Microstructure and electrochemical behavior of contemporary Ti6Al4V implant alloys

Abstract

Ti6Al4V is the most common titanium alloy within the biomaterial field. While material standards for different variations of this alloy exist, there are only minimal requirements with respect to its microstructure which is directly related to the alloy’s properties. Thus, a better understanding of the Ti6Al4V microstructure of common contemporary implant components and its effect on the electrochemical behavior is needed; including additively manufactured (AM) devices. Therefore, this study aimed at characterizing the microstructures of conventional and AM total joint replacement components, and to identify the effect of microstructure on the electrochemical behavior. Thus, 22 components from conventional (surgically retrieved cast and wrought implants) and AM implants (not previously implanted) were analysed to characterize microstructure by means of electron backscatter diffraction (EBSD) and energy dispersive X-Ray spectroscopy (EDS), and tested to determine its electrochemical behavior (potentiodynamic polarization and EIS). The microstructure of the conventional implants varied broadly but could be categorized into four groups as to their grain size and shape: fine equiaxed, coarse equiaxed, bimodal, and lamellar. The AM components exhibited a fifth category: lath-type. The AM components had a network of β-phase along the α-phase grain boundaries, prior β-grains, and manufacturing voids. Finally, the electrochemical study showed that the equiaxed coarse grains and lath-type grains (AM components) had inferior electrochemical behavior, whereas cast alloys had superior electrochemical behaviour; fine-grained wrought alloys likely provide the best compromise between electrochemical and mechanical properties.

1. Introduction

Titanium-6Aluminum-4Vanadium (Ti6Al4V) alloy is the most common titanium alloy on the market, with approximately 50% market share ^1–3. This alloy is generally specified by the biomedical standards ASTM F1472 and ISO 6832-3, and more specifically by ASTM F1108 (cast alloy) and ASTM F136 (wrought extra low interstitial alloy, ELI). Originating from the aerospace industry, Ti6Al4V is now used in the biomedical field as one of the most important implant alloys due to its low density, superior corrosion resistance, good osseointegration properties, and lower elastic modulus when compared to other commonly used alloys such as Cobalt-Chromium-Molybdenum (CoCrMo) and stainless steel^4,5. While orthopaedic implants exhibit overall satisfactory longevity, implant failure can occur in some cases due to fracture, oxide induced stress corrosion cracking, hydrogen embrittlement, or fretting corrosion ^6–9. It is well known that the mechanical and electrochemical properties of any given alloy are governed by the microstructure^10,11. Generally, the microstructure of Ti6Al4V is bi-phasic, consisting mostly of the hexagonal-closed packed (hcp) α phase, and the more ductile, body-centered cubic (bcc) β phase ¹². Microstructural features such as grain size and shape, as well as phase content and distribution directly result from the chemical composition, processing parameters (cast vs. wrought alloy), and processing sequence (applied cooling rate, the subsequently applied type(s) of heat or thermo-mechanical treatments)^13–16.

Most orthopaedic implants are currently made from either wrought or cast alloys. Previous studies have shown that the microstructure of CoCrMo alloy can differ widely, even within the same implant type and manufacturer, with implications on their corrosion behavior ^17,18. While microstructures of Ti6Al4V alloys were well studied in orthopaedic implants in the past ^19–23, lately less attention has been given to the role of microstructure in contemporary Ti6Al4V alloys used in orthopaedic implants. Moreover, the additive manufacturing (AM) process has emerged in the biomedical field. This manufacturing process enables new solutions such as the affordable and quick production of custom-made and personalised implants to fit specific patient needs and anatomy with highly complex geometries^24,25. This technology is already available on the market for total joint arthroplasties (TJA) and spinal implants, where modulated laser powder bed fusion (e.g., Selective Laser Melting, SLM) is one of the most common AM processes, by which metal implants are basically built by targeted micro-welding of fine metal alloy powders. However, the impact on the microstructure due to frequent and fast local solidification, heating, and cooling sequences still needs further research ^1,26. Although the metallurgy of AM parts is well known in the aerospace community ^2,27, it is unclear if AM components in the medical industry have a comparable microstructure and microstructural variability to commonly used wrought and cast Ti6Al4V alloy implants, and if this microstructure and microstructural variability has any implications on implant performance.

Therefore, to start with one property of clinical relevance, the goal of this study was to characterize the microstructure of contemporary TJA implant components made of Ti6Al4V alloy and determine its effect on the implant’s electrochemical behavior. The electrochemical properties of the alloy do not only determine its risk of corrosion-driven damage mode, e.g. oxide-induced stress corrosion cracking ⁷, but also the implant surfaces’ interaction with biological constituents of the periprosthetic environment. We aimed to answer the following research questions: (1) What is the variability in microstructure of Ti6Al4V implant components made from conventional (wrought and cast alloy) and AM alloy? (2) Do microstructural features influence the electrochemical behavior of different Ti6Al4V alloys?

2. Materials and Methods

For this study, we selected 18 implant components that included acetabular cups, femoral stems, dual modular femoral necks, and tibial plateaus from the common contemporary total hip and knee replacements available in our implant retrieval repository. Additionally, we tested three acetabular cups from two different designs that were not previously implanted and were specifically advertised as being additively manufactured. For the retrieved implants, information on the applied manufacturing process or heat treatment of the bulk alloy was not available. Similarly, for the AM devices, no details on the applied AM process or subsequent treatment were known. Three of the 21 different implant components were tested in duplicate to determine intra-manufacturer consistency. Finally, a Ti6Al4V Extra Low Interstitial (ELI) bar stock by Allegheny Technologies Incorporated (ATI) was used as a reference (control) alloy. Overall, 22 alloy samples (21 implant components + 1 control alloy) were analysed (Table 1).

