TRANSAT CORP.

Application Note: Automatic Etch Control Using the CEM-1000

This application note describes a system that monitors and controls the etching of quartz blanks by measuring the frequency of a monitor blank that is immersed in the etch tank together with the etch load.

I. PRINCIPLE
The system monitors the etching of quartz crystals to a target thickness and provides a signal upon reaching target. A "Monitor Blank" is mounted in a "Monitor Probe" and immersed in an etch tank together with the etch load. The probe exposes one face of the monitor blank to the etchant. Since, during etching, all blank surfaces undergo the same thickness removal, the monitor thickness removal is half the load thickness removal, and etching of the blanks can be controlled by measuring the frequency of the monitor blank. Since both monitor and load blank are exposed to the same environment, the etch control is independent of etch rate variations.

The thickness change DT of the load blanks is expressed in terms of the initial frequency F1 and the final frequency F2 of the monitor blanks as:

DT/2 = (F2 - F1)/(F1 x F2) [kHz/MHz] (1)

The basic arrangement is shown in Fig. 1. A "Crystal Etch Monitor" (CEM) is used to monitor the frequency.

.

II. MONITOR PROBE
Fig. 2 shows the monitor probe. It consists of two parts, a "Monitor Cartridge" that hold the monitor blank, and a "Signal Adapter" that includes electronic circuitry. It is housed in a Teflon-coated metal tube that is partially immersed in the etchant and serves as the electrode through capacitive coupling to the etchant. The probe is suitable for all commercially used etchants except (currently) high temperature sodium hydroxide.

The CEM uses two different probes: a medium-frequency probe with a frequency range to about 45 MHz (for "shallow" etch, with DT < 2) and a high-frequency probe, with a frequency range from 20 to 150 MHz (for "deep" etch). For the medium frequency probe, the typical repeatability is r=20 ppm. For the high-frequency probe, r=20 ppm in the middle of the range and up to 60 ppm toward the upper and lower frequency limits.

III. MONITOR BLANK
The monitor blank is mounted in the cartridge and is easily replaced by removing a cap nut. In principle, the blank dimensions are arbitrary, but the present cartridge is designed for a blank diameter of 8.2 mm. The thickness is a compromise between two conflicting goals; to minimize thickness in order to maximize measurement accuracy, yet avoid extreme fragility for handling and mounting purposes. A practical compromise is a monitor frequency of about 30 MHz. Monitor and load blanks are preferably of the same quartz material. For etching in the nonlinear range, they also must have the same surface finish. For etching in the linear range, both monitor and load blanks must be pre-etched into the linear range.

IV. ETCH CURVE
In the following, reference is made to the familiar etch curves that show thickness removal versus time. They include an initial "Nonlinear" etch rate, during which the lapping debris is removed and the etching changes from a fast rate toward a slower "Linear" rate. "Shallow Etch" covers the initial etch up into the linear range. "Deep Etch" covers the linear range up into large values of DT.

V. CEM OPERATION
The operator enters the initial and target frequencies of the etch load. When he pushes the "Start" button, the CEM measures the initial frequency of the monitor blank and computes the target frequency for monitor and load. During etching, the CEM displays the current frequency and the frequency change (DF) of etch load or monitor. Upon reaching target, it provides audio and electronic signals for terminating etching.

All CEM functions can be monitored and controlled by an external computer via an RS-232 communications interface. The CEM can be used in different operating modes, including:

A. Frequency Monitor Mode: At the start, etch load and monitor probe are immersed in the etchant, and etching is terminated when the monitor blank reaches its target frequency.

For lapped blanks, the immersion must be simultaneous, since the initial etch rate changes very fast. The monitor blank must be replaced after each etch run.

For pre-etched blanks, the monitor can be immersed before the etch load, and it can be re-used until it becomes nonlinear.

