trib_135_2_021504.pdf

Julian Le Rouzice-mail: [email protected]

Tom Reddyhoffe-mail: [email protected]

Tribology Group,

Department of Mechanical Engineering,

Imperial College,

London SW7 2AZ, United Kingdom

Development of InfraredMicroscopy for MeasuringAsperity Contact TemperaturesSurface temperature measurements within sliding contacts are useful since interfacial heatdissipation is closely linked to tribological behavior. One of the most powerful techniquesfor such measurements is in-contact temperature mapping whereby a sliding contact islocated beneath an infrared microscope. In this approach, one of the specimens must betransparent to infrared and coated such that radiation components can be distinguishedand isolated from background values. Despite its effectiveness, a number of practical con-straints prevent this technique from being applied to rough surfacesa research areawhere temperature maps could provide much needed two-dimension input data to informmixed and boundary friction models. The research described in this paper is aimed atimproving the infrared temperature mapping technique in terms of validity, robustness, andspatial resolution, so that measurements of rough surfaces contacts can be made. First,Plancks law is applied in order to validate the use of surface coating as a means of remov-ing background radiation. Second, a refined method of calibration is put forward andtested, which negates the need for a soft aluminum coating and hence enables rough surfa-ces to be measured. Finally, the use of super-resolution algorithms is assessed in orderextend spatial resolution beyond the current limit of 6 lm. [DOI: 10.1115/1.4023148]

1 IntroductionTemperature in TribologicalContacts

When friction occurs between sliding surfaces, mechanicalenergy is converted into heat [1]. The resulting increase in temper-ature is important since it can influence the behavior of the slidingsystem in a number of ways. For dry interfaces, friction-inducedtemperature rises can cause components to melt [2], crack [3],oxidize [4,5], deform and wear rapidly [6], and change theirmicrostructure [7]. If a liquid lubricant is present between thesurfaces, contact temperatures affect fluid film thickness [8,9],traction [10], and can cause lubricant degradation [11], desorptionof boundary films [12], and the onset of scuffing failure [13].These effects are often severe and result in contact conditions thatdiffer from those which may have been anticipated or desired [1].As a result, many studies have predicted these temperatures ana-lytically [1,1416].In addition to the problems described above, friction-induced

temperature rises can also be useful in providing information toelucidate in-contact behavior since contact temperatures areclosely related to the frictional mechanisms that dissipate heat.For example, contact temperature measurements have been usedto test lubricant rheological models [17,18], and models that pre-dict cutting-tool/work-piece interactions [1921].In order to analyze and control the interfacial behavior

described above, it is necessary to develop methods able to mea-sure accurately the temperature of surfaces within sliding con-tacts. Thermocouples are the most commonly used for thispurpose [2224]; however a number of different experimentaltechniques are possible, as described in the following reviewarticles [25,26]. Particularly useful are those techniques that givetwo-dimensional data, rather than ensemble average values, sincethe phenomena that cause/result from increased temperaturesthemselves vary locally over the interface. For instance, the condi-tions at the contacting high spots, or asperities, differ significantlycompared to conditions taken as average over the contact as a

whole. As Ludema states in his review paper for instance:friction should be represented in some more fundamental andlocally distributed manner than the coefficient of friction [27].Despite its importance, to the authors knowledge, no experi-

mental technique has been developed that is able to map effec-tively the temperature of contacting asperities. A significantbenefit of such measurements is that they can be coupled withmoving heat source theory [18,2831] to obtain the local variationin friction occurring over rough surface contacts. Local frictionvalues, found in this way, would provide invaluable input data forthe many mixed and boundary models that currently relying onestimated values of friction (e.g., [32,33]).The focus of this paper is on developing an accurate and reli-

able experimental technique with sufficient resolution to map thetemperature of contacting asperities. The advances necessary forsuch a measurement would have useful implications for otherareas of technology such as microelectromechanical systems(MEMS), where the asperity contact temperatures affect the elec-trical resistance and play a significant role in radio frequency (rf)MEMS switches [34].

2 Background

2.1 Two-Dimensional Temperature MeasurementTechniques. Thermal measurement techniques with high spatialresolution have been developed and applied to many areas ofphysics, engineering science, and biology. A short review of thesetechniques follows, in which their suitability to measure interfa-cial asperity temperatures is assessed.

Microthermocouples are a low-cost measurement methodwith excellent thermal resolution (0.01K). However, spatialresolution is limited as thermocouple wires are typicallylarger than 15 lm in diameter [35]. Furthermore, imagingwith thermocouples requires the development of complicatedtranslation stages [36].

Scanning thermal microscopy is a near-field technique pro-viding temperature measurements with very high spatial reso-lution. In this case, a setup similar to an atomic forcemicroscope is used, with a small thermocouple or resistive

Contributed by the Tribology Division of ASME for publication in the JOURNALOF TRIBOLOGY. Manuscript received August 29, 2012; final manuscript receivedDecember 2, 2012; published online March 18, 2013. Assoc. Editor: Dong Zhu.

Journal of Tribology APRIL 2013, Vol. 135 / 021504-1CopyrightVC 2013 by ASME

Downloaded From: http://tribology.asmedigitalcollection.asme.org/ on 04/10/2015 Terms of Use: http://asme.org/terms

probe that functions as a temperature sensing tip. Althoughscanning thermal microscopy can achieve spatial resolutionsof 50 nm and thermal resolutions of 0.1K [37], a number oflimitations arise. First, measurement procedures are complexand time consuming. Second, coupling between thermal andsurface roughness effects make tip/sample interactions diffi-cult to interpret. Finally, due to the close proximity requiredbetween probe and surface, interfaces cannot be accessed. Areview of near-field techniques can be found in Ref. [38].