Table 1 –

Microstructural features of the titanium implants.

Sample	Microstructure type	α%	β%	ECD mean	β ECD mean	AR mean	β AR
				(SD) [μm]	(SD) [μm]	(SD)	mean (SD
1 (control)	A: Equiaxed fine	98	2	1.6 (±1.17)	0.58 (±0.21)	1.84 (±0.68)	1.98 (±0.71)
2	A: Equiaxed fine	99	1	1.69 (±1.33)	0.77 (±0.22)	1.9 (±0.65)	1.83 (±0.47)
3	A: Equiaxed fine	98.2	1.8	1.27 (±1.13)	0.59 (±0.22)	1.79 (±0.59)	1.82 (±0.54)
4	A: Equiaxed fine	97	3	1.53 (±1.28)	0.74 (±0.31)	1.75 (±0.53)	1.84 (±0.49)
5	A: Equiaxed fine	99	1	1.52 (±1.34)	0.55 (±0.18)	1.74 (±0.59)	1.66 (±0.38)
6	B: Equiaxed coarse	93.2	6.8	9.58 (±7.03)	4.35 (±1.85)	1.96 (±0.82)	2.19 (±0.91)
7	B: Equiaxed coarse	93.2	6.8	9.58 (±7.03)	4.35 (±1.85)	1.96 (±0.82)	2.19 (±0.91)
8	C: Bimodal	98.4	1.6	2.9 (±2.53)	0.9 (±0.01)	1.8 (±0.74)	1.71 (±0.4)
9	C: Bimodal	96.67	3.33	3.04 (±2.1)	1.18 (±0.29)	1.93 (±0.81)	2.02 (±0.7)
10	C: Bimodal	94	6	2.03 (±1.55)	0.93 (±0.4)	1.78 (±0.62)	1.85 (±0.55)
11	C: Bimodal	96.8	3.2	3.26 (±2.27)	1.29 (±0.29)	1.95 (±0.72)	2.08 (±0.64)
12	C: Bimodal	98.25	1.75	2.04 (±1.58)	0.81 (±0.25)	1.88 (±0.91)	1.87 (±0.52)
13	C: Bimodal	98.8	1.2	2.81 (±2.1)	1.12 (±0.18)	1.85 (±0.57)	1.81 (±0.42)
14	C: Bimodal	99.4	0.6	1.82 (±1.5)	0.81 (±0.17)	1.84 (±0.58)	2.02 (±1.27)
15	C: Bimodal	96.5	3.5	3.03 (±2.18)	1.44 (±0.31)	1.89 (±0.73)	1.9 (±0.54)
16	C: Bimodal	96.5	3.5	2.09 (±1.85)	0.96 (±0.49)	1.81 (±0.65)	1.89 (±0.65)
17	D: Lamellar	99.9	0.1	18.98 (±17.83)	-	2.13 (±0.92)	-
18	D: Lamellar	99.9	0.1	10.55 (±10.83)	-	2.34 (±1.24)	-
19	D: Lamellar	99.9	0.1	10.55 (±10.83)	-	2.34 (±1.24)	-
20	E: Lath-type	94.53	5.47	6.06 (±5.99)	2.67 (±1.22)	3.06 (±1.98)	3.37 (±2.51)
21	E: Lath-type	95.67	4.33	10.05 (±6.1)	4.87 (±0.82)	2.86 (±1.54)	3.12 (±1.98)
22	E: Lath-type	96.33	3.67	7.59 (±5.35)	3.2 (±0.74)	2.81 (±1.39)	2.55 (±0.91)

Sample preparation

Alloy samples were directly sectioned from the implant components. To ensure that the alloy microstructure was homogeneous, samples were cut along the longitudinal and the transverse direction using a cut-off machine (Secotom, Struers). The longitudinally and transverse oriented samples were then embedded in an epoxy resin with a hot mounting press (CitoPress, Struers). The metallographic preparation was performed with an automatic grinding/polishing machine (Tegramin, Struers) using the following steps: sequential grinding using SiC paper with 220 grit (2 minutes) and 320 grit (4 minutes), followed by mechanical polishing with 9 μm diamond suspension (10 minutes) and by 0.04 μm colloidal silica suspension (8 minutes). A final ion polishing (Cross-section polisher IB-19530 CP, JEOL) step at 4kV for 10 min, followed by 3kV for 15 min at 4°, was also performed to provide optimal conditions for microstructural analysis.

Characterization of microstructure

Alloy samples were analysed by means of a scanning electron microscope (IT500HR, JEOL) equipped with an electron backscattered diffraction (EBSD) detector (C-nano detector, Oxford Instruments) and an energy dispersive X-ray spectroscopy (EDS) detector (Ultim-Max detector, Oxford Instruments). EBSD renders microstructural information, while EDS allows for simultaneous analysis of the local chemical composition. Combined EBSD/EDS analysis was performed on 6 different areas of each alloy sample: 3 areas on the longitudinally oriented sample surface and 3 areas on the transverse oriented sample surface. Measurements were conducted at magnifications between 350x to 2000x for all samples. Lower magnifications were applied if a coarse grain size was observed to ensure that a minimum of 85 grains was captured per image. The electron beam scanning step size was 0.07 μm at high magnifications and 0.6 μm at low magnifications. These conditions were chosen to achieve a good balance between scanning speed (0.3 to 1.5 hours) and resolution. Postprocessing was conducted using AZtec and Channel 5 software (Oxford Instruments). The following microstructural metrics were recorded: grain size as measured by the equivalent circle diameter (ECD), grain aspect ratio (AR), α/β ratio, β grain size (β ECD), β phase aspect ratio (β AR). Additionally, EDS provided co-localized chemical maps for all alloy elements. Finally, all samples were categorized in groups according to their grain size and shape.

