The Effect of Boron on the Microstructure and Properties of Refractory Metal Intermetallic Composites (RM(Nb)ICs) Based on Nb-24Ti-xSi (x = 16, 17 or 18 at.%) with Additions of Al, Cr or Mo

This paper is about metallic ultra-high temperature materials, in particular, refractory metal intermetallic composites based on Nb, i.e., RM(Nb)ICs, with the addition of boron, which are compared with refractory metal high entropy alloys (RHEAs) or refractory metal complex concentrated alloys (RCCAs). We studied the effect of B addition on the density, macrosegregation, microstructure, hardness and oxidation of four RM(Nb)IC alloys, namely the alloys TT2, TT3, TT4 and TT8 with nominal compositions (at.%) Nb-24Ti-16Si-5Cr-7B, Nb-24Ti-16Si-5Al-7B, Nb-24Ti-18Si-5Al-5Cr-8B and Nb-24Ti-17Si-3.5Al-5Cr-6B-2Mo, respectively. The alloys made it possible to compare the effect of B addition on density, hardness or oxidation with that of Ge or Sn addition. The alloys were made using arc melting and their microstructures were characterised in the as cast and heat-treated conditions. The B macrosegregation was highest in TT8. The macrosegregation of Si or Ti increased with the addition of B and was lowest in TT8. The alloy TT8 had the lowest density of 6.41 g/cm3 and the highest specific strength at room temperature, which was also higher than that of RCCAs and RHEAs. The Nbss and T2 silicide were stable in the alloys TT2 and TT3, whereas in TT4 and TT8 the stable phases were the Nbss and the T2 and D88 silicides. Compared with the Ge or Sn addition in the same reference alloy, the B and Ge addition was the least and most effective at 800 °C (i.e., in the pest regime), when no other RM was present in the alloy. Like Ge or Sn, the B addition in TT2, TT3 and TT4 did not suppress scale spallation at 1200 °C. Only the alloy TT8 did not pest and its scales did not spall off at 800 and 1200 °C. The macrosegregation of Si and Ti, the chemical composition of Nbss and T2, the microhardness of Nbss and the hardness of alloys, and the oxidation of the alloys at 800 and 1200 °C were also viewed from the perspective of the alloy design methodology NICE and relationships with the alloy or phase parameters VEC, δ and Δχ. The trends of these parameters and the location of alloys and phases in parameter maps were found to be in agreement with NICE.


Introduction
Research is underway worldwide to develop new metallic ultra-high temperature materials (UHTMs) to replace Ni superalloys in the hottest parts of future aero-engines to enable the latter to meet environmental and performance targets. Candidate metallic UHTMs include refractory metal (RM) intermetallic composites (RMICs), RM high entropy alloys (RHEAs) and RM complex concentrated alloys (RCCAs) (unnecessarily, in our opinion, another term has been added for RHEAs and RCCAs to further muddle the names used for metallic UHTMs, namely refractory chemically complex alloy (RCCAs)). The new materials must meet specific property targets (goals) for toughness, creep and oxidation The structure of the paper is as follows: the description of experimental procedures is followed by the presentation of the results for the microstructures, density and hardness of the alloys. Preliminary results of the oxidation of the alloys are presented to show the effect of the addition of B on oxidation at 800 and 1200 • C. The discussion of the results includes their scrutiny from the perspective of NICE.

Experimental Section
Large button ingots (0.6 kg) of the alloys (Table 1) were prepared from high purity (min. 99.99 wt.%) Al, Cr, Mo, Nb, Si, Ti and B (99.5 wt.%) in an argon atmosphere using arc melting with a non-consumable tungsten electrode and a water-cooled copper crucible. Each alloy was melted five times. The alloys were heat-treated for 100 h at 1500 • C (1200 • C for TT2, see next section). Each alloy was wrapped in Ta foil and was heat-treated in a tube furnace under a continuous flow of Ti gettered argon followed by furnace cooling to room temperature. The microstructures of the alloys were characterised with XRD (Philips diffractometer, Hiltonbrooks Ltd., Crewe, UK, Cu radiation, solid specimens, JCPDS database) and microanalysis using EPMA (JEOL 8600 electron probe micro analyser, Tokyo, Japan, operating conditions 9 kV with a beam current of 30 nA [18]) equipped with WDX and EDX spectrometers, elemental standards and BN and TiN standards. The samples were prepared following standard metallographic procedures, mounted in Bakelite and ground using SiC grinding papers (320 to 2400 grit) and polishing with diamond DUR clothes (6 and 1 µm finish).
The density of the alloys was measured using Archimedes principle and a Sartorious electronic precision balance, Sartorius Lab Instruments GmbH & Co. KG, Göttingen, Germany, equipped with a density measurement kit. The isothermal oxidation of the as cast (AC) alloys was studied at 800 and 1200 • C using Stanton Redcroft thermo-balances (Thermal Scientific plc., Odessa, TX, USA) and specimens from the as cast alloys. The weight of each sample was measured at the start and end of each oxidation experiment. A Mitutoyo HM-101 hardness testing machine (Mitutoyo America, Aurora, IL, USA) with Vickers indenter was used to measure the hardness (0.5 kg load) of the as cast (AC) and heat-treated (HT) alloys. The microhardness of the solid solution and T2 silicide was measured in the AC and HT alloys using a 0.025 kg load. At least ten measurements were done for each alloy and condition, and each phase. The area fraction of the solid solution was measured using the software KS Run 3 in a Polyvar Met microscope. The SEM images of the large areas (×250) of the as cast (AC) and heat-treated (HT) alloys that were analysed by EPMA were used for the measurements.