Near-field scanning optical microscopy (NSOM) is anadvanced technique in which a sample is illuminated by lightfrom a tip with a 50 nm aperture. This scanning tip thendetects the resulting, exponentially decreasing, evanescentwaves that are emitted from the sample surface and changesin reflectivity are processed to obtain temperature. Similar toother near-field techniques, NSOM resolution is excellent(20 nm); however acquisition rates are very slow and interfa-cial measurements are not possible. Furthermore, the tip ge-ometry can change with surface heating and distortmeasurements, while the short tip/sample distance can causeprobe damage [39].

Raman spectroscopy can also be used to map surface temper-ature. Here a sample is illuminated with a laser and the differ-ence in energy between incident and scattered photons ismeasured by a shift in frequency. This effect is attributed tothe inelastic scatterings and energy exchange in the samplephenomena, which are sensitive to temperature. By monitor-ing the frequency shift and correlating it to temperature, ther-mal images with spatial resolutions of less than 1 lm arepossible. However, temperature resolution is generally poor(between 5 and 20K) [40].

Fluorescent thermography techniques exploit the temperaturedependence of fluorescence quantum efficiency. This requiresa fluorescent material to be incorporated into the sample thatis then illuminated with UV light. In this way it is possible toobtain a thermal image of the sample with an excellent ther-mal resolution (0.01K) and a spatial resolution of a fewmicrometers [41]. This approach is potentially very useful foranalyzing fluid lubricated contacts; however the requirementfor a fluorescer probably precludes its application to roughsurface contact measurements.

Interferometry can also be used to image surface tempera-tures. Here two lasers beams are focused on a sample using aMichelson interferometer. The measured dynamic deforma-tions of the surface can then be used to determine the surfacetemperature. This technique is limited to reflective samples;however spatial resolutions of 1 lm are possible [42].

Thermoreflectance is a method that utilizes the change in re-fractive index, and hence reflectivity of a material that occurswith temperature. Spatial and thermal resolution as low as0.3 lm and 0.01K, respectively, have been reported [43,44].Although thermoreflectance has excellent spatial and thermalresolution, it is unlikely that this method could be used tostudy rough surfaces.

Liquid crystal thermography uses a thin layer of liquid crystaldeposited on a sample that is then illuminated with whitelight. The resulting change of color can be correlated to thetemperature producing thermal images with spatial and ther-mal resolution of 2 lm and 0.5K, respectively [45].

Infrared (IR) thermography is currently the most used tempera-ture mapping technique due to recent advances in camera sen-sitivity and reductions in cost. This approach uses a camerawith a detector sensitive to IR light to collect thermal radiationof the sample of known emissivity, which according toPlancks law, can be related to the temperature [46]. Typicallysuch detectors are sensitive to wavelengths between 3 and 10lm, which limit spatial resolution to around these values. Forhot samples (>500K), it is possible to use a CCD camera todetect near-infrared light and hence obtain submicron spatialresolution [47]. Being noncontacting, sensitive to temperature

fluctuations down to around 0.01K [48], and robust, IR ther-mography is a particularly suitable means of monitoring inter-facial temperatures. There are two main limitations associatedwith this technique. First, one of the two rubbing specimensmust be transparent to IR in order for the camera to view thecontact. The second limitation is that the spatial resolution isusually limited to the wavelength of IR. The use of a transpar-ent window is unavoidable with this technique; however thelimited spatial resolution is addressed in this study.

2.2 Development of Infrared Thermography forTribological Applications. The first use of an IR thermographyin tribology was proposed in the 1970 s by Winer and co-workersto study elastohydrodynamic lubrication (EHL) [4951]. Sincefull field measurements were not possible, a point measuringmicroscope was scanned over a contact formed between a steelball and sapphire disk in order to map the balls temperature. Adecade later, Keping and Shizhu used a similar technique, theresults of which they compared to numerical predictions of filmtemperature rise [52,53]. Spikes and co-workers then advancedthe technique by applying coatings to the disk in order to measurethe temperature of both specimen surfaces separated by the fluidfilm [54] and also applied moving source theory to calculate shearstress maps from the measured temperatures [2931]. Followingthis, Yagi et al. used an IR camera to measure the temperature risein an EHL contact associated with rough surface features [55],dimple formation [56], and the effects of slip ratio [57].Recently, Reddyhoff et al. incorporated a high specification IR

camera and custom-built microscope lens into their experimentalsetup in order to map contact temperatures down to 0.01K [18]with a diffraction-limited spatial resolution of 6 lm [58,59]. Usingthe approach put forward by Spikes et al., they calculated in-contact shear stresses from measured temperature and so showeddifferences between lubricants which were not evident when aver-aging friction measurement were made.As described above, the majority of infrared thermography

experiments have been restricted to liquid-lubricated contacts.This has been because surface coatings can be used without beingworn away and contact sizes are large compared to the techni-ques resolution. There are some exceptions where maps of inter-facial temperatures in dry contacts have been obtained. Mostnotably, Quinn and Winer used a camera with photographic filmto monitor hotspots resulting from asperity contact between slid-ing specimens [60]. Although this method did not allow for thequantitative measurement of temperatures, useful observationsregarding the size, number, and transient nature of the hotspotswere reported. More recent research on hotspots has focused ontheir occurrence in disk brakes [61,62]. Hotspots typically resultfrom thermoelastic instability, whereby positive feedback betweenfrictional heating thermal expansion leads to very high tempera-tures and large contact areasconditions well suited to IR ther-mography. Very few attempts have been made to measure thetemperature of asperities in stable conditions; however the inter-pretation of thermal images was hampered by insufficient spatialresolution and unmeasureable variations in emissivity [62,63].Both of these limitations are addressed in this paper.Despite the advances in IR imaging, reliable experimental tem-

perature measurements of asperity contacts, and hence frictiondata, are not available. The aim of this paper is to extend theexperiment technique reported by Reddyhoff et al. [58] with thegoal of making such measurements possible. To do this, threeissues are addressed.

