FOX12 Calibration Procedure

t t t

R D =

= . (4-1)

Using thicker material layers, longer mill times and averaging on the Dektak measurement data improves accuracy. Note that the total mill time is somehow limited due to resist cross-linking that may occur for longer mill times. When the resist gets ‘burned’ it may become difficult to remove and the Dektak readings become unreliable.

4.3.2. FOX12 Calibration Procedure

The standard calibration technique described in the previous section is not applicable to the determination of the ion mill etch rate of cured FOX12. Since ion milling is in principle a non-selective etch process (the energetic plasma ions sputter away whatever material they impinge on), a uniform layer of material is required to perform an ion mill rate calibration. Because FOX12 resist hardens into a SiO2-like hard mask upon e-beam exposure, in order to obtain a uniform layer of cured FOX12, the entire substrate should be scanned with the focused e-beam spot, which is not feasible due to the slow scanning nature of the e-beam exposure. Therefore, an alternative (indirect) approach is proposed.

Figure 4-9: The strips (A) and triangles (B) from which the step heights are obtained using Dektak surface profilometry.

A relatively thick uniform material layer (e.g. Au) is deposited on a substrate, which is then spin-coated with a FOX12 resist layer. Using e-beam lithography, a design containing rectangular strips of 50x100 μm² and triangles of comparable size (see Figure 4-9 A and B) is transferred onto the substrate in the form of a cured FOX12 hard mask.

After exposure and development, the initial thickness F0 of the cured FOX12 hard mask is obtained by Dektak surface profilometry, performed on the strips and triangles. Subsequently, the substrate is subjected to a series of ion mill steps.

Between every mill step, the substrate is taken out of the ion mill system and the height difference between the top of the cured FOX12 layer and the substrate Au level, D(t) (see Figure 4-10), is recorded using Dektak surface profilometry.

Figure 4-10: Cross-sectional view of a cured FOX12 hard mask feature on top of a uniform Au layer and the effect of an ion-mill step.

From Figure 4-10, it becomes clear how the FOX12 etch rate can be deduced from the step heights D(t) recorded as a function of time. During every subsequent ion mill step, both cured FOX12 and Au are etched. The initial FOX12 thickness F0 will thus be reduced by an amount RFOX12·t, with RFOX12 the etch rate of the cured FOX12 and t the ion mill time. At the same time, an amount RAu·t is milled away in the Au layer, where RAu is the mill rate of Au, which can be obtained with the standard calibration procedure described in Section 4.3.1. The resulting step height can now be written as

t Since D(t) is provided by the Dektak measurements and RAu·t can be calculated from the standard Au mill rate calibration, the left hand side of this equation is directly obtainable from a measurement of the step heights as a function of time. The obtained values should follow the linear trend given by the right hand side (where F0 is also

Ti (10) Cross Section

Au (180)

FOX₁₂ D(t)

F₀

known from the initial Dektak measurement). The FOX12 etch rate, RFOX12, can then be determined as the negative slope of the linear function in the right hand side.

Now, the actual value of the FOX12 ion mill etch rate is determined. The substrate used is the full spin valve stack described in Section 4.2.1, but with a relatively thick uniform Ti(10.0)/Au(180.0) top layer. The thickness of the FOX12 resist after exposure and development is determined by an initial Dektak measurement, yielding an average value of F0 = 75 nm. Using the standard calibration procedure, the ion mill rate of Au is determined to be 35 nm/min. The substrate is then subjected to seven one-minute ion mill steps. After each step, the average step height is recorded. The left hand side of equation (4-3) can then be calculated and the result is plotted in Figure 4-11.

Figure 4-11: The left hand side of equation (4-3) is plotted as a function of mill time. A linear fit generates the mill rate for cured FOX12. After 5.1 minutes (t0), the slope abruptly changes, indicating that the 180 nm Au layer has been milled away completely. Thereafter, the slope is related to the mill rate of the Ti layer under the Au.

From this figure, the ion mill etch rate of cured FOX12 is determined as the negative slope of a linear fit through the first 5 measurement points. This yields an average 15 nm/min for the strips and triangles. Interestingly, after 5.1 minutes, the slope abruptly changes to an increased negative value. This behavior is attributed to the fact that either the FOX12 has been removed completely or that an interface to another material under the 180 nm Au layer has been reached. Note that in the first case, provided the Au layer is thick enough so that ion milling continues into an uniform Au layer, at the intercept with the x-axis (t = t0), the curve would resume a slope given

A.

t0

by the mill rate of Au. This can be seen from equation (4-3), since setting the left side

The first statement indicates that at the crossing point, the step height is exactly the amount of Au that has been milled away up to that time, which is to be expected, since at that precise time, all FOX12 has been removed. The second statement is the analogue of equation (4-1), since in the time t0 the entire FOX12 layer is exactly removed. From that moment on, the patterned structure and the surrounding layer get ion milled at the same rate and the step height remains constant. This implies that the left hand side of equation (4-3) will indeed further scale linearly with time, with a slope given by the mill rate of Au (replace FOX12 by Au in equation (4-3)).

However, the slope observed in Figure 4-11 after t = t0 (54 nm/min) clearly does not correspond to the mill rate of Au which was determined to be only 35 nm/min.

Therefore, the second explanation for the abrupt slope change, the encounter of a material interface, is thought to be applicable here. Indeed, based on the RAu = 35 nm/min mill rate, t0 = 5.1 minutes corresponds exactly to the time necessary to mill away the entire 180 nm Au layer, after which ion milling continues into the underlying Ti(10.0) layer.

In this case, the procedure outlined above can be repeated, with the understanding that the roles of FOX12 and Au are now fulfilled by Au and Ti. The governing equation becomes

with the prime indicating a time after t = t0. Rearranging this expression gives (keep in mind that now the mill rate of the top instead of the bottom layer is known, leading to the plus sign)

Casting this into a similar form as equation (4-3), so that it can be applied to Figure 4-11, yields

(

)

' ' 2 ' '

' )

' (

' t R t F₀^' R t R t F₀^' R R t

D − _Au⋅ = + _Ti⋅ − _Au⋅ = + _Ti − _Au ⋅ . (4-8) The slope is now given by RTi - 2RAu. Note that, since the time t0 (coincidentally) also corresponds to the time necessary to mill away the entire FOX12 layer, only the last two points are considered in the linear fit, leading to a value of 10.4 nm/min for the ion mill rate of Ti. The value obtained here is expected to be only approximately accurate, since only two points were considered for the linear fit. Moreover, since the Ti layer is only 10 nm thick, the last point may have encountered another transition to the thin film layer under Ti. A regular calibration of the ion mill rate for Ti resulted in a value of 7.8 nm/min.