B. Rate Monitor Mode: The etch rate is monitored with the monitor probe immersed continuously or periodically. Etch rate variations can be accounted for either manually or automatically through an external computer. The monitor blank must be pre-etched into the linear range and can be used for repeated etch runs.

C. Multi-batch Mode: A pre-etched monitor blank is immersed in the etchant. An external computer is used to:

1. Monitor the thickness change DT
2. Accept as input the initial and target frequencies of "n" etch loads
3. Start monitoring the first load when the operator immerses it and enters "start" for load 1
4. Signal etch termination when load 1 reaches its target thickness removal
5. Repeat steps 3) and 4) for loads 2,3,....n. The loads may be etched simultaneously or sequentially.

D) Deep Etching: In deep etching of inverted mesa blanks, the monitor is preferably also a mesa. One reason for this is the protective metal coat on the blank periphery, which prevents the etchant from penetrating under the seal in the monitor cartridge. Another reason is to maximize the monitor frequency while avoiding extreme fragility for handling the blank. Table 1 gives a selection of initial and final monitor frequencies (F1M and F2M) for different etch load conditions (F1L and F2L).

Table I - Deep etch monitor blank frequency selection

Load range F1L F2L	DT	F1M = 20 MHz	F1M = 30 MHz	F1M = 33 MHz	F1M = 35 MHz
20-200 MHz	45.0	F2M = 36 MHz	F2M = 92 MHz	F2M = 128 MHz	F2M = 165 MHz
20-150 MHz	43.3	F2M = 35 MHz	F2M = 86 MHz	F2M = 116 MHz	F2M = 145 MHz
20-100 MHz	40.0	F2M = 33 MHz	F2M= 75 MHz	F2M = 97 MHz	F2M = 117 MHz
20- 60 MHz	33.3	F2M = 30 MHz	F2M = 60 MHz	F2M = 73 MHz	F2M = 84 MHz
20- 40 MHz	25.0	F2M = 27 MHz	F2M = 48 MHz	F2M = 56 MHz	F2M = 62 MHz

E) SC-cut Etching: The C-mode of SC-cut blanks is a quasi-shear mode, where part of the vibrating motion is perpendicular to the blank surface and therefore severely damped by the etchant. This inhibits reliable frequency measurements.

If the frequency relation between C-mode and B-mode is known, the C-mode can be controlled via the B-mode, provided there is good temperature control to account for the relatively large frequency-temperature coefficient (about 25 ppm/°C) of the B-mode.

VI. ETCH CONTROL ACCURACY

Generally, there is a "Load Frequency Error" between the final load frequency (after etching) and the targeted load frequency. The main sources for the error are: A) the repeatability of monitor and load frequency measurements; and B) a difference in etch rates for monitor and load blanks.

A) Error Due to Frequency Measurement Repeatability

For perfect repeatability, the final load frequency is:

	F1L x F1M x F2M
F2L =	----------------------------------------	(2)
	(F1M x F2M) - (2 x F1L x (F2M - F1M))
where,	F1L = Initial load frequency F2L = Final load frequency F1M = Initial monitor blank frequency F2M = Final monitor blank frequency

The maximum possible load frequency error due to each of these frequencies can be calculated from equation (2) by associating the monitor frequencies with a repeatability error "r" and the load frequencies with the repeatability error "R".

Errors due to Monitor frequency measurement repeatability.The load frequency errors due to the monitor frequency measurements are shown in Fig. 3a, 3b, and 3c. The figures show that the errors increase with DT and with increasing ratio F1L/F1M. Therefore, it is important to choose the highest feasible monitor frequency.

For shallow etch, the error due to the monitor frequency measurement is: Load Error (ppm) = 4 x r x F1L/F1M.

Errors due to Load frequency measurement repeatability. The load frequency error due to the initial load frequency (F1L) measurement repeatability error, R (ppm), is:

Load Frequency Error (MHz)	= F2L (Target)	-	F2L (Actual)
			1000 x (F1L x (1 + R))
	= F2L (Target)	-	---------------------------------
			1000 - DT x (F1L x (1 + R))

This error occurs whether the CEM is used or not!