(i) First, Plancks law is applied to existing experimental data toassess the validity of the procedure in which radiation com-ponents are isolated and background radiation is removed.

(ii) Despites its proven effectiveness in smooth, liquid lubri-cated, contacts, the current method of use of an aluminumcoating is not viable since it is worn away when roughsurfaces are tested. Therefore, the second part of the paper

021504-2 / Vol. 135, APRIL 2013 Transactions of the ASME


tests a new calibration procedure that allows backgroundradiation to be isolated and removed without the use of analuminum coating.

(iii) The current spatial resolution of the IR camera, with the5 objective lens, is close to 6 lm. Although this resolu-tion is close to the diffraction limit of IR light, it is insuffi-cient to map asperities with contact dimensions close to12 lm. To address this, the final section of the paperdescribes the use of super-resolution algorithms that areapplied to IR images in order to increase spatial resolutionbeyond 6 lm.

3 Description of Technique

The sliding contact used to assess the IR mapping techniquewas produced by loading a steel ball or roller against a sapphiredisk using a conventional EHL rig (manufactured by PCS Instru-ments, Acton UK) as shown in Fig. 1. In all tests, the ball or rollerused was 19.05mm in diameter, made from AISI 52,100 steelwith an Ra roughness

4 Results and Discussion

4.1 Theoretical Validation of Measurement. The use ofsurface coatings, described above, is aimed at distinguishingbetween different components of radiation. To test the effective ofthe approach, this section describes how Plancks law of thermalradiation can be applied to measured data. Plancks law quantifiesthe spectral radiance of a blackbody (in units of W/m2 sr1 m1)as a function of its temperature [64]:

BkT 2hc2

k51

exphc

kkBT

1

(3)

where k is the wavelength of emission, c is the speed of light, h isPlancks constant, and kB is the Boltzmann constant.Then, if the energy of a single photon E is considered:

E hck

(4)

Dividing Eq. (3) by Eq. (4) gives the radiance of emitted light inunits of photons/s m2 sr1 m1:

bkT 2c

k41

exphc

kkBT

1

(5)

This expression can then be used to estimate the number ofcounts (i.e., photons) that the infrared camera should record whenfocused on a sample of temperature T. However, a number of pa-rameters need to be introduced before this calculation can bemade. These are the surface area of the sample A, the emissivityof the sample e (constant with k, i.e., assuming specimen is a graybody), the solid angle X (i.e., the spatial fraction of emitted lightcollected by the objective lens), a transmission coefficient F (toaccount for absorption by components of the system such as theobjective lens), the quantum efficiency of the detector g (numberof e/holes created for each incident photon), the spectral sensi-tivity of the detector defined by upper and lower wavelength lim-its k1 and k2, and the cameras integration time ti (IR equivalentto shutter speed). Once these coefficients have been included, Eq.(6) can be integrated between k1 and k2 to give the number ofcounts C, recorded by each pixel of the camera during one inte-gration time:

C AeXFgti2ck2k1

dk

k4 exphc

kkBT

1

(6)

In order to apply this equation to experimental data, valuesmust be found for the coefficients defined above. Specifically, thesurface area of one pixel A 6:3 106 2 (m2) (obtained fromcamera specifications), emissivity of steel specimen e 0:4 [65],transmission coefficient F 0:05 (this is an approximate estimatebased on the number of lenses in the system), quantum efficiencyg 0:9 photons1 (obtained from camera specifications), spectralsensitivity k1 3 106m and k2 5 106m, and integrationtime ti 6 103s. Finally, the solid angle X can be estimatedfrom the focal distance f 2 104 m and the radius of theobjective lens R 2 104 m using the following two equations:

h tan1 Rf

(7)

X 2ph0

sin h0dh0 (8)

Values were substituted into Eq. (6), which was then solvednumerically using Matlab for a surface temperature of 293K, togive a theoretical estimate of the number of the camera reading ofCth 2:54 105 counts. This is substantially different from theactual measured value. As shown by Fig. 3, approximately 5000counts are recorded by the camera during a calibration in whichthe specimens are heated to 293K.This discrepancy is unsurprising considering the number of

estimated parameters required in Eq. (6); with the most likelysources of error being an overestimation of the transmission coef-ficient and quantum efficiency of the system (due to the age of theequipment and contamination of lenses). If however each set ofresults are normalized by their maximum value, theory andexperiment can be compared, as shown in Fig. 4. This is a validapproach since it removes the discrepancy due to poorly estimatedcoefficients to leave only spectral variation in actual and predictedmeasurements.

