CN100535755C - Adaptive optimization method for coaxial-aligning vertical-scanning length - Google Patents

Abstract

The invention discloses an adaptive optimum method of coaxial aligning vertical length scanning, which comprises the following steps: (a) affirming the maximum root-mean-square value and minimum root-mean-square value at the vertical direction; (b) proceeding the aligning scanning to obtain the sampling point number of vertical direction; (c) calculating the image maximum value and minimum value according to the data gathered by the sensor; (d) updating the maximum and minimum root-mean-square value of image vertical direction then returning to the step (b) when the image maximum and minimum root-mean-square value is within the normal value scale; repeating the aligning scanning; (e) returning the step (b) directly when the image maximum and minimum root-mean-square value is not within the normal value scale; repeating the aligning scanning.

Description

The adaptive optimization method of the vertical sweep length of coaxial alignment

Technical field

The present invention relates to a kind of technique of alignment disposal route in the SIC (semiconductor integrated circuit) photoetching production equipment, the adaptive optimization method of the vertical sweep length of particularly a kind of coaxial alignment.

Background technology

Adopting litho machine to carry out in the IC Chip Production process, in order to realize litho machine expected accuracy index, need accurately set up the relation between each coordinate system of litho machine, mask, mask platform, object lens, silicon chip, silicon chip platform can be worked in a unified scaling system.Aim at is exactly to find the definite position that is marked at lens below imaging on mask or the mask platform by the alignment mark on the silicon chip platform.Use sensor with vertical direction light intensity to be carried out scanning sample in the horizontal direction,, find out largest light intensity point position by light intensity data is carried out The Fitting Calculation.The light intensity maximum position of the imaging that finds by aligning just is called aligned position.

When sampling, need abundant sampled point so that provide enough information for The Fitting Calculation, but because the increase of sampled point, may bring information superfluous in, physics realization simultaneously need be spent the longer time, it should be noted that especially increases a sampled point in vertical direction, and physics realization take time increases a sampled point and Yan Gengchang with respect to horizontal direction.In order to boost productivity, the calculating of vertical direction sampled point should be tried one's best accurately, except the enough information of trying one's best is provided to The Fitting Calculation, does not spend oversize sweep time simultaneously, promptly think this moment the sampled point optimum.

Summary of the invention

The object of the present invention is to provide the adaptive optimization method of the vertical sweep length of a kind of coaxial alignment, make the sampling number that calculates to approach standard sample fast and count.

The present invention is achieved by the following technical solutions: the adaptive optimization method of the vertical sweep length of a kind of coaxial alignment comprises the steps: that (a) determines maximum effective value of vertical direction image and minimum effective value; (b) carry out alignment scanning, obtain the vertical direction sampling number; (c) after alignment scanning finishes, calculate, determine the maximal value and the minimum value of measuring image according to the data that sensor sample obtains; (d) when measuring image maximal value and minimum value all in range of normal value, then return step (b) behind maximum effective value of update image vertical direction image and the minimum effective value, carry out alignment scanning again; (e) when measuring image maximal value and minimum value not in range of normal value, then directly return step (b), carry out alignment scanning again.

Wherein, maximum effective value of the vertical direction image in the step (a) and minimum effective value obtain by experimental formula.

Vertical direction sampling number in the step (b) is to pass through formula

Nr_of_vert_samples=N* (eff_max+const)/eff_min determines, wherein, Nr_of_vert_samples is the vertical direction sampling number, and N is that vertical minimum sampled point coefficient, const are that capture range, eff_max are that the maximum effective value of vertical direction image, eff_min are the minimum effective values of vertical direction image.

The data that sensor sample obtains in the step (c) comprise position data x and light intensity data I, and the relation between them is I=β ₁X ²+ β ₂X+ β ₃Largest light intensity 50% pairing picture traverse is imaging size, i.e. I _Max/ 2=β ₁X ²+ β ₂X+ β ₃, I wherein _MaxIt is largest light intensity.

