ION AND PLASMA BEAMS IN OPTICAL TECHNOLOGIES

Sputtering Parameters of Optical Materials

A target's sputtering rate and sputtering yield are the main parameters for ion beam milling. Nevertheless the published data for optical materials are non numerous, uncoordinated and hard comparable because of different experimental conditions. Besides the most of theories of sputtering describe a process in the approximation of monatomic target, but even for monoatomic targets a difference between theoretical and experimental values of sputtering yield (rate) may be big enough (2 - 3 times). A successful analysis of multicomponent materials sputtering (glasses, for instance) is much more problematic.Therefore the best way for the practical goals is to obtain the sputtering parameters in experiments with comparable conditions. I run these investigations for a number of optical glasses, fused silica, germanium (optical grade) and zinc selenide (optical grade). I used the following Russian optical glasses: LK5 (analog Pyrex), K8 (analog BK7), TK16 (analog SK4), STK119 (analog LAK28), TF3 (analog SF1), and BF112 (analog BASF1).IMPORTANT NOTE: These Russian and western glasses have the same optical parameters but I'm not sure they have the same composition, structure, and physical parameters defining their behavior under the ion bombardment!

The samples' surfaces to be sputtered have been optically polished to R_z less than 0.05 micron. The ion beam source "Arctur" has been used in these experiments. It was installed inside the 0.3 m³ vacuum chamber with turbomolecular pumping system (pumping speed about 1000 l/s for N₂). The ultimate pressure was less than 10^-3 Pa.

I used Ar, N₂ and mix Ar + O₂ (3 : 1) gases. The working pressure was usually in the range (2 ... 6)x10^-2 Pa. The ion energy was about 180 eV for all the gases. I sputtered in the same time a few target surfaces, one of which was normal to the incident ion flux, and others were located at different angles to the ion flux (Fig.1, patented). The part of each surface S_i has been covered with a film mask M so only uncovered areas of surfaces S_i in the neighborhood of common for each couple of surfaces S₀ - S_i points O_i, were sputtered.

Fig.1. Sputtering rate measurements.

In general, for a given material and a given ions with a given energy the sputtering rate is described as follows:

V = kS(a,E)jcosa;

where a - incident angle, S(a ,E) - sputtering yield for a given ions and their energy, j - ion current density, k - proportional rate. Therefore the sputtered depth h = Vt, where t is a sputtering time. Supposing that in the neighborhood of point O_i a ion current density j = Const, we can define the sputtered depth of S₀ surface during the time t as:

h₀ = kS(a = 0)jt,

and the sputtered depth of S_i surface as:

h_i = kS(a_i)jtcos a_i.

Now we can define the relative sputtering rate V_rel(a_i) and sputtering yield S_rel(a_i) as;

V_rel(a_i) = h_i/h₀

S_rel(a_i) = h_i/h₀cosa_i.

The sputtered depth was measured using micro interferometer by the shift of interference stripes on the step between masked and unmasked areas of sputtered surface. These steps were located at the distance about 1.5 - 2 mm from point O_i. Therefore we may suppose that the current density and ion energy are the same at so close located points. It has been shown that for all investigated glasses the sputtering rate is increasing of 1.5 - 2 times with the incident angle, and reaches maximum values at a = 40^o - 50^o. The germanium sputtering rate is increasing too, but not so much (about 1.1 times); ZnSe sputtering rate stays constant up to a = 40^o and decreases at incident angles more than 40^o.The sputtering yields show the same but more strong dependencies due to the presence of cosa_i in denominator. So the glasses sputtering yields are increasing of 3 - 4.5 times with the incident angle and reach maximum values at a = 55^o - 65^o. The germanium sputtering yield is increasing about 2.3 times and reaches this maximum at a = 65^o, the ZnSe sputtering yield is increasing about 1.5 times at the same maximal angle. Typical sputtering yield angular dependencies for the different gases are shown on Fig.2 (glass TF3). For all other materials the curves are the same (I didn't sputter Ge and ZnSe with ions of Ar - O₂ mix). The divergence of these three curves actually is in the error of measurements limits and we can't claim that there are different S(a) dependencies for the different gases. Anyway all curves for all the investigated materials are typical for the case of polycrystalline and amorphous materials sputtered with the ions of average and low energies. The same curve for monocrystaline germanium might be explained by the well known fact that even for low energy ions the monocrystaline surface becomes amorphous if the irradiation dose is big enough (in my experiments about 10²⁰ cm^-2.

