Stress reduction in a-C:H coatings through the addition of nitrogen to the feed gas

Stress reduction in a-C:H coatings through the addition of nitrogen to the feed gas F. Rabbani a, *, R. Escobar Galindo b , W.M. Arnoldbik c , S. van der Zwaag d , A. van Veen b, F , H. Schut b a Department of Materials Science, Delft University of Technology, Rotterdamsweg. 137, 2628 AL Delft, The Netherlands b Interfaculty Reactor Institute, Delft University of Technology, Mekelweg. 15, 2629 JB Delft, The Netherlands c Surfaces, Interfaces and Devices, Debye Institute, Utrecht University, P.O. Box 80.000, 3508 TA, Utrecht, The Netherlands d Faculty of Aerospace Engineering, Delft University of Technology, Kluyverweg. 1, 2629 HS Delft, The Netherlands Abstract Intrinsic stress in amorphous hydrogenated carbon (a-C:H) coatings was reduced through addition of 10–20 sccm N 2 to the feed gas. The compressive stresses observed in this study were in the range 0.91 to 1.6 GPa. Approximately 2–3 at.% nitrogen was incorporated into the coatings as determined using elastic recoil detection (ERD). Raman spectroscopy shows that the G peak of samples made with nitrogen is shifted to higher wave numbers, and that there is a more pronounced ‘shoulder’ at the D peak position. It is hypothesized that nitrogen addition to the feed gas at a flow rate of 20 sccm enhances the formation of aromatic rings associated with the D peak. The stress reduction noted for nitrogen addition at 10 sccm was contributed to a reduction in the mean coordination number of the network. Positron beam analysis (PBA) was used to show that in a high temperature deposition the interface of the nitrogen-containing coating has more open space. This phenomenon can contribute to a reduction of the compressive stress by reducing the interfacial stress. Annealing tests were performed to monitor this open volume using Raman and PBA analysis. Keywords: Amorphous hydrogenated carbon; Vibrational properties characterization; Defect characterization 1. Introduction Some of the attractive properties of amorphous hydrogenated carbon (a-C:H) coatings are their low coefficient of friction, good wear rate, chemical inertness and low surface energy. A problem in the manufacture of these coatings is their high intrinsic stress, which results in coating failure. In this research the adhesion of a-C:H coatings onto steel and Si substrates was noted to improve with the addition of nitrogen to the gas plasma. Other researchers have also reported better adhesion and/or a reduction of the coating stress with nitrogen incorporation [1–6]. The following systematic study was undertaken to investigate the reduction in the intrinsic stress of a-C:H coatings deposited onto Si wafers with nitrogen in the gas feed at 10 and 20 sccm flow rates. These a-C:H coatings were produced by a nonconventional process: the decomposition of a reactive plasma was achieved in a PVD chamber by applying a DC-bias voltage to the substrate table while gases were introduced into the evacuated chamber at set flow rates. The gas composition was varied from pure CH 4 ,CH 4 +Ar, to CH 4 +H 2 , in each case a sample was also deposited with nitrogen in the gas feed. It was seen that addition of 10 sccm nitrogen had some effect but not as much as 20 sccm in reducing the intrinsic stress. This stress was in all the cases compressive and in the range of stresses reported for these coatings in the literature [7]. As the deposition temperature is also a variable that affects film growth, two samples were made at 300 jC to isolate the effect of nitrogen addition. In all cases, the compressive stress in the coatings decreased with the addition of nitrogen. In depo- * Corresponding author. Tel.: +31-1527-89518; fax: +31-1527-86730. E-mail address: [email protected] (F. Rabbani). F Deceased 3 January 2004. sitions where heating was not applied to the process, the elastic modulus was also reduced with nitrogen addition at 20 sccm. In the case of the high temperature depositions, the reduction in the elastic modulus was noted for both samples. The stress in the coatings was determined through radius of curvature measurements made on the Si substrate before and after deposition. The mechanical properties of the coatings such as the elastic modulus and hardness were measured using nano-indentation testing. Elastic recoil detection yielded information on the composition of the films and show that 2–3 at.