Electrochemical study

Three samples from each microstructure group were selected for the electrochemical study, with the exception of one group for which only two samples were available, for a total of 14 alloy samples. Each sample was tested 5 times, for a total of 70 corrosion tests. Samples were sectioned and embedded, and repeatedly mirror-polished. Additionally, samples were sonicated in Tergazyme® soap solution for 5 min, generously rinsed under double-distilled water, rinsed with 70 vol% ethanol solution, and dried with N₂ gas flow before each test. The electrochemical measurements were performed in a three-electrode cell kept in a water bath at 37 ± 2°C, with the sample as working electrode (WE), a graphite rod as counter electrode (CE), and a Ag/AgCl Skinny Reference Electrode (Gamry Instruments) as reference electrode (RE), using a G750 potentiostat (Gamry Inc.) and Gamry Framework 6.2 software. The previously described rectangular corrosion cell held 200ml of electrolyte¹⁷. The polished metal sample was placed vertically against a round opening with an exposed area of 0.38 cm². The electrolyte, simulating the joint fluid, was a new-born calf serum solution, which was buffered with 27 g/L tris(hydroxymethyl)aminomethane (TRIS), diluted in saline solution (0.9 g/L) for a final total protein content of 30 g/L, and adjusted with HCl for a final pH of 7.4. The electrolyte was not de-aeriated during testing. Two electrochemical measurements were performed in sequence: 1) Electrochemical Impedance Spectroscopy (EIS), and 2) potentiodynamic polarization. These measurements were preceded by cathodic cleaning (−0.9 V vs. Ag/AgCl for 10 min), and one hour of open circuit potential (OCP) stabilization. The EIS measurements were performed in the frequency range from 100 kHz to 50 mHz, with signal amplitude ±10mV vs OCP. Equivalent electric circuit analysis was performed by fitting the EIS data to the response of a modified Randles circuit (Gamry Echem Analyst Software 7.8.2), in line with the work of Bijukumar et al. ^28,29. The modified Randles circuit consisted of a solution resistance (R_S), in series with a Constant Phase Element (CPE, represented by the capacitance element ‘Q’ and the exponent ‘a’) in parallel with a resistance (R_p). The potentiodynamic tests were performed from −0.25 V +0.25 V vs OCP, at a scan rate of 1 mV/s, to assess corrosion potential (E_corr) and corrosion current density (i_corr) by Tafel’s method. A superior corrosion resistance of a given sample was related to relatively: higher resistance ‘R_p’; lower capacitance ‘Q’; higher ‘E_corr’; lower ‘i_corr’. Larger deviations of ‘a’ values from 1 indicate more surface heterogeneities disturbing the current response behavior from an ideal capacitance circuit element³⁰.

Statistical analysis

Values of microstructural metrics (ECD, AR, β content, and ΔV) are given as mean and standard deviation (SD). Comparisons between alloy groups were conducted by one-way-ANOVA and independent variable t-tests. Electrochemical metrics are listed as median. Comparisons were conducted with Kruskal-Wallis tests, and pair-wise comparisons were conducted with Mann-Whitney tests. We also conducted single and multiple linear regression models to determine the effect of different microstructural features on each electrochemical metric. Significance was set at p=0.05 for all statistical tests.

3. Results

Ti6Al4V microstructure of contemporary orthopaedic implant components

Overall, five different groups of microstructures could be distinguished based on grain size (equivalent circle diameter - ECD) and grain shape (Figure 1). Groups A-C were different wrought alloys, Group D were cast alloys, and the AM alloys fell into Group E. The microstructure Groups A-E were defined as follows. Group A had a microstructure with fine equiaxed grains with an ECD ranging from 1 to 2 μm, grain aspect ratio (AR) ranging from 1.6 to 1.9 and β content ranging from 1 to 3%. We defined grains with an AR standard deviation (SD) of less than 1.5μm as equiaxed. Group B had coarse equiaxed grains with an ECD of more than 9 μm, AR ranging from 1.8 to 2, and β content ranging from 1.8 to 7%. Only two implant components fell into this group. Group C had bimodal (fine and coarse equiaxed grains) grains with an ECD of 2 to 4 μm, AR ranging from 1.7 to 2, and β content ranging from 1.8 to 6%. Group D had lamellar grains with characteristic coarse lamellar-like grains. This microstructure exhibited minimal β content (<0.5%) and an AR ranging from 2 to 2.4. Finally, Group E exhibited a lath-type grain structure with and an ECD of 6 to 10 μm, elongated grains (or laths) with an AR of 2.8 to 3. Although this group exhibited a similar grain size (ECD = 7.9 ± 5.81 μm) with Group B, it had a significantly larger aspect ratio due to the lath-type microstructure grain shape (AR = 2.91 ± 1.63) (Figure 2). Lath-type microstructure is usually not characterized by grain size as measured by the ECD, but rather by the lath width (effective grain size). The mean lath width was 8.36 ± 7.5 μm. Within Group E, there was also a significant difference in lath width (p<0. 01) between cups from different manufacturers, where samples 18 and 20 had a smaller lath width compared to sample 19, with 4.93 ± 3.98 μm, 5.71 ± 4.44 μm and 10.94 ± 8.6 μm, respectively. Similar to Group B, Group E alloys exhibited a network-like distribution of beta phase along the alpha grain boundaries. The β content ranged from 3.5 - 5.5 % in those alloys. Additionally, a local preferential orientation (texture) of the β phase was observed in the pole figures (Figure 3e) of all AM alloys, prompting analysis at lower magnification. The red spots on the pole figures represent the preferential grain orientation. Orientation maps were plotted separately for co-localized α and β phase. The β phase orientation map illustrates areas of several β grains with the same crystal orientation giving the appearance of an underlying grain structure (Figure 3). Finally, the lath-type microstructure (Group E) also exhibited fine cavity-like voids, which could be mainly observed close to the surface. Such defects were either round with a mean size of 15 μm (range: 2.5-56 μm) or irregularly shaped with a size of 40 μm (range: 16-104 μm). Round voids are likely caused by retained gas from the raw metal powder or during the melting process while irregularly shaped voids result from a lack of fusion of powder particles (Figure 4) ^2,31–33. Differences in grain size and shape, as well as other substructures, can easily be appreciated in Figure 5.