Results
The actual composition of each alloy in the AC or HT condition is given in Table 1. The densities of the alloys are given in Table 2. The alloys TT4 and TT2 had the lowest and highest density, respectively, and the densities of all four alloys were lower than 7 g/cm 3 . The macrosegregation of solutes (MACX, X = Al, B, Cr, Si, Ti) in the alloys is given in Table 3. Note that this table has data for the Al and Cr free alloy TT1, to help us understand the role of boron with other alloying additions on MACX. The highest macrosegregation of B and Si was observed respectively in the alloys TT8 and TT4 whereas the macrosegregation of Ti was essentially similar in all four alloys. The phases that were observed in the alloys using XRD and EPMA are summarised in Table 4. The chemical compositions of the phases that were observed in all parts of the ingots are given in Table 5.
indicates the presence also of Ti-rich phase (see Table 5), + 1500 • C/100 h, ♦ 1200 • C/100 h, o indicates that alloy was contaminated by nitrogen.

As Cast Alloys
In all parts of the ingot of the alloy TT2, were observed the Nb ss , T2, a Cr and Tirich phase and a eutectic of the Nb ss and T2, whereas in specific parts of the ingot other phases were also present, namely a Ti-rich phase in the top and bottom, a C14-Laves phase in the top and bulk and Nb 3 Si silicide in the bulk (Tables 4 and 5, Figure 1a and Figure S1a in Supplementary Materials). Ti-rich Nb ss and Ti-rich T2 were also observed. In Figure 1a, the grey contrast facetted phase is the T2, the bright contrast phase is the Nb ss , darker contrast areas at the edges of Nb ss are Ti-rich Nbss and darker grey contrast areas in between Nb ss are Ti-rich T2. The Ti-rich phase, C14-Laves phase, Cr and Ti-rich phase and Nb 3 Si were observed in the dark and very contrast inter-dendritic Nb ss areas. The average compositions of the Ti-rich phase, the Laves phase and the Nb 3 Si silicide, respectively, were 23. It should be noted that the Laves and Nb 3 Si compounds were B free and that the Si concentration in the silicide would suggest also the presence of metastable Nb 3 Si [11], which however was not indicated by the XRD (Figure S1a in Supplementary Materials). In the Nb ss , with increasing Ti concentration, the concentrations of B and Cr respectively decreased and increased, whereas in the T2 the B concentration decreased with increasing Ti concentration ( Table 5). The average <Si> content and the <Nb>/<Si> and Si/B ratios of the T2 and Ti-rich T2 were 38.2 and 37.1 at.%, 1.6 and 1.7, and 2 and 3.8, respectively (where <Nb> = Nb + Cr + Ti and <Si> = B + Si). The T2 silicide was facetted, this was particularly noticeable in the top and bottom of the ingot. The vol.% of the eutectic was lower in the bulk. In all parts of the ingot, the microstructure of the alloy TT3 consisted of the Nb ss , T2, and eutectic of these two phases (Tables 4 and 5, Figure 1c and Figure S1c in Supplementary Materials). The XRD suggested the presence of the D8 8 silicide, however, its presence was not confirmed by EPMA. Ti-rich Nb ss and T2 were also observed but in this alloy the partitioning of Ti to T2 was very strong and resulted in distinct separate T2 grains very rich in Ti (and Al) in the top and bulk of the ingot, with average composition 34.2(29.8-37.2)Nb-28.2(24.7-32.2)Ti-27.8(26.6-28.8)Si-4.2(3.7-4.9)Al-5.6(3.6-7.4)B. In Figure 1c, the light contrast phase is the Nb ss , the grey contrast phase is the T2 and the darker grey contrast is the Ti-rich T2. In the Nb ss and T2, with increasing Ti concentration, the concentrations of both B and Al decreased in the solid solution, and respectively decreased and increased in the T2. The average <Si> content, and the <Nb>/<Si> and Si/B ratios of the T2, Ti-rich T2 and T2 very rich in Ti were 38.6, 39 and 37.5 at.%, 1.6, 1.56 and 1.67, and 1.9, 4 and 5, respectively (where <Nb> = Nb + Ti and <Si> = Al + B + Si). The T2 was facetted only in the bulk of the ingot. Cracks that had formed in large Ti-rich T2 grains were stopped by the surrounding solid solution.
In all parts of the ingot of the alloy TT4, the Nb ss , T2, D8 8 and eutectic of Nb ss and T2 were observed. The D8 8 silicide exhibited very bright contrast (Tables 4 and 5, Figure 1e and Figure S1e in Supplementary Materials). Ti-rich Nb ss and Ti-rich T2 were also observed but there was no evidence of Ti-rich D8 8 . The latter was Al free. In Figure 1e, the T2 is the grey contrast phase, darker grey contrast areas correspond to Ti-rich T2, the very bright contrast phase with no darker contrast areas is the D8 8 (e.g., the very bright contrast thin long phase at the bottom left-hand side, middle and upper right-hand side of the image "running" at about 45 degrees from right to left) and the bright phase is the Nb ss with darker areas corresponding to the Ti-rich Nb ss .
In the Nb ss and T2, with increasing Ti concentration, the concentrations of B and Cr decreased and increased, respectively, and the concentration of Al essentially did not change in the solid solution, whereas in the T2 the Al and Cr increased and the B decreased. The average <Si> content, and the <Nb>/<Si> and Si/B ratios of the T2, Ti-rich T2 and D8 8 were 38.2, 36.7 and 40.8 at.%, 1.62, 1.72 and 1.45, and 4.3, 6.4 and 0.6, respectively (where <Nb> = Nb + Ti + Cr and <Si> = Al + B + Si). The T2 was facetted in the top and the bulk of the ingot and Ti-rich T2 grains were cracked with the cracks parallel to each other and not extending in the surrounding Nb ss .
