Differential method for layer-to-layer registration

[0001] The invention related to the precisely alignment between layers. The application which exist in semiconductor wafer manufacturing layer to layer alignment, or any other place need precise alignment.

[0002] When need precisely to align one object to another object, usual way is to put align mark to each object, then to measure the shift between these two marks to decide the shiftiness of the two object, then to align them. Unfortunately, when the precision requirement reaches nm level, it could not be accurately measured optically because optical wavelets limitation. It can use light diffraction such as applying grating structure, but because the line thickness also affect the light diffraction, to use grating diffracting to measure layer to layer shiftiness still has problems.

[0003] The invention is to use different pitch grating for two layers. By compare left and right reflected/diffracted or any combined light signal, to decide the mis-aligned information. By this method, all unrelated things such as grating thickness or line width signals will be canceled each other, and only the shiftiness signal will be enlarged dozens or hundreds times, even the stiffness is sub-nm, the final signal from this invention will be hundred nm large, and can be actually measured optically.

[0008] Referring now to the drawing, and in particular to FIG. 1, grating 100 has a pitch n, for example, 50 um, grating 101 has a pitch n+1, in this example is 51 um. 102 are overlapped grating.

[0009] Referring now to FIG. 2, a system catch the reflecting/diffracting/interference light from the overlapped gratings, to compare left and right deference.

[0010] Let d₁ be line pitch of previous layer mark, d₂ be line pitch of current layer mark, c is a constant, Δs be actual shift between layers, the overlapped grating image intensity meets the following equation

[0011] $\frac{x_p}{d_1}\left(d_2-d_1\right)-\Delta s=\frac{x_0}{d_1}\left(d_2-d_1\right)$

[0012] The FIG. 2 system 103 will see the shift Δx=x_p−x₀:

[0013] $\Delta x=x_p-x_0=\left(\frac{d_1}{d_2-d_1}\right)\Delta s$

[0014] This means that the overlaid image distribution shift Δx (compare left and right) is d₁/(d₂−d₁) times large then actual lay shift Δs. Use d₁=0.5 um, d₂=0.51 um value here, the enlargement here is 50 times, as shown in FIG. 3. The picture shows a 10 nm shift between two layers, the microscopy observed 500 nm shiftiness, which is very easy to measure.