% nitrogen is incorporated. This research uses positron beam analysis (PBA) and Raman spectroscopy as complementary techniques to study a-C:H coatings. As hydrogen present at voids essentially masks these defects for PBA detection [8], a series of annealing tests were carried out to ‘isolate’ open space after out diffusion of hydrogen. Unique to this study is the finding that the D peak becomes more pronounced, significant of augmentation in size or number of aromatic ring structures. 2. Experimental A Hauzer PVD machine was used to generate a reactive glow made from the decomposition of a range of gas mixtures with and without N 2 , where CH 4 was the carbon containing gas. Even though a PVD chamber was used the process is unique in that target sputtering was not employed. The PVD chamber was first pumped down to a vacuum of approximately 10 4 Pa. Prior to the coating procedure an etch step was included using Ar gas to remove the surface oxide layer on the Si substrate. The plasma was generated through the application of a bias voltage of 550 V-DC to a rotating substrate table, while the gases were introduced at specific flow rates high enough to generate a glow discharge. The coatings were deposited onto Si (100) single crystal substrates, for a deposition time of 1 h. During deposition the process temperature rises gradually to some equilibrium value less than 300 jC due to the heat generated by ion bombardment. Two samples were made with temperature regulation, i.e. the deposition was started at 300 jC and this temperature was maintained during the test by applying external heating not generated by the plasma itself. The annealing tests were conducted at 150, 300, 400, 500 and 600 jC in a vacuum of 10 5 Pa for 30-min intervals. PBA and Raman spectroscopy were used to monitor the changes that occurred in the coatings between each annealing experiment. 2.1. Stress measurements Stress measurements were made by determining the radius of curvature of the Si wafer before and after deposition, using a bending laser beam method. Details of the experimental set-up can be found in Ref. [9]. The stress was calculated using a modified Stoney equation [10]: r¼½ESi=ð1mSiÞðt2 Si=6tcÞð1=Rc1=RSiÞð1Þ where ris the stress in the coating, E Si the elastic modulus of the Si wafer, m Si the Possion’s ratio of the Si wafer, t Si the thickness of the Si wafer, t c the thickness of the coating (determined by weight gain measurements and assuming a density for the a-C:H coating of 2.010 +3 kg m 3 ). R Si is the radius of curvature of the Si substrate before deposition and R c the radius of curvature after deposition. 2.2. Nano-indentation testing The elastic modulus and hardness of the coatings were measured with a Hysitron TriboScope R nanomechanical test instrument equipped with a Berkovich diamond tip. The software calculates the elastic modulus by taking the linear portion of the unloading curve. Simultaneously, hardness is calculated by subtracting the elastic displacement from the load-displacement data [11]. 2.3. Raman A Renishaw Raman microscope system 2000, using the 514.5 nm line of an Ar ion laser was used for the spectroscopy analysis. The measurements were made at a laser power setting of 2 mW, and the system was calibrated with a Si specimen. Grams 32 software was applied to subtract a linear background from the spectra and to fit two curves of a mix Gaussian–Lorentzian function in the region 1000– 1750 cm 1 . All the fit parameters such as linewidth, positions and areas were allowed to vary. The G peak is due to the relative stretching motion of sp 2 carbon atoms in rings or chains; and the D peak is due to the breathing modes of aromatic rings [12]. The I d /I g ratio was determined based on peak areas. 2.4. PBA The PBA experiments were performed with the Delft Variable Energy Positron beam (VEP) [13]. The positrons were injected in the samples with energies tuned between 100 eV and 30 keV. The maximum implantation energy corresponds to a typical mean implantation depth of f4Am in materials with a density of f3gcm 3 . All experiments were carried out at room temperature under a vacuum of approximately 10 6 Pa. PBA results are described in terms of two parameters describing the Doppler broadening of the 511 keV annihilation photo-peak. The Sparameter indicates the fraction of positrons that annihilate with low momentum electrons (small Doppler broadening) such as valence or conduction electrons. This parameter is related to the open volume defects present in the sample such as vacancy clusters and/or interfaces with misfit. Sincreases as the open-volume defects in a material increase, and larger values of Sindicate that the material has more open-volume defects [8].TheWparameter indicates the fraction of positrons that annihilate with