The microstructure of the alloy TT8 consisted of Nb ss , T2, D8 8 , Nb ss + T2 eutectic in all parts of the ingot and Nb 3 Si that was observed only in the top and bulk of the ingot (Tables 4 and 5, Figure 1g and Figure S1g in Supplementary Materials). The average composition of Nb 3 Si was 46Nb-23Ti-20.1Si-2.8Al-10Cr-5.8B-1.3Mo. The D8 8 exhibited very bright contrast and was Al free. In Figure 1g, the T2 is the grey contrast phase, darker grey contrast areas correspond to Ti-rich T2, the very bright contrast phase with no darker contrast areas around it is the D8 8 , and the bright phase with darker areas is the Nb ss . The Nb 3 Si exhibits very dark contrast in the inter-dendritic Nb ss areas. The concentration of B in the solid solution varied significantly. It should be noted that there were solid solution grains that were B free. Ti-rich Nb ss and Ti-rich T2 were also observed, and no Ti-rich D8 8 . In the Nb ss and T2, with increasing Ti concentration, the concentrations of Al, B and Mo decreased and that of Cr increased in the solid solution whereas in the T2 the B decreased but the changes of Al, Cr and Mo were minimal. The average <Si> content, and the <Nb>/<Si> and Si/B ratios of the T2, Ti-rich T2 and D8 8 were 38.9, 37.7 and 42.6 at.%, 1.57, 1.65 and 1.35, and 2.5, 3.3 and 0.4, respectively (where <Nb> = Nb + Ti + Cr and <Si> = Al + B + Si). Both the T2 and D8 8 were facetted, particularly the latter and there were cracks in T2 that run parallel to each other and stopped at the Nb ss .
The vol.% of the Nb ss was essentially the same in the alloys TT2, TT3 and TT8, about 38% and significantly lower in the alloy TT4 ( Table 2). The microhardness of the solid solution was lowest and highest respectively in the alloys TT2 and TT4 (472 and 563 HV, respectively), the microhardness of T2 was highest and lowest in the alloys TT2 and TT3, respectively (1346 and 1232 HV, respectively), and the highest and lowest alloy hardness was measured for the alloys TT8 and TT3, respectively (744 and 630 HV, respectively), see Table 2.

Heat Treated Alloys
All the alloys were heat-treated at 1500 • C with the exception of the alloy TT2 that exhibited liquation at this temperature and was heat-treated at 1200 • C. The phases that were present in their microstructures are given in Table 4. The alloys TT2 and TT8 were contaminated by nitrogen during the heat treatment.
In the alloy TT2 the Nb 3 Si silicide, the Ti-rich phase, the Cr and Ti-rich phase and the Laves phase that were present in the cast alloy were not observed. The Nb ss and T2 were present with Ti-rich areas in the latter (Tables 4 and 5, Figure 1b and Figure S1b in Supplementary Materials). In Figure 1b, the light contrast areas are the Nb ss , the grey contrast phase is the T2, the darker grey contrast is the Ti-rich T2 and the very dark areas are TiN. The chemical inhomogeneity that was characteristic of this alloy in the as cast condition resulted in T2 grains in the heat-treated microstructure with <Nb>/<Si> and Si/B ratios covering a wider range than in the other alloys. The chemical composition of T2 and Ti-rich T2 did not change significantly compared with the cast alloy (Table 5) The microstructures of the alloys TT3 and TT4 consisted of the Nb ss , T2 with the D8 8 present only in the latter alloy and no TiN in both alloys. Ti-rich T2 was observed in TT3 but not in TT4 and the D8 8 was Al free, as was the case in the cast alloy (Tables 4 and 5, Figure 1d,f and Figure S1d,f in Supplementary Materials). The microstructure of TT3-HT is shown in Figure 1d, where the light contrast phase is the Nb ss , the grey contrast phase is the T2 and the darker grey contrast is the Ti-rich T2. In the alloy TT3, the average <Si> content, and the <Nb>/<Si> and Si/B ratios of the T2 and Ti-rich T2 were 38.5 and 37.7 at.%, 1.6 and 1.65, and 2.6 and 6.1, respectively (where <Nb> = Nb + Ti and <Si> = Al + B + Si). In the TT4-HT there were no facetted T2 grains and often the D8 8 silicide was observed inside (surrounded by) T2 grains. The microstructure of TT4-HT is shown in Figure 1f, where the bright phase is the Nb ss , the grey phase is the D8 8 and the darker grey contrast phase is the T2. The T2 had <Si> = 39 at.%, <Nb>/<Si> = 1.56, Si/B = 5.5 (<Si> = Al + B + Si, <Nb> = Nb + Cr + Ti), whereas the D8 8 had <Si> = 42.7 at.%, <Nb>/<Si> = 1.3 and Si/B = 0.5. In the alloy TT8 the microstructure consisted of Nb ss , the T2, D8 8 and Ti-rich T2 silicides, was contaminated by nitrogen and the Nb 3 Si was not present (Tables 4 and 5, Figure 1h and Figure S1h in Supplementary Materials). The D8 8 was Al free. The microstructure of TT8-HT is shown in Figure 1h, where the bright phase is the Nb ss , the grey phase is the D8 8 , the darker grey contrast phase is the T2 and dark areas around the latter correspond to Ti-rich T2. The average <Si> content, and the <Nb>/<Si> and Si/B ratios of the T2, Ti-rich T2 and D8 8 were 37.3, 38.3 and 40 at.%, 1.68, 1.61 and 1.5, and 6.2, 4.1 and 0.51, respectively (<Nb> = Nb + Cr + Mo + Ti). Some grains for the Nb ss were B free, as in TT8-AC.
After the heat treatment, the vol.% of Nb ss increased slightly in the alloys TT2, TT3 and TT8 and significantly in TT4 and was similar in all alloys, about 39% ( Table 2). The microhardness of the Nb ss was reduced in all alloys, most noticeably in TT3. Similarly, the microhardness of T2 was reduced in all alloys, but the smallest reduction was in TT4. The hardness of all alloys was reduced, the TT8 still had the highest hardness but the smallest hardness reduction was measured for TT2 (Table 2).

