Micro structure engineering

Interface microstructure engineering by high power impulse magnetron sputtering forthe enhancement of adhesionA. P. Ehiasarian, J. G. Wen, and I. Petrov Citation: Journal of Applied Physics 101, 054301 (2007); doi: 10.1063/1.2697052 View online: http://dx.doi.org/10.1063/1.2697052 View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/101/5?ver=pdfcov Published by the AIP Publishing Articles you may be interested in ZrN coatings deposited by high power impulse magnetron sputtering and cathodic arc techniques J. Vac. Sci. Technol. A 32, 031507 (2014); 10.1116/1.4869975 Steady state discharge optimization in high-power impulse magnetron sputtering through the control of themagnetic field J. Appl. Phys. 111, 023301 (2012); 10.1063/1.3673871 Structure and properties of ZrN coatings deposited by high power impulse magnetron sputtering technology J. Vac. Sci. Technol. A 29, 011004 (2011); 10.1116/1.3520640 Evolution of the plasma composition of a high power impulse magnetron sputtering system studied with a time-of-flight spectrometer J. Appl. Phys. 105, 093304 (2009); 10.1063/1.3125443 Influence of the bias voltage on the structure and mechanical performance of nanoscale multilayer Cr Al Y N ∕ CrN physical vapor deposition coatings J. Vac. Sci. Technol. A 27, 174 (2009); 10.1116/1.3065675

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:

129.78.233.210 On: Sun, 08 Feb 2015 22:42:48

Interface microstructure engineering by high power impulsemagnetron sputtering for the enhancement of adhesion

A. P. Ehiasariana�

Materials and Engineering Research Institute, Sheffield Hallam University, Howard Street,Sheffield S1 1WB, United Kingdom

J. G. Wen and I. PetrovFrederick Seitz Materials Research Laboratory, University of Illinois, Urbana, Illinois 61801and Materials Science Department, University of Illinois, Urbana, Illinois 61801

�Received 29 November 2006; accepted 4 January 2007; published online 2 March 2007�

An excellent adhesion of hard coatings to steel substrates is paramount in practically all applicationareas. Conventional methods utilize Ar glow etching or cathodic arc discharge pretreatments thathave the disadvantage of producing weak interfaces or adding droplets, respectively. One tool forinterface engineering is high power impulse magnetron sputtering �HIPIMS�. HIPIMS is based onconventional sputtering with extremely high peak power densities reaching 3 kW cm−2 at currentdensities of �2 A cm−2. HIPIMS of Cr and Nb was used to prepare interfaces on 304 stainless steeland M2 high speed steel �HSS�. During the pretreatment, the substrates were biased to Ubias

=−600 V and Ubias=−1000 V in the environment of a HIPIMS of Cr and Nb plasma. Thebombarding flux density reached peak values of 300 mA cm−2 and consisted of highly ionized metalplasma containing a high proportion of Cr1+ and Nb1+. Pretreatments were also carried out with Arglow discharge and filtered cathodic arc as comparison. The adhesion was evaluated for coatingsconsisting of a 0.3 �m thick CrN base layer and a 4 �m thick nanolayer stack of CrN/NbN witha period of 3.4 nm, hardness of HK0.025=3100, and residual stress of −1.8 GPa. For HIPIMS of Crpretreatment, the adhesion values on M2 HSS reached scratch test critical load values of LC

=70 N, thus comparing well to LC=51 N for interfaces pretreated by arc discharge plasmas and toLC=25 N for Ar etching. Cross sectional transmission electron microscopy studies revealed a cleaninterface and large areas of epitaxial growth in the case of HIPIMS pretreatment. The HIPIMSpretreatment promoted strong registry between the orientation of the coating and polycrystallinesubstrate grains due to the incorporation of metal ions and the preservation of crystallinity of thesubstrate. Evidence and conditions for the formation of cube-on-cube epitaxy and axiotaxy on steeland �-TiAl substrates are presented. © 2007 American Institute of Physics.�DOI: 10.1063/1.2697052�

INTRODUCTION

The adhesion of coatings to the substrates is a key factordetermining their performance in a specific application andsuccessful implementation in industrial production. The ad-hesion on substrates such as low- and high-carbon steels,metallic alloys, and carbides is influenced by a number offactors. One of the most important is surface contaminationin the form of surface oxides and organic substances fromthe environment which may prevent direct contact and bond-ing between the film and the substrate. Films growing on topof contaminated areas have poor adhesion as the bonding tothe substrates relies predominantly on the weak van derWaals forces. Furthermore, islands of contamination causeshadowing, formation of large-scale growth defects, and as-sociated film porosity. When contaminants are removed, thecrystalline substrate provides a defined chemistry that allowsmetallic, ionic, and covalent bonds to be established directlywith the growing film.

The adhesion strength of the interface can be furtherimproved by incorporation of metal into the substrate. It has

been shown1,2 that even 5–15 nm thick metal implantedzones can improve the adhesion of nitride films on steel andWC substrates by providing a gradient change in stress.