high momentum electrons (core electrons) and thus cause larger Doppler broadening. This parameter is related to the chemical environment where the annihilation takes place. Both parameters can be combined in SWmaps where the different annihilation sites can be distinguished. The data were analyzed with the VEPFIT program [14]. The Sand Wparameters of the Si substrate were used to normalize the data (as these values are identical for all samples, it allows for comparisons to be made between coatings and among annealing tests). Although positron annihilation sites can occur at the surface, in the bulk of the coating, at the coating-substrate interface, and in the silicon substrate, only the coating and the interface values are of interest in this study. These annihilation sites can be distinguished by selecting the positron implantation energy. 2.5. ERD The elastic recoil detection [15] (ERD) measurements were carried out employing 50 MeV Cu 8+ ions produced by the 6 MV EN Tandem van de Graaff accelerator at Utrecht University. These ions can profile all elements from hydrogen to silicon to a depth of a few hundred nanometer in one single measurement. However, in the case of large hydrogen concentrations, hydrogen tends to desorb from the layers under heavy ion irradiation [16]. Therefore, the hydrogen concentration was determined in a separate, short, measurement using a large opening angle. During this measurement the hydrogen content in the film was monitored as a function of ion dose, and this curve is extrapolated to its initial value to determine the hydrogen content. For the hydrogen measurements a solid-state detector at an angle of u=30jwith the beam direction was used. A 29 Am Mylar absorber foil prevents particles other than hydrogen from entering the detector. The incidence angle between the ion beam and the sample surface was set to 20j. Subsequently, the other elements were measured under the same geometry, using a DE-Eionization chamber with a Frisch grid as the particle detector. No Ar was detected in the coatings using energy dispersive X-ray spectroscopy analysis. 3. Results 3.1. Nano-indentation testing and radius of curvature measurements Table 1 summarizes the gas compositions and flow rates that were used to generate the plasma, with the corresponding values of elastic modulus (E), hardness (H), coating thickness (t) and compressive stress of the resultant coatings. The samples were given a code symbolizing the differences in the gas plasmas: R, N2, Ar, ArN2, Ar2N2, H2, H2N2 (where R is the reference plasma consisting of CH 4 gas only). Nano-indentation testing was used to arrive at the Eand Hvalues, and as described in the experimental section, stress in the coatings was determined by the radius of curvature method. As the average particle energy of the impinging ions or neutrals is directly proportional to the term V B /P 1/2 [17], the development of compressive stress is associated with the gas pressure ( P) and the bias voltage applied (V B ). In this study, V B was kept constant and Pwas varied between 7.7 and 12 Pa, however, no relationship between these variables and the stress in the film was observed. Different gas compositions produce coatings with different stress levels, and the only trend in stress reduction was related to the addition of nitrogen. Fig. 1 shows that in all cases, the measured stress in the coatings is lower with nitrogen Table 1 Sample references corresponding with set process variables that produce a-C:H coatings with the given coating thickness (t), Eand Hvalues, and compressive stress Sample Pressure (Pa) Trange (jC) t(nm) E(GPa) H(GPa) Stress (GPa) 100 sccm CH 4 9.6 50–145 582 125.5F2.3 14.3F0.6 1.4 N2 100 sccm CH 4 +20 sccm N 2 12 53–226 441 115.3F2.6 14.4F0.6 1.1 Ar 50 sccm CH 4 +30 sccm Ar 7.8 103–180 345 132.1F2.9 15.5F0.8 1.6 ArN2 50 sccm CH 4 +30 sccm Ar+10 sccm N 2 7.8 105–134 365 121.1F3.2 15.3F1.0 1.3 Ar2N2 50 sccm CH 4 +30 sccm Ar+20 sccm N 2 8.3 108–180 335 109.5F4.3 13.0F0.9 0.91 H2 50 sccm CH 4 +50 sccm H 2 7.7 300–300 182 106.4F5.1 9.7F2.2 1.5 H2N2 50 sccm CH 4 +50 sccm H 2 +20 sccm N 2 8.3 300–300 122 109.8F4.4 9.7F3.0 1.2 addition, with more pronounced effects occurring at 20 sccm flow rate. In the case of depositions made at temperatures lower than 300 jC, there is a concurrent reduction in the elastic modulus with nitrogen addition, but a clear trend does not emerge for a reduction in the hardness values. At high process temperature, i.e. 300 jC, there is little difference between the Eand Hvalues of the nitrogen containing coatings and the control, however, compared to the low temperature depositions these variables are greatly reduced. The stress reduction associated with nitrogen inclusion is also evident for this test. 3.2. ERD The compositional variation of the a-C:H coatings with and without nitrogen inclusion has been determined with ERD and the results are summarized in Table 2. It can be seen that all the samples made with nitrogen in the plasma are doped, including the sample made at a nitrogen flow rate of 10 sccm. The nitrogen incorporation varies approximately between 2 and 3 at.