Oxidation
The mass change data is shown in Figure 2a,b for isothermal oxidation at 800 and 1200 • C and typical examples of scale spallation at the two temperatures are shown in Figure 2c,d. The Figure 2a,b include data for the MASC alloy (Nb-25Ti-16Si-2Al-2Cr-8Hf [20]) and the alloy TT1 [12]. The oxidation rate constants are given in Table 6. At 800 and 1200 • C, all four alloys followed respectively linear and parabolic oxidation kinetics. None of the alloys with B addition suffered from pest oxidation at 800 • C, meaning each alloy did not convert into powder, but the scales that formed on the alloys TT1, TT2, TT3 and TT4 spalled off. At 1200 • C scale spallation was also observed for the alloys TT1, TT2, TT3 and TT4, and the MASC alloy. The 1200 • C specimen of TT8 looked exactly like the specimen shown in Figure 2d.

Macrosegregation
The addition of B increased the macrosegregation of Ti and Si compared with the reference B free alloys, with the exception of TT8 where MACSi decreased compared with JG3 (Table 7). TT4 had the highest MACSi, an increase of 3.4 at.% compared with KZ5, and TT1 had the highest MACTi. The highest and lowest MACB was exhibited respectively by TT8 and TT4. With the exception of the alloy TT4, the spread of MACB was narrow and around 4.4 at.%. Table 7. Comparison of the macrosegregation (MACX) of X = B, Si or Ti in the B containing alloys of this work and their basis/reference alloys KZ3, KZ4, KZ5, KZ7 [16] and JG3 [17] (see Appendix A). Large button ingots (0.6 kg) of all alloys were prepared using arc melting. Note that the table includes data for the alloy TT1 (Nb-24Ti-18Si-8B [12]) and its basis alloy KZ3 (Nb-24Ti-18Si [16] The macrosegregation of Si and Ti in the alloys of this study and their reference alloys is shown in Figure 3a-p. This figure shows plots of MACSi and MACTi versus specific parameters that describe macrosegregation in RM(Nb)ICs [19]. In agreement with previous research on B free RM(Nb)ICs [21][22][23][24], RM(Nb)ICs/RCCAs [25] and HEAs [26] [19]). This means that alloy design in NICE can optimise both MACSi and MACTi in B containing alloys using the same parameters. It should be noted that in the MACSi plots the data for the alloy TT8 and its basis JG3 does not follow the aforementioned trend. Further research is essential to confirm MACSi in TT8 and to study RC(Nb)ICs based on TT4 with RM addition(s) as both alloys (i.e., TT4 and TT8) point to a "pathway" for developing RCCAs with B addition [1,2].     The data for MACB is limited only to the alloys of this study and the alloy TT1 [12]. Figure 3q to s show a good parabolic fit of this limited data of MACB versus the parameters ∆H m sd , ∆H m sd /∆H m sp and T m sd , and suggests "maximum" MACB between five and six at.% for ∆H m sd ≈ 18 kJ/mol, ∆H m sd ≈ 1.535 kJ/mol and T m sd ≈ 1840 K. More data about MACB is needed before we can firmly ascertain that the same parameters and trends also describe the macrosegregation of B in RM(Nb)ICs, and RHEAs or RCCAs with B addition.

Microstructures
The alloy microstructures were chemically inhomogeneous, particularly in the case of the alloy TT2. In the reference alloys KZ7, KZ5 and JG3, the high-temperature tetragonal βNb 5 Si 3 had formed from the melt [16,17], whereas with the addition of B, in the alloys TT2, TT3, TT4 and TT8 the tetragonal T2 silicide (same structure as the low-temperature αNb 5 Si 3 , see introduction) formed from the melt as did the hexagonal D8 8 silicide in the alloys TT4 and TT8. In the reference alloy KZ4, the αNb 5 Si 3 formed via the eutectoid transformation of Nb 3 Si [16]. Solute partitioning made the identification of phases, in particular the D8 8 silicide in the alloys TT4 and TT8, difficult. In all four alloys, the Nb ss and T2 were stable phases and the D8 8 in TT4 and TT8.
In the alloy TT2 the Ti-rich phase, Nb 3 Si and C14 Laves phase formed in competition with the Cr and Ti-rich phase (Table 4) in specific parts of the ingot of the cast alloy. The Ti-rich phase was observed only in the top and bottom of the ingot, the Laves phase in the top and bulk, similarly to KZ4-AC [16], and the Nb 3 Si silicide only in the bulk. The aforementioned four phases were not stable in the heat-treated alloy. Note that the C14 Laves phase was also not stable in KZ4-HT. The Nb ss + Nb 3 Si eutectic that formed in KZ4-AC was replaced by the Nb ss + T2 eutectic in TT2-AC. In the latter, the area fraction of the eutectic was significantly reduced in the bulk of the ingot rather than at the bottom, which was the case in TT1-AC [12], and would suggest sensitivity of the Nb ss + T2 eutectic to cooling rate and the presence or not of the Nb 3 Si silicide. The Laves phase formed from Cr rich melt and thus depended on the partitioning of Cr during solidification. This would explain why the Laves phase was not observed in the bottom of the ingot. Compared with KZ4-AC, no Ti-rich areas were observed in the Laves phase in TT2-AC and the Ti concentration was significantly higher than in the Laves phase in KZ4-AC. This was attributed to the Laves phase "competing for Ti" with the Cr and Ti-rich phase (which formed everywhere in the ingot) and the Ti-rich phase (which formed in the top and bottom of the ingot). Chemical inhomogeneity existed in TT2-HT, with a range of composition of the Ti-rich T2 (see Table A2 in the Appendix A) indicating that equilibrium was not reached in this alloy after 100 h at 1200 • C.
