Characterisation of Carbide Particles in S-65 Beryllium by Scanning Kelvin Probe Force Microscopy

Christopher F. Mallinson*a, Ann Harveyb, John F. Wattsa

aThe Surface Analysis Laboratory, Department of Mechanical Engineering Sciences, University of Surrey, Guildford, Surrey, GU2 7XH, UK,

bLife Prediction Team, Metallurgy Group, Science Function, AWE, Aldermaston, Reading, RG7 4PR*c.mallinson@surrey.ac.uk

Abstract

Scanning Kelvin probe force microscopy has been employed to examine the behaviour of second phase carbide particles in beryllium at different relative humidity levels and after exposure to deionised water. Carbides are believed to have a role in the localised corrosion of beryllium as a result of their hydrolysis when exposed at the metal surface. The presence of beryllium carbide was confirmed by means of Auger electron spectroscopy and the particles were further characterised by scanning electron microscopy, energy/wavelength dispersive x-ray spectroscopy and scanning Kelvin probe force microscopy. The particles were found to have a more noble Volta potential than the beryllium matrix and a decrease in the Volta potential difference between the second phase particles and the matrix was observed as the humidity was increased. A thick beryllium oxide/hydroxide layer then formed on the particles following exposure to water significantly reducing their potential.

Introduction

Beryllium, like aluminium, is passivated by a native oxide layer ~3 nm thick [1], resulting in a susceptibility to localised corrosion in the form of pitting [2–4]. The initiation of pitting corrosion is believed to be associated with second phase particles of mixed size and composition [2]. Intermetallic particles are the most commonly reported phases to play a role in the localised corrosion of the metal [2,5]. Corrosion investigations giving attention to beryllium carbide particles are less common in the literature than those relating to other second phase particles [6]. This is likely a consequence of the improved manufacturing methods used for the current commercial grades of beryllium that results in a metal containing far fewer carbides [7].

One of the earliest studies on second phase particles in beryllium was based upon observations at Oak Ridge National Lab in 1947 [8]. Corrosion occurred on several pieces of beryllium which were exposed to the local atmosphere for approximately 6 months. During this time, deposits of white corrosion products formed which were examined by X-ray diffraction and found to be a mixture of hydrated beryllium oxide (BeO.xH2O) and/or beryllium hydroxide (Be(OH)2). These were caused by the hydrolysis of exposed beryllium carbide inclusions at the machined metal surface [8]. This type of corrosion has also been duplicated with carbide seeded beryllium [2].

Any beryllium carbide particles exposed at the metal surface are susceptible to hydrolysis. With the carbides reacting with moisture in the environment to form beryllium oxide/hydroxide as well as methane gas as indicated in Equations 1 and 2 [2].

Be2C + 2H2O → 2BeO + CH4 (1)

Be2C + 4H2O → 2Be(OH)2 + CH4 (2)

The hydrated beryllium oxides occupy a volume approximately four times greater than that of the carbide resulting in the formation of large blisters on the beryllium surface around the carbide. This

1

1

2345678

910111213141516171819

20

2122232425262728

293031323334353637383940414243444546

can cause precision components to be scrapped because of the disruption to their finely prepared surface [2]. One particularly interesting study examined the effect of an unusually large carbide particle present in a precision beryllium part [9]. The pocket of material containing the particle was ruptured during machining exposing part of the carbide at the surface. During storage the carbide reacted in accordance with Equations 1 and 2 leading to volume expansion and fracture of the matrix around the particle. This led to a large section of beryllium breaking away from the part resulting it being scrapped.

While the chances of a significantly large carbide particle occurring in material produced today are exceedingly low, beryllium carbide particles are still thought to be present in almost all available beryllium [7]. This is because carbon is present at concentrations of 700-1500 ppm in all commercial grades of beryllium, as shown by the values in Table 1 [10]. The resulting beryllium carbide has been observed to be evenly distributed as fine particles (0.25-4 µm) or as fewer coarse particles (5-10 µm) [6]. In this work the carbide particles were identified using energy dispersive x-ray spectroscopy (EDX) to analyse the particles composition. The particles contained significant amounts of carbon and low levels of other metals. Little or no beryllium was detected as a result of the poor sensitivity of EDX for Be Kα x-rays. The particles were identified as a mix of Be2C and mixed metal carbides [6]. Scanning electron microscopy performed before and after electrochemical polarisation of beryllium containing these carbides revealed that some corrosion pits initiated at the particles with the apparent corrosion of the particles. However, it was noted that pit initiation appeared to be more common at grain boundaries than carbide particles.

Recently scanning Kelvin probe force microscopy (SKPFM) has been used to investigate the Volta potential difference between various second phase particles and the matrix for S-65 beryllium [11]. However, because of their rarity, no carbide particles were included within this investigation as they could not be located in the test specimens. The results from the investigation showed that all of the second phase particles types present in the metal were nobler than the matrix. Additionally, it was shown that a number of factors affected the Volta potential of the particles which resulted in scatter of the potential values for a particular particle type. The standard deviation in the Volta potential for all particle types was found to be ~50 mV in this study [11]. These factors included: compositional variations, oxide layer thickness and the presence of a beryllium over-layer. It is unknown as to whether the carbide particles present in beryllium act as cathodes in a galvanic couple with the matrix as has been observed for other second phase particle compositions or if they only corrode by the hydrolysis mechanisms of Equations 1 and 2.

The presence of hydrocarbon contamination on the surface of a specimen has been found to be highly detrimental to SKPFM studies that involve the prior investigation of areas of interest by electron beam based techniques [12,13]. The contamination becomes pyrolised under the action of the beam resulting in a “black box” covering the area of interest. This carbon has a significant cathodic potential that prevents accurate determination of the true Volta potential [13]. As a consequence of this effect small carbide particles were initially “forfeited” for in-depth electron and x-ray analysis to confirm that the particles in this manuscript were carbidic in nature. This enabled the surface of the two largest carbide particles that were identified to remain free of contamination for the SKPFM analysis [14].