%, and the relationship between nitrogen content and stress reduction is displayed in Fig. 2. There is a 5–8 at.% decrease in hydrogen content for coatings containing nitrogen as compared with the control counterparts; the greatest effect is observed for the depositions at lower temperatures. ERD shows that after annealing to 600 jC, the hydrogen content has decreased by 50–58% of its original value. 3.3. Raman spectroscopy of as deposited coatings The Raman spectra of the as deposited coatings, (not including the high temperature deposition), are shown in Fig. 3. There is a more distinct ‘shoulder’ in the spectra for samples made with 20 sccm nitrogen. These results are summarized in Table 3 for all coatings and include the G and D peak positions, the G and D peak line widths (measured at full width half maximum (FWHM)), I d /I g ratios, and approximate sp 3 content. In amorphous carbons, the development of a D peak indicates ordering, and I d /I g is proportional to the number and clustering of rings [12]. The intensity maximum of the D peak relative to the G peak is directly related to the existence of six-fold aromatic rings, while a broadening of the D peak can be correlated with ring orders other than six [12]. The width of the G peak is proportional to bond-angle disorder at sp 2 sites [12].Fig. 4 is a plot of the G peak linewidth as a function of the measured stress for coatings made without nitrogen, and with nitrogen at a flow rate of 20 sccm in the gas feed. It can be seen from this figure and the data in Table 3 that the line width of the G peak decreases when nitrogen is included at a flow rate of 20 sccm relative to the gas composition without nitrogen. This is indicative of less bond angle disorder at sp 2 sites which in turn means that the system is less constrained and, therefore under less stress. Schwan et al. [18] show that as the intrinsic stress increases Table 2 Compositional variation of the samples as measured with ERD including the nitrogen to carbon ratio (N/C) Sample H (at.%) N (at.%) N/C R 27 0.07 0.001 N2 (20 sccm N 2 ) 19 2.1 0.026 Ar 27 0.18 0.002 ArN2 (10 sccm N 2 ) 21 2.2 0.029 Ar2N2 (20 sccm N 2 ) 22 2.9 0.038 H2 31 0.42 0.006 H2N2 (20 sccm N 2 ) 26 2.9 0.041 H2 after final anneal 12 – – H2N2 (20 sccm N 2 )13 – – After final anneal Fig. 1. Bar graphs showing the reduction in the compressive stress with the addition of nitrogen for different sets of gas compositions used to generate the plasma. Fig. 3. Raman spectra for as deposited coatings with and without nitrogen addition. (a) Raman spectra for coatings R and N 2 showing development of enhanced ‘shoulder’ with the addition of nitrogen, (schematic of the fitted D and G peaks have been included). (b) Samples Ar, ArN2 and Ar2N2 compose the series made with CH 4 /Ar, with 10 and 20 sccm nitrogen inclusion-a broadening of the ‘shoulder’ only occurs with 20 sccm nitrogen addition. Fig. 2. Relationship between the compressive stresses measured for the a-C:H coatings and their nitrogen content. (Unfilled symbols correspond to coatings made with nitrogen in the gas plasma). the G peak line width increases; therefore coatings under lower intrinsic stress have a smaller G peak line width. The variation in the D peak width between coatings made in a plasma containing nitrogen as compared with its corresponding counterpart containing no nitrogen was only significant for samples R and N2, (the differences among the other sets were too close to the error in the peak fit analysis to establish conclusive findings). The width of the D peak is greater for the coating deposited using a methane/ nitrogen (N2) mixture as compared with the pure methane plasma (R), indicating that the coating contains a greater diversity of aromatic ring orders [12]. The data in Table 3 illustrate that with nitrogen addition the G peak shifts to higher wave numbers, and the I d /I g ratio increases. It has been shown using EELS spectroscopy that sp 2 bonding increases as a function of nitrogen content for films of a-C:H [19]. As it was not possible to do EELS spectroscopy in this study, an sp 3 