The as solidified microstructure in TT3 was similar to that in KZ7-AC [16], the Nb ss + βNb 5 Si 3 eutectic in the latter was replaced by the Nb ss + T2 eutectic in TT3-AC. Compared with TT1-AC, the addition of Al in TT3 suppressed the formation of Nb 3 Si, which was also the case in KZ7-AC compared with KZ3-AC [16]. The suppression of the Nb 3 Si was also linked with the strong partitioning of Ti in the T2 that "starved" the melt of Ti, which stabilises the Nb 3 Si. Therefore, the synergy of B and Al in TT3 furthered the formation of primary tetragonal T2 (as did the addition of Al in KZ7 regarding the formation of primary βNb 5 Si 3 [16]) and of the Nb ss + T2 eutectic. The partitioning of Ti in T2 was sensitive to cooling rate as the very Ti-rich T2 was not observed in the bottom of the ingot of TT3-AC.
In the alloy TT4, the C14 Laves phase was not formed, differently with KZ5-AC [16]. In the alloy TT8-AC, the Nb 3 Si silicide was formed but not the C14 Laves, similarly with JG3-AC [17], in which Nb ss + Nb 3 Si eutectic was formed compared with the Nb ss + T2 eutectic in TT8-AC. The Nb 3 Si was not observed in the bottom of the ingot of TT8-AC. Comparison with the data for JG3-AC [17] and TT8-AC would suggest sensitivity of the formation of Nb 3 Si to cooling rate that was attributed to the synergy of B, Cr and Mo in which B and Cr lead. Three-phase equilibria between Nb ss , Nb 5 Si 3 and Nb 4 Si (metastable silicide) would exist for an alloy of composition Nb-17.5Si-2Mo (at.%) according to the ternary Nb-Si-Mo system by Savitskiym et al. [27]. The Nb 3 Si in TT8-AC had Si + B + Al ≈ 28.7 at.%, which is close to the composition of the metastable Nb 3 Si-m in [13], and the Si content (20.1 at.%) corresponds to Nb 4 Si (metastable silicide, [13]). Thus, it is likely that with the addition of Mo both the stable (i.e., tetragonal) and metastable Nb 3 Si silicide(s) formed in TT8-AC. Compared with JG3-AC, the Nb 3 Si was leaner in Si and Al and richer in Cr; this was attributed to the synergy of Mo and B in TT8. The concentration of Mo was the same in the Nb 3 Si in JG3-AC and TT8-AC, which would suggest that B did not affect the concentration of Mo in the 3-1 silicide. Compared with TT1-AC [12], the Nb 3 Si in TT8-AC was leaner in Si but richer in B. Compared with TT2-AC, the Nb 3 Si in TT8-AC was leaner in Si but richer in Cr and B. Given that in TT2-AC the Nb 3 Si was B free, it was concluded that in the presence of Mo the solubility of B in Nb 3 Si is increased and that Mo leads in the Mo and Cr synergy and eliminates the negative effect of Cr on the solubility of B in Nb 3 Si.
In both the TT4 and TT8 alloys (Ti/Si = 1.5 based on nominal compositions or average Ti/Si = 1.42 based on actual chemical compositions, Table 1), with the addition of B, a new B rich phase formed and was stable, namely the D8 8 silicide. The latter was observed in the areas surrounding the T2 and exhibited very strong bright contrast, which was stronger than that of Nb ss (Figure 1e,g). There was no solubility of Al in D8 8 and its Ti concentration was in the range 10.2 < Ti < 12.5 at.% (average 11.7 at.%, Table 5). In all the alloys of this study, whether in the AC or HT condition, the Si + B or the Si + B + Al content of the T2 was essentially the same (37 and 38.1 at.%, respectively, average 37.6 at.%) whereas the Si + B content of the D8 8 was higher and in the range 40 to 42.7 at.% (average 41.5 at.%). Furthermore, the Si/B ratio in the T2 was in the range 1 to 6.4 (average 4) compared with 0.5 for the D8 8 silicide (range 0.4 to 0.6), and the average <Nb>/<Si> ratio was 1.7 in the T2 compared with 1.4 in the D8 8 (Table A2 in Appendix A). The Ti concentration of the T2 and D8 8 , respectively, was in the ranges 14.3 to 30.7 at.% (average 21.2 at.%) and 10.2 to 12.5 at.% (average 11.7 at.%), Table 5. The above data for the T2 and D8 8 silicides is in agreement with the Nb-Si-B ternary [10], see introduction.
In the 1600 • C ternary Nb-Si-B section, tentative equilibrium between T2 and Nb 3 B 2 was indicated by Nowotny et al. [10]. Thermodynamic modelling of the Nb-Si-B ternary [28] calculated the Nb 3 B 2 to be stable at 1600 • C and showed that it formed two three-phase equilibria Nb + T2 + Nb 3 B 2 and T2 + Nb 3 B 2 + NbB with neighbouring phases. Sun et al. [28] stated "there are no direct experimental evidences to support the existence of these two three-phase equilibria" and "further experimental results are needed to accurately define the stability region of Nb 3 B 2 ".
The Si content in the Nb ss increased with its Ti concentration ( Figure 4a) and was essentially the same for all the HT alloys, about 0.5 at.%, the latter in agreement with other work on B-free RM(Nb)ICs e.g., [16,17]. Whereas the Si concentration was the same, the B content on the HT alloys varied depending on element(s) in synergy with B (Figure 4b). In the Ti-rich Nb ss and "normal" Nb ss the maximum B concentration respectively was about 6.1 and 3.2 at.%, which corresponded respectively to 1.7 and 1.9 at.%Si. The concentrations of B, Cr and Ti depended on each other, the Ti and B concentrations in the Nb ss respectively increased and decreased with increasing Cr content in the Nb ss (Figure 4c). As the Al + B + Ti content of the Nb ss decreased the VEC of the solid solution increased, whereas the opposite was the case regarding B + Cr + Ti (Figure 4d). The VEC of the Nb ss increases with Ti + Cr or Ti content in the alloys TT2, TT4 and TT8 (Figure 4e,f). The B or Mo content of the Nb ss decreased with increasing Ti concentration in the Nb ss (Table 5).  In the T2 silicide, the B content decreased and the Al and Cr concentrations increased as the Ti concentration increased ( [6] and Table 5). Figure 5 shows maps of the T2 silicide in the AC and HT alloys TT2, TT3, TT4 and TT8 and of the Nb 5 Si 3 silicide in the reference B-free alloys KZ4, KZ5, KZ7 and JG3. The T2 occupies a distinct, separate area in the ∆χ versus VEC map (Figure 5a,b) in agreement with [2,6]. Both VEC and ∆χ increase with increasing <Nb> and the linear fit of the data for T2 and Nb 5 Si 3 is remarkably good (Figure 5c,d).