In this paper we investigate the behaviour and nobility of beryllium carbide particles with increasing humidity and following exposure to deionised water by means of SKPFM for S-65 beryllium.

Experimental

2

47484950515253545556575859606162636465666768697071727374757677787980

818283848586878889

9091

92

The sample used in this work was a cross sectioned 10 mm disc of S-65 beryllium bar, mounted in conductive Bakelite. The sample surface was prepared for analysis by grinding and polishing using diamond impregnated discs with a final 1 μm diamond paste. The typical chemical composition of the S-65 beryllium bar sample used in this work is given in Table 1 together with the compositions of other grades of beryllium.

All SKPFM analysis was performed in air. The humidity within the atomic force microscope (AFM) chamber was controlled by using a bubbler filled with deionised water and adjusting the flow rate of air through the water to between 2 and 10 l/min to achieve 55 and 75% relative humidity. To obtain the lowest humidity of 10% dry compressed air was directly fed into the AFM chamber at a rate of 10 l/min. To obtain 35% RH the AFM chamber door was opened to the ambient laboratory environment. The humidity levels were all kept to within 3% of the desired value using these methods. All measurements were acquired at 20 ±1°C and prior to analysis the sample was left to equilibrate in the chamber for 1 hour at each humidity level.

The morphology and the composition of the carbide particles was studied using a JEOL JSM-7100F scanning electron microscope (SEM) equipped with a Thermo Scientific Ultradry energy dispersive x-ray detector (EDX). Micrographs and EDX spectra were acquired using a primary beam energy of 15 keV and a specimen current of 5 nA. For the initial surface imaging to identify the regions of the surface containing particles, a short image acquisition time and a large field of view (FOV) was used to minimise the charge density at the sample surface [13].

Auger electron spectroscopy (AES) of the particles was performed using a Thermo Scientific MICROLAB 350 scanning Auger microscope. A primary beam energy of 10 keV and a beam current of 5 nA was used for the acquisition of point Auger spectra. AES survey and high resolution spectra were recorded with a retard ratio of 4.

The Volta potential maps of the sample surface were acquired with a Bruker Dimension Edge AFM using silicon tips coated with a conductive platinum-iridium layer. These were SCM-PIT-V2 tips purchased from Bruker. A scan rate of 0.2 Hz was used for the acquisition of all maps and a tip-sample distance of 100 nm was used during the Volta potential scans. The maps displayed in this paper were acquired using 512 lines. The Volta potential difference between the carbide particles and the matrix was measured by extracting values from a cross section analysis on the Volta potential maps.

All of the raw data showed the particles to have a lower Volta potential than the matrix which is in agreement with previous authors for more noble materials [15]. However, SKPFM potential data is often inverted to obtain the same polarity as the electrochemical potentials. As such here all the Volta potential maps and potential values are presented in their inverted form, in line with other authors [14,16–18].

Results and Discussion

The detection and confirmation of beryllium carbide particles is particularly challenging in x-ray analysis as a result of the difficulty in quantification of the light elements, because of their poor X-ray emission cross-sections. To ensure that the particles in this study were carbide particles AES was used. AES is particularly suited to the identification of carbides as the C KLL Auger transition reveals chemical state information, with the peak from a carbide showing a different shape compared to the peak from organic carbon or hydrocarbon contamination. Fig. 1 shows the C KLL Auger transitions from carbon contamination, a beryllium carbide reference spectrum [19], the spectrum acquired

3

9394959697

9899100101102103104105

106107108109110111

112113114115

116117118119120121122

123124125126127

128

129130131132133134135

from the surface of a possible carbide particle and a confirmed carbide particle. The spectrum from hydrocarbon contamination shows a broad single peak while the spectrum from the reference carbide shows a narrower primary peak with two additional smaller peaks at a lower kinetic energy, as noted by the small black arrows in the figure. The presence of these three peaks is characteristic for carbides in AES [20]. The spectrum from the particle is a close match to the carbide reference spectrum with a narrow primary intense peak and two weaker satellite peaks. The two satellite peaks in the spectrum from the confirmed carbide particle have a slightly larger energy spacing compared to those in the reference spectrum. This is likely a result of the oxide present on the carbide particle which affects the valence band density of states from which the electrons involved in the formation of these peaks originate [21].

Following an initial imaging of the surface using a low beam current and large imaging FOV a number of possible carbide particles were identified in a region of the sample surface. These ranged in size from 2 - 80 μm in diameter. The smaller FOV micrographs shown in this paper were acquired after SKPFM analysis. Fig. 21 (a) and (b) show secondary electron (SE) and backscattered electron (BSE) micrographs from one of these carbide particles. BSE imaging was used to highlight the compositional variations within the bulk of the particle. EDX analysis was performed on this region to identify the carbide particle and the discrete particle in the bottom left of the images. The elemental maps are shown in Fig. 2. In both the micrographs the contrast between the beryllium matrix and the carbide particle is quite low, while the 5 µm silicon particle (as determined by EDX mapping and point spectroscopy) in the bottom left of the micrograph is noticeably brighter. This is a result of the low atomic mass difference between the beryllium matrix and the carbide particle.

The micrographs highlight the difficulty that can be faced when attempting to locate these particles within the matrix compared to other higher mass second phase particles. Previously reported results of an automated SEM/EDX analysis routine designed to characterise and count the secondary phase particle distribution in beryllium have not revealed beryllium carbide particles to be present [13,19]. The reason for this is now clear. While the carbides are an uncommon particle type, the contrast image thresholding required to perform the automated analysis will blur the low contrast carbides into the background of the image preventing them from being accurately identified and counted.