fraction was approximated using the analysis of Robertson and Ferrari [12] based on the G peak position and the I d /I g ratio. They obtained a relationship between sp 3 content and these Raman parameters for as deposited a-C:H, by fitting a line to the experimentally obtained data of sp 2 content and optical gap [12]. In this study, this analysis was further extended to estimating an sp 3 content for the doped coatings since the nitrogen content in the films was too low to establish the presence of a CN phase, and any changes observed in the optical gap were attributed to variation in the sp 2 content. This estimate shows that the sp 3 content is lower for samples made with 20 sccm nitrogen. In addition, the high temperature deposition at 300 jC (H2) also has a lower sp 3 content or higher sp 2 content than control coatings made at lower temperatures. The cluster diameter or in-plane correlation length of aromatic clusters, L a , was calculated based on the following relation: Id=Ig¼CVkðÞL2 að2Þ Where CV(514 nm)=0.0055 (A ˚ 2 ) (Although this equation is used to arrive at L a for coatings studied in this research, the comparison is comparative, i.e. the value for L a is most likely overestimated as suggested by Schwan et al. [18]). The resulting L a values are included in Table 3. The plot of L a against the G peak position and G peak width is illustrated in Fig. 5. It can be seen that in general L a is bigger (e.g. H2N2) and/or that there are more aromatic clusters (e.g. N2) with nitrogen inclusion, and as the value of L a Table 3 G and D peak position, line widths, I d /I g ratios, sp 3 fraction and cluster diameter (L a ) for a-C:H coatings made with and without N 2 Sample Nitrogen flow G peak Line width D peak Line width I d /I g Approximate L a (sccm) (cm 1 ) of G peak (cm 1 ) of D peak sp 3 content (A ˚) (cm 1 ) (cm 1 ) (%) [12] R 0 1552 154 1389 364 1.5 35 16.5 N2 20 1564 131 1407 399 3.0 25 23.3 Ar 0 1554 156 1396 380 1.7 30 17.6 ArN2 10 1553 155 1394 386 1.8 30 18.1 Ar2N2 20 1562 142 1404 382 2.5 25 21.3 H2 0 1566 142 1410 384 2.4 25 20.9 H2N2 20 1574 126 1412 374 3.1 25 23.7 Fig. 4. Changes in G peak line width as a function of stress for coatings made with various plasma compositions with and without nitrogen addition at 20 sccm. (The dotted line was arbitrarily drawn to show the demarcation in stress level between coatings made with nitrogen and those made without nitrogen). increases the G peak position moves to higher wave numbers and the width decreases. The latter observation, i.e. decrease in width of the G peak, has been linked with stress reduction [18]. As noted before from the variation in the D peak width, sample N2 has a larger diversity of aromatic ring orders so the larger value of L a in this case is attributed to an increase in number of clusters. The larger value of L a for H2N2 corresponds with an increase in the six-membered aromatic cluster size as the D peak width has not increased and the G peak has shifted to higher wave numbers. 3.4. Annealing study: using Raman spectroscopy and positron beam analysis 3.4.1. Raman spectroscopy Two samples were made at the regulated and elevated temperature of 300 jC, to isolate the effect of nitrogen on stress development at high temperatures. That is, to observe if a high temperature deposition, which may be conducive to the generation of a higher thermal stress component, is influenced by the presence of nitrogen in the gas plasma. As seen in Table 1, the stress is reduced with nitrogen inclusion and the Raman spectra show that although both samples have a ‘shoulder’, the sample manufactured with 20 sccm nitrogen has a more pronounced ‘shoulder’. Fig. 6 displays the Raman spectra of the as deposited coatings and their final evolution after annealing to 600 jC. As mentioned before, the annealing experiments were performed to identify open volume in the samples. The evolution of the G and D peak positions, and the changes in their line widths, as a function of annealing temperature is listed in Table 4. The graphitization process is underway at 300 jC for the nitrogen-containing sample, H2N2, as the G peak position is at 1580 cm 1 . However, for the coating made without nitrogen, H2, 300 jC marks the start of the process as seen by the movement of the G peak to higher wave numbers. This behavior is linked with the fact that the as deposited coating with nitrogen in the gas plasma has a G peak that is shifted to higher wave numbers, 1574 cm 1 as compared Fig. 5. Changes in the cluster size with nitrogen inclusion: (a) Shows variation of cluster size with G peak position for samples