The value of determining measuring image is to pass through formula $\mathrm{aiz}\_\mathrm{meas}=|x_1-x_2|=\left|\frac{\sqrt{b^2-4\·a\·c}}{a}\right|$ Determine that wherein, aiz_meas is the measuring image value, a=β ₁, b=β ₂, c=β ₃-I _Max/ 2, maximal value max (aiz_meas) that gets measuring image respectively and minimum value min (aiz_meas).

Upgrading maximum effective value of vertical direction image and minimum effective value in the step (d) is by formula eff_max=η max (aiz_meas)+(1-η) eff_max _Old, eff_min=η min (aiz_meas)+(1-η) eff_min _OldDetermine that wherein, eff_max is the maximum effective value of vertical direction image, eff_min is the minimum effective value of vertical direction image, and η is the index forgetting factor, and max (aiz_meas) is the maximal value of measuring image, min (aiz_meas) is the minimum value of measuring image, eff_max _OldAnd eff_min _OldIt is respectively the vertical direction image minimum effective value before upgrading the maximum effective value of preceding vertical direction image and upgrading.

Described index forgetting factor η passes through formula

$\mathrm{\η}=\min (1,C_{\mathrm{\η}}+\min (K*C_{\mathrm{\η}}*\frac{\max (\mathrm{aiz}\_\mathrm{meas})-\min (\mathrm{aiz}\_\mathrm{meas})}{\max (\mathrm{aiz}\_\mathrm{meas})+\min (\mathrm{aiz}\_\mathrm{meas})},\max ((L-C_{\mathrm{\η}}),0)))$ Determine, wherein, C _ηFor minimum is upgraded the factor, the maximum factor of upgrading of L, K is a coefficient of balance, and max (aiz_meas) is the maximal value of measuring image, and min (aiz_meas) is the minimum value of measuring image.

The adaptive optimization method of the vertical sweep length of a kind of coaxial alignment of the present invention, by adaptive method, make the vertical direction sampling number that calculates to approach standard sample fast and count, guaranteed the computational accuracy of aligned position, improved production efficiency greatly.

Description of drawings

Fig. 1 is the process flow diagram of the adaptive optimization method of the vertical sweep length of coaxial alignment of the present invention.

Embodiment

See also Fig. 1, the step of the adaptive optimization method of the vertical sweep length of a kind of coaxial alignment of the present invention is as follows:

At first execution in step 101, determine maximum effective value of vertical direction image and minimum effective value.Maximum effective value eff_max of vertical direction image and the minimum effective value eff_min of vertical direction image calculate by experimental formula when using for the first time.

Secondly, execution in step 102 is carried out alignment scanning, obtains the vertical direction sampling number.When alignment scanning, the vertical direction sampling number calculates by formula 1,

Nr_of_vert_samples=N* (eff_max+const)/eff_min (formula 1)

Wherein, the vertical minimum sampled point coefficient N of parameter, capture range const are the empirical value of statistics acquisition.

It has been generally acknowledged that under current system state, exist a standard sample to count, satisfy above-mentioned situation, both provided the enough information of trying one's best, can not spend oversize sweep time simultaneously to The Fitting Calculation.Because there are various errors in system, cause the sampling number that calculates excessive or too small for standard sample is counted, therefore just need to make the sampling number that calculates approach standard sample fast and count by a kind of adaptive algorithm.

Secondly, execution in step 103 after alignment scanning finishes, is carried out The Fitting Calculation by position data x and the light intensity data I that certain sensor sample is obtained, except obtaining aligned position, can also obtain the size of the pairing alignment mark of this sensor imaging below lens.Through later position data and the light intensity data of The Fitting Calculation following relation is arranged, as shown in Equation 2.

I=β ₁X ²+ β ₂X+ β ₃(formula 2)

Wherein, x is a position data, and I is a light intensity data, calculates largest light intensity I by formula 2 _MaxIt is generally acknowledged that largest light intensity 50% pairing picture traverse is the picture size, promptly formula 3.