As observed in the micrographs in Fig. 2 the carbide particles contain regions of isolated higher mass material which are responsible for the spots of brighter contrast in the images. The presence of these large (0.5 - 1 μm) and small (<0.1 μm) inclusions within the bulk of the carbide particles explain the results from point EDX analysis results in the literature which appear to show the particles as mixed carbides [6]. The analysis volume of EDX within the particle is sufficiently large (>> 1µm3) so as to envelope a number of these higher mass particles including them within the resulting spectrum making the carbide particle appear to contain these impurity elements. To ensure that the small iron and aluminium containing particles were not ternary iron-aluminium carbides, low current high resolution AES was performed on a number of the particles [20]. This revealed them to be AlFeBe4 precipitates and did not appear to contain carbon.

The EDX spectrum from the centre of the carbide from Fig. 1, is shown in Fig. 3 and the analysis position is shown by the white circle in the SEM micrograph in Fig. 42. The spectrum reveals how difficult positive confirmation of the particle can be using x-ray analysis. The Be Kα x-rays are readily absorbed and attenuated within the sample and so the Be Kα peak is not observed in the particle spectrum. The spectrum does show intense C Kα and O Kα peaks with moderately intense Al Kα and Si Kα peaks. The presence of the intense oxygen peak as well as the weaker aluminium and silicon peaks and absence of the beryllium peak add to the difficulty in confirming the particle to be a carbide. The positive identification of the particle as a carbide is only possible by the combination of

4

136137138139140141142143144145

146147148149150151152153154155156

157158159160161162163

164165166167168169170171172173

174175176177178179180181

AES and EDX in addition to the EDX maps in Fig. 4. The correct conclusions from the EDX spectrum are taken as a beryllium carbide particle covered with an oxide/hydroxide layer which contains small oxide particles, silicon particles and AlFeBe4 precipitates which are dispersed throughout the bulk. In comparison to the literature which suggests that these particles are mixed metal carbides [6].

EDX mapping reveals the constituent elements of the larger particles. The EDX maps acquired from the carbide particle shown in Fig. 12 are shown in Fig. 24. The maps highlight the outline of the particle in the carbon x-ray map, revealing that the particle is not a compact mass but is irregular and contains voids. The beryllium map shows the extent of the attenuation of the particularly low kinetic energy Be Kα x-rays as no counts are observed from the centre of the particle while the matrix shows moderate intensity. The oxygen map tracks the carbide particle outline which is likely as a result of the oxide layer on the particle. The grain boundaries in the same region are also highlighted because of the oxide particles present at them. There are also discrete points of high intensity. These may be caused by oxide particles within the carbide bulk or by oxygen associated with the other small second phase particles within the carbide. The aluminium and iron maps match closely with overlapping points of intensity at the locations of small AlFeBe4 precipitates within the bulk carbide. A weak aluminium and iron signal is observed in both of these maps from the entire majority of the carbide region. This is likely caused by the <100 nm particles that are well dispersed within the particle but too small to pinpoint by EDX, contributing to the blurring of the element x-ray maps. The silicon map also shows discrete points of intensity at the positions of small particles within the carbide.

The detection and confirmation of beryllium carbide particles is particularly challenging in x-ray analysis as a result of the difficulty in quantification of the light elements, because of their poor X-ray emission cross-sections. To ensure that the particles in this study were carbide particles AES was used. AES is particularly suited to the identification of carbides as the C KLL Auger transition reveals chemical state information, with the peak from a carbide showing a different shape compared to the peak from organic carbon or hydrocarbon contamination. Fig. 4 shows the C KLL Auger transitions from carbon contamination, a beryllium carbide reference spectrum from the literature [21], the spectrum acquired from the surface of a suspected carbide particle is shown in Fig. 5a and a spectrum from a confirmed carbide particle which is shown in Fig. 1 The spectrum from hydrocarbon contamination shows a broad single peak while the spectrum from the reference carbide shows a narrower primary peak with two additional smaller peaks at a lower kinetic energy, as noted by the small black arrows in the figure. The presence of these three peaks is characteristic for carbides in AES [22]. The spectrum from the particle is a close match to the carbide reference spectrum with a narrow primary intense peak and two weaker satellite peaks. The two satellite peaks in the spectrum from the confirmed carbide particle have a slightly larger energy spacing compared to those in the reference spectrum. This is likely a result of the oxide present on the carbide particle which affects the valence band density of states from which the electrons involved in the formation of these peaks originate [19].

The detection and confirmation of beryllium carbide particles is particularly challenging in x-ray analysis as a result of the difficulty in quantification of the light elements, because of their poor X-ray emission cross-sections. To ensure that the particles in this study were carbide particles AES was used. AES is particularly suited to the identification of carbides as the C KLL Auger transition reveals chemical state information, with the peak from a carbide showing a different shape compared to the peak from organic carbon or hydrocarbon contamination. Fig. 1 shows the C KLL Auger transitions from carbon contamination, a beryllium carbide reference spectrum [19], the spectrum acquired from the surface of a possible carbide particle and a confirmed carbide particle. The spectrum from

5

182183184185

186187188189190191192193194195196197198199200201

202203204205206207208209210211212213214215216217218219

220221222223224225226227

hydrocarbon contamination shows a broad single peak while the spectrum from the reference carbide shows a narrower primary peak with two additional smaller peaks at a lower kinetic energy, as noted by the small black arrows in the figure. The presence of these three peaks is characteristic for carbides in AES [20]. The spectrum from the particle is a close match to the carbide reference spectrum with a narrow primary intense peak and two weaker satellite peaks. The two satellite peaks in the spectrum from the confirmed carbide particle have a slightly larger energy spacing compared to those in the reference spectrum. This is likely a result of the oxide present on the carbide particle which affects the valence band density of states from which the electrons involved in the formation of these peaks originate [21].