Hardness
The hardness of the Nb ss decreased or increased with the parameter VEC Nbss when the alloys TT4 and TT8 were grouped respectively with the Cr containing alloy TT2 (Figure 6a) or the Al-containing alloy TT3 (Figure 6b). The same was the case for the reference alloys KZ5 and JG3 when grouped with the Cr or Al-containing alloys KZ4 and KZ7, respectively ( Figure 6). The hardness of the Nb ss in the AC and HT alloys TT2, TT4 and TT8 decreased with increasing VEC Nbss (Figure 6c).
The dependence of the hardness of the Nb ss on solutes in the alloys is shown in Figure 7. The hardness decreased with increasing Ti + B + X or Ti + X content (X = Al, Cr), and this was linked with changes of the parameter VEC Nbss (Figures 4 and 6), and also decreased with increasing Ti content in the Nb ss , the same trend as for the reference alloys (Figure 7d). Note the link between Ti in the solid solution and its VEC (Figure 4f) and the change of hardness with VEC (Figure 6a).   The VEC T2 or ∆χ T2 increased with increasing <Nb> in T2 (Figure 5c,d) and the hardness of T2 increased with VEC T2 (in [2]). The synergy of Cr with B and Ti decreased slightly the hardness of T2 in TT2-AC, which was the highest amongst the AC alloys of this study, compared with TT1-AC, but the T2 in TT2-HT did not retain its hardness after heat treatment, which decreased by 13% compared with 6% in TT1-HT ( [12] and Table 2) and was essentially the same as the hardness of T2 in TT3-HT and the second-lowest amongst all the HT alloys (Table 2). Best hardness retention of the T2 and highest hardness after heat treatment was observed in the case of the alloy TT4. Compared with the alloys ZF4 [30], ZF5 [31] and ZF6 [32], which are similar respectively to TT2, TT3 and TT4 with the addition of Ge instead of B, the T2 had significantly lower hardness than Nb 5 Si 3 in these three alloys. Compared with TT1, the T2 hardness in TT4-HT was not significantly different than that in TT1-HT [12], which was also lower than the Nb 5 Si 3 in ZF3 [33]. It should be noted that the relationship between the hardness of T2 and VEC T2 is opposite that of Nb 5 Si 3 and VEC Nb5Si3 (in [2]).
Alloy hardness versus the parameters VEC alloy , δ alloy and ∆χ alloy is shown in Figure 8a-c. Room temperature specific strength calculated from hardness (σ HV /ρ) alloy is shown in Figure 8d. The addition of B to the reference alloys caused a decrease in VEC alloy and an increase in HV alloy and (σ HV /ρ) alloy . Notice that the B containing alloys occupy specific distinct areas in the HV alloy versus δ alloy or ∆χ alloy maps.  Furthermore, notice the location of the JG3 alloy in the HV alloy versus ∆χ alloy map. The maximum specific strength of 385 MPa cm 3 g −1 for VEC alloy = 4.45 from the parabola for B containing alloys in Figure 8d is significantly higher than those reported for RCCAs to date [1,2], and only slightly lower than that of the RM(Nb)ICs-RCCAs JZ4 and JZ5 [34].
The parameters VEC alloy , δ alloy and ∆χ alloy link with specific solutes and thus allow us (i) to link the hardness and specific strength of alloys with solute additions and; (ii) to use these relationships for the design of alloys using the alloy design methodology NICE [1][2][3]. We demonstrate such relationships for the alloys TT2, TT4 and TT8 and the solutes B, Cr and Ti in Figure 9.   The parameter VEC alloy increased with (Ti + Cr) alloy (Figure 9a) whereas the parameters δ alloy and ∆χ alloy decreased (Figure 9b,c). The shift towards higher HV alloy with decreased VEC alloy for the B-containing alloys (Figure 8a) was thus attributed to the lower (Ti + Cr) content in the alloys TT2, TT4 and TT8. The shift of the latter in separate distinct areas in the maps in Figure 8b,c was attributed to the increase of δ alloy and ∆χ alloy with the decrease of (Ti + Cr) alloy (Figure 9b,c). The parameter VEC alloy decreased and the δ alloy and ∆χ alloy increased with increasing (Ti/Cr) alloy (Figure 9d,e,f). The alloys were separated in the VEC alloy or δ alloy versus (Ti/Cr) alloy maps, as was the case in the VEC alloy or δ alloy versus (Ti + Cr) alloy maps, but in the former two maps the case for Mo-containing alloys JG3 and TT8 was different, with VEC alloy minimum at about 4.45 and δ alloy maximum at about 15 for (Ti/Cr) alloy = 4.75. Note that VEC alloy ≈ 4.45 corresponds to maximum HV alloy and specific strength in Figure 8a,d. Similarly to the ∆χ alloy versus (Ti + Cr) alloy map, the alloys were not separated in the ∆χ alloy versus (Ti/Cr) alloy map (Figure 9c,f). Thus, even though there are relationships of the parameters VEC alloy , δ alloy and ∆χ alloy with (Ti + Cr) alloy , or (Ti/Cr) alloy , and the parameters VEC alloy , δ alloy and ∆χ alloy are used to design new alloys using NICE [1][2][3], only the parameter VEC alloy links hardness or specific strength with specific solutes in NICE, (Ti + Cr) alloy and (Ti/Cr) alloy in the case demonstrated above. Furthermore, the parameters VEC alloy , δ alloy and ∆χ alloy also link with the (Ti + Cr + B) content of the alloys TT2, TT4 and TT8, as shown in the maps in Figure 9g,h,i. VEC alloy is maximum for (Ti + Cr + B) alloy = 35.5 at.%. The HV alloy versus (Ti + Cr + B) alloy map (not shown) shows minimum (685 HV alloy ) and maximum (740 HV alloy ) hardness, respectively, for AC and HT alloys for Ti + Cr + B = 35.5 at.