In order to remove contaminants, physical vapor deposi-tion �PVD� processes typically incorporate a pretreatmentstep prior to coating deposition. Traditionally, the pretreat-ment is performed in Ar glow discharge plasmas wherebycontamination is sputtered away by bombardment of gas ionswhich are accelerated to several hundred eV by a high biasvoltage on the substrates. The plasma density npl generated isof the order of 108 cm−3 and the ion flux to the substrates jS

is in the range �0.1 mA cm−2. The method can be imple-mented easily in industrial processes as Ar is often one of theprocess gases already installed. However, one disadvantageis that carbon-based contaminants have a low yield of physi-cal sputtering in Ar. Chemical sputtering with oxygen is farsuperior since carbon is released into the chamber as volatileCO and CO2. The physical sputtering yield of carbon is typi-cally an order of magnitude lower than that of metals, andtherefore surface roughening may develop as contaminatedregions are sputtered slower and thus remain higher thaninitially clean regions. Another disadvantage is the incorpo-a�Electronic mail: [email protected]

JOURNAL OF APPLIED PHYSICS 101, 054301 �2007�

0021-8979/2007/101�5�/054301/10/$23.00 © 2007 American Institute of Physics101, 054301-1

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:

129.78.233.210 On: Sun, 08 Feb 2015 22:42:48

ration of the gas ions into the substrate. As Ar is inert itoccupies interstitial sites whose presence induces increasedstrains in the substrate lattice. The high strain embrittles thesubstrate material by bringing it closer to its yield stress.Furthermore, when heated during pretreatment or under ex-ploitation, Ar diffuses and agglomerates into bubbles, whichintroduce porosity and weakening of the interface.

In order to create sufficient Ar bombardment flux, theplasma density during Ar pretreatment is often enhanced byworking at high gas pressures of the order of 10−2 mbarand/or by utilizing a radio frequency �rf� or midfrequency�MF� bipolar pulsing discharge. The plasma density npl

achieved with rf and bipolar pulsing is of the order of109 cm−3, and the ion flux to the substrates jS is in the rangeof 0.1–1 mA cm−2. The higher flux allows faster and moreefficient cleaning of the substrate and avoids oxide forma-tion. A disadvantage in the use of rf discharge systems is thattheir operation depends strongly on the load of the system.This complicates process control and may result in reproduc-ibility issues.

Superior adhesion can be obtained when etching in ca-thodic arc �CA� discharge environment.2,3,15 Cathodicvacuum arc discharges are characterized with high plasmadensities at the substrate of npl=1010 cm−3 and jS is in therange of 1–10 mA cm−2. CAs produce a highly ionizedmetal flux �up to 100% when combined with a filter� andmay be operated without intentionally introduced processgas. It has been shown that interfaces prepared with arc etch-ing have a high density and contain a 3–15 nm thick implan-tation zone in the substrate surface for substrate bias Ubias

�1.2 kV. Such interfaces have been shown to promote lo-calized epitaxial growth of the coating over individual grainsof polycrystalline steel and WC substrates.4 The disadvan-tage of unfiltered cathodic vacuum arc discharges is the pro-duction of liquid metal droplets with sizes of 100–1000 nmat the cathode. When embedded in the growing films, thesemacroparticles shadow the deposition flux, thus creating po-rosity and nucleating a larger scale growth defect as the coat-ing thickness increases.5

High power impulse magnetron sputtering6 �HIPIMS�has been shown to produce high ion fluxes with a high metalcontent similar to an arc discharge. HIPIMS is an impulse�short pulse� sputtering method where the peak power on thetarget reaches 3 kW cm−2 at a target current density of2 A cm−2. During the peak of the pulse jS

=50–500 mA cm−2 and npl=1013 cm−3. Fluxes generated byHIPIMS comprise high fractions of metal ions7 with gas ionto metal ion ratios of 1:1.8 Because it is a sputtering method,HIPIMS has the advantage that it does not produce droplets.It has been shown that single layer CrN films pretreated byHIPIMS have an excellent adhesion and superior perfor-mance in mechanical testing.9,10 Nanoscale multilayerCrN/NbN films have also been deposited on industrial scalemachines with improved adhesion and corrosion protectionproperties following HIPIMS pretreatment.11

HIPIMS has been shown to provide a powerful tool forsubstrate etching9,12 and interface engineering. However,atomic level insight on the interfacial structure and chemistryproduced by this method is still missing. The current paper

presents results from high-resolution transmission electronmicroscope �TEM� analysis of the microstructure of inter-faces prepared by HIPIMS and the subsequent growth of thefilm, which are compared with interfaces prepared by filteredcathodic vacuum arc. A CrN/NbN nanoscale multilayercoating is utilized as a model system. The adhesion of HIP-IMS pretreated coatings is compared with that of Ar glowand cathodic arc pretreated benchmark coatings. We showthat the crystallinity of the substrate grains is preserved dur-ing cleaning. The substrate surface is highly reactive andallows nucleation to proceed uniformly and in an orientedfashion over large areas where the bonding chemistry be-tween coating and substrate is not disturbed by contamina-tion. A high degree of local epitaxial growth is achievedusing the HIPIMS technique for pretreatment and depositionthat is comparable in nature to the interfaces obtained byCAs, but without the droplet problem encountered in an un-filtered arc and the limited productivity and upscalability of afiltered arc.

EXPERIMENTAL DETAILS

Pretreatment by high power impulse magnetronsputtering „HIPIMS…

The experiments with HIPIMS pretreatment were carriedout in an industrial size HTC-1000/4 ABS system �HauzerTechno Coating, The Netherlands� shown schematically inFig. 1. The chamber was equipped with four rectangular tar-gets with an area of 1200 cm2, which can be used in eitherCA or unbalanced magnetron �UBM� sputtering mode byadjustment of permanent magnets behind the targets.13 AHIPIMS power supply HMP 2/4 �Advanced Converters�AC�, Warsaw, Poland� with a peak power capability of8 MW at 2 kV and arc suppression was additionally con-

FIG. 1. Cross sectional schematic of the Hauzer HTC 1000/4 ABS coatingdeposition unit.

