Microstructural development during diffusion bonding of α-silicon carbide to molybdenum

ELSEVIER

Microstructural

Materials

Science and Engineering

Al 91 (1995) 239-247

development during diffusion bonding of a-silicon carbide to molybdenum A.E. Martinelli, R.A.L. Drew

Department of Mining and Metallurgical Engineeting, McGill University 3450 University, Montreal, Quebec H2A 3A 7, Canada Received 3 1 May 1994

Abstract Silicon carbide was joined to molybdenum by solid state bonding at temperatures ranging from 1200 “C to 1700 “C. The interfaces were characterized by scanning electron microscopy, electron probe microanalysis, and X-ray diffraction. Diffusion of Si and C into MO resulted in a reaction layer with two main phases: Mo,Si, and MO&. For temperatures higher than 1400 “C a ternary phase of composition Mo,Si,C was also formed, and at 1700 “C nucleation of MoC was observed. Keywords: Diffusion; Silicon; Carbides; Molybdenum

1. Introduction Advanced ceramics are potential structural and electrical candidates for high temperatures and severe environmental conditions. Among several options, silicon carbide and silicon nitride are the most attractive materials for their low density combined with good mechanical properties and chemical stability at high temperatures [ 11. For most applications, ceramics are, however, incorporated to metals as components for engines, in the fabrication of composites, or as coatings. Therefore, the use of advanced ceramics depends on the reliability of ceramic-metal joining processes and the properties of the resulting interfaces. The choice of a ceramic-metal combination is based on the design and properties required for the joint. For example, Sic-Mo joints have been studied for applications in fusion reactor technology [2] and as electrical contacts [3]. As a consequence of their covalent bonding nature, ceramics have generally lower coefficient of thermal expansion (CTE) than metals. This mismatch can result in high thermal stresses which originate during cooling from the joining temperature and, consequently, form mechanically weak joints. This problem occurs even 0921-5093/95/$9.50 0 1995 - Elsevier Science S.A. All rights reserved SSDI 0921-5093(94)09633-g

with refractory metals with low CTEs. Therefore, for each ceramic-metal system, it is important to select the proper joining technique and its corresponding parameters. This is the only way to ensure the production of reliable joints, characterized by the formation of continuous and flaw-free interfaces. Among the techniques employed to join ceramics to metals, active metal brazing and diffusion bonding are the most common. In this work, the solid state diffusion bonding method was applied to join SIC to MO. For this process, low melting brazed interlayers are not required, resulting in interfaces able to withstand high temperatures, severe service conditions, and high mechanical stresses [4].

2. Experimental procedure Sic-Mo joints were prepared by solid state diffusion bonding using a hot press equipment. The starting materials used were pressureless sintered a-Sic Hexoloy grade SATM (Carborundum Co., Niagara Falls) and MO sheet (99.95%, AESAR Division, Johnson & Matthey, Toronto). Sic was supplied as 150 x 150 x 6 mm3 plates and MO as 50 X 50 X 2.5 mm3 sheets.

240

A.E. Martinelli, R.A.L. Drew

/

Materials Science and EngineeringA

The experimental samples had cross-section 9.0 x 9.0 mm* and thickness of the supplied blocks. The surfaces to be joined were ground and polished with 1.O pm diamond paste finish and cleaned with isopropanol in an ultrasound bath for 5 min. Joining was carried out under a vacuum of 150 mTorr for temperatures varying from 1200 “C to 1700 “C. The holding time ranged from 10 min to 4 h and the uniaxial pressure applied on the samples from 5 to 100 MPa. The heating rate was 15 “C mm-’ and cooling was 5 “C min- ’ during the first 500 “C and 15 “C min-’ during the remainder of the cycle. The temperature and pressure profiles are presented in Fig. 1. The samples were placed in the furnace inside a graphite die (SPEER, Canada, grade 3499) surrounded by a powder bed consisting of pure boron nitride (99.5%, AESAR Division, Johnson & Matthey, Toronto). The applied pressure was maintained within a range of 10% of the nominal value and the temperature within a range of 5 “C. Cross-sections of each joint were cut using a low speed diamond saw and mounted in a cold-setting epoxy resin. Specimens were then polished with diamond paste and alumina powder (0.05 pm finish). Micrographs were obtained from a JEOL JSM-840A scanning electron microscope and interfacial phase analyses were performed in a CAMEBEX electron microprobe with wavelength-dispersive system. X-ray line analysis was performed in a JEOL-8900 superprobe. For scanning electron microscopy (SEM) and electron probe microanalysis (EPMA) characterization, the samples were coated with carbon using an EDWARDS E306A coating system. X-ray diffraction was performed on fracture surfaces of the joints using Cu Ka radiation on a