Fig. 3 Single-pixel camera counts, as a function of tempera-ture, from an interface between a steel ball and an uncoated(Un), a chromium coated (Cr), and an aluminum (Al) coated disk

Fig. 4 Normalized single-pixel camera counts, as a function oftemperature, from an interface between a steel ball and anuncoated (Un), a chromium coated (Cr), and an aluminum (Al)coated disk



As expected, Fig. 4 shows that the raw experimental measure-ments do not follow Plancks law. This is to be expected sinceeach of the measurements are contaminated with the spectrallycomplex background radiation from the bulk of the sapphire disk(i.e., the gray body assumption is not valid). If however, thisunwanted background radiation is removed by subtracting the Alspecimen radiation from that of the uncoated and Cr specimens[i.e., if Eqs. (2a) and (2b) are applied] excellent agreement withPlancks theory is found, as illustrated by Fig. 5.Several conclusions can be drawn from this result. First, it vali-

dates the use of coatings as means to isolate radiation componentsand measure the temperature of surfaces within a sliding interface.Conversely, it shows that experimental methods that do not isolatebackground radiation are not obtaining correct interfacial temper-atures. It also shows that, under these conditions, both ball anddisk specimens can be safely considered as gray bodies. Conse-quently, it can be noticed that the ratio of radiation componentsUn Al=Cr Al is approximately equal to the ratio of speci-men emissivities esteel=eCr 2:5.

4.2 Reliable Measurement Without Al Coating. The aimof this section of work is to develop an approach that uses thesame principles of radiation component isolation validated abovebut which negates the requirement to test with a soft aluminumcoating. The most obvious way to achieve this is to obtain calibra-tion curves, using only an uncoated and a Cr coated disk, thatdirectly relate camera counts to surface temperature. Here therelationship for the disk surface temperature, obtained from thepure rolling calibration, is given by

Tcalibrationbath g1CCr pure rolling (9)

This is equivalent to Eq. (1a), however no Al is used (and thecorresponding ball calibration equation has been omitted for brev-ity). Again, pure rolling conditions are used in calibration to pro-duce negligible frictional heating; so that surface temperaturescan be controlled by heating of the lubricant bath. Once functionsg1 and g2 are obtained, they are intended to be used to calculatesurface temperatures during subsequent sliding tests:

Tdisk g1CCr sliding (10)The approach assumes that the background radiation from the

bulk of the disk is the same under steady heating during calibration

as it is when locally heated by a sliding contact. This calibrationmethod was tested by applying it, along with the establishedapproach, to a set of radiation measurements. These measurementswere obtained from a contact between a steel ball and sapphiredisk, lubricated with Santotrac50, loaded with 20N, at a range ofentrainment speeds with a slide-roll ratio of 1.0. The resulting max-imum contact temperatures from each method are shown in Fig. 6,where there is clear disagreement between the two. This shows thatthe Cr-only calibration must be incorrect since the original methodhas been validated in a previous study [18]. The difference betweenthe sets of values show that, for the same contact temperature, moreradiation is emitted from the bulk of the disk during steady stateheating than during a sliding test. During the calibration, the wholesystem (oil, ball, disk, rig, etc.) are held at a constant temperature,while during a sliding test the temperature rise occurs only withinthe contact.To solve this problem a two-step calibration procedure is pro-

posed. In the first step, the radiation from the contact is monitoredfor both Al and Cr coated disks, under pure rolling conditions, asthe specimens are steadily heated using the temperature controlledbath. This is a standard calibration and, as described in Sec. 3,provides a relationship between contact temperature and the radia-tion component from the disk surface:

Tcalibrationbath f1CCr pure rolling CAl pure rolling (11)

The second calibration step involves keeping the temperatureof the lubricant bath constant while increasing the sliding speed ofthe contact. As the sliding speed increases, frictional heatingcauses the contact temperature to rise. Radiation from both the Aland the Cr disks are recorded in the test and correlated to providea relationship between background and surface radiation compo-nents under actual test conditions:

CAl sliding f2CCr sliding (12)

Once functions f1 and f2 are known, they can be used in subse-quent experiments to calculate the temperature of the disk by thefollowing equation:

Tdisk f1 CCr sliding f2CCr sliding

(13)

This equation can be used in experiments to calculate contact tem-perature from recorded radiation values; it accounts for back-ground radiation and does not require the use of an aluminum

Fig. 5 Normalized single-pixel camera counts, as a function oftemperature, from an interface between a steel ball and anuncoated (Un) and a chromium coated (Cr), with subtraction ofcamera counts against an aluminum coated disk (Al)

Fig. 6 Maximum contact temperature versus speed for a con-tact between ball and disk, lubricated with Santotrac50. Thecontact is loaded with 20N load and a slide-roll ratio of 1.0 isapplied. The temperatures shown have been calculated usingeither Eq. (2a) or (10).

Journal of Tribology APRIL 2013, Vol. 135 / 021504-5


coating. In order to test this new approach, functions f1 and f2were determined (using the pure rolling and the sliding calibrationsteps described above). Then Eq. (13) was applied to data from aseparate sliding test (for a steel roller sliding against sapphiredisk). Figure 7 compares the calculated temperature found in thisway with that obtained using the established approach. This timethere is good agreement between approaches. Furthermore, the setof values obtained using the new Cr-only method show less noise,which is probably due to effects of the worn Al coating not beingpresent. This is an important result as it shows that interfacial tem-peratures can be measured accurately without using an Al coating.This approach can also be used with an uncoated disk in order toobtain the temperature of the ball specimen.