054301-2 Ehiasarian, Wen, and Petrov J. Appl. Phys. 101, 054301 �2007�

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:

129.78.233.210 On: Sun, 08 Feb 2015 22:42:48

nected to one of the targets, allowing switching between CA/UBM and HIPIMS. The substrates were subject to threefoldrotation.

The coating deposition sequence is shown in Fig. 2. TheHIPIMS of Cr and Nb pretreatments utilized pulses withpeak power densities of 1600 W cm−2 applied at the target ata duty cycle of 1%. The 304 stainless steel and M2 highspeed steel substrates were biased to Ubias=−600 V andUbias=−1000 V. The substrate temperature was maintainedat 400 °C throughout the process by additional heaters andthermocouple monitoring.

The coating step was carried out by direct current �dc�sputtering in a reactive atmosphere of Ar and N2 and com-prised the deposition of a 0.3 �m thick CrN base layer fol-lowed by a 4 �m thick nanolayer structured CrN/NbN witha period of 3.4 nm. During deposition the total pressure was0.4 Pa and the substrate bias was Ubias=−75 V. The sub-strates were heated to 400 °C. Further details are given byHovsepian et al.14

Pretreatment in Ar glow discharge

The Ar pretreament was performed at a dc bias voltageof Ubias=−1000 V and an Ar pressure of 8�10−3 mbar. Thefour magnetrons were operated behind shutters at a lowpower of 0.5 kW to increase the plasma density near thesubstrates. The deposition of the CrN/NbN nanoscalemultilayer coating was performed in the same equipment andat identical conditions as for HIPIMS pretreatment.

Pretreatment by unfiltered steered cathodic arc

The CA pretreatment was carried out at an arc current of100 A dc. The substrates were biased to Ubias=−1200 V forsubstrate surface cleaning. The CA was operated in pure Aratmosphere at a pressure of 0.12 Pa. The substrates wereheated to a temperature of 400 °C, which increased to500 °C during CA pretreatment. The deposition of theCrN/NbN nanoscale multilayer coating was performed inthe same equipment and at identical conditions as for HIP-IMS pretreatment.

Pretreatment by filtered cathodic arc

CrN coatings were prepared by a process comprisingthree steps—heating, pretreatment by Cr filtered cathodic arc�FCA�, and coating deposition of CrN by reactive magnetronsputtering as outlined in the flow chart in Fig. 2. The coatingdeposition process was carried out in a laboratory vacuumchamber equipped with one dc FCA source and one dc mag-netron sputtering source �3 in. diameter�. The base pressurewas 10−4 Pa �10−6 mbar�. The arc filter was a 90° bend withan internal coil diameter of 80 mm. The substrate was sta-tionary during pretreatment and deposition. The substratewas attached to a holder box and was heated from the back.The temperature was monitored in situ with a thermocoupleattached to the side of the sample.

The samples were heated to 250 °C. The pretreatmentwas carried out for 4 min with an arc current of 40 A dc anda filter current of 350 A dc. The substrate bias was Ubias

=−1200 V and the total substrate current density was2 mA cm−2. Argon was introduced at a low pressure of0.1 Pa�1�10−3 mbar� to improve stability of the discharge.After the pretreatment step the substrate holder box was ro-tated such that the substrate faced the magnetron source andthe deposition step was started. During deposition, theAr–N2 partial pressure ratio was PAr: PN2=1:1. The flow ofN2 was adjusted to maintain a constant total pressure of0.4 Pa�4�10−3 mbar�. The substrate bias was −75 V.

Further details are given in Ref. 15.

Plasma and materials analyses

Diagnostics of the plasma composition at the substrateposition at 200 mm from the cathode was carried out byenergy-resolved mass spectroscopy utilizing a PSM003 in-strument �Hiden Analytical Ltd.�. The spectrometer wasgated to collect data through the entire on time of the HIP-IMS power pulse while off times were excluded.

FIG. 2. Coating deposition sequence with pretreatment by HIPIMS �upper�and filtered cathodic arc �lower�.

054301-3 Ehiasarian, Wen, and Petrov J. Appl. Phys. 101, 054301 �2007�

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:

129.78.233.210 On: Sun, 08 Feb 2015 22:42:48

The ion flux to the substrates was measured with a flatelectrostatic probe with a diameter of ø20 mm and was re-corded with an oscilloscope to estimate the peak current dur-ing the pulse.

The chemistry of the coating-substrate interface was in-vestigated with a scanning transmission electron microscopyenergy dispersive x-ray �STEM-EDX� analysis technique. Adedicated STEM VG HB501 system with a field emissiongun was used. The electron beam width was �0.3 nm, pro-viding a spatial resolution for the analysis of 1 nm with pro-vision for some spread within the sample.

The microstructure of the coating-substrate interface wasobserved by cross-sectional transmission electron micros-copy �XTEM� �high-resolution TEM JEOL 2010F�. Thecoating adhesion was determined in scratch testing �CSEM,Revetest� by optical observation of the scratch. The criticalload LC was defined as the first occurrence of the adhesivefailure �spallation� of the film.

High speed steel �HSS� disks polished to Ra=0.03 �mwere used to evaluate adhesion. TEM cross sections werecarried out on polished 304 stainless steel squares.