(1995) 239-247

Rigaku RU-200B diffractometer. The angular range between 5” and 100” was investigated with an angular velocity of 0.6” min I, accelerating voltage of 50 kV, and current of 150 mA. The results were compared with theoretical predictions obtained from the F*A+C*T program [5]. In this software, the routine REACTION calculates changes in extensive thermochemical functions for a balanced chemical equation. This routine was used to obtain the variation in the Gibbs energy for the formation of molybdenum silicides and carbides as a function of the temperature. These values were compared with that of silicon carbide to investigate the products possibly formed from the reaction between molybdenum and silicon carbide.

3. Results and discussion 3.1. Thermodynamic

evaluation of the SC/MO system

When Sic is in contact with a metal (Me), the reaction follows one of three routes [6]: Me + Sic -

Me silicide + Me carbide

(1)

Me + Sic -Me silicide + C

(2)

_2:; -40 G

zQ i;j ‘El Z z 3 0q -lOO-

-

MO& + +++++++++++++

+

Mo,Si, ~~xxxxxxxx~xx)(x

x *

xx .

.

*****..+++~+~**

-120-

Mo,Si I

-140 0

I 1200

600

1600

T (“C) Time (h) Fig. 1. Profile of temperature of Sic/MO.

and pressure

during hot-pressing

Fig. 2. Thermodynamic silicides relative to F*A*C*T ).

stability of molybdenum carbides and Sic (data obtained from program

A.E. Martinelk, R.A.L. Drew

Me + Sic -

Si + Me carbide

/

Materials Science and Engineering A191 (199&i)239-247

(3)

Comparing the thermodynamic stability of Sic with those of metal silicides and carbides it is possible to predict the most probable products to be formed. When SIC is in contact with MO, it decomposes into Si and C. According to Fig. 2, the reaction should follow route (2), initially forming Mo,Si and free C. In a second step, diffusion of Si into Mo,Si would stabilize Mo,Si,. For temperatures higher than 1600 “C, route (1) is more favourable, because Mo,C becomes more stable than Sic. The silicide present would be MoSi,, formed from the diffusion of Si into Mo,Si,. Although expected only for high temperatures, the formation of Mo,C as a reaction product between MO and Sic is well established for temperatures lower than 1700 “C [2,7,8]. Okamoto [6] suggested that commercial Sic is somewhat less stable than that used in performing the thermodynamic calculations. No thermodynamic information was available from the F*A*C*T databases for ternary phases in the Mo-Si-C system, although the formation of Mo,Si,C has been reported [ 21. 3.2. Sample characterization Diffusion couples of Sic and MO were hot pressed in the temperature range from 1200 “C to 1700 “C. The resulting interfaces were formed in the MO side of the sample by diffusion of C and Si. For temperatures ranging from 1200 to 1300 “C, the interfaces consisted of one layer with two intermixed phases. EPMA performed on a sample produced at 1250 “C for 2 h (Fig. 3) indicated that these two phases were Mo,Si, and MO&.

Fig. 3. Backscattered image of joint produced at 1250 “C for 2 h under 10 MPa.

241

Si probably diffused into MO faster than C, despite its higher atomic radius. It has been suggested that the rate of diffusion of non-metals in transition metals is related to the ionization potential of the non-metal rather than the size of the atom [9]. The lower the ionization potential of the non-metal atom, the easier the diffusion in the transition metal. Following this argument, Si diffuses into MO easier than C, as a result of its lower ionization potential (8.28 eV) as compared with that of C ( 11.24 eV). However, the high affinity of MO for Si resulted in the formation of a silicide, initially MO,!%, preventing Si from diffusing extensively into the MO. Mo,Si transformed rapidly into Mo,Si, which indicates considerable decomposition of Sic and a large supply of Si. The diffusion of C into MO resulted in Mo,C. Although surface and grain boundary diffusion usually have lower activation energies as compared with lattice diffusion, the formation of MoC, _x showed that the latter process also took place, especially at high temperatures. No flaws were observed in the interfaces formed at 1200 “C, possibly because of the relatively low joining temperature and amount of reaction. Increasing the temperature to 1250 “C resulted in a thicker reaction layer and formation of flaws (Fig. 3); however, the overall interface was not damaged. Fig. 4 shows the sample produced at 1300 “C for 1 h. The interface was once more formed of MosSi and MO&. However, it was observed that the formation of Mo,C extended even deeper inside the MO, and at 1400 “C (Fig. 5) it consisted of a separate interfacial layer. Backscattered imaging showed a contrast effect in the Mo,C region probably related to a texture effect. A different contrast was also observed in the Mo,Si,

