lable at ScienceDirect

Journal of Nuclear Materials 486 (2017) 111e121

Contents lists avai

Journal of Nuclear Materials

journal homepage: www.elsevier .com/locate/ jnucmat

Experimental measurements of surface damage and residual stressesin micro-engineered plasma facing materials

David Rivera, Richard E. Wirz, Nasr M. Ghoniem*

a Department of Mechanical & Aerospace Engineering, University of California at Los Angeles (UCLA), 420 Westwood Plaza, Los Angeles, CA 90095-1597,USA

h i g h l i g h t s

� Refractory metals with micro-engineered surfaces are developed as heat shields for high heat flux applications.� Tungsten with micro-pillar type surface architecture is shown to have significantly reduced residual stresses after plasma exposure.� X-ray diffraction (XRD) spectra of the W-(110) peak reveal that broadening of the FWHM for micro-engineered samples is small.� Spectral shifts of XRD signals show that cyclic plasma heat loading anneals out built-up residual stresses in micro-engineered surfaces.

a r t i c l e i n f o

Article history:Received 24 September 2016Received in revised form20 November 2016Accepted 28 December 2016Available online 4 January 2017

Keywords:TungstenThermomechanicsPlasma transientsMicro-engineered surfacesResidual stress

* Corresponding author.E-mail address: david.rivera.ucla@gmail.com (D. R

http://dx.doi.org/10.1016/j.jnucmat.2016.12.0350022-3115/© 2017 Elsevier B.V. All rights reserved.

a b s t r a c t

The thermomechanical damage and residual stresses in plasma-facing materials operating at high heatflux are experimentally investigated. Materials with micro-surfaces are found to be more resilient, whenexposed to cyclic high heat flux generated by an arc-jet plasma. An experimental facility, dedicated toHigh Energy Flux Testing (HEFTY), is developed for testing cyclic heat flux in excess of 10 MW/m2. Weshow that plastic deformation and subsequent fracture of the surface can be controlled by samplecooling. We demonstrate that W surfaces with micro-pillar type surface architecture have significantlyreduced residual thermal stresses after plasma exposure, as compared to those with flat surfaces. X-raydiffraction (XRD) spectra of the W-(110) peak reveal that broadening of the Full Width at Half Maximum(FWHM) for micro-engineered samples is substantially smaller than corresponding flat surfaces. Spectralshifts of XRD signals indicate that residual stresses due to plasma exposure of micro-engineered surfacesbuild up in the first few cycles of exposure. Subsequent cyclic plasma heat loading is shown to anneal outmost of the built-up residual stresses in micro-engineered surfaces. These findings are consistent withrelaxation of residual thermal stresses in surfaces with micro-engineered features. The initial residualstress state of highly polished flat W samples is compressive (z -1.3 GPa). After exposure to 50 plasmacycles, the surface stress relaxes to �1.0 GPa. Micro-engineered samples exposed to the same thermalcycling show that the initial residual stress state is compressive at (- 250 MPa), and remains largelyunchanged after plasma exposure.

© 2017 Elsevier B.V. All rights reserved.

1. Introduction

Plasma-facing materials (PFM) are required for a great numberof technological applications, encompassing space electric propul-sion devices, micro-electronics fabrication, fusion energy conver-sion, and pulsed power devices, just to name a few. Continueddevelopment of space Electric Propulsion (EP), fusion energy, and

ivera).

Pulsed Power (PP) systems relies on fundamental advances in ourunderstanding of material performance and survival in extraordi-narily severe environments. The demand of higher performancematerials is even greater in future technologies that will requirematerials to operate in substantially more aggressive environ-ments. In such applications, PFM encounter unprecedented severethermomechanical environments, as plasma ions and electronsslam onto the surface. Many physical degradation phenomenaensue, includingmaterial loss by ablation, blistering, sputtering andevaporation, as well as genuine thermoemchanical damage in theform of extensive plastic deformation and complex surface

Fig. 1. Lifetime Figure-Of-Merit (FOM) metric of plasma-facing materials measured inkWhr/cm2, as a function of the incident heat flux (rough indicator of environmentalseverity).

D. Rivera et al. / Journal of Nuclear Materials 486 (2017) 111e121112

cracking. The environment severity is correlated with two mainparameters. The first is the total amount of absorbed energy densitythat passes through the surface, Gt (kWh/cm2); a measure of thematerial lifetime. The second load-related parameter is theinstantaneous heat flux, G (MW/m2). While Gt is indicative of theduration throughwhich thematerial will have to survive successiveamounts of energy, delivered to its surface either continuously orintermittently, G is associated with the response time scale of thematerial itself to instantaneous pulses of energy. Generally, if G ishigh, the material lifetime is low, and vice-versa.

If the incident heat flux is below z 0.1 MW/m2, active coolingmay not be required and many of today's EP devices (e.g. Hallthrusters) are passively cooled. However, for more demanding ap-plications and future devices, active cooling is required andsignificantly higher heat fluxes are expected. In Hall thrusters, theion power is 0.05 MW/m2, and the electron power is 0.8 MW/m2,resulting in rapid erosion of the insulating boron nitride rings, andthermo-mechanically induced fractures. The current limits of theenergy fluence are on the order of 3 TJ/m2 and 50 TJ/m2 for ions andelectrons, respectively. In ion thrusters, the cathode is a highlystressed component, receiving ion powers of 0.05 MW/m2, andreaching an energy fluence of 7.5 TJ/m2 at the end of life. On theother hand, the screen grids in ion thrusters receive only0.0025 MW/m2 of ion flux, and because of their critical designlocation, survive only to an energy fluence of 0.4 TJ/m2. Theaggressive environments of Magneto Plasma Dynamics (MPD)thrusters requires instantaneous electron power of 8 MW/m2 andion power of 1.6 MW/m2, and because of the pulsed operation, thelifetime energy fluence is 0.5 TJ/m2 for electrons and 1.0 TJ/m2 for Liions.