RESULTS

Plasma properties

The magnitude of the ion flux was estimated with flatelectrostatic probe measurements. During Ar pretreatmentthe current to the substrates was jS=0.2 mA cm−2. For theCA pretreatment jS=3 mA cm−2. During HIPIMS pretreat-ment, the peak pulse current to the substrates reached jS

=300 mA cm−2. The average current was jS=3 mA cm−2

�duty cycle of 1%�. During coating the current was againjS=3 mA cm−2.

The chemistry of the ion flux during the HIPIMS pre-treatment and UBM deposition process was estimated withenergy-resolved mass spectroscopy. Figure 3 shows a mass

spectrum collected at the substrate position during pretreat-ment by HIPIMS of Cr. The strongest peaks are from thesingly ionized Cr1+ and Ar1+. It is interesting to note that themetal ion to gas ion ratio is approximately 1. In contrast, forconventional UBM sputtering, the ratio was measured to be0.1 for the same average power and gas pressure. Doublycharged species of Cr2+ and Ar2+ were also found in HIPIMSat levels of 10% and 5%, respectively. The metal-to-gas ionratio observed for 2+ is higher than that for 1+ due to thelower ionization potential of Cr2+ compared to Ar2+. Theinset in Fig. 3 shows the energy distribution for singlycharged ions; the doubly charged species had similar distri-butions �not shown�. The average energy of all ions was�2 eV with 95% of ions having energy less than 10 eV.

A similar high metal ion content was observed in HIP-IMS of Nb targets at the power densities discussed in thispaper. Optical emission spectroscopy �OES� observations10

have demonstrated that the metal-to-neutral ion ratio wasconsiderably higher for HIPIMS than for conventional UBM.

During coating deposition by conventional UBM, the ionflux was dominated by gas species with the following con-centrations: N+ �63%�, N2

+ �22%�, and Ar1+ �1%�. The Crmetal ion content at the substrate position was �1%.

Interface: Chemical composition

The chemical composition of the coating-substrate inter-face was investigated for a system comprising three compo-nents: SS substrate, HIPIMS-Nb pretreatment, and CrNdeposition. In this combination the incorporated element�Nb� is not present in the coating or substrate, which allowsthe spatial distribution to be well defined. The substrateswere biased to Ubias=−1000 V in a HIPIMS of the Nbplasma environment.

Figure 4 shows the atomic concentration of elements asa function of distance from the interface. At the very inter-

FIG. 3. Mass and energy spectra of the substrate ion flux generated by aHIPIMS of Cr discharge.

FIG. 4. Chemical composition of steel-CrN interface pretreated with HIP-IMS of Nb at Ubias=−1000 V. Hollow symbols mark concentration in at. %.Stars represent the Nb:Fe ratio.

054301-4 Ehiasarian, Wen, and Petrov J. Appl. Phys. 101, 054301 �2007�

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:

129.78.233.210 On: Sun, 08 Feb 2015 22:42:48

face is a region with width of �5 nm, where Nb is incorpo-rated. The concentration of Ar is very low at �1 at. %.

The apparent ion range and content of Nb can be ex-plained by the magnitude and chemistry of the ion flux. At abombarding voltage of Ubias=−1000 V two phenomenaoccur—substrate resputtering and ion implantation. The sub-strate resputtering rate is significant because the sputter yieldat 1000 eV is high, and at the same time the ion bombard-ment flux at the 300 mA cm−2 peak is substantial.

In conjunction, bombarding ions may be incorporatedinto the substrate to a significant depth of a few nanometers.Gas and metal ions have different behaviors when implanted.Implanted Ar gas atoms are inert and may be incorporated asinterstitials in the substrate lattice or may interact with va-cancies generated by the high energy bombardment and thusbecome substitutional. In either case the incorporation of Arcauses local increases in stress, which in turn embrittles theinterface as it brings it closer to the yield stress.

In contrast, metal ions have a high bonding affinity andcan be incorporated at lattice sites of the substrate as replace-

ments. In the case of Nb implantation into Fe, the phases�-Fe2Nb and �-FeNb are likely to form. In some systemssuch as Cr–Fe the solubility is full and there is no limit of theconcentration that can be incorporated.

The ratio of Nb:Fe is plotted in Fig. 4. The maximumratio found at the interface is 1.5, corresponding to a solidsolution of the �-FeNb phase and pure Nb. It should benoted that in this measurement Nb and Fe are always presentat the same time, indicating that a pure Nb layer is not de-tected.

To elucidate the experimental observations of implantedNb, dynamic transport of ions in matter �TRIM� simulationswith the TRIDYN software16 were performed. The resultsshowed that for a metal-to-gas ion ratio J�Ar1+� :J�Nb1+�=1:1 and Ubias=−1000 V, no layer of Nb is formed. BothNb and Ar are implanted in the steel substrate to a maximumion range of approximately 4 nm. The simulation also showsthat the probability of backscattering for Ar is a factor of 100greater than for Nb. For Cr+Ar and V+Ar combinations thedifference is a factor of 3. The number of reflected projectiles

FIG. 5. Cross sectional views of interfaces prepared by �a� HIPIMS Cr pretreatment, Ubias=−600 V, �b� HIPIMS Cr pretreatment, Ubias

=−1000 V, �c� HIPIMS Nb pretreatment, Ubias=−1000 V, and �d� Cr FCA pretreatment, Ubias=−1200 V.