4.3 Super Resolution Algorithms. The current experimentalsetup, involving an IR camera (Flir Phoenix 9705) combined witha 5 IR objective lens, gives a spatial resolution of 6.3 lm/pixel.Since this value is close to the wavelength of IR it represents the

maximum resolution which can be achieved with such a camera(scanning techniques described in Sec. 2.1 have higher spatial re-solution but are unable to probe sliding interfaces). However, 6.3lm/pixel is insufficient to differentiate the flash temperaturescaused by asperity contacts, which are of the order of 1 lm.With the aim of improving spatial resolution sufficiently to map

asperity temperatures, the use of super-resolution computing [66]is now investigated. Super-resolution processes are those in whichseveral low resolution (LR) images are combined in such a wayso as to produce a single high resolution (HR) image. For thiscomputation to function effectively, the set of LR images musthave been obtained with a range of different lateral displacementsbetween the camera and subject. In this way, the LR images con-tain sufficient information to build a HR image. The super-resolution (SR) process is outlined in Fig. 8 and consists of twodistinct steps. The first step is to estimate the displacements andthe rotations between the different LR imagesa process knownas motion recognition, for which a number of algorithms can beused [67]. The second step involves superimposing the alignedLR images onto a HR grid to produce a reconstructed HR imageof the subject. A range of algorithms are also available for this lat-ter image reconstruction process [67]. The few examples of thistechnique being applied to IR thermography data include Saka-gami et al., who used it to increase resolution to detect cracks instructures [68], and Teyssieux et al., to enhance thermal detectionwith CCD camera thermography [69].In order to assess whether SR algorithms can tackle asperity

contacts that are poorly defined in their shape distribution, it isnecessary to validate the technique using small scale surface fea-tures of known geometry. To do this, we focused the IR apparatusonto the surface of photolithography masks (normally used in sili-con fabrication). The surfaces of these specimens consist of a reg-ular repeating pattern of lines or dots, with dimensions close tothe current resolution limit of the camera/lens. Once the camerawas focused, a range of small displacements (approximately

Two examples follow, which demonstrate the effectiveness ofthe SR approach. As shown in Fig. 9, the surface of the first sam-ple consists of repeated lines of 20 and 40 lm in thickness. A LRimage of this sample is displayed in Fig. 10, alongside a HRimage obtained by applying the robust super resolution algo-rithm [70] to 50 such LR images.It is evident that the HR image in Fig. 10 has better resolution

than the LR image. Specifically, it can be seen that aliasing isreduced on the edges of the lines in the HR image. This compari-son however is hindered by the fact that the lines were alreadyeasily detected on the LR image. In order to further test the tech-nique, samples with smaller details were used.As shown in Fig. 11, the features on the second photolithogra-

phy specimen consisted of a lattice of a lattice of 10 lm diameterdots spaced 30 lm apart. A LR image of this sample is displayedin Fig. 12(a), alongside two HR images obtained by applying dif-ferent SR algorithms; namely the fast robust [71] and theinterpolation (bicubic interpolation with Matlab built-in func-

tion) [72] algorithms shown in Figs. 12(c) and 12(d), respectively.Figure 12(b) shows a simple interpolation of a LR image obtainedusing Matlab software (i.e., a single LR image interpolated line-arly onto a HR grid).The difference in quality between LR and HR images in Fig. 12

is very obvious, with the dots on the LR image being barely visi-ble, whereas in the HR images they appear clearly. It can also beseen that the simple linear Matlab interpolation of a LR image isof poorer quality than those obtained using the super resolution. Itshould be noted that the pixel values in LR and HR images aresimilar (Fig. 13). This observation shows that increasing the reso-lution with this technique does not distort the data. This is critical

Fig. 11 Geometry of surface pattern on photolithographyspecimen (a lattice of 30 lm spaced dots of 10 lm diameter)

Fig. 12 (a) LR image. (b) LR image with simple linear interpolation. (c) HR image with fast robustsuper resolution of 50 images. (d) HR image with interpolation super resolution of 50 images.

Fig. 13 Pixel values for a line across the LR and HR infraredimages shown in Figs. 12(a) and 12(b)



since the measured pixels values should correspond to actual con-tact temperatures.To study and improve the behavior of the SR technique further,

it is necessary to devise a means of quantifying the resolution ofan image (i.e., to give a measure of how faithful an image is to theactual geometry of the specimen surface). For this, a circularHough transform [73] was applied that detects the number and di-ameter of circles in an image, as illustrated by Fig. 14 (resolutionis measured by how close the diameter, predicted by Hough trans-form, is to that of the actual surface features).It is now possible to study a relationship between the number of

LR images used in an SR algorithm and the resulting improve-ment in resolution. To do this, the robust and interpolation SRalgorithms were applied to a varying number of LR input imagesand the resulting HR images analyzed using the Hugh transform.These results are summarized in Fig. 15 revealing a number ofobservations. First, only 5 to 10 LR images are required beforethe SR output stabilizes. Second, the predicted circle diameter of10.4 lm is very close to the actual diameter 10 lm, again showingthe accuracy of this method. It can also be noticed that no circle isdetected for a single LR image.The improved calibration and resolution of infrared microscopy

outlined in this paper should enable the measurement of interfacialtemperatures between rough surfaces. Before this can be done how-ever, a number of experimental practicalities should be considered,such as how to obtain a set of low resolution images under each testcondition. The most obvious way is to load a stationary rough ballspecimen against a sliding sapphire disk. A shaker can then be usedto apply a small displacement to the camera relative to the contact.Alternatively, the ball could be rotated in order to enable rolling/sliding conditions. This would require the camera to be triggered toalways record images of the same portion of the rough specimen.

5 Conclusions

Infrared microscopy is an effective tool for mapping interfacialcontact temperatures and its use is becoming more widespreaddue to increasing camera sensitivity and falling costs. Moreover,by applying coatings to the transparent specimen, it is possible tomeasure accurately both of the surfaces within the contact.To date, two significant restrictions have prevented the applica-

tion of IR microscopy to rough surface contacts; namely the dif-fraction limited spatial resolution and the requirement for an Alcoated specimen. The current study has put forward and testedrefinements to the calibration approach that negate the require-ment for an Al coated specimen. This offers the possibility toaccurately measure unlubricated and rough interfaces for the firsttime. Sufficient detail has been provided that these refinementscan be adopted by other researchers using this technique. It hasalso been shown that IR microscopy resolution can be increasedby applying super resolution algorithms to recorded images. Pre-viously invisible details, smaller than 10 lm in diameter, wereenhanced and detected, without distortion. These advances makethe IR microscopy a more powerful tool in tribology and pave theway for temperature mapping of contacting asperities.