Fig. 1 shows the dependence of the lifetime FOM (Gt, kWh/cm2)on the instantaneous heat flux (G, MW/m2), carried by plasma ions,photons, and electrons, in a number of space electric propulsionand energy applications. The figure shows approximate FOM limitsfor current state-of-the art technologies (light green region). Thisincludes some common technologies, like turbine blades of jetengines, photo-voltaic solar cells, as well as comparable technolo-gies under development (e.g. fusion energy). The FOM is in therange 100e600 kW h/cm2 for the most demanding present daytechnologies. To achieve greater lifetimes in the future (e.g. inexcess of 1000 kWh/cm2), newmaterials with special surfacesmustbe developed.

Severe plasma transients are expected to greatly influence theintegrity of plasma-facing components in fusion energy systems,notably the tungsten armor in most of existing divertor designs.Typical design-base transients contain an amount of energy inbetween 0.1 and 0.5 MJ/m2 for the Joint European Torus (JET) andbetween 1 and 5 MJ/m2 for the International Tokamak Experi-mental Reactor (ITER), lasting a duration between 0.1 and 1 ms,respectively. Such transients are expected to cause surface fracture,plasticity, and grain motion in polycrystalline tungsten. Among theoperational scenarios of interest is the response of tungsten to edgelocalized modes, which carry energy densities on the order of 1 MJ/m2 and time durations in the range of 0.1e0.5 ms, which can resultin severe damage to the tungsten armor in the form of cracking andlocalized melting of the material [1]. Cyclic thermal transients aregenerally associated with complex plasticity and fracture phe-nomena. In particular, the response of a material facing cyclic highheat flux loading can be either elastic, plastic, ratcheting, shake-down, or fracture. Ratcheting is the continuous accumulation ofplastic strain with each passing cycle up to the point of failure, andis considered detrimental. Shakedown describes an initial degree ofplastic deformation, quickly reaching equilibrium in subsequentloading cycles. A detailed analysis of the residual stress state of the

micro features comprising the coating is necessary to provide anavenue for improvement in the coating process. If the surface layeris structured so as to allow expansion and contraction withoutthese large residual stresses, one might be able to develop a ma-terial that is more resistant to high heat flux thermomechanicaleffects. We have recently developed such amaterial concept, wherean “armor” coating is deposited by CVD on top of the main struc-ture, with the main function being thermomechanical damagemitigation [2]. The present work extends the experimental in-vestigations presented in Ref. [2].

A number of researchers have attempted designing refractorymetal coatings from a traditional standpoint where a flat surface isdeposited onto a substrate and exposed to a thermally harshenvironment [3e6]. Although efforts along this path have provedsuccessful in protecting the underling structure, improvements arestill achievable through the advent of castellated armor concepts[2]. In such designs, free-standing structures, such as micro-pillarsare grown onto a surface, and are capable of protecting it throughreduction of residual thermal stresses. Similar concepts have beenproposed at a macroscopic scale, yet function largely in the sameway [7,8]. A unique multiphase material system is developed andinvestigated here. Experimental results characterizing the me-chanical response of surfaces designed with coatings of uniformmicro-pillar geometry, each comprised of a rhenium core sur-rounded by a tungsten shell, will be discussed here. The compositenature of microarchitected surfaces, coupled with the mechanicalstrength and ductility furnished by small feature size, make thesematerial concepts suitable for shielding the underlying structurefrom the effects of severe plasma transients. The experimentaleffort presented in this article will aim to show that the plasticdistortion induced by thermal loading of microarchitected surfacesis mitigated relative their planar counterpart. The reduction ofplastic distortion in microarchitected surfaces brings to light thetailorable properties of such coatings, providing designers with amaterial capable of withstanding a range of extreme environmentsand not limited to the intrinsic materials properties of a monolithicplanar surface. The introduction of new surface design variables,such as geometry of constituent features, in addition to materialselection offers an opportunity in thermal shielding technology.

The objective of this work is to develop plasma-resilient, micro-engineered, refractory metal armor (coating) to resist thermalshock and harmful thermal residual stresses during severe plasmatransients. The focus is on manifestations of thermomechanicaldamage in the form of surface fracture and residual plastic stresses,

D. Rivera et al. / Journal of Nuclear Materials 486 (2017) 111e121 113

recognizing that other factors may have to be taken into account,such as vacuum compatibility, erosion, and fuel retention, whenthese materials are to be used in practice. We first present themethod of sample fabrication by the Chemical Vapor Deposition(CVD) process in section 2. We then discuss the experimental setupfor testing high heat flux samples in section 3, and show results ofsurface damage characterization of tungsten samples exposed torepeated plasma transients in section 4. Observations of surfacefracture and how the fracture mode depends on surface architec-ture are shown in subsection 4.1, while results of residual stressevolution using X-ray diffraction techniques are given in subsection4.2. Here, we study the evolution of residual thermal stresses insamples that have smooth surfaces. and compare the results tosamples that have micro-engineered surfaces. Finally conclusionsof part 1 of this article are given in section 5.

2. Micro-engineered material fabrication

Textured and foam-type plasma-resilient materials are fabri-cated by the Chemical Vapor Deposition (CVD) process at Ultramet,Inc. Purity levels in excess of 99.99% are achievable. In addition, theCVD process exhibits the greatest throwing power, or ability touniformly deposit materials onto/into intricately shaped ortextured substrates, thus allowing fabrication of near-net shapeparts with geometries that cannot be produced through conven-tional processing and machining. Metal deposits are formed on thesubstrate at the molecular level, while the gaseous reactionbyproducts are exhausted from the system. The typical depositionrate for metals is 0.010 in/hr. CVD tungsten is deposited at just 550 +

C through the hydrogen reduction of its hexafluoride under vac-uum, flowed over a heated substrate. Deposition of other refractorymetals such as molybdenum and rhenium is similarly performedthrough the thermal decomposition and reduction of metal halides.