I _Max/ 2=β ₁X ²+ β ₂X+ β ₃(formula 3)

The absolute value of two difference in the formula 3 is exactly the size of measuring image, as shown in Equation 4,

$\mathrm{aiz}\_\mathrm{meas}=|x_1-x_2|=\left|\frac{\sqrt{b^2-4\·a\·c}}{a}\right|$ (formula 4)

Wherein, a=β ₁, b=β ₂, c=β ₃-I _Max/ 2

Sampled data by a mark can The Fitting Calculation goes out the size of the pairing alignment mark of this sensor imaging below lens, be the value aiz_meas of measuring image, a plurality of marks just exist maximum value and minimal value, might as well be called measuring image maximal value max (aiz_meas) and measuring image minimum value min (aiz_meas).

Execution in step 104 then, judge that measuring image maximal value max (aiz_meas) and minimum value min (aiz_meas) are whether in range of normal value.

If measuring image maximal value max (aiz_meas) and measuring image minimum value min (aiz_meas) be not in range of normal value, can think the failure of this alignment scanning, not maximum effective value eff_max of update image vertical direction image and minimum effective value eff_min, directly return step 102, carry out alignment scanning once more, recomputate the vertical direction sampling number.

If all in range of normal value, can think needs maximum effective value eff_max of update image vertical direction image and minimum effective value eff_min for measuring image maximal value max (aiz_meas) and measuring image minimum value min (aiz_meas).

And maximum effective value eff_max of renewal vertical direction image and the minimum effective value eff_min of vertical direction image need determine the index forgetting factor, and promptly execution in step 105.Parameter η is the index forgetting factor, promptly upgrades the factor, and the speed of convergence that the factor is directly connected to the vertical direction sampling number is upgraded in 0＜η＜1, and it calculates as shown in Equation 5:

$\mathrm{\η}=\min (1,C_{\mathrm{\η}}+\min (K*C_{\mathrm{\η}}*\frac{\max (\mathrm{aiz}\_\mathrm{meas})-\min (\mathrm{aiz}\_\mathrm{meas})}{\max (\mathrm{aiz}\_\mathrm{meas})+\min (\mathrm{aiz}\_\mathrm{meas})},\max ((L-C_{\mathrm{\η}}),0)))$

(formula 5)

Wherein, constant C _ηFor minimum is upgraded the factor, the maximum factor of upgrading of L, K is a coefficient of balance, is the system statistics constant, can suitably adjust according to statistical conditions, and L＞C _η, C _η＜η＜L.

Execution in step 106 then, upgrade maximum effective value eff_max of vertical direction image and the minimum effective value eff_min of vertical direction image according to the index forgetting factor that obtains.Can upgrade by maximum effective value eff_max of 6 pairs of vertical direction images of formula and the minimum effective value eff_min of vertical direction image.

eff_max＝η·max(aiz_meas)+(1-η)·eff_max _old，

Eff_min=η min (aiz_meas)+(1-η) eff_min _Old(formula 6)

Wherein, eff_max _OldAnd eff_min _OldIt is respectively the vertical direction image minimum effective value before upgrading the maximum effective value of preceding vertical direction image and upgrading.

Then return step 102, carry out alignment scanning once more, recomputate the vertical direction sampling number.

The present invention mainly upgrades maximum effective value eff_max of vertical direction image and the minimum effective value eff_min of vertical direction image according to the difference size between measuring image maximal value max (aiz_meas) and the measuring image minimum value min (aiz_meas), thereby realizes the purpose that vertical direction sampling number self-adaptation is adjusted.After successful several times alignment scanning, the vertical direction sampling number also can be through self-adaptation adjustment several times, and finally can converge to a suitable value.

The adaptive optimization method of the vertical sweep length of a kind of coaxial alignment of the present invention, by adaptive method, make the vertical direction sampling number that calculates to approach standard sample fast and count, guaranteed the computational accuracy of aligned position, improved production efficiency greatly.

Claims (8)

1, the adaptive optimization method of the vertical sweep length of a kind of coaxial alignment is characterized in that comprising the steps:

(a) determine maximum effective value of vertical direction image and minimum effective value;

(b) carry out alignment scanning, obtain the vertical direction sampling number;

(c) after alignment scanning finishes, calculate, determine the maximal value and the minimum value of measuring image according to the data that sensor sample obtains;

(d) when measuring image maximal value and minimum value all in range of normal value, then return step (b) behind maximum effective value of update image vertical direction image and the minimum effective value, carry out alignment scanning again;

(e) when measuring image maximal value and minimum value not in range of normal value, then directly return step (b), carry out alignment scanning again.

