JPS6320015B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a one-field alignment method in vector scan electron beam lithography.

An example of a conventional one-field alignment method will be described below.

When drawing a pattern on a sample using an electron beam deflection means using a computer such as an electron beam drawing device,
Due to distortion caused by the control circuit and distortion caused by the deflection means, the coordinates (U, V) on the sample surface to which the electron beam is irradiated do not match the coordinates (x, y) of the CPU field specified by the computer, and both The following relationship (1) holds true between them.

U=(1+A)x+By+Cxy+D V=Ex+(1+F)y+Gxy+H...(1) However, A to H are distortion coefficients indicating the type and magnitude of distortion. Therefore, it is necessary to correct the amount of distortion expressed by ΔU=Ax+By+Cxy+D ΔV=Ex+Fy+Gxy+H (1)′ using the control system.

FIG. 1 shows a block diagram of the control system of a vector scan type electron beam lithography system. In addition, in Figure 1, x
Only the axis control system is shown; the y-axis control system is
This is omitted because it is the same as the axis control system.

In the figure, the coordinates (x, y) of the CPU field of the computer 1 are set by the graphic generator 2 to the D/A converter 3 as the starting and ending points of the line elements of the graphic by vector scanning. The D/A converter 3 converts the above start point and end point into analog values, and the linear generator 4
A beam scanning signal is formed by the integrating circuit.
This signal is corrected by the amount of distortion expressed by (1)' above in the automatic correction circuit 5. That is, correction is performed by setting correction coefficients a to h, which determine the above-mentioned A to H, from the computer 1 to the multiplier 11 and the adder 12. Thereafter, the corrected scanning signal passes through a distortion correction circuit 6 for higher-order distortion terms and a deflection amplifier 7, and excites a deflection coil 8 to deflect an electron beam 9. At this time, the electron beam 9 is scanned from the starting point to the ending point designated by the computer 1, and drawing is executed. Note that 13 indicates a DC power supply for performing shift correction.

In this system, A to H in equation (1) are the distortion coefficient gains (Gain) of the control system and deflection system.
It is approximately expressed by the following equation using components α _a to α _h , offset components β _a to β _h of distortion coefficients, and parameters a to h.

A= _a α _a + β _a (amplitude correction term) B= _b α _b + β _b (rotation correction term) C= _c α _c +β _c (keystone correction term) D= _d α _d + β _d (shift correction term) E= _e α _e +β _e (rotation correction term) F= _f α _f +β _r (amplitude correction term) G= _g α _g +β _g (keystone correction term) H= _h α _h +β _h (shift correction term)……(2 ) Therefore, instead of directly setting the values A to H to the automatic correction circuit 5, the parameters a to h are input from the computer 1 as correction coefficients.

By the way, in the conventional one-field alignment method, the gain components α _a to α _h of the distortion coefficients of the system are calculated for each chip to determine the automatic correction coefficients a to h.
The specific calculation method is as follows.

Now, let us consider a case where there are alignment marks on the four corners of the chip to be aligned.

First, find α _a to α _h . Automatic correction circuit 5 a~
Set h=+1, scan the electron beam, and find the coordinates (x ⁺ , y ⁺ ) of the CPU field when the electron beam passes over the alignment mark. Do this operation 4
Do this for each corner alignment mark, making 4
Coordinates of CPU field (x ⁺ _i , y ⁺ _i ) (i = 1 to 4)
seek.

Next, set a to h = -1 in the automatic correction circuit 5 and perform the same operation to further obtain the coordinates (x ^- _i , y ^- _i ) (i = 1 to 4) of the four CPU fields. .

Note that the mark position is calculated using a known method in which the position of the center of gravity of the detection signal pattern obtained by scanning the mark with an electron beam is expressed with the coordinates of the CPU field as a reference.

Here, if the coordinates of the four alignment marks on the sample surface are (U _i , V _i ) (i=1 to 4), then equation (3) is obtained from equation (1).