Acknowledgment

The authors are very grateful to EURAMET for supporting thisproject EMRP Researcher Grant IND11-REG1 MADES associ-ated with the project Metrology to assess durability and functionof engineered surfaces.

Nomenclature

A surface area of a pixelBk spectral radiance of a black body

c light velocityC camera countsE energy of a photonf lens focal distance

f1, f2, g1, g2 functions relating temperature to camera countsF transmission coefficienth Plancks constant

kB Boltzmanns constantRa roughness of the steel ball

ti integration timeT temperaturebk spectral radiance in photons of a black bodye emissivityg quantum efficiency of InSb detectorh half angle between sample and lensk wavelength

k1, k2 lower and upper wavelengths of the sensitivityrange of detector

X solid angle between sample and lens

References[1] Kennedy, F. E., 1984, Thermal and Thermomechanical Effects in Dry

Sliding, Wear, 100, pp. 453476.[2] Carignan, F. J., and Rabinowicz, E., 1980, Friction and Wear at High Sliding

Speeds, ASLE Trans., 23, pp. 451459.[3] Dow, T. A., 1980, Thermoelastic Effects in Brakes, Wear, 59, pp. 213221.[4] Quinn, T. F. J., 1983, Review of Oxidational Wear, Part I, Tribol. Int., 16, pp.

257271.[5] Quinn, T. F. J., 1983, Review of Oxidational Wear, Part II, Tribol. Int., 16,

pp. 305315.[6] Chen, C. P., and Burton, R. A., 1980, Thermoelastic Effects in Brushes With

High Current and High Sliding Speeds, Wear, 59, pp 277288.[7] Bill, C. B., and Wisander, D., 1977, Recrystallization As a Controlling Process

in the Wear of Some F.C.C. Metals, Wear, 41, pp. 351363.[8] Crook, A. W., 1961, The Lubrication of Rollers III. A Theoretical Discussion

of Friction and the Temperatures in the Oil Film, Philos. Trans. R. Soc. Lon-don, 254, pp. 237258.

[9] Hili, J., Olver, A. V., Edwards, S., and Jacobs, L., 2010, Experimental Investi-gation of Elastohydrodynamic (EHD) Film Thickness Behaviour at HighSpeeds, Trib. Trans., 53, pp. 658666.

Fig. 14 (a) Circles detection of on HR image by circular Houghtransform. (b) High magnification of (a).

Fig. 15 Number of circles detected and their average diameterversus the number of LR images used for (a) and (b) fast robustSR and (c) and (d) interpolation SR algorithms



[10] Olver, A. V., and Spikes, H. A., 2001. Prediction of Traction in Elastohydro-dynamic Lubrication, Proc. Inst. Mech. Eng., 215, pp. 309310.

[11] Santos, J. C. O., Garcia dos Santos, I. M., Souza, A. G., Sobrinho, E. V., Fer-nandes, Jr., V. J., and Silva, A. J. N., 2004, Thermoanalytical and RheologicalCharacterization of Automotive Mineral Lubricants After Thermal Degrada-tion, Fuel, 83, pp. 23932399.

[12] Grew, W. J. S., and Cameron, A., 1972. Thermodynamics of Boundary Lubri-cation and Scuffing, Proc. R. Soc. London, 327, pp. 4759.

[13] Enthoven, J. C., Cann, P. M., and Spikes, H. A., 1993, Temperature andScuffing, Trib. Trans., 32, pp. 277288.

[14] Archard, J. F., 1959, The Temperature of Rubbing Surfaces, Wear, 2, pp.438455.

[15] Dayson, C., 1967, Surface Temperatures at Unlubricated Sliding Contacts,Trans. ASLE, 10, pp. 169174.

[16] Ling, F. F., 1969, On Temperature Transients in Sliding Interface, ASME J.Lubr. Tech., 91, pp. 397405.

[17] Clarke, A., Sharif, K. J., Evans, H. P., and Snidle, R. W., 2006, Heat Partitionin Rolling/Sliding Elastohydrodynamic Contacts, ASME J. Tribol., 128, pp.6778.

[18] Reddyhoff, T., Spikes, H. A., and Olver, A. V., 2009, Compression Heatingand Cooling in Elastohydrodynamic Contacts, Tribol. Lett., 36, pp. 6980.

[19] Biermann, D., and Schneider, M., 1997, Modeling and Simulation of Work-piece Temperature in Grinding by Finite Element Analysis, Mach. Sci. Tech-nol., 1, pp. 173183.

[20] Anderson, D., Warkentin, A., and Bauer, R., 2008, Experimental Validation ofNumerical Thermal Models for Shallow and Deep Dry Grinding, J. Mater.Process. Tech., 204, pp. 269278.

[21] Dinc, C., Lazoglu, I., and Serpenguzel, A., 2008, Analysis of Thermal Fieldsin Orthogonal Machining With Infrared Imaging, J. Mater. Process. Technol.,198, pp. 147154.

[22] Kennedy, F. E., Frusescu, D., and Li, J., 1997, Thin Film ThermocoupleArrays for Sliding Surface Temperature Measurement, Wear, 207, pp. 4654.

[23] Tian, X., and Kennedy, F. E., 1995, Prediction and Measurement of SurfaceTemperature Rise at the Contact Interface for Oscillatory Sliding, Proc. Inst.Mech. Eng. J J. Eng. Tribol., 209, pp. 4151.

