4 r
Vloss ∼
= (A.9) Transactions of the Faraday Society 43 (1947) 429–434.
4 R R [23] J.A. Bares, N. Argibay, N. Mauntler, G.J. Dudder, S.S. Perry, G.R. Bourne, W.G.
Sawyer, High current density copper-on-copper sliding electrical contacts at
The precision of (A.9) increases for increasingly smaller values low sliding velocities, Wear 267 (2009) 417–424.
of the ratio a/R, a characteristic of the toroidal geometry (e.g. for [24] L. Brown, D. Kuhlmann-Wilsdorf, W. Jesser, Testing and evaluation of metal
R = 3.0 mm, a = 100 ␮m, a/R ∼ 0.03). For the example of a spheri- fiber brush operation on slip rings and commutators, components and packag-
ing technologies, IEEE Transactions 31 (2008) 485–494.
cal cap with r/R = 1 and a/R = 0.1 the error in expression (A.9) is
[25] P.V. Nazarenko, Y.I. Korolenko, Effect of oxide films on the variation in disloca-
approximately 0.3%. tion structures and friction forces, Materials Science 6 (1973) 209–212.
[26] D.H. Buckley, Surface Effects in Adhesion, Friction, Wear, and Lubrication, Tri-
bology Series 5, Elsevier, 1981.
References
[27] R. Wang, H. Mughrabi, S. McGovern, M. Rapp, Fatigue of copper single crystals
in vacuum and in air. I. Persistent slip bands and dislocation microstructures,
[1] I.R. McNab, Recent advances in electrical current collection, Wear 59 (1980) Materials Science and Engineering 65 (1984) 219–233.
259–276. [28] N. Thompson, N. Wadsworth, N. Louat, Xi. The origin of fatigue fracture in
[2] P.W. Kendall, I.R. McNab, G.A. Wilkin, Recent developments in current collec- copper, Philosophical Magazine 1 (1956) 113–126.
tion, Physics in Technology 6 (1975) 117. [29] D. Duquette, M. Gell, The effect of environment on the mechanism of
[3] D. Kuhlmann-Wilsdorf, Metal Fiber Brushes, in: P.G. Slade (Ed.), Electrical Stage I fatigue fracture, Metallurgical and Materials Transactions B 2 (1971)
Contacts Principles and Applications, Marcel Dekker, New York, 1999, pp. 1325–1331.
943–1017. [30] H. Shen, S.E. Podlaseck, I.R. Kramer, Effect of vacuum on the fatigue life of
[4] R. Holm, Electric Contacts: Theory and Application, 4th ed., Springer-Verlag, aluminum, Acta Metallurgica 14 (1966) 341–346.
New York, 1967. [31] N.P. Suh, The delamination theory of wear, Wear 25 (1973) 111–124.
[5] L. Boyer, S. Noel, J.P. Chabrerie, Electrochemically activated wear of metal fibre [32] N.P. Suh, An overview of the delamination theory of wear, Wear 44 (1977) 1–16.
brushes, Wear 116 (1987) 43–57. [33] J.R. Fleming, N.P. Suh, Mechanics of crack propagation in delamination wear,
[6] D. DeNardis, D. Rosales-Yeomans, L. Borucki, A. Philipossian, Characterization Wear 44 (1977) 39–56.
of copper–hydrogen peroxide film growth kinetics, Thin Solid Films 513 (2006) [34] J.R. Fleming, N.P. Suh, The relationship between crack propagation rates and
311–318. wear rates, Wear 44 (1977) 57–64.
[7] M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, 2nd ed., [35] S. Jahanmir, N.P. Suh, Mechanics of subsurface void nucleation in delamination
National Association of Corrosion Engineers, Houston, TX, 1974. wear, Wear 44 (1977) 17–38.
[8] N. Argibay, J.A. Bares, J.H. Keith, G.R. Bourne, W.G. Sawyer, Copper–beryllium [36] S. Jahanmir, N.P. Suh, E.P. Abrahamson, Microscopic observations of the wear
metal fiber brushes in high current density sliding electrical contacts, Wear 268 sheet formation by delamination, Wear 28 (1974) 235–249.
(2010) 1230–1236. [37] S. Dillich, D. Kuhlmann-Wilsdorf, Effects of surface films on the performance of
[9] N. Argibay, J.A. Bares, W.G. Sawyer, Asymmetric wear behavior of self-mated silver–graphite (75 wt% Ag, 25 wt% C), Electric Brushes, Components, Hybrids,
copper fiber brush and slip-ring sliding electrical contacts in a humid carbon and manufacturing Technology, IEEE Transactions 3 (1980) 37–41.
dioxide environment, Wear 268 (2010) 455–463.