An example of a dendritic/pillar-like coating can be seen inFig. 2. The height of the pillars generally extend 15e30 mm from thebase and can range in diameter from 2 to 5 mm. They are fabricatedthrough a chemical vapor deposition process in which rheniumdendrites are first formed on the surface of a given substrate(tungsten in the case seen here), and subsequently given a thinprotective layer of W on the order of 0.3e1 mm thick. The combi-nation of materials (Re and W) provides these geometric featureswith a composite structure and thus their stress response duringthermal loading will differ relative to that seen in traditional planarsurfaces. The differing stress response can be further attributed tothe small scale of the geometry. Designing structures at the micronand nano scales has been shown to induce the well known “sizeeffect”in which traditionally brittle materials can sustain largeamounts of plastic strain without fracture [9e11]. Microarchitectedcoatings harness the superior mechanical properties afforded bythe size effect into the design scheme, this results in a heat shieldwhich benefits from both the freedom of expansion provided byindependently moving microfeatures as well as a betterment ofinherent mechanical strength common to micron and submicronscale structures. Additionally the selection of tungsten and rheniumto form the basis for the micro composite structure allows forsignificant interdiffusion between the two metals at elevatedtemperature [12], effectively alleviating any detrimental interfacialstresses which may lead to premature failure. While the currentresearch will show the superior performance of thin layers ofmicro-engineered surfaces in response to plasma transients, it maybe feasible in the future to develop a thick structure composed ofmany layers that have the same micro-engineered features.

3. High heat flux experimental setup

In order to test the response of microarchitected surfaces undera thermal load, two W substrates were coated with a dendriticsurface consisting of Re pillars covered with a thin layer of W. SEMimages of these coatings, fabricated by Ultramet Inc., are shown inFig. 2. The samples were then exposed to a cyclic thermal loading,using a high heat flux testing apparatus, designed at UCLA anddescribed in Ref. [11]. The specimens, given internal identifiers ofPW60 4 and PW60 5, are shown in their pristine (unexposed)condition in Fig. 4. The approach here is to show the possibility ofmicro engineering surfaces to mitigate plasma damage. Once this isproven, other material combinations and engineered armor can bedeveloped to enhance functional performance in other areas (e.g.neutron radioactivity, vacuum and tritium retention, fuzz forma-tion, blistering, and sputtering erosion.

3.1. Thermomechanical plasma testing facility

The High Energy Flux Testing facilitY (HEFTY), designed andbuilt at UCLA, provides for the ability to study thermomechanicaldamage using a commercial arc-jet plasma gun (Praxair SG-100). Inorder to study the effects of long duration plasma transients (>10 MW/m2), samples that are thermally exposed in the system aresimultaneously water cooled via a specially designed flange ontowhich they are mechanically fastened during the course of a test.The cooling circuit is a closed loop, where traditional calorimetrymethods can be used to calibrate the magnitude of the heat fluxreceived by the sample. The heat flux provided by the gun wasmeasured with a heat flux sensor, and calibrated through calo-rimetry, was determined to be in the range 13e16 MW/m2 at a2.54 cm distance from the gun nozzle, as seen in Fig. 3.

The energy flux impinging upon a sample can be controlled bysimply modifying the distance between the sample face and thegun nozzle exit, as seen in Fig. 3. In this manner, a range of damagemodes caused by different heat flux values can be studied. Inaddition, the high cooling rates afforded by water jet impingementallow the investigation of damage effects caused by long durationplasma transients (e.g. recrystallization, dislocation reorganization,grain growth etc.). The HEFTY facility can be used for studies ofthermal shock [13e15], thermomechanical fatigue, and fatigue-creep interactions. The system is controlled via a LabView pro-gram, thus allowing controlled pulsed operation capabilities. Animage of a sample being exposed in the HEFTY facility is shown inFig. 5(a). Before exposure, the chamber is pumped down to lowvacuum levels and back-filled with argon gas to mitigate oxidationof the tungsten surface at high temperature. The samples them-selves are mechanically attached to a specialized cooling flange viaa series of bolts. An illustration of the sample geometry used in theexperimental setup reported here can be seen in Fig. 5(b). Thesamples are typically 5.08 cm in diameter and 1.5 mm thick. Largecompressive stresses are generated upon pulsing as a result of thesample thermal exposure area being smaller than the sample sizeitself, as seen in Fig. 5(b). The generated large compressive radialstress, the magnitude of which can be controlled by cooling, is thedriving force for quantitative thermomechanical damage that ispresented in this work.

3.2. Heat flux measurement

Heat flux measurement techniques include differential tem-perature sensors (e.g. thermopiles, layered resistance temperaturedevices or thermocouples), calorimetric methods involving a heatbalance analysis, energy supply or removal methods (e.g. using aheater to generate a thermal balance), and indirectly through

Fig. 2. Example of a micro-engineered armor surface consisting of densely packed W-coated Re pillars.

Fig. 3. Schematic of the experimental set-up during plasma exposure. A sample wasplaced a distance of 2.54 cm from the gun nozzle, and the heat flux at this distance wasmeasured as 14.5± 1.5 MW/m2.

D. Rivera et al. / Journal of Nuclear Materials 486 (2017) 111e121114

measurement of mass transfer that can be correlated with heattransfer [16]. Each of these methods has a specific design andapplication. In our experimental setup, we determine the heat fluxon the sample surface by two independent techniques. The first is acalorimetry method, where the temperature rise of the coolingwater stream is measured during a fixed number of thermal cycles.Measurement of the temperature rise, together with the water flowrate, results in determination of an average value of the heat flux byconserving thermal power on the sample. The input heating ratefrom the plasma arc-jet is balanced by the absorbed energy in thewater stream. The second method is based on the development of aheat flux sensor, shown in Fig. 6(b), and using it in an inverseproblem of determining an unknown heat flux from transienttemperature measurements. The developed sensor is a variant of anull-point calorimeter, and is comprised of a copper cone with asmall bore hole drilled through the back. The hole stops just short