with and without nitrogen. (b) Illustrates that as the cluster size increases, the G peak width decreases. (Unfilled symbols represent samples made with nitrogen in the plasma.) with the G peak of the methane/hydrogen mixture which is at 1566 cm 1 . The transformation to a graphitic structure with annealing has been linked to the release of bounded hydrogen between 400 and 600 jC[20]. In this temperature range sp 3 carbon bonds are transformed to sp 2 bonds [21]. As seen before, the width of the G peak is less for the nitrogen containing samples—therefore, less bond angle disorder at sp 2 sites. A general trend of decreasing bond angle disorder with increasing annealing temperature is recorded for both samples. There is a shift of the D peak position with annealing temperature to lower wavelengths. The wavelengths 1353 and 1598 cm 1 are associated with small crystallites of graphite, and in the proximity of 1353 cm 1 it can be concluded that the crystallites formed are threefold coordinated [22]. A fine-grained polycrystalline structure emerges as the coating becomes more graphitic in nature, i.e. the clusters will lose their molecular like properties [19]. The changes in the D peak width with annealing temperature shows that ordered clusters of six-membered rings (crystallites) dominate as the sample is heated. These transformations can also be seen in the PBA data. 3.4.2. PBA The positron beam analysis data (Table 5) show that the open volume, as seen in the value of the Sparameter is greater at the interface for both coatings as compared with Table 4 Shift of the G and D peaks and changes in peak widths as a function of annealing temperature TG peak position G peak line width D peak position D peak line width (jC) (cm 1 ) (cm 1 ) (cm 1 ) (cm 1 ) H2N2 H2 H2N2 H2 H2N2 H2 H2N2 H2 As deposited 1574 1566 126 142 1412 1410 374 384 150 1573 1565 123 142 1419 1409 406 381 300 1580 1571 115 131 1411 1403 365 375 400 1580 1576 109 120 1401 1405 352 366 500 1587 1590 103 103 1372 1375 333 310 600 1585 1585 107 101 1355 1349 316 314 Fig. 6. Raman spectra for the high temperature depositions: (a) As deposited a-C:H films (H2N2 contains nitrogen); (b) Shows the final evolution of the Raman spectra after successive annealing procedures to 600 jC. the bulk. The Wparameter is also significantly different at the interface suggesting the existence of a mixed layer, which is created during the sub-plantation of the ions during deposition. The changes in the Wparameter within the coating show that the chemical environment is changing with annealing. Both the Sand Wparameters have altered after the final anneal: the open volume has decreased, and the value of the Wparameter has increased. The change in the Wparameter of the coating for H2, (the sample made without nitrogen), is significantly greater than H2N2. At 600 jC PBA shows that there is ‘delamination’ of both coatings as the Sand Wparameters take on the value of the surface. Fig. 7 shows the changes in the Sand Wparameters with annealing, and it becomes obvious that there are vacancy defects at the interface (Fig. 7a) of the nitrogencontaining sample (H2N2) since the Sparameter is high and the Wparameter low [23]. During the annealing procedure to 400 jC, the Sparameter increases. This rise in the value of the Sparameter can be seen more clearly in Fig. 8a, coincides with removal of physisorbed hydrogen [8]. However, at approximately 500 jC the gas released during the annealing is not only the release of ‘unbound’ hydrogen, but also the H bounded at tetrahedral sites to carbon, then there is a need for internal restructuring of the ‘lattice’, hence the decrease in the value of Sat the interface. The chemical environment within the coatings, and at their interface, change with temperature as can be noted in the variation of the Wparameter (Fig. 8b). Also evident in the plot of Fig. 8a is that the Sparameter of the nitrogen containing coating, H2N2, is generally larger than that of the reference sample indicating that overall this coating has more open volume which is characteristic of a doped layer. Fig. 7. S–Wmaps for the coatings made at high temperature, H2N2 and H2 showing the effects of annealing: (a) interface (b) coating. (Note: Fig. 7a and b are not to the same scale). Table 5 VEPFIT results of samples studied at room temperature and after annealing to 600 jC(S Si =1, W Si =1). (The errors are DSf0.002 and DWf0.03) Sample Before annealing After annealing S coat W coat S interf W interf S coat W coat S delam W delam H2N2 0.932 1.40 0.964 1.33 0.929 1.48 0.906 1.68 H2 0.931 1.30 0.946 1.43 0.919 1.54 0.907 1.67