Unambiguous identification of the carbide particles in this work was made using the combination of SEM/EDX/WDX and AES. Reliance upon SEM and EDX results for the identification of beryllium carbides can produce inaccurate results. This is believed to be the case for the coarse carbides previously discussed in the literature [6], which do not appear to consist of beryllium carbide. A number of example carbides were identified at AWE using the same method as previously outlined for a billet sample of S-65 beryllium [6]. SEM micrographs from four such coarse particles identified in this manner are shown in Fig. 5. These particles were subsequently investigated using a combination of SEM/EDX/WDX and AES at the University of Surrey. In each case they were later confirmed not be carbidic in nature. Instead each one was found to contain carbon, in an organic form as revealed by the C KLL spectrum “possible carbide Fig. 5a”, shown in Fig. 4, as well as beryllium oxide and other inorganic contaminates. They are believed to be artefacts of the mounting and polishing procedure.

The EDX maps of carbon, oxygen and beryllium for the particle shown in Fig. 5a, are shown in Fig. 6 and appear different to the EDX maps acquired from the confirmed carbide particle, in Fig. 1. Areas of the particle are rich in carbon while others are rich in oxygen compared to the more uniform intensity of carbon and isolated patches of oxygen intensity observed for the confirmed carbide. This is observed for a number of the particles that are initially identified as possible carbides. The EDX and WDX spectra from the carbon rich region of the particle are shown in Fig. 7. The EDX spectrum shows an intense C Kα peak and a weak O Kα peak. The overlaid WDX spectrum also shows an intense C Kα peak. The Be Kα peak region was also acquired using WDX, however the intensity of the peak was so low that it has been shown as an inset in the figure. Despite the high sensitivity for beryllium in WDX few Be Kα counts were acquired from this region showing that little beryllium is present in the carbon rich region of the particle. To further investigate the nature of the carbon in the particle AES was performed and high resolution carbon Auger spectrum acquired from particle is shown in Fig. 4 and is labelled as “possible carbide Fig. 5a”1. In contrast to the carbide reference spectrum, from the literature, and the spectrum acquired from the confirmed carbide (Fig. 1) the spectrum from the carbon rich region of the particle appears the same as that acquired from hydrocarbon contamination. This combined with the WDX data reveals that the carbon rich region of particle shown in Fig. 5a is not carbidic but consists of organic carbon.

The series of potential maps acquired from the an largest 80 µm carbide particle, which was the largest carbide particle identified by SEM during this investigation, are shown in Fig. 8 a - d. The four maps show a reduction in the contrast between the particle and the surrounding matrix as the humidity increases from 10% to 75%. This is a result of the decreasing Volta potential difference between the particle and the matrix. The position at which the Volta potential was measured is shown in (a) by the white arrows. The Volta potential was extracted using a cross section analysis tool on the map. For each subsequent map the position of the cross section was maintained. The

6

228229230231232233234235236

237

238239240241242243244245246247248249

250251252253254255256257258259260261262263264265266

267268269270271272273

cross section analysis showed that there is a gradual reduction in the Volta potential difference between the matrix and the particle as the humidity increases. The Volta potential difference changes from 403 mV, 359 mV, 310 mV and 291 mV as the humidity is increased from 10, 35, 55 and 75 % RH respectively. The Volta potential values from the carbide particles investigated in this work are summarised in Table 2. The regions of particularly intense signal on the right and left hand sides of the particle are holes. The holes appear as intense regions of cathodic activity which has previously been noted in the literature [23]. The topography maps acquired from the carbide particles revealed that all of the particles were slightly recessed by 5-10 nm from the surrounding matrix, indicating that they are preferentially polished during sample preparation.

The Volta potential maps acquired from a 30 µm carbide particle, the second largest carbide identified on the sample during this investigation, are shown in Fig. 9 a - d. They show the same trend as the 80 µm 80 µm carbide particle in Fig. 8, with a reduced Volta potential difference between the particle and the matrix as the humidity is increased. With the Volta potential difference changing from 322 mV, 295 mV, 246 mV and 155 mV as the humidity is increased from 10, 35, 55 and 75 % RH respectively. Areas of increased Volta potential compared to the bulk of the particles are observed at a number of points in the maps from each the carbides. These <1 µm points are caused by silicon particles and AlFeBe4 precipitates within the bulk of the carbide.

The values in Table 2 highlight the change in Volta potential of the carbide particles as a function of increasing humidity. The shift of >110 mV for each of the particles is greater than the scatter in potential ~50 mV from a given particle composition as well as the reproducibility, < 30 mV, of Volta potential measurements previously performed on beryllium [11]. These humidity results suggest the particle becomes less cathodically active as the humidity increases. This may be a result of the changing nature of the oxide layer covering the particle as the humidity is increased. This behaviour has previously been observed on clean iron. Kelvin probe analysis showed that the Volta potential went to a minimum (most cathodic) when the sample was purged with dry nitrogen, upon purging with dry oxygen the Volta potential increased to become less cathodic and further purging with moist oxygen led to the least cathodic surface [24]. The move to less cathodic surface under dry oxygen was believed to be a result of oxidation of the iron surface with the further shift to less cathodic behaviour in moist oxygen caused by the formation of a hydrated oxide on the surface.

Unfortunately, as a result of the deleterious effect of hydrocarbon contamination on the SKPFM measurements, it is not possible to assess the state of the oxide following exposure to each humidity level on the two large carbide particles. As such AES analysis was performed on smaller carbide particles located near to the larger carbides before and after the humidity study. These spectra are shown in Fig. 10.