U _i = (1+A ⁺ )x _i ⁺ +B ⁺ y _i ⁺ +C ⁺ x _i ⁺ y _i ⁺ +D ⁺ V _i =E ⁺ x _i ⁺ +(1+F ⁺ )y ⁺ +G ⁺ x _i ⁺ y _i ⁺ +H ⁺ (3a) U _i = (1+A ^- )x _i ^- +B ^- y _i ^- +C ^- x _i ^- y _i ^- +D ^- V _i =E ^- x _i ^- +(1+F ^- )y _i ^- +G ^- x _i ^- y _i ^- +H ^- (3b) However, A ⁺ ~H ⁺ and A ^- ~H ^- in equation (2),
The values are a to h=+1 and a to h=-1, respectively.

Furthermore, the following equation can be obtained from equation (3).

(1+A ⁺ )x _i ⁺ +B ⁺ y _i ⁺ +C ⁺ x _i ⁺ y _i ⁺ +D ⁺ = (1+A ^- )x _i ^- +B ^- y _i ^- +C ^- x _i ^- y _i ^- +D ^- (4a) E ⁺ x _i ⁺ + (1+F ⁺ )y _i ⁺ +G ⁺ x _i ⁺ y _i ⁺ +H ⁺ =E ^- x _i ^- +(1+F ^- )y _i ^- +G ^- x _i ^- y _i ^- +H ^- (4b) (2 ) and the system satisfies α>β, then β
Since the term can be ignored, equation (4) becomes equation (5).

α _a (x _i ⁺ +x _i ^- ) + α _b (y _i ⁺ +y _i ^- ) + α _c (x _i ⁺ y _i ⁺ +x _i ^-
y _i ^- ) + 2α _d = x _i ^- −x _i ⁺ α _a (x _i ⁺ +x _i ^- ) + α _b (y _i ⁺ +y _i ^- ) + α _c (x _i ⁺ y _i ⁺ +x _i ^-
y _i ^- ) + 2α _d = x _i ^- −x _i ⁺ α _e (x _i ⁺ +x _i ^- ) + α _f (y _i ⁺ +y _i ^- ) + α _g (x _i ⁺ y _i ⁺ +x _i ^-
y _i ^- ) + 2α _h = y _i ^- −y _i ⁺ (5) Since equation (5) is (i = 1,...4), eight equations are obtained and α _a ~ α _h are calculated. .

Next, a to h are calculated when α _a to α _h are determined. If the coordinates of the alignment marks at the four corners in the CPU field are (x _i , y _i ) (i = 1 to 4), then the coordinates of each alignment mark on the sample are U _i '= if no correction is performed. (1+A)x _i +By _i +Cx _i y _i +D V _i ′=Ex _i +(1+F)y _i +Gx _i y _i +H (5)′. A to H are determined so that (U _i ′, V _i ′) match the actual coordinates of the registration mark. In other words, since it is only necessary to determine a to h, from equations (3) and (5'), as U _i = U _i ′ and V = V _i ′, x _i α _a a + y _i α _b + x _i y _i α _c C + α _d d=1/2 {(x _i ⁺ +x
_i ^- −2x _i ) + α _a (x _i ⁺ −x _i ^- ) + α _b (y _i ⁺ −y _i ^- ) + α _c (x _i ⁺ y _i ⁺ −x _i ^- y _T ^- )} (6a) x _i α _e a+y _i α _f +x _i y _i α _g C+α _h d=1/2 {(y _i ⁺ +y
_i ^- −2y _i )+α _e (x _i ⁺ −x _i ^- ) +α _f (y _i ⁺ −y _i ^- )+α _g (x _i ⁺ y _i ⁺ −x _i ^- y _i ^- )}(6b) ( i=1,...,4) is obtained.

However, the β term was ignored due to the relationship α>β.

Therefore, since eight equations are obtained in equation (6), the values of a to h to be set in the automatic correction circuit 5 are calculated.

The above-mentioned method has been employed in the past. However, since this method performs mark detection eight times for each chip to calculate the distortion coefficient components α _a to α _h of the system, the number of mark detections becomes extremely large, making it extremely difficult to achieve one field alignment. It has disadvantages such as being time consuming.

The present invention has been made with attention to the above points, and is a method for correcting a control system so as to match a CPU field and a drawing field that control a vector scan type electron beam drawing apparatus. The purpose of this is to shorten and improve accuracy.