% whereas the HV alloy versus (Ti + Al + B) alloy map (not shown) shows minima of 615 HV alloy and 560 HV alloy with (Ti + Al + B) alloy equal to 34.5 and 35.5 at.%, respectively, for the AC and HT alloys TT3, TT4 and TT8. It should be noted that also the hardness of the solid solution (HV Nbss ) links with VEC Nbss ( Figure 6) and with solutes, namely Ti + X, Ti + B + X (X = Al, Cr), and Ti/Cr (Figure 7) and that solutes in the Nb ss link with its parameters (see Figure 4 for relationships with VEC Nbss for B containing alloys and Figure A1 in the Appendix A for relationships in the reference alloys). Such relationships allow the design methodology NICE to optimise alloy composition for the balance of strength and oxidation properties, see [1][2][3]. Figure 10 compares the density of the base alloys KZ4, KZ7 and KZ5 with that of the equivalent alloys with Ge addition, namely ZF4, ZF5 and ZF6 [30][31][32], and with B addition. It is clear that the B addition had the most significant effect on reducing alloy density. Figure 11 compares the room temperature hardness of AC and HT alloys ZF4, ZF5, ZF6 [30][31][32] and TT2, TT3 and TT4. The addition of Ge to the reference alloys increased the alloy hardness compared with the addition of B. This was attributed (i) to the increase of the hardness of the Nb 5 Si 3 when alloyed with Ge; and (ii) to the lower vol.% of Nb ss in the Ge containing alloys [6,[30][31][32] compared with the alloys TT2, TT3 and TT4. Figure 11. Room temperature Vickers hardness of AC and HT alloys. Odd numbers AC alloys, even numbers HT alloys. TT2 (1, 2), ZF4 (3,4), TT3 (5, 6), ZF5 (7,8), TT4 (9, 10), ZF6 (11,12). See Appendix A for nominal compositions of alloys.

Oxidation
The data for the alloys TT1, TT2, TT3, TT4 and TT8 can help us to infer what the effect of B addition to the reference alloys was on isothermal oxidation at 800 and 1200 • C and to compare the effect of B addition with that of Ge or Sn to the same reference alloys. Amongst the reference alloys KZ4, KZ5, KZ7 and JG3, the alloy JG3-AC had the best oxidation behaviour at both temperatures, did not pest at 800 • C, and followed parabolic oxidation at 800 • C and liner oxidation at 1200 • C (the JG3-HT followed linear oxidation at both temperatures) [35]. The rate constant of the alloys TT1, TT2 and TT3 at 800 • C ( Table 6) was one order of magnitude higher than that of JG3-HT, and of the alloys TT4 and TT8 similar or one order of magnitude lower, respectively. The alloys KZ4, KZ5 and KZ7 formed Maltese crosses at 800 • C [36], like the MASC alloy, and their scales spalled off at 1200 • C, as did the scale of JG3.
At 800 • C, the mass change of TT1 was significantly lower than that of KZ3, which gained more than about 200 mg/cm 2 after just 10 h. The synergy of B and Cr decreased the mass change compared with KZ4 (about 60 mg/cm 2 after 100 h). The Al addition in TT3 also improved the oxidation behaviour compared with TT1 (Table 6). Compared with KZ7, the synergy of Al with B was only marginally beneficial for the first 20 h, after this time the mass change increased. The mass of the alloy TT4 did not change for the first 10 h and after this time it was lower than that of KZ5, which gained about 32.5 mg/cm 2 . The Mo addition in TT8 significantly improved the oxidation behaviour compared with JG3, which gained about 3.5 mg/cm 2 .
At 1200 • C, the synergy of Al and Cr in TT4 had a strong effect on oxidation behaviour (mass change after 90 h: TT1 = 80.3, TT2 = 105.6, TT3 = 66.1 and TT4 = 34.0 mg/cm 2 ). The addition of Mo in the TT8 alloy resulted in a marginal improvement of the oxidation behaviour of TT8 compared with TT4. Figure 12 compares the mass changes of the alloys of this study with their reference alloys after 90 h at 800 or 1200 • C. In each part of this figure, the effect of B addition is indicated by an arrow. With the exception of TT3, whose mass increased more than KZ7 at 800 • C (Figure 12b), the addition of B resulted to lower mass changes at both temperatures, particularly in the case of the alloy TT8. Plots of the mass changes of the B-containing alloys and their reference alloys versus the alloy parameters VEC and δ are shown in Figure 13. According to the alloy design methodology NICE, for improved oxidation behaviour, the alloy design should aim to lower VEC alloy and increase δ alloy [1][2][3]. Figure 13 shows that this was the case for all alloys at 1200 • C, and also at 800 • C where the alloy TT3 was the exception as its mass change was higher than that of KZ7 (Figures 12b and 13a,b).  Compared with the Ge containing alloys ZF4, ZF5 and ZF6 [37] and the Sn containing alloys ZX4, ZX6 and ZX8 [22] (for nominal compositions see Table A1 in the Appendix A), the oxidation of the B-containing alloys TT2, TT3 and TT4 was inferior at 800 • C. Indeed, the above ZF and ZX series alloys did not pest, their scales did not spall off, and followed parabolic oxidation kinetics [37], or parabolic with/without linear kinetics [22]. At 1200 • C the ZF and ZX series alloys followed either parabolic plus linear kinetics [37] or parabolic or linear kinetics [22] and their scales spalled off, the same as the KZ series reference alloys. Only the alloy TT8 did not pest at 800 • C and its scale did not spall off at 800 and 1200 • C.