The survey spectra from before and after exposure to the high humidity environment show the presence of the Be KLL, C KLL and O KLL Auger transitions. The high resolution C KLL , inset Fig. 10, from each of the analyses from each of the analyses (not shown) reveals the triplet of peaks representing a carbide indicating that the oxide layer is thinner than the depth of analysis, ~6 nm, for both analyses. The intensity of the carbon Auger transition in the spectrum from before exposure to high humidity is significantly greater than the intensity after exposure. The reverse is also true of the oxygen peak intensity, with a significant increase in peak intensity after exposure to higher humidity. This is in keeping with what is expected based upon Equations 1 and 2 and shows that the oxide/hydroxide layer has increased in thickness. This is also supported by examination of the high resolution beryllium spectrum, inset Fig. 10, which shows a significant reduction in the intensity of

7

274275276277278279280281282

283284285286287288289290

291292293294295296297298299300301302

303304305306307

308309310311312313314315316317

the carbide peak component ~100 eV on the shoulder of the oxide peak component ~95 eV [25][24].

The reducing potential of the carbide particles as the humidity increases suggests that the activity of the particles reduces with the increase in the oxide layer thickness, in agreement with results from previous SKPFM analysis [11]. By comparing previous SKPFM analysis of second phase particles in S-65 beryllium billet and bar samples to the results obtained in this work it becomes apparent that the carbide particles have the highest Volta potential of all the second phase particle types except for alumina type particles in bar grade beryllium [11]. This is particularly important as the bar grade matrix has a more cathodic Volta potential, by ~100 mV, than billet grade matrix. This is likely to result in the increased activity of cathodic particles in billet beryllium. A conclusion which is supported by previous electrochemical results [6]. As a result of the more cathodic matrix surrounding the carbide particles in this work it can be assumed that a similar 100 mV shift to a more cathodic particle potential would be observed for carbide particles in billet beryllium. . Carbides in billet sample would then represent the highest Volta potential difference between any particle type and the matrix indicating that despite their rarity in the matrix they may be the most severe contributors to the pitting corrosion of beryllium.

Following the completion of the humidity exposure the sample was lightly repolished using 1 µm diamond solution to refresh the sample surface. SKPFM/AFM analysis was then performed prior to immersing the sample in halide-free deionised, distilled water for time periods of 1, 2, 4, 8 and 48 hours cumulative. SKPFM was repeated following each immersion. SEM analysis was not utilised to avoid the influence of the pyrolised carbon layer on the SKPFM results. Topographic AFM maps and SKPFM Volta potential maps of the particle at time zero and after 1, 8 and 48 hours are shown in Fig. 11. They show that Volta potential difference between the particle and the matrix decreases with increasing exposure time to water. The largest drop in potential is after 1 hour, this is accompanied by a significant change in the topographic map. The potential of the particle decreased from 295 mV to 82 mV during this exposure.

The particle which was slightly recessed from the surface grows a substantial oxide/hydroxide layer and becomes proud of the surface by ~100 nm. At each point where the carbide was exposed at the surface the oxide grows. The layer thickness reached a plateau of after 2 hours at 150 nm thick. Despite the significant exposure time no evidence for galvanic activity was observed. Additionally, no evidence for crevice corrosion at the interface between the particle and the matrix was observed for the 80 µm carbide (Fig. 8) or the 30 µm carbide particle (Fig. 9). As a result of this apparent inactivity the sample was subsequently exposed to a pH 7, 0.5 M solution of sodium chloride for a period of 4 hours. Despite the more corrosive media no further attack was observed on or around the carbide particles.

Conclusions

The present work has investigated the Volta potential difference between S-65 beryllium and second phase carbide particles present in the metal by means of SKPFM with increasing humidity. As a result of this work it has been found that:

1. Carbides do not appear to be mixed metal carbides but contain isolated inclusions of higher mass elements dispersed throughout their bulk.

2. Carbide particles show a more noble potential than the beryllium matrix in agreement with all other particle types in S-65 beryllium.

8

318319

320321322323324325326327328329330331332333

334335336337338339340341342343

344345346347348349350351352

353

354355356

357358359360

3. The carbide particles showed a decreasing Volta potential difference with the matrix as the humidity was increased. This is believed to be a result of an increasing thickness of the oxide/hydroxide layer present on the particles.

4. Under ambient conditions the carbide particles showed a Volta potential that was significantly nobler than all particles types except alumina type particles in bar grade beryllium.

5. Exposure of the carbides to halide free water resulted in a thick oxide/hydroxide layer forming on the carbide particles and despite the cathodic Volta potential of the carbides no evidence of galvanic coupling with the matrix was observed. An additional exposure to 0.5 M NaCl showed no further attack.

Acknowledgments

The authors wish to thank AWE Aldermaston for funding this work.

References

[1] C.F. Mallinson, S. Kozlowski, A. Harvey, J.F. Watts, An investigation of the effect of chlorinated solvents on surface characteristics of S-65 beryllium, Surf. Interface Anal. 48 (2016) 689–693. doi:10.1002/sia.5912.

[2] J.J. Mueller, D.R. Adolphson, Beryllium Science and Technology 2, in: D.R.F. and J.N. Lowe (Ed.), Plenum Press, New York, 1979: pp. 417–433.

[3] M.A. Hill, D.P. Butt, R.S. Lillard, The Passivity and Breakdown of Beryllium in Aqueous Solutions, J. Electrochem. Soc. 145 (1998) 2799–2806. doi:10.1149/1.1838717.

[4] P.D. Miller, W.K. Boyd, Corrosion of Beryllium, Defence Metals Information Centre, Ft. Belvoir :, 1967. http://oai.dtic.mil/oai/oai?verb=getRecord&metadataPrefix=html&identifier=AD0824446.

[5] M.A. Hill, R.J. Hanrahan, C.L. Haertling, R.K. Schulze, R.S. Lillard, Influence of beryllides on the corrosion of commercial grades of beryllium, Corrosion. 59 (2003) 424–435. doi:10.5006/1.3277574.