054301-5 Ehiasarian, Wen, and Petrov J. Appl. Phys. 101, 054301 �2007�

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:

129.78.233.210 On: Sun, 08 Feb 2015 22:42:48

depends on the atomic weight of the substrate material andthe implanted layer with heavier substrate materials reflect-ing more. In addition to higher reflection, the inert nature ofAr and the presence of high energy metal ion bombardmentfavor radiation enhanced outdiffusion of the gas. Thus, ex-perimentally, no detectable amounts of Ar are retained in theimplanted layer. On the other hand, Nb—and also other met-als in general—is highly reactive and is retained morereadily. The simulation results on ion range and incorpora-tion probability are consistent with the experimental mea-surement of interface chemistry presented in Fig. 4. Themeasured Nb concentration profile may thus be explained bythe material implanted in the substrate. This implantation is aresult of the high metal content in the ion flux and the highenergy of bombardment given by the substrate voltage. Themicrostructure resulting from this incorporation is discussedin the following sections.

Interface: Microstructure

The microstructure of interfaces pretreated by HIPIMSof Cr and Nb was compared to the pretreatment by FCA ofCr.

Figure 5�a� shows the interface between the CrN baselayer and the stainless steel substrate pretreated with Ubias

=−600 V, which appears abrupt with a well defined bound-ary. At higher magnifications a band of a speckled contrastwith a thickness of �5–10 nm can be discerned within thesubstrate immediately below the interface, which is outlinedin the figure. This band is produced during the pretreatmentstep and is a result of the high energy ion bombardment. Thespeckled contrast arises from lattice strain due to radiationdamage induced by the metal ion implantation. The interfaceis generally clean and contains no foreign or amorphousphases due to contamination.

A similar structure is also observed for the substratespretreated at Ubias=−1000 V in HIPIMS of the Cr environ-ment. Figure 5�b� shows the interface region and a band withspeckled contrast. The band here is larger �10–15 nm� thanfor the pretreatment at Ubias=−600 V, which is due to thehigher energy and an associated longer range of implantationof bombarding ions.

Figure 5�c� is a cross section of the interface prepared bypretreatment in HIPIMS of the Nb environment at Ubias

=−1000 V. The interface is clean and dense. A band ofspeckled contrast similar in width as HIPIMS Cr pretreat-ment at Ubias=−1000 V is visible.

The pretreatment by FCA resulted again in a clean inter-face with a 10 nm thick band with speckled contrast, asshown in Fig. 5�d�. The thickness of the modified area isconsistent with the high energy of bombardment and indi-cates a zone of high defect density.

The four interfaces presented in Figs. 5�a�–5�d� and dis-cussed above are clean and dense as a result of the highlyintensive bombardment flux containing high fractions ofmetal ions. It can be expected that these interfaces provide adirect contact between the coating and the substrate and thus

promote the formation of the metallic, covalent, or ionicbonds as opposed to the significantly weaker van der Waalsbonds.

Local epitaxy on large areas: HIPIMS pretreatment

Larger lengths of the substrate-coating interface were ex-amined. The interfaces considered were HIPIMS of Nb pre-treated at Ubias=−1000 V. Figure 6�a� shows a sample regioncontaining a grain boundary in the substrate. The sample istilted such that a zone axis of the left grain is aligned withthe beam, thus exciting that grain. At the same time the rightgrain remains away from its zone axis and is therefore notexcited. The dark field image of this area �Fig. 6�b�� showsclearly a bright contrast from the left substrate grain, whilethe one on the right is completely dark. The white contrast ofthe excited left substrate grain is transferred across the inter-face to the coating grains, indicating that the coating has asimilar orientation as the substrate. The right hand side sub-strate grain is fully dark �no excitation�, and so is the coatingstructure above. From the dark field image it is evident thatthe switch in coating orientation is localized exactly at thesubstrate grain boundary where the switch in substrate orien-tation also occurs.

FIG. 6. TEM cross section of a CrN base layer grown on 304 stainless steelafter HIPIMS Nb pretreatment at Ubias=−1000 V. �a� Bright- and �b� darkfield images at the interface show the influence growth from the substrate.�c� The left and right selected-area diffraction patterns are from A+B regionand C+D regions, respectively.

054301-6 Ehiasarian, Wen, and Petrov J. Appl. Phys. 101, 054301 �2007�

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:

129.78.233.210 On: Sun, 08 Feb 2015 22:42:48

The particular orientation of the substrate and coatinggrains was evaluated with selected-area diffraction patterns�SADPs�. Figure 6�c� is SADP taken from the interface be-tween grains A and B marked in Fig. 6�a�, i.e., to the left ofthe substrate grain boundary. The diffraction pattern shows

double reflections indexed as �11̄2� zone axis of both filmand substrate grains. The sharp reflections at the larger dis-tance from the transmitted beam �smaller lattice size aSS

=0.359 nm� belong to the fcc stainless steel substrate. Theset of reflections at a smaller distance �larger lattice sizeaCrN=0.415 nm� belongs to the CrN base layer. In contrast tothe substrate reflections, the coating ones are more arclikewith a relatively small full width at half maximum �FWHM�intensity of 2.5° due to mosaicity in the in the film grain. Thediffraction pattern shows unambiguously the cube-on-cubeepitaxial orientation between the substrate and coating withgood alignment between substrate and film, which is particu-larly interesting given the large lattice mismatch of 15%. Thesubstrate grain B is oriented with a low index direction of�111� parallel to the substrate normal which may facilitatethe good local epitaxy. Figure 6�d� is a SADP taken fromgrains C and D in Fig. 6�a�, i.e., to the right of the substrategrain boundary. Here, the substrate grain is oriented with a

�332̄� zone axis parallel to the beam direction, while the lowindex 113 direction is at 21° with respect to the surface nor-mal. The film zone axis is 001, with a parallel beam direction

of �332̄�. The 220 reflections of the coating and the substrate

FIG. 7. Low magnification cross sectional dark field image of the interface.