Figure 1 -.

5 different types of microstructure were categorized based on grain size and shape: A) fine equiaxed grains (control), B) coarse equiaxed grains (sample 6), C) bimodal (sample 13), D) lamellar dendritic grains (sample 17) and E) coarse lath-type microstructure grains (sample 20).

Figure 2 –

Comparison of the (a) grain size, (b) grain aspect ratio and (c) β phase content of A) fine equiaxed grains, B) coarse equiaxed grains, C) bimodal, D) lamellar dendritic grains and E) lath-type microstructure grains.

Figure 3 –

Sample 20: (a) phase map of α phase (red) and its grain orientation (IPF) map (b); (c) phase map of β phase (blue), its grain orientation map (d). Different colors in (b) and (d) indicated different grain orientations. While the α phase orientation appears to be homogenous, the β phase orientation map exhibits areas with the same crystal orientation as illustrated by areas with the same color. These areas correspond to the prior β grains. The resulting texture is also confirmed by the pole figures for (e) α and (f) β phase. Texture is confirmed by the symmetry of hot sports in each quadrant.

Figure 4 -.

Manufacturing defects observed on additive manufactured acetabular cups (samples 20-22). The hole shapes are characteristic of lack of fusion (blue) and retained gas from the raw metal powder or during the melting process (yellow).

Figure 5 -.

Microstructural difference of the acetabular cups according to the manufacturing process where (a) sample 3, (b) sample 6, (c) sample 18, (d) sample 20 and (e) sample 22. The black arrows show the prior β phase. The orientation is represented by the colors on the inverse pole figure for each phase.

Comparisons between alloy microstructure groups exhibited differences in grain size (Figure 2a); however, the β content and aspect ratio (Figure 2c) were variable across and within groups. The chemical distribution was inherently inhomogeneous for all alloys with a notable β content due to the affinity of aluminium and vanadium with α and β phase, respectively. Only the lamellar alloys exhibited a homogeneous microstructure due to the lack of a notable β content. Alloys with a β content of > 0.5 vol% usually exhibited slightly larger content of vanadium within the β phase and of aluminium within the α phase. However, in some bimodal cases (Group C), vanadium occurred independent of the β phase, but was concentrated within smaller grains (Figure 6). We used the gradient of vanadium (ΔV) across α/β phase boundaries as a criterion (Figure 7) for the chemical distribution of the alloying elements as will be explained later. Group D exhibited a ΔV of approximately 0 due to the absence of β phase. Groups A and C exhibited a moderate ΔV of 4.5 % and 2.5 %, while Groups B and E had the highest ΔV of 11.3 % and 11.6 %, respectively.

Figure 6 –

Band contrast, phase map (red=α, blue=β phase), titanium map, aluminum map and vanadium map of the control alloy, sample 12 and sample 6. The phase maps illustrate different fraction sizes and distributions of beta phase. The EDS elemental maps show differences in the chemical distribution of alloy elements. Vanadium occurred predominantly within small grains in sample 12 and aluminum within the larger grains of the bimodal microstructure, whereas in the control and sample 6 vanadium is concentrated within the β phase. The network-like β phase distribution is even better illustrated within the vanadium map than in the actual phase map. The comparably low intensity of titanium within the β phase in sample 6 suggests a strong decline of titanium content in favor of vanadium.

Figure 7 –

(a) and (b) vanadium and titanium variation between α and β led to galvanic corrosion in the equiaxed coarse grains group and lath-type grains group (AM); (c) vanadium variation (ΔV) across all groups.

With respect to the type of implant component, 2 acetabular cups, 1 femoral stem, and 1 tibial plateau belonged in Group A, together with the control sample; 2 acetabular cups belonged in Group B; 4 tibial plateaus, 3 femoral stems, and 2 modular necks fell into Group C, 2 femoral stems and 1 acetabular cup from the same manufacturer belonged to Group D, and all 3 AM cups belonged in Group E.