Fig. 4. Interface of sample hot pressed at 1300 “C for 1 h under 5 MPa (backscattered electron image) (BEI).

242

A.E. Martinelli, R.A.L. Drew

/

Materials Science and Engineering A191 (1995) 239-247

Fig. 5. Interface produced at 1400 “C for 1 h under 5 MPa shows MozC as a separate layer and formation of Mo,Si,C (BEI).

region in contact with Sic. EPMA showed a composition which corresponded to Mo,Si, but contained a trace amount of C. It is believed that this was the ternary phase Mo,Si,C. The formation of flaws at 1300 “C occurred preferentially in the Mo,Si, region (Fig. 4). Cracking in the Mo,Si, regions was attributed to an increase in volume (in relation to the volume of the metal) during the formation of the silicide and to thermal expansion mismatch between the Mo,Si, and SIC. The formation of Mo,Si, from the original MO involves a 46% volume increase, while the formation of Mo,C resulted in only a 20% increase. Because of the brittle nature of Mo,Si,, cracking resulted from the residual stresses and was due to the volume expansion. In addition, the interface had SIC in contact with Mo,Si, and MO in contact with Mo,C. The CTE mismatch Aa between Sic and Mo,Si, is 2 X lo-” C- ’ and between MO and Mo,C is 1 X 10-6”C-‘. The high thermal expansion mismatch associtated with the brittle Sic and Mo,Si, was another factor that contributed to the formation of flaws in the silicide. All the interfaces observed showed an irregular pattern with wavy phase boundaries between Mo,Si, and Mo,C and between Mo,C and MO. It has been suggested that this is a consequence of different concentrations of the compounds in the interface [2]. The reaction to form Mo,C may have.different kinetics at different points along the interface, depending on the distance that C has to travel through the Mo,Si,. This fact reflects in a diffusion path that crosses the tie-lines of the Mo-Si-C phase diagram. At 1500 “C a sample was hot pressed for 1 h. Excessive residual stresses resulted in debonding of the joint on cooling. The corresponding fracture surfaces of Sic

Fig. 6. Secondary electron images of (a) SIC and (b) MO fracture surfaces of interface produced at 1500 “C for 1 h under 5 MPa.

and MO are shown in Fig. 6. Fragments of the interface can be observed on the Sic surface (Fig. 6(a)); however, EPMA analysis indicated that most of the reaction layer remained on the MO side of the joint. The MO fracture surface (Fig. 6(b)) showed a distribution of SIC particles. The concentration of these particles was non-uniform and higher close to the corners of the sample as a consequence of the stress distribution curve. As a result of its geometry, and the corresponding formation of regions in tension and compression in the ceramic, stresses developed preferentially closer to the corners of the rectangular surface, near the interface on the ceramic side [lo]. To study the effect of the interface thickness on the residual stresses, a second sample was produced at 1500 “C, limiting the holding time to only 10 min. Fig. 7 shows a backscattered image of the resulting interface. Cracking was severely reduced and a reliable joint was obtained. An Mo,Si,C layer of 4 pm in average thickness was formed in contact with Sic and identified

A.E. Murtinelli, R.A.L. Drew

/

24.;

Materials Science and Engineering A191 (1995) 239-217

Mo,Si,C

Fig. 7. Sample hot pressed at 1500 “C for 10 min under 10 MPa (BEI).