FIG. 8. Cross sectional lattice imaging of interfaces prepared by �a� HIPIMS of Cr pretreatment, Ubias=−600 V, �b� HIPIMS of Cr pretreatment, Ubias

=−1000 V, �c� HIPIMS of Nb pretreatment, Ubias=−1000 V, and �d� Cr FCA pretreatment, Ubias=−1200 V.

054301-7 Ehiasarian, Wen, and Petrov J. Appl. Phys. 101, 054301 �2007�

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:

129.78.233.210 On: Sun, 08 Feb 2015 22:42:48

are mutually aligned. The film reflection in this case are pro-nounced arcs with a disorientation of �5°. We find that thesituation presented in Fig. 6 is typical, where substrate grainsexposing low index planes promote cube-on-cube epitaxywhile high index planes promote epitaxy on one coincidentset of planes. The latter case is known as axiotaxy.19 It hasbeen shown that in cases when the substrate and coatingplanes have a high lattice mismatch, a clean interface withintimate bonding can still force the first monolayers to crys-tallize in a structure similar to the substrate. The growthproceeds in a tilted direction at an angle � to the normal suchthat the atomic distance in the coating will match that of thesubstrate. The angle is given by cos���=b /a, where a is thelattice spacing of the substrate and b is the lattice spacing ofthe coating.

Figure 7 shows a larger area of the interface region in adark field imaging mode taken with an aligned substrate-filmpair of reflections, similar to the one presented in Fig. 6�d�.The interface region is highly excited and shows a whitecontrast that encompasses the coating and substrate. Signifi-cantly, the contrast is distributed over the whole length ofobservation of 3 �m, indicating a good alignment betweencoating and substrate along the full length of the observedarea. Above several hundred nanometers into the coating thevaried contrast indicates that the defective epitaxial grain hasbroken into polycrystalline columnar grains.

Lattice imaging of the interface region reveals furtherdetails of bonding.

Lattice imaging of the interface and local epitaxy

Figure 8�a� shows a lattice image of an interface pre-pared by HIPIMS of Cr with Ubias=−600 V. Atomic columnsare resolved in both the coating and the substrate, indicatingatomic registry between the two lattices. The interface can bedistinguished by the apparent difference in angle of orienta-tion between the lattice fringes of the substrate and the coat-ing. The crystalline structure is preserved throughout the in-terface region. The atomic planes of the coating are seen tobe in direct contact with the atomic planes of the SS sub-strates throughout the imaged area, thus signifying a fullydense interface. This level of contact can be compared to agrain boundary in bulk materials. No apparent gas accumu-lation and bubble formation were present.

Figure 8�b� shows a lattice image of the interface pre-pared at Ubias=−1000 V in HIPIMS of the Cr environment.The interface microstructure is modified compared to theUbias=−600 V case. In the substrate a damage zone of5–10 nm is visible by a slightly disturbed order in the latticestructure near the interface. Still, looking at the coating, theatomic columns are clearly resolved along both lattice vec-tors, thus indicating crystallographic registry to the substrate.Such a close match can rarely be obtained by chance espe-cially in randomly oriented substrates and when, for ex-ample, an amorphous �or amorphized� layer prevents the in-fluence of substrate on the coating.

The orientation alignment shown above was also ob-served for HIPIMS Nb pretreatment. As shown in Fig. 8�c�,the lattice fringes of the SS substrate are transferred throughthe HIPIMS interface to the CrN coating.

The registry in crystal orientation between the substrateand coating was visible over large areas of micrometer sizeof the sample, and the images are representative of the longrange structure despite showing an area of some tens of na-nometers only. This is strong evidence pointing towards localepitaxy.

Clean interfaces were also generated by pretreatment byFCA. Figure 8�d� presents a lattice image of the substrate-coating interface. The interface is clean and no amorphouslayer is observed. The substrate lattice planes are disturbedand bent in the first 6 nm adjacent to the interface. A numberof misfit dislocations are observed at the very interface. Ingeneral, however, the interface is fully dense and is expectedto have a high bonding strength.

In all cases the substrate grain lattice is more ordered,while the film lattice fringes are somewhat disordered, indi-cating defective epitaxy. While Fig. 8�c� is from a regionwith a cube-on-cube epitaxy, the rest of the lattice-resolutionimages indicate a rotation of the coating planes to match theplanes of the substrate grains exposed to the surface in anaxiotaxial fashion.

Adhesion

The large-scale epitaxy discussed in the previous sectionis expected to translate into strong adhesion of the overallcoating as measured with scratch testing. Figure 9 presents acomparison between scratch test critical loads LC for identi-cal CrN/NbN nanolayered coatings where the interface wasprepared by three methods of pretreatment: Ar glow dis-charge, Cr CA, and HIPIMS of Cr. The LC values for HIP-IMS pretreatment are higher than those for arc pretreatment.

FIG. 9. Adhesion of CrN/NbN coatings on HSS substrate pretreated by Arglow discharge, cathodic arc, and HIPIMS discharges.

054301-8 Ehiasarian, Wen, and Petrov J. Appl. Phys. 101, 054301 �2007�

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:

129.78.233.210 On: Sun, 08 Feb 2015 22:42:48

The difference may be attributed in part to the more dis-turbed nature of the arc interface as described in the previoussection �Fig. 6�d��. The Ar pretreatment resulted in a lowadhesion due to the insufficient cleanliness of the interfaceand the embrittlement caused by the incorporation of Ar.