At 800 • C, the scales that formed on the alloys TT1 and TT2 consisted of SiO 2 , TiO 2 , Nb 2 O 5 , 3Nb 2 O 5 .TiO 2 , and 5Nb 2 O 5 .2TiO2, which are typical oxides in the scales of RM(Nb)ICs [3,35,36], plus B 2 O 3 and B 2 SiO 5 in TT1 plus CrNbO 4 in TT2 ( Figure S2a in Supplemental data). The scales that formed on TT3 and TT4 consisted of the same oxides as TT1, plus AlNbO 4 in TT3 and CrNbO 4 and AlNbO 4 in TT4 (data not shown). At 1200 • C, the scale on TT1 consisted of SiO 2 , TiO 2 , Nb 2 O 5 and 3Nb 2 O 5 .TiO 2, 5Nb 2 O 5 .2TiO 2 , plus B 2 O 3 and B 2 SiO 5 whereas the scale on TT2 consisted of the same oxides plus NbO, and CrNbO 4 ( Figure S2b in Supplementary Materials). The scale of TT3 consisted of the same oxides as TT1 plus AlNbO 4 (data not shown). The presence of CrNbO 4 and AlNbO 4 in the scales of the alloys TT2, TT3 or TT4 is in agreement with the data for the B free alloys KZ4, KZ7 and KZ5 [3,36] and the Ge containing alloys ZF4, ZF5 and ZF6, where also SiO 2 , Nb 2 O 5 , TiNb 2 O 7 (Nb 2 O 5 .TiO 2 ) and GeO 2 were observed [37].
Comparison of the oxidation of the reference alloys KZ4, KZ7, KZ5 [36] with the alloys with B addition, namely TT2, TT3 and TT4, or Ge addition (ZF4, ZF5, ZF6 [37]) or Sn addition (ZX4, ZX6, ZX8 [22]) shows that at 800 • C (i.e., in pest regime) Ge and B, respectively, was most and least effective and that at 1200 • C all three elements were not effective as scale spallation was not suppressed. However, only when B was in synergy with Al, Cr and Mo was the oxidation improved compared with the basis alloy JG3 [35], as the alloy TT8 did not pest and its scale did not spall off at 800 and 1200 • C. Thus, the aforementioned synergy of elements in TT8 had the same effect at both temperatures as that of Al and Cr with Ge plus Sn in the alloy OHS1 [23]. In other words, the alloy TT8 had outstanding (for RMICs, and RHEAs and RCCAs) oxidation behaviour at 800 and 1200 • C with density 6.4 g/cm 3 and specific strength of 377 and 337 MPacm 3 g −1 , respectively, in AC and HT conditions. Another characteristic difference of the alloys with B or Ge addition [37] with the Sn containing alloys ZX4, ZX6 and ZX8 [22] was the fact that in the latter, tin oxide was rarely (if at all) observed in the scale, in contrast with the former alloys where B 2 O 3 or GeO 2 were formed. The latter two oxides form glass with silica.

Conclusions
We studied the effect of B addition on the density, macrosegregation, microstructure, hardness and oxidation of four RM(Nb)IC alloys, namely the alloys TT2, TT3, TT4 and TT8. In actual fact, these alloys were based on four KZ series alloys (the basis/reference alloys) to which B was added. This choice of alloys made it possible to also compare the effect of B addition on density, hardness or oxidation with that of Ge or Sn addition.
The addition of B resulted in B macrosegregation that was highest in TT8 and in increased macrosegregation of Si and Ti, both of which were lowest in TT8. With the B addition, the density of the alloys was decreased significantly and was lower than the alloys with Ge addition. The alloy TT8 had the lowest density of 6.41 g/cm 3 and the highest specific strength at room temperature, which was also higher than that of RCCAs and RHEAs. The Nb ss and T2 silicide were stable in the alloys TT2 and TT3, whereas in TT4 and TT8 the stable phases were the Nb ss and the T2 and D8 8 silicides. The latter silicide was Al free. The T2 had Si + B + Al ≈ 37.5 at.%, Si/B ≈ 4 and <Nb>/<Si> ≈ 1.7 compared with Si + B ≈ 41.5 at.%, Si/B ≈ 0.5 and <Nb>/<Si> ≈ 1.4 of the D8 8 . Compared with the Ge or Sn addition in the same reference alloy, the B and Ge addition was the least and most effective at 800 • C (i.e., in the pest regime), when no other RM was present in the alloy. Like Ge or Sn, the B addition in TT2, TT3 and TT4 did not suppress scale spallation at 1200 • C. Only the alloy TT8 did not pest and its scales did not spall off at 800 and 1200 • C.
The macrosegregation of Si and Ti, the chemical composition of Nb ss and T2, the microhardness of Nb ss and the hardness of alloys, and the oxidation of the alloys at 800 and 1200 • C were also viewed from the perspective of the alloy design methodology NICE and relationships with the alloy or phase parameters VEC, δ and ∆χ. The trends of these parameters and the location of alloys and phases in parameter maps were found to be in agreement with NICE, the one exception being the macrosegregation of Si in TT8 when compared with its basis/reference alloy.
Author Contributions: Experimental work, T.T.; funding, P.T.; supervision, P.T.; formal analysis, T.T. and P.T.; draft preparation, T.T.; review, P.T., final paper, P.T. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: All the data for this paper is given in the paper and its Supplementary Materials, other data cannot be made available to the public.
where c i , r i , χ i , (VEC) i and T mi , respectively, are atomic percentage, atomic radius, Pauling electronegativity, VEC and melting point of ith element, ∆ mix AB is the mixing enthalpy of binary liquid AB alloy and R is the gas constant.