by EPMA. Fig. 8 shows a wavelength-dispersive spectroscopy (WDS) line analysis with an X-ray profile of the atomic species across the interface. The different reaction layers can be identified through the variation in the X-ray intensities as a function of the distance from the original interface. When MO is saturated with Si, cubic Mo,Si is formed. Further addition of Si stabilizes tetragonal MO, Si,, The next step in the reaction should be the formation of MoSi, as was observed in siliconizing of MO [9]. However, the presence of C prevented the formation of MoSi,. C may diffuse preferentially through the grain boundaries of Mo,Si,; however, lattice diffusion also occurred, and a fraction of C atoms diffusing through Mo,Si, remained in the structure stabilizing the hexagonal Mo,Si,C. In the new structure, C occupies interstitial positions at the centre of octahedra formed by MO atoms [ 111. The failure of the joint produced at 1500 “C for 1 h confirmed that the amount of reaction is an important factor in determining the mechanical strength of the interface. Cracking resulted not only from thermal stresses generated during cooling from the joining temperature but also from excessive interfacial reaction. The interface formed for 10 min was relatively thin compared with that formed for 1 h, resulting in a reliable joint with little cracking, although residual thermal stresses were the same for the two samples. Joining carried out for temperatures higher than 1500 “C resulted in failure of the joints. Fig. 9 shows the results from X-ray diffraction performed on the SIC and MO fracture surfaces of a sample hot pressed at 1650 “C. Mo,Si, was the main phase present on the MO surface (Fig. 9(a)). The structure of this phase was identified as tetragonal and the main peaks were (411)

Distance

(pm)

-10pm

Fig. 8. WDS line analysis of interface hot pressed at 1500 “C for 10 min under 10 MPa.

and (32 1), although the (32 1) peak overlapped with the (102) peak of Mo,Si,C. In addition, hexagonal Mo,C was observed. The relative intensities of the carbide peaks increased with the angle of diffraction owing to the higher penetration of X-rays. The MO fracture surfaces were characterized by a change in composition with distance from the surface, and therefore diffraction at high angles included a deeper region of the interface. For angles greater than 60”, diffraction originated from the interface layer rich in Mo,C and, depending on the local thickness of the interface, even the MO layer diffracted, as could be seen by the MO (220) peak at 28 = 87.65”. An intense peak was observed at 20 = 35.7”. Comparing Figs. 9(a) and 9(b), this peak was identified as the (006) plane of hexagonal a-Sic and originated from Sic particles dispersed on the MO surface. Another relatively intense line was observed at 2 0 = 20.9”. It was attributed to free carbon also present on the MO fracture surface. The diffraction pattern obtained from the Sic fracture surface (Fig. 9(b)) matched the a-Sic standard with residual Mo,Si, and Mo,Si,C. Although overlapping of Mo,Si,C with Mo,Si, and Sic occurred, the presence of Mo,Si,C was established by the (112) peak. Mo,C peaks were not observed showing that the joint did fail along the interface between Mo,Si,C and SIC.

244

A.E. Martin&

R.A.L. Drew

4-o

/

Materials Science and Engineering A191 (1995) 239-247

*

6.0

.

90.

10-o

SO.

1c

3

at 1650 “C for 30 min under 5 MPa: A, Sic;

A,

7.0

.

6.0

.

A

‘9’

I 1.-..,....‘.---,-.--‘...~.,.--~....-,.-.-..-..,...-.-...,-..~...-.~.--~~~~~-[~--.~~~-’ 20.

30.

40.

(b)

Fig. 9. X-ray diffraction of (a) MO and (b) Sic fracture MO; ?? , MO,%,; 0, Mo,Si,C; 0, Mo,C; 0, MoC.

60.

50. 28

70.

60.

(“1

surfaces of sample hot pressed

At 1700 “C, samples were hot pressed for 1 h under different pressures. X-ray diffraction of the MO fracture surface of the sample prepared under 100 MPa showed the presence of Mo,Si,C and Mo,Si, (Fig. 10). The crystal structure for Mo,Si,C was identified as hexagonal. The most intense peak originated from the ( 112) planes. MO and MO carbide peaks were observed at high angles of diffraction. The most probable carbide formed was hexagonal B-Mo,C; however, the peak at 28 = 75.7” showed an anomalous high intensity. This could be related to a variation in the penetration of X-rays accompanied by strong texture of the (20 1) direction, or this peak may correspond to the (203) direction of hexagonal r-MoC. The relative intensity of the (203) diffraction of q-MoC is much higher than that of /3-Mo,C (20 1). In addition, other peaks may also correspond to hexagonal r-MoC. However, their angular positions overlapped with peaks of other phases. In order to elucidate whether q-MoC formed at 1700 “C, X-ray diffraction was performed after