[6] J.S. Punni, M.J. Cox, The effect of impurity inclusions on the pitting corrosion behaviour of beryllium, Corros. Sci. 52 (2010) 2535–2546. doi:10.1016/j.corsci.2010.03.024.

[7] K.A. Walsh, E.N.C. Dalder, E.E. Vidal, Beryllium Chemistry and Processing, ASM International, United States, 2009.

[8] J.L. English, The Metal Beryllium, American Society for Metals, Cleveland Ohio, 1955.

[9] A.J. Stonehouse, W.W. Beaver, Beryllium Corrosion and how to Prevent it, Mater. Prot. 4 (1965) 24–28.

[10] B.I. Materion, Beryllium Product Uses & Descriptions, Materion Corporation, United States, 2011.

[11] C.F. Mallinson, A. Harvey, J.F. Watts, The Nobility of Second Phase Particles in S-65 Beryllium Studied by Scanning Kelvin Probe Force Microscopy, Corros. Sci. 112 (2016) 669–678. doi:10.1016/j.corsci.2016.09.004.

[12] M.F. Hurley, C.M. Efaw, P.H. Davis, J.R. Croteau, E. Graugnard, N. Birbilis, Volta Potentials Measured by Scanning Kelvin Probe Force Microscopy as Relevant to Corrosion of Magnesium Alloys, Corrosion. 71 (2015) 160–170. doi:10.5006/1432.

[13] C.F. Mallinson, J.F. Watts, Communication—The Effect of Hydrocarbon Contamination on the Volta Potential of Second Phase Particles in Beryllium, J. Electrochem. Soc. 163 (2016) C420–C422. doi:10.1149/2.0471608jes.

[14] P. Schmutz, G.S. Frankel, Corrosion Study of AA2024-T3 by Scanning Kelvin Probe Force Microscopy and In Situ Atomic Force Microscopy Scratching, J. Electrochem. Soc. 145 (1998) 2295–2306. doi:10.1149/1.1838634.

9

361362363364365366367368369370

371

372

373

374375376

377378

379380

381382

383384

385386

387388

389

390

391

392393394

395396397

398399400

401402403

[15] L. Lacroix, L. Ressier, C. Blanc, G. Mankowski, Statistical Study of the Corrosion Behavior of Al2CuMg Intermetallics in AA2024-T351 by SKPFM, J. Electrochem. Soc. 155 (2008) C8–C15. doi:10.1149/1.2799089.

[16] F. Andreatta, M.M. Lohrengel, H. Terryn, J.H.W. de Wit, Electrochemical characterisation of aluminium AA7075-T6 and solution heat treated AA7075 using a micro-capillary cell, Electrochim. Acta. 48 (2003) 3239–3247. doi:10.1016/S0013-4686(03)00379-7.

[17] F. Andreatta, H. Terryn, J.H.. de Wit, Corrosion behaviour of different tempers of AA7075 aluminium alloy, Electrochim. Acta. 49 (2004) 2851–2862. doi:10.1016/j.electacta.2004.01.046.

[18] A.B. Cook, Z. Barrett, S.B. Lyon, H.N. McMurray, J. Walton, G. Williams, Calibration of the scanning Kelvin probe force microscope under controlled environmental conditions, Electrochim. Acta. 66 (2012) 100–105. doi:10.1016/j.electacta.2012.01.054.

[19] C.F. Mallinson, The Chloride Induced Localised Corrosion of Aluminium and Beryllium: A Study by Electron and X-ray Spectroscopies, University of Surrey, 2015. http://epubs.surrey.ac.uk/id/eprint/809467.

[20] J.A. Jiménez, G. Frommeyer, The ternary iron aluminum carbides, J. Alloys Compd. 509 (2011) 2729–2733. doi:10.1016/j.jallcom.2010.12.017.

[21] C.F. Mallinson, J.E. Castle, J.F. Watts, Analysis of the Be KLL Auger Transition of Beryllium Nitride and Beryllium Carbide by AES, Surf. Sci. Spectra. 22 (2015). doi:10.1116/1.4922819.

[22] T.W. Haas, J.T. Grant, G.J.D. III, Chemical Effects in Auger Electron Spectroscopy, J. Appl. Phys. 43 (1972) 1853–1860. doi:10.1063/1.1661409.

[23] P. Schmutz, G. Frankel, Characterization of AA2024-T3 by Scanning Kelvin Probe Force Microscopy, J. Electrochem. Soc. 145 (1998) 2285–2295. doi:10.1149/1.1838633.

[24] S. Yee, R.A. Oriani, M. Stratmann, Application of a kelvin microprobe to the corrosion of metals in humid atmospheres, J. Electrochem. Soc. 138 (1991) 55–61. doi:10.1149/1.2085578.

[25] C.F. Mallinson, J.E. Castle, J.F. Watts, Analysis of the Be KLL Auger Transition on Beryllium and Beryllium Oxide by AES, Surf. Sci. Spectra. 20 (2013) 97–112. doi:10.1116/11.20130801.

[1] C.F. Mallinson, S. Kozlowski, A. Harvey, J.F. Watts, An investigation of the effect of chlorinated solvents on surface characteristics of S-65 beryllium, Surf. Interface Anal. 48 (2016) 689–693. doi:10.1002/sia.5912.

[2] J.J. Mueller, D.R. Adolphson, Beryllium Science and Technology 2, in: D.R.F. and J.N. Lowe (Ed.), Plenum Press, New York, 1979: pp. 417–433.

[3] M.A. Hill, D.P. Butt, R.S. Lillard, The Passivity and Breakdown of Beryllium in Aqueous Solutions, J. Electrochem. Soc. 145 (1998) 2799–2806. doi:10.1149/1.1838717.