In order to achieve the above object, the present invention includes:
Before performing 1-field alignment to match the CPU field and drawing field,
A method of matching the CPU field and the drawing field in absolute dimensions in the control system, that is, absolute dimension calibration, is applied, and the gain and offset components of each correction term of the control system and beam deflection system are calculated, and the CPU field is adjusted using the correction circuit setting parameters. Calibrate with absolute dimensions. After completing the above method, 1 chip above 1
The position of the alignment mark for field alignment is measured and the set position of the alignment mark and the actual measured position are detected, and control is performed using the gain component and offset component values of each correction term obtained previously from the error value. A method is adopted in which a correction coefficient for each correction term in the system is calculated and set.

Hereinafter, the present invention will be explained in detail with reference to Examples.

In the present invention, each component α _a ~α _h of the strain coefficient of the system, β _a ~
Note that β _h is unique to the system of the lithography system, and once determined, it remains a constant steady value until the direct lithography within one wafer is completed. Therefore, first, using one standard mark installed at a location other than the wafer 10, the CPU field and the desired drawing field are connected to the moving stage 1, which is a part of the drawing apparatus.
4 and a laser length measurement system to perform absolute dimension calibration. At this time, the gain components and offset components α _a to α _h and β _a to β _h of the system-specific distortion coefficients are determined as follows.

(1) Move the movable base 14 on which one standard mark is installed to a desired position, and measure the position of the movable base 14 after the movement using a known laser length measurement system.
The position (U _i , V _i ) of the standard mark on the moving table 14 is calculated from the measured value at this time. On the other hand, in the automatic correction circuit 5, the correction coefficients a to h are set to ±1, the electron beam is scanned, and the coordinate position of the CPU field (x _i
⁺ , y _i ⁺ ). The measurement method will be the same as the conventional method. The series of operations of movement and standard mark position detection described above is performed a plurality of times (n times) without bias within the electron beam deflection field on the same surface as the sample. As a result, the following equation holds between the obtained (U _i , V _i ) and (x _i ⁺ , y _i ⁺ ) from equation (1).

U _i = (1+A ^± )x _i ^± +B ^± y _i ^± +C ^± x _i ^± y _i ^± +D ^± V _i =E ^± x _i ^± +(1+F ^± )y _i ^± +G ^± x _i ^± y _i ^± +H ^± ...(7) (a~h=±1, compound same order, i=1,...n) However, A ⁺ =±α _a +β _a H ⁺ =±α _h +β _h (a~h=±1, Composite same order) ...(8) Calculate the values from A ⁺ to H ⁺ from formulas (2) and (7). In order to uniquely obtain these values, it is necessary to move the standard mark to at least four locations and perform mark detection. Normally, to ensure accuracy, the deflection field is divided into 25 (5 x 5) points, 81 (9 x 9) measurement points, etc., and (U _i , V _i ) and (x _i ⁺ , y _i
⁺ ) is measured. In these cases, the number of equations for calculating A ⁺ to H ⁺ becomes very large, and it becomes necessary to calculate by the method of least squares.

(3) From the obtained A ⁺ ~H ⁺ , α _a ~ using equation (8)
α _h , β _a to β _h are calculated as follows.

α _a =A ⁺ −A ^- /2, β _a =A ⁺ +A ^- /2 〓 〓 ...(9) α _h H ⁺ −H ^- /2, β _h =H ⁺ +H ^- /2 (4) Then , conditions for a to h such that the distortion term becomes zero for these α _a to α _h and β _a to β _h are determined. That is, as shown in the formula below, a _p ~ h _p
becomes.

a _p = -β _a / α _a (A = 0) 〓 〓 〓 ...(10) h _p = -β _h / α _h (H = 0) In this case, α _a ~ α _h , β _a ~ β The value of _h is stored in the computer 1.

(5) Next, set α _p to h _p in the automatic correction circuit 5, move the chip into the drawing field to perform field alignment, scan the alignment mark in the chip with the electron beam, and By detecting the position of the alignment mark, the amount of error between the measured alignment mark position and the CPU field alignment mark position (designed position) is determined.

FIG. 2 shows the relationship between the mark position 22 in a chip having four alignment marks and its error amount. Since the amount of error is measured after absolute dimension calibration, it can be expressed by the following equation (11).