of penetrating through the length of the cone, and leaves a thinsection between the exposed tip and the attached thermocouple, asseen in Fig. 6(a). This set-up allows for the transient temperaturesignal to be comparedwith Finite Element Analysis (FEA) data, so asto match the incident plasma heat flux. The simulated temperaturesignal is generated by importing the sensor geometry to a com-mercial FEA solver (Comsol), and applying an estimated heat fluxvalue to the tip of the cone where the heat flux is impinging. Theinverse problem is solved iteratively, where various values of theheat flux are assumed to be constant within the 6 s period of asingle pulse, and the FEA simulated signal is matched with themeasured temperature transient signal. Fig. 6(b) presents a pictureof the fabricated sensor prior to a test. The two methods of mea-surements (water calorimetry, and transient temperature match-ing) have produced results within 10%. When the plasma gunnozzle distance is 2.54 cm from the sample surface, the measuredheat flux is 14.5± 1.5 MW/m2.

4. Characterization of surface damage

A series of plasma exposure tests were performed on uncoatedand coated samples in order to examine the nature of thermo-mechanical damage, and relate it to the surface architecture. Inboth cases of smooth sample surfaces and micro-engineered sur-faces, the cooling rate of the sample is varied. When the samplesare uncooled and are subjected to a limited number of plasmapulses, extensive damage is generally observed. On the other hand,by varying the water flow rate, the rate of heat removal from theback of the samples can be adjusted, and the sample stress state canthus be controlled. We will first discuss the results of plasmaexposure of micro-engineered samples, followed by observations ofthermomechanical damage of smooth uncoated samples.

The pillar architecture of the coated surfaces of samples PW604 and PW60 5 is comprised of a rhenium core that has a hexagonalcross-section, and is coated by a thin layer of tungsten. Typicalmeasured dimensions of the pillar are as follows: Re core diameterat the base¼ 5 mm, at the tip ¼ 3 mm, pillar height ¼ 20 mm, Wcoating thickness ¼ 0.3 mm, pillar areal surface density ¼ 8 �1010 m�2. Samples PW60 4 and PW60 5 were subjected to a totalof 200 and 250 plasma pulses, respectively (G¼ 14.5± 1.5 MW/m2).Each pulse had a 10 s duration (ton ¼ 6 s, toff ¼ 4 s). The specimenswere characterized using scanning electron microscopy (SEM) andX-ray diffraction (XRD) techniques at 50 pulse intervals. Broadeningof the full width at half-max (FWHM) and the shift in the X-raydiffraction peaks are correlated with SEM observations of surface

Fig. 4. Un-exposed microarchitected samples of the dendritic variety, both consisting of Re dendrites coated with W, a) PW60 4. Note the oriented nature of the pillars grownnormal to the surface, b) PW60 5, pillars appears to be grown in a more random nature as compared to PW60 4.

Fig. 5. a) Demonstration of the plasma arc-jet firing onto a tungsten sample, b) A sample being exposed to a plasma column in HEFTY. Note the holes on the perimeter of the sampleare used to fasten the specimen onto a specialized flange so as to provide for active cooling. The exposed area on the sample is smaller than the overall sample size, which leads tothe build-up of large compressive thermal stresses.

Fig. 6. Heat flux sensor used for heat flux determination. a) The principle of operation relies on matching a transient temperature signal to FEA simulations. The thermocouple isattached to the back of the exposed sensor tip. b) Fabricated sensor prior to testing.

D. Rivera et al. / Journal of Nuclear Materials 486 (2017) 111e121 115

D. Rivera et al. / Journal of Nuclear Materials 486 (2017) 111e121116

damage. The FWHM was used as a measure of plastic distortioninduced by thermal loading. XRD was also employed to investigatethe texturing of the coatings prior to exposure and account forrecrystallization induced by the high surface temperatures duringtesting. The initial XRD pattern for both PW60 4 and PW60 5 canbe seen in Fig. 7. The absence of the (211) peak in PW60 4 is anindication of a larger degree of texturing in the sample relative tothe PW60 5 sample. This can be further confirmed by SEM ex-aminations, as shown in Fig. 4, where the more random nature ofpillars is observed in the PW60 5 sample.

4.1. Observations of surface fracture

4.1.1. Micro-engineered samplesFollowing the initial 50 thermal plasma cycles, sample PW60 4

exhibited light cracking along many of its constituent micropillars,as shown in Fig. 8. This cracking pattern can be explained byconsidering the source of stress evolution during plasma cycles inthe micro-pillars. The cross-section of an individual pillar can beseen in Fig. 9, which reveals the composite-like nature of themicro-pillar. The Re core of the pillar is coated by a thin W layer (z0.2e0.5 mm) on the outside. During the heating phase of the plasmapulse, the Re core expands against the W coating, which effectivelyacts as a casing for the Re dendrite. Considering that the thermalexpansion coefficient of Re (9.1� 10�6 K�1) is almost twice as largeas that of W (4.3� 10�6 K�1), the expansion/contraction of Reduring thermal cycling will put the W coating under tension/compression, respectively. The tensile stress in the W coating canbe shown to be sufficient to cause plastic flow during thermalloading, and depending on the thickness of the coating (W) can leadto either compressive or tensile states of residual stress. Thisfinding is used to design pillars that are more resistant to thermalfracture by appropriate selection of the pillar geometry.

4.1.2. Smooth planar samplesBefore turning the attention to a description of thermo-

mechanical damage in smooth, planar samples, we point out theimportance of controlling the sample cooling rate. Adjustments inthe cooling rate provide an opportunity to study distinct damagemodes, manifest as either fracture, plastic deformation, or a com-bination of both. High cooling rates lead to a controlled

Fig. 7. Comparison of W over Re microarchitected specimens, both diffraction patternsrepresent two samples in pristine condition, note the large degree of texturing present,strong preferred orientation in the < 110> direction, substrate thickness is 0.060 in(1.5 mm), the increased degree of texturing seen in PW60 4 is apparent from themissing (211) peak relative to PW60 5.

temperature rise in the sample, thus allowing control over residualthermal stresses. On the other hand, when cooling is totally shut off,the sudden rise of sample temperature leads to thermal shock typefracture.