[4] P.D. Miller, W.K. Boyd, Corrosion of Beryllium, Defence Metals Information Centre, Ft. Belvoir :, 1967. http://oai.dtic.mil/oai/oai?verb=getRecord&metadataPrefix=html&identifier=AD0824446.

[5] M.A. Hill, R.J. Hanrahan, C.L. Haertling, R.K. Schulze, R.S. Lillard, Influence of beryllides on the corrosion of commercial grades of beryllium, Corrosion. 59 (2003) 424–435. doi:10.5006/1.3277574.

[6] J.S. Punni, M.J. Cox, The effect of impurity inclusions on the pitting corrosion behaviour of beryllium, Corros. Sci. 52 (2010) 2535–2546. doi:10.1016/j.corsci.2010.03.024.

[7] K.A. Walsh, E.N.C. Dalder, E.E. Vidal, Beryllium Chemistry and Processing, ASM International, United States, 2009.

[8] J.L. English, The Metal Beryllium, American Society for Metals, Cleveland Ohio, 1955.

[9] A.J. Stonehouse, W.W. Beaver, Beryllium Corrosion and how to Prevent it, Mater. Prot. 4 (1965) 24–28.

[10] B.I. Materion, Beryllium Product Uses & Descriptions, Materion Corporation, United States, 2011.

10

404405406

407408409

410411

412413414

415416417

418419

420421

422423

424425

426427

428429

430431432

433434

435436

437438

439440

441442

443444

445

446

447

[11] C.F. Mallinson, A. Harvey, J.F. Watts, The Nobility of Second Phase Particles in S-65 Beryllium Studied by Scanning Kelvin Probe Force Microscopy, Corros. Sci. 112 (2016) 669–678. doi:10.1016/j.corsci.2016.09.004.

[12] M.F. Hurley, C.M. Efaw, P.H. Davis, J.R. Croteau, E. Graugnard, N. Birbilis, Volta Potentials Measured by Scanning Kelvin Probe Force Microscopy as Relevant to Corrosion of Magnesium Alloys, Corrosion. 71 (2015) 160–170. doi:10.5006/1432.

[13] C.F. Mallinson, J.F. Watts, Communication—The Effect of Hydrocarbon Contamination on the Volta Potential of Second Phase Particles in Beryllium, J. Electrochem. Soc. 163 (2016) C420–C422. doi:10.1149/2.0471608jes.

[14] P. Schmutz, G.S. Frankel, Corrosion Study of AA2024-T3 by Scanning Kelvin Probe Force Microscopy and In Situ Atomic Force Microscopy Scratching, J. Electrochem. Soc. 145 (1998) 2295–2306. doi:10.1149/1.1838634.

[15] L. Lacroix, L. Ressier, C. Blanc, G. Mankowski, Statistical Study of the Corrosion Behavior of Al2CuMg Intermetallics in AA2024-T351 by SKPFM, J. Electrochem. Soc. 155 (2008) C8–C15. doi:10.1149/1.2799089.

[16] F. Andreatta, M.M. Lohrengel, H. Terryn, J.H.W. de Wit, Electrochemical characterisation of aluminium AA7075-T6 and solution heat treated AA7075 using a micro-capillary cell, Electrochim. Acta. 48 (2003) 3239–3247. doi:10.1016/S0013-4686(03)00379-7.

[17] F. Andreatta, H. Terryn, J.H.. de Wit, Corrosion behaviour of different tempers of AA7075 aluminium alloy, Electrochim. Acta. 49 (2004) 2851–2862. doi:10.1016/j.electacta.2004.01.046.

[18] A.B. Cook, Z. Barrett, S.B. Lyon, H.N. McMurray, J. Walton, G. Williams, Calibration of the scanning Kelvin probe force microscope under controlled environmental conditions, Electrochim. Acta. 66 (2012) 100–105. doi:10.1016/j.electacta.2012.01.054.

[19] C.F. Mallinson, The Chloride Induced Localised Corrosion of Aluminium and Beryllium: A Study by Electron and X-ray Spectroscopies, University of Surrey, 2015. http://epubs.surrey.ac.uk/id/eprint/809467.

[20] J.A. Jiménez, G. Frommeyer, The ternary iron aluminum carbides, J. Alloys Compd. 509 (2011) 2729–2733. doi:10.1016/j.jallcom.2010.12.017.

[21] C.F. Mallinson, J.E. Castle, J.F. Watts, Analysis of the Be KLL Auger Transition of Beryllium Nitride and Beryllium Carbide by AES, Surf. Sci. Spectra. 22 (2015). doi:10.1116/1.4922819.

[22] T.W. Haas, J.T. Grant, G.J.D. III, Chemical Effects in Auger Electron Spectroscopy, J. Appl. Phys. 43 (1972) 1853–1860. doi:10.1063/1.1661409.

[23] P. Schmutz, G. Frankel, Characterization of AA2024-T3 by Scanning Kelvin Probe Force Microscopy, J. Electrochem. Soc. 145 (1998) 2285–2295. doi:10.1149/1.1838633.

[24] S. Yee, R.A. Oriani, M. Stratmann, Application of a kelvin microprobe to the corrosion of metals in humid atmospheres, J. Electrochem. Soc. 138 (1991) 55–61. doi:10.1149/1.2085578.

[25] C.F. Mallinson, J.E. Castle, J.F. Watts, Analysis of the Be KLL Auger Transition on Beryllium and Beryllium Oxide by AES, Surf. Sci. Spectra. 20 (2013) 97–112. doi:10.1116/11.20130801.