Electrochemical behavior of alloys with varying microstructure

Representative outcomes of the electrochemical tests for each group are illustrated by polarization curves (Figure 8a), Nyquist Plots (Figure 8b), Bode Plots Z Module (Figure 8c) and Bode Plots Z Phase (Figure 8d). The median E_corr (V) and i_corr (A/cm²) were −0.18 (−0.27 to −0.05) and 3.6 E-8 (1.7 E-8 to 1.2 E-6), −0.24 (−0.38 to −0.08) and 7.3 E-8 (2.4 E-8 to 1.1 E-6), −0.14 (−0.30 to −0.06) and 2.6 E-8 (1.1 E-8 to 3.9 E-7), −0.20 (−0.27 to −0.09) and 2.9 E-8 (1.2 E-8 to 5.5 E-8), −0.13 (−0.27 to −0.06) and 5.2 E-8 (9.9 E-9 to 1.5 E-6) for Groups A-E, respectively (Figure 9a–b) (Table 2). Regarding the equivalent circuit parameters obtained from fitting the EIS data, the median values for Q (μS·s^a/cm²) and R_p (kΩ·cm²) were 28.9 (0.8 to 48.2) and 792 (76 to 3039), 40.8 (32.6 to 163) and 522 (8 to 1380), 29.2 (27.1 to 31.2) and 834 (98 to 2114), 30 (26.5 to 37.6) and 1314 (942 to 2633), 32.5 (29.3 to 43.4) and 745 (12 to 5449) for Groups A-E, respectively (Figure 9c–d) (Table 3). Non-parametric statistical comparison between alloy groups exhibited statistical differences for E_corr (p=0.014), i_corr (p=0.001), Q (p<0.001), and R_p (p=0.003). The lowest value for the corrosion potential (E_corr) was observed for Group B. Pairwise comparisons exhibited lower (p<0.05) E_corr values for Group B compared to all groups with the exception of Group A, where only a trend was observed (p=0.052). The other groups did not differ from one another (Figure 9a). The corrosion current density (i_corr) had its highest median value and range for Groups B and D, with greater values compared to Groups C (p=0.012, p=0.019) and D (p<0.001, p=0.002) (Figure 7b). Group B showed the highest median value for the capacitance element (Q). Despite a broad variability, pairwise comparison showed that Q values of Group B were higher compared to Groups A and C-E (p=0.007, p<0.001, p<0.001, p=0.003, respectively) (Figure 9c). Group E also exhibited a greater Q value compared to Groups C (p<0.001) and D (p=0.01). Finally, the value for the polarization resistance (R_p) for Group D was higher compared to Groups A (p=0.014), B (p<0.001), C (p=0.025), and E (p=0.001) (Figure 9d).

Figure 8 –

(a) polarization curves, (b) Nyquist Plot, (c) Bode Plot Z Module, (d) Bode Plot Z Phase and (e) modified Randles circuit.

Figure 9 –

(a) E_corr, (b) i_corr, (c) Q and (d) R_p of A: Equiaxed fine, B: Equiaxed coarse, C: Bimodal, D: Dendritic and C: Lath-type. Equiaxed coarse and lath-type groups showed poor corrosion current, polarization resistance and capacitance.

Table 2 –

Electrochemical parameters from potentiodynamic polarization of each group (median, N = 5)

Group	ECORR, E-01∙V vs. Ag/AgCl			iCORR, E-08∙A/cm2			Q			Rp
	Median	Min	Max	Median	Min	Max	Median	Min	Max	Median	Min	Max
A: Equiaxed fine	−0.18	−0.27	−0.05	3.6	1.7	120	28.9	0.8	48.2	792	76	3039
B: Equiaxed coarse	−0.24	−0.38	−0.08	7.3	2.4	110	40.8	32.6	163	522	8	1380
C: Bimodal	−0.14	−0.3	−0.06	2.6	1.1	39	29.2	27.1	31.2	834	98	2114
D: Lamellar	−0.2	−0.27	−0.09	2.9	1.2	5.5	30	26.5	37.6	1314	942	2633
E: Lath-type	−0.13	−0.27	−0.06	5.2	0.99	150	32.5	29.3	43.4	745	12	5449

Table 3 –

Electrochemical parameters from potentiodynamic polarization of each sample (average ± standard deviation, N = 5)

Sample	Group	ECORR, E-01∙V vs. Ag/AgCl			iCORR, E-08∙A/cm2			Q			Rp
		Av.	Min	Max	Av.	Min	Max	Av.	Min	Max	Av.	Min	Max
1 (Control)	A: Equiaxed fine	−1.68 (±0.06)	−2.62	−1.16	4.49 (±3.37)	1.66	8.94	21.2 (±20.37)	0.83	48	1883.4 (±927.19)	639	3039
4	A: Equiaxed fine	−1.63 (±0.06)	−2.21	−0.72	2.86 (±0.57)	2.15	3.6	28.3 (±0.56)	28	29	891.8 (±139.87)	752	1152
5	A: Equiaxed fine	−1.87 (±0.10)	−2.65	−0.54	36.6 (±46.5)	5.3	118	36.1 (±2.54)	33	39	160 (±99.55)	76	356
6	B: Equiaxed coarse	−2.58 (±0.04)	−3.22	−2.31	5.34 (±1.73)	2.37	6.68	33.4 (±1)	33	35	1033.4 (±244.18)	655	1380
7	B: Equiaxed coarse	−2.16 (±0.11)	−3.76	−0.82	42.5 (±42.9)	7.85	111	96 (±44.88)	46	163	184.1 (±149.21)	8	389
10	C: Bimodal	−1.37 (±0.06)	−2	−0.56	13.8 (±14.7)	1.6	38.9	29.9 (±0.95)	29	31	795.8 (±688.73)	98	2113
11	C: Bimodal	−1.64 (±0.09)	−3.04	−0.67	3.17 (±4.02)	1.1	10.3	27.9 (±0.73)	27	29	1263.8 (±559.61)	383	2063
12	C: Bimodal	−1.36 (±0.02)	−1.58	−1.09	4.05 (±2.88)	1.8	8.7	29.8 (±1.28)	28	31	939.7 (±296.35)	526	1398
17	D: Lamellar	−1.42 (±0.08)	−2.74	−0.89	3.58 (±1.35)	1.54	5.18	29.5 (±2.71)	26	32	1195 (±142.34)	942	1380
18	D: Lamellar	−2.11 (±0.01)	−2.33	−2.01	1.62 (±0.36)	1.16	2.11	28.4 (±0.38)	28	29	1594.6 (±540.95)	1140	2632
19	D: Lamellar	−1.67 (±0.05)	−2.42	−1.23	3.79 (±1.15)	2.84	5.5	33.5 (±3.29)	30	37	1566.7 (±426.30)	1027	2286
20	E: Lath-type	−1.1 (±0.07)	−2.28	−0.59	68.3 (±44.7)	43.4	148	38.1 (±4.16)	32	43	45.9 (±32.75)	12	87
21	E: Lath-type	−1.27 (±0.06)	−2.16	−0.68	4.74 (±2.88)	1.81	9.55	31.1 (±1.68)	29	34	875.8 (±300.19)	570	1395
22	E: Lath-type	−1.69 (±0.06)	−2.7	−1.18	3.63 (±2.03)	0.99	5.29	32.3 (±0.71)	31	33	1787 (±1837.74)	745	5449