Experiments on planar tungsten samples under high coolingand long duration plasma transients (ton ¼ 6 s, toff ¼ 4 s) haveshown that the samples begin to recrystallize after roughly 1100cycles of G ¼ 14.5± 1.5 MW/m2. The appearance of recrystallizationleads to the generation of micro-cracks on the specimen surface[11]. X-ray diffraction (XRD) measurements of pre-exposure andpost-exposure planer tungsten can be seen in Fig. 10. The presenceof freshly nucleated grains is evident from the emergence of pre-viously absent diffraction peaks before plasma exposure. Recrys-tallization is known to have deleterious effects on the fracturetoughness of tungsten [17], and results in surface micro-cracks.

To study thermal shock response, samples with smooth planarsurfaces were suddenly exposed to several plasma pulses withoutany active cooling of the sample, and surface fracture was observed.Large residual tensile stresses, generated upon sample cool-down,cause specimen fracture in the central exposed area. Radialconfinement of the hot exposed central region of the sample by thecooler surrounding material, as illustrated in Fig. 5(b), is the pri-mary reason of observed thermal shock fracture. The combinationof large compressive stress seen by the exposed area of the sampleand the reduced flow stress of tungsten at elevated temperature(only 290 MPa at 727 + C for polycrystalline tungsten) results insignificant residual tensile stresses at the exposed region of thespecimen upon cool-down. The effective outcome of this sequenceof events is a complex surface fracture pattern of the sample, asshown in Fig. 11(a). SEM images of a sample tested without activecooling is shown in Fig. 11(b), where fracture is clearly seen span-ning the surface of the specimen.

A remarkable aspect of thermomechanical fracture in W is thedual nature of its response, being either brittle or ductile,depending on the local temperature at any given region. When thefracture surface is more closely examined, it is interesting to notethat the central exposed region of the specimen fails by ductiletearing, while the sample periphery fractures in a completelybrittle manner. This is evidenced by a dimpled fracture surface,shown in Fig. 12(a). The periphery regions of the sample, which areunexposed to the incoming plasma fail in a brittle manner. This canbe observed by reference to Fig. 12(b), which shows the relativelyflat cleavage planes of the fracture surface. The transition fromductile tearing to brittle fracture within one sample is attributableto the radial temperature gradient in the sample as a result ofplasma heating being supplied only to the central region. Theappearance of both ductile and brittle failure modes in the samespecimen presents an opportunity for further studies of theDuctile-to-Brittle-Transition (DBTT) phenomenon in W underthermomechanical conditions.

The observed thermomechanical damage mode can range fromfracture at low cooling conditions to controlled plastic deformationwhen samples are vigorously cooled. Under conditions with noactive cooling, excessive thermal stresses coupled with a low flowstress at elevated temperature lead to thermal shock fracture.Nevertheless, this type of fracture is not entirely of a brittle manner,but shows a transition from a ductile tearing mode in regionswhere the plasma deposits its energy to a classical brittle mode insample zones that are not in contact with the plasma source.However, when the temperature is controlled via active cooling,thermal stresses are correspondingly reduced, causing only plasticdeformationwithout initiation of fracture. This is further confirmedthrough X-ray diffraction studies of residual stresses, as presentednext in subsection 4.2. Plastic deformation is indeed evident in

Fig. 8. W coated Re pillar showing light cracking along its facets.

Fig. 9. Cross-section of a W coated Re pillar.

Fig. 10. Recrystallization in planar tungsten exposed to long duration plasma transients a) unexposed XRD spectrum, b) exposed XRD spectrum showing the emergence of newdiffraction peaks (110) and (220) caused by the nucleation of stress free grains.

D. Rivera et al. / Journal of Nuclear Materials 486 (2017) 111e121 117

highly cooled samples, yet fracture is avoided or largely mitigatedin comparison with non-cooled specimens.

4.2. X-ray diffraction analysis of residual plastic strain

Residual plastic strain can be qualitatively, and in some casesquantitatively measuredwith the X-ray diffraction (XRD) technique

[18e21]. The fundamental basis of the method is Bragg's law ofdiffraction: nl ¼ 2dsinq, where l is a fixed X-ray wavelength, d theinteratomic spacing, and q the diffraction angle. Measurements ofthe diffracted radiation intensity gives an XRD spectrum, wherepeaks in a stress-free condition are associated with specific crystalplanes for single crystals. A shift in the peak intensity to largerdiffraction angles is correlated with a decrease in the interatomicspacing, d. During plastic deformation, dislocation motion on glideplanes results in local lattice rotations and curvature, while thedisplacement field emanating from dislocation cores expands orcontracts nearby interatomic spacings. The first effect (latticerotation and curvature) is manifest in broadening of the XRD signal,while the second (dilatation/contraction) appears as a shift in thediffraction angle. Since the measurement is basically a surfacemethod, the normal stress is zero and one has a condition of planestress on the surface. The in-plane stress at an angle f from apredetermined principal direction on the surface (sf) is obtained

from a set of measurements of the inter-planar spacings, dj, and dnat angles j to the surface normal and perpendicular to the surface,respectively. The residual stress is then given by:

sf ¼ Eð1þnÞsin2j

�dj�dn

dn

�. The XRD intensity signal is fitted to a curve,

where both the peak angle and the full signal width at halfmaximum (FWHM) are obtained. Both FWHM and the shift angle

Fig. 11. Surface crack pattern in the smooth planar W sample exposed to a 16 MW/m2 heat flux (not cooled), a) Cracks were initiated at the sample center and spanned the diameterof the sample splitting the specimen in two pieces of roughly equal size, cracks are indicated by the red arrows, b) SEM image of the crack pattern showing possible melting of thesurface as indicated by the splat pattern near the crack opening. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of thisarticle.)