[24] Qiu, M., Zhang, Y. Z., Shangguan, B., Du, S. M., and Yan, Z. W., 2007, TheRelationships Between Tribological Behaviour and Heat-Transfer Capability ofTi6Al4V Alloys, Wear, 263, pp. 653657.

[25] Kalin, M., 2004, Influence of Flash Temperatures on the Tribological Behav-iour in Low-Speed Sliding: A Review, J. MSEA A, 374, pp. 390397.

[26] Komanduri, R., and Hou, Z. B., 2001, A Review of the Experimental Techni-ques for the Measurement of Heat and Temperatures Generated in Some Manu-facturing Processes and Tribology, Tribol. Int., 34, pp. 653682.

[27] Meng, H. C., and Ludema, K. C., 1995, Wear Models and Predictive Equa-tions: Their Form and Content, Wear, 181183, pp. 443457.

[28] Jaeger, J. C., 1942, Moving Sources of Heat and the Temperatures at SlidingContacts, Proc. R. Soc. NSW, 76, pp. 203224.

[29] Cann, P. M., and Spikes, H. A., 1989, Determination of the Shear Stresses ofLubricants in Elastohydrodynamic Contacts, Tribol. Trans., 32, pp. 414422.

[30] Glovnea, R., and Spikes, H. A., 1995, Mapping Shear Stress in Elastohydrody-namic Contacts, Tribol. Trans., 38, pp. 932940.

[31] Grieve, R. S. A., and Spikes, H. A., 2000, Temperature and Shear Stress inRolling Sliding Elastohydrodynamic Contacts, Proceedings of the Leeds-Lyon Symposium on Thinning films and Tribological Interfaces Leeds, pp.512522.

[32] Kadiric, A., Sayles, R. S., and Ioannides, E., 2008, Thermo-Mechanical Modelfor Moving Layered Rough Surface Contacts, ASME J. Tribol., 130, pp. 114.

[33] Andersson, J., Larsson, R., Almqvist, A., Grahn, M., and Minami I., 2012,Semi-deterministic Chemo-Mechanical Model of Boundary Lubrication, Far-aday Discuss., 156, pp. 343360.

[34] Brown, C., Rezvanian, O., Zikry, M. A., and Krim, J., 2009, Temperature De-pendence of Asperity Contact and Contact Resistance in Gold RF MEMSSwitches, J. Micromech. Microeng., 19, p. 025006.

[35] Christofferson, J., Maize, K., Ezzahri, Y., Shabani, J., Wang, X., and Shakouri,A., 2008, Microscale and Nanoscale Thermal Characterization Techniques,ASME J. Electron. Packag., 130, p. 041101.

[36] Ho, H. P., Lo, K. C., and Wu, S. Y., 2001, A Scanning ThermocoupleProbe for Temperature Mapping, IEEE Trans. Instrum. Meas., 50, pp.11671170.

[37] Shi, L., Kwon, O., Miner, A. C., and Majumdar, A., 2001, Design and BatchFabrication of Probes for Sub-100 nm Scanning Thermal Microscopy, J.Microelectromech. Syst., 10, pp. 370378.

[38] Pollock, H. M., and Hammiche, A., 2001, Micro-thermal Analysis: Techniquesand Applications, J. Phys. D Appl. Phys., 34, pp. R23R53.

[39] Vertikov, A., Kuball, M., Nurmikko, A. V., and Maris, H. J., 1996, Time-Resolved Pump-Probe Experiments With Subwavelength Lateral Resolution,Appl. Phys. Lett., 69, pp. 24652467.

[40] LaPlant, F., Laurence, G., and Ben-Amotz, D., 1996, Theoretical and Experi-mental Uncertainty in Temperature Measurement of Materials by RamanSpectroscopy, Appl. Spectrosc., 50, pp. 10341038.

[41] Kolodner, P., and Tyson, J. A., 1982, Microscopic Fluorescent Imaging of Sur-face Temperature Profiles With 0.01 C Resolution, Appl. Phys. Lett., 40, pp.782784.

[42] Claeys, W., Dilhaire, S., Jorez, S., and Patino-Lopez, L. D., 2001, LaserProbes for the Thermal and Thermomechanical Characterisation of Microelec-tronic Devices, Microelectron. J., 32, pp. 891898.

[43] Christofferson, J., and Shakouri, A., 2004, Thermal Measurements of ActiveSemiconductor Micro-structures Acquired Through the Substrate Using NearIR Thermoreflectance, Microelectron. J., 35, pp. 791796.

[44] Farzaneh, M., Maize, K., Lueren, D., Summers, J. A., Mayer, P. M., Raad, P.E., Pipe, K. P., Shakouri, A., Ram, R. J., and Hudgings, J. A., 2009. CCD-Based Thermoreflectance Microscopy: Principles and Applications, J. Phys. DAppl. Phys., 42, p. 143001.

[45] Csendes, A., Szekely, V., and Rencz, M., 1996, Thermal Mapping With LiquidCrystal Method, Microelectron. Eng., 31, pp. 281290.

[46] Liu, W., and Yang, B., 2007, Thermography Techniques for Integrated Cir-cuits and Semiconductor Devices, Sensor Rev., 27, pp. 298309.

[47] Teyssieux, D., Thiery, L., and Cretin, B., 2007, Near-Infrared ThermographyUsing a Charge-Coupled Device Camera: Application to Microsystems, Rev.Sci. Instrum., 78, p. 034902.