ΔU _i =Ax _i +By _i +Cx _i y _i +D ΔV _i =Ex _i +Fy _i +Gx _i y _i +H...(11) (i=1~4) (6) In the case of Fig. 2, measurement at 4 locations is required. Yes, total 8
The equations A to H are obtained and the distortion coefficients A to H are calculated by solving the simultaneous equations. Note that when there are four or more measurement points, A to H can be determined by using the least squares method or the like.

(7) By substituting the distortion coefficients A to H into equation (2),
α _a that has already been calculated and stored in calculator 1
The values of correction coefficients a to h are calculated from ~α _h and β _a to β _h as follows. From formula (2) and A to H, a to h
is determined by the following formula.

a=(A-β _a )/α _a 〓 ...(12) h=(H-β _h )/α _h (8) Next, set the correction coefficients a to h from the calculator 1 to the automatic correction circuit 5. This completes one field alignment.

(9) Furthermore, from the subsequent 1-field alignment, the step of absolute dimension calibration is not performed, but the correction coefficients a to h are determined using the initially determined a _p to h _p , and the 1-field alignment is performed.

In the example described above, the shift terms D and H are added in the automatic correction circuit 5, but the computer 1 or the graphic generator 2 may also have a function of correcting the shift terms. That is, α _d , α _h , β _d , β _h d,
Even if the remaining correction term excluding the shift component of h is determined by the method of the present invention, and the shift correction amount is determined by another method,
1-field alignment can be performed.
Furthermore, as another example of the above, absolute dimension calibration is performed at arbitrary time intervals, and the gain components, offset components α _a to α _h , β _a of the distortion coefficients of the deflection system and control system are
.about.β _h can be newly determined and one-field alignment can be performed while updating α _a to α _h and β _a to β _h already stored in the computer 1. This method is effective when there is a large change over time.

Further, calculation of α _a to α _h and β _a to β _h is a task that must be performed when operating the system, and the values of the operating time may always be employed. This is actually more desirable in order to increase speed. In addition, a to h, a _p
It is preferable that the calculations of ~h _p , α _a ~α _h , β _a ~ β _h , and A ~ H be processed within the computer 1, but it does not deviate from the present invention even if they are implemented as a circuit.

As described above, in the method according to the present invention, the number of mark detections is halved to 4 times compared to 8 times in the conventional method, and in addition, unlike the conventional method, each component α of the distortion coefficient of the system is Since it is not necessary to calculate _a to α _h and β _a to β _h , the time required for one field alignment is significantly shortened. In addition, in the present invention, a mark other than the alignment mark, that is, a standard with _a high S/N provided at a location other than the wafer, is used as a sample for determining each component α _{a to} α _h and β a to β h of the strain coefficient of the system. By using marks, the distortion coefficient components are calculated accurately, so the CPU field and the drawing field can be perfectly matched, and it is expected that the precision in field alignment will be improved compared to conventional methods.

Therefore, by employing the present invention for direct writing, higher precision and faster writing can be achieved.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing the control system (x-axis) of a vector scan type electron beam lithography system, and Fig. 2 shows the mark position error within one chip from the coordinate point in the CPU field in the case of four marks. FIG. In the figure, 1... Computer, 2... Graphic generator, 3... D/A converter, 4... Direct generator, 5
... Automatic correction circuit, 6 ... Distortion correction circuit, 7 ... Deflection amplifier, 8 ... Deflection coil, 9 ... Electron beam, 10 ... Sample, 11 ... Multiplier, 12 ... Adder, 13 ... ...DC power supply, 14...Moving table, 20
...CPU field, 21...Drawing field,
22...Mark position.

Claims (1)

[Claims] 1 One image field and computer for wafer direct writing in vector scan electron beam lithography
In the one-field alignment method, which corrects the control system for field alignment so that the CPU fields match, before detecting the positions of multiple alignment marks within one chip, Calibrate the CPU field in absolute dimensions using a standard mark with a large N ratio, then measure the position of each alignment mark on the above chip, and compare the CPU field calibrated with the measured alignment mark position and absolute dimension. A one-field alignment method characterized by calculating the amount of error between the alignment mark and the design position, and correcting the control system for field alignment based on this.