DISCUSSION

The structure and microchemistry of the pretreated sur-face are determined in part by the balance of metal ion im-plantation and resputtering. Diffusion plays a significant role.However, it is mainly affected by the temperature of thesubstrate20 and thus can be considered constant when theenergy and flux of ions change. The implantation depths ofsingly charged ions at the 600–1000 eV energies discussedare in the range of 5–8 nm. An analytical model by Carr etal.20 predicts that the in-depth distribution of implanted ionconcentration can be described with the logarithm of an errorfunction. The authors show that the surface and maximumconcentration is a function of resputtering efficiency and re-tention probability. The depth is controlled by the energy ofthe implanted species. In real processes, the rate of resput-tering is governed by the magnitude of the ion flux. Theenergy of ions is given by the product of the bias voltage andthe charge state. Thus, treatments that are carried out in thepresence of highly charged metal ions can produce deeperimplantation profiles. A typical example for discharges con-taining highly charged metal ions are CAs, where the aver-age charge state is 2.05 for Cr and 2.5 for Nb with the maxi-mum charge states for Nb of 5+. When electromagneticmacroparticle filters are used, higher ionization states are fa-vored due to the limited expansion of the cathode spotplasma by the magnetic field, leading to an extended time ofionization before equilibrium described as freezing of thecharge state distribution is reached.17 The average chargestate in the presence of magnetic field with shape similar tothat in a filter can increase to 3.5 and 4 for Cr and Nb,respectively.17 These highly charged metal ions gain signifi-

cant energy in the biased substrate sheath and can have avery high range. This is illustrated by the large implantedzone shown in Fig. 6�d�. Dynamic TRIM simulations16 predicta penetration depth of 8 nm for Cr2+ in steel, which propor-tionally increases with charge state, and for Cr4+ the ionrange is 15 nm. In this case the implanted species may causeirradiation damage relatively deep into the substrate. In somecases a nanocrystalline/amorphous layer can be formed.18

Such layers act as a crystallographically neutral substrate andthus inhibits epitaxy.

The apparent coherency of the coating lattice to that ofthe substrate presented in Figs. 6–8 may be explained by theintimate bonding between the nitride coating and the sub-strate. As the nitride layer nucleates, the clean surface pro-motes direct bonding of the adatoms to the substrate withindividual nuclei highly aligned over individual substrategrains due to crystallographic templating. During coales-cence the film nuclei merge without forming grain bound-aries. Thus, the grain size of the coating closely duplicatesthe structure of the substrate which typically is on the orderof micrometers. This mode of interface formation is mark-edly different from cases when the coating nucleates on sub-strate grains covered with a layer that is amorphized or finegrained because of contamination or excessive ion damage.In addition to forming weaker bonding across the interface,the coating nuclei are randomly oriented in the plane of thefilm and form high-angle grain boundaries during coales-cence, thus defining a column size in the nanometer or tensof nanometer range, as shown in Refs. 5 and 2.

Crystallographic templating by cube-on-cube epitaxy oraxiotaxy can be observed for a wide range of crystallinesubstrate-coating pairs as long as the interface is free of im-purities and amorphous contamination. For example, tem-plating was observed for several coating-substrate combina-tions, e.g., Ti0.5Al0.5N base layer on stainless steel withpretreatment by HIPIMS of V shown in Fig. 10�a� as well asfor a CrAlN layer grown on the �-TiAl substrate after HIP-IMS Cr pretreatment shown in Figs. 10�b� and 10�c�. The

FIG. 10. Cross sectional views of the interface region of �a� TiAlN grown on SS304 after HIPMS of V pretreatment Ubias=−1000 V and �b� CrAlN coatingon �-TiAl with pretreatment of HIPIMS of Cr at Ubias=−600 V.

054301-9 Ehiasarian, Wen, and Petrov J. Appl. Phys. 101, 054301 �2007�

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:

129.78.233.210 On: Sun, 08 Feb 2015 22:42:48

Ti0.5Al0.5N appears with uniform diffraction contrast in thebright-field TEM image �Fig. 10�a�� indicative of a largecrystal grain growth as an extension to the substrate grainwhich is analogous to the interface structure shown in Fig. 7.The dark bands within the film are bend contours within thesingle crystal. The STEM image in Fig. 10�b� shows that thecoating adjacent to the interface appears as a mirror image ofthe substrate. Figure 10�c� shows the crystalline epitaxial mi-crostructure at the interface with cube-on-cube epitaxial re-lationship. The �-TiAl substrate has a tetragonal structurewith parameters a=0.398 nm and c=0.407 nm, which areclose to the lattice parameters of CrN of aCrN=0.415 nm.Although no crystallographic data were found for the CrAlNphase, thin film growth studies using cathodic arc evapora-tion have shown that the structure is similar to CrN up to aCr:Al ratio of 2:3 and the lattice parameter is decreased to0.413 nm for arc deposited coatings.21 The small mismatchof 4% and 2% for a and c, respectively, is a strong conditionpromoting epitaxial growth.

In practice, substrate surfaces are contaminated andplasma pretreatment is a necessary step to achieve a highlevel of cleanness. It is beneficial to avoid inert gas ion etch-ing due to incorporation in the substrate and the associatedembrittlement due to high strains and coalescence into gasbubbles. When metal ions are used during pretreatment, theycan be implanted and incorporated in the substrate as re-placements, sometimes forming phases. Metal ion irradiationprovides intense mixing in the near surface region whichpromotes radiation enhanced diffusion which provides an es-cape path for the implanted inert ion which in the case ofHIPIMS are with a ratio of 1:1 with the metal gas ions. Metalion incorporation in a HIPIMS plasma environment pre-serves the crystallinity and provides a high density interfacewith a good adhesion strength.