Fig.2. The TF3 glass sputtering yield S angular dependence for three different ions.

For the all investigated materials the angles a_max are in the range of 55^o - 70^o when S(a) reaches its maximum values. We can see some correlation between a_max and material density (Fig.3), but I couldn't find out any correlation with others known macro parameters of materials describing their hardness, elasticity, heat capacity, heat conductivity, and so on.

Fig.3. Correlation between the material bulk density and the angle a_max. 1 - LK5, 2 - Fused silica, 3 - K8, 4 - TK16, 5 - BF112, 6 - STK119, 7 - TF3, 8 - ZnSe, 9 - Ge.

Application the following well known Lindhard formula for a_max calculation doesn't look correct in case of multicomponents materials.

Here a = 0.53 A - the first Bore's radius of hydrogen atom; N - atomic density of target material, A^-3; z_m and z_i- atomic numbers of target material and incident ions accordingly; E_R= 13.6 eV (first ionization potential of hydrogen), E_i - ion's energy. First, what atomic numbers must we use? For the simple oxides like SiO₂ we may (with some limitations) use the average atomic number, but for really multicomponent material it is not obviously. Second, we can't exclude the selective sputtering when the target surface layer changes its chemical composition due to the preferable sputtering and losses of one or a few components. It is known at the same time that for high enough ion energies (more than 350 eV) calculation a_maxusing Lindhard formula gives results that are in good correlation with experimental data. Third, for low energies the Lindhard formula shows a strong a_max decreasing (Fig.4), for instance, at the argon ions energy 180 eV the value a_max for SiO₂ should be 38^o, and for Ge - 25^o. However we have got in experiment a_max = 57.5^o for SiO₂ and a_max = 67^o for Ge. So, it is a question.... To my regret, I couldn't find the original Lindhard's article (and it seems to me it is in Dutch....) to check up his approximations and initial conditions. Probably, his formula is working well only for high energy ions and simplest materials....

Fig.4. The energy dependence of a_max for Ge(1) and SiO₂ (2) accordingly to Lindhard formula.

In the Table you can see the relative sputtering rates of investigated materials for the normal incident angle of ions (the K8 sputtering rate is taken as 1):

Usage of Ar - O₂ mix decreases the sputtering rates approximately at 30%, N₂- in two times comparing with Ar.

On the Fig.5 you can see the sputtering yields angular dependencies normalized to K8 sputtering yield. Here are: fused silica (1), crowns (2 - 5), flints (6,7) and crystals (Ge - 8, ZnSe - 9). You can see that inside the given type of glasses (crowns and flints) the sputtering yield curves are close enough.

Fig.5. The relative sputtering yield angular dependencies for fused silica (1), LK5 (2), K8 (3), TK16 (4), STK119 (5), BF112 (6), TF3 (7), Ge (8), and ZnSe (9).

Obtained results show that lead containing glasses have the larger sputtering rates. This fact has been noted and in other investigations, but we can't say that the sputtering rate directly depends on PbO concentration. The PbO concentration in BF112 is of 5 times less than in TF3, however the difference between sputtering rates is just a few percents. There is an opinion that a sputtering rates of glasses increase with the SiO₂ concentration decreasing. But again - the SiO₂ concentration in STK119 is of 20 times less than in LK5, practically equal in BF112, TK16, and TF3 glasses, and it doesn't correlate with sputtering rates of these glasses. At the same time it has been shown in other investigations that the doping of quartz even with a few percents of Na₂O may change a sputtering rate to a considerable extent.

Therefore we can make the following conclusion:

• The optical glasses, ZnSe, and monocrystaline Ge are sputtered by low energy ions like polycrystalline and amorphous materials.
• The glasses of the same type (crowns or flints) have close sputtering parameters.
• There are no close correlation between sputtering parameters and physical properties of a bulk material.
• It looks like to obtain a trustworthy information about multicomponent material sputtering we need to run combined experiments including investigation of sputtered particles, surface layer composition and concentration profiles in a surface layer, stechiometry and structure of a surface layer, processes of diffusion and segregation, and so on.

Not too much....

Nevertheless because of a strong sputtering yield angular dependence, big differences between the sputtering rates of optical materials we must know these parameters for each material we are going to work with. Especially it is important for high precise ion beam milling technology. I could get repeatable and predictable results in a good agreement with calculated data only using carefully defined sputtering yields angular dependencies for the material of part to be milled.