Fig 6 SEM cross section of 20 l m thick sputtered columnar AlN on Pt with Mo on

Fig. 6 SEM cross section of 2.0 l m thick sputtered columnar AlN on Pt with Mo on top Pt Ti AlN (0112) - (0112) - (0112) - (0112) - [0001] [0001] 100 nm Fig. 7 STEM Z-Contrast image of a vertical slice of AlN cones on Ti–Pt substrate. The growth direction and the lateral surfaces of the cones are marked ii i iii 0110 - 111 002 113 - 0002 0111 - - 0002 0111 - 0110 - Pt AlN ii i iii [2110] -- - [2110] - [110] 5 nm Fig. 8 HRTEM micrograph of the Pt/AlN interface with the corre- sponding FFT. ( i ) Pt along the [110] zone axis, ( ii ) ? ( iii ) AlN along the [2–1–10] zone axis Microsyst Technol (2012) 18:787–795 791 123

3.5 Intrinsic layer stress The collected experimental results of different sputter processes versus the applied rf bias power levels demon- strated that the intrinsic stress in AlN layers could be tuned over a wide range of about 2 GPa (Fig. 9 ). This is in good agreement with earlier reported results (Lanz et al. 2006 ; Medjani et al. 2006 ; Martin and Muralt 2006 ; Zhang et al. 2005 ). The reason for the intrinsic AlN stress is caused by energy of the ions impacting on the substrate surface (Dubois and Muralt 2001 ; Lee and Lee 1997 ). The film stress curve of AlN films sputtered at 200 ° C is strongly shifted compared to curves recorded at 400 ° C. 3.6 Piezoelectric measurements The effect of the rf bias power level variation on the AlN material parameter e 31,f , d 33,f , dielectric permittivity and FWHM r. c. are summarized in Fig. 10 . Usually three test samples from a 200 mm wafer were measured. The vertical bars represent the 1 r standard deviation. Figure 10 a shows that the effective transverse coeffi- cient has no dependency of the rf bias power. This is in good agreement with earlier published results from (Karakaya et al. 2008 ). The mean transverse coefficients are between - 1.0 and - 1.2 C/m 2 . Maximum values of - 1.25 C/m 2 were reached and are little less than the best published results (Lanz et al. 2006 ; Muralt et al. 2005 ). Also, for the corresponding effective longitudinal coeffi- cient no dependency on the applied rf power could be observed (Fig. 10 b). Dubois and Muralt ( 2001 ) noticed that the d 33, f is independent from the bias voltage when a threshold values is exceeded. Therefore we assume that in all our experiments we are beyond this threshold. Mean d 33,f values up to 4.7 pm/V were reached. The maximum measured d 33,f was 5.2 pm/V and is comparable with the highest published results (Lanz et al. 2006 ; Martin et al. 2004 ). As for the piezoelectric coefficients the dielectric per- mittivity is constant throughout the complete rf bias power range (Fig. 10 c). In the latter case the values are about 10.5 Fig. 9 Intrinsic stress of AlN films with a thickness of 0.5, 1.0, 2.0 l m at 400 ° C and for 1.0 l m at 200 ° C sputtered on Pt as a function of rf bias power level (a) (b) (c) (d) Fig. 10 AlN material parameter a |e 31,f | , b d 33,f , c dielectric permit- tivity and d FWHM r. c. as a function of the rf bias power level sputtered on Pt at 400 ° C. Error bars indicate the 1 r standard deviation 792 Microsyst Technol (2012) 18:787–795 123