gradually removing layers of reaction material from a sample hot pressed under 20 MPa. Fig. 11 shows a cross-section of the MO fracture surface of that sample and contains all the interfacial layers described so far. First, the original interface was analysed and the presence of Mo,Si,C and Mo,Si, as the main phases was confirmed (Fig. 12(a)). Mo,C was identified at high angles. Then, the sample was ground to eliminate the ternary phase and X-ray diffraction was performed on the mixed layer (Fig. 12(b)). Mo,Si, was identified as the main phase although Mo,Si,C was still present. At 28 = 36.7”, the ( 102) peak of hexagonal q-MoC was observed. However, this position overlapped with the (002) peak of Mo,Si,. The last grinding step was performed until optical microscopy of the cross-section showed only the carbide layer in contact with MO, at which point about 70 pm had been removed from the sample. X-ray diffraction was then performed resulting in a more accurate determination of the type of carbide present (Fig. 12(c)). Although residual Mo,Si, was still detected, carbide peaks were observed over the whole

A.E. Martinelli, R.A.L. Drew

/

245

Materials Science and Engineering A191 (1995) 239-247

a

b i

20.

30.

40.

60.

50. 20

diffraction of MO fracture surface of sample hot pressed at 1700 “C for 1 h i, Mo?C; 0, MoC.

angular range. /3-Mo,C was identified as the main carbide formed; however, peaks of q-MoC were also observed. The most intense peak for /3-Mo,C was observed at 28= 39.49” corresponding to the (101) planes, followed by the (002) which showed some texturing. The presence of q-MoC was determined by its most intense peaks: (102) and (103). Although the (102) peak was superimposed with Mo,Si, (002), comparing the relative intensities of Mo,Si, (002) and (202) in Fig. 12(b) it could be easily seen that the peak at 2 19= 36.78” (Fig. 12~) could not be explained soley by Mo,Si, (002). Moreover, the diffraction pattern for r-MoC shows this peak as its most intense. Other peaks also corresponded to v-MoC, such as the (203) and the ( 1011) reflections. For these peaks there was no overlapping, and therefore their presence could not be related to any other phase. The transformation from /3-Mo,C to r]-MoC can be explained based on the MO-C phase diagram [ 121. MO carbides are of interstitial nature. When C diffuses into MO, it tends to be accommodated in the octahedral interstitial sites of the MO b.c.c. structure. However, these sites are too small for C, especially in the c direction [ 131. This situation forces a change in crystal structure of MO from b.c.c. to h.c.p. which contains larger

80.

90.

100

(“I

Fig. 10. X-ray o.Mo,Si,C;

70.

under 100 MPa: A , Sic;

A, MO; w, Mo,Si,;

Mo,C --+

Mo,Si, +-_) Mo,C

Mo,Si,C-+

Fig. 11. Cross-section of MO fracture surface indicating layers analysed by X-ray diffraction. Sample prepared at 1700 “C for 1 h under 2- MPa (BEI).

octahedral interstitial sites arranged in a simple hexagonal structure. In the Mo,C structure, C occupies statistically half of the available interstitial sites. However, the actual amount of C present can vary drastically, resulting in a wide range of compositions for these compounds. Diffusion of C into the Mo,C structure at high temperatures stabilizes hexagonal qMoC, -x. /3-Mo,C crystallizes in the L3’ structure

246

A.E. Martinelli, R.A.L. Drew

/

Materials Science and Engineering A191 (1995) 239-247

Fig. 12. X-ray diffraction of MO fracture surface of sample hot pressed at 1700 “C for 1 h under 20 MPa: (a) complete (b) mixed layer; (c) carbide layer; A, Sic; A, MO; m, MosSi,; 0, Mo,Si,C; 0, Mo,C; 0, MoC.

(4

(b)

reaction layer;

rat.%, B-Mo,C transforms into substoichiometric MoC,_,, provided that the temperature is higher than 1657 “C. The similarity of crystal structures observed between the two carbides may also confirm this fact, assuming that /~-MO& undergoes a displacive transformation into q-MoC, --x. In the new structure, C occupies statistically two-thirds of the octahedral interstitial sites of the MO h.c.p. structure. The stacking sequence of MO is ABCACBABCACB . . . and the complete stacking sequence AXBX’CX”AX”CX’BXAXBX’CX”AX”CX’BX.. . (Fig. 13(b)).