Table 1. Chemical compositions of commercial grades of beryllium [10] *Other Ni, Cu, Ti, Zr, Ca, Mn, Ag, Co, Pb, Mo, Ca, U

GradeChemical composition (at.% max)

Be % min BeO Al C Fe Mg Si Other*I-70H 99.0 0.7 0.07 0.07 0.10 0.07 0.07 0.04I-220H 98.0 2.2 0.10 0.15 0.15 0.08 0.08 0.04O-30H 99.0 0.5 0.07 0.07 0.12 0.12 0.07 0.04

11

448449450

451452453

454455456

457458459

460461462

463464465

466467

468469470

471472473

474475

476477

478479

480481

482483

484485

486487

488

S-200F 98.5 1.5 0.10 0.15 0.13 0.13 0.08 0.04S-200FH 98.5 1.5 0.10 0.15 0.13 0.08 0.08 0.04SR-200 98.0 2.0 0.16 0.15 0.18 0.08 0.08 0.04S-200D 98.0 2.0 0.16 0.15 0.13 0.18 0.08 0.04

S-65 99.2 0.9 0.05 0.09 0.08 0.01 0.045 0.045

S-65 Bar 99.2 0.69 0.03 0.09 0.06 unknown 0.04

Table 2 Volta potential values from the two carbide particles at each humidity level

Relative humidity

80 μm carbide 30 μm carbide Volta potential difference to matrix (mV)

10 % 403 32235 % 359 29555 % 310 24675 % 291 155

Fig. 1 SE micrograph (a) and BSE micrograph (b) of the entire structure of a beryllium carbide particle in S-65 beryllium. A higher magnification BSE micrograph shows the presence of higher atomic mass particles within the carbide bulk (c) and a higher magnification SEM micrograph of particularly small higher mass inclusions

in the carbide particle (d). The brighter particle in the bottom left of micrographs (a) and (b) is a silicon second phase particle.

12

489

490

491

492

493494495496497

Fig. 1 High resolution AES spectra of the C KLL Auger transition from hydrocarbon contamination, a beryllium carbide reference spectrum [19] a confirmed beryllium carbide particle studied in this

work and a possible coarse carbide particle.

Fig. 2 SE micrograph (a) and BSE micrograph (b) of the entire structure of a beryllium carbide particle in S-65 beryllium. A higher magnification BSE micrograph shows the presence of higher atomic mass particles within the carbide bulk (c) and a higher magnification SEM micrograph of particularly small higher mass inclusions

in the carbide particle (d). The brighter particle in the bottom left of micrographs (a) and (b) is a silicon second phase particle.

13

498

499500501

502

503504505506507

Fig. 2 SEM micrograph and elemental EDX maps from a carbide particle. Maps for carbon, beryllium, oxygen aluminium, iron and silicon are shown. The 5 μm scale bar is for all images. The white circle in the SEM

micrograph shows the position of the EDX point spectrum acquisition of Fig. 3.

Fig. 3 EDX spectrum from the centre of the carbide particle with overlaid WDX spectra of the Be Kα and C Kα transitions. A high resolution Be Kα spectrum is also inset.

14

508

509510511

512

513514

Fig. 3 EDX spectrum from the centre of the carbide particle, shown in Fig.1, with overlaid WDX spectra of the Be Kα and C Kα transitions. A high resolution Be Kα spectrum is also inset.

Fig. 4 High resolution AES spectra of the C KLL Auger transition from hydrocarbon contamination, a beryllium carbide reference spectrum [21] a confirmed beryllium carbide particle studied in this

work (Fig. 1) and a possible carbide particle which is shown in Fig. 5a.

15

515

516517

518

519520521

Fig. 4 SEM micrograph and elemental EDX maps from a carbide particle. Maps for carbon, beryllium, oxygen aluminium, iron and silicon are shown. The 5 μm scale bar is for all images. The white circle in the SEM

micrograph shows the position of the EDX point spectrum acquisition of Fig. 3.

Fig. 5 Examples of SEM micrographs acquired from previously identified coarse carbides in S-65 beryllium.

16

522

523524525

526

527

528529

Fig. 6 SEM micrograph and carbon, oxygen and beryllium EDX maps of the possible coarse carbide shown in Fig. 5a.

Fig. 7 EDX spectrum and overlaid WDX spectrum of the C Kα transition from the carbon rich region of the possible coarse carbide shown in Fig. 5a. Inset WDX spectrum of the Be Kα transition. The

peak intensity is too weak to observe if overlaid on the EDX spectrum.

17

530

531532

533

534535536

537

Fig. 8 Volta potential maps of an 80 µm carbide particle at different humidity: (a) 10%, (b) 35%, (c) 55% and (d) 75%. All plotted with the same Z scale of 1.5 V and range 0.5 to -1.0 V. The white arrows in (a) denote the position of the cross sectional analysis from which potentials were measured. This position was used for all

of the measurements. Volta potential data shown inverted.

18

538

539

540541542543

Fig. 9 Volta potential maps of the a 30 µm carbide particle at different humidity: (a) 10%, (b) 35%, (c) 55% and (d) 75%. All plotted with the same Z scale of 1.1 V and range 400 to -700 mV. The white arrows in (a)

denote the positions of the cross sectional analysis from which potentials were measured. This position was used for all of the measurements. Volta potential data shown inverted.

19

544

545

546547548549

Fig. 10 AES survey spectra with high resolution Be KLL and C KLL spectra (inset) from a carbide particle before and after the humidity study.

Fig. 10 AES survey spectra and high resolution Be KLL spectra (inset) from a carbide particle before and after the humidity study.

20

550

551

552553

554555

Fig. 11 Volta potential maps (a-d) and topographic maps (e-h) of the 80 µm carbide particle (Fig. 8) following exposure to deionised water for time zero, 1 hour, 8 hours and 48 hours exposure. Potential scale -500 – 300 mV, height scale -200 – 100 nm.

Fig. 11 Topographic maps (a-d) and Volta potential maps (e-h) of the large carbide particle following exposure to deionised water for time zero, 1 hour, 8 hours and 48 hours exposure. Potential scale -500 – 300 mV, height scale -200 – 100 nm.

21

556

557

558559560

561562563

564

565