Fig. 12. The fracture surface of a smooth planar specimen tested without active cooling. a) Dimpled ductile failure at the exposed region of the sample, where temperatures arehighest, b) Brittle failure at the periphery of the sample, which is not exposed to the incoming plasma source.

D. Rivera et al. / Journal of Nuclear Materials 486 (2017) 111e121118

can be used to analyze the state of plastic deformation near thesurface. Although peak broadening data can give informationregarding the amount of plastic strain, it is not ideal for the quan-titative determination of the residual stress, in particular in caseswhen the grain size exceeds 100 nm [19,22]. In such cases it is bestto resort to alternative XRD tools; one particular analytical tech-

nique is the sin2ðjÞ method. There, the average residual surfacestress is obtained via the relationship sf ¼ E

ð1þnÞ1d0

vdvsin2ðjÞ, where d0

is the unstressed lattice parameter of the material, j is the tilt anglethe sample has been rotated through relative to the incoming X-rayradiation [18,19]. The stress sf is then proportional to vd

vsin2ðjÞ. This

later quantity is obtained as the slope of a line relationship between

the lattice parameter and sin2ðjÞ. By plotting a series of latticeparameter values from XRD profiles over a range of tilt angles j, alinear relationship can be established between the lattice param-

eter d and sin2ðjÞ, from which the slope vdvsin2ðjÞ can be quickly

determined. Finding values of lattice parameters d for each angle oftilt j makes use of the fact the position of the diffraction peak will

shift if elastic strain is present in the material.In order to study the effects of repeated plasma pulsing on the

plastic deformation of tungsten at high temperature, a smoothplanar W sample was exposed to 1100 cycles with G ¼ 25 MW/m2,ton ¼ 4 s, toff ¼ 6 s, resulting in a total surface energy density of110 GJ/m2 (z 3 kWh). The sample was actively cooled during thisexperiment. Broadening of the (200) peak after plasma exposure isclearly shown in Fig. 13. Because the high surface temperature re-sults in recrystallization and recovery of the material, peak broad-ening is a qualitative measure of the dislocation density, and henceresidual plastic strain near the surface layer.

To determine the relative change in the residual surface stresswhen the material is micro-engineered to have s pillar architecturecoating, X-Ray Diffraction (XRD) profiles of specimens PW60 4 andPW60 5 were examined. XRD spectra of micro-engineered sam-ples were contrasted with those of planar W specimens subjectedto a similar plasma exposure. Since the FWHM can be regarded as aqualitative measure of the degree of plastic deformation present inthe surface, the FWHM of the (110) peak for all specimens wasrecorded after every 50 cycles, and plotted against the absorbed

Fig. 13. (200) peak FWHM before and after plasma exposure. Peak broadening isobserved as a function of repeated plasma exposure, and is a qualitative measure ofresidual plastic strain.

D. Rivera et al. / Journal of Nuclear Materials 486 (2017) 111e121 119

energy density (GJ/m2). The results of the comparison are displayedin Fig.14. The plot reveals a marked reduction in the plastic strain inthe micro-engineered material relative to that of planarW. The rateof FWHM increase (slope) with absorbed plasma thermal energy isreduced by almost a factor of 4 for micro-engineered samples ascompared to smooth planar samples.

Following the comparison of residual plastic strain betweenplanar and micro-engineered samples, we further investigatechanges in the surface stress in micro-engineered samples. The(110) diffraction peaks for specimens PW60 5 and PW60 4 areplotted in Fig. 15(a) and (b), respectively. The shift in the peaks,shown in Fig. 16(a) and (b), after the initial 50 pulse exposure, isclearly noted in both samples. We note that in the PW60 5 sample,the peak shifts towards smaller Bragg angles, indicating a tensilestress, while the PW60 4 (110) peak moves towards larger Braggangles, consistent with a residual compressive stress. This differ-ence in the nature of the residual stress is attributed to pillar ge-ometry, where the Re core size and the W coating thicknessdetermine the stress state following plastic deformation. ThinnerW coatings on the Re core will tend to remain in a state ofcompressive surface stress, while the thicker variants result in atensile surface stress.

The XRD peaks move back towards their initial position with

Fig. 14. Dependence of the FWHM of the (110) peak on absorbed energy density. Notethe tendency of the microarchitected material PW60 4 and PW60 5 to retain smallerplastic strain as compared to the planar counterpart. 0.508 and 1.5 are the respectivethicknesses (mm) of planar samples tested in HEFTY (PW).

increasing plasma exposure, indicating that the dislocationmicrostructure in theW coating is annealed, gradually relieving theplastic strain built up during the first set of plasma pulses (seeFig. 16(a) and (b)). This behavior is a consequence of the competi-tion between plastic hardening during deformation and a simul-taneous annealing or annihilation of dislocation density at elevatedtemperatures. The more gradual relaxation of PW60 5 towards itsinitial Bragg angle as well as its increased peak broadening relativeto PW60 4 (as seen in Fig. 14) is thought be due to the highlyoriented nature of sample PW60 4. The majority of the plasticstrain is concentrated along the length of dendrites, because the X-ray beam generating the diffraction pattern is primarily striking thetop or caps of the pillars. Thus, the plastic strain in PW60 4 may bereduced relative to PW60 5, where an increased portion of thesignal is generated from locations along the length of the pillarswhere strain is expected to be greatest.

The residual stress state of both planar and microarchitected

surfaces was examined through the use of the sin2ðjÞ technique,both before and after thermal pulsing (with active cooling). Theresults show that the initial residual stress state for highly polishedflat tungsten samples is compressive, and on the order of 1.3 GPa.The compressive surface stress is attributed to the effect of surfacepolishing before plasma exposure. After 50 plasma cycles, thestresses relax to a compressive value of 1.0 GPa, representing astress relaxation of 300 MPa. In contrast, micro-engineered mate-rials (sample PW60 6) exposed to the same thermal cyclingenvironment show that the initial residual stress state for thesurface is compressive at 250 MPa, and remains largely unchangedafter plasma exposure. For reference, tungsten powder wasanalyzed using the same methodology, and showed a tensile stressvalue of 300 MPa. This comparison is evidence of the essentiallystress-free state of the micro-engineered surface after plasmaexposure. This particular sample (PW60 6) was highly tapered,with an average tip diameter of 1 mm. In some cases, the base of thepillars widened to nearly 5 mm, and do not show strong preferredorientation, as can be observed in Fig. 17.