Linear correlation of microstructural features with electrochemical metrics across all samples and alloy groups was conducted. The chosen microstructural features included ECD (grain size), AR (grain shape), β content, and ΔV, the gradient of V across the α/β phase boundaries. The results exhibited no significant correlation with ECD and only a weak positive correlation between AR and E_corr (R²=0.062, p=0.04). The β content correlated positively with i_corr (R²=0.1, p=0.01), Q (R²=0.17, p<0.001). It also exhibited a weak positive correlation with R_p (R²=0.076, p=0.026). Finally, ΔV correlated positively with i_corr (R²=0.30, p=0.001) and Q (R²=0.14, p=0.001). It also exhibited a weak negative correlation with R_p (R²=0.076, p=0.026). Multi-linear regression (MLR) revealed that AR was positively correlated with E_corr (B=0.06, p=0.008), while ECD exhibited a trend towards a negative correlation (B=−0.04, p=0.054). For i_corr, MLR confirmed a strong positive correlation with ΔV (B=3.26x10⁻⁸, p<0.001). For Q, MLR revealed a strong positive correlation with ECD (B=2.2, p<0.001) and β content (B=6.238, p<0.001), as well as a negative correlation with AR (B=−19.41, p=0.001). For R_p, the previously proposed correlation with ΔV was confirmed.

4. Discussion

This study aimed to characterize the microstructure of contemporary TJA implant components made of Ti6Al4V alloy, and its effect on the electrochemical behavior. Among the implants characterized, we found five distinctly different microstructures with varying grain size and shape, beta phase content and the gradient of vanadium across α/β phase boundaries (ΔV). Regarding electrochemical properties, we found that ΔV correlated with increased corrosion rate and lower polarization resistance, while a smaller grain size and larger grain aspect ratio seemed to have a beneficial influence on the corrosion behavior.

Differences in Ti6Al4V microstructure and implications on its properties

The five different microstructural groups unsurprisingly fell into the three manufacturing process categories: wrought, cast, and additively manufactured; however, wrought alloys could be further distinguished in three sub-types: Groups A and C were generally similar with respect to mean grain size and shape, and beta content. Differences between A and C are likely to be explained by the thermomechanical processing parameters ^12,34. Electrochemically, there was no difference observed between those two groups. However, from a mechanical point of view, the larger grains present within the bimodal microstructure of group C may negatively influence the alloy’s mechanical behavior by decreasing ductility and introducing the risk of early crack initiation and fast propagation ^35–43.

The equiaxed coarse microstructure differed clearly from Groups A and C, both in mean grain size, β content and phase distribution. Grain growth and network-like distribution of the β phase are likely the results of heat treatment, such as annealing, over a long time period and potentially slow cooling rates. The same could be said for the AM alloy components (Group E). This alloy differs mainly from Group B due to its characteristic lath-type grains. The closest resemblance of this lath-type microstructure may be that of the so-called Widmanstätten microstructure ^40,44. This type of microstructure commonly occurs in steels, titanium and zirconium alloys. It is known to enhance fracture toughness, but it is also associated with lower strength and ductility in titanium alloys ⁴⁵. Group E also had similarities to Group B, considering the large mean grain size and the considerable β phase fraction distributed in a network-like fashion. There are several examples in literature showing that this lath-type microstructure is the result of annealing of the as-printed Ti6Al4V microstructure ^1,26,27,46. The initial form of this alloy would usually exhibit a similar grain size to Group A wrought alloys, but also characteristic needle shaped grains with little to no beta phase. Also, as-printed Ti6Al4V alloy consists for the most part of the meta stable α prime (α’) phase, whereas the heat-treated form represented here in Group E consists of α and β phase. The prominent β content and its network-like distribution along the α grain boundaries could enable crack propagation along the α-β interface. Under fretting conditions, this could be a preferential site of oxide induced stress-corrosion-cracking, hydrogen embrittlement, or preferential corrosion of the β phase as previously reported for other types of Ti6Al4V alloys ^6–9. Besides the grain shape, Group E differs from all other groups by the presence of voids (built-defects) and a prior-β grain structure. Both may potentially affect the mechanical properties of this alloy ⁴⁶. The clearly visible prior β grain structure is a unique feature in AM components. Residual prior β grain sub-structures originate from a fast solidification process that is likely inherent to the laser powder bed fusion process^2,26,27,46. Such prior β grains may increase the ductility of the alloy but could also provide additional sites for cracks⁴⁷. It needs to be investigated if additional heat-treatments of the alloy may be able to alter this microstructure. The presence of voids observed within AM acetabular cups may impact the fatigue behavior since they can act as crack nucleation sites. According to Li ⁴⁸, crack nucleation can occur from voids of > 40 μm in Ti6Al4V alloy. Therefore, several voids within the AM components studied here lie right around this threshold. In addition to that, crack nucleation relies on a) the number of voids, where a low void density is desirable, b) void shape, where an uneven shape is more prone to crack nucleation, c) void location, where subsurface voids may be more likely to lead to crack nucleation, and d) void size. The process of Hot Isostatic Pressing (HIP) can address residual voids by subjecting the part to an isostatic inert gas pressure at a high temperature, subsequently diminishing sub-surface defects ⁴⁹. However, HIP is not able to eliminate voids caused by entrapped gas, and cannot remove other voids entirely, but rather substantially reduce their size ^50,51. A side effect of the HIP process can be grain coarsening ^48,52.