[48] Schulz, M., Gross, W., and Scheuerpflug, H., 2000, High-Resolution Thermo-physical Measurements Using Staring Infrared Detector Arrays, High. Temp.-High Press., 32, pp. 547556.

[49] Sanborn, D. M., and Winer, W. O., 1971, Fluid Rheological Effects in SlidingElastohydrodynamic Point Contacts, ASME J. Tribol., 93, pp. 262271.

[50] Turchina, V., Sanborn, D. M., and Winer, W. O., 1973, Temperature Measure-ments in Sliding Elastohydrodynamic Point Contacts, ASME J. Lubr. Tech-nol., 96, pp. 464471.

[51] Ausherman, V. K., Nagaraj, H. S., Sanborn, D. M., and Winer, W. O., 1976,Infrared Temperature Mapping in Elastohydrodynamic Lubrication, ASME J.Lubr. Technol., 98, pp. 236243.

[52] Keping, H., and Shizhu, W., 1988, Temperature Measurement in Elastohydro-dynamic Lubrication Contacts Using an Infrared Technique, Tribol. Int., 21,pp. 287289.

[53] Keping, H., and Shizhu, W., 1991, Analysis of Maximum Temperature forThermoelastohydrodynamic Lubrication in Point Contacts, Wear, 150, pp.110.

[54] Spikes, H. A., Anghel, V., and Glovnea, R., 2004, Measurement of the Rheol-ogy of Lubricant Films in Elastohydrodynamic Contacts, Tribol. Lett., 17, pp.593605.

[55] Yagi, K., Kyogoku, K., and Nakahara, T., 2006, Measurements of Tempera-ture Distributions Around Longitudinally Grooved Rough Surfaces in SlidingElastohydrodynamic Point Contacts, Tribol. Trans., 49, pp. 282289.

[56] Yagi, K., Kyogoku, K., and Nakahara, T., 2005, Relationship Between Tem-perature Distribution in EHL Film and Dimple Formation, ASME J. Tribol.,127, pp. 658665.

[57] Nakahara, T., and Yagi, K., 2007, Influence of Temperature Distributions inEHL Film on Its Thickness Under High Slip Ratio Conditions, Tribol. Int., 40,pp. 632637.

[58] Reddyhoff, T., Spikes, H. A., and Olver, A. V., 2009, Improved Infrared Tem-perature Mapping of EHL Contacts, Proc. Inst. Mech. Eng. J J. Eng., 223, pp.11651177.

[59] Ingram, M., Reddyhoff, T., and Spikes, H. A., 2010, Thermal Behaviour of aSlipping Wet Clutch Contact, Tribol. Lett., 42, pp. 2332.

[60] Quinn, T. F. J., and Winer, W., 1987, An Experimental Study of the Hot-SpotsOccurring During the Oxidational Wear of Tool Steel on Sapphire, ASME J.Tribol., 109, pp. 315320.

[61] Majcherczak, D., Dufrenoy, P., and Berthier, Y., 2007, Tribological, Thermaland Mechanical Coupling Aspects of the Dry Sliding Contact, Tribol. Int., 40,pp. 834843.

[62] Panier, S., Dufrenoy, P., and Weichert, D., 2004, An Experimental Investiga-tion of Hot Spots in Railway Disc Brakes, Wear, 256, pp. 764773.

[63] Sudipto, R., and Chowdhury S. K. R., 2011, Prediction of Contact SurfaceTemperature Between Rough Sliding BodiesNumerical Analysis andExperiments, Ind. Lub. Tribol., 63, pp. 327343.

[64] Vollmer, M., and Mollman, K. P., 2010, Infrared Thermal Imaging: Fundamen-tals, Research, and Applications, Wiley VHC, Weinheim.

[65] Nayer, A., 1997, The Metals Data Book,. McGraw-Hill, New York.[66] Borman, S., and Stevenson, R. L., 1998, Superresolution From Image

SequencesA Review, Proc. Midwest Symp. Circuits and Systems (IEEE),pp. 374378.

[67] http://lcav.epfl.ch/software/superresolution[68] Sakagami, T., Matsumoto, T., Kubo, S., and Sato, D., 2009, Nondestructive

Testing by Super-Resolution Infrared Thermography, Proc. SPIE OrlandoThermosense XXXI 7299 72990V.

[69] Teyssieux, D., Euphrasie, S., and Cretin, B., 2009, Thermal DetectivityEnhancement of Visible and Near Infrared Thermography by Using Super-Resolution Algorithm: Possibility to Generalize the Method to Other Domains,J. Appl. Phys., 105, p. 064911.

[70] Zomet, A., Rav-Acha, A., and Peleg, S., 2001, Robust Super-Resolution, Pro-ceedings International Conference on Computer Vision and Pattern Recognition(CVPR).

[71] Farsiu, S., Robinson, M. D., Elad, M., and Milanfar P., 2004, Fast and RobustMultiframe Super Resolution, IEEE Trans. Image Process., 13, pp. 13271344.

[72] http://www.mathworks.co.uk/help/techdoc/ref/griddata.html[73] http://www.mathworks.co.uk/help/toolbox/images/ref/hough.html



s1s2s2Als2Bs3E1aE1bE2aE2bF1F2s4s4AE3E4E5E6E7E8F3F4s4BE9E10E11E12E13F5F6s4CF7F8F9F10F11F12F13s5B1B2B3B4B5B6B7B8B9F14F15B10B11B12B13B14B15B16B17B18B19B20B21B22B23B24B25B26B27B28B29B30B31B32B33B34B35B36B37B38B39B40B41B42B43B44B45B46B47B48B49B50B51B52B53B54B55B56B57B58B59B60B61B62B63B64B65B66B67B68B69B70B71B72B73