The largely stress-free condition of the micro-engineered ma-terial relative to its planar counterpart allows the plasma-exposedsurface to sustain large thermal loads without catastrophic failure.Two factors aid the surface in sustaining thermomechanicallyinduced stresses without failure. First, the strain tolerant nature ofthe structure itself gives the top layer the ability to freely expandand contract , without damaging thermal strains. This is similar towhat has been achieved in the design of Thermal Barrier Coatings(TBC) that were successfully introduced for thermal protection ofjet engine turbine blades. Second, the fundamentally ductile natureof the tungsten - rhenium alloy system results in a greater degree ofplastic flow accommodation. As is well documented, W-Re alloysare among the most favorable materials for high temperature ap-plications, as they posses a desirable combination of high flowstress at elevated temperature in addition to good ductility, even atroom temperature [23].

The reduction in residual stress seen in micro-engineered sur-faces can in part be attributed to the highly ductile nature of theW-Re alloy system. Micro-pillars comprising a micro-engineered sur-face are composite structures consisting of a Re core and a thin Wcoating. However, due to the high inter-diffusivity of Wand Re [12],this initial composite structurewill quickly lose the sharp interfacesbetweenWand Re, and form a more diffuse bond between the twometals. Since the pillars making up the micro-engineered surfaceare of such small dimensions (only 5 mm in diameter with the Wovercoat being 0.5 mm in thickness), they will inter-diffuse forminga gradedW-Re alloy. This effect has the benefit of quickly alleviatingany stresses in the composite, which arise from sharp changes in

Fig. 15. XRD spectra before and after plasma exposures. a) Sample PW60 5, residual tensile stress, dendrites were � 2.1 mm at the tip and showed significant tapering/increase indiameter towards the base, b) PW60 4 residual compressive stress, dendrites were � 3.6 mm in diameter at the tip and showed limited tapering relative to PW60 5.

Fig. 16. XRD (110) peak position as function of absorbed plasma energy density. a) sample PW60 5, initially the peak shifts to lower Bragg angles indicating a tensile residual stress.Residual stress relaxation is evident upon repeated exposure, as the Bragg angle begins to move back toward its initial position, dendrites were � 2.1 mm at the tip and showedsignificant tapering/increase in diameter towards the base, b) Peak shift in sample PW60 4 indicating a residual compressive stress. Dendrites were � 3.6 mm in diameter at the tipand showed limited tapering relative to PW60 5.

Fig. 17. SEM image of sample PW60 6. Measured residual stress before and afterplasma exposure remained largely unchanged at a compressive value of 250 MPa. Theexposure parameters were 50 cycles at G ¼ 16 MW/m2, ton ¼ 6 s, toff ¼ 4 s.

D. Rivera et al. / Journal of Nuclear Materials 486 (2017) 111e121120

the coefficient of thermal expansion. Effectively, the compositerelaxes the thermal stresses after the first several minutes ofoperation to come to an equilibrium configuration consisting of aductile refractory metal alloy.

5. Conclusions

The potential for development of materials systems with tail-orable properties has been a point of interest in materials sciencefor many years [24,25]. Throughmodifications in surface geometry,it has been shown that materials can bemade tomove beyond theirapparent intrinsic limitations and produce structures withimproved mechanical properties, such as increased ductility andstrength. Surfaces with micro-engineered architecture are effectivein reducing plastic strain in situations of extreme thermal loadingdue to plasma exposure.

Micro-engineered surfaces have been shown in this work to bemore resistant to plastic deformation in plasma facing materials.This behavior was observed through the analysis of XRD peakbroadening data, which demonstrated reduced levels of plasticstrain for micro-engineered surfaces relative to their planar coun-terparts. The reduction in plastic deformation is further confirmed

by residual stress determination via the sin2ðjÞ method. Analysisusing this technique shows that residual stresses in micro-engineered surfaces are largely absent after thermal plasmacycling, whilst planar surfaces exposed to the same test conditionsretain residual stresses of up to 1.0 GPa inmagnitude. The reductionin residual stress is a direct result of the strain tolerant nature of themicro-engineered surface, which allows free expansion of thegeometric features comprising the surface. Additionally, the strong

D. Rivera et al. / Journal of Nuclear Materials 486 (2017) 111e121 121

tendency for interdiffusion between W and Re allows the micro-structure to quickly relax any stresses associated with coefficient ofthermal expansionmismatch at the interface of the twometals. Theinterdiffusion betweenWand Re at elevated temperature graduallytransitions the micro-pillars comprising the surface from acomposite-like structure with sharp interfaces between constitu-ent parts (W and Re) to an increasingly homogeneous distributionof material, effectively forming an alloy. The inherently favorablenature of W-Re alloys for high temperature applications providesthe micro-engineered surface with a ductile characteristic, anecessary condition for damage resistance under pulsed high heatflux conditions. Pillar geometry is shown to play a key role in theperformance of the micro-engineered surface. When the Re corediameter is large relative to the W shell thickness, cracking canensue along the height of the structure as tensile hoop stresses areeasily generated from the thermal expansionmismatch betweenWand Re. To prevent fracture of the pillars, theW shell thicknessmustbe sufficient to resist the expansion of the Re core, resulting in acompressive residual stress in the W shel.

Acknowledgements

This work is partially supported by the Air Force Office of Sci-entific Research (AFOSR), Award No: FA9550-16-1-0444with UCLA,and by the U.S. Department of Energy, United States, Award Num-ber DE-FG02-03ER54708 with UCLA. The technical assistance ofPhil Chiu, Nick Tarasi, and Dan Goebel with the HEFTY controlsystem, and that of Chao Li and Professor Mark Goorsky with XRDanalysis is greatly appreciated.