4. Conclusions Fig. 13. Stacking sequence of MO and C atoms in (a) hexagonal Mo2C and (b) hexagonal MoC, --I: 0, MO atoms; 0, C sites.

whereas MO forms an h.c.p. lattice with stacking sequence ABAB . . . and the carbon atoms forms a simple hexagonal structure by occupying octahedral interstitial sites of the MO h.c.p. structure. The complete stacking sequence is, AXBXAXBX.. . (Fig. 13a). The homogeneity of /3-Mo,C ranges from 31.2 to 33.3 at.% of carbon. If the amount of carbon reaches 39

The results obtained from SEM, EPMA and X-ray diffraction showed that two main phases are formed by solid state reaction between Sic and MO: Mo,Si, and Mo,C. The presence of Mo,C was observed from 1200 “C to 1600 “C, despite the fact that its formation is unlikely to occur for temperatures lower than 1600 “C, according to thermodynamic predictions. A ternary phase of composition Mo,Si,C was also observed. This phase was identified for temperatures higher than 1500 “C; however, it is believed that traces

A.E. Murtinelli, R.A.L. Drew

/

Materials Science and EngineeringA

of this phase were also present at lower temperatures. The kinetics of formation of this phase is strongly dependent on the temperature. The suggested sequence of reaction is that, after the formation of tetragonal Mo,Si,, C diffuses into the octahedral interstitial sites of this phase causing it to transform to hexagonal Mo,Si,C. The main carbide formed was identified as hexagonal /3-Mo,C. At 1700 “C, however, MoC was also observed. Carbon accommodates in the Mo,C structure precipitating non-stoichiometric hexagonal q-MoC, -_i. The amount of reaction was an important factor in determining the mechanical properties of the interface, particularly when temperatures of the order of 1500 “C or higher were used. Residual stresses generated during cooling from the joining temperature and the difference in volume resulting from the formation of Mo,Si, led to failure of the joint, by debonding the interface between Mo,Si,C and SIC.

Acknowledgements

The authors would like to thank the Natural Science Research Council of Canada and Engineering (NSERC) for partial support of this research. A.E. Martinelli would like to express his gratitude to CNPq

(199.5) 239-247

247

Brazil for the scholarship grant. Pratt & Whitney Canada are also gratefully acknowledged for the supply of ceramic materials.

References

[ll

T.C. Chou, A. Joshi and T. Wadsworth, J. Vat. Sci. Technol. A, 9(3)(1991) 1525-1534. [21 F.J.J. Loo, EM. Smet, C.D. Rieck and G. Versoi, High Temp. High Pressures, 14 ( 1982) 25-3 1. [31 K.L. Moazed, Metall. Trans. A, 2S (7) ( 1992) 1999-2006. 141 G. Ellsner and G. Petzow, ISIJ Zm., 30 (12) (1990 101 l-1032. [51 W.T. Thompson, A.D. Pelton and C.W. Bale, F*A+C* TGuide to Operations, &ole Polytechnique, Montreal, 1985. T. Okamoto, ZSZJZnt.,30(12) (1990) 1033-1034. ;;I A. Horiguchi, K. Suganuma, Y. Miyamoto, M. Shimada and M. Koizumi, J. Sot. Muter. Sci. Jpn., 35 (388) (1986) 35-40. [81 K. Kurokawa, Nippon Kinzoku Gakkai Kaiho, 29 (11 I (1990)931-938. [91G.V. Samsonov and A.P. Epik, Coatings of High-Temperuture Materials, Plenum, New York, 1966. [lOI K. Suganuma, T. Okamoto and M. Koizumi. J. Mater. Sci.. 22(1987)3561-3565. [Ill L.E. Toth. Transition Metal Carbides and Nitrides. Academic Press, New York, 197 1. [I21 J. Kouvetakis and L. Brewer, ./. Phase Equilib., 13 (hi (1992)601-604. [131 C.S. Barrett and T.B. Massalski. Structure qf A4etals. McGraw-Hill, New York, 1966.