2, the adaptive optimization method of the vertical sweep length of coaxial alignment as claimed in claim 1 is characterized in that: maximum effective value of the vertical direction image in the step (a) and minimum effective value obtain by experimental formula.

3, the adaptive optimization method of the vertical sweep length of coaxial alignment as claimed in claim 1 is characterized in that: the vertical direction sampling number in the step (b) is to pass through formula

Nr_of_vert_samples=N* (eff_max+const)/eff_min determines, wherein, Nr_of_vert_samples is the vertical direction sampling number, and N is that vertical minimum sampled point coefficient, const are that capture range, eff_max are that the maximum effective value of vertical direction image, eff_min are the minimum effective values of vertical direction image.

4, the adaptive optimization method of the vertical sweep length of coaxial alignment as claimed in claim 1 is characterized in that: the data that sensor sample obtains in the step (c) comprise position data x and light intensity data I, and the relation between them is I=β ₁X ²+ β ₂X+ β ₃, wherein, β ₁, β ₂, β ₃Be the fitting coefficient of I about the quadratic equation of x.

5, the adaptive optimization method of the vertical sweep length of coaxial alignment as claimed in claim 4 is characterized in that: largest light intensity 50% pairing picture traverse is imaging size, i.e. I _Max/ 2=β ₁X ²+ β ₂X+ β ₃, wherein, I _MaxBe largest light intensity, β ₁, β ₂, β ₃Be the fitting coefficient of I about the quadratic equation of x.

6, the adaptive optimization method of the vertical sweep length of coaxial alignment as claimed in claim 5, it is characterized in that: the value of determining measuring image is to pass through formula $\mathrm{aiz}\_\mathrm{meas}=|x_1-x_2|=\left|\frac{\sqrt{b^2-4\·a\·c}}{a}\right|$ Determine that wherein, aiz_meas is the measuring image value, a=β ₁, b=β ₂, c=β ₃-I _Max/ 2, maximal value max (aiz_meas) that gets measuring image respectively and minimum value min (aiz_meas), wherein, x ₁, x ₂Be equation I=β ₁X ²+ β ₂X+ β ₃Two roots.

7, the adaptive optimization method of the vertical sweep length of coaxial alignment as claimed in claim 1 is characterized in that: upgrade maximum effective value of vertical direction image and minimum effective value in the step (d) and be by

eff_max＝η·max(aiz_meas)+(1-η)·eff_max _old，

Formula eff_min=η min (aiz_meas)+(1-η) eff_min _OldDetermine that wherein, eff_max is the maximum effective value of vertical direction image, eff_min is the minimum effective value of vertical direction image, and η is the index forgetting factor, and max (aiz_meas) is the maximal value of measuring image, min (aiz_meas) is the minimum value of measuring image, eff_max _OldAnd eff_min _OldIt is respectively the vertical direction image minimum effective value before upgrading the maximum effective value of preceding vertical direction image and upgrading.

8, the adaptive optimization method of the vertical sweep length of coaxial alignment as claimed in claim 7 is characterized in that: described index forgetting factor η passes through formula

$\mathrm{\η}=\min (1,C_{\mathrm{\η}}+\min (K*C_{\mathrm{\η}}*\frac{\max (\mathrm{aiz}\_\mathrm{meas})-\min (\mathrm{aiz}\_\mathrm{meas})}{\max (\mathrm{aiz}\_\mathrm{meas})+\min (\mathrm{aiz}\_\mathrm{meas})},\max ((L-C_{\mathrm{\η}}),0)))$ Determine, wherein, C _ηFor minimum is upgraded the factor, the maximum factor of upgrading of L, K is a coefficient of balance, and max (aiz_meas) is the maximal value of measuring image, and min (aiz_meas) is the minimum value of measuring image.

Images