The error due to the final load frequency measurement (F2L) repeatability error is a constant R. Fig.4 shows the error due to the measurement of both F1L and F2L. The error increases with increasing etch removal.

B) Errors Due to Different Etch Rates

The etch rates of monitor and load blanks are affected by the etchant's access to the blanks. The access to the monitor is limited because the blank is recessed in the probe head. The access of the load is limited by the mounting rack. Depending on the type of mounting rack, the etch rate for the load may be higher or lower than the monitor etch rate. Different rates produce a consistent offset between the final load frequency and the targeted load frequency. Once the user knows the offset, he can compensate by offsetting the target frequency.

An additional consistent frequency offset may occur if the monitor quartz material is different from the load material.

VII. A PRACTICAL EXAMPLE

The example is based on an evaluation report supplied by a customer (CTS-Reeves). Part of this report is quoted verbatim:

"... All tests were performed on 44.736 MHz, .250" diameter inverted Mesa blanks. The blanks were ready for final "bumping" to the production frequency window of 46.205 to 46.269 MHz. A 31 MHz, .320" diameter blank was used as a monitor. The monitor blank had been pre-etched into the linear etch range. The results of the test groups are as follows:

	Group #1	Group #2	Group #3	Group #4
Freq. In	44.665 - .690	44.553 - .579	44.628 - .656	44.418 - .445
Target	46.205	46.180	46.180	46.192
Freq. Out	46.247 - .274	46.208 - .248	46.188 - .225	46.224 - .254
Average	46.262	46.225	46.210	46.239
Adjusted Avg.	46.249	46.237	46.222	46.239

In previous evaluations we had determined that these blanks typically overshoot the target frequency when monitored by the CEM. We adjusted the target frequency to compensate for this overshoot. In order to determine the repeatability of the CEM, the differences in target frequency must be removed from the average of each group (adjusted average). You can then see that the CEM maintains a repeatability of 580 ppm. This would enable us to "bump" the blanks into the final etch window in one step...."

In this example, the customer used the same monitor blank for all etch runs. He also used the "high frequency" probe.

The overshoot in the example indicates that the load blanks were etched faster than the monitor was, because the etchant apparently had better access to the load blanks than to the monitor blanks. Depending on which type of mounting rack a customer uses, he could also get an undershoot. In either case, once the offset is known, it can be compensated as shown in the example.

The remaining error of 580 ppm (or +/- 290 ppm if the target were centered) is mainly due to the repeatability of the frequency measurements. This can be shown by determining the theoretical error from Figures 3c and 4:

The example has an initial load frequency of F1L = 44 MHz and a DT of about 1. The monitor frequency is 31 MHz. The closest curve in Fig. 3c is that for F1M = 30 MHz. For a DF = 1, it shows an error of about 5r. Assuming r = 40, one obtains a monitor frequency error of 200 ppm.

From Fig. 4, one obtains a load frequency error of 2R. Assuming R = 50, one obtains a load frequency error of 100 ppm. Hence, the total error is +/- 300 ppm, i.e. approximately the same value as in the practical example.

VIII. DEEP ETCH EXAMPLE

In this theoretical example, it is desired to etch the load blanks from 20 MHz to 150 MHz (DT = 43.3) in one step. An initial monitor blank frequency of 35 MHz is selected from Table 1. The final monitor frequency is 144 MHz.

From Fig. 3c the error due to the CEM repeatability is about 11r. Assuming a CEM repeatability of 40 ppm, the error is 440 ppm.

From Fig. 4 the error due to the load frequency measurements is about 8.5R. Assuming R=50, the error is 425 ppm. This error occurs independent of the etch control method, i.e. when etching using your present method without CEM.

The total error due to frequency measurement repeatability errors is about 865 ppm.