Group D exhibited all the characteristic hallmarks of a cast alloy, including large grain size and lamellar grains, which would also indicate a lower strength compared to wrought alloys. Another main characteristic of this group was the virtual absence of β phase, and subsequently, a more homogeneous chemical distribution compared to the other alloys. With respect to chemical distribution, it appears that subsequent heat treatment of the alloy comes along with increased beta phase fraction, beta phase size, and the inherent chemical inhomogeneity of bi-phasic alloys across the α-β boundaries, which is quantified here as ΔV.

Absolute certainty of each alloys’ mechanical performance can only be achieved by testing. For example, the microstructure of a cast and AM alloy is largely driven by the cooling sequence inherent to the implant’s specific geometry. The reproduction of the exact same microstructure to a different geometry, such as a standardized sample, needs to be verified to ensure that in vitro results may accurately predict in vivo performance. Therefore, a fundamental understanding of the alloy microstructure-property relationship will help to evaluate the electrochemical behavior accurately.

Impact of Ti6Al4V microstructure on the electrochemical behavior

The coarse equiaxed microstructure (Group B) had the least favorable electrochemical behavior by every metric. The AM microstructure (Group E) exhibited a similar trend, with the exception of E_corr, where this group experienced the noblest median values. The similarity in corrosion behavior between these two groups can be linked to similarities in microstructure. Both exhibited a network-like distribution of beta phase, and even more importantly, a large value for ΔV. The latter was strongly associated with a higher corrosion rate (i_corr) and a lower polarization resistance (R_p). In contrast, the lamellar microstructure (Group D) had the lowest corrosion rate which can be explained by the minimal β phase content and the resulting chemical homogeneity which in turn eliminates local galvanic effects⁵³ across α and β phases. There might also be an effect of grain orientation ⁵⁴, however it was not the focus of this study and further research is needed. The importance of chemical homogeneity in preventing in vivo corrosion of orthopaedic implants was recently demonstrated for CoCrMo alloy as well ¹⁷. Segregation of alloying elements - manifested by longitudinal bands depleted in Mo and Cr and introduced during the thermo-mechanical processing - were shown to increase the corrosion rate both in vitro and in vivo ^17,55,56. Carbides within the alloy were also associated with pitting corrosion ¹⁷. Inhomogeneity of the alloy was shown to provide preferential corrosion sites that can be the target of a severe chemical attack in vivo in some cases, especially within the confined space of modular junctions. In CoCrMo implant components, corrosion is directly associated with a severe adverse local tissue reaction and implant failure in many cases⁵⁷, while corrosion manifests with less severity in Ti6Al4V implants. Yet, electrochemical processes on the surface of Ti-alloy implants, especially when coupled with fretting, can lead to implant fracture⁷.

The EIS analysis coupled with multi-linear regression further suggests that grain size and beta content were both strongly associated with a higher value of the capacitance element Q. Studies on other alloys have shown that a smaller grain size can lead to a more stable passive film, because fine grains result in a larger area of grain boundaries, that act as passive film nucleation sites and provide stronger passive film adherence ^58,59, which may explain the preferred corrosion kinetics observed with finer grains. It is unclear whether the presence of the bcc lattice structure itself or the larger ΔV is responsible for this observed relationship between Q and β content. Q was also negatively associated with AR in the MLR model. It is not clear whether this relationship has a direct physical meaning or is coinsidental. Further tests would be needed to confirm these correlations. Finally, the exponential factor ‘a’ did not present substantial variations among samples (Appendix A), and therefore it was not concluded for the statistical analysis and discussion.

Limitations

The investigated devices and manufacturers only reflect a few of the most common implants used at our institution. A single device per implant type was investigated, and intra-manufacturer repeatability is unclear. However, the three implants tested in duplicate suggest there is good repeatability within the same implant design. Another limitation was that group B (N=2) was underpowered to conduct statistical comparisons because alloy groups were retrospectively assigned after the microstructural analysis. Additionally, we did not examine the morphology and composition of the protective passive film, which we intend to study in more detail in the future. Also, the authors had no knowledge of the manufacturing process and process parameters applied. Any judgment of the alloy microstructure could only be made retroactively. Most importantly, this study only provides knowledge of differences in alloy microstructure and its effect on the implant’s electrochemical behavior. No direct correlations could be made as to how the microstructure related to the in vivo performance.

In this study we used a combination of new and surgically devices. While this approach may not be ideal, we do not consider it a limitation, because we examined the bulk microstructure of the alloys. Any potential effects due to wear and corrosion would have only affected the immediate subsurface zone and the observed difference in microstructure are characteristic for specific manufacturing processes and subsequent heat-treatments. Finally, even though specific alloy microstructures are associated with better electrochemical behavior, it does not necessarily follow that other microstructures are not sufficient for its specific application.

5. Conclusion

Ti6Al4V implant alloys exhibit a wide range of variability in microstructure that, in turn, have an impact on electrochemical properties. Additively manufactured implants exhibited unique microstructural features, which varied in several ways from the established wrought and cast alloys. Moreover, equiaxed coarse grains and lath-type grains (AM) groups had the worst corrosion behavior, attributed to the high composition gradient between α and β phase leading to local galvanic corrosion. Also, the lath-type microstructure of the AM alloys may promote local crevice corrosion due to the presence of voids from lack of fusion and gas porosity on the implant surface. Finally, while the lamellar microstructure associated with cast alloys exhibits the overall best corrosion behavior, the equiaxed fine microstructure likely provides the best compromise between good electrochemical and mechanical behaviour based on a moderate ΔV and the fine grain size, respectively. Considering the increase of AM implants on the market, it is of great importance to monitor their performance in vivo and ensure that these unique microstructural features do not introduce increased risk of short- or long-term mechanical or electrochemical failure. Additional research is needed to determine best AM process parameters and subsequent heat treatments to provide the best suited alloy microstructure for specific implant applications.