CONCLUSIONS

The paper illustrates the utilization of the high powerimpulse magnetron sputtering �HIPIMS� technology for thepretreatment of substrates prior to nitride coating deposition.Substrates were pretreated under high energy bombardmentin a HIPIMS environment prior to coating deposition. Theinterface chemistry and microstructure were discussed in re-lation to the adhesion of the coatings.

HIPIMS was found to produce a highly ionized plasmacontaining equal amounts of metal and gas ions. The inter-face was free of contamination and amorphous phases suchas native oxides. This provided a fully crystalline microstruc-ture at the coating-substrate interface. The high content ofmetal ions in the bombarding flux modified the interface toinclude a metal implantation zone that promoted a strongbond between the substrate and the subsequently depositedcoating. The clean interface and metal implantation zonepromoted the alignment of the coating growth orientation to

the crystal orientation of the substrate in at least one direc-tion �axiotaxial growth�. The coating orientation was ob-served to follow that of the substrate over large areas ofseveral micrometers. It was also shown that the coating ori-entation mirrored changes in orientation at substrate grainboundaries.

Overall coatings deposited after HIPIMS pretreatmentexhibited superior adhesion in comparison to pretreatmentsin Ar glow discharge and cathodic vacuum arc environments.

ACKNOWLEDGMENTS

The authors gratefully acknowledge Professor PapkenHovsepian from Sheffield Hallam University, UK and Pro-fessor Peter Barna, MFA, Hungary for stimulating discus-sions. Dr. Andre Anders of Lawrence Berkeley NationalLaboratory is acknowledged for the pretreatment by filteredcathodic arc. The financial support of the EU project INNO-VATIAL is acknowledged. Two of the authors �J.G.W. andI.P.� are supported by U.S. Department of Energy underGrant No. DEFG02-91-ER45439.

1P. Hovsepian, Arch. Metall. 33, 4 �1988�.2C. Schonjahn, A. P. Ehiasarian, D. B. Lewis, R. New, W. D. Munz, R. D.Twesten, and I. Petrov, J. Vac. Sci. Technol. A 19, 4 �2001�.

3G. Hakansson, L. Hultman, J.-E. Sundgren, J. E. Greene, and W.-D.Muenz, Surf. Coat. Technol. 48, 1 �1991�.

4C. Schonjahn, L. A. Donohue, D. B. Lewis, W.-D. Munz, R. D. Twesten,and I. Petrov, J. Vac. Sci. Technol. A 18, 4 �2000�.

5I. Petrov, P. Losbichler, D. Bergstrom, J. E. Greene, W.-D. Munz, T. Hurk-mans, and T. Trinh, Thin Solid Films 302, 1 �1997�.

6V. Kouznetsov, K. Macak, M. Schneider, J. U. Helmersson, and I. Petrov,Surf. Coat. Technol. 122, 2 �1999�.

7A. P. Ehiasarian, R. New, W.-D. Munz, L. Hultman, U. Helmersson, andV. Kouznetsov, Vacuum 65, 2 �2002�.

8A. P. Ehiasarian, Y. A. Gonzalvo, and T. Whitmore, Plasma ProcessesPolym. �to be published�.

9A. P. Ehiasarian, W.-D. Munz, L. Hultman, U. Helmersson, and I. Petrov,Surf. Coat. Technol. 163–164, 267 �2003�.

10A. P. Ehiasarian, P. E. Hovsepian, L. Hultman, and U. Helmersson, ThinSolid Films 457, 2 �2004�.

11C. Reinhard, A. P. Ehiasarian, and P. Eh. Hovsepian, Thin Solid Films515, 3685 �2007�.

12W.-D. Munz, A. P. Ehiasarian, and P. E. Hovsepian, European Patent No.EP1260603B1 �2006�.

13W.-D. Munz, D. Schulze, and F. J. M. Hauzer, Surf. Coat. Technol. 50, 2�1992�.

14P. E. Hovsepian, D. B. Lewis, and W. D. Munz, Surf. Coat. Technol.133–134, 166 �2000�.

15A. P. Ehiasarian, A. Anders, and I. Petrov, J. Vac. Sci. Technol. A �sub-mitted�.

16W. Moller, W. Eckstein, and J. P. Biersack, Comput. Phys. Commun. 51, 3�1988�.

17E. M. Oks, A. Anders, M. R. Dickinson, and R. A. MacGill, IEEE Trans.Plasma Sci. 24, 1174 �1996�.

18C. Schonjahn, H. Paritong, W.-D. Munz, R. D. Twesten, and I. Petrov, J.Vac. Sci. Technol. A 19, 4 �2001�.

19C. Detavernier, A. S. Ozcan, J. Jordan-Sweet, E. A. Stach, J. Tersoff, F. M.Ross, and C. Lavoie, Nature �London� 426, 6967 �2003�.

20W. Carr, M. Seidl, G. S. Tompa, and A. Souzis, J. Vac. Sci. Technol. A 5,1250 �1987�.

21H. Hasegawa, M. Kawate, and T. Suzuki, Surf. Coat. Technol. 200, 2409�2005�.

054301-10 Ehiasarian, Wen, and Petrov J. Appl. Phys. 101, 054301 �2007�

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:

129.78.233.210 On: Sun, 08 Feb 2015 22:42:48