References

[1] S. Pestchanyi, I. Garkusha, I. Landman, Simulation of residual thermostress intungsten after repetitive ELM-like heat loads, Fusion Eng. Des. 86 (9) (2011)1681e1684.

[2] S. Sharafat, N.M. Ghoniem, M. Anderson, B. Williams, J. Blanchard, L. Snead,Micro-engineered first wall tungsten armor for high average power laserfusion energy systems, J. Nucl. Mater. 347 (3) (2005) 217e243.

[3] X. Liu, L. Yang, S. Tamura, K. Tokunaga, N. Yoshida, N. Noda, Z. Xu, Thermalresponse of plasma sprayed tungsten coating to high heat flux, Fusion Eng.Des. 70 (4) (2004) 341e349.

[4] X. Liu, S. Tamura, K. Tokunaga, N. Yoshida, N. Noda, L. Yang, Z. Xu, High heatflux properties of pure tungsten and plasma sprayed tungsten coatings,J. Nucl. Mater. 329 (2004) 687e691.

[5] Kazutoshi Tokunaga, Yusuke Kubota, Nobuaki Noda, Yoshio Imamura,Akira Kurumada, Nobuaki Yoshida, Toshiaki Sogabe, Toshiyuki Kato,Bertram Schedler, Behavior of actively cooled mock-ups with plasma sprayed

tungsten coating under high heat flux conditions, Fusion Eng. Des. 81 (1)(2006) 133e138.

[6] Y. Yahiro, M. Mitsuhara, K. Tokunakga, N. Yoshida, T. Hirai, K. Ezato, S. Suzuki,M. Akiba, H. Nakashima, Characterization of thick plasma spray tungstencoating on ferritic/martensitic steel f82h for high heat flux armor, J. Nucl.Mater. 386 (2009) 784e788.

[7] A. Rene Raffray, Siegfried Malang, Xueren Wang, Optimizing the overallconfiguration of a he-cooled w-alloy divertor for a power plant, Fusion Eng.Des. 84 (7) (2009) 1553e1557.

[8] E.A. Mogahed, I.N. Sviatoslavsky, 3d thermo-mechanical analysis of the iterlimiter small scale specimen, Fusion Eng. 1997. 17th IEEE/NPSS Symp. 1(1997) 381e384. IEEE.

[9] Ladislas Kubin, Dislocations, Mesoscale Simulations and Plastic Flow, vol. 5,Oxford University Press, 2013.

[10] M. Pozuelo, Y.W. Chang, J.-M. Yang, In-situ microcompression study ofnanostructured mg alloy micropillars, Mater. Lett. 108 (2013) 320e323.

[11] David Rivera, Tamer Crosby, Andrew Sheng, Nasr M. Ghoniem, Characteriza-tion of thermomechanical damage on tungsten surfaces during long-durationplasma transients, J. Nucl. Mater. 455 (1) (2014) 500e506.

[12] Y.F. Hua, Z.X. Li, X. Zhang, J.H. Du, C.L. Huang, M.H. Du, Inter-diffusion analysisof joint interface of tungstenerhenium couple, J. Nucl. Mater. 416 (3) (2011)270e272.

[13] G. Pintsuk, A. Prokhodtseva, Inge Uytdenhouwen, Thermal shock character-ization of tungsten deformed in two orthogonal directions, J. Nucl. Mater. 417(1) (2011) 481e486.

[14] Muyuan Li, Mathias Sommerer, Ewald Werner, Stefan Lampenscherf,Thorsten Steinkopff, Philipp Wolfrum, Jeong-Ha You, Experimental andcomputational study of damage behavior of tungsten under high energyelectron beam irradiation, Eng. Fract. Mech. 135 (2015) 64e80.

[15] M. Wirtz, J. Linke, G. Pintsuk, L. Singheiser, M. Zlobinski, Comparison ofthermal shock damages induced by different simulation methods on tungsten,J. Nucl. Mater. 438 (2013) S833eS836.

[16] P.R.N. Childs, J.R. Greenwood, C.A. Long, Heat flux measurement techniques,Proc. Institution Mech. Eng. Part C J. Mech. Eng. Sci. 213 (7) (1999) 655e677.

[17] A.V. Babak, Effect of recrystallization on the fracture toughness of tungsten,Powder Metallurgy Metal Ceram. 22 (4) (1983) 316e318.

[18] ASM International Handbook Committee, ASM Handbook, Volume 10-Mate-rials Characterization, ASM International, 1986.

[19] B.D. Cullity, Elements of X-ray Diffraction, Adisson-Wesley Publishing, U. S. A.,1978.

[20] I. Groma, F. Sz�ekely, Analysis of the asymptotic properties of x-ray linebroadening caused by dislocations, J. Appl. Crystallogr. 33 (6) (2000)1329e1334.

[21] B.E. Warren, B.L. Averbach, The effect of cold-work distortion on x-ray pat-terns, J. Appl. Phys. 21 (6) (1950) 595e599.

[22] V.D. Mote, Y. Purushotham, B.N. Dole, Williamson-hall analysis in estimationof lattice strain in nanometer-sized zno particles, J. Theor. Appl. Phys. 6 (1)(2012) 1e8.

[23] P.L. Raffo, T.E. Mitchell, Yielding, Work Hardening, And Cleavage In Tanta-lumeRhenium Alloy Single Crystals, Technical report, Lewis Research Center,Cleveland, 1968.

[24] Amit Misra, Ludovic Thilly, Structural metals at extremes, MRS Bull. 35 (12)(2010) 965e976.

[25] Roger Naslain, Design, preparation and properties of non-oxide cmcs forapplication in engines and nuclear reactors: an overview, Compos. Sci.Technol. 64 (2) (2004) 155e170.