Appendix A –

Equivalent electric circuit parameters from fitting of EIS data

Sample	Group	Rp, kΩ∙cm2	Q, μS∙sa/cm2	a	Goodness of Fit
Control	A. Equiaxed fine	2068	26.0	0.92	9.3E-04
Control	A. Equiaxed fine	1006	0.8	0.73	5.0E-02
Control	A. Equiaxed fine	3039	0.8	0.56	4.8E-03
Control	A. Equiaxed fine	639	48.2	0.91	2.4E-04
Control	A. Equiaxed fine	2665	30.1	0.93	3.6E-04
1	A. Equiaxed fine	1153	27.5	0.93	2.5E-04
1	A. Equiaxed fine	752	28.0	0.93	2.6E-04
1	A. Equiaxed fine	869	28.7	0.93	3.0E-04
1	A. Equiaxed fine	792	28.4	0.93	3.1E-04
1	A. Equiaxed fine	892	28.9	0.93	1.9E-04
2	A. Equiaxed fine	356	38.5	0.89	1.9E-03
2	A. Equiaxed fine	121	37.2	0.90	1.1E-03
2	A. Equiaxed fine	127	38.1	0.90	1.2E-03
2	A. Equiaxed fine	121	33.7	0.92	7.8E-04
2	A. Equiaxed fine	76	33.1	0.92	3.7E-04
3	B. Equiaxed coarse	1113	32.6	0.94	2.7E-04
3	B. Equiaxed coarse	1127	33.2	0.93	2.8E-04
3	B. Equiaxed coarse	893	32.9	0.93	3.8E-04
3	B. Equiaxed coarse	655	33.2	0.93	3.1E-04
3	B. Equiaxed coarse	1380	35.2	0.93	2.1E-04
4	B. Equiaxed coarse	52	115.3	0.31	6.1E-03
4	B. Equiaxed coarse	324	82.9	0.82	2.9E-03
4	B. Equiaxed coarse	8	163.0	0.63	9.8E-03
4	B. Equiaxed coarse	147	46.4	0.84	6.7E-03
4	B. Equiaxed coarse	389	72.3	0.81	3.0E-03
5	C. Bimodal	2114	28.8	0.93	9.6E-04
5	C. Bimodal	476	29.2	0.94	3.1E-04
5	C. Bimodal	98	31.1	0.90	3.0E-03
5	C. Bimodal	637	30.4	0.92	1.6E-04
5	C. Bimodal	654	30.0	0.92	1.3E-04
6	C. Bimodal	1261	27.1	0.92	4.3E-04
6	C. Bimodal	1035	27.7	0.92	4.8E-04
6	C. Bimodal	1577	27.9	0.92	4.0E-04
6	C. Bimodal	2063	29.1	0.92	3.5E-04
6	C. Bimodal	383	27.8	0.92	4.9E-04
7	C. Bimodal	820	27.9	0.93	4.1E-04
7	C. Bimodal	1119	30.2	0.93	4.8E-04
7	C. Bimodal	834	31.2	0.93	4.9E-04
7	C. Bimodal	526	30.5	0.93	4.7E-04
7	C. Bimodal	1398	29.2	0.94	2.6E-04
8	D. Lamellar	942	31.9	0.94	1.4E-04
8	D. Lamellar	1381	30.0	0.93	2.0E-04
8	D. Lamellar	1205	32.1	0.92	1.2E-04
8	D. Lamellar	1249	26.4	0.93	2.4E-04
8	D. Lamellar	1198	26.8	0.94	1.9E-04
9	D. Lamellar	1498	27.9	0.93	1.6E-04
9	D. Lamellar	1511	28.9	0.93	1.9E-04
9	D. Lamellar	1140	28.4	0.93	2.8E-04
9	D. Lamellar	1192	28.7	0.93	2.0E-04
9	D. Lamellar	2633	28.1	0.93	3.0E-04
10	D. Lamellar	1473	37.6	0.94	3.7E-04
10	D. Lamellar	1733	36.4	0.93	5.4E-04
10	D. Lamellar	1028	31.9	0.94	2.4E-04
10	D. Lamellar	2286	31.4	0.93	3.2E-04
10	D. Lamellar	1314	30.1	0.93	2.2E-04
11	E. Lath-type	87	37.6	0.87	1.8E-03
11	E. Lath-type	33	31.9	0.90	2.2E-03
11	E. Lath-type	15	38.0	0.91	2.7E-03
11	E. Lath-type	12	43.4	0.88	1.1E-02
11	E. Lath-type	83	39.6	0.90	2.0E-03
12	E. Lath-type	1395	33.6	0.93	3.5E-04
12	E. Lath-type	1015	31.8	0.92	1.0E-03
12	E. Lath-type	758	30.1	0.93	4.7E-04
12	E. Lath-type	641	30.6	0.92	5.0E-04
12	E. Lath-type	570	29.3	0.93	3.3E-04
13	E. Lath-type	1172	32.4	0.93	3.2E-04
13	E. Lath-type	781	32.8	0.93	3.2E-04
13	E. Lath-type	788	32.8	0.93	4.2E-04
13	E. Lath-type	745	32.5	0.93	4.4E-04
13	E. Lath